Background

Myasthenia gravis (MG) is an autoimmune disease characterized by fatigability and weakness in skeletal muscles [1]. As an autoimmune disease that mainly affects the postsynaptic membrane at the neuromuscular junction, MG is categorized by the pathogenic autoantibodies targeting various post-synaptic components in neuromuscular junctions, including acetylcholine receptor (AChR), muscle-specific kinase (MuSK), and low-density lipoprotein receptor-related protein (LRP4) [2]. The overall prevalence of MG is 12.4 people per million, with a refractory rate of 10% and a relatively heavy social burden [3, 4]. Myasthenic crisis (MC) is a life-threatening state of MG patients [5]. Within the first 2 years after the diagnosis of MG, around 15–20% of the patients may develop MC, presenting rapid worsening dysphagia and respiratory failure which entails intubation or noninvasive ventilation [6]. The mortality rates of MC are heterogeneous among varied regions, reported from 5 to 22% [7, 8]. Though the AChR antibody is the most prevalent (~ 80%) autoantibody in MG, all serotypes of MG still have the potential to develop MC [9]. To date, a variety of precipitants have been demonstrated to be associated with MC, including a concurrent infection, thymectomy, pregnancy, childbirth, or tapering of immunotherapies [10].

The pathogenesis of MG is multifaceted, involving autoimmune cells, pro-inflammatory cytokines, complement factors, and autoantibodies. While the direct cause is often attributed to autoantibodies, such as the AChR antibody, attacking the target receptors on the postsynaptic membrane at the neuromuscular junction, other factors also contribute to this impairment. For instance, autoreactive T cells of MG were characterized by an increased production of interleukin-17 (IL-17), Interferon-gamma (IFN-γ), granulocyte–macrophage colony-stimulating factor (GM-CSF), a decreased production of interleukin-10 (IL-10) [11]. A shift has been observed in circulating follicular helper T cells towards T helper 2 (Th2) and T helper 17 (Th17) dominance over T helper 1 (Th1) in MG [12]. Additionally, serum levels of B-cell-activating factor were elevated in MG patients [13]. Immunoglobulin G (IgG) and complement 3 (C3) have been identified as co-localizing on the postsynaptic membrane, and membrane attack complexes (MACs) have been detected at the muscle end-plates [14].

Consistent with the immunological findings in MG patients, pro-inflammatory CD4+ T signatures were identified during MC, in particular elevated Th1 and Th17 subsets, from our previous study [15]. Notably, hypercytokinemia is a newly identified immune feature that dominates the crisis, with pan-elevation of cytokines associated with Th1, Th2, Th17, Th9, and Tregs. Hypercytokinemia is also a special feature that is commonly seen in the acute phase or the most worsening stages in immune-dysregulated diseases, including severe systemic inflammation, acute respiratory syndrome coronavirus 2 (COVID-19), and system lupus erythematosus [16, 17]. The mainstay fast-acting immunomodulatory therapies to benefit patients with MC include plasma exchange (PE) and intravenous immunoglobulin (IVIG). The superior efficacy of PE in rapidly alleviating respiratory failure compared to intravenous IVIG was observed in a prospective MC cohort, a finding also seen in other diseases characterized by a cytokine storm, such as COVID-19 [15, 18, 19]. This observation supports the hypothesis that hypercytokinemia might play an important role in the pathogenesis of MC. Theoretically, PE can directly remove inflammatory cytokines, whereas IVIG works indirectly by modulating the immune system to suppress inflammation [20]. However, the development of hypercytokinemia during MC remained unknown and in-hospital mortality is still high even though timely immunotherapies were provided.

Explorations in the immune pathogenesis of MG were mainly restrained to the selected immune cell types, while numerous components, particularly those from the innate immune system, remained largely unexplored [21]. Single-cell transcriptomic sequencing has emerged in recent years with a superiority to traditional bulk RNA sequencing for its higher resolution and larger data size in acquiring an immune landscape [22]. The first application of this technique in the research field of MG was recently conducted [23], in which CD180− B cells were highlighted as to be associated with disease activity and antibody titer. However, it only included female early-onset MG patients. As the most devastating state of MG, MC has been rarely investigated from a single-cell perspective.

In this study, we profile peripheral blood mononuclear cells (PBMCs) isolated from AChR + MC patients from the acute phase to three months after MC at single-cell resolution. By integrating the changes in individual cells using single-cell transcriptomic and immune repertoire sequencing, we attempt to delineate the immune landscape in MC and explore the cell type which was the most pro-inflammatory during the crisis. We identify a novel monocyte subset, monocytes 3, as the potential key driver for activating the downstream pro-inflammatory cells including neutrophils, T, B, and natural killer (NK) cells. The existence of a high neutrophil-to-lymphocyte (NLR) ratio is then validated in a prospective AChR + MC cohort and further supports this hypothesis.

Methods

Patients

Two AChR + MC patients (male, 70 and 73 years old) were treated and recruited from Huashan Hospital from October 2021 to June 2022. The study was reviewed and approved by the institutional review board of General Huashan hospital Fudan University (2020–883). Informed consent was obtained from each participant. All AChR + MG patients were diagnosed according to the 2020 MGFA guidance, while MC was defined as an exacerbation of myasthenic symptoms with bulbar or general weakness requiring mechanical ventilation [6, 24]. The participants included for the single-cell transcriptomic and immune repertoire analysis received corticosteroids and immunomodulatory therapies during MC (details in Additional file 2: Table S1). The precipitating factors for MC in these patients were post-thymectomy MC and upper respiratory infection, respectively. To analyze the overall cell proportional changes, the hemogram of 22 AChR + MC patients during MC and three months after MC was analyzed (Additional file 2: Table S2). The average age of these 22 patients was 52.77 ± 15.58 years old with a female-to-male ratio of 6:5.

Sample preparation and single-cell sequencing

For each MC participant, 2  mL venous blood was collected in EDTA anticoagulant tubes and transferred to the laboratory with ice 2 times. The first time point was the initiation of intubation or mechanical ventilation, and the second time point was 3 months after the crisis. PBMCs were isolated by density gradient centrifugation using the Ficoll-Paque medium. The cell viability should exceed 90% which was determined with trypan blue staining. An appropriate volume of cell suspension was calculated to contain ~ 13,000 cells for each sample.

The PBMCs were loaded onto the 10 × Chromium Single Cell Platform (10X Genomics) (5' and V(D)J Enrichment) as described in the manufacturer’s protocol. Generation of gel beads in emulsion (GEMs), barcoding, GEM-RT clean-up, complementary DNA amplification, and library construction were all performed as per the manufacturer’s protocol. Qubit was used for library quantification before pooling. The final library pool was sequenced on the Illumina NovaSeq 6000 using 150 base-pair paired-end reads. The cellranger pipeline (version 6.1.1) was applied to analyze the sequencing raw data, with references of the transcriptome (GRCh38-3.0.0) and VDJ repertoire (vdj_GRCh38_alts_ensembl-4.0.0). An overall summary of sorted cells is recorded in Additional file 2: Table S3.

scRNA-seq data analysis

The analysis of the single-cell dataset was conducted using Scanpy (version 1.9.1) [25], running on Python (version 3.8.13). The batch effect across different samples was adjusted using the “harmony_integrate” function in Scanpy (Additional file 1: Fig. S1) [26]. Doublet cells were detected and removed with the use of scvi-tools (version 0.20.3) [27]. The preprocessing was performed according to standard procedures detailed on Scanpy’s website (https://scanpy-tutorials.readthedocs.io/en/latest/pbmc3k.html). These included: removing cells with fewer than 200 expressed genes, eliminating genes expressed in fewer than 3 cells, discarding cells where the proportion of mitochondrial gene count exceeded 20%, and retaining gene counts ranging from 0.02 to 0.98. The expression data were then normalized, logarithmized, and the effects of total counts per cell were regressed out, followed by scaling to the same range. RNA velocity was calculated using velocyto and scvelo, applying the bam file from raw data for pseudotime-trajectory analysis [28]. After these, a recently published automated cell type annotation tool “Celltypist” was applied to help classify cell types in the PBMCs [29]. Two MC related single-cell RNA (ScRNA) cell-prediction models were used to help annotate cell types in our data. One is the build-in high-resolution immune cell annotation model (Immune_All_Low.pkl) in the Celltypist, and the other is a transfer-learning model generated by Celltypist using a recent published MG ScRNA dataset from Japan (only the PBMCs dataset was used) [30] (Additional file 1: Fig. S2). The marker genes for different cell types had consulted those used in previously published high-quality ScRNA articles (monocytes [31], T cells [30], and other cells [29, 30]) and are listed in Additional file 2: Table S4.

Differentially expressed genes (DEG) were defined by adjusted p value less than 0.05 and log2foldchange larger than 0.5. DEG gene sets were further enriched in different databases to explain their biological functions, including Ingenuity Pathway Analysis (IPA), CellMarker, Gene Ontology (GO), WikiPathways, and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases [32,33,34,35,36]. To explore cell-to-cell interactions, CellChat (version 1.6.0) running on R (Version 4.1.0) was applied to investigate the communication among PBMCs at MC, by utilizing their ligands, receptors and cofactors information (Additional file 1: Fig. S8, S9) [37]. To compare the monoclonal proliferations in B and T cells, Scirpy (version 0.11) was applied in the VDJ sequencing B-cell receptor (ScBCR) and T-cell receptor (ScTCR) data, according to its tutorials (https://scverse.org/scirpy/latest/tutorials/tutorial_3k_tcr.html). Only cells with single pair, extra VJ, extra VDJ, and two full chains were preserved in the VDJ analysis. Clonotypes were defined using the scirpy.tl.define_clonotypes function.

Statistical analysis

Continuous variables are reported as mean (SD) and compared by paired Wilcoxon or T tests accordingly. Correlation was conducted by the Pearson method. When multiple comparisons were involved, Bonferroni method was applied to adjust the p values. Other R packages used for figure generating included tidyverse, ggpubr, cowplot, RColorBrewer. Other Python packages used for data analysis included gseapy, researchpy, numpy, pandas, seaborn, matplotlib.

Results

Study design and single-cell RNA profiling of PBMCs during and after MC

We longitudinally collected 4 fresh peripheral blood samples derived from 2 AChR + patients during and after the MC (P1-P2, Additional file 2: Table S1). The patients were diagnosed according to the 2020 MGFA guidelines. For each participant, the first blood sample was taken within 3 days after intubation for respiratory support. The second sample was obtained 3 months after the crisis. All 2 patients respond well to the immunosuppressive/immunomodulatory therapies.

For these PBMCs, 33577 high-quality cells were identified in ScRNA-seq using the 10 × Genomics platform (Additional file 2: Table S3, Additional file 1: Fig. S1, S2). Additionally, 1159 B cells of single-cell scBCR-seq and 6845 T cells of ScTCR-seq were identified from the same samples as well. A total of 25 cell types were identified in ScRNA data, including (1) CD8 Tnaive (1.36%), (2) CD8 Tem (4.49%), (3) CD4 Tnaive (3.29%), (4) CD4 Tcm (Th0) (5.56%), (5) CD4 Tcm (Th2) (0.22%), (6) CD4 Tcm (Th17) (1.88%), (7) CD4 Tem (Th1/17) (0.34%), (8) CD4 Tem (Th1) (3.24%), (9) CD4 Temra (Th1) (6.38%), (10) Naive Treg (0.70%), (11) Activated Treg (1.14%), (12) Monocytes 1 (34.26%), (13) Monocytes 2 (5.81%), (14) Monocytes 3 (6.39%), (15) Monocytes 4 (4.70%), (16) Age-associated B (0.38%), (17) Memory B (2.03%), (18) Naive B (1.19%), (19) Plasma cells (0.19%), (20) Neutrophils (1.73%), (21) Neutrophil-myeloid progenitor (0.23%), (22) NK (13.13%), (23) DC2 (0.17%), (24) Megakaryocytes (1.08%), (25) Hematopoietic stem cells and multipotent progenitors (HSC/MPP) (0.08%) (Fig. 1). Monocytes, especially monocytes 1, exhibited the highest difference during the crisis, compared with that at three months after MC (Additional file 1: Fig. S3).

Fig. 1
figure 1

Study design and single-cell landscape of cell types identified from myasthenic crisis (MC). a. The design of the study. This is a self-comparative study. The single-cell RNA sequencing (ScRNA-seq) was performed on peripheral blood mononuclear cells (PBMCs) from three (acetylcholine receptor antibody positive) AChR + myasthenic crisis (MC) patients. The sequencing was conducted at the most acute stage (on ventilation) and three months after the crisis. Hemogram results from 22 AChR + MC patients were also analyzed to obtain an overview of the whole blood. b. A total of 33577 PBMC cells are identified in ScRNA-seq, among which 18114 are from the acute stage of MC and 15463 from three months after MC. The dimensional reduction is performed with the uniform manifold approximation and projection (UMAP). c. The single-cell landscape at the acute stage of MC. d. The single-cell landscape after three months after the crisis. Treg regulatory T; Tcm central memory T; Tem effector memory T; Temra terminally differentiated effector memory T; DC dendritic cell; NK natural killer

Full size image

Given the relatively small sample size of the current single-cell RNA study, we mainly focused on identifying the primary cell type with pro-inflammatory function. This could potentially serve as a therapeutic target in clinical settings.

Innate immune activation identified at MC in PBMC ScRNA data

The analysis of differentially expressed genes (DEG) in all PBMCs revealed an activation of the innate immune system and simultaneously, a suppression of the adaptive immune system during the crisis, compared with those three months after MC (Fig. 2a, Additional file 2: Table S5). The highly ranked pathogen-induced cytokine storm signaling pathway (-log p-value 8.44) might suggest that elevated cytokines could be a potential contributing factor during the crisis. The suppressed Th1 and Th2 activation signaling (-log p-value 10.50) implied a T cell exhaustion state at MC, while activated neutrophil pathways implied a potential neutrophil-induced inflammation at MC. The top-ranked genes by fold changes also revealed an activation state of immune-regulator genes at MC (Additional file 1: Fig. S4a), including VSIG4 (an immune checkpoint) [38] and SIGLEC1 (a type I interferon biomarker) [39]. However, the excessive expressions of pro-inflammatory genes, such as C1QA, C1QB, C1QC (complement activation) [40], indicated sustained inflammatory activation even under the background of immunosuppressive treatments. To better characterize the condition of MC, up-regulated (n = 440, Fig. 2b) and down-regulated (n = 903, Fig. 2c) gene sets were independently analyzed in further databases. CellMarker database [32] indicated monocyte activation and lymphocytes suppression at MC. Additionally, up-regulated neutrophil activation pathways, cytokines-mediated signaling pathway, and toll-like receptor signaling pathway of monocytes also supported an activated innate immune response. Conversely, genes related to T cell activation and function were enriched in the suppressed pathways, as shown in the KEGG and Wikipathway results (Additional file 1: Fig. S4b, c).

Fig. 2
figure 2

Hemogram results and overall differentially expressed genes (DEG) analysis of all PBMCs. a. The DEG analysis of all MC PBMCs (at MC vs. three months after MC). Myeloid cell pathways activation indicated an innate immune activation, while lymphoid cell pathways suppression indicated an adaptive immune inhibition. b. Enrichment analysis of MC up-regulated genes (n = 440) in different databases. The CellMarker enrichment indicates monocytes were the most activated cell type. c. Enrichment analysis of MC down-regulated genes (n = 903) in different databases. The CellMarker enrichment indicates lymphocytes were the most repressed cell types. d. the hemogram results of 22 AChR+ MC patients between the acute stage at MC and three months after MC (Bonferroni adjusted p-values). Extremely neutrophilia and lymphopenia comprise the drastically higher neutrophil–lymphocyte ratio (NLR) at MC. e. The correlation relationships between NLR and MG-related clinical scores (Bonferroni adjusted p-values). Three (ADL, MMT, QMG) of four scores show the positive relationships between NLR and clinical severity. QMG, Myasthenia Gravis Foundation of America quantitative myasthenia gravis score; MMT, MG-specific manual muscle testing; ADL, Myasthenia Gravis Activities of Daily Living Scale; QOL-15, Myasthenia Gravis quality of life questionnaire

Full size image

NLR was higher at MC and associated with the clinical scores

To achieve a general picture of the peripheral cellular changes, we retrospectively reviewed the hemogram changes in AChR + MC patients (n = 22) at MC and three months after MC. Consistent with previous results in MG studies [41, 42], a significantly higher NLR (0.18 ± 0.14%) was identified at the acute stage of MC patients, compared to that at 3 months after MC (0.05 ± 0.03%, adjusted p = 4.0e-05) (Fig. 2d, Additional file 2: Table S2). This result can be explained by the concurrent neutrophilia (10.13 ± 4.20 10^9/L) and lymphopenia (0.92 ± 0.79 10^9/L) at the MC. Though the lymphocyte counts gradually restored to a normal range after the crisis, the neutrophil counts still exceeded the normal range. To potentially rule out the influence of concurrent infections on the escalation of NLR during MC, we conducted a subgroup analysis by dividing these 22 patients into those with concurrent infectious diseases (n = 13) and those without (n = 9) upon admission. The non-infectious group still exhibited a higher NLR during MC (from 15.18 ± 14.05% to 4.51 ± 1.83%, p = 0.05), suggesting that this innate immune activation might not be solely due to concurrent infections. Besides, NLR also showed a positive correlation with MG-related clinical scores (Fig. 2e), including the Myasthenia Gravis Foundation of America quantitative myasthenia gravis (QMG) score (r = 0.543, adjusted p = 0.001), MG-specific manual muscle testing (MMT) (r = 0.399, adjusted p = 0.037), and Myasthenia Gravis Activities of Daily Living Scale (ADL) score (r = 0.456, adjusted p = 0.010). These findings aligned with the enhanced activation of cellular components in the innate immune system observed in the initial DEG analysis.

Monocytes 3 was identified as a pro-inflammation cell type in MC PBMCs

Given that monocytes, a type of innate immune cell, were highlighted as the most activated cell type in the overall DEG analysis by CellMarker, we initially sub-clustered the monocytes. This was done in accordance with a high-quality ScRNA study specifically focused on monocytes [31] (Fig. 3a). The trajectory of RNA velocity in monocytes hinted in the differentiation from monocytes 1 to monocytes 2 and 3 at MC (Fig. 3b). Notably, the monocytes 3 were distinct from the traditional monocytes type 1 and were closely related with the neutrophils (Fig. 1). FCGR3B, one of the cell markers used to classify monocyte 3, was found to be largely expressed in this group compared to other monocytes (Fig. 3c and Additional file 1: Fig. S5). Traditional markers (CD14 and FCGR3A) which were used to distinguish CD14+ and CD16+ monocytes were not abundantly expressed in monocytes 3. Hence these FCGR3B+ monocytes are a novel monocytes subset, as is evidenced in the reference ScRNA study [31]. Other genetic signatures of monocytes 3 included expressions of NAMPT, NEAT1, and CSF3R.

Fig. 3
figure 3

Single-cell analysis of monocytes from MC PBMCs. a. The distribution of all monocytes (at MC and after MC). b. The RNA velocity tendency in different monocytes at MC, where monocytes 1 were differentiated into monocytes 2, 3 and 4. c. The cell markers used to classify monocytes. d. The DEG analysis of all monocytes (MC vs. after MC). e. The DEG analysis of monocytes 3 compared to other monocytes (including monocytes from both at MC and after MC states). f. The DEG analysis of only monocytes 3 (MC vs. after MC)

Full size image

We then performed the DEG analysis (MC vs. 3 months after) across all monocytes and monocytes 3 alone to better characterize their differences. The overall monocytes showed higher inflammation activation status at MC, such as the activated coronavirus pathogenesis pathway (-log p-value 11.80) and interferon signaling pathway (-log p-value 6.25) (Fig. 3d and Additional file 2: Table S6). When compared to other monocytes 1, 2, and 4, monocytes 3 also exhibited higher granzyme A signaling (-log p-value 22.70) and coronavirus pathogenesis pathway (-log p-value 15.90) (Fig. 3e and Additional file 2: Table S7). Additionally in monocytes 3, up-regulated proinflammatory pathways such as multiple sclerosis signaling (-log p-value 12.80) and neuroinflammation signaling (-log p-value 12.3) were highlighted at MC compared to three months after MC (Fig. 3f and Additional file 2: Table S8).

To further explore the origin of the pro-inflammatory cytokines that were identified in the serum of MC patients, we analyzed the gene expression levels of previously identified cytokines (IL-1, IL-2, IL-4, IL-5, IL-6, IL-9, IL-13, IL-17A, IL-22, IL27, TNF-α, and GM-CSF) [15], along with other 9 chemokines that were representative in autoimmune diseases. Except for four cytokines (IL-9, IL-13, IL-17A, IL-22) which were not identified in the ScRNA dataset, the expression levels of the rest 17 cytokines and chemokines were all significantly revealed in Fig. 4a. Notably, a pro-inflammatory cytokine IL-1B and the most potent neutrophil chemokine CXCL8 were mostly expressed in monocytes 3, which were increased by 2.84 and 2.01 folds respectively, at the MC compared with that in 3 months after MC (p = 0.0000) (Additional file 1: Fig. S5). The gene expressions of the corresponding receptors of IL-1B and CXCL8 were also analyzed in all cell types at MC (Fig. 4b, c). IL1R1 was mainly expressed on neutrophils and activated regulatory T (Treg) cells, and CXCR1 and CXCR2 were mainly expressed on natural killer (NK) cells and monocytes 3. The pathway scores in inflammasome signaling and interferon signaling were also the highest in monocytes 3 at MC (Fig. 4d, e). Taken together, monocyte 3 might be the main cause of neutrophilia which leads to a drastically high NLR at MC.

Fig. 4
figure 4

The characterization of monocytes 3 from MC PBMCs. a. The expression of relevant cytokines and chemokines across all cell types at MC. The eight chemokines were in fact found surging in plasma at the acute stage of MC in our previous follow-up study [15]. The nine chemokines are representative CC (induce migration of lymphocytes and monocytes) and CXC (promote neutrophil migration) chemokines. The monocytes 3 exhibited high expression of IL1B and CXCL8, which potentially might augment the MC inflammation. b. The receptors expression of (IL-1 and CXCL8) in different cells at MC. c. Expression of CXCL8, IL1B, and RNASE2 in different monocytes. d. The inflammasome signaling scores in all cell types. The monocyte 3 exhibited the highest activation in inflammasome signaling at MC. e. The interferon signaling scores in all cell types. The interferon signaling was highest activated in monocytes at MC. f. The cell–cell communications between monocytes 3 and other cell types at MC. g. The B activation (BAFF) signaling communications among all cell types at MC

Full size image

Subsequently, the cell-communication networks were investigated on monocytes 3. The monocytes 3 have higher weighted communications with T cells, in particular, CD8+ Tnaive, CD8+ Tem, CD4+ Temra, CD4+ Tem Th1, and NK cells (Fig. 4f). While the B cell activation factor (BAFF) signaling network showed that monocytes 1, 2, and 3 were more communicative than other cell types with age-associated B, memory B, naïve B, and plasma cells (Fig. 4g). These findings indicated that monoctyes 3 and other monocytes might have interactions with both T and B subsets.

Major lymphoid cell suppression and activated T/B subsets during MC

According to the Celltypist annotation model [29], four types of B cells were classified (Fig. 5a, b). DEG analysis of B cells at MC highlighted two apoptosis-related pathways at MC (Fig. 5c and Additional file 2: Table S9). The pro-inflammatory coronavirus pathogenesis pathway was upregulated in all B cells at MC (-log p-value 13.90) compared to three months after MC. Conversely, B cell receptor signaling was downregulated at MC (-log p-value 5.35), indicating a suppressed state of B cells at this acute stage. Notably, a unique population of memory B cells, named age-associated B cells, was identified with clonal expansion at MC. Age-associated B cell expansion has been previously identified in autoimmune diseases, which were able to augment the immune response via cytokine production and T-cell stimulation [43].

Fig. 5
figure 5

Single-cell analysis of B cells from MC patients. a. The distribution of all B cells (MC and after MC). b. The RNA velocity tendency in different B cells at MC. c. The DEG analysis of all B cells (MC vs. after MC), in which the B cell receptor signaling was inhibited. d. ScBCR clonal expansion in all B cells. Age-associated B, naive B, and plasma cells were clonally expanded at MC. e. VDJ combination of BCRs in all B cells

Full size image

A total of 1159 B cells with 1117 clonal types were identified in the ScBCR sequencing (Additional file 1: Fig. S7). The comparisons of clonal BCRs at MC and 3 months after MC revealed marked clonal expansions in age-associated B cells, naive B, and plasma cell expansions with less BCR repertoire heterogeneity (Fig. 5d). Since IL-10 signaling pathway was up-regulated in overall monocytes at MC (Additional file 2: Table S9), we next analyzed the expressions of ribonuclease A family member 2 (RNASE2) which can induce IL-10 secretion from monocytes and thus augment the age-associated B cells expansion in systemic lupus erythematosus [40]. RNASE2 genes were mainly expressed in all monocytes (Fig. 4c) and were significantly increased at MC (Additional file 1: Fig. S5). VDJ genes that compose the BCR repertoire mostly included V3-20, J1, V3-23, D3-9, and J4 (Fig. 5e). No B cell clone was identified in both patients at and after MC, though IGLV5-45 + IGHV3-11 and IGLV3-1 + IGHV3-23 were the abundant B cell clones at MC.

A total of 11 T cell subsets were identified from the ScRNA dataset (Fig. 6a, b), which was classified according to a recent ScRNA dataset of PBMCs derived from MG patients [29] (Additional file 1: Fig. S2). The DEG analysis of T cells indicated TCR signaling inhibition (-log p-value 15.8) and T exhaustion (CTLA4 signaling and PD-1 signaling) during MC (Fig. 6c and Additional file 2: Table S10). At both the MC stage and three months after MC, T activation markers were upregulated in CD4+ Tem (Th1), CD4+ Temra (Th1), and activated Treg, while T exhaustion markers were upregulated in Tem (Th1), CD4+ Temra (Th1), CD8+ Tem, and activated Treg. These observations may suggest that T cell exhaustion could begin at the acute stage of MC and persist for three months. A total of 6845 T cells were identified in the ScTCR sequencing (Additional file 1: Fig. S7). The TCR repertoire analysis also supported the activation and expansion across almost all T subsets in certain clonotypes (e.g., CD8+ tem) (Fig. 6f). Similar to the B cell clone results, no T cell clones were shared between these two patients during or after MC. However, the most abundant T cell clones at MC were TRAV1-2 + TRBV9, TRAV21 + TRBV5-4, and TRAV17 + TRBV12-4.

Fig. 6
figure 6

Single-cell analysis of T cells from MC patients. a. The distribution of all T cells (MC and after MC). b. The RNA velocity tendency in different T cells at MC. c. The DEG analysis of all T cells (MC vs. after MC), in which T exhaustion signaling (CTL4 and PD-1) were mostly activated. d. T activation scores in different T cells. e. T exhaustion scores in different T cells. f. ScTCR clonal expansion in all T cells. Most T cells were clonally expanded at MC

Full size image

DEG analyses in NK cells revealed a consistent suppressive state at MC which is in line with that in B and T cell subsets (Additional file 2: Table S11). The pro-inflammatory EIF2 signaling was down-regulated (-log p-value 24.1) and the anti-inflammatory IL-10 signaling was up-regulated (-log p-value 7.55). The enriched glucocorticoid receptor signaling with uncertain activation state (-log p-value 7.9) may indicate the disturbance caused by the immune-modulatory treatments (e.g., corticosteroid) at MC.

Taken together, these observations suggest an overall suppression of lymphoid cells during MC, a conclusion consistent with the high NLR observed in the MC hemograms.

Discussion

Although great efforts have been made in the research and treatment of MG, many essential questions remain to be clarified, such as the immune status during the most devastating state, MC. Our previous prospective follow-ups of MC patients identified hypercytokinemia as a significant peripheral immune feature at the most acute stage [15]. This finding is consistent with the innate immune activation at MC revealed in the current study and is further supported by the drastically elevated NLR and its correlation with MG-related clinical scores (ADL, MMT, QMG). Another QOL-15 scale was not correlated with the NLR, which can be explained by its tighter relevance to life quality instead of clinical symptoms [44].

These findings are commonly observed at the acute stage of many systemic inflammations, such as COVID-19 [16, 45]. COVID-19, a disease well-known for causing a cytokine storm [46], has been reported to precede MC in several MG patients [47, 48]. Moreover, PE was observed to be more efficient than IVIG in ameliorating MC symptoms in a prospective cohort, possibly due to its direct action in removing pro-inflammatory cytokines [18, 49]. The NLR has been reported as a peripheral marker of systemic inflammation [50]. And in the clinical context of MG, the NLR has been previously identified as an index correlating with disease activity and as an in-hospital mortality predictor [41, 42]. Although steroid treatment may induce this NLR escalation, it cannot be fully explained by steroid treatment alone, as previous pharmacodynamic studies of steroids have shown that the duration of neutrophilia and lymphopenia peaks at around four hours and returns to normal within 24 h when steroids are administered intravenously or orally in humans [51, 52]. Therefore, we propose that innate immune activation, characterized by hypercytokinemia, might be a key feature in the pathogenesis of MC, potentially leading to the observed elevated NLR in the hemogram.

The triggers leading to the transition from a non-crisis state to a crisis in MG patients can be primarily categorized into two types: (1) infection and (2) surgery [53]. These correspond to the pathogen-associated molecular pattern (PAMP) and damage-associated molecular pattern (DAMP), respectively, which are characteristics of monocyte toll-like receptor activation [54]. Previous studies have identified higher monocyte percentages in normal MG patients than controls using mass cytometry [55, 56]. CD14 + monocytes also have been reported in MG patients and are known to exhibit high inflammatory activity [23, 56, 57]. As the first line of defense, innate immunity rapidly activates monocytes and neutrophils, which can then release cytokines and chemokines to attract more effective T and B cells from the adaptive immune system [58]. The single-cell sequencing data in the current study were obtained from patients undergoing immunosuppressive/immunomodulatory therapies. Such therapies would inevitably interfere with pro-inflammatory pathways. Despite this interference, we observed that pro-inflammatory pathways, such as coronavirus pathogenesis signaling and interferon signaling, were notably activated in monocytes 3 (FCGR3B+ monocytes). This activation might lead to the release of MC-specific pro-inflammatory cytokines (IL-1 and CXCL8) previously identified, thereby augmenting the immune response. Contrary to the classical (CD14+) and non-classical monocytes (CD16+), monocyte 3 is a relatively novel entity with an unknown function [31]. Interestingly, a single-cell RNA study identified high expression of FCGR3B in alveolar macrophages in severe COVID-19 patients [59]. This finding suggests that FCGR3B+ monocytes/macrophages might play pivotal roles in systemic inflammation. FCGR3B encodes the Fc gamma receptor 3B, a surface marker for neutrophils, which can cooperate with other Fc gamma receptors to promote phagocytosis of antibody-opsonized microbes by neutrophils [60]. Consequently, monocytes 3 might possess higher phagocytic capacity and differentiate into tissue macrophages to amplify inflammation. However, we should not overlook the possibility that the activation of monocyte 3 could also be a result of an insidious infection during MC. This is because all invasive treatments significantly elevate the risk of infection [61].

Other cell types might also participate in MC pathogenesis and associate with the monocytes. Although T, B, and NK cells were inhibited in MC, some lymphoid subtypes with clonal expansion might still promote inflammation. We found significant expansions in aged-associated B and CD8+ tem cells at MC. The aged-associated B is a well-known autoimmune-contributing cell type, and our results also showed its BAFF activation was connected with monocytes 1, 2, and 3. The BAFF signaling promotes B-cell survival, maturation, and differentiation, and excess of which promotes the development of autoreactive B cells [62]. The age-associated B expansion might also be explained by more IL-10 derived from the monocytes, for RNASE2 upregulation can activate IL-10-producing signaling in monocytes [63]. CD8+ T cell exhaustion was found to be associated with a good outcome in autoimmune diseases [64]. Consistently, the exhausted CD8+ tem/temra in our study were most dominant after three months from discharge instead of at MC, implying T cell activation at MC may augment the inflammation. The IL-1 potentially secreted by monocytes 3 might also contribute to this process, since IL-1 is an innate mediator of T cell activation [65]. Besides, other CD4+ T cells (CD4 tem Th1/Th17) also expand at MC, which is in accordance with our in-depth peripheral CD4+ T profile study at MC [15].

There are several limitations to this study. First, more functional studies are needed to validate the causality between monocytes 3 and MC. Blood samples from those infection-free patients with other causes will be valuable to explore the effects of different MC triggers on monocytes 3 activation. Secondly, numerous pro-inflammatory cytokines, which were surging in the plasma during MC, could not be traced back to originating cell types. Usually, macrophages and dendritic cells are more powerful in releasing pro-inflammatory cytokines than monocytes [66]. However, these are mainly tissue-resident that we can not access for further evaluations. Third, immunosuppressive treatment during MC may interfere with the interpretation of the immune datasets, thus immunosuppressing signaling being more likely activated (e.g., IL-10 signaling [25]). Last, clustering based on the mRNA transcriptome may not be fully adherent to the analysis based on the protein surface markers.

Conclusions

In summary, we depicted the landscape of the PBMCs from the devastating stage of MC in MG patients from a single-cell perspective. Hypercytokinemia driven by the innate immune system might be the underlying mechanism in the development of MC. The cells highlighted in this study might be future targets for developing new treatment.