Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 May 5.
Published in final edited form as: Nat Cardiovasc Res. 2023 Oct 30;2(11):1003–1014. doi: 10.1038/s44161-023-00356-1

Human circulating CD24hi marginal zone B cells produce IgM targeting atherogenic antigens and confer protection from vascular disease

Tanyaporn Pattarabanjird 1,2,3,4, Anh Tram Nguyen 2, Chantel McSkimming 1, Huy Q Dinh 5, Melissa A Marshall 1, Yanal Ghosheh 6, Rishab Gulati 6, Chistopher Durant 6, Jenifer Vallejo 6, Ryosuke Saigusa 6, Fabrizio Drago 1,2, Thomas V Guy 7, Katherine Premo 7, Angela M Taylor 2,4, Soumen Paul 8, Bijoy Kundu 3,8, Stuart Berr 3,8, Ayelet Gonen 9, Sotirios Tsimikas 9, Yury Miller 9, Shiv Pillai 7, Klaus Ley 10, Catherine C Hedrick 10, Coleen A McNamara 1,2,4,✉
PMCID: PMC13138113  NIHMSID: NIHMS2160349  PMID: 39196097

Abstract

IgMs that inactivate oxidation-specific epitopes (IgMOSE), which are secondary products of lipid peroxidization, protect against inflammatory diseases, including diet-induced atherosclerosis. However, the human B cell subtype that produces IgMOSE remains unknown. In this study, we used single-cell mass cytometry and adoptive transfer of B cell subtypes to NOD. Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice to identify B27+IgM+CD24hi cells as the major producers of IgMOSE in humans. Notably, these cells have characteristics of human circulatory marginal zone B (MZB) cells, which are known to be atheoroprotective IgM producers in mice. CD24 antibody treatment to reduce MZB cells and IgM in a hyperlipidemic humanized mouse model provides the evidence that MZB cells protect against vascular inflammation. Consistent with these findings, the frequency of B27+IgM+CD24hi cells (MZB) in patients inversely correlates with coronary artery disease severity.

Oxidized phospholipids (OxPLs) and their secondary lipid peroxidation-derived adducts, known as oxidation-specific epitopes (OSEs), are formed in inflamed tissues, such as atherosclerotic lesions in the artery wall1–5, and can mediate further tissue inflammation and damage4,6–9. B-1 cell-derived IgM that recognizes and binds to these OSEs can block their pro-inflammatory effects10,11. Epidemiological studies in humans clearly demonstrate that levels of IgM to the OSE malondialdehyde-modified low-density lipoprotein (MDA-LDL) are inversely associated with coronary artery disease (CAD) severity and cardiovascular events, such as heart attacks and death6,11–15. Immunostaining and in vivo imaging techniques reveal MDA-LDL as a dominant neoepitope in atherosclerotic plaques16,17. In mice, IgM to MDA-LDL (IgMMDA-LDL) is produced by both subtypes of atherosclerosis-attenuating B-1 cells (B-1a and B-1b)18,19. Although a putative human B-1 cell that spontaneously produces IgM has been reported, the cellular source of IgMMDA-LDL in humans is unknown.

To identify the human B cell subtype(s) that produce IgMMDA-LDL, mass cytometry (CyTOF) using a customized panel containing both established human B cell subtype markers and chemokine receptors known to regulate murine IgMMDA-LDL production and B-cell-mediated atheroprotection19–21 was applied on purified B cells from human subjects with high versus low plasma levels of IgMMDA-LDL. This unbiased approach along with further in vivo characterization, using a humanized mouse model, revealed that human circulating B cells of the subtype uniquely marked by CD27, IgM and high expression of CD24 were major producers of IgMMDA-LDL. Additional functional and marker characterizations revealed that these human IgMMDA-LDL-producing B cells bear characteristics of circulating marginal zone B (MZB) cells. Lastly, our study also unveils a promising role of these circulating MZB cells in protecting against vascular inflammation and human atherosclerosis.

Results

B cell subsets enriched in humans with high IgMMDA-LDL

A 24-antibody mass cytometry panel was used to subtype circulating B cells in 28 patients with cardiovascular disease with high plasma IgMMDA-LDL (n = 14) and low plasma IgMMDA-LDL (n = 14), otherwise matched for cardiovascular risk factors (CRFs). Results of self-organizing map clustering (FlowSOM) and consensus clustering of CD19+ B cells identified 11 distinct clusters of B cells represented by nonlinear embedding space using uniform manifold approximation and projection (UMAP). Unexpectedly, cluster frequency analysis revealed one individual with a markedly expanded B cell cluster (CD20+CD27+CD25hi) not seen in any other individual (Extended Data Fig. 1). CD20+CD27+CD25hi B cells are a special type of B cell with low capability for immunoglobulin secretion but efficient antigen presentation with potential roles in autoimmune diseases22,23. Due to the clearly aberrant nature of this individual’s B cell pool, they were excluded from further analysis. FlowSOM clustering analysis was re-performed on the remaining 27 CRF-matched individuals with cardiovascular disease with high (n = 13) and low (n = 14) plasma IgMMDA-LDL (Fig. 1a). Notably, individuals with high IgMMDA-LDL also had significantly more IgM to an MDA-LDL mimotope, phosphoryl choline (PC)-BSA (IgMPC-BSA) and oxidized cholesterol esther (OxCE) (IgMOxCE) (Supplementary Table 1). Consistent with these findings, levels of plasma IgMMDA-LDL significantly correlated with both IgMPC-BSA and IgMOxCE levels (Extended Data Fig. 2). Measures of IgG to these same epitopes were not different between groups (Supplementary Table 1). UMAP visualization of this clustering analysis indicated 11 distinct clusters (Fig. 1b). Cluster frequency analysis of these 27 individuals demonstrated similar frequency of each cluster without disproportionate expansion of specific B cell subtypes (Fig. 1c). Each cluster was annotated based on expression of B cell markers represented in the heat map (Fig. 1d). Of all the clusters, only clusters 1 and 8 had elevated frequency in individuals with high IgMMDA-LDL (a statistically significant association with cluster 1’s frequency and a trending association with cluster 8’s frequency) (Fig. 1e), although the association with cluster 8 was not statistically significant. Rainbow plots revealed that clusters 1 and 8 were the only clusters that were CD27+IgM+ (B27+IgM+), which is a marker of innate IgM-producing B cells24–26.

Fig. 1 |. Humans with high levels of IgM specific to MDA-LDL have a greater frequency of CD27+IgM+ B cells (B27+IgM+).

Fig. 1 |

Open in a new tab

a, Cohort of 27 individuals with high (IgM to MDA-LDL > 5,000 RLU, n = 13) and low (IgM to MDA-LDL < 5,000 RLU, n = 14) IgM specific to MDA-LDL with matched age, body mass index (BMI), hypertension (HTN), percent female, total cholesterol (TC), HDL cholesterol and LDL cholesterol (P values were calculated using two-sided Mann–Whitney Wilcoxon test without corrections for multiple comparisons). b, Representative meta-cluster UMAP showing 11 distinguished B cell subsets by using FlowSOM clustering and representative UMAP plots of distinguishing surface markers used for immunophenotyping B cell subsets. c, Bar chart showing percentage of B cells in all subsets for all donors colored as in meta-clustering UMAP. d, Heatmap showing median expression of 24 surface markers from meta-clustering and potential phenotypes of 11 B cell subsets. e, Biaxial plots to compare frequency of each cluster as a percentage of total B cells in individuals with high (n = 13) and low (n = 14) IgM MDA-LDL. Data in Fig. 5e were analyzed using two-sided Mann–Whitney Wilcoxon test. Values are mean ± s.d. *P < 0.05. Biaxial plots in e are defined as the following: minima = lowest value, maxima = highest value, center = median, lower bound of box = quartile 1 and upper bound of box = quartile 3. RLU, relative light unit.

To better understand potential subtypes of IgM-producing B cells, clusters 1 and 8 were combined and re-clustered using the same 24-antibody mass cytometry panel as prior. Unsupervised clustering using a Louvain algorithm revealed four subtypes of innate IgM-producing B cells (Extended Data Fig. 3a). Each IgM-producing B cell subtype was annotated based on expression of B cell markers (Extended Data Fig. 4b,c), consistent with prior literature24,27–29. UMAP of CD24 expression across these four subtypes of innate IgM-producing B cells was also demonstrated (Extended Data Fig. 3d). Subtype 1 demonstrated a phenotype consistent with pre-switched B cells and was found to be enriched in individuals with low plasma IgMMDA-LDL. In contrast, subtype 3 demonstrated an MZB1 cell phenotype given high CD24, CCR7 and CD25 expression27 and was found increased in frequency in individuals with high plasma IgMMDA-LDL (Extended Data Fig. 3d).

CD24 augments the atheroprotective IgM production of B27+IgM+ cells

To unbiasedly identify surface proteins that potentially regulate IgMMDA-LDL production in CD27+IgM+ B cells, we analyzed our mass cytometry (CyTOF) data on the circulating human B cells obtained from the 27 individuals to correlate expression of 21 surface markers, known to regulate B cell functions and trafficking with the level of plasma IgMMDA-LDL. The analysis indicated that CD24 was the only marker with a direct significant correlation (r = 0.52, P = 0.006) (Fig. 2a and Extended Data Fig. 4). Notably, analysis of the association of IgM to other antigens linked to cardiovascular diseases demonstrated that CD24 expression on CD27+IgM+ B cells was also directly associated with IgM to OxCE, IgM to high mobility group 1 protein (HMGB1) (a damage-associated molecular pattern found in atherosclerotic plaques) as well as IgM to ALDH4a1 (an autoantigen increased in atherosclerosis) (Extended Data Fig. 5).

Fig. 2 |. CD24 expression on B27+IgM+ cells marks a subtype with greater total and antigen-induced MDA-LDL-specific IgM.

Fig. 2 |

Open in a new tab

a, Pearson correlation between human plasma IgM to MDA-LDL level and CD24 expression measured by geometric mean (GM) on CD27+IgM+ B cells. b, Schematics of B27+IgM+CD24lo/− (n = 7), B27+IgM+CD24hi (n = 8) and B27− (n = 4) cells then adoptively transferred into humanized mice. c, Plasma level of total human IgM quantified by ELISA 1 week after transfer of B27+IgM+CD24lo/−, B27+IgM+CD24hi and B27− cells. d, Schematics of B27+IgM+CD24lo/− and B27+IgM+CD24hi pre-treated with MDA mimotope (n = 4 for B27+IgM+CD24lo/− group and n = 5 for B27+IgM+CD24hi groups) or scrambled peptide control (n = 7 for B27+IgM+CD24lo/− group and n = 8 for B27+IgM+CD24hi groups) then adoptively transferred into humanized mice. e,f, Plasma level of total human IgM (e) and IgM to MDA mimotope (f) quantified by ELISA 1 week after transfer of either B27+IgM+CD24lo/− or B27+IgM+CD24hi with either MDA mimotope stimulation or scrambled peptide control (n = 7 for B27+IgM+CD24lo/− + scrambled peptide; n = 4 for B27+IgM+CD24lo/− + MDA mimotope; n = 8 for B27+IgM+CD24hi + scrambled peptide; n = 5 for B27+IgM+CD24hi + MDA mimotope). Data were analyzed using one-way ANOVA with multiple comparisons correction, *P < 0.05, **P < 0.01 and ****P < 0.0001. Values are mean ± s.d. NS, not significant. n.d., not detected.

Considering a direct significant correlation between CD24 expression on CD27+IgM+ B cells and IgM to OSEs and atherosclerosis-related antigens, we hypothesized that CD24 might play a role in augmenting IgM production by CD27+IgM+ B cells. To test this hypothesis, we sort-purified the B27+IgM+ cells that were CD24lo (B27+IgM+CD24lo/−) and CD24hi (B27+IgM+CD24hi) expression. We included CD27− B cells (B27−) to test the importance of CD27 for IgM production (Extended Data Fig. 6). Adoptive transfer (AT) via intraperitoneal (IP) injection of equal numbers of human sort-purified B27+IgM+CD24lo/−, B27+IgM+CD24hi and B27− cells into NOD. Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) (B, T and functional natural killer (NK) cell deficient) mice (Fig. 2b) resulted in undetectable plasma level of IgM in NSG mice injected with B27− cells. In contrast, AT of B27+IgM+CD24hi and B27+IgM+CD24lo/− cells into NSG mice resulted in measureable plasma levels of IgM (Fig. 2b). Notably, the IgM plasma levels were significantly higher in the mice injected with B27+IgM+CD24hi cells compared to those injected with B27+IgM+CD24lo/− cells (Fig. 2c). Additionally, IP injection of B27+IgM+CD24lo/− and B27+IgM+CD24hi cells into NSG mice after stimulation with an MDA mimotope or the scrambled peptide control (Fig. 2d) revealed that MDA stimulation led to both higher total and MDA-specific IgM in the plasma of NSG mice injected with B27+IgM+CD24hi cells compared to the B27+IgM+CD24lo/− cells (Fig. 2e,f). Lastly, we also found that frequencies of both B27+IgM+CD24lo/− and B27+IgM+CD24hi cells in 60 individuals with CAD directly correlated with plasma level of IgMMDA-LDL (Extended Data Fig. 7).

B27+IgM+CD24hi cells have features of MZB cells

IP injection of equal numbers of B27+IgM+CD24lo/− and B27+IgM+CD24hi cells into NSG mice (Fig. 3a) revealed no significant difference in cell number in the peritoneal cavity (PerC) but significantly more B27+IgM+CD24hi cells in the spleen and bone marrow compared to B27+IgM+CD24lo/− cells 7 d after transfer (Fig. 3b). Additionally, a higher percentage of B27+IgM+CD24hi cells than B27+IgM+CD24lo/− cells were induced by MDA mimotope stimulation to express CD138 (Fig. 3c), suggesting that CD24 promotes antigeninduced plasma cell differentiation. Consistent with findings of greater splenic localization and plasma cell differentiation of B27+IgM+CD24hi cells, flow cytometry analysis using additional markers of MZB cells identified B27+IgM+CD24hi cells as CD21midCD23lo/−CD1c+ and having significantly higher expression measured by geometric mean (GM) of CD1C and lower expression of CD23 when compared to B27+IgM+CD24lo/− cells, consistent with the phenotype of human circulating MZB cells24,30 (Fig. 3d).

Fig. 3 |. B27+IgM+CD24hi cells have characteristics of activated MZB cells.

Fig. 3 |

Open in a new tab

a, Schematics of AT for trafficking studies. b, Cell numbers of transferred B27+IgM+CD24lo/− (n = 4) and B27+IgM+CD24hi (n = 5) recovered from PerC, spleen and bone marrow 1 week after AT (data analyzed using two-sided Mann–Whitney Wilcoxon test, *P < 0.05). c, % CD138+ of B27+IgM+CD24lo/− (n = 4) and B27+IgM+CD24hi (n = 4) after 24 h of unstimulated and MDA-mimotope-stimulated conditions (data were analyzed using one-way ANOVA with multiple comparisons correction, **P < 0.01). d, Histogram of one representative individual with CAD and GM of surface CD21, CD23 and CD1c expression on B27+IgM+CD24lo/− (n = 35) and B27+IgM+CD24hi (n = 35) cells (data were analyzed using two-sided Mann–Whitney Wilcoxon test). Values are mean ± s.d.

MZB cells are highly activated and bind to MDA antigen

As we previously demonstrated that B27+IgM+CD24hi cells can be stimulated by MDA mimotope to produce more total IgM and IgMMDA-LDL (Fig. 2d,e), we further explored the potential underlying mechanism of this MDA stimulation. Through flow cytometry labeling assay, MDA mimotope was found to bind significantly more to MZB/B27+IgM+CD24hi cells compared to B27+IgM+CD24lo/− cells (Fig. 4a,b). Scrambled peptide was used as a control, which demonstrated low non-specific binding to B27+IgM+CD24hi cells and B27+IgM+CD24lo/− cells (Fig. 4c,d). In addition, MDA-mimotope-stimulated circulatory MZB/B27+IgM+CD24hi cells were found to have a higher percent of p-BTK+, a downstream marker of B cell receptor (BCR) activation, compared to B27+IgM+CD24lo/− cells, suggesting that these MZB cells can be directly activated with MDA antigen through the BCR pathway (Fig. 4e). We also further demonstrated that other antigens, in this case phosphatidyl choline (PC), did not stimulate MZB cells as robustly as the MDA antigen (Fig. 4f).

Fig. 4 |. MZB-like B cells are more activating and bound to MDA compared to B27+IgM+CD24lo/−.

Fig. 4 |

Open in a new tab

a,b, Representative flow cytometry plots (a) and percent of cells bound to MDA mimotope (n = 4 for B27+IgM+CD24lo/− group and n = 4 for B27+IgM+CD24hi group) (b). c,d, Representative flow cytometry plots (c) and percent of cells bound to scrambled peptide (n = 4 for B27+IgM+CD24lo/− group and n = 4 for B27+IgM+CD24hi group) (d). e, % p-BTK+ on B27+IgM+CD24hi (n = 4) and B27+IgM+CD24lo/− (n = 4) from flow cytometry MDA mimotope stimulation. f, % p-BTK+ on B27+IgM+CD24hi (n = 4) and B27+IgM+CD24lo/− (n = 4) from flow cytometry with MDA and PC activation. g, % sequence of VH and JH usage from BCR sequencing of B27+IgM+CD24hi (n = 3) and B27+IgM+CD24lo/− (n = 3) obtained from human donors. h, Volcano plot indicating 117 DE genes (FDR < 0.05) between B27+IgM+CD24lo/− (blue, n = 4) and B27+IgM+CD24hi (orange, n = 4) out of 7,196 aligned genes (n = 4 per group). i,j, Pathway analysis to identify the top 30 cellular processes (i) and the top 40 canonical pathways (j) upregulated in sort-purified B27+IgM+CD24hi. Data were analyzed using two-sided Mann–Whitney Wilcoxon test. Values are mean ± s.d. *P < 0.05, **P < 0.01.

BCR sequencing of human donors revealed that MZB/B27+IgM+CD24hi had different VH and JH usage compared to B27+IgM+CD24lo/− cells. Specifically, MZB/B27+IgM+CD24hi cells used more VH3 and JH6, whereas B27+IgM+CD24lo/− cells used more VH1, JH4 and JH5 (Fig. 4g).

To identify differences in gene expression between B27+IgM+CD24lo/− and MZB/B27+IgM+CD24hi cells that might further explain these functional differences, we performed bulk RNA sequencing on sort-purified human B27+IgM+CD24lo/− and B27+IgM+CD24hi cells. Out of 7,196 aligned genes, there were 117 genes differentially expressed (DE) (false discovery rate (FDR) < 0.05 and log2 fold change (FC) < 1) in B27+IgM+CD24hi cells compared to B27+IgM+CD24lo/− cells (Supplementary Table 2). The annotated volcano plot (Fig. 4h) highlights increased expression of chemokine receptors, such as CCR6 and CXCR4, as well as IgM and signals downstream of the BCR (LYN, SYK, PLC γ2) in B27+IgM+CD24hi cells compared to B27+IgM+CD24lo/− cells. These RNA sequencing result was also found to be consistent with previous CyTOF data with cluster 8, a surrogate of B27+IgM+CD24hi, which reflected higher expression of CXCR4, CCR6 and CCR7, CD24 and CD43 when compared to cluster 1 (Fig. 1d). Ingenuity pathway analysis of DE genes identified processes such as lymphocyte survival/proliferation, migration and IgM production (Fig. 4i). Canonical pathways predominated by cytokine, chemokine, BAFF/April, CD40 and BCR signaling (Fig. 4j) were more highly expressed in B27+IgM+CD24hi cells compared to B27+IgM+CD24lo/− cells.

CD24 knockdown reduces MZB numbers and IgM plasma levels

To determine the impact of reduction in CD24 on B27+IgM+ cells, CD24 was knocked down in MZB/B27+IgM+CD24hi cells by nucleotransfecting Cas9/CD24-targeted gRNA. Efficiency of CD24 knockdown was evaluated by both flow cytometry and CD24 RT–qPCR as depicted in Fig. 5a and Extended Data Fig. 8, respectively. CD24 wild-type (WT) B27+IgM+ cells as well as CD24 knocked-down B27+IgM+ cells were IP injected into NSG mice (Fig. 5a). CD24 knocked-down MZB cells were found to have reduction in number in all harvested tissue compartments, including PerC, spleen and bone marrow (Fig. 5b). Plasma total IgM level was also found to be lower in adoptively transferred MZB cells with CD24 knocked down (Fig. 5c), suggesting that CD24 increases MZB survival, leading to higher IgM production. This result was also consistent with the previous result of B27+IgM+CD24hi cells producing higher level of total IgM compared to B27+IgM+CD24lo/− cells (Fig. 2b).

Fig. 5 |. CD24 knockdown reduces MZB numbers and IgM plasma levels.

Fig. 5 |

Open in a new tab

a, Experimental scheme for AT of CD24KO and CD24WT B27+IgM+CD24hi cells into humanized mice. b,c, Number of CD24KO (n = 4) and CD24WT (n = 4) B27+IgM+CD24hi cells recovered from PerC, spleen or bone marrow (BM) (b) and human plasma total IgM measured by ELISA (c) 1 week after transfer. Data were analyzed by using two-sided Mann–Whitney Wilcoxon test. Values are mean ± s.d. *P < 0.05.

Blocking CD24 on MZB increases vascular inflammation

As loss of CD24 on MZB cells resulted in reduction in IgM production, we investigated the impact of CD24 in vascular inflammation. NSG mice were engrafted with human peripheral blood mononuclear cells (PBMCs), injected with AAV8-PCSK9 and fed 10 weeks of Western diet (WD) to induce hyperlipidemia, similar to a study conducted by Proto et al.31. These hyperlipidemic NSG mice were then treated with a CD24-blocking monoclonal antibody (CD24mAb) or IgG isotype control for 3 weeks via intravascular injection (Fig. 6a). CD24mAb treatment led to a significantly lower number of B27+IgM+CD24hi cells in the spleen and bone marrow (Fig. 6b) as well as lower plasma level of total human IgM (Fig. 6c) and IgMMDA-LDL (Fig. 6d). 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) imaging was used to evaluate vascular inflammation. The imaging result indicated that CD24mAb treatment led to a higher FDG uptake in the aorta area (Fig. 6e,f) despite similar plasma total cholesterol level in both IgG isotype control and CD24mAb treatment groups (Fig. 6g).

Fig. 6 |. Reduction of B27+IgM+CD24hi (MZB) cells through using CD24mAb increases level of vascular inflammation.

Fig. 6 |

Open in a new tab

a, Schematics of hyperlipidemic humanized mice treated with 200 μg of CD24mAb or IgG isotype control two times per week for 3 weeks. b–d, Number of B27+IgM+CD24hi cells recovered from PerC, spleen or bone marrow (BM) (b) and human plasma total IgM (c) and IgM to MDA mimotope (d) measured by ELISA after 3 weeks of CD24mAb (n = 5) or IgG isotype control (n = 5) treatment. e,f, Representative images (e) and mean SUV obtained from FDG-PET (f) to evaluate aortic vascular inflammation of hyperlipidemic humanized mice treated with CD24mAb (n = 5) or IgG isotype control (n = 5) treatment. g, plasma total cholesterol in hyperlipidemic humanized mice after 10 weeks of WD and either CD24mAb or IgG isotype control treatment. Data were analyzed by two-sided using Mann–Whitney Wilcoxon test. Values are mean ± s.d. *P < 0.05. Ctrl, control; ID/ml, injected dose per milliliter; vg, viral genomes; wks, weeks.

Humans with severe atherosclerosis have low MZB frequency

As we demonstrated that treatment with the inhibitory CD24mAb reduced MZB cell number and increased vascular inflammation in hyperlipidemic humanized mice, we evaluated the relationship between human CAD severity and the number of circulatory MZB cells. Individuals in our study had their CAD severity assessed by quantitative coronary angiography (QCA) at the time of their blood draw, and CAD severity was measured by the well-established Gensini scoring system32 (Fig. 7a). Through reinvestigating 27 individuals whom we previously evaluated using CyTOF, we found an increase in frequency of MZB/B27+IgM+CD24hi cells in human subjects with lower CAD severity assessed by the Gensini scoring system (Fig. 7b). We also further validated this result with another separate cohort of 60 individuals with CAD. These individuals had either high CAD severity measured by the Gensini scoring system (score >30, n = 30) or low CAD severity measured by the Gensini scoring system (score <5, n = 30), otherwise matched for traditional CRFs (Fig. 7c). Individuals with low CAD severity were found to have significantly higher frequency of MZB/B27+IgM+CD24hi cells (Fig. 7c).

Fig. 7 |. Increase in frequency of B27+IgM+CD24hi (MZB) cells is associated with low CAD severity in two separate human cohorts.

Fig. 7 |

Open in a new tab

a, Schematics demonstrating the Gensini scoring system to evaluate CAD severity in both cohort 1 and cohort 2. b, Pearson correlation between frequency of circulatory B27+IgM+CD24hi (MZB) cells and Gensini CAD severity score in 27 individuals obtained from the CyTOF cohort, as depicted in Fig. 1a. c, Validating cohort of 60 individuals selected based on high (Gensini score >30, n = 30) and low (Gensini score <5, n = 30) Gensini CAD severity score, demonstrating no significant differences in age, body mass index (BMI), hypertension (HTN), statin use, percent female, total cholesterol (TC), HDL cholesterol and LDL cholesterol and smoking, with the exception of a significant difference in frequency of circulatory B27+IgM+CD24hi (MZB) cells (data were analyzed by using two-sided Mann–Whitney Wilcoxon test). Values are mean ± s.d. *P < 0.05.

Discussion

There is clear evidence that IgMOSE limits inflammation in pre-clinical models of inflammatory diseases, including atherosclerosis10,33–38. The main source of IgMOSE in mice is B-1 cells37,39,40, a cell type known to limit diet-induced inflammation and inflammation-related diseases, such as insulin resistance and atherosclerosis18,19,38,41,42. However, the cellular source of IgM to OSE in humans remains elusive. Oxidation of cellular membranes, lipids and proteins can produce a wide range of OSEs, but aldehyde adducts on LDL, such as MDA, are prominent OSEs in atherosclerotic plaques17. A wealth of epidemiological data demonstrate that individuals with advanced atherosclerotic disease and major adverse cardiac events (MACEs) have low levels of IgMMDA-LDL (refs. 1–6). In addition, immunization studies with MDA-LDL in both LDLR−/− mice and rabbits showed a decrease in atherosclerosis plaque burden35,43,44, and a recent study showed that germinal center B cell response to MDA antigen contributes to atheroprotection45. The importance of identifiying the cellular source of IgMMDA-LDL is highlighted by these findings and the likelihood that this same cell produces IgMOSE of other specificities, such as MAA, ApoB100, OxCE and PC43,46, as well as IgM to other autoantigens found in atherosclerosis plaque47,48. Indeed, our findings demonstrate a strong correlation among IgMMDA-LDL, IgMPC-BSA and IgMOxCE in humans (Extended Data Fig. 2), and the frequency of MZB/B27+IgM+CD24hi cells was significantly associated with IgMMDA-LDL. As such, identification of the cellular source of IgMMDA-LDL-producing cells in humans has the potential to inform the development of therapeutics aimed at cellular targeting that may allow for enhancing production of many atheroprotective IgM subtypes to treat the inflammatory state of atherosclerosis.

In this study, we used CyTOF of human B cells to unbiasedly identify B cell subytpes associated with plasma IgMOSE level. These low and high IgMOSE individuals were matched for CRFs to allow an associative study to be performed using a cost-intensive and labor-intensive single-cell technology. Given the limitation of a small cohort size, we also coupled our hypothesis-generating associative study with in vitro and in vivo studies to identify circulating B27+IgM+ cells as producers of IgMOSE in humans. These B27+IgM+ cells have been known as innate-like B cells, which play crucial roles in primary immune response through producing IgM against T-independent antigens and protecting against autoimmune and inflammatory diseases49,50, whereas germinal center-derived plasma cells are the IgG and IgM producers against T-cell-dependent antigens51,52. B-1 and MZB are the two subtypes of innate-like B cells that have been well studied. We provide the evidence that CD24 expression divides B27+IgM+ cells into two subtypes with high CD24 expression associated with plasma levels of IgMOSE in humans and higher plasma IgMMDA-LDL in NSG mice when injected with human B27+IgM+CD24hi cells. We identified key cellular processes and canonical pathways in B27+IgM+CD24hi cells, suggesting that they are more highly activated and proliferative (Fig. 4g,h). Additional flow cytometry suggests that these B27+IgM+CD24hi cells are circulating MZB cells as opposed to B-1 cells. Like B-1 cells in mice, human MZB cells are producers of natural IgM53 in response to innate stimuli (for example, OxPL and OSE)54–57. However, in contrast to mice, human MZB cells can be found in circulation, whereas murine MZB cells are confined to the spleen30,58. Our B27+IgM+CD24hi cells do, in fact, express markers traditionally used to identify human circulating splenic MZB cells (CD21midCD23−CD1C+; Fig. 3d)26,30,58.They also differentiate into antibody-producing cells/plasma cells, migrate to the spleen, produce IgM in a T-cell-independent manner and increase total and MDA-specific IgM production in response to MDA antigen stimulation, which is indicative of human MZB cell functionality53,55,58,59.

Clear identification of the subtypes of CD27+IgM+ B cells in human has been elusive, but they have included IgM memory, unswitched memory, B-1 and MZB cells. Multiple studies have explored putative human MZB cells through single-cell transcriptomic and BCR sequencing and found that human MZB cells can be marked with CD27+IgM+IgD+—consistent with our study54—and differentiated from IgM high transitional B cells by increased CD27 and CD1C expression, maintained CD24 expression as well as enhanced cell activation and migration once differentiated to MZB cells55. Consistent with this, our study demonstrated that B27+IgM+CD24hi/MZB cells had higher BCR activation and chemokine receptors (for example, CCR6 and CXCR4) expression55 than B27+IgM+CD24lo cells. In addition, BCR sequencing analysis of MZB/B27+IgM+CD24hi cells also indicated higher utilization of VH3 and VH4 compared to B27+IgM+CD24lo cells, which is consistent with the previous study by Bagnara et al.56.

Our study is the first, to our knowledge, to show that CD24 augmented T-independent IgM and IgMMDA-LDL production of CD27+IgM+ B cells (Fig. 2). Notably, CD24 expression on CD27+IgM+ B cells was also associated with other oxidation-specific neoantigens (for example, OxCE) as well as other atherosclerosis-related antigens (HMGB1 and ALDH4a1), as shown in Extended Data Fig. 5. This increase in IgM production was likely a result from higher BCR activation and survivability in B27+IgM+CD24hi cells (Figs. 4b,c and 5b,c). MZB cells are known to produce IgM in a T-cell-independent manner through antigen activation of the polyreactive BCRs, leading to clonal expansion53. It is possible that MZB cells produce IgM to each oxidation-specific neoantigen, and atherosclerosis-related antigens might be a different subpopulation within MZB cells. Our current BCR sequencing has a limited sequencing depth with a lack of CDR3 sequence to address this hypothesis. Followon study with a deeper BCR sequencing allowing clonotype analysis to investigate clonal expansion and sensitization to atheroscleroticrelated antigens is needed to further validate this hypothesis.

Murine MZB cells were atheroprotective, likely through suppressing the activation of T follicular helper cells60,61, and Yanyi et al. demonstrated that murine MZB cells prevented cardiac remodeling after myocardial infarction62. Our study demonstrated that human MZB cells produce atheroprotective IgMs and that these cells occur more frequently in individuals with less severe CAD. In addition to a potential protective effect of B27+IgM+CD24hi/MZB cells against severe CAD, we used a hyperlipidemic humanized model to test whether inhibition of CD24 using a CD24mAb reduces B27+IgM+CD24hi cells and the IgMMDA-LDL that they produce. These hyperlipidemic humanized mice have their cholesterol level mimicking human level. As such, there is no obvious atherosclerotic plaque formed in the aorta after 10 weeks of diet. However, a previous study performed by Jarr et al.63 found that vascular inflammation evaluated by FDG-PET was well correlated with atherosclerotic lesion vulnerability index, allowing us to use inflammation of vasculature as a surrogate to evaluate potential early atherosclerotic lesions. Analysis of in vivo vascular inflammation using FDG-PET provided evidence that CD24mAb treatment can lead to worsening diet-induced vascular inflammation. It is likely that the blockage of CD24 on B27+IgM+CD24hi/MZB cells with mAb is one of the underlying mechanisms leading to worsening vascular inflammation. However, CD24 is also expressed on other immune cells64, and that might also contribute to its atheroprotective effect. Notably, CD24mAb was recently shown to have immunomodulatory benefits on triple-negative breast cancer through blocking the CD24 ‘don’t eat me’ signal63 and is among the more recent immunotherapy regimens being tested in various trials. The present study provides, to our knowledge, the first insight into possible cardiovascular toxicity of this immunotherapy.

In summary, the present study identifies B27+IgM+ cells as IgMMDA-LDL-producing B cells in humans. Deeper evaluation of the CD24hi B27+IgM+ cells revealed a phenotype consistent with human circulating MZB cells and demonstrated that they could be induced to produce atheroprotective IgMMDA-LDL through specific antigen stimulation. This feature holds promise for modulating this subtype to bolster innate immune protection from OSEs in atherosclerosis. Highly proliferative, migratory and activated characteristics of these B27+IgM+CD24hi cells require further investigation to help understand how to manipulate these cells and identify other novel targets to boost their atheroprotective property. In addition, study of a role of B27+IgM+CD24hi cells in a large population cohort with rigorous multi-parametric analyses for clinical covariates is also needed to implicate any clinical significance. Lastly, as anti-CD24 is now under development as a cancer immunotherapy, the effect of CD24mAb in depleting IgMMDA-LDL-producing B cells and increasing vascular inflammation highlights a potentially important adverse effect of this approach for cancer therapy. Therefore, a deeper understanding of CD24 in subclinical and clinical cardiovascular disease and other OSE-related diseases is warranted.

Methods

Our research complies with all relevant ethical regulations. All human study protocols were approved by the Human Institutional Review Board (IRB) at the University of Virginia under IRB number 15328 for individuals with CAD and IRB number 16017 for healthy human volunteers. All murine study protocols were approved by the Animal Care and Use Committee at the University of Virginia.

Human CAD subjects

All human CAD subjects (n = 87) from the Coronary Assessment in Virginia (CAVA) cohort were recruited for the study through the Cardiac Catheterization Laboratory at the University of Virginia. All participants provided written informed consent before enrollment. No compensation was provided. The study was approved by the Human IRB under IRB number 15328. Peripheral blood was obtained from these participants before catherization. Demographics of all individuals with CAD are provided in Figs. 1a and 7c. Individuals were excluded if they had any of the following: any acute illness, type 1 diabetes, current acute coronary syndrome (ACS), autoimmune disease or on immunosuppressive therapy, prior organ transplantation, anemia, pregnancy and HIV infection. After cardiac catheterization, the extent of atherosclerotic disease was quantified by using QCA and Gensini score. Traditional risk factors were also assessed through demographic status, physical and physiological measurement and laboratory values. A detailed description of how Gensini score and each risk factor was quantified is presented below.

Healthy human volunteers

Peripheral blood from healthy volunteers was obtained after written informed consent. The study was approved by the Human IRB at the University of Virginia under IRB number 16017.

Mice

All animal protocols were ethically approved by the Animal Care and Use Committee at the University of Virginia. NSG mice were purchased from The Jackson Laboratory. Mice used in all the experiments in this study were both males and females, 8–10-week-old NSG mice, with 40–60% female mice in each experiment. All mice were housed at 68–79 °F with 40–70% humidity and 14-h light/10-h dark cycle. Mice were fed with a standard chow or WD.

QCA

Patients underwent standard cardiac catheterization with two orthogonal views of the right coronary artery and four of the left coronary artery according to accepted standards. QCA was performed using automatic edge detection at an end-diastolic frame. For each lesion, the frame was selected based on the most severe stenosis with minimal foreshortening and branch overlap. Computer software was used to calculate the minimum lumen diameter, reference diameter, percent diameter stenosis and stenosis length. Analysis was performed by blinded experienced investigators. The Gensini score was used to determine disease burden for each individual. In brief, each artery segment was assigned a score of 0–32 based on stenosis percentage. The severity score for each segment was multiplied by 0.5–5, depending on stenosis location. Scores for all segments were then added together to give a final score of angiographic disease burden. Score adjustment for collateral was not performed for this study.

PBMC isolation

Blood from individuals with CAD and healthy volunteers was drawn into BD K2 EDTA vacutainer tubes and processed at room temperature within 1 h of collection. Whole blood in vacutainers was centrifuged at 400g for 10 min at room temperature to remove platelet-rich plasma. Plasma was cryopreserved at −80 °C. PBMCs were further separated by Ficoll-Paque density gradient centrifugation (Ficoll-Paque PLUS (GE Healthcare Biosciences) and SepMate-50 (STEMCELL Technologies)) following the manufacturers’ protocols. Trypan blue staining was performed to acquire live cell counts. PBMCs were cryopreserved in freezing solution (90% FBS/10% DMSO) or used fresh. PBMC vials were stored at −80 °C in a Mr. Frosty (Thermo Fisher Scientific) for 48 h and then stored in liquid nitrogen until use.

ELISA to quantify total and anti-OSE IgM or IgG isotypes in mice and humans

Total IgM or IgG subtypes in mouse or human plasma were measured using colorimetric ELISA as described previously18. Levels of IgM antibodies specific for MDA-LDL and OxCE in the serum or plasma were determined as previously described18,43,65. Levels of IgM or IgG against MDA-LDL and OxCE in human plasma were measured by chemiluminescent ELISA as previously described18,43,65. In brief, 96-well plates were coated with 5 μg ml−1 MDA-LDL, MDA mimotope or OxCE and incubated overnight at 4 °C. Plates were then blocked with 1% BSA-PBS before incubating with plasma either from humans or mice for 1.5 h. Goat anti-human or goat anti-mouse IgM or IgG were then used as detecting antibody.

CyTOF optimization and staining

All metal-conjugated antibodies were purchased from Fluidigm, and purified unlabeled antibodies were purchased from BioLegend. Unlabeled antibodies were conjugated in-house using the Maxpar Antibody Labeling Kit (Fluidigm) according to the manufacturer’s protocol and stored at 4 °C. Cryopreserved PBMCs obtained from healthy volunteers were used for antibody titration to determine optimal concentration.

Cryopreserved PBMCs obtained from 27 individuals with CAD with high and low plasma IgMMDA-LDL were thawed and washed twice with warm complete media (RPMI supplemented with 5% FBS, 1 mM sodium pyruvate and penicillin–streptomycin). All PBMC samples had 85–98% viability. PBMCs from each individual subject were barcoded using the palladium-based 20-Plex Pd Barcoding Kit (Fluidigm) according to the manufacturer’s protocol. Barcoded cells were then combined into a single tube before Fc receptor blocking (BD Biosciences) and staining with a cocktail of metal-conjugated antibodies against cell surface markers (Supplementary Table 3) for 30 min at room temperature. Stained cells were washed with CSB buffer and chilled on ice for 5 min. After washing, cells were fixed with 2% paraformaldehyde (PFA) for 10 min at room temperature, washed again and stored overnight at 4 °C. The next day, before running mass cytometry, cells were stained with iridium DNA intercalator (Fluidigm) in Maxpar Fix and Perm Buffer (Fluidigm) for 20 min at room temperature, washed once with CSB followed by two washes with the Maxpar Cell Acquisition Solution (Fluidigm) and filtered through a 40-μm membrane. Cells were then acquired on a Helios mass cytometer (Fluidigm).

CyTOF data pre-processing and analysis

FCS data files were obtained from the Helios instrument. Data were normalized using the Nolan laboratory MATLAB normalizer version 0.3 (http://github.com/nolanlab/bead-normalization/releases) and debarcoded using the Zunder laboratory debarcoder 24 (https://github.com/zunderlab/single-cell-debarcoder). Normalized and debarcoded files were further gated based on barcode stringency parameters and iridium DNA intercalator to remove non-cell debris and cellular aggregates. B cells were identified by manual gating determined by CD19+CD3−CD14−. CyTOF data were analyzed using R framework (version 3.5.0) and Bioconductor (version 3.7). First, protein expression was normalized using arcsinh (cofactor = 5) transformation. Then, we used the self-organizing map method66 for clustering with the number of clusters from two to 30. Cluster robustness was evaluated using the relative change in area under cumulative distribution function using consensus clustering67 that identified 11 clusters. Heat map of average expression for each cluster used a pheatmap R package with expression scale from 0 to 1.

Sort-purifying B cells

Cryopreserved PBMCs were thawed and washed twice with warm complete media (RPMI supplemented with 5% FBS, 1 mM sodium pyruvate and penicillin–streptomycin). Cells were then incubated with Fc receptor blocking (BD Biosciences) and stained with a cocktail of fluorophore-conjugated antibodies against cell surface markers (Supplementary Table 4) for 30 min at room temperature and then washed with FACS buffer (1% FBS, 0.1% NaN3 in PBS). Before fluorescence-activated cell sorting (FACS), cells were stained with 7AAD and filtered through a 40-μm membrane. Cells were sorted on an Influx cell sorter (BD Biosciences) into 2% FBS RPMI media. Sortgating for B27+IgM+CD24hi and B27+IgM+CD24lo/− cells is shown in Extended Data Fig. 6.

Sample preparation for bulk RNA sequencing

Sort-purified B27+IgM+CD24hi (n = 4) and B27+IgM+CD24lo/− (n = 4) cells obtained from healthy volunteers were RNA extracted using the Qiagen RNeasy Plus Kit. The purified RNA was stored at −80 °C and sent to a third-party vendor for sequencing.

DE genes and pathway analysis

RNA sequences in raw FASTQ data files were obtained from a third-party vendor. Sequencing reads were aligned to the reference genome (GRCh38) using HISAT2. The annotated sequences were then quantified and assembled using StringTie, and DE genes were analyzed using the R Ballgown package. Volcano plots of DE genes were visualized using the Python bioinfokit package. Ingenuity pathway analysis was performed on all annotated RNA to analyze differentially regulated cellular processes and canonical pathways.

BCR sequencing of B27+IgM+CD24lo and B27+IgM+CD24hi cells

This BCR sequencing was performed as a part of TotalSeq multi-omics sequencing of PBMCs obtained from three human donors. A detailed protocol on cell staining and optimization can be found in Vallejo et al.68. This TotalSeq panel comprised 192 surface antibodies, 487 genes and V-J usage of B cells. Out of total PBMCs, B cells were first identified by manual gating determined by CD19+CD3−. B27+IgM+CD24lo and B27+IgM+CD24hi cells were then identified with manual gating using CD27+IgM+CD24lo (total of 338 cells) and CD24hi (total of 1,221 cells), respectively. Numbers of reads of V and J regions between B27+IgM+CD24lo and B27+IgM+CD24hi cells were then compared.

MDA mimotope peptide binding assay

PBMCs obtained from human donors were enriched for B cells. First, 200 μg ml−1 engineered MDA dual mimotope peptide conjugated with Strep-tag II (HSWTNSWMATFLSAHSWTNSWMATFLSAWSHPQFEK) and 200 μg ml−1 scrambled peptide were pre-incubated with Strep-Tactin APC and Strep-Tactin PE for 15 min at room temperature. The enriched B cells were then incubated with the pre-incubated engineered peptide or scrambled peptide for 2 h at room temperature. The engineered peptide was synthesized at the Mass General Brigham Peptide/Protein Core at Massachusetts General Hospital. The Strep-Tactin APC and Strep-Tactin PE were purchased from IBA Lifesciences. After the peptide labeling, the cells were blocked with Fc receptor blocker and stained with a cocktail of fluorophore-conjugated antibodies to help identify subtypes of B cells.

AT to NSG mice

B27+IgM+CD24hi and B27+IgM+CD24lo/− cells were sort-purified. After sorting, these cells were rested in 20% FBS RPMI, incubated with either 20 μg ml−1 IgG isotype control or 20 μg ml−1 CD24mAb or stimulated with 100 μg ml−1 MDA peptide mimotope or scrambled peptide69 for 1 h at 37 °C. After in vitro stimulation/treatment, 200,000 of these cells were adoptively IP transferred into NSG mice. Each subclass of B27+IgM+CD24hi or B27+IgM+CD24lo/− cells from each donor was transferred into a unique recipient NSG mouse. These cells were also found to have 95–98% viability measured by methylene blue staining before AT. These NSG mice are on the NOD/ShiLtJ genetic background and carry mutations in the severe combined immune deficiency (scid) and a complete null allele of the IL2 receptor common gamma chain (IL2rgnull). The scid mutation renders the mice B and T cell deficient. The IL2rgnull mutation prevents cytokine signaling through multiple receptors, leading to a deficiency in functional NK cells. Such severe immunodeficiency allows the mice to be humanized by engrafting human PBMCs70. One week after AT, mice were killed, and the PerC, bone marrow and spleen were processed for flow cytometry as previously described18. Harvested cells were blocked with Fc receptor blocker and stained with a cocktail of fluorophore-conjugated antibodies (Supplementary Table 5) and then washed with FACS buffer. Flow cytometry was used to quantify human B27+IgM+CD24hi and B27+IgM+CD24lo/− cells in each compartment of humanized mice.

Human PBMC CD24 knockdown

gRNA molecule-targeting exon of CD24 (sequence GGTGCATCCATACACCTCG) was purchased from Integrated DNA Technologies and conjugated with tracRNA ATTO. Total human PBMCs were enriched for B cells using the EasySep Human Pan-B Cell Enrichment Kit purchased from STEMCELL Technologies and stimulated with 1 μg ml−1 human IL-4 for 2 h. The enriched B cells were then nucleotransfected with Cas9 ribonuclear protein and pre-conjugated CD24 gRNA-tracRNA ATTO using the P2 Primary Cells Nucleofection Kit purchased from Lonza. B cells were collected for analysis and sorting 24 h after nucleofection.

Generation of hyperlipidemic humanized mice

NSG mice were engrafted with 1 × 107 human PBMCs obtained from healthy donors via tail vein injection. A day after PBMC engraftment, mice were intravitreally injected with 1 × 1012 AAV8 (D377Y) PCSK9 to induce gain-of-function PCSK9, allowing induction of hyperlipidmia. Humanized mice were then fed a WD beginning at the time of AA8-PCSK9 delivery for 10 weeks. Blood was drawn after 10 weeks of WD to measure total cholesterol using the Infinity cholesterol colorimetric kit (Thermo Fisher Scientific).

In vivo CD24mAb treatment and 18F-FDG-PET/CT imaging

Hyperlipidemic humanized mice after 10 weeks of WD feeding were treated with 200 μg of CD24mAb or IgG isotype control two times a week for 3 weeks via intravitreal injection. At the end of 3 weeks of treatment, 18F-FDG-PET/CT images were acquired on a Bruker Albira Si scanner. The mice were fasted overnight before the scan. During anesthesia with isoflurane, 10 MBq of 18F-FDG was tail vein injected to the mice. A static PET scan was obtained 30 min after 18F-FDG injection. FDG uptake was quantified using average standardized uptake volume (SUV) performed by PMOD 3.9 software. The region of interest (ROI) for the aorta area was determined using the CT images. The ROI was then transferred to the co-registered PET images to measure SUV.

Statistics

Statistics were calculated using GraphPad Prism version 7.0a (GraphPad Software.), Python 3.0, R 3.6.1 or SAS 9.4. Results from all replicated experiments are displayed, and bar graphs display mean ± s.e.m.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. Unsupervised metalouvain clustering of CD19+ B cells obtained from 28 subjects with high and low plasma levels of IgM to MDA-LDL (n = 14 high IgM and n = 14 low IgM).

Extended Data Fig. 1 |

Open in a new tab

Clustering result indicates 11 distinct clusters with a subject from high IgM to MDA-LDL group representing a potential clonal expansion of cluster 1 (CD20+ CD25hi B cells).

Extended Data Fig. 2 |. Plasma level of IgMMDA-LDL strongly associates with Plasma level of IgMPC-BSA and IgMOxCE.

Extended Data Fig. 2 |

Open in a new tab

a, Pearson correlations between plasma IgMPC-BSA and plasma IgMMDA-LDL. b, Pearson correlations between plasma IgMOxCE and plasma IgMMDA-LDL. Error bands indicate values within 95% confidence interval. p-values were calculated with two-tailed Pearson correlations.

Extended Data Fig. 3 |. Unsupervised clustering of IgM producing B cells (clusters 1 and 8) obtained from 27 subjects with high and low plasma levels of IgM to MDA-LDL.

Extended Data Fig. 3 |

Open in a new tab

a, Representative metacluster UMAP showing 4 distinguished innate B cell subtypes by using Louvain clustering. b-c, Heatmap showing median expression of 24 surface markers from metaclustering (b) and potential phenotypes of 4 innate B cell subtypes (c). d, Representative UMAP showing CD24 expression across 4 innate B cell subtypes, e, Biaxial plots to compare frequency of each subtype as a percentage of total CD27+IgM+ B cells in subjects with high (n = 13) and low (n = 14) IgM MDA-LDL. Data were analyzed using two-sided Mann-Whitney Wilcoxon test. Values are mean ± s.d. Exact p-values were provided above each graph. Boxplots in Extended Data Fig. 3e are defined as the followings: minima = Quartile 1 – 1.5x(interquartile range), maxima = Quartile 3 + 1.5x(interquartile range), center = median, lower bound of box = Quartile 1, and upper bound of box = Quartile 3.

Extended Data Fig. 4 |. Pearson correlations between geometric mean (GM) of all CyTOF panel’s surface markers on B27+IgM+ and plasma IgMMDA-LDL.

Extended Data Fig. 4 |

Open in a new tab

Error bands indicate values within 95% confidence interval. p-values were calculated with two-tailed Pearson correlations.

Extended Data Fig. 5 |. Pearson correlations between geometric mean (GM) of CD24 on B27+IgM+ and plasma total IgM, IgM to OxCE, IgM to HMGB1, and IgM to ALDH4a1.

Extended Data Fig. 5 |

Open in a new tab

Error bands indicate values within 95% confidence interval. p-values were calculated with two-tailed Pearson correlations.

Extended Data Fig. 6 |. Sorting strategy for B27−, B27+IgM+CD24l°/−, and B27+IgM+CD24hi.

Extended Data Fig. 6 |

Open in a new tab

Human PBMCs were enriched for B cells and sorted for CD20CD27−, CD20 + CD27+IgM+CD24lo, and CD20 + CD27+IgM+CD24hi.

Extended Data Fig. 7 |. Pearson correlations between frequency of B27+IgM+CD24lo/− (a) and B27+IgM+CD24hi (b) and plasma IgMMDA-LDL in coronary artery disease (CAD) patients.

Extended Data Fig. 7 |

Open in a new tab

Error bands indicate values within 95% confidence interval. p-values were calculated with two-tailed Pearson correlations.

Extended Data Fig. 8 |. Confirmation of CD24 knockdown by RT-qPCR.

Extended Data Fig. 8 |

Open in a new tab

B27+IgM+CD24hi cells from 5 human donors were nucleo-transfected by Cas9/CD24 non-targeted or CD24 targeted trancrRNA-crRNA. Relative expression of CD24 mRNA was measured by RT-qPCR and corrected by GAPDH housekeeping gene. Data were analyzed by using two-sided Mann-Whitney Wilcoxon test. Values are mean ± s.d. Exact p-values were provided above each graph.

Supplementary Material

Reporting Summary
Sup. Table 1
Source Data 1
Source Data 2
Source Data 3
Source Data 4
Source Data 5
Source Data 6

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s44161-023-00356-1.

Acknowledgements

We thank M. Solga and C. Chew from the University of Virginia Flow Cytometry Core for their excellent technical assistance. We thank L. Erickson (University of Virginia Carter Immunology Center) for his invaluable advice on B cell phenotypes and S. Bekiranov (University of Virginia Biochemistry and Molecular Genetics) for his insights on RNA sequencing analysis. This work was supported by National Institutes of Health grants R01HL136098 and R01-HL148109 (C.A.M.) and P01HL136275 (C.C.H., K.L., C.A.M., Y.M., A.M.T. and S.T.) and a LeDucq Foundation Transatlantic Network of Excellence grant: ‘B cells in cardiovascular disease’ (C.A.M.). T.P. is a recipient of an American Heart Association pre-doctoral fellowship.

Footnotes

Competing interests

The authors declare no competing interests.

Peer review information Nature Cardiovascular Research thanks Dennis Wolf and the other, anonymous, reviewers for their contribution to the peer review of this work. Primary Handling Editor: Vesna Todorovic, in collaboration with the Nature Cardiovascular Research team.

Data availability

The CyTOF raw data and RNA sequencing raw data are available at https://zenodo.org/record/8318928. Human genome assembly GRCh38, used to align RNA sequencing data, is available at https://genome.ucsc.edu/cgi-bin/hgGateway.

Other additional CAVA cohort surface marker data for Fig. 7c are available at the Gene Expression Omnibus under accession number GSE190570. De-identified human subjects clinical and imaging data will be released to others after the establishment of a data transfer agreement between the University of Virginia Health System and the requesting institution. The contact author of this manuscript, C.A.M., will serve as the contact for access requests and will respond to requests within 2 weeks. The shared human subjects clinical and imaging data will need to be treated confidentially and used only for research purposes.

Code availability

CyTOF data were normalized using the Nolan laboratory MATLAB normalizer version 0.3 with the code available at http://github.com/nolanlab/bead-normalization/releases. CyTOF data were also de-barcoded using the Zunder laboratory de-barcoder 24, with the code available at https://github.com/zunderlab/single-cell-debarcoder.

References

  • 1.Ikura Y et al. Localization of oxidized phosphatidylcholine in nonalcoholic fatty liver disease: impact on disease progression. Hepatology 43, 506–514 (2006). [DOI] [PubMed] [Google Scholar]
  • 2.Bossola M et al. Oxidized low-density lipoprotein biomarkers in patients with end-stage renal failure: acute effects of hemodialysis. Blood Purif. 25, 457–465 (2008). [DOI] [PubMed] [Google Scholar]
  • 3.Birukova AA et al. Polar head groups are important for barrier-protective effects of oxidized phospholipids on pulmonary endothelium. Am. J. Physiol. Lung Cell. Mol. Physiol. 292, L924–L935 (2007). [DOI] [PubMed] [Google Scholar]
  • 4.Tsimikas S & Witztum JL The role of oxidized phospholipids in mediating lipoprotein(a) atherogenicity. Curr. Opin. Lipidol. 19, 369–377 (2008). [DOI] [PubMed] [Google Scholar]
  • 5.Sottero B, Rossin D, Poli G & Biasi F Lipid oxidation products in the pathogenesis of inflammation-related gut diseases. Curr. Med. Chem. 25, 1311–1326 (2017). [DOI] [PubMed] [Google Scholar]
  • 6.Fraley AE et al. Relationship of oxidized phospholipids and biomarkers of oxidized low-density lipoprotein with cardiovascular risk factors, inflammatory biomarkers, and effect of statin therapy in patients with acute coronary syndromes. Results from the MIRACL (Myocardial Ischemia Reduction With Aggressive Cholesterol Lowering) trial. J. Am. Coll. Cardiol. 53, 2186–2196 (2009). [DOI] [PubMed] [Google Scholar]
  • 7.Feng X et al. ApoE−/−Fas−/− C57BL/6 mice: a novel murine model simultaneously exhibits lupus nephritis, atherosclerosis, and osteopenia. J. Lipid Res. 48, 794–805 (2007). [DOI] [PubMed] [Google Scholar]
  • 8.Qin J, Goswami R, Balabanov R & Dawson G Oxidized phosphatidylcholine is a marker for neuroinflammation in multiple sclerosis brain. J. Neurosci. Res. 85, 977–984 (2007). [DOI] [PubMed] [Google Scholar]
  • 9.Ross BM, Mamalias N, Moszczynska A, Rajput AH & Kish SJ Elevated activity of phospholipid biosynthetic enzymes in substantia nigra of patients with Parkinson’s disease. Neuroscience 102, 899–904 (2001). [DOI] [PubMed] [Google Scholar]
  • 10.Que X et al. Oxidized phospholipids are proinflammatory and proatherogenic in hypercholesterolaemic mice. Nature 558, 301–306 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Ravandi A et al. Relationship of IgG and IgM autoantibodies and immune complexes to oxidized LDL with markers of oxidation and inflammation and cardiovascular events: results from the EPIC-Norfolk study. J. Lipid Res. 52, 1829–1836 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Tsimikas S et al. Oxidation-specific biomarkers, prospective 15-year cardiovascular and stroke outcomes, and net reclassification of cardiovascular events. J. Am. Coll. Cardiol. 60, 2218–2229 (2012). [DOI] [PubMed] [Google Scholar]
  • 13.Tsimikas S et al. Relationship of IgG and IgM autoantibodies to oxidized low density lipoprotein with coronary artery disease and cardiovascular events. J. Lipid Res. 48, 425–433 (2007). [DOI] [PubMed] [Google Scholar]
  • 14.Srikakulapu P et al. Perivascular adipose tissue harbors atheroprotective IgM-producing B cells. Front. Physiol. 8, 719 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.van den Berg VJ et al. IgM anti-malondialdehyde low density lipoprotein antibody levels indicate coronary heart disease and necrotic core characteristics in the Nordic Diltiazem (NORDIL) study and the Integrated Imaging and Biomarker Study 3 (IBIS-3). EBioMedicine 36, 63–72 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Rosenfeld ME, Palinski W, Ylä-Herttuaia S, Butler S & Witztum JL Distribution of oxidation specific lipid-protein adducts and apolipoprotein B in atherosclerotic lesions of varying severity from WHHL rabbits. Arterioscler. Thromb. Vasc. Biol. 10, 336–349 (1990). [DOI] [PubMed] [Google Scholar]
  • 17.Senders ML et al. PET/MR imaging of malondialdehyde-acetaldehyde epitopes with a human antibody detects clinically relevant atherothrombosis. J. Am. Coll. Cardiol. 71, 321–335 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Rosenfeld SM et al. B-1b cells secrete atheroprotective IgM and attenuate atherosclerosis. Circ. Res. 117, e28–e39 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Upadhye A et al. Diversification and CXCR4-dependent establishment of the bone marrow B-1a cell pool governs atheroprotective IgM production linked to human coronary atherosclerosis. Circ. Res. 125, e55–e70 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Döring Y et al. B-cell-specific CXCR4 protects against atherosclerosis development and increases plasma IgM levels. Circ. Res. 126, 787–788 (2020). [DOI] [PubMed] [Google Scholar]
  • 21.Doran AC et al. B-cell aortic homing and atheroprotection depend on Id3. Circ. Res. 110, e1 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Amu S, Tarkowski A, Dörner T, Bokarewa M & Brisslert M The human immunomodulatory CD25+ B cell population belongs to the memory B cell pool. Scand. J. Immunol. 66, 77–86 (2007). [DOI] [PubMed] [Google Scholar]
  • 23.Brisslert M et al. Phenotypic and functional characterization of human CD25+ B cells. Immunology 117, 548–557 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Kaminski DA, Wei C, Qian Y, Rosenberg AF & Sanz I Advances in human B cell phenotypic profiling. Front. Immunol. 3, 302 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Weill JC & Reynaud CA IgM memory B cells: specific effectors of innate-like and adaptive responses. Curr. Opin. Immunol. 63, 1–6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Appelgren D, Eriksson P, Ernerudh J & Segelmark M Marginal-zone B-cells are main producers of IgM in humans, and are reduced in patients with autoimmune vasculitis. Front. Immunol. 9, 2242 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Tull TJ et al. Human marginal zone B cell development from early T2 progenitors. J. Exp. Med. 218, e20202001 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Siu JHY et al. Two subsets of human marginal zone B cells resolved by global analysis of lymphoid tissues and blood. Sci. Immunol. 7, eabm9060 (2022). [DOI] [PubMed] [Google Scholar]
  • 29.Glass DR et al. An integrated multi-omic single-cell atlas of human B cell identity. Immunity 53, 217–232 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Weller S et al. Human blood IgM ‘memory’ B cells are circulating splenic marginal zone B cells harboring a prediversified immunoglobulin repertoire. Blood 104, 3647–3654 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Proto JD et al. Hypercholesterolemia induces T cell expansion in humanized immune mice. J. Clin. Invest. 128, 2370–2375 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Rampidis GP, Benetos G, Benz DC, Giannopoulos AA & Buechel RR A guide for Gensini Score calculation. Atherosclerosis 287, 181–183 (2019). [DOI] [PubMed] [Google Scholar]
  • 33.Binder CJ et al. Pneumococcal vaccination decreases atherosclerotic lesion formation: molecular mimicry between Streptococcus pneumoniae and oxidized LDL. Nat. Med. 9, 736–743 (2003). [DOI] [PubMed] [Google Scholar]
  • 34.Shaw PX et al. Natural antibodies with the T15 idiotype may act in atherosclerosis, apoptotic clearance, and protective immunity. J. Clin. Invest. 105, 1731–1740 (2000). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Palinski W, Miller E & Witztum JL Immunization of low density lipoprotein (LDL) receptor-deficient rabbits with homologous malondialdehyde-modified LDL reduces atherogenesis. Proc. Natl Acad. Sci. USA 92, 821–825 (1995). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Faria-Neto JR et al. Passive immunization with monoclonal IgM antibodies against phosphorylcholine reduces accelerated vein graft atherosclerosis in apolipoprotein E-null mice. Atherosclerosis 189, 83–90 (2006). [DOI] [PubMed] [Google Scholar]
  • 37.Chou MY et al. Oxidation-specific epitopes are dominant targets of innate natural antibodies in mice and humans. J. Clin. Invest. 119, 1335–1349 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Harmon DB et al. Protective role for B-1b B cells and IgM in obesity-associated inflammation, glucose intolerance, and insulin resistance. Arterioscler. Thromb. Vasc. Biol. 36, 682–691 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Masmoudi H, Mota-Santos T, Huetz F, Coutinho A & Cazenave PA All T15 Id-positive antibodies (but not the majority of VHT15+ antibodies) are produced by peritoneal CD5+ B lymphocytes. Int. Immunol. 2, 515–520 (1990). [DOI] [PubMed] [Google Scholar]
  • 40.Tsiantoulas D, Gruber S & Binder CJ B-1 cell immunoglobulin directed against oxidation-specific epitopes. Front. Immunol. 3, 415 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Kyaw T et al. B1a B lymphocytes are atheroprotective by secreting natural IgM that increases IgM deposits and reduces necrotic cores in atherosclerotic lesions. Circ. Res. 109, 830–840 (2011). [DOI] [PubMed] [Google Scholar]
  • 42.Shen L et al. B-1a lymphocytes attenuate insulin resistance. Diabetes 64, 593–603 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Gonen A et al. Atheroprotective immunization with malondialdehyde-modified LDL is hapten specific and dependent on advanced MDA adducts: implications for development of an atheroprotective vaccine. J. Lipid Res. 55, 2137–2155 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.George J et al. Hyperimmunization of apo-E-deficient mice with homologous malondialdehyde low-density lipoprotein suppresses early atherogenesis. Atherosclerosis 138, 147–152 (1998). [DOI] [PubMed] [Google Scholar]
  • 45.Martos-Folgado I et al. MDA-LDL vaccination induces atheroprotective germinal-center-derived antibody responses. Cell Rep. 41, 111468 (2022). [DOI] [PubMed] [Google Scholar]
  • 46.Björkbacka H et al. Low levels of apolipoprotein B-100 autoantibodies are associated with increased risk of coronary events. Arterioscler. Thromb. Vasc. Biol. 36, 765–771 (2016). [DOI] [PubMed] [Google Scholar]
  • 47.Kalinina N et al. Increased expression of the DNA-binding cytokine HMGB1 in human atherosclerotic lesions. Arterioscler. Thromb. Vasc. Biol. 24, 2320–2325 (2004). [DOI] [PubMed] [Google Scholar]
  • 48.Lorenzo C et al. ALDH4A1 is an atherosclerosis auto-antigen targeted by protective antibodies. Nature 589, 287–292 (2021). [DOI] [PubMed] [Google Scholar]
  • 49.Martin F & Kearney JF B1 cells: similarities and differences with other B cell subsets. Curr. Opin. Immunol. 13, 195–201 (2001). [DOI] [PubMed] [Google Scholar]
  • 50.Berkowska MA et al. Human memory B cells originate from three distinct germinal center-dependent and -independent maturation pathways. Blood 118, 2150–2158 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Victora GD & Nussenzweig MC Germinal centers. Annu. Rev. Immunol. 30, 429–457 (2012). [DOI] [PubMed] [Google Scholar]
  • 52.Mesin L, Ersching J & Victora GD Germinal center B cell dynamics. Immunity 45, 471–482 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Cerutti A, Cols M & Puga I Marginal zone B cells: virtues of innate-like antibody-producing lymphocytes. Nat. Rev. Immunol. 13, 118–132 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Palm AKE & Kleinau S Marginal zone B cells: from housekeeping function to autoimmunity? J. Autoimmun. 119, 102627 (2021). [DOI] [PubMed] [Google Scholar]
  • 55.Martin F, Oliver AM & Kearney JF Marginal zone and B1 B cells unite in the early response against T-independent blood-borne particulate antigens. Immunity 14, 617–629 (2001). [DOI] [PubMed] [Google Scholar]
  • 56.Grasset EK et al. Sterile inflammation in the spleen during atherosclerosis provides oxidation-specific epitopes that induce a protective B-cell response. Proc. Natl Acad. Sci. USA 112, E2030–E2038 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Shaw PX, Goodyear CS, Chang MK, Witztum JL & Silverman GJ The autoreactivity of anti-phosphorylcholine antibodies for atherosclerosis-associated neo-antigens and apoptotic cells. J. Immunol. 170, 6151–6157 (2003). [DOI] [PubMed] [Google Scholar]
  • 58.Steiniger B, Timphus EM & Barth PJ The splenic marginal zone in humans and rodents: an enigmatic compartment and its inhabitants. Histochem. Cell Biol. 126, 641–648 (2006). [DOI] [PubMed] [Google Scholar]
  • 59.Balázs M, Martin F, Zhou T & Kearney JF Blood dendritic cells interact with splenic marginal zone B cells to initiate T-independent immune responses. Immunity 17, 341–352 (2002). [DOI] [PubMed] [Google Scholar]
  • 60.Nus M et al. Marginal zone B cells control the response of follicular helper T cells to a high-cholesterol diet. Nat. Med. 23, 601–610 (2017). [DOI] [PubMed] [Google Scholar]
  • 61.Nus M et al. NR4A1 deletion in marginal zone B cells exacerbates atherosclerosis in mice—Brief Report. Arterioscler. Thromb. Vasc. Biol. 40, 2598–2604 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Yanyi S et al. Splenic marginal zone B lymphocytes regulate cardiac remodeling after acute myocardial infarction in mice. J. Am. Coll. Cardiol. 79, 632–647, 10.1016/j.jacc.2021.11.051 (2022). [DOI] [PubMed] [Google Scholar]
  • 63.Jarr KU et al. 18F-Fluorodeoxyglucose-positron emission tomography imaging detects response to therapeutic intervention and plaque vulnerability in a murine model of advanced atherosclerotic disease—brief report. Arterioscler. Thromb. Vasc. Biol. 40, 2821–2828 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Fang X, Zheng P, Tang J & Liu Y CD24: from A to Z. Cell. Mol. Immunol. 7, 100–103 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Gonen A et al. A monoclonal antibody to assess oxidized cholesteryl esters associated with apoAI and apoB-100 lipoproteins in human plasma1. J. Lipid Res. 60, 436–445 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 66.Van Gassen S et al. FlowSOM: using self-organizing maps for visualization and interpretation of cytometry data. Cytometry A 87, 636–645 (2015). [DOI] [PubMed] [Google Scholar]
  • 67.Wilkerson MD & Hayes DN ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics 26, 1572–1573 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Vallejo J et al. Combined protein and transcript single-cell RNA sequencing in human peripheral blood mononuclear cells. BMC Biol. 20, 193 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Amir S et al. Peptide mimotopes of malondialdehyde epitopes for clinical applications in cardiovascular disease. J. Lipid Res. 53, 1316–1326, 10.1194/jlr.M025445 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Shultz LD et al. Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2Rγnull mice engrafted with mobilized human hemopoietic stem cells. J. Immunol. 174, 6477–6489 (2005). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Reporting Summary
NIHMS2160349-supplement-Reporting_Summary.pdf (93.3KB, pdf)
Sup. Table 1
NIHMS2160349-supplement-Sup__Table_1.xlsx (19.6KB, xlsx)
Source Data 1
NIHMS2160349-supplement-Source_Data_1.xlsx (13.1KB, xlsx)
Source Data 2
NIHMS2160349-supplement-Source_Data_2.xlsx (9.4KB, xlsx)
Source Data 3
NIHMS2160349-supplement-Source_Data_3.xlsx (11.2KB, xlsx)
Source Data 4
NIHMS2160349-supplement-Source_Data_4.xlsx (9.8KB, xlsx)
Source Data 5
NIHMS2160349-supplement-Source_Data_5.xlsx (8.9KB, xlsx)
Source Data 6
NIHMS2160349-supplement-Source_Data_6.xlsx (9.3KB, xlsx)

Data Availability Statement

The CyTOF raw data and RNA sequencing raw data are available at https://zenodo.org/record/8318928. Human genome assembly GRCh38, used to align RNA sequencing data, is available at https://genome.ucsc.edu/cgi-bin/hgGateway.

Other additional CAVA cohort surface marker data for Fig. 7c are available at the Gene Expression Omnibus under accession number GSE190570. De-identified human subjects clinical and imaging data will be released to others after the establishment of a data transfer agreement between the University of Virginia Health System and the requesting institution. The contact author of this manuscript, C.A.M., will serve as the contact for access requests and will respond to requests within 2 weeks. The shared human subjects clinical and imaging data will need to be treated confidentially and used only for research purposes.

CyTOF data were normalized using the Nolan laboratory MATLAB normalizer version 0.3 with the code available at http://github.com/nolanlab/bead-normalization/releases. CyTOF data were also de-barcoded using the Zunder laboratory de-barcoder 24, with the code available at https://github.com/zunderlab/single-cell-debarcoder.

RESOURCES