The clinical significance of the Mediterranean fever gene MEFV variants in Castleman disease

Abstract

Background

Idiopathic Multicentric Castleman Disease, Thrombocytopenia, Anasarca, Fever, Reticulin Fibrosis, Organomegaly (iMCD-TAFRO) is a rare cytokine storm syndrome with high mortality. Pathogenesis of Castleman disease (CD) remain largely unknown. We aim to unravel the role of Mediterranean fever gene MEFV variants, the key gene variants implicated in familial Mediterranean fever, in CD.

Methods

Clinical data from a retrospective cohort of 37 patients with CD were collected. Blood and/or lymph node biopsy specimens were obtained for whole-exome sequencing. In vitro lipopolysaccharide stimulation experiment and single-cell RNA-sequencing (scRNA-seq) were performed to characterize the immune signature using peripheral blood mononuclear cells from an adolescent TAFRO patient, his asymptomatic parents, and a healthy control. The relative gene expression were examined by quantitative PCR. Cytokine levels were assessed using Luminex. Statistics were performed by SPSS and GraphPad Prism.

Results

Here we show an adolescent TAFRO patient with familial MEFV mutations demonstrates responsiveness to anti-IL-6 containing therapy. Through comprehensive analysis of a cohort of 37 CD patients, we observe a high prevalence of MEFV mutations (76%, 28/37). Notably, the MEFV E148Q-P369S-R408Q variant is present in 19% (7/37) of all patients and 50% (2/4) of the TAFRO subtype, with variant carriers exhibiting more severe disease course. Inflammation responses experiments and scRNA landscape reveal that MEFV expression is dominant in CD16⁺ monocytes and correlated with IL-6 pathway activation likely via the interaction with naïve B/memory B cells in the TAFRO.

Conclusions

This study presents one of the largest cohorts demonstrating the high prevalence of MEFV variants in CD, providing important insights for understanding and treating CD, particularly TAFRO.

Plain language summary

Castleman disease (CD) is a rare disease about which our understanding is still limited. In this study, we comprehensively analyzed an adolescent patient with the most severe subtype of CD, known as TAFRO, alongside further investigation of samples from 37 CD patients collected over 11 years. We found that many of the CD patients carried changes in their DNA, specifically in a gene called Mediterranean Fever (MEFV), which could play a large role in this rare disease. Targeting MEFV-related inflammation might be an effective intervention for treating CD. Our study thus provides important insights for understanding the cause of Castleman disease and ways in which it might be treated.

Du, Xie, Dai, Jia, et al. assess the role of MEFV variants in Castleman Disease using both in vitro experiments and patient samples. The inflammatory response, scRNA landscape, and high prevalence of MEFV variants propose a critical role for MEFV in Castleman disease.

Introduction

Castleman disease (CD) comprises a group of rare and heterogeneous systemic lymph-proliferative disorders, classified into unicentric (UCD), multicentric (MCD)^1,2, and oligocentric CD^3,4. MCD encompasses several distinct variants, including iMCD-TAFRO—a potentially life-threatening subtype first reported in 2010⁵. Although TAFRO has been well documented, there are only about 15 reported cases in patients younger than 18 years old all over the world^6–10. While the international consensus has recommended anti-IL-6 treatment as first line for all iMCD¹¹, very little evidence is currently available for TAFRO¹². Efficacy of siltuximab in pediatric TAFRO remains unknown¹, since children with iMCD have not been enrolled in clinical trials of siltuximab¹³.

The pathogenesis of TAFRO remains largely unknown¹⁴. Recently, small cohort studies^15–19 showed an association between Mediterranean fever gene MEFV variants, the crucial gene variants implicated in familial Mediterranean fever, and iMCD. However, only about 20 iMCD without any TAFRO were described in all these studies, and functional analysis of MEFV variants in iMCD, particularly in TAFRO are largely lacking.

Here, we add to the very limited literature by describing the unique characteristics of an adolescent TAFRO and the responsiveness to anti-IL-6 containing therapy. MEFV variants appear to be potent in promoting inflammation. A high incidence of MEFV variants, together with a more severe disease course in the cohort of 37 CD, one of the largest cohorts to date, suggests a potential role in the pathogenesis of CD. Inflammation responses experiments and single-cell RNA-sequencing (scRNA-seq) unravel MEFV-enriched CD16⁺ monocytes likely interact with naïve B/memory B cells, contributing to IL-6 pathway over-activation. Furthermore, we provide the potential link of megakaryocytes implicated in inflammation via the CXCL signaling network in TAFRO.

Methods

Clinical course of the patients

This study was approved by the Institutional Review Boards, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University. All participants or parents for those under the age of 18 provided written informed consent for publication. The data of 37 CD patients were collected, and the follow-up visit lasted till Mar 31st, 2025. Blood and/or lymph node biopsy specimens were obtained for whole-exome sequencing (WES). In vitro lipopolysaccharide (LPS) stimulation experiments and scRNA-seq were performed to characterize the immune signature using peripheral blood mononuclear cells (PBMCs) from an adolescent TAFRO patient (during flare and remission), his asymptomatic parents, and a healthy control.

Cell culture

PBMCs were extracted from heparinized blood of the patient, his asymptomatic parents, and a healthy control by gradient centrifugation using Ficoll-Paque. Cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum.

Whole-exome sequencing (WES)

WES was done in Sun Yat-sen Memorial Hospital, Sun Yat-sen University, and HaploX Genomics Center, Jiangxi HaploX Medical Laboratory Co., Ltd. Notably, as a retrospective study, matched normal tissue control specimens (such as fibrofatty tissue) were available for WES in only 9/37 cases. Whether the detected variants were of germ-line or somatic origin was undetermined, except for one with blood/saliva/lymph node specimens. The main procedures of WES included library generation, reference alignment, and variant calling. Genomic DNA was extracted from lymph node tissues by using the QIAamp DNA FFPE tissue kit (Qiagen), followed by library construction. Library sequencing was then performed on the Illumina Novaseq 6000 system according to the manufacturer’s instructions after quality analysis. Preprocessing the Raw Data obtained from the Illumina sequencing platform using fastp (V0.12.6, https://github.com/OpenGene/fastp) was conducted to acquire the Clean Data for subsequent analysis. Reads mapping to the reference genome was done as usual. The raw SNP/Indel sets are called by Sentieon DNAseq (https://support.sentieon.com/manual/DNAseq_usage/dnaseq/). Main steps included (1) Realigner, (2) BQSR, (3) Haplotyper, and (4) VariantFiltration. All types of alterations were evaluated, including single-nucleotide variants, short insertions and deletions (indels), copy number alterations, gene fusions, and rearrangements. ANNOVAR was used for functional annotation of variants. Germ-line variants in MEFV were further validated using Sanger sequencing.

Sanger sequencing

We performed Polymerase chain reaction (PCR) for the MEFV mutational aberrance using primers as listed in Table S4. For MEFV exon2F/R, degeneration with 94 °C for 5 min, followed by annealing for 35 cycles (94 °C for 30 s, 57 °C for 30 s, 72 °C for 90 s), and elongation with 72 °C for 5 min. For MEFV exon3F/R, degeneration with 94 °C for 5 min, followed by annealing for 35 cycles (94 °C for 30 s, 59 °C for 30 s, 72 °C for 30 s), and elongation with 72 °C for 3 min. The PCR products were purified after gel electrophoresis, followed by Sanger sequencing. The PCR and Sanger sequencing were performed by the IGEbio company, Guangzhou.

Quantitative PCR

Total RNA was extracted using FastPure® Cell/Tissue Total RNA Isolation Kit V2, subsequently followed by reverse-transcription into cDNA, using HiScript® III RT SuperMix for qPCR. Quantitative PCR for MEFV, IL-6, and IL-1β was done by using ChamQ SYBR Color qPCR Master Mix. Primers were listed in Table S4. All kits were purchased from Vazyme Biotech Co., Ltd. Experiments were performed according to the manual. All standards and samples were examined in triplicate. The mRNA values for all tested samples were determined by LightCycler analysis software. Data were presented as the mean level of expression of targeted genes relative to GAPDH expression.

In vitro inflammasome stimulation by LPS

PBMCs were incubated for 3 h with 1000 ng/ml lipopolysaccharide (LPS) or RPMI 1640, after which the medium was removed. Cells were incubated for an additional 15 min with either RPMI 1640 or 2 mM ATP. Pictures of cells were taken under a microscope. Supernatants were collected by spinning and frozen at −80 °C until the cytokine assay. mRNA was isolated for subsequent quantitative PCR analysis.

Colchicine inhibition experiments

Colchicine (100 ng/ml) was added to the LPS plus ATP incubated cells for 30 min. Pictures of cells were taken under a microscope. Collection and storage or subsequent analysis of supernatants and mRNA were done as mentioned above.

Cytokine production assay by Luminex (Bio-Plex Pro Human Cytokine Assay)

Luminex technology using microspheres (beads) conjugated with specific antibodies was employed to detect multiple cytokines in a single sample. The experiments were performed by Shanghai Universal Biotech Co., Ltd. Generally, the experiments were done including the following steps: Sample collection and preparation, Reagent preparation, Bead preparation, Incubation and washing, Sample and standard addition, Washes, Detection antibody incubation, Streptavidin-PE Conjugation, Resuspension and reading, and Data Analysis.

Library preparation for scRNA-seq

Fresh PBMCs were isolated from peripheral blood, rapidly frozen, and stored in liquid nitrogen. ScRNA-seq libraries were prepared using a DNBelab C Series High-throughput Single-Cell RNA Library Preparation Set (MGI, #940-000519-00), as previously described in our work²⁰. Briefly, Single-cell suspensions were processed into barcoded scRNA-seq libraries through a series of steps: droplet encapsulation, emulsion breakage, mRNA capture on beads, reverse-transcription, cDNA amplification, and purification. The resulting cDNA was sheared into short fragments ranging from 300 to 500 bp. Indexed sequencing libraries were then constructed following the manufacturer’s protocol. Library quality was assessed using the Qubit ssDNA Assay Kit (Thermo Fisher Scientific) and Agilent Bioanalyzer 2100. Finally, all libraries were sequenced on the MGISEQ-2000 platform using pair-end sequencing. The sequencing was performed by BGI Research, Shenzhen, China.

scRNA-seq data analysis

(1) Raw data processing: Raw sequencing data were processed using the open-source DNBelab C Series scRNA analysis pipeline (https://github.com/MGI-tech-bioinformatics/DNBelab_C_Series_scRNA-analysis-software). In brief, data demultiplexing, barcode processing, and single-cell 3′ UMI (Unique Molecular Identifiers) counting were performed using default parameters. The filtered reads were aligned to the hg38 genome using STAR²¹ (v2.5.1b). Valid cells were then identified by the barcodeRanks function from DropletUtils, excluding those with UMI counts below the threshold. Cell versus gene UMI count matrices were generated using PISA. (2) Quality control: Count matrices were processed using the Seurat package²² (v3.1.4) in R (v3.6.0). To ensure data quality, genes expressed in fewer than three cells were removed, and cells with less than 15% mitochondrial reads, fewer than 500 genes, or fewer than 3,000 UMIs were excluded. Doublets were identified and filtered out using DoubletFinder²³ with default parameters. (3) Analysis of scRNA-seq data: Seurat package (v3.1.4) in R (v3.6.0) was used to create Seurat objects for analysis. Seurat objects were scaled using the following functions in sequence: NormalizeData, FindVariableFeatures, and ScaleData. Dimension reduction was then initiated with PCA on the 3000 most significantly variable genes, followed by UMAP for visualization in 2D space. Cell types were identified based on their distinct gene expression profiles within each cluster. Cells expressing MEFV (count >0) were defined as MEFV⁺ cells. Differentially expressed genes (DEGs) between clusters or cell types were identified using the FindAllMarkers function from the Seurat package, with settings only.pos = TRUE, min.pct = 0.25, and logfc.threshold = 0.25. DEGs were used for functional annotation performed by Metascape²⁴ (http://metascape.org/). The AverageExpression function from the Seurat package was used to calculate the average expression level of each gene within different clusters or samples. Correlation analysis was performed by the matrix of the 3000 most hypervariable genes in Seurat objects, utilizing the cor() function in R. Plots were generated by DimPlot, DotPlot, and VlnPlot from the Seurat package. Bar plots were generated using ggplot2 (v3.3.3). Heatmaps were generated by pheatmap (v1.0.12) in R. CellChat²⁵ (v1.6.1) was used to identify significant interactions between different cell types. The raw gene expression matrix was formatted for CellChat, and the default ligand-receptor database provided by CellChat was used. The analysis was conducted using default parameters, and the results were visualized through heatmaps and circle plots generated by CellChat.

Statistics and reproducibility

Statistics were performed by SPSS 22.0 and GraphPad Prism 9. For continuous variables that conformed to the normal distribution, one-way ANOVA and t-test were used. For those that did not conform to the normal distribution, the Mann–Whitney U test (two independent sample rank sum test) was used. P ≤ 0.05 was considered a statistically significant difference. *P < 0.05, **P < 0.01, ***P < 0.001. Experiments were done at least twice for reproducibility. Sample sizes and number of replicates were provided in detail as indicated in the figure legends.

Results

The responsiveness of the adolescent TAFRO patient to IL-6 blockade-containing therapy

A previously healthy 15-year-old Chinese male was admitted with the chief complaint of abdominal pain and fever. Laboratory tests showed anemia, thrombocytopenia, hypo-albumin, renal insufficiency, and elevations in C-reactive protein (CRP), IL-6, and vascular endothelial growth factor (VEGF) (Fig. 1G–I). CT, MRI, and PET-CT demonstrated pleural effusions, severe ascites, multiple enlarged lymph nodes, and splenomegaly with increased FDG uptake (Fig. 1A, B, J). Excisional lymph node biopsy revealed features of a mixed type of CD (Figs. 1C, D and S1A). Similar to the recent study²⁶, bone marrow biopsy detected megakaryocytic atypia and reticulin fibrosis (MF-2/3) (Figs. 1E and S1B). Infectious and autoantibodies workup appeared negative. The patient fulfilled the diagnostic criteria⁶ of TAFRO, with 10 points falling into the severe grade⁵.

Fig. 1. Adolescent TAFRO patient showed responsiveness to anti-IL-6-containing therapy.

A Magnetic resonance imaging (MRI) revealed pancreatic enlargement, consistent with pancreatitis (red triangles), subcutaneous soft-tissue edema of the abdominal wall (red arrowheads), perihepatic and perisplenic effusion (red arrows), and mild bilateral pleural effusion (black asterisks). The red dagger indicated an enlarged spleen before treatment, with a major axis of 152 mm. B (Left) Maximum intensity projection (MIP) image of 18-FDG positron emission tomography-computed tomography (PET-CT) demonstrated multiple enlarged lymph nodes and spleen with increased fluorodeoxyglucose (FDG) uptake. Bilateral cervical (upper middle), axillary (upper right), inguinal (lower right) lymphadenopathy, and splenomegaly (lower middle)with increased FDG uptake. All with white arrows. C HE staining showed expanded concentric mantle zones (onion skin sign) with polytypic plasmacytosis in some follicles (left), regressed germinal centers (middle), and proliferation of highly dense endothelial vessels and moderate cell degeneration in the interfollicular area (right). Scale bar size 40×. D Immunohistochemical analysis revealed positivity for CD38 in plasma cells (upper left), a disrupted pattern of follicular dendritic cells (FDC) by CD21 (upper right), mild expression for CD 123 (lower left), and negativity for HHV-8 (lower right). Scale bar size 40×. E Bone marrow biopsy demonstrated hypercellularity with progressive multi-lineage hematopoiesis (left), increased megakaryocytes with megakaryocytic atypia (middle), and remarkable fibrosis (MF-2) on reticulin staining (right). Scale bar size 160×. F The patient received siltuximab (11 mg/kg at every 3-week intervals for 5 doses), corticosteroids (high-dose methylprednisolone 500 mg/day for 3 consecutive days with slowly tapering), as well as ciclosporin (75–50 mg every 12 h) replaced by tacrolimus (0.5 mg every 12 h to every day) and achieved complete remission (CR) 3 months post onset. TCP-like regimen (thalidomide 50–25 mg per day, cyclophosphamide 400 mg per week, three times every month to 50 mg per day, prednisone 1 mg/kg/day with slow tapering) was employed as maintenance therapy. m month. G–I Laboratory findings showed rapid and significant improvement after treatment. IL-6 interleukin 6, IL-1β interleukin-1β, Hb hemoglobin, Alb albumin, PLT platelet, VEGF vascular endothelial growth factor, Cr creatinine, ESR erythrocyte sedimentation rate, CRP C-reactive protein. J Chest computed tomography scans showing mild bilateral pleural effusion before treatment (left, red asterisks), reduction at 1 week post-treatment (middle, red asterisks), and complete resolution 3 months post-treatment (right). The red dotted circles indicate enlarged lymph nodes in the bilateral axillary region before treatment (left), smaller lymph nodes at 1 week post-treatment (middle), and normal size 3 months post-treatment (right). K Three-month post-treatment CT imaging showed no pancreatic swelling (red triangles), with complete resolution of subcutaneous soft-tissue edema, perihepatic and perisplenic effusion, and mild bilateral pleural effusion. The red dagger revealed a smaller spleen post-treatment compared to that before treatment, with the major axis extending from 152 mm (B, lower right, red dagger) to 123 mm. L Bone marrow biopsy did not detect any alterations, with normal multi-lineage hematopoiesis (upper) and no fibrosis (MF-0) (lower) 3 months post-treatment. Scale bar size 160×.

Treatment for this TAFRO adolescent included siltuximab, corticosteroids, and immunomodulators (Fig. 1F). After five doses of siltuximab around 3 months post-treatment, he achieved complete remission²⁷, followed by treatment of TCP (thalidomide, cyclophosphamide, prednisone) regimen²⁸. Siltuximab was discontinued after five doses due to financial issues. Importantly, even a slight rebound of IL-6 could be observed at around the 4th month (Fig. 1G), a sustained response, without serious adverse events, was demonstrated (Figs. 1F–L and S1B, Table S2). Similar to a recent study²⁹ showing a portion of TAFRO patients could be curable, our patient discontinued treatment after 18 months, maintaining complete remission for 32 months at the time of manuscript preparation. This special case also supports the notion that pediatric iMCD is often severe and responds to siltuximab³⁰.

WES identified a unique pattern of MEFV germ-line variants in the family of an adolescent TAFRO

Since gene analysis in CD, especially in TAFRO, was largely lacking, we performed WES using the peripheral blood of the patient (the only child in the family) and his parents, together with a healthy control. We identified the MEFV complex alleles E148Q/E148Q-P369S/WT-R408Q/R408Q variant in the patient, inherited from his parents, while the healthy control did not show any alteration of MEFV (Figs. 2A, B and S2). To further validate the MEFV mutations as germ-line variants, the lymph nodes and saliva from the patient were also used for WES. We found a similar variant allele frequency (VAF) (E148Q 1.0, P369S 0.5, R408Q 1.0) across blood, lymph node, and saliva. In addition, we did not identify other variants previously included as being potentially pathogenic variants in the ClinVar (http://www.ncbi.nlm.nih.gov/clinvar/) or Infevers (http://infevers.umai-montpellier.fr) database. Sanger sequencing confirmed single-nucleotide mutations detected in WES, with concordance of 100% (Figs. 2A and S2).

Fig. 2. The MEFV variant of TAFRO exhibited potent inflammation activation in vitro, and its clinical significance in a cohort of 37 patients with CD.

A Sanger sequencing confirmed the presence of MEFV variants in accordance with WES. Sequence alignment was performed using the BioEdit 7.0.9 software. B Pedigree chart of affected iMCD-TAFRO adolescence, diagnosed with iMCD-TAFRO at 15 years old, rectangle in red. C Quantitative PCR showed that the expression of MEFV mRNA in peripheral blood mononuclear cells (PBMCs) decreased as the number of mutations increased. GAPDH was used as an internal control. N = 3, PBMCs of individuals were collected and tested in triplicate. Means and standard deviations were shown. Turkey was used for multiple comparisons following one-way ANOVA analysis. *p = 0.013, ***p < 0.001. D LPS plus ATP stimulation induced cell aggregation of PBMCs. More remarkable aggregations were observed as the number of MEFV mutations increased, as the MEFV E148Q/E148Q-P369S/WT-R408Q/R408Q PBMCs of the TAFRO patient showed the most severe aggregations. LPS, lipopolysaccharide. ATP, Adenosine Triphosphate. Scale bar size 100×. E Quantitative PCR revealed a sharp increase in IL-6 and IL-1βmRNA levels following LPS plus ATP stimulus. MEFV variants further amplified the up-regulation of IL-6 and IL-1β, as cells from a TAFRO patient bearing MEFV E148Q/E148Q-P369S/WT-R408Q/R408Q achieved the highest mRNA expression level. N = 3, stimulation experiments were performed in triplicate. Means and standard deviations were shown. Turkey was used for multiple comparisons following one-way ANOVA analysis. **P = 0.008, *** P < 0.001, NS not significant P = 1.000. F Cytokines assay using Luminex demonstrated inflammatory activation in PBMCs with MEFV variants, particularly in the TAFRO patient. N = 2, means were shown. VEGF vascular endothelial growth factor. G Colchicine inhibited LPS-induced cytokine release in PBMCs of TAFRO. N = 2, means were shown. TNF-α tumor necrosis factor-α. H Pie chart showing the high prevalence of MEFV variations in 37 CD. I Details of the MEFV variants in 37 CD. UCD Unicentric Castleman Disease, aMCD asymptomatic Multicentric Castleman Disease, iMCD idiopathic Multicentric Castleman Disease, NOS Not Otherwise Specified, IPL Idiopathic Plasmacytic Lymphadenopathy, TAFRO thrombocytopenia, anasarca, fever, reticulin fibrosis, organomegaly. J E148Q, P369S, and R408Q ranked the top 3 variants of MEFV among 6 variants identified in 37 CD. The allele frequency of MEFV variants from East Asian populations was shown in parallel. The binomial test was used for comparison. E148Q ***P < 0.001, P369S **P = 0.002, R408Q ***P < 0.001, L110P **P = 0.006, R202Q P = 0.601, G304R P = 0.564. NS not significant. a, data from gnomAD v4.1.0. K MEFV E148Q-P369S-R408Q-positive CD exhibited a more severe disease course compared with the negative ones. Means were shown. For those that conform to the normal distribution, the independent samples two-sided t-test was used; otherwise, the Mann–Whitney U test was adopted. WBC **P = 0.001, Hb **P = 0.002, PLT *P = 0.018, CRP *P = 0.049, ALB *P = 0.014, eGFR *P = 0.014. WBC white blood cell, Hb hemoglobin, PLT platelet, CRP C-reactive protein, ALB albumin, eGFR estimated Glomerular Filtration Rate.

PBMCs of TAFRO carrying MEFV E148Q-P369S-R408Q variant exhibited hyperactive inflammasome signaling induced by LPS, which was partly abrogated by colchicine in vitro

The MEFV variant in the TAFRO patient and his asymptomatic parents led us to explore the genotype-phenotype correlation. A study had shown that decreased MEFV mRNA expression was correlated with an increased number of mutations and higher clinical severity in FMF³¹. Accordingly, our quantitative PCR data showed that expression of MEFV mRNA was decreased as the number of mutations increased (Fig. 2C).

To further illustrate the response of different MEFV variants to inflammation activation, an LPS plus ATP stimulation model was established. Morphologically, LPS-induced cell aggregation compared to the medium control in all groups. Cells bearing MEFV E148Q/E148Q-P369S/WT-R408Q/R408Q from the TAFRO patient showed the most pronounced aggregations (Fig. 2D). The cell aggregation might be the link to inflammation, as documented in autoimmune disorders^32,33.

The expression of IL-6 and IL-1β was increased sharply in LPS-stimulated samples, with peak transcriptional activation being particularly pronounced in PBMCs of a TAFRO patient, as indicated by qPCR experiments (Fig. 2E).

A panel of multiple cytokines was examined by using Luminex. Consistent with qPCR data, up-regulation of IL-6 and IL-1β induced by LPS was observed (Fig. 2F). CXCL/CCL signaling played a crucial role in iMCD³⁴. Notably, we found CXCL/CCL signaling cytokines (CXCL9/CXCL12, CCL2/CCL3/CCL7/CCL27) were remarkably activated in the TAFRO patient after LPS stimulation (Figs. 2F, S3 and S4). Similar trends were found in the other pro-inflammation factors.

Given the critical role of colchicine in treating FMF, which is normally symbolized by MEFV mutations, we investigated whether colchicine could inhibit the inflammation driven by MEFV in our TAFRO patient in vitro. The addition of colchicine inhibited the cell aggregation (Fig. S4) induced by LPS in the TAFRO patient. Furthermore, colchicine repressed the release of cytokines, including IL-6, IL-1β, and CXCL/CCL signaling (Figs. 2G and S4).

MEFV variants were highly prevailed and associated with more severe disease manifestations in 37 CD patients

Given the certain role of the MEFV variant in the TAFRO patient, we next investigated the clinical significance of MEFV in a large cohort of CD. Available lymph nodes from 37 out of 77 CD patients (from 1st January 2012 to 31st December 2022) were sequenced by WES. MEFV mutations were identified in 76% (28/37) of patients, among which the triple-mutation variant E148Q-P369S-R408Q accounted for 19% (7/37) of all cases and 25% (7/28) of MEFV mutant patients (Fig. 2H, I). We identified 6 variants of MEFV (L110P, E148Q, R202Q, G304R, P369S, R408Q), with E148Q, P369S, and R408Q being the three most prevalent variants (Fig. 2J) and functionally predicted to be probably damaging/disease-causing automatic by different software (Table 1). Importantly, the frequencies of these variants (E148Q, P369S, R408Q, L110P) in our cohort were significantly higher than those in the East Asian population from gnomAD v4.1.0 (https://gnomad.broadinstitute.org) (Fig. 2J).

Table 1.

Functional prediction of MEFV variants by different softwares

Nucleotide change	Amino acid change	SIFT _score	SIFT _pred	Polyphen2_HDIV _score	Polyphen2_HDIV _pred	Mutation Taster _score	Mutation Taster _pred	CADD _ C-score
c.T329C	L110P	0.04	Damaging	0.652	Possibly damaging	1	Polymorphism	9.59
c.G442C	E148Q	0.01	Damaging	0.995	Probably damaging	0.913	Polymorphism automatic	22.6
c.G605A	R202Q	1	Tolerated	0.002	Benign	1	Polymorphism automatic	0.001
c.G910A	G304R	NA	NA	0.001	Benign	1	Polymorphism	22.9
c.C1105T	P369S	0.05	Damaging	0.959	Probably damaging	0	Disease-causing automatic	22.7
c.G1223A	R408Q	0.23	Tolerated	0.259	Benign	0	Disease-causing automatic	3.6

SIFT (sorts intolerant from tolerant): score ≤ 0.05, damaging; score > 0.05, Tolerated.

PolyPhen-2 (Polymorphism Phenotyping v2) _HDIV: score ≥ 0.957, D: probably damaging; 0.453 < score ≤ 0.956, P: possibly damaging; Score ≤ 0.452, B: benign.

Mutation taster: A: disease-causing automatic; D: disease-causing, N: polymorphism, P: polymorphism automatic.

CADD (Combined Annotation Dependent Depletion): a scaled C-score of greater than or equal to 10 indicates that these are predicted to be the 10% most deleterious substitutions that you can do to the human genome, a score of greater than or equal to 20 indicates the 1% most deleterious, and so on.

NA not available.

Interestingly, we could see a trend toward increased MEFV E148Q-P369S-R408Q-positive rate and the severity of disease as following UCD (4/25, 16%), MCD (3/12, 25%), iMCD (3/10, 30%), and TAFRO (2/4, 50%) (Figs. 2I and S5). Recently, MEFV E148Q-P369S-R408Q was detected in 14% (2/14) of the Japanese iMCD cohort, which was associated with more severe disease manifestations¹⁹. Our study identified a higher frequency among iMCD patients (3/10, 30%) and TAFRO patients (2/4, 50%) (Fig. 2H, I). When looking into the 7 patients positive for E148Q-P369S-R408Q, 29% (2/7) were TAFRO subtype (Fig. S5B). However, due to the relatively small sample size of each entity of disease, the association of MEFV E148Q-P369S-R408Q positivity and subtype of disease requires further validation.

Besides, MEFV E148Q-P369S-R408Q-positive CD exhibited a more severe disease course compared with non-MEFV E148Q-P369S-R408Q-positive patients, with a higher number of WBC, CRP, and a lower number of Hb, PLT, Alb, and inferior renal function (Fig. 2K, Table S3). Details of 7 patients carrying the MEFV E148Q-P369S-R408Q variant were provided in Table 2.

Table 2.

Clinical features of 7 CD patients positive for MEFV E148Q-P369S- R408Q

	Patient 1a	Patient 2a	Patient 10a	Patient 13ab	Patient 16a	Patient 29a	Patient 31a
Sex	M	M	M	F	F	F	M
Age (years)	10–20	30–40	40–50	10–20	10–20	40–50	40–50
Clinical type	TAFRO	TAFRO	iMCD-NOS	UCD	UCD	UCD	UCD
Histological type	Mixed	PC	PC	HV	HV	PC	HVc
Chief complaint	Abdominal pain, fever	Mass in the left neck, edema in the lower limb	Abdominal pain	Mass in the left supraclavicular region	Mass in the right neck	Dizzy	Mass in upper abdomen
Location of mass	Left cervical region	Left cervical region	Pancreas, retroperitoneal	Left supraclavicular	Right cervical	Right parapharyngeal	Anterior pancreas
ECOG scores	3	1	1	0	0	1	2
PLT (×109/L)	40	72	239	282	348	317	187
Hb (g/L)	78	108	138	127	122	136	142
Anasarca	Yes	Yes	No	No	No	No	No
Fever	Yes	Yes	No	No	No	No	No
eGFR [ml/(min ×  1.73 m2)]	23.88	45.72	87.55	163.14	129.65	145.23	72.66
Urine protein (g/24 h)	8.648	11.08	0.06	Trace	ND	ND	ND
Bone marrow fibrosis	MF-2/3	MF-0/3	ND	ND	ND	ND	ND
Organomegaly	Yes	No	No	No	No	No	No
CRP (mg/L)	74.6	31.9	3.23	<2.98	ND	ND	0.85
ESR (mm/h)	100	68	16	22	ND	ND	3
IL-6 (pg/ml)	35.2	9.06	2.94	ND	ND	ND	ND
IL-1β (pg/ml)	5.87	<5	<5	ND	ND	ND	ND
WBC (×109/L)	4.23	7.02	5.02	5.91	6.05	6.28	6.13
Albumin (g/L)	29	27	37.3	38.8	48.4	38.2	45.8
ALP (U/L)	123	138	71	70	79	79	88
IgG (g/L)	17.8	12.6	13.4	ND	ND	ND	ND
Treatment	Siltuximab, corticosteroids, ciclosporin, thalidomide, cyclophosphamide	Rituximab, prednisone	Resection, cyclophosphamide, meprednisone	Resection	Resection	Resection	Resection, corticosteroids, Rituximab, chemotherapy (CHOP+GemOx)
Duration of visit (days)	960	1719	483	920	1961	522	1621
Outcome	Survival	Survival	Survival	Survival	Survival	Survival	Died

Lymph node biopsy specimens were obtained for whole-exome sequencing (WES) from 37 patients with CD. The MEFV E148Q-P369S- R408Q variant was identified in seven patients. Clinical data were collected and analyzed for these patients.

ECOG Eastern Cooperative Oncology Group, PLT platelet, Hb hemoglobin, eGFR estimated glomerular filtration rate, CRP C-reactive protein, ESR erythrocyte sedimentation rate, WBC white blood cell, ALP alkaline phosphatase, PC plasmacytosis, HV hyaline-vascular, ND not determined.

^apatient no. correlated with Figs. 2I and S6.

^bPatient 13 carried MEFV L110P- E148Q-P369S- R408Q variant, while the other 6 patients were MEFV E148Q-P369S- R408Q mutated.

^cAccompanied by follicular dendritic cell sarcoma. CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; GemOx, gemcitabine and oxaliplatin.

In sum, the high prevalence of MEFV variants in our large CD cohort suggested the involvement of auto-inflammatory mechanisms in CD. Particularly, the activated innate response modified by the MEFV E148Q-P369S-R408Q variant in TAFRO deserves further elucidation.

ScRNA-seq unraveled MEFV-enriched CD16⁺ monocytes likely interacted with naïve B/memory B cells, contributing to IL-6 pathway activation

To investigate cell type variation and MEFV mutation-associated mechanisms, we performed scRNA-seq on PBMCs from a healthy control, the adolescent TAFRO patient (in flare; in remission: 3 months and 5 months post-treatment), and his parents, who carried fewer MEFV mutations but lacked disease manifestations. Cell populations were annotated as 20 cell types based on marker gene expression (Figs. 3A, B and S7A). We first analyzed the proportional distribution of each cell type (Figs. 3C and S7C). Notably, the percentage of naïve B/memory B cells, CD14⁺/CD16⁺ monocytes, and megakaryocytes correlated with MEFV mutation amount, suggesting a functional link. During disease flare, the patient exhibited elevated naïve B/memory B cells and megakaryocytes compared to his parents and the healthy control, which decreased upon remission (Fig. 3C). Conversely, CD14+/CD16+ monocytes showed an inverse trend.

Fig. 3. ScRNA seq unraveled MEFV-enriched CD16⁺ monocytes likely interacted with naïve B/memory B cells, contributing to IL-6 pathway activation, and megakaryocytes implicated in inflammation via CXCL signaling network.

A UMAP of the identified cell types in scRNA-seq of PBMCs. B UMAP visualization based on (A), highlighting the identified cell types in the scRNA-seq dataset generated from the indicated samples. C Bar plots showing the changes in the proportions of representative cell types across different samples. D Bubble plot representing the frequency of expression and scaled average expression of representative IL-6 pathway-related genes (left) and Castleman disease-related genes (right) in all samples. E UMAP visualization based on (A), highlighting the expression of MEFV. F Violin plots showing the log-normalized expression of MEFV in MEFV+ cells. G Heatmap showing the expression patterns of differentially expressed genes (DEGs) in MEFV+ cells from healthy control (Ctrl) and TAFRO patient in flare (P-f) across all samples. GO analysis for each group is shown. H Pseudobulk correlation analysis of MEFV+ cells from all samples. I Violin plots showing the log-normalized expression of MEFV in P-f. CD16⁺ monocytes were highlighted in red. J Violin plots showing the inflammatory score in P-f. K Cell interactions among various cell types in P-f were visualized using CellChat. The scale number represents the strength of interactions between cells. CD16⁺ monocytes were highlighted in red. L Heatmap showing the expression of IL-6 in P-f. M Violin plots showing the IL-6 pathway score in CD16+ monocytes across all samples. N Heatmap showing the expression of CXCL signaling-related genes in megakaryocytes across all samples. O Circle plots highlighting the cell interactions mediated by CXCL signaling networks in P-f. Ctrl healthy control, M mother, F father, P-f patient-flare, P-m3 patient at 3 months since the onset of disease, in remission, P-m5 patient at 5 months since the onset of disease, in remission, CM central memory, EM effector memory.

Analysis of gene expression profiles revealed a strong correlation between disease-associated genes and clinical phenotypes, with flare samples from the patient showing significantly elevated IL-6 pathway and CD-related gene expression (Fig. 3D). To investigate the role of MEFV mutations in disease pathogenesis, we analyzed cell populations with enriched MEFV expression. Consistent with previous reports³⁵, MEFV was predominantly expressed in monocytes (Figs. 3E and S7D, E), and its expression levels were reduced during disease flares (Fig. 3F). Notably, MEFV⁺ cells displayed significant transcriptomic dysregulation during disease flares relative to controls, with restoration of normal expression patterns post-treatment (Fig. 3G, H), suggesting disease-associated functional impairment.

We then examined the patient-flare sample in detail to explore potential disease mechanisms. MEFV-enriched CD16⁺ monocytes exhibited the most robust inflammatory transcriptional profiles (Fig. 3I, J). CellChat analysis identified these monocytes as primary signal receivers, with naïve and memory B cells serving as dominant signal inputs (Fig. 3K), a finding concordant with elevated IL-6 expression in B cells (Fig. 3L). Notably, CD16⁺ monocytes from flare-phase patients displayed the highest IL-6 pathway activity (Fig. 3M). Collectively, these results suggest that MEFV mutations drive functional impairment in CD16⁺ monocytes, rendering them hyper-responsive to B-cell-derived IL-6 signals and exacerbating inflammatory activation. Therapy-induced remission restored the altered immune condition in our TAFRO patient, corrected the dysregulated immune cell compositions, and alleviated immune malfunction.

Beyond monocyte alterations, we observed a marked increase in megakaryocyte frequency in the adolescent TAFRO patient (Fig. 3C), accompanied by elevated inflammatory scores (Fig. 3J). Mechanistically, megakaryocytes primarily function as signal senders (Fig. 3K), exhibiting high expression of CXCL family genes (Fig. 3N). These CXCL signals predominantly target NK T cells and CD8+ central memory T cells (Fig. 3O). Accordingly, CXCL/CCL signaling was remarkably up-regulated in the TAFRO patient after LPS stimulation, as shown in Figs. 2F, S3 and S4. Thus, we propose the potential link between megakaryocytes and TAFRO pathogenesis, where megakaryocyte-derived CXCL chemokines may drive disease-associated inflammation.

Discussion

The present study depicted and comprehensively analyzed the unique adolescent TAFRO patient, covering the procedure of diagnosis, treatment, and follow-up. Pedigree gene investigation, together with in vitro functional experiments and single-cell landscape analysis, identified the MEFV E148Q-P369S-R408Q variant as the potential pathogenic gene event for this rare disease entity. The role of the MEFV variant in CD was further confirmed in a retrospective cohort comprising 37 patients over 11 years from a single center. Our study presents one of the largest cohorts demonstrating the high prevalence of MEFV variants in CD, providing important insights for understanding and treating CD, particularly TAFRO.

Implications of auto-inflammatory mechanisms have been addressed in the pathogenesis of CD. MEFV encoded pyrin, an inflammasome sensor mainly expressed in monocytes and dendritic cells³⁵. Pyrin has been suggested to be crucial in innate immunity via regulation of interleukin-1β (IL-1β) activation³⁵. Most recently, MEFV mutations, the symbolic gene event for FMF, have been implicated in CD^17–19. MEFV mutations, including E148Q/P369S/R408Q, could be sufficient to drive an inflammatory storm in response to foreign stimuli³⁶, prompting us to investigate the role of MEFV variants in our patient with TAFRO and a further large cohort.

Six MEFV variants were identified, with E148Q, P369S, and R408Q being the three most prevalent variants (Fig. 2I, J). Due to the high allele frequency in healthy individuals, whether E148Q (exon 2) is a simple polymorphism or a disease-causing mutation remains unknown. As shown in Fig. 2J, the frequency of E148Q was 25.46% in the East Asian population (data from gnomAD v4.1.0). However, the frequency of E148Q in our CD cohort was as high as 64.86%, significantly higher than that in the East Asian population, suggesting its pathogenesis potential.

While E148Q was one of the most common variants in patients with FMF worldwide, P369S/R408Q (exon 3) was only frequently seen in Asia, particularly in Japan and China³⁷, but not in Mediterranean regions³⁸. Notably, we found that the P369S and R408Q variant haplotypes appeared in strong linkage disequilibrium in CD. Moreover, P369S-R408Q was detected only in E148Q-mutated individuals except one (Fig. 2I, P24). The first case of MEFV heterozygosity in CD was reported in 2009¹⁵. The patient bearing MEFV M694V and M694I¹⁵ or Ile729Met¹⁸ was successfully treated with an IL-6 inhibitor. Similar to our finding, the MEFV variants in 10/14 Japanese patients with iMCD included L110P, E148Q, R202Q, P369S, and R408Q¹⁹. It had been suggested that pyrin dysfunction caused by MEFV germ-line mutations may induce IL-6 production via inflammasome signaling and modify the pathology of iMCD¹⁹. However, evidence from functional experiments is largely lacking. Using the LPS stimulation model and scRNA-seq, we found MEFV expression was dominant in CD16⁺ monocytes and correlated with IL-6 pathway activation, likely via the interaction with naïve B/memory B cells, contributing to the pathogenesis of TAFRO.

While MEFV E148Q-P369S-R408Q was identified in one Chinese³⁹, one Korean⁴⁰, three Japanese^38,41, two French¹⁵, one Armenian⁴² adult with FMF, and two Japanese adults with iMCD¹⁹, to the best of our knowledge, ours is the first report showing a Chinese adolescent with TAFRO carrying complex alleles E148Q-E148Q/P369S-WT/R408Q-R408Q. Furthermore, MEFV E148Q-P369S-R408Q was identified in our cohort as following CD (7/37, 19%), UCD (4/25, 16%), MCD (3/12, 25%), iMCD (3/10, 30%), and TAFRO (2/4, 50%) (Figs. 2I and S5). Additionally, the MEFV E148Q-P369S-R408Q variant conferred more severe disease manifestations in CD. A link between the incidence of MEFV E148Q-P369S-R408Q and disease severity appears to exist, but needs further validation.

According to the recent expert perspective⁴³, UCD comprises asymptomatic UCD and symptomatic UCD, including UCD with systemic inflammation. Indeed, our cohort included 25 UCD, among whom some had systemic inflammation (Fig. 2K). This might explain the presence of MEFV variants in UCD.

The gene dose effect could be proposed. Bloch et al. have recently shown that the outcome of adult hemophagocytic lymphohistiocytosis (HLHa) may have a heterozygous genetic basis⁴⁴. To note, the increased number of variant alleles was correlated with more severe disease conditions, suggesting the gene dose effect on disease severity. In fact, the gene dose effect was also involved in FMF⁴⁵ and CD¹⁹ as well. In our study, we could show gene dose effect of MEFV variants in CD characterized by (1) the adolescent TAFRO patient carried variants of MEFV E148Q/E148Q, P369S/WT, R408Q/R408Q, while his asymptomatic parents had lower MEFV variant allele frequencies (Fig. 2A, B); (2) Patients with CD carrying triple mutations of MEFV exhibited higher disease severity, as shown in Fig. 2K.

Although our patient showed responsiveness to a siltuximab and corticosteroids-containing regimen, siltuximab was discontinued after 5 doses due to financial issues. While the CD patient bearing MEFV M694V and M694I¹⁵ or Ile729Met¹⁸ was successfully treated with an IL-6 inhibitor, another iMCD patient carrying MEFV L110P-E148Q¹⁷ did not respond to anti-IL-6 treatment. Whether a CD with MEFV variants predicts a good response to anti-IL-6 therapy remains to be further investigated.

Colchicine, the common drug used for FMF characterized by MEFV variants, showed partial¹⁷ or no¹⁶ response in patients with iMCD carrying MEFV variants. Accordingly, our in vitro experiments indicated that colchicine inhibited LPS-induced cell aggregation and cytokine release in TAFRO cells with MEFV E148Q-P369S-R408Q (Figs. 2G and S4), suggesting its therapeutic potential. Considering the high prevalence of MEFV variants in CD and the critical role of colchicine in treating FMF, interventions of CD, particularly TAFRO with colchicine, may deserve further investigation, especially for those resistant to or inaccessible to anti-IL-6 therapy. The unresponsiveness or refractoriness of iMCD to IL-6 blockade remains unresolved. Potential novel strategies include thalidomide²⁸, BTK inhibition⁴⁶, TNF inhibition⁴⁷, and others. Our study, together with previous studies¹⁷, implied the possibility of auto-inflammation inhibition for treating iMCD.

The candidate dysregulated cell types implicated in CD, especially in TAFRO, still remain unclear^11,14.To our knowledge, scRNA-seq for TAFRO was reported in only four recentstudies^47–50. While our data were consistent with recent studies showing naïve B/memory B cells as the main producers of IL-6 in iMCD⁴⁹/TAFRO⁴⁸, the role of monocytes in iMCD, particularly in TAFRO were still controversial^48–51. The discrepancy between these studies highlighted the heterogeneous immune background of this rare disease entity. In our unique adolescent TAFRO with MEFV variant, we found a reduced fraction of CD16⁺ monocytes in flare, characterized by high inflammatory and IL-6 pathway scores (Fig. 3J, M), likely contributing to the amplification of the inflammatory response via the strong interaction with naïve B/memory B cells (Fig. 3K). However, the co-culture of these monocytes with B cells is required for further function validation. Interestingly, MEFV was exclusively enriched in CD16⁺ but not in CD16⁻ monocytes (Figs. 3I and S7E), suggesting its unique role in inflammasome signaling. In fact, it has been proposed that pyrin malfunction resulting from MEFV mutations could induce the production of IL-6 by inflammasome signaling and contribute to the development of iMCD¹⁸. The precise interplay between MEFV and CD16 requires further elucidation.

Additionally, we found that megakaryocytes were extremely highly expressed in flare of the TAFRO patient (Fig. 3N), consistent with the most recent report⁴⁸. However, no more details of megakaryocytes were described in that study. Furthermore, circle plots highlighted a remarkable correlation between the megakaryocyte and CXCL signaling network contributing to the inflammation in TAFRO (Fig. 3O). In line with this, CXCL/CCL signaling was robustly activated in the TAFRO patient after LPS stimulation (Figs. 2F, S3 and S4). We hypothesized that the severe bone marrow fibrosis in our TAFRO case compromised the integrity of the bone-blood barrier, facilitating megakaryocyte egress into the peripheral circulation and potentiating systemic inflammatory responses. To note, the role of megakaryocytes in immune activation in other diseases has been well illustrated^52,53.

Furthermore, CCL3, one of the key cytokines secreted from NK T cells, was extremely elevated in the patient after LPS stimulation (Fig. S3). Our data thus suggested a potential role of NK T cells in the pathogenesis of CD. However, the precise effect of NK T cells in the disease remains to be further investigated.

Our study has some limitations. First, the sample size of our cohort was relatively small due to the rarity of the disease. Second, functional tests were performed only in the adolescent TAFRO and his family, but not in the other retrospectively analyzed patients, limiting generalizability. More functional tests, such as flow cytometry and co-culture experiments, are encouraged to be performed for further validation. Third, the iMCD pathogenesis is quite complex. Other cellular factors or variants in other genes are also involved in inflammatory regulation, such as CXCL³⁴, TNF⁴⁷, IFN-γ⁴⁸, NCOA4^51,54. Our in vitro experiments also revealed the role of CXCL, TNF, and IFN-γ in the inflammation response (Figs. 2F, G, S3 and S4B). As shown in our present study and other reports^15–19, MEFV may act as a modifier gene in the pathogenesis of iMCD. However, a direct causal link has not been established. Last, the potential impact of therapies (e.g., steroids, IL-6 blockade) on the cytokine and single-cell analyses can significantly alter cellular phenotypes.

In conclusion, we report MEFV variant E148Q/E148Q-P369S/WT-R408Q/R408Q in the unique adolescent TAFRO showing good responsiveness to IL-6 blockade-containing treatment. In vitro stimulation experiment and scRNA-seq revealed MEFV-enriched CD16⁺ monocytes likely interacted with naïve B/memory B cells, contributing to IL-6 pathway activation. We also provided the potential link of megakaryocytes implicated in inflammation via the CXCL signaling network in TAFRO. A high incidence of MEFV variants, together with more severe disease manifestations in the MEFV E148Q-P369S-R408Q-positive CD from one of the largest cohorts, suggested a potential role in the pathogenesis of CD.

Author contributions

Y.D., Z.D., W.J., W.L., K.H. analyzed the data and prepared the manuscript; L.Z. provided pathological assessment of tissues; G.D. assisted clinical imaging analysis and drafted the figures; S.X., M.W., Y.L., W.Y., J.X., Y.Z., J.L., W.Z. provided clinical information and samples and performed follow up of patients; Y.D., K.H. performed experiments; H.Z. contributed to language editing and manuscript improvement; Q.L. contributed to data analysis; K.H., Y.D., Z.D., S.J., D.N. conceptualized the study, interpreted the data, and drafted the manuscript; all authors agreed to submit the manuscript, approved the final draft, and take full responsibility for its content.

Peer review

Peer review information

Communications Medicine thanks Luke Y. C. Chen, David Boutboul, Ruth-Anne Langan Pai, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The source data for Figs. 1G–I, 2C, E–G, J, K, 3A, B, J, L, M, S3 and S4B are in Supplementary Data. Sequencing data have been uploaded via http://db.cngb.org/cnsa/project/CNP0007486_f4cf18fa/reviewlink/. All relevant data in the manuscript and Supplementary Information are available from the authors upon reasonable request to K.H. No restrictions exist on data availability. If it refers to generated data, 1 month for response is required.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-026-01392-1.