Natural killer cells as biomarkers for disease activity, lupus nephritis, and time to remission in treatment-naïve childhood-onset systemic lupus erythematosus: a cohort study

Abstract

Objectives

Adaptive immunity is well-established in lupus pathogenesis, yet the clinical significance of innate immune components, particularly natural killer (NK) cells, remains underexplored. This study aims to evaluate the clinical relevance of NK cell levels in treatment-naïve, infection-free patients with childhood-onset systemic lupus erythematosus (cSLE).

Methods

We conducted a retrospective cohort study among 96 treatment-naïve, infection-free cSLE patients diagnosed during 2015–2025. Logistic regression was applied to compare patients of moderate-to-severe disease activity (SLEDAI > 9) versus low disease activity (SLEDAI ≤ 9), and to assess the associations between NK cell levels and organ involvement. Cox regression analysis was performed to evaluate the effect of baseline NK cell counts on the time to first remission. To evaluate whether longitudinal changes in absolute NK cell counts are associated with the timing of initial remission, we performed a linear mixed-effects model analysis in the subset of patients with serial NK cell data.

Results

Both absolute NK cell counts and percentages were lower than healthy references. Anti-dsDNA positivity (OR = 5.107, 95% CI: 1.257–20.757, P = 0.023), neuropsychiatric involvement (OR = 33.306, 95% CI: 6.79–163.32, P < 0.001), and absolute NK cell count (OR = 0.982, 95% CI: 0.969–0.995, P = 0.007) were independent factors associated with higher disease activity. Similarly, independent factors for lupus nephritis included anti-dsDNA positivity (OR = 8.555, 95% CI: 2.302–31.791, P = 0.001) and absolute NK cell count (OR = 0.991, 95% CI: 0.982–1.000, P = 0.041). During follow-up, in multivariate Cox regression with forward likelihood ratio, higher absolute NK cell count was independently associated with a shorter time to first remission (HR 1.005, 95% CI 1.001–1.009, P = 0.015), whereas hypocomplementemia was independently associated with a longer time to first remission (HR 0.472, 95% CI 0.244–0.912, P = 0.026). The optimal cut-off was 126 cells/µL.

Conclusions

Absolute NK cell count is a clinically relevant biomarker in cSLE. Compared with healthy controls, cSLE exhibit consistently reduced NK cell count and percentage across all age groups. Absolute NK cell count independently stratifies patients with cSLE by disease activity, clinical diagnosis of lupus nephritis, and time to first remission.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12969-026-01204-9.

Introduction

Systemic lupus erythematosus (SLE) is a complex chronic autoimmune disease. Childhood-onset SLE (cSLE) follows a more aggressive course than adult-onset SLE, characterized by heightened disease activity, increased risks of nephritis and neuropsychiatric disease, and elevated mortality [1, 2]. The disease is driven by a constellation of immunologic aberrations, foremost among them a loss of self-tolerance, the expansion of autoreactive T and B cell clones, and polyclonal B cell activation leading to hypergammaglobulinemia [3]. These events precipitate a cascade of autoantibody production, immune complex deposition, and systemic inflammation. Despite progress in elucidating the adaptive immune mechanisms of tissue injury in SLE, the roles of innate immune cells, including natural killer (NK) cells, remains a significant knowledge gap.

NK cells, constituent members of the innate lymphoid cell family, are lymphocytes that play a critical role in the initial immune response to infection and neoplastic transformation [4]. They constitute a heterogeneous population with a broad spectrum of functional attributes, including varied proliferative potential, homing characteristics, functional capacities, and cytokine responsiveness. Current definition of human NK cells is based on the surface expression of CD56 and CD16 [5], categorizing into two major subsets: CD56^brightCD16– and CD56^dimCD16⁺ [6].

Current research on NK cells and cSLE remains relatively limited. Adult studies have reported reductions in NK cell counts or percentages in SLE patients [7–9], and NK cell levels may be associated with lupus activity and nephritis in children in cross-sectional analyses [10–12]. However, these observations are often confounded by infections commonly complicated in SLE (such as CMV [13] or EBV [14]), as well as immunosuppressive treatments, leaving intrinsic effects of SLE on NK cells or vice versa unexplored. Additionally, there is a lack of longitudinal analysis on the prognostic value of baseline NK cells and disease remission in cSLE.

To address these critical gaps, we established a longitudinal cSLE cohort, exclusively enrolling patients who were infection-free and had not yet started immunosuppressive therapy at diagnosis. This study aimed to determine whether NK cell counts influence disease activity, organ involvement, and time to first remission. This study is intended to serve as a starting point to contribute to future efforts in using NK cell metrics for cSLE in both research and clinical contexts.

Methods

Patients

This retrospective cohort study included children diagnosed with SLE at Peking Union Medical College Hospital between January 2015 and May 2025. Inclusion criteria: (1) Age of diagnosis between 6 and 18 years; (2) Patients met the 1997 [15] American College of Rheumatology (ACR) or 2019 [16] European League Against Rheumatism (EULAR)/ACR SLE classification criteria. Exclusion criteria: (1) infection, including EBV (defined as abnormalities in EBV VCA-IgM or EBV-DNA), CMV (abnormalities in CMV-IgM, CMV-DNA, or CMV-PP65), or other pathogens; (2) monogenic lupus; (3) those who have been treated with glucocorticoids, immunosuppressants or biologics medications; (4) incomplete data.

Data collection

We collected baseline demographic, clinical, and laboratory data from the electronic health record system. The extracted variables included sex, age at initial diagnosis, visit date, organ-specific manifestations (including cutaneous, mucosal, renal, neuropsychiatric, and hematologic articular, serosal), SLEDAI-2 K score, anti-dsDNA antibody status, medication usage including glucocorticoids, immunosuppressants, or biologics. NK cells were determined from clinical flow cytometry assay as CD3^–CD16⁺ or CD3^–CD56⁺.

Virological screening for EBV and CMV was performed on the same day as NK cell measurement, both at initial diagnosis and during follow-up. Only samples confirmed negative for both viruses were included in the analysis to rule out confounding by active viral infection.

Flow cytometry analysis of NK cells

Peripheral blood samples were stained with the following monoclonal antibodies: CD45-FITC, CD3-APC-A750, CD56-ECD, and CD16-ECD. Flow cytometry data were acquired on a Beckman Coulter NAVIOS flow cytometer and analyzed using Kaluza Analysis Software. NK cells were identified using a sequential gating approach. First, lymphocytes were gated based on CD45 expression and side scatter (SSC). Within the CD45 + lymphocyte gate, CD3- cells were selected. NK cells were then defined as CD3- cells that expressed either CD56 or CD16. This definition captures both CD56 + NK subsets as well as CD16 + NK cells. The absolute NK cell count was derived by multiplying the percentage of NK cells (within the lymphocyte gate) by the absolute lymphocyte count obtained from a concurrent complete blood count (CBC) performed on the same day. Absolute counts are expressed as cells/µL.

Definitions and outcomes

SLE remission was defined according to the 2021 Definition of Remission in SLE (DORIS): [17] Clinical SLE disease activity index (SLEDAI-2 K) = 0 and Physician Global Assessment (PGA) < 0.5 (0–3), irrespective of serology; The patient may be on antimalarials, low-dose glucocorticoids (prednisolone ≤ 5 mg/day), and/or stable immunosuppressives including biologics.

Lupus nephritis was clinically diagnosed in accordance with the Chinese guideline on diagnosis and treatment of lupus nephritis [18]. Hypocomplementemia was defined as C3 or C4 below the laboratory reference range. The time to first remission was defined as the time from the initial diagnosis to the first documented remission after regular treatment. The follow-up time was calculated from the initial diagnosis.

Statistical analysis

Continuous variables were summarized as median with interquartile range (IQR) and compared using the Mann-Whitney U test. The association between baseline NK cell counts and SLE disease activity was assessed using logistic regression. The model’s discriminatory ability was evaluated with receiver operating characteristic (ROC) analysis, and the optimal threshold was determined by maximizing the Youden’s index. For the analysis of time to remission, a Cox proportional-hazards model with stepwise variable selection was used to assess the independent impact of NK cell counts on the time to first remission in cSLE. Patients were stratified by an NK cell count cut-off that maximized discrimination in Kaplan-Meier curves. Sensitivity analysis was conducted using different endpoints (Supplementary data). To evaluate whether longitudinal changes in absolute NK cell counts are associated with the timing of initial remission, we performed a linear mixed-effects model analysis in the subset of patients with serial NK cell data. All analyses were performed using R software (version 4.4.3). P values below 0.05 were considered statistically significant.

Results

Demographics and clinical characteristics of cSLE

A total of 96 newly diagnosed children with cSLE free of infection were enrolled in this study at Peking Union Medical College Hospital based on the inclusion and exclusion criteria. Of the 96 cases, 84 (87.5%) cases were girls. The girl to boy ratio was 7:1. The study cohort had a median age of 13.5 years (IQR, 10.8–15.1). The demographics and clinical characteristics of cSLE were shown in Table 1.

Table 1.

Demographics and clinical characteristics of cSLE (N = 96)

Variables	No. of cases	Percentage (%)
Demographic
Sex (girl)	84	87.5
Organ involvement
Mucocutaneous	27	28.1
Renal	34	35.4
Hematological	74	77.1
Pulmonary	22	22.9
Neuropsychiatric	31	32.3
Articular	21	21.9
Pericardial/pleural	18	18.8
Laboratory
Leukopenia (WBC < 3 × 109/L)	22	22.9
Thrombocytopenia (PLT < 100 × 109/L)	22	22.9
Proteinuria (24-hour urine protein > 0.5 g/24 h)	22	22.9
Anti-dsDNA positivity	62	64.6
Anti-Smith antibody positivity	29	30.2
Hypocomplementemia (C3 or C4)	72	75.0
Disease activity (SLEDAI-2 K)
0–4	25	26.0
5–9	34	35.4
10–15	19	19.8
>15	18	18.8

cSLE, childhood-onset systemic lupus erythematosus; WBC, white blood cell; PLT, platelet; SLEDAI-2 K, systemic lupus erythematosus disease activity index 2000

Peripheral blood NK cell counts and percentages in cSLE were below healthy references

We compared NK cell parameters among cSLE and healthy Chinese children [19]. The median absolute NK cell counts with 10–90th percentile ranges in cSLE patients were: 135 (98–257) cells/µL for 6–8 years, 71 (50–150) cells/µL for 8–12 years, and 47 (28–76) cells/µL for 12–18 years. Similarly, median NK cell percentages with 10–90th percentile ranges were: 6.2% (3.2–10.1) in 6–8 years, 4.3% (3.4–8.1) in 8–12 years, and 4.7% (2.6–7.7) in 12–18 years. A total of 96 cases were analyzed. A low NK cell percentage (< 10th percentile) was observed in 78 cases (81.2%), and a low absolute NK cell count (< 10th percentile) was found in 91 cases (94.8%) (Fig. 1).

Fig. 1.

NK cell levels in cSLE versus healthy children. Healthy references are presented with 10th and 90th percentiles. The dots represent the absolute count (A) and the percentage (B) of NK cells in peripheral blood

Associations between NK cell count and cSLE disease activity

Univariate analysis stratified by disease activity revealed impaired NK cell parameters in cSLE with moderate-to-severe disease activity (SLEDAI > 9) versus cSLE with mild activity (SLEDAI ≤ 9). The moderate-to-severe group exhibited markedly reduced absolute NK cell counts (median [IQR] 44 [26–65] vs. 66 [46–127] cells/µL; U = 659.5, P = 0.001) and lower NK cell percentages (median: 3.2% [1.9–6.7] vs. 5.7% [3.8–8.1]; U = 741.5, P = 0.009), as shown in Fig. 2A–B. In univariate logistic regression, lower absolute NK cell count was associated with higher disease activity (OR = 0.987, 95% CI: 0.977–0.996; P = 0.006). ROC curve analysis revealed a modest discriminatory ability (AUC = 0.698, 95% CI: 0.590–0.805; P < 0.001) at the optimal cut-off value of 44.5 cells/µL determined by the maximum Youden index (0.303), with 54.1% sensitivity and 76.3% specificity (Fig. 2E). NK cell percentages did not demonstrate statistical significance (OR = 0.959, 95% CI: 0.875–1.049; P = 0.36). Other variables associated with disease activity (moderate-severe vs. mild) in cSLE included sex, age, proteinuria, leukopenia, thrombocytopenia, anti-dsDNA positivity, low complement levels, cutaneous manifestations, lupus arthritis, neuropsychiatric involvement, and serosal effusions. In multivariate logistic regression using forward stepwise likelihood ratio selection, the model containing absolute NK cell count was statistically significant (McFadden R² = 0.42, P < 0.001), with a sensitivity of 81.1% and specificity of 83.1%. Independent factors included anti-dsDNA positivity (OR = 5.107, 95% CI: 1.257–20.757, P = 0.023), neuropsychiatric (OR = 33.306, 95% CI: 6.79–163.32, P < 0.001), and absolute NK cell count (OR = 0.982, 95% CI: 0.969–0.995, P = 0.007). When absolute NK cell count was excluded, the model remained significant (McFadden R² = 0.31, P < 0.001) but showed reduced sensitivity (59.5%) and increased specificity (96.6%), with anti-dsDNA positivity (OR = 6.622, 95% CI: 1.857–23.609, P = 0.004) and neuropsychiatric (OR = 14.369, 95% CI: 4.581–45.067, P < 0.001) as significant factors. The model containing absolute NK cell count (AUC = 0.884) displayed higher discriminative ability than that excluded absolute NK cell count (AUC = 0.825), as shown in Fig. 2F. The difference in AUC between the two models was statistically significant (Z = 2.53, P = 0.011), underscoring the significant value of absolute NK cell count in influencing moderate-to-severe disease activity in cSLE.

Fig. 2.

NK cell levels as biomarkers for cSLE activity and renal involvement. (A) Absolute NK cell count stratified by SLEDAI score. (B) NK cell percentage stratified by SLEDAI score. (C) Comparison of absolute NK cell count between LN and non-LN groups. (D) Comparison of NK cell percentage between LN and non-LN groups. (E) Univariate analysis of absolute NK cell count for discriminating moderate-to-severe from mild cSLE. (F) Stepwise logistic regression model incorporating NK cell count shows improved discriminative ability for moderate-to-severe cSLE compared with the model without NK cell count. (G) Univariate analysis of absolute NK cell count for identifying LN. (H) The predictive model for LN was improved by the inclusion of absolute NK cell count

Association between NK cell count and lupus nephritis in cSLE

Absolute NK cell counts and percentages were compared across groups stratified by organ involvement status. LN was associated with lower absolute NK cell counts versus non-LN patients (median 36.5 vs. 72.5 cells/µL, U = 448.5, P < 0.001), as well as lower NK cell percentages (median 3.2% vs. 5.7%; U = 697, P = 0.006) (Fig. 2C–D). However, no significant differences in absolute NK cell counts or percentages were observed for other organ involvement (Table 2). In univariate logistic regression analysis, lower absolute NK cell count was associated with a higher likelihood of developing LN (OR = 0.986, 95% CI: 0.976–0.996; P = 0.007). ROC curve analysis revealed a modest discriminatory ability (AUC = 0.787, 95% CI: 0.691–0.884; P < 0.001) at the optimal cut-off value of 49.5 cells/µL determined by the maximum Youden index (0.552), with 79.4% sensitivity and 75.8% specificity (Fig. 2G). The NK cell percentages did not demonstrate statistical significance (OR = 0.950, 95% CI: 0.862–1.047; P = 0.298). Other variables associated with renal involvement included age, anti-dsDNA positivity, hypocomplementemia, and absolute NK cell count (Supplementary Table 1). In multivariate logistic regression using forward stepwise likelihood ratio selection, the model containing absolute NK cell count was statistically significant (McFadden R² = 0.197, P < 0.001), with a sensitivity of 76.5% and specificity of 83.9%. The independent factors included anti-dsDNA positivity (OR = 8.555, 95% CI: 2.302–31.791, P = 0.001) and absolute NK cell count (OR = 0.991, 95% CI: 0.982–1.000, P = 0.041). When absolute NK cell count was excluded, the model remained significant (McFadden R² = 0.185, P < 0.001), but showed increased sensitivity (88.2%) and decreased specificity (64.4%), with anti-dsDNA positivity (OR = 8.421, 95% CI: 2.209–32.107, P = 0.002), age (OR = 1.215, 95% CI: 1.027–1.437, P = 0.02)as the significant influencing factors. The model containing absolute NK cell count showed higher discriminative ability (AUC = 0.817) than that excluded absolute NK cell count (AUC = 0.755), as shown in Fig. 2H. The difference in AUC between the two models was not statistically significant (Z = 1.349, P = 0.177).

Table 2.

Comparision of absolute NK cell count and percentage by organ involvement

Organ	Absolute NK cell count (cells/µL), median (IQR)	P value	NK cell percentages (%), median (IQR)	P value
Mucocutaneous
with	54 (44, 96)	0.928	4.8 (3.5, 7.15)	0.461
without	55 (30, 114)		4.3 (2.4, 8)
Renal
with	36.5 (23, 49)	< 0.001	3.2 (1.7, 6.1)	0.006
without	72.5 (50, 129)		5.7 (3.7, 8.2)
Neuropsychiatric
with	48 (27.5, 114.5)	0.743	4.6 (3, 9.7)	0.476
without	56 (39, 98)		4.7 (2.9, 7.2)
Articular
with	65 (28, 107)	0.546	5.7 (3, 7.2)	0.954
without	53 (36.5, 103)		4.6 (2.95, 8.05)
Pulmonary
with	60 (37, 114)	0.373	4.45 (3.4, 8.2)	0.403
without	52 (33, 114)		4.65 (2.6, 7.7)
Pericardial
with	44.5 (26, 98)	0.274	3.4 (3, 7.2)	0.321
without	54.5 (36, 107)		4.75 (2.9, 8.1)
Pleural
with	52 (31.5, 90.5)	0.764	4.15 (3.3, 7.55)	0.978
without	54 (34, 104.5)		4.65 (2.9, 7.95)
Hematological
with	52 (32, 98)	0.459	4.6 (2.9, 7.9)	0.641
without	59 (41, 125)		5.85 (3, 8.2)

We performed univariate analyses for anti-dsDNA positivity and hypocomplementemia and constructed ROC curves to evaluate their predictive performance. The discriminative ability of absolute NK cell counts was comparable to that of anti-dsDNA positivity and hypocomplementemia (Supplementary Fig. 3).

Association of baseline NK cell count with time to remission in cSLE

We next evaluated the predictive value of baseline NK cell count in longitudinal cSLE outcomes. The longitudinal analysis included 92 participants (95.8% of the original cohort) after excluding four patients who were lost to follow-up or lacked treatment response data. The median follow-up time was 15 months. Therefore, we used time to remission during the 15-month follow up as the primary endpoint for survival analysis. Among the 92 patients with follow-up data, treatment regimens included glucocorticoid pulse therapy (32/92, 34.8%), cyclophosphamide pulse therapy (27/92, 29.3%), belimumab (4/92, 4.3%), mycophenolate mofetil (23/92, 25.0%), hydroxychloroquine (54/92, 58.7%), ciclosporin (3/92, 3.3%), methotrexate (6/92, 6.5%), rituximab (3/92, 3.3%), and tacrolimus (3/92, 3.3%). We further compared baseline SLEDAI scores and time to first remission between patients who received specific treatments and those who did not. As shown in Supplementary Table 2, patients who received glucocorticoid pulse therapy or cyclophosphamide pulse therapy had significantly higher baseline SLEDAI scores than those who did not, indicating more severe disease activity. However, no significant difference in time to first remission was observed between treatment groups. In addition, univariate Cox regression analysis was performed for each treatment variable using time to remission within the 15-month follow-up period as the endpoint. As summarized in Supplementary Table 3, none of the treatment variables showed a statistically significant association with remission time in univariate analysis.

In univariate Cox regression, younger age and higher absolute NK cell count were associated with a shorter time to first remission, whereas hypocomplementemia was associated with a longer time to first remission (Supplementary Table 3). However, we considered treatment intensity to be a strong potential confounding factor. Therefore, we also included the receipt of cyclophosphamide pulse therapy or glucocorticoid pulse therapy as covariates in the stepwise regression model. In multivariate Cox regression with forward likelihood ratio, higher absolute NK cell count was independently associated with a shorter time to first remission (HR 1.005, 95% CI 1.001–1.009, P = 0.015), whereas hypocomplementemia was independently associated with a longer time to first remission (HR 0.472, 95% CI 0.244–0.912, P = 0.026). The final model exhibited excellent fit (χ² = 12.113, P = 0.002). To support clinical application, patients were stratified into low (NK ≤ 126 cells/µL, N = 76) and high (NK > 126 cells/µL, N = 16) NK groups according to the optimal cut-off value from the maximally selected rank statistic. Kaplan-Meier analysis showed significantly longer median time to remission in the low-NK group (14.2 months) compared to the high-NK group (9.7 months, log-rank P = 0.0031) (Fig. 3). Using Cox regression, we found that patients in the high-NK group had a shorter time to first remission (HR = 2.853, 95% CI: 1.379–5.904, P = 0.005), verifying the significance of baseline NK in cSLE prognosis evaluation. To assess the robustness of our findings, we performed sensitivity analyses at 12 and 18 months, which confirmed the primary findings (Supplementary Fig. 2).

Fig. 3.

Kaplan-Meier analysis of time to remission stratified by absolute NK cell count. The optimal cut-off value was identified from the maximally selected rank statistic

To assess whether longitudinal NK cell measurements correlate with therapy response, we performed linear mixed‑effects modeling in patients with serial NK cell data collected during follow‑up. The model included time, baseline NK value, and remission status (remission achieved within 12, 15, or 18 months) as fixed effects. Taking the 15-month endpoint as an example: NK cell counts showed a significant recovery over time (Beta coefficient [95% CI]: 0.49 [0.24–0.75], P < 0.001). The longitudinal NK values were positively associated with baseline NK levels (Beta: 5.55 [0.20–10.90], P = 0.048). However, the rate of NK cell recovery did not significantly differ between patients who achieved remission and those who did not (Supplementary Table 4).

Discussion

SLE is a chronic, progressive autoimmune disease with marked heterogeneity. Despite knowledge of adaptive immune cells in disease pathogenesis, the clinical values of innate lymphoid cells, like NK cells, have been rarely investigated. Existing studies were either confounded by infections or immunosuppressive treatments. By stringently enrolling newly diagnosed, treatment-naïve, infection-free cSLE patients this study suggests that absolute NK cell counts are clinically significant in stratifying patients with cSLE by disease activity (high vs. low), clinical diagnosis of LN, and time to first remission.

Based on the established reference ranges for peripheral blood lymphocyte subsets in healthy Chinese children [19], the counts and percentages of NK cells in cSLE patients were consistently below the normal range across all pediatric age groups. This finding aligns with the report from pediatric population [20]. Thangjam et al. reported a marked reduction in the percentage of CD16⁺ cells in immunosuppressant-treated patients, dropping to two-fifths of the level in controls [21]. Both studies were confounded by immunosuppressive treatments. In our study, notably reduced counts and percentages of NK cells were observed in treatment-naïve patients with new-onset cSLE, suggesting an intrinsic disease mechanism rather than a therapy-related effect. The decline, particularly in the mature CD56^dimCD16⁺ subset, could indicate that these highly cytotoxic cells are migrating from the circulation to target organs, thereby exacerbating local tissue damage [10].

This study identified absolute NK cell count as an independent biomarker associated with cSLE disease activity. The SLEDAI-2 K score, while widely used, relies heavily on clinician-reported and patient-reported variables such as rash, alopecia, and lupus headache, which are inherently subjective and prone to inter-observer variability. In contrast, absolute NK cell counts represent a quantifiable and objective immunological parameter. We demonstrated that the model incorporating NK cell counts showed higher discriminative ability than that without NK cell counts, supporting its clinical application as a standardized, reproducible biomarker. Furthermore, in the significance of NK cell subpopulations, Li et al. demonstrated an increased frequency of immature NK cells (CD56^brightCD16^–) and a concomitant decrease in mature, differentiated NK cells (CD57⁺CD56^dimCD16^–) in high-activity SLE, resulting in a significantly higher immature/mature NK ratio compared with that among patients with low-activity disease [22]. However, we could not ascertain a mechanism correlation between NK cells and disease activity, which warrants further experimental study.

This study observed a reduction in absolute NK cell counts in children with LN compared to non-LN. Moreover, the absolute NK cell count was established as an independent factor associated with LN. This finding is consistent with previous reports suggesting a correlation between NK cells and LN [12]. The protective role of NK cells in autoimmunity has been related to its downregulation of autoreactive adaptive immune responses [23]. In this way, NK cell numeric and functional defects have been observed in patients with autoimmune diseases [9, 23]. A study extends to NK cells an altered profile of TRAIL, Bcl-2, TNFR1, Fas, FasL, Bax, Bim, and caspase-3 proteins in patients with jSLE, particularly in nephritis [12]. This change in apoptosis-related protein expressions may contribute to the defective functions of NK cells and, consequently, to lupus development. The full clarification of the role of NK cells in jSLE pathogenesis may pave the way for new therapies like those of NK cell-based.

One important consideration when interpreting our findings is the scale of measurement for absolute NK cell counts. Given that these counts typically range from tens to several hundred cells/µL, the odds ratio (OR) or hazard ratio (HR) estimated per one-unit increase (i.e., per 1 cell/µL) is mathematically expected to be infinitesimally close to 1.0. This statistical artifact reflects the unit of analysis rather than a lack of clinical relevance. Indeed, a decrease of even a few dozen cells/µL—a modest absolute change—may correspond to a substantial shift in risk, which would be masked if one focuses solely on the per-unit effect size. Crucially, when evaluated at a clinically meaningful scale, the predictive performance of baseline NK cell count for both disease activity and lupus nephritis was comparable to that of established serological biomarkers, including anti-dsDNA positivity and hypocomplementemia. These findings suggest that baseline NK cell count may offer additional value in risk stratification. By capturing a dimension of immune dysregulation distinct from autoantibody production or complement consumption, it could potentially complement conventional serological markers in identifying pediatric patients at risk for active disease or LN.

A body of evidence from the literature has implicated NK cells in various autoimmune processes, including rheumatoid arthritis [24] and even the partial remission of type 1 diabetes in children [25], highlighting their potential role in disease remission. To our knowledge, the present study is the first to establish the role of NK cell in the remission in SLE patients. This study demonstrated absolute NK cell count as an independent factor of the time to first remission in cSLE. Additionally, we identified an optimal prognostic cut-off of 126 cells/µL using the maximally selected rank statistic: patients with NK cell counts at or below this threshold experienced a significantly longer time to first remission compared to those above it, highlighting its value in stratifying time to first remission in cSLE. While baseline NK cell proportion is independently associated with time to first remission and serves as a robust stratified indicator at diagnosis, the trajectory of NK cell recovery during follow-up does not appear to reflect therapeutic outcome in our cohort.

Our study design incorporated specific measures to mitigate potential sources of bias. Selection and confounding biases were rigorously addressed through our stringent inclusion and exclusion criteria. Crucially, we excluded patients with prior exposure to immunosuppressive therapies or with active EBV/CMV infections, as these factors are known to significantly modulate NK cell function and could thus confound their association with disease activity. Furthermore, by employing internationally standardized outcome definitions (e.g., DORIS for remission) and extracting data from structured electronic health records, we minimized information bias with objective and consistent measurements. Collectively, these strategies enhance the internal validity and reliability of our findings regarding the role of NK cells in childhood-onset SLE. The inter-assay coefficient of variation (CV), which was consistently < 6% for NK cell enumeration, confirming the robustness and reproducibility of our measurements.

This study is limited by its single-center design. Future multi-center investigations are encouraged to validate the role of NK cells in predicting disease flares and to verify the cut-off values established in this research. This research offers a meaningful contribution to pediatric rheumatology by demonstrating that the absolute NK cell count at diagnosis is a clinically significant biomarker in cSLE. It is associated with more severe disease activity, the presence of lupus nephritis, and most importantly serves as an independent factor of the time to achieve remission. The findings suggest that measuring NK cell counts at diagnosis could help stratify patients, identifying those at risk for a more aggressive disease course (low NK) and those likely to respond more quickly to treatment (high NK). This could potentially guide treatment intensity and monitoring strategies, laying the groundwork for standardizing NK cell metrics in cSLE research and clinical practice. Whether NK cell dynamics can serve as a reliable follow-up parameter remains uncertain. The cut-off value of 126 cells/µL for predicting remission was statistically derived from our cohort and should be regarded as exploratory. The threshold may be influenced by platform-specific variability, population heterogeneity, and differences in sample processing. Therefore, external validation in independent, multicenter, and ethnically diverse cohorts is essential before clinical application. As this is a retrospective study based on routine clinical flow cytometry data, our gating strategy was predefined for clinical reporting, which used the same fluorochrome (ECD) to detect CD56 and CD16, precluding the separation of CD56^bright, CD56^dim, or CD16⁺ NK subsets. Further precise research on subgroups will contribute to a more comprehensive understanding of the relationship between NK cells and cSLE. The diagnosis of lupus nephritis was based on retrospective medical records and included both biopsy-confirmed and clinically diagnosed cases. This heterogeneity in diagnostic criteria reflects the real-world nature of our cohort, particularly in the pediatric setting where renal biopsy is not always feasible. However, we recognize that the absence of uniform pathological confirmation may introduce misclassification bias.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study conception and design. TL contributed to the initial analyses and writing the first draft of the manuscript, SG checked the initial analyses and revised the manuscript, TZ provided statistical support and revised the manuscript, YZ and ZY provided the statistical analysis, TL, SP, YJ, CZ and MM managed patients and collected data, HS designed the study, revised the manuscript and provided fund support. All authors read and approved the final manuscript.

Funding

This work was supported by National Key R&D Program of China (2021YFC2702005) and National High Level Hospital Clinical Research Funding (2025-PUMCH-C-047).

Data availability

Data are available from the corresponding author on reasonable request.

Declarations

Ethics approval

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (JS-3362D). As this was a retrospective study, the requirement for informed consent was waived by the ethics committee.

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Competing interests

The authors declare no competing interests.