Single-cell RNA sequencing reveals that kimchi dietary intervention modulates human antigen-presenting and CD4⁺ T cells

Abstract

Kimchi, a traditional Korean fermented food, is recognized for its metabolic benefits; however, its effects on human immune function remain poorly defined. In this study, we performed single-cell RNA-seq on peripheral blood mononuclear cells from 13 overweight adults using a paired pre/post design in a 12-week randomized, double-blind, placebo-controlled trial. The participants consumed a placebo, spontaneously fermented kimchi powder, or starter culture-fermented kimchi powder. Kimchi enhanced intercellular signaling mediated by antigen-presenting cells, increased antigen uptake, and promoted the upregulation of MHC class II–related genes through the JAK/STAT1–CIITA axis. Single-cell trajectory analysis revealed accelerated CD4⁺ T cell differentiation toward effector and regulatory phenotypes, whereas CD8⁺ T cells, B cells, and NK cells remained stable, indicating preserved systemic immune homeostasis. In summary, 12 weeks of kimchi dietary intervention enhanced antigen presentation and remodeled CD4⁺ T cells without broad systemic activation, providing single-cell evidence of kimchi-induced dietary immunomodulation in humans. The clinical trial was registered at ClinicalTrials.gov (https://www.clinicaltrials.gov/) under the identifier NCT05898802.

Introduction

Kimchi, a traditional Korean fermented vegetable dish, has attracted global interest due to its health-promoting properties^1,2. Its fermentation is primarily driven by lactic acid bacteria (LAB), including Leuconostoc, Weissella, and Lactobacillus, which produce diverse bioactive metabolites, including organic acids, vitamins, peptides, and exopolysaccharides^3,4. These microbial and metabolic constituents exert antioxidant, anti-obesity, and immunomodulatory effects^5,6.

The functional benefits of kimchi stem from a complex dietary matrix comprising LAB, fermentation-derived metabolites, and plant-derived bioactives^7–9. These components act synergistically to influence host physiology, including immune-related responses^10,11. Similar synergistic effects have also been reported for other fermented foods, in which the microbial diversity and metabolic complexity contribute to immune modulation^12–15. The compositional diversity of kimchi may differentially modulate immune cell activation, differentiation, and function, given that the immune system is highly responsive to dietary microbes and their byproducts.

Despite this potential, clinical studies directly examining the immune effects of kimchi are limited. Most human intervention trials have focused on metabolic parameters or gut microbiota composition, offering only indirect evidence of systemic immunomodulation^16,17. Furthermore, while probiotic studies using isolated kimchi-derived strains have demonstrated immunological benefits, such reductionist approaches fail to capture the complex diet–microbiota–host interactions that occur during the consumption of whole kimchi^18–20. Concerns have been raised in global nutritional immunology, where single-strain probiotic studies often fail to reflect the complexity of whole-diet interventions²¹. Given that the gut–immune axis is influenced by diverse dietary, microbial, and metabolic inputs, a holistic evaluation of kimchi’s immunological impact is warranted^22,23.

Current immunological assessment tools—such as serum cytokine profiling, flow cytometry, and ex vivo stimulation—often lack the resolution needed to detect subtle or cell-type–specific shifts, largely because of inter-individual variability and reliance on predefined surface markers²⁴. Although bulk transcriptomics offers broader insights into immune-related gene programs, it averages gene expression across heterogeneous cell populations, obscuring cell-specific remodeling and rare immune subsets²⁵.

Single-cell RNA sequencing (scRNA-seq) enables high-resolution profiling of transcriptional activity at the level of individual cell types and states^26–28. This approach enables the identification of subtle immune responses and intercellular signaling networks that conventional techniques often fail to capture. Such resolution is particularly valuable in nutritional immunology, where biologically meaningful effects may occur within rare or functionally specialized immune populations²⁹. Given the microbial and metabolic complexity of kimchi, scRNA-seq provides a powerful platform to uncover cell-specific immune modulation associated with its consumption³⁰. Indeed, this strategy has also been applied to dietary and microbiome interventions globally, revealing immune dynamics not captured by traditional methods³¹.

In this study, we performed scRNA-seq on peripheral blood mononuclear cells (PBMCs) collected from overweight adults (BMI 23–30 kg/m²) who were otherwise healthy, before and after a 12-week kimchi dietary intervention. We sought to comprehensively profile systemic immune remodeling at single-cell resolution. Additionally, by comparing two fermentation approaches—spontaneous fermentation (S-K) and starter-culture fermentation (LMS-K)—we investigated whether the method of fermentation differentially affects immunological outcomes. This system-level analysis was designed to elucidate how sustained kimchi intake modulates human immune responses.

Results

Single-cell transcriptomic landscape of PBMCs following intervention

The randomized, double-blind, placebo-controlled trial involved 90 overweight adults (BMI 23–30 kg/m²) and primarily examined the effects of kimchi powder intake on obesity and gut microbiota composition³². Building on this cohort, we conducted a single-cell transcriptomic analysis of PBMCs collected before and after the 12-week intervention to investigate the immunological effects of kimchi.

For this analysis, 13 participants were randomly selected and distributed across the three intervention groups: placebo (n = 4), S-K (n = 5), and LMS-K (n = 4). The baseline characteristics did not significantly differ between the groups, as summarized in Table 1.

Table 1.

Baseline characteristics of participants

Variable	Placebo (n = 4)	S-K (n = 5)	LMS-K (n = 4)	p value
Age (years)	48 (9.42)	42.4 (14.36)	59.5 (2.89)	0.098
Sex				0.263
Female	4 (100%)	3 (60%)	2 (50%)
Male	0 (0%)	2 (40%)	2 (50%)
BMI (kg/m2)	25.63 (0.95)	24.44 (1.83)	24.8 (2.13)	0.594
Glucose (mg/dL)	100 (7.62)	93.4 (5.5)	95 (6.06)	0.326
HbA1c (%)	5.82 (0.37)	5.68 (0.38)	5.83 (0.39)	0.811
Total cholesterol (mg/dL)	230.25 (35.22)	200.6 (24.4)	200.75 (50.41)	0.444
Triglyceride (mg/dL)	106.5 (40.45)	102.4 (42.78)	163 (56.16)	0.160
HDL (mg/dL)	59.5 (10.66)	55.4 (10.71)	50.5 (7.14)	0.455
LDL (mg/dL)	163.75 (29.1)	135.8 (20.95)	138.75 (46.18)	0.427

Data are expressed as the mean (SD) or percentage.

P values were calculated using ANOVA for continuous variables and chi-square tests as an approximation to the Fisher–Freeman–Halton exact test for categorical variables.

The BD Rhapsody platform was used to analyze PBMCs, yielding 88,403 high-quality single-cell transcriptomes (Fig. 1A). Unsupervised clustering followed by Uniform Manifold Approximation and Projection (UMAP) revealed nine distinct immune cell populations (Fig. 1B). The clustering structure was consistently maintained across groups and timepoints, indicating successful data integration and normalization with minimal batch effects (Supplementary Fig. 1). Cell types were annotated using canonical marker gene expression (Fig. 1C, D) and validated using SingleR³³. CD4⁺ T cells expressed TCF7 and LEF1, consistent with a helper-like memory or naïve phenotype. CD8⁺ T cells expressed CD8A/CD8B but lacked strong cytotoxic gene expression, suggesting a non-terminal effector state. NK-featured T (NKFT) cells expressed TRGC2 and NKG7, indicating an innate or NKT-like phenotype. NK cells showed high GNLY and IL2RB expression, typical of cytotoxic effector function. B cells were identified by the expression of MS4A1 and CD79A, while a subset lacking MS4A1 and IGHM but expressing innate-like genes was detected. Monocytes showed typical S100A8, S100A9, and LYZ expression, with a neighboring cluster expressing VCAN, NAMPT, and CSF3R, indicative of a non-classical or intermediate profile. Dendritic cells (DCs) expressed human leukocyte antigen-DR alpha (HLA-DRA), CD74, and transcriptional regulators, such as IRF8 and TCF4, consistent with both conventional and plasmacytoid dendritic cell (pDC)-like subtypes. A comprehensive list of marker genes used for annotation is provided in Supplementary Table 1.

Fig. 1. Overview of single-cell transcriptomic profiling and immune cell annotation following intervention.

A Peripheral blood mononuclear cells (PBMCs) from 13 participants (placebo, n = 4; S-K, spontaneously fermented kimchi, n = 5; LMS-K, starter kimchi fermented with Leuconostoc mesenteroides KCKM0828, n = 4) in the randomized, placebo-controlled trial were collected before and after the 12-week intervention. Samples were barcoded with sample-specific tags, pooled (multiplexed), and processed using the BD Rhapsody system to generate single-cell gene expression libraries for next-generation sequencing. B Uniform Manifold Approximation and Projection (UMAP) of 88,403 high-quality single-cell transcriptomes, colored by custom cell-type annotations. A total of nine transcriptionally distinct immune cell populations were identified, with cell counts indicated for each cluster. C Dot plot showing the expression of canonical marker genes across annotated immune cell types. The size of the dots indicates the percentage of cells expressing the gene, and the color intensity indicates the average expression level. D Heatmap of the marker genes for each cell type, ordered by their relative proportions. E Stacked bar plots showing the relative proportions of major annotated immune cell types before and after the 12-week intervention in each group (placebo, S-K, and LMS-K).

The distribution of PBMC subtypes remained within physiological ranges across all groups and timepoints, with only modest changes observed following kimchi or placebo intake. CD4⁺ T cells were the most prevalent (50.3–62.6%), followed by NK-like T cells (13.0–20.3%), B cells (5.5–15.7%), NK cells (4.3–10.1%), and CD8⁺ T cells (0.8–3.0%) (Fig. 1E). Overall immune cell composition was largely maintained regardless of kimchi or placebo intake.

Kimchi intake reinforced antigen-presenting cell–driven immune interactions

To evaluate how kimchi intake modulates peripheral immune communication, we conducted paired CellChat analysis^34,35 on single-cell transcriptomes of PBMCs collected before and after intervention across the three groups: placebo (Fig. 2A, B), S-K (Fig. 2C, D), and LMS-K (Fig. 2E, F). The resulting networks revealed group-specific differences, where the number of interactions (Fig. 2A, C, E) indicated network connectivity and signaling strength (Fig. 2B, D, F) reflected the intensity of intercellular communication.

Fig. 2. Remodeling intercellular immune communication in PBMCs following the intervention.

Circle plots show the differential number of ligand–receptor interactions (A, C, E) and the differential interaction strength (B, D, F) between nine major immune cell populations, comparing before versus after intervention in the placebo (A, B), S-K (C, D), and LMS-K (E, F) groups. Each node represents a distinct immune cell type, with CD4⁺ T cells shown in red, CD8⁺ T cells in orange, NK-featured T cells in blue, NK cells in purple, B cells in green, innate-like B cells in brown, monocytes in pink, monocyte-like cells in lavender, and dendritic cells in teal. Edges denote ligand–receptor-mediated interactions. The width of the edges reflects the magnitude of change, and the color of the edges indicates the direction of change, with red representing increased interactions and blue representing decreased interactions when comparing before versus after intervention.

In the placebo group, the immune cell communication network exhibited both selective decreases and increases in connectivity, suggesting that intercellular interactions were reorganized rather than uniformly altered. Notably, DCs exhibited a reduced number of interactions with monocytes, NK cells, and NKFT cells, whereas NK cells exhibited increased connectivity with other immune cell types (Fig. 2A). Furthermore, differential interaction strength analysis (Fig. 2B) revealed a broad reduction in signaling strength in DCs, monocytes, and innate-like B cells. In the S-K group, kimchi consumption selectively expanded the intercellular immune communication network. Innate-like B cells showed the most substantial increase in interaction number, engaging in newly formed or strengthened connections with most immune cell types, except B cells (Fig. 2C). In parallel, the interaction strength analysis (Fig. 2D) revealed a broader increase in signaling intensity across the network. Notably, DCs, monocytes, and innate-like B cells exhibited marked increases in both the interaction number and the signaling strength, indicating their substantial contribution to the expanded intercellular communication network. Similarly, kimchi consumption in the LMS-K group remodeled the immune communication network. Interaction number analysis (Fig. 2E) indicated that NK cells exhibited the most substantial gain in connectivity, whereas DCs, monocytes, and innate-like B cells showed modest increases. Consistently, the interaction strength analysis (Fig. 2F) revealed marked increases in signaling intensity from these cell types, indicating functional enhancement of their immunological interactions.

In summary, kimchi consumption, irrespective of fermentation method, selectively enhanced immune communication through monocytes, DCs, and innate-like B cells, thereby increasing the connectivity and strength of peripheral signaling networks. In contrast, the same cell types in the placebo group exhibited reduced activity, suggesting that they serve as key mediators of kimchi-induced immune modulation.

Kimchi intake enhanced major histocompatibility complex (MHC)-II–mediated signaling in monocytes and DCs

To explore how kimchi intake shapes immune communication, we focused on monocytes, DCs, and innate-like B cells—the three key populations consistently affected by the intervention. Outgoing and incoming interactions, defined by ligand and receptor expression, respectively, revealed how kimchi modulates intercellular signaling.

In monocytes, the placebo group exhibited a marked reduction in incoming MHC-II–mediated signals (Δ in ≈ −0.63), accompanied by a modest increase in outgoing MHC-I–associated communication (Δ out ≈ +0.36) (Fig. 3A). This pattern was reversed in the S-K group, where MHC-II signaling was upregulated in both directions (Δ out ≈ +0.65; Δ in ≈ +0.61) and MHC-I showed a strong unidirectional increase in outgoing strength (Δ out ≈ +0.9; Δ in ≈ −0.02) (Fig. 3D). The LMS-K group displayed a similar trend, with MHC-II upregulated in both directions (Δ out ≈ +0.61; Δ in ≈ +0.24) and a modest enhancement of MHC-I–mediated outgoing communication (Δ out ≈ +0.27; Δ in ≈ +0.02) (Fig. 3G). Apart from these dominant axes, the majority of the other signaling pathways, including those associated with ICAM and GALECTIN, showed moderate increases in outgoing signaling.

Fig. 3. Differential interaction strength of ligand–receptor pairs in monocytes, dendritic cells, and innate-like B cells before and after intervention.

Differential ligand–receptor interaction strengths were analyzed in monocytes (A, D, G), dendritic cells (B, E, H), and innate-like B cells (C, F, I) across the three experimental groups: placebo (A–C), S-K (D–F), and LMS-K (G–I). Each panel plots ligand–receptor pairs by the change (Δ = after−before) in outgoing signaling strength (x-axis) and incoming signaling strength (y-axis) between before and after intervention. Dashed lines at x = 0 and y = 0 indicate no change. Each point is colored by temporal specificity: black for shared, red for before-specific, and teal for after-specific. Point shapes indicate directionality—circles for no directional change (shared), squares for incoming-specific, triangles for outgoing-specific, and diamonds for changes in both directions. This categorical encoding complements the continuous values on the x- and y-axes.

A comparable modulation profile was observed in the DCs (Fig. 3B, E, H). In the placebo group, MHC-II exhibited the largest bidirectional decrease among all signaling programs (Δ out ≈ −0.36; Δ in ≈ −0.21), along with reductions in APP signaling (Fig. 3B). In contrast, MHC-I exhibited a mild increase in outgoing activity (Δ out ≈ +0.26; Δ in ≈ −0.04). In the S-K group, DCs exhibited robust bidirectional enhancement of MHC-II (Δ out ≈ +1.35; Δ in ≈ +2.1), along with increased APP and GALECTIN signaling (Fig. 3E). MHC-I also demonstrated a substantial unidirectional gain (Δ out ≈ +0.88). The LMS-K group showed a similar but attenuated profile, with MHC-II (Δ out ≈ +0.52; Δ in ≈ +0.08) and MHC-I (Δ out ≈ +0.32) moderately upregulated (Fig. 3H). The other communication channels remained relatively unchanged.

In innate-like B cells, the placebo group displayed divergent behavior in MHC-I–mediated signaling, with a slight increase in outgoing activity (Δ out ≈ +0.15) but a marked reduction in incoming strength (Δ in ≈ −0.50) (Fig. 3C). MHC-II signaling also declined in both directions (Δ out ≈ −0.09; Δ in ≈ −0.49). Following S-K intake, MHC-I was strongly upregulated bidirectionally (Δ out ≈ +0.95; Δ in ≈ +4.1), while MHC-II showed a unidirectional increase in outgoing strength (Δ out ≈ +0.65) (Fig. 3F). In the LMS-K group, MHC-I showed a bidirectional increase (Δ out ≈ +0.26; Δ in ≈ +1.55), and MHC-II signaling increased in both directions (Δ out ≈ +0.40; Δ in ≈ +0.37) (Fig. 3I). Other signaling interactions remained largely unchanged.

Taken together, these findings indicate that kimchi intake selectively remodeled intercellular immune communication across distinct cell types and pathways. Monocytes and DCs consistently exhibited increased MHC-II–mediated signaling, reflecting enhanced antigen-presenting activity. In contrast, innate-like B cells showed selective upregulation of MHC-I–mediated interactions, particularly in the incoming direction.

Dietary kimchi promoted DC antigen uptake and facilitated MHC class II–mediated antigen presentation

Considering that kimchi intake enhances MHC class II-mediated signaling in DCs, we subsequently investigated whether these effects correspond to the transcriptional regulation of MHC II–related genes (Fig. 4A, B). Gene expression profiling of DCs within the PBMC population revealed that the expression of major MHC II genes was unaltered or marginally reduced in the placebo group. In contrast, the S-K group exhibited robust upregulation across multiple components of the MHC II pathway, including marked increases in HLA-DRA (Δ + 9.6) and human leukocyte antigen–DR beta1 (HLA-DRB1) (Δ + 3.1), accompanied by the reactivation of additional isoforms such as HLA-DRB5, HLA-DPA1, and HLA-DPB1 (Fig. 4B). These findings indicate a coordinated transcriptional enhancement of the antigen presentation axis following S-K intake. The LMS-K group also demonstrated increased expression of HLA-DRA (Δ + 4.7) and HLA-DRB1 (Δ + 0.1), although the magnitude and diversity of gene induction were consistently lower than those in the S-K group (Fig. 4B).

Fig. 4. Functional validation of antigen uptake and MHC class II induction by kimchi treatment.

A Heatmap showing log-normalized expression of MHC class II genes and CIITA in dendritic cells across groups (placebo, S-K, LMS-K; before vs. after intervention). B Heatmap depicting differential expression (after–before) of MHC class II genes and CIITA per group. C, D Flow cytometry analysis of antigen uptake in PMA-differentiated THP-1 macrophages treated with S-K or LMS-K compared with vehicle controls. C Representative histograms of DQ-ovalbumin uptake, with vehicle samples shown in gray, S-K samples shown in green (left), and LMS-K samples shown in green (right). D Quantification of DQ-ovalbumin uptake expressed as the geometric mean fluorescence intensity (MFI). E, F Flow cytometry analysis of MHC class II surface expression in the same setting. E Representative histograms of HLA-DR expression, with vehicle samples shown in gray, S-K samples in red (left), and LMS-K samples in red (right). F Quantification of HLA-DR expression expressed as the geometric MFI across the indicated conditions. Statistical significance was determined using the two-sided Mann–Whitney U test (****p < 0.0001).

To functionally validate the transcriptomic changes induced by kimchi intake, we assessed antigen uptake using DQ-ovalbumin in phorbol 12-myristate 13-acetate (PMA)-differentiated THP-1 macrophages and measured HLA-DR surface expression using flow cytometry (Fig. 4C–F). To rule out potential cytotoxicity, we initially evaluated cell viability and confirmed that neither S-K nor LMS-K affected the viability of lipopolysaccharide (LPS)-treated THP-1 cells (Supplementary Fig. 2A, B). Under these non-cytotoxic conditions, both treatments enhanced antigen uptake and proteolytic processing of DQ-ovalbumin compared with vehicle controls (Fig. 4C), with quantitative analysis of geometric mean fluorescence intensity (MFI) showing significant increases (Fig. 4D; both S-K and LMS-K, p < 0.0001). Consistently, kimchi-treated cells displayed elevated HLA-DR surface expression (Fig. 4E), and statistical analysis also revealed significant increases relative to controls (Fig. 4F; both S-K and LMS-K, p < 0.0001).

Together, these results confirm that kimchi-derived factors enhance antigen uptake, processing, and overall antigen-presenting activity in vitro, consistent with the transcriptomic upregulation of MHC class II genes observed in vivo.

Kimchi-induced MHC-II upregulation was associated with IFN-γ–responsive pathways

To assess whether the transcriptional upregulation of MHC-II genes observed in vivo is functionally linked to broader immunoregulatory programs, we examined the correlations between HLA-DRA and HLA-DRB1 expression in DCs and hallmark-derived pathway module scores (Fig. 5A). Pathway scores—including IFN-α/γ, complement, TNF-α, IL6/JAK/STAT3, inflammatory response, and TGF-β—were generated using the AddModuleScore function in the Seurat package³⁶. In both the S-K and LMS-K groups, HLA-DRA and HLA-DRB1 expressions exhibited strong positive correlations with IFN-γ response (ρ = 0.41–0.57), IL6/JAK/STAT3 signaling (ρ = 0.48–0.57), and TNF-α signaling (ρ = 0.37–0.48). The placebo group followed a similar pattern, albeit with lower correlation coefficients. These results suggest an association between kimchi intake–induced upregulation of MHC-II–related genes and multiple immune activation pathways.

Fig. 5. Association of kimchi-induced MHC-II regulation with IFN-responsive signaling pathways.

A Association analysis including Spearman correlation coefficients (ρ) between HLA-DRA or HLA-DRB1 expression in dendritic cells and selected hallmark pathway module scores. All coefficients are shown; statistical significance is indicated by asterisks (*p < 0.05, **p < 0.01, ***p < 0.001) based on Spearman’s rank correlation with Benjamini–Hochberg adjustment. B, C Scatter plots of qPCR array measurements for interferon-responsive genes in PMA-differentiated THP-1 macrophages treated with LPS alone (x-axis) or LPS + kimchi supernatant (y-axis): B S-K and C LMS-K. The gray diagonal denotes y = x; red and blue dashed lines indicate two-fold differences relative to LPS alone. Data represent the means of biological duplicates.

Among the correlated pathways, IFN-γ is a well-established upstream regulator of MHC-II and CIITA expression. To investigate whether IFN-γ–related signaling is involved in the transcriptional changes observed in vivo, we conducted in vitro experiments using PMA-differentiated THP-1 macrophages. The cells were treated with LPS alone or in combination with S-K or LMS-K, and IFN-related gene expression was analyzed using a qPCR array (Fig. 5B, C, Supplementary Table 4). Both supernatants upregulated the expression of IFNGR1 and IRF1 while suppressing type I IFN-associated genes, with LMS-K showing a somewhat higher expression of several IFN-γ-related genes.

These results suggest that kimchi supernatants activate IFN-γ-responsive genes while suppressing type I IFN-associated programs, thereby linking the in vivo upregulation of MHC-II genes to IFN-γ-related signaling.

Kimchi enhanced antigen-presenting activity in THP-1 macrophages via the JAK/STAT1–CIITA–MHC II axis

Subsequently, we examined JAK1/2 and STAT1 phosphorylation, a downstream event of IFN-γ signaling and a key upstream regulator of MHC class II gene expression, to determine whether kimchi supernatant-induced MHC-II upregulation is mediated through this pathway.

Western blot analysis showed that LPS stimulation markedly increased the phosphorylation of JAK1/2 and STAT1 in PMA-differentiated THP-1 macrophages compared with unstimulated controls. Treatment with either S-K or LMS-K supernatants attenuated this LPS-induced hyperphosphorylation, while maintaining phosphorylation levels above those of unstimulated cells (Fig. 6A–D). In parallel, qPCR analysis confirmed that both S-K and LMS-K supernatants significantly enhanced the expression of CIITA, HLA-DRA, and HLA-DRB1, with LMS-K consistently inducing higher expression levels across all three genes (Fig. 6E–G; all p < 0.01).

Fig. 6. Kimchi-mediated MHC class II regulation via JAK/STAT–CIITA in THP-1 macrophages.

A–D Western blot and densitometry of JAK/STAT signaling in PMA-differentiated THP-1 macrophages (LPS-stimulated; ±S-K or LMS-K supernatants; ±ruxolitinib). A Representative blots for p-JAK1/JAK1 and p-JAK2/JAK2; GAPDH as the loading control. B Densitometric quantification of (A) (phospho/total ratios); traces without ruxolitinib are shown in green and with ruxolitinib in purple; circles indicate JAK1 and squares indicate JAK2. C Representative blots for p-STAT1(Tyr701), p-STAT1(Ser727), and STAT1 under the same conditions as in (A). D Densitometric quantification of (C) (p-STAT1/STAT1 at Tyr701 and Ser727; phospho/total ratios); traces without ruxolitinib in green and with ruxolitinib in purple; circles indicate Ser727 and squares indicate Tyr701. qRT-PCR of CIITA (E), HLA-DRA (F), and HLA-DRB1 (G) in LPS-stimulated THP-1 macrophages under the same conditions (LPS-stimulated; ±S-K or LMS-K supernatants; ±ruxolitinib). Bars show the mean ± SEM of three biological replicates. Statistical significance was determined using the two-sided Student’s t test (**p < 0.01, ***p < 0.001).

To assess the role of JAK/STAT1 signaling in kimchi-induced gene regulation, the cells were pretreated with the JAK1/2 inhibitor ruxolitinib (Fig. 6A–D). Ruxolitinib treatment effectively suppressed JAK phosphorylation and markedly reduced downstream STAT1 phosphorylation (Fig. 6A–D), thereby completely mitigating the kimchi-induced upregulation of CIITA, HLA-DRA, and HLA-DRB1 (Fig. 6E–G; all p < 0.001). These findings indicate that the transcriptional effects of kimchi are dependent on JAK/STAT1 signaling.

To functionally assess the JAK/STAT1 dependency of kimchi-induced antigen-presenting activity, we evaluated antigen uptake/processing (DQ-ovalbumin assay) and HLA-DR surface expression in the presence or absence of ruxolitinib pretreatment (Fig. 7A–D). Both S-K and LMS-K supernatants significantly enhanced DQ-ovalbumin uptake (Fig. 7A, B; p < 0.0001 for both S-K and LMS-K) and HLA-DR expression (Fig. 7C, D; p < 0.0001 for both S-K and LMS-K). Ruxolitinib pretreatment markedly reduced these effects (Fig. 7A–D; p < 0.0001 for both S-K and LMS-K). Nevertheless, DQ-ovalbumin uptake and HLA-DR expression remained significantly above baseline despite ruxolitinib pretreatment (Fig. 7A–D; p < 0.0001 for both S-K and LMS-K), indicating residual activity independent of JAK/STAT1.

Fig. 7. Functional validation of JAK/STAT1-dependent antigen-presenting activity by kimchi supernatants.

A Representative histograms of DQ-ovalbumin uptake in PMA-differentiated THP-1 macrophages treated with vehicle (gray), S-K (green, left), or LMS-K (green, right); ruxolitinib traces are shown in purple. B Quantification of DQ-ovalbumin uptake expressed as geometric mean fluorescence intensity (MFI). C Representative histograms of HLA-DR surface expression measured in the same setting, with vehicle in gray, S-K (red, left), LMS-K (red, right), and ruxolitinib pretreatment in blue. D Quantification of HLA-DR expression expressed as geometric MFI across the indicated conditions. Data are presented as mean ± SEM. Statistical significance was determined using the two-sided Mann–Whitney U test (****p < 0.0001).

Together, these results demonstrate that kimchi-induced enhancement of antigen-presenting capacity is primarily mediated through the JAK/STAT1–CIITA–MHC II axis, although additional pathways appear to contribute to the residual functional activity.

Kimchi intake was associated with selective modulation of CD4⁺ T cell differentiation without broad changes in CD8⁺ T cell states

Building on the observed antigen-presenting cells (APCs) remodeling, we subsequently examined how sustained kimchi intake influences downstream T cell subsets. To this end, we conducted a focused analysis of CD4⁺ and CD8⁺ T cells using integrated clustering and pseudotime-based inference. Re-clustering of T cells from the integrated PBMC dataset revealed seven transcriptionally distinct subtypes (Fig. 8A–C), including CD4⁺ effector memory, naïve/memory-like, RORA⁺ effector, and activated effector T cells, as well as proliferating, regulatory (Treg), and early-activated CD8⁺ T cells. Annotation was based on canonical marker combinations (e.g., TCF7, FOXP3, and CD8A), and full gene lists are provided in Supplementary Table 2.

Fig. 8. Functional heterogeneity and trajectory dynamics of CD4⁺ and CD8⁺ T cells in response to the intervention.

A UMAP of the following seven T cell subclusters identified from integrated CD4⁺ and CD8⁺ T cell populations: CD4 activated effector, CD4 effector memory, CD4 naïve/memory-like, CD4 RORA⁺ effector, early-activated CD8⁺ T cells, proliferating T cells, and Treg cells. B Dot plot displaying the top five marker genes for each subcluster. Dot size reflects the percentage of cells expressing each gene, and color intensity indicates average expression level. C Heatmap of the marker genes per subcluster. Yellow indicates high expression; purple indicates low expression. D Paired box plots comparing the proportions of each T cell subcluster before (blue) and after (red) intervention across the three groups (placebo, S-K, LMS-K). Paired Wilcoxon signed-rank tests were used to assess significance. Asterisks indicate *p < 0.05; “n.s.” indicates non-significant differences. E Pseudotime density plots of CD4⁺ T cell functional subsets—Th1-like (top), Th2-like (middle), and Treg (bottom)—in each group. Blue and orange curves represent before and after intervention, respectively. Corresponding jitter plots below each panel display single-cell pseudotime values. P values from two-sample Kolmogorov–Smirnov tests are shown. F Pseudotime density plots for CD8⁺ T cell subsets—pre-activated (top), cytotoxic (middle), and exhausted (bottom)—using the same layout and statistical evaluation as in (E).

Quantitative analysis of subtype proportions before and after the intervention revealed selective changes (Fig. 8D). Proliferating T cells were significantly increased in the S-K (p = 2.1 × 10⁻⁴) and LMS-K (p = 4.5 × 10⁻³) groups, but not in the placebo group, indicating kimchi-specific promotion of T cell proliferation. In contrast, CD4⁺ RORA⁺ effector and naïve/memory-like subsets were consistently reduced across all groups (p < 0.05), suggesting nonspecific effects independent of intervention. Other subtypes—including effector memory, cytotoxic CD8⁺ T cells, and Tregs—showed minimal or inconsistent shifts in abundance.

To evaluate functional differentiation, pseudotime density analysis was conducted on CD4⁺ T subtypes (Th1-like, Th2-like, and Tregs) using subtype-specific marker genes, such as TBX21 (Th1), GATA3 (Th2), and FOXP3 (Tregs) (Fig. 8E; full gene sets in Supplementary Table 3). In the placebo group, Th1-like cells maintained a bimodal pseudotime distribution (peaks at ~12 and ~20; p = 4.3 × 10⁻¹¹), whereas the S-K group showed a more consolidated distribution toward the terminal stage (p = 1.3 × 10⁻²⁵). LMS-K also induced a rightward shift (p = 5.4 × 10⁻¹⁰), albeit less pronounced. Th2-like cells followed similar patterns, with terminal-stage enrichment observed in both kimchi groups (S-K, p = 7.5 × 10⁻⁴³; LMS-K, p = 1.9 × 10⁻¹⁷) but not in the placebo group (p = 1.3 × 10⁻²³). Tregs also showed pseudotime advancement in all groups, but only the kimchi groups exhibited unimodal density curves, suggesting more convergent differentiation patterns (S-K, p = 4.3 × 10⁻³⁷; LMS-K, p = 2.5 × 10⁻³²).

CD8⁺ T cells were subclassified into cytotoxic, pre-activated, and exhausted states based on well-defined markers (e.g., GZMB, PDCD1) (Fig. 8F; full marker sets in Supplementary Table 3). Pseudotime distributions before and after the intervention showed no significant shifts in any group, suggesting that CD8⁺ T cell differentiation was largely stable regardless of kimchi intake.

Collectively, these findings suggest that kimchi intake is associated with modest but selective modulation of CD4⁺ T cell differentiation, particularly along the effector and regulatory lineages, while CD8⁺ T cell states remained largely unchanged. These results imply the potential of kimchi to influence helper T cell dynamics without inducing broad immune activation.

Kimchi intake preserved B cell subset stability with modest functional modulation

To further evaluate the impact of kimchi on immune homeostasis, we extended our analysis to B cells, assessing both their compositional and functional aspects. Integrated clustering revealed seven transcriptionally distinct B cell subsets: naïve, activated, cytotoxic-like, plasma-like, resting memory, lineage-ambiguous, and transitional/unknown populations (Supplementary Fig. 4A–C and Supplementary Table 2). The subset proportions remained largely consistent across the groups, with only minor variation observed in the transitional/unknown population (Supplementary Fig. 4D).

Pseudotime analysis further revealed early-state enrichment in naïve and plasma-like B cells following kimchi intake—most prominently in the LMS-K group (p = 1.6 × 10⁻¹⁹ and 5.4 × 10⁻⁹, respectively)—with similar but less pronounced trends in the S-K and placebo groups (Supplementary Fig. 4E). Memory B cells exhibited a modest leftward shift in LMS-K (p = 8.4 × 10⁻⁷), whereas changes in the other groups were minimal or non-significant.

Taken together, these results indicate that kimchi intake preserves B cell subset composition while modestly modulating differentiation dynamics toward earlier or intermediate transcriptional states. This pattern suggests a homeostatic rather than immunostimulatory effect, potentially contributing to balanced immune regulation.

Kimchi intake preserved NK and NKFT cell stability with limited functional changes

To broaden our assessment of the immunomodulatory effects of kimchi, we analyzed NKFT and NK cells (Supplementary Figs. 5 and 6). Marker-based clustering resolved five NKFT (Supplementary Fig. 5A–C) and six NK (Supplementary Fig. 6A–C) subtypes. Subset proportions were generally stable across the groups, with a modest enrichment of memory-like NK cells in the LMS-K group (Supplementary Fig. 6D). Pseudotime analysis indicated limited functional changes, and while some subsets reached statistical significance (p < 0.05), overall trajectories remained stable before and after the intervention.

These findings suggest that, while kimchi intake selectively modulated CD4⁺ T cells and modestly influenced B cells, cytotoxic lymphocyte populations remained stable, reflecting a targeted and restrained immunomodulatory effect.

Discussion

This study demonstrated that kimchi intake induced selective changes in the human immune system at the single-cell level. During the 12-week dietary intervention, kimchi consumption led to transcriptional remodeling primarily in APCs and CD4⁺ T cells, while causing no substantial alterations in overall immune cell composition or systemic immune balance. These findings suggest that kimchi may act as a modulator that targets specific immune pathways rather than as a broad immune stimulant.

Mechanistically, our findings suggest that kimchi intake enhances antigen-presenting capacity by modulating the JAK/STAT1–CIITA axis rather than globally inhibiting it (Fig. 6). LPS induced hyperphosphorylation of JAK1/2 and STAT1, whereas kimchi supernatants attenuated this excessive activation and maintained signaling at levels sufficient to support CIITA and MHC-II expression. This dependency was confirmed by ruxolitinib, which strongly suppressed JAK/STAT1 phosphorylation and abolished CIITA and HLA-DRA/B1 induction. Functionally, kimchi increased antigen uptake and HLA-DR expression, and these effects were reduced but not completely eliminated by ruxolitinib, suggesting that residual activity may involve auxiliary or JAK-independent pathways. Together, these results support a model in which kimchi calibrates JAK/STAT1 signaling to optimize APC function.

Building on this foundation, we performed exploratory pseudotime trajectory analysis to assess the downstream consequences of APC activation. Single-cell data revealed accelerated differentiation of CD4⁺ T cells toward late-stage effector and regulatory subsets, suggesting enhanced T cell priming and immune adaptation. Although functional validation through co-culture or cytokine assays was not performed in this study, these transcriptomic shifts align with the known immunological role of APC-driven MHC-II signaling in orchestrating CD4⁺ T cell differentiation and adaptive immunity^37–42. Importantly, while these adaptive shifts were prominent in the CD4⁺ T cell compartment, the transcriptional landscape and relative abundance of CD8⁺ T cells, B cells, and NK cells remained largely unchanged. This cell-type specificity suggests that kimchi intake does not broadly activate the immune system but instead exerts a targeted effect on antigen presentation and helper T cell function.

To further explore the influence of the fermentation method, we examined S-K with LMS-K. Both types shared common features, including enhanced MHC class II expression and CD4⁺ T cell remodeling, yet LMS-K more strongly suppressed JAK1/2–STAT1 phosphorylation in vitro and induced higher expression of MHC-II–related genes than S-K. These functional differences were less pronounced in vivo, where single-cell transcriptomics revealed broadly similar cell–cell interaction and CD4⁺ T cell trajectory patterns between S-K and LMS-K. This discrepancy indicates that, although fermentation conditions can modulate the immunological properties of kimchi in vitro, such differences may be buffered in vivo by host- and microbiota-dependent factors, thereby highlighting the complexity of diet–microbiota–immune interactions.

Kimchi intake was associated with increased MHC class II expression and accelerated CD4⁺ T cell differentiation. These changes highlight potential immunological benefits, including enhanced antigen-specific priming⁴³, balanced effector polarization⁴⁴, and regulatory T cell expansion⁴⁵ that may support inflammatory control and immune homeostasis. Similar concepts have been emphasized in broader nutritional immunology, where diet-induced modulation of APC–T cell interactions contributes to immune regulation^46,47. In parallel, the upregulation of IFN-γ–responsive genes with attenuation of type I IFN–related signals suggests a selectively tuned immune state that could favor antiviral defense and potentially improve vaccine responsiveness^48,49. This is consistent with global findings that dietary and microbiota cues can modulate vaccine outcomes and antiviral immunity^50,51. Taken together, these observations provide a tentative mechanistic interpretation that links kimchi intake to improved immune regulation.

This study has some limitations. First, the sample size (n = 13) was relatively small and the cohort was demographically homogeneous, which may limit the generalizability of the results. While the paired, within-subject design increased sensitivity to individual immune changes, unmeasured behavioral or environmental factors may still have introduced residual confounding. Second, the conclusions rely primarily on transcriptomic profiling. Systemic immune parameters, such as circulating cytokines, were not assessed, and functional validation was confined to targeted in vitro assays of APC activation. Future investigations should therefore include larger and more diverse cohorts with longitudinal follow-up and incorporate expanded immune assays—such as flow cytometry, multiplex cytokine profiling, and T cell co-culture experiments—to provide a more comprehensive understanding of kimchi’s immunomodulatory effects.

In summary, kimchi intake modulates human immune function by enhancing antigen presentation in APCs and orchestrating CD4⁺ T cell differentiation through precisely regulated immune signaling. These data support the concept of kimchi as a functional dietary immunomodulator with potential relevance for immune health maintenance and the prevention of immune-related disorders.

Methods

Study design and ethical approval

This study is an exploratory post hoc single-cell transcriptomic analysis of PBMC samples obtained from participants in a previously conducted 12-week randomized, double-blind, placebo-controlled clinical trial. The parent trial evaluated the metabolic effects of kimchi powder supplementation in 90 overweight but otherwise healthy adults (BMI 23–30 kg/m²; ClinicalTrials.gov Identifier: NCT05898802)³². The participants were randomly assigned to receive either placebo, S-K, or LMS-K and consumed their assigned intervention at a dose of 3000 mg/day (equivalent to 30 g of fresh kimchi) daily for 12 weeks.

Twenty-two participants provided additional informed consent for immunological assessments. Due to the resource-intensive nature of scRNA-seq analysis, 13 samples (placebo n = 4; S-K n = 5; LMS-K n = 4) were randomly selected from those that met predefined quality thresholds, including PBMC viability ≥95%. After randomization for analytical purposes, no participants were excluded, and no additional PBMC samples were collected outside the scope of the original trial. The study protocol was approved by the Institutional Review Board of Pusan National University Hospital (Approval No. 2210-037-119), and all procedures were conducted in accordance with the ethical principles of the Declaration of Helsinki.

Preparation of S-K and LMS-K capsules

The preparation protocol for S-K and LMS-K capsules followed a previously described method by Lee (2024). In brief, kimchi was produced using salted kimchi cabbage (Brassica rapa subsp. pekinensis) and a standardized seasoning mixture containing red pepper, fish sauce, garlic, ginger, radish, and onion. For LMS-K, the kimchi mix was inoculated with the starter strain, Leuconostoc mesenteroides KCKM0828, at 10⁶ CFU/g, whereas S-K was prepared through spontaneous fermentation without any starter. Fermentation was performed at 6°C for 14 days. The final pH values were 4.1 ± 0.01 for S-K and 3.88 ± 0.18 for LMS-K, confirming successful fermentation. Both variants were freeze-dried, powdered, and formulated into capsules. Each capsule contained 333 mg of either S-K or LMS-K powder, combined with 63 mg of lactose and 4 mg of kimchi flavoring (FK140217), to constitute 400 mg. The placebo capsule matched in weight and appearance and included lactose, caramel coloring (P212), red coloring RR (red), and kimchi flavor (FK140217).

PBMC isolation and preservation

Peripheral blood samples were collected in K₂EDTA BD Vacutainer tubes (#366643; BD) and processed within 2 h of collection. Mononuclear cells were isolated using Ficoll-Paque PLUS (#17-1440-020, Millipore Sigma) with SepMate-50 tubes (#85450, Stemcell Technologies), according to the manufacturer’s protocol. Briefly, blood samples were diluted 1:1 with phosphate-buffered saline (PBS) (#21-040-CV, Corning-Cellgro), carefully layered over Ficoll, and centrifuged at 1200 × g for 10 min. The mononuclear cell layer was harvested, washed twice with PBS, and pelleted by centrifugation at 250 × g for 10 min at room temperature. Cell pellets were resuspended in PBS, and cell counts and viability were assessed using a hemocytometer.

For cryopreservation, cells were first pelleted at 300 × g for 5 min at 4 °C. After separating the supernatant, cell pellets were resuspended in chilled Iscove’s Modified Dulbecco’s Medium (#12440-053, Gibco) supplemented with 10% fetal bovine serum (FBS) (#97068-085, Seradigm) to a concentration of 20 × 10⁶ cells/mL. An equal volume of 2× freezing medium (20% DMSO in FBS; #ICN19141880, Fisher Scientific) was then added to reach a final concentration of 10 × 10⁶ cells/mL. Cell suspensions were aliquoted into pre-cooled cryovials (#368632, Nunc) and placed in a controlled-rate freezing container at −80°C for at least 4 h. The samples were then transferred to liquid nitrogen for long-term storage.

Single-cell library preparation and sequencing

Cryopreserved PBMC samples were thawed in Dulbecco’s Modified Eagle Medium containing 10% FBS at 37°C and washed twice with calcium- and magnesium-free PBS supplemented with 0.04% BSA (w/v). After centrifugation at 300 × g for 5 min at 4°C, the cells were gently resuspended in cold Stain Buffer (#554656, BD Biosciences). Cell viability and concentration were assessed using the LUNA-FX7™ Automated Fluorescence Cell Counter (Logos Biosystems) with acridine orange and propidium iodide (AO/PI) staining (#F23001, Logos Biosystems).

To enable multiplexing, cells were labeled with antibody-conjugated DNA barcodes according to the manufacturer’s protocol. Samples were incubated with multiplexing antibodies for 20 min at room temperature and washed three times with Stain Buffer. After the final wash, the cells were resuspended in cold Sample Buffer (#664887, BD Biosciences), recounted, and pooled.

Single-cell capture and mRNA barcoding were performed using the BD Rhapsody HT Xpress System with an eight-lane cartridge (#666262, BD Biosciences). Following cell lysis, mRNA transcripts were captured on magnetic beads containing unique cell barcodes. First-strand cDNA synthesis and exonuclease treatment were performed out on-bead using the BD Rhapsody cDNA Synthesis Kit (#633773, BD Biosciences).

Whole transcriptome amplification was performed using the BD Rhapsody WTA Amplification Kit (#633801, BD Biosciences), following the manufacturer’s protocol. For gene expression library construction, cDNA was subjected to random priming and extension (RPE), RPE amplification, and index PCR. For sample tag library construction, nested PCR (PCR1 and PCR2) and index PCR were conducted.

The final libraries were quantified using qPCR according to the qPCR Quantification Protocol Guide (KAPA Biosystems) and assessed for quality using the Agilent 4200 TapeStation system (Agilent Technologies). Sequencing was performed on the Illumina NovaSeq X Plus platform (Illumina).

Data preprocessing and quality control

Raw sequencing data (FASTQ format) were processed using the BD Rhapsody WTA Analysis Pipeline (v2.2.1). Reads were aligned to the human reference genome (GRCh38), and digital gene expression matrices were generated for downstream single-cell transcriptomic analysis. Raw single-cell gene expression data were imported using the Read10X function in v4.3.0³⁶, and individual sample-level Seurat objects were created using CreateSeuratObject, applying thresholds of a minimum of 200 detected genes and three cells per feature. Cells originating from 13 PBMC samples (across placebo and kimchi intervention groups) were merged into a single object using the merge function. For quality control, the percentage of mitochondrial transcripts per cell was calculated using PercentageFeatureSet with the pattern “^MT-.” Cells were retained if they expressed 200–6000 genes, and less than 20% mitochondrial content, a common threshold to exclude apoptotic or stressed cells. Cells that did not meet these criteria were filtered out. Gene expression data were log-normalized using Seurat’s NormalizeData function. Highly variable genes were identified using FindVariableFeatures, and the data were scaled using ScaleData before dimensionality reduction. Principal component analysis was performed using the top 20 components, which were used for the neighborhood graph construction (FindNeighbors) and clustering (FindClusters, resolution = 0.5). UMAP was applied to the same 20 components for two-dimensional visualization. After filtering and preprocessing, 88,403 high-quality cells were retained for downstream analyses, including clustering, cell-type annotation, and differential expression analysis. Immune cell types were annotated using the following two complementary approaches: (1) automated reference-based labeling with SingleR using the MonacoImmuneData reference and (2) manual curation based on clustering results and canonical marker expression (FindAllMarkers). Clusters with fewer than 50 cells were excluded from the analysis.

Cell–cell communication analysis

For each intervention group (Placebo, S-K, and LMS-K), PBMC single-cell transcriptomic data collected before and after intervention were processed using the CellChat R package (v 1.6.1) to infer and compare intercellular communication networks^34,35. Seurat objects corresponding to each group and timepoint were generated by subsetting the merged single-cell dataset. Each subset was converted into a CellChat object using cell-type annotations embedded in the metadata. The human ligand–receptor interaction database (CellChatDB.human) was used as the reference for signaling inference. Each CellChat object was subjected to a standardized preprocessing pipeline, including the detection of overexpressed genes and interactions, computation of communication probabilities, and aggregation of pathway-level signaling using default parameters. Subsequently, communication networks were merged within each group to compare pre- and post-intervention states using the mergeCellChat function. Changes in intercellular communication were quantified and visualized by comparing both the number and strength of signaling interactions using the compareInteractions and netVisual_diffInteraction functions, respectively. Circle plots were generated to illustrate global signaling networks among immune cell subsets. For targeted analysis of APCs, including monocytes, DCs, and innate-like B cells, scatter plots of differential incoming and outgoing signals were generated using the netAnalysis_signalingChanges_scatter function.

Pseudotime trajectory and functional state analysis of immune cell subsets

Pseudotime analysis was conducted to explore the transcriptional dynamics of immune cell subsets following kimchi intake. For each functional subtype, cells were classified based on module scores derived from curated gene sets (Supplementary Table 3), and functional states were assigned using Gaussian Mixture Modeling via the mclust package⁵². Pseudotime trajectories were inferred using the Slingshot⁵³ algorithm, based on low-dimensional embeddings from UMAP and subtype-based clustering. To compare overall distributional changes in pseudotime values before and after intervention within each group, the two-sample Kolmogorov–Smirnov test was applied. P-values were directly annotated on density plots, with non-significant results labeled as “n.s.” All pseudotime-related analyses and visualizations were conducted in R (v 4.4.2), using the Seurat (v 4.3.0), ggplot2 (v 3.5.2), dplyr (v 1.1.4), slingshot (v 2.14.0), mclust (v 6.1.1), and patchwork (v 1.3.0) packages.

Pathway–gene correlation analysis in DCs

A correlation analysis was performed between the expression of MHC class II gene expression and immune pathway activity scores in DCs using scRNA-seq data. DCs were extracted from PBMC-derived Seurat objects and normalized using the NormalizeData function. Immune pathway module scores were calculated using AddModuleScore, based on gene sets curated from the Molecular Signatures Database (MSigDB, Hallmark category) and relevant literature⁵⁴. The following immune-related hallmark pathways were included in the analysis: IFN-α/γ response, complement system, TNF-α, IL6/JAK/STAT3, inflammatory response, and TGF-β signaling. HLA-DRA and HLA-DRB1 were selected as canonical markers of MHC class II expression. For each group (placebo, S-K, LMS-K), the Spearman correlation coefficient (ρ) was calculated between gene expression and immune pathway scores. P- values were adjusted using the Benjamini–Hochberg method. Associations were considered significant if ρ > 0.3 or ρ < −0.3 with adjusted p < 0.05. All analyses were conducted in R, and visualizations were generated using the ggplot2 package (v 3.5.2). For group-wise comparisons, DCs from each intervention group were analyzed separately, and correlations were computed within each group independently. Heatmaps were used to display the results, with significance levels annotated as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.

Cell viability measurement

THP-1 cells were obtained from the Korean Cell Line Bank (Seoul, Korea) and cultured in Roswell Park Memorial Institute Medium 1640 supplemented with 10% FBS and 1% penicillin/streptomycin. The cells were differentiated into macrophage-like cells by treatment with 100 nM PMA for 48 h. For viability assessment, PMA-differentiated THP-1 cells (3 × 10⁵ cells/200 µL per well) were treated for 16 h with LPS (500 ng/mL) alone or in combination with various concentrations of kimchi supernatants (S-K or LMS-K; 0.01–0.25 mg/mL). After treatment, 20 µL of CellTiter 96® AQueous One Solution Reagent (Promega, Madison, WI, USA) was added to each well and incubated at 37°C for an additional 2 h. Cell viability was determined by measuring the absorbance at 490 nm using a microplate reader (Tecan, Zurich, Switzerland).

Antigen uptake and HLA-DR expression assay

To assess antigen uptake and processing, THP-1 monocytes were differentiated with PMA (100 ng/mL, 48 h) and subsequently treated with kimchi supernatants (S-K or LMS-K, 0.25 mg/mL) or vehicle control for 24 h. Where indicated, the cells were pretreated with the JAK1/2 inhibitor ruxolitinib (1 μM; #S1378, Selleckchem, Houston, TX, USA) for 1 h prior to exposure to the kimchi supernatant. The cells were then incubated with DQ™-ovalbumin (10 μg/mL; #D12053, Thermo Fisher Scientific, Waltham, MA, USA) at 37°C for 1 h to allow uptake and proteolytic processing. After incubation, the cells were washed with ice-cold PBS, harvested, and stained with anti–HLA-DR (clone LN3) Monoclonal antibody, PE (#12-9956-42, Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at 4°C. Following staining, cells were washed with PBS to remove unbound antibody. Fluorescence signals from processed DQ™-ovalbumin (FITC channel) and HLA-DR surface expression (PE channel) were acquired using a BD FACSCanto™ II flow cytometer (BD Biosciences, San Jose, CA, USA) and analyzed with FlowJo software (BD Biosciences, Ashland, OR, USA). Dead cells were excluded by LIVE/DEAD™ Fixable Blue Dead Cell Stain, PB450 (#L34962, Invitrogen, Carlsbad, CA, USA). For visualization (histograms), at least 10,000 viable events per sample were used. For statistical comparisons, 250 events were randomly subsampled from each sample to ensure equal representation across groups.

qPCR array profiling of interferon-related genes

PMA-differentiated THP-1 cells were stimulated with LPS (500 ng/mL) in the absence or presence of 0.25 mg/mL kimchi supernatant (S-K or LMS-K) for 16 h. Total RNA was extracted using the Direct-zol RNA MiniPrep Plus Kit (#R2072, Zymo Research, Irvine, CA, USA), and cDNA was synthesized using the QuantiTect Reverse Transcription Kit (#205311, Qiagen, Hilden, Germany). Interferon-related gene expression was profiled using a targeted qPCR array (AccuTarget™ Human qPCR Screening Kit, Interferons and Receptors, #SH-000-20, Bioneer, Korea). The fold changes were calculated using the 2^−ΔCt method. The design of the target gene panel is presented in Supplementary Table 4. Scatter plots were generated to visualize gene expression relative to the LPS-only control, with red and blue dashed lines indicating ±two-fold thresholds.

Western blot analysis

For western blot analysis, PMA-differentiated THP-1 cells were stimulated with LPS (500 ng/mL) in the presence or absence of 0.25 mg/mL kimchi supernatant (S-K or LMS-K). For inhibitor experiments, the cells were pre-treated for 1 h with the JAK1/2 inhibitor ruxolitinib (1 μM; #S1378, Selleckchem, Houston, TX, USA) before LPS and supernatant exposure. Following stimulation, the cells were lysed in RIPA buffer (#89901, Thermo Fisher Scientific, USA) supplemented with protease and phosphatase inhibitors (#78440, Thermo Fisher Scientific, USA). Protein concentrations were determined using a BCA assay (#23225, Thermo Fisher Scientific, USA), and equal amounts of protein (20 μg) were resolved using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes (#IPVH08100, Millipore).

The membranes were blocked with 5% skim milk in PBS-T (20 mM Tris-HCl, 150 mM NaCl, 0.1% Tween-20, pH 7.6) for 1 h at room temperature and incubated overnight at 4°C with primary antibodies against JAK1 (#3344), phospho-JAK1 (Tyr1034/1035, #74129), JAK2 (#3230), phospho-JAK2 (Tyr1007/1008, #3776), STAT1 (#14994), phospho-STAT1 (Tyr701, #7649), and phospho-STAT1 (Ser727, #8826) (all from Cell Signaling Technology). After washing, membranes were incubated with HRP-conjugated anti-rabbit IgG secondary antibody (#7074, Cell Signaling Technology) for 1 h at room temperature. Protein bands were visualized using enhanced chemiluminescence (#sc-2048, Santa Cruz Biotechnology, Dallas, TX, USA) and imaged with a ChemiDoc imaging system (Amersham™ Imager 600, GE Healthcare, Buckinghamshire, UK). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (#sc-32233, Santa Cruz Biotechnology, USA) was used as a loading control. Densitometric analysis was performed in ImageJ (National Institutes of Health) and normalized to total JAK1, JAK2, or STAT1.

Quantitative real-time polymerase chain reaction (qRT-PCR) validation of MHC class II gene expression

qRT-PCR was performed to measure the expression levels of HLA-DRA, HLA-DRB1, and CIITA in PMA-differentiated THP-1 cells treated as described above. When required, cells were pre-treated for 1 h with ruxolitinib (1 μM; Cat. No. S1378, Selleckchem, Houston, TX, USA) before kimchi supernatant treatment. Total RNA was extracted and reverse-transcribed as described previously. qRT-PCR was conducted using the SYBR Green PCR Master Mix (Enzynomics, Cat. No. RT500M, Daejeon, Korea) on a CFX96 Real-Time PCR Detection System (Bio-Rad, USA). The primer sequences used are listed in Supplementary Table 5. Gene expression levels were normalized to GAPDH levels, and relative fold changes were calculated using the ΔΔCt method. Data represent the mean ± SEM from three independent experiments (n = 3). Statistical significance was determined using the two-sided Student’s t test.

Statistical analyses

Baseline characteristics were compared using one-way ANOVA for continuous variables and χ² tests as an approximation to the Fisher–Freeman–Halton exact test for categorical variables, given the limited sample size. ANOVA assumes independence, normality, and homogeneity of variances, while χ² tests assume independence and adequate expected counts. All baseline analyses were conducted in IBM SPSS Statistics v30.0.0.0. Correlations between gene expression and pathway scores were assessed using Spearman’s rank correlation with Benjamini–Hochberg adjustment to control the false discovery rate. Statistical significance was defined as adjusted p < 0.05 for multiple testing. Flow cytometry data were summarized as MFI and compared using the two-sided Mann–Whitney U test, a non-parametric test that does not assume normality. qRT-PCR gene expression differences were evaluated using the two-sided Student’s t test, assuming independence, normality, and equal variances. Within-group comparisons of immune cell proportions were tested with the paired Wilcoxon signed-rank test, which assumes paired observations and symmetric distributions of differences. Pseudotime distributions were compared using the two-sample Kolmogorov–Smirnov test, which assumes independent samples from continuous distributions.

Author contributions

W.L. conceived and designed the study, conducted formal analysis, and interpreted the data. H.M. played a central role in conducting experiments and generating key datasets. H.J.K., Y.R.Y., and M.S.K. contributed to study design and coordinated data curation. H.C. and H.J.L. assisted in experimental work. W.L. and S.W.H. jointly drafted the manuscript and shared primary responsibility for the final content. All authors read and approved the final manuscript.

Data availability

Single-cell RNA sequencing data described in this manuscript are publicly available in the Gene Expression Omnibus (GEO) repository under accession number GSE301131.

Code availability

The code used for data analysis is available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-025-00593-7.