RRM2B is associated with low-level viremia during entecavir therapy in pediatric chronic hepatitis B and promotes profibrotic signaling in vitro

Abstract

Background

Low-level viremia (LLV) during nucleos(t)ide analogue therapy for chronic hepatitis B (CHB) is clinically relevant, but its biological basis in children remains unclear. We investigated the clinical features of pediatric LLV and sought host factors linking incomplete virological suppression to stress adaptation and profibrotic signaling.

Methods

Children with CHB who initiated entecavir therapy between 2019 and 2024 were retrospectively enrolled. LLV and maintained virological response (MVR) were defined by serum HBV DNA at week 48. Longitudinal trajectories were analyzed using generalized estimating equation models. Public liver transcriptomic datasets were integrated to identify CHB-associated host candidates, which were prioritized by machine learning and validated in pediatric samples. Functional studies were performed in an entecavir-based LLV-like HepG2.2.15 model.

Results

Among 109 children with CHB, 27.5% developed LLV. Compared with the MVR group, LLV was associated with higher baseline HBV pregenomic RNA (HBV pgRNA) and smaller declines in HBV DNA, aminotransferases, and APRI, with the earliest divergence at week 12. Integrated transcriptomic and machine-learning analyses identified 5,080 differentially expressed genes enriched in oxidative stress- and DNA damage-related pathways and prioritized RRM2B as a candidate host factor. RRM2B was elevated in pediatric LLV samples and correlated positively with HBV pgRNA. In vitro, low-dose entecavir induced an LLV-like state characterized by increased RRM2B, reactive oxygen species, γ-H2AX, and PI3K/AKT signaling. RRM2B silencing further suppressed HBV DNA replication, reduced oxidative stress and DNA damage, and attenuated LLV-like hepatocyte-derived TGF-β1 secretion and subsequent profibrotic activation of LX-2 cells.

Conclusions

RRM2B is associated with pediatric LLV and may be involved in stress-related and profibrotic host responses under conditions of incomplete virological suppression.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12985-026-03246-7.

Background

Hepatitis B virus (HBV) infection remains a major global health burden and a leading cause of cirrhosis, hepatocellular carcinoma (HCC), and liver-related mortality [1]. This burden is particularly pronounced in children, in whom immune immaturity confers a substantially increased risk of chronic infection following early-life exposure [2]. Despite effective vaccination and perinatal prevention strategies, a proportion of children still develop chronic hepatitis B (CHB) and remain at risk for fibrosis, cirrhosis, and HCC [3]. Accordingly, beyond preventing vertical transmission, improving long-term antiviral control and disease outcomes in pediatric CHB remains an important unmet clinical need.

Nucleos(t)ide analogues (NAs) constitute the cornerstone of antiviral therapy for CHB, and entecavir (ETV) is widely recommended as a first-line agent because of its potent antiviral activity and high genetic barrier to resistance [4]. However, durable and complete virological suppression is not achieved in all treated patients. In clinical practice, some patients develop low-level viremia (LLV), defined as detectable HBV DNA levels > 10 but < 2,000 IU/mL after 48 weeks of therapy [5, 6]. LLV has been associated with ongoing hepatic inflammation, fibrosis progression, and increased HCC risk [7], indicating clinically meaningful incomplete viral suppression. Nevertheless, current understanding of LLV has been derived predominantly from adult cohorts, whereas its clinical characteristics, biological implications, and mechanistic basis in pediatric CHB remain insufficiently defined.

The pathobiology of LLV is likely multifactorial and cannot be fully explained by residual covalently closed circular DNA (cccDNA) activity alone [8, 9]. Accumulating evidence suggests that host regulatory networks also contribute to viral persistence under antiviral pressure [10, 11]. In particular, host metabolic reprogramming and stress-adaptive responses have emerged as important determinants of persistent viral transcription and hepatocellular injury [12]. Previous studies have shown that HBV transcription and replication are modulated by host cellular factors [13–15]. Together, these findings suggest that LLV may be associated with adaptive host–virus interactions under antiviral pressure, potentially involving cellular stress-response and metabolic pathways.

Glutathione (GSH) metabolism regulates cellular redox balance and oxidative stress responses. Dysregulated GSH metabolism may promote oxidative injury, DNA damage responses, and profibrotic signaling, thereby contributing to chronic liver injury [16, 17]. Given the links among oxidative stress, HBV persistence, and hepatic injury, GSH-related pathways may also contribute to the maintenance of LLV. However, their role in LLV, particularly in pediatric CHB, remains unclear.

Accordingly, defining the host molecular determinants of LLV is of both mechanistic and clinical importance. A clearer understanding of LLV-associated host programs may improve risk stratification and identify potential host-directed therapeutic targets. In the present study, we combined longitudinal clinical phenotyping, transcriptomic discovery, and functional validation in an LLV-like hepatocyte model to identify host factors associated with persistent low-level replication. We prioritized RRM2B and explored its potential links to redox imbalance, DNA damage signaling, PI3K/AKT activation, and profibrotic crosstalk with hepatic stellate cells (HSCs).

Materials and methods

Patients

This retrospective cohort study included pediatric patients with CHB who initiated ETV therapy at the Children’s Hospital of Chongqing Medical University between January 2019 and December 2024. Eligible patients were 1–18 years of age, met guideline-based diagnostic criteria for CHB with persistent hepatitis B surface antigen (HBsAg) positivity for at least 6 months, had a baseline HBV DNA level ≥ 2,000 IU/mL, and were antiviral-treatment naïve. Exclusion criteria included co-infection with other hepatitis viruses or human immunodeficiency virus, liver disease of other etiologies, incomplete clinical records, poor treatment adherence, or HBV DNA ≥ 2,000 IU/mL after 48 weeks of therapy.

All enrolled patients received oral ETV once daily according to weight-based pediatric dosing recommendations (0.015 mg/kg/day for children weighing 10–30 kg and 0.5 mg/day for those weighing > 30 kg). Virological response was assessed after at least 48 weeks of continuous treatment. Patients were subsequently classified into the LLV group (HBV DNA > 10 and < 2,000 IU/mL) or the MVR group (HBV DNA ≤ 10 IU/mL) based on serum HBV DNA levels at week 48.

The study was approved by the Ethics Committee of the Children’s Hospital of Chongqing Medical University (No. 2025154) and was conducted in accordance with the Declaration of Helsinki.

Data collection

Baseline variables included age, sex, HBV DNA, HBV pgRNA, routine hematological parameters, and biochemical indices. Follow-up data were collected at weeks 12, 24, and 48 to evaluate longitudinal changes in virological markers, liver biochemical parameters, and non-invasive fibrosis-related indices between groups.

Public transcriptomic data collection and bioinformatic screening

Gene expression datasets related to CHB, liver fibrosis, and hepatocellular carcinoma were retrieved from the Gene Expression Omnibus (GEO) database, including GSE65359, GSE61260, GSE84044, GSE230397, and GSE94660. For CHB liver tissue datasets, batch effects were corrected using the ComBat algorithm and evaluated by principal component analysis (PCA). Differential expression analysis between CHB and normal liver tissues was performed using the limma package. Differentially expressed genes (DEGs) were identified using an FDR-adjusted P value ≤ 0.05 and |log₂FC| ≥ 0.3. DEGs were intersected with the GSH metabolism-related gene set from KEGG, followed by Gene Ontology (GO) and KEGG enrichment analyses. Candidate genes were further prioritized by LASSO regression, machine-learning models, and Shapley additive explanations (SHAP) analysis. Because liver transcriptomic datasets directly stratified by LLV and MVR are currently unavailable, this discovery step was designed to identify CHB-associated host candidates rather than LLV-specific markers.

Immunohistochemistry and ELISA

Liver specimens from normal controls and pediatric patients with CHB (including the LLV and MVR groups) were collected. Paraffin-embedded sections were deparaffinized, rehydrated, subjected to antigen retrieval, and incubated overnight at 4 °C with an anti-RRM2B or anti-γ-H2AX primary antibody, followed by the corresponding secondary antibody at room temperature. Immunoreactivity was visualized with diaminobenzidine and sections were counterstained with hematoxylin.

Peripheral blood RRM2B levels were measured using a commercial ELISA kit according to the manufacturer’s instructions. Absorbance was read at 450 nm, and concentrations were calculated from the standard curve.

Cell culture and transfection

The HBV-producing human hepatoma cell line HepG2.2.15 was obtained from the Shanghai Cell Bank. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Procell, China) supplemented with 10% fetal bovine serum (ExCell, USA), 1% penicillin-streptomycin (Solarbio, China), and 0.5 mg/mL G418 (MCE, China) at 37 °C in 5% CO₂.

Small interfering RNA (siRNA) targeting RRM2B and negative control siRNA were synthesized by Morzan Biotech (Shanghai, China) (Supplementary Table S1). Transfection was performed with Lipofectamine 3000 (Invitrogen, USA) according to the manufacturer’s protocol. After 24 h, the medium was replaced with fresh medium containing ETV (0, 1, or 5 µg/ml), and cells were harvested 72 h later.

In vitro model of graded virological suppression with an LLV-like phenotype

To establish an in vitro LLV-like model, HepG2.2.15 cells were treated with graded concentrations of ETV, as described in previous NAs-treated HBV replication models [14, 18, 19]. Untreated cells served as the high-replication control. Low-dose ETV (1 µg/ml) was used to generate partial virological suppression with detectable residual HBV DNA and was operationally defined as an LLV-like state. High-dose ETV (5 µg/ml) was used to model a more profound suppression state analogous to MVR. After 72 h, culture supernatants and cell lysates were collected for HBV DNA quantification, HBsAg/HBeAg measurement, and protein analyses.

RT-qPCR and western blotting

Total RNA was extracted with TRIzol reagent, and first-strand cDNA was synthesized from 1 µg RNA. Quantitative real-time PCR was performed using SYBR Green Master Mix on an ABI 7300 Real-Time PCR System. GAPDH was used as the internal control, and relative mRNA expression was calculated using the 2^−ΔΔCt method. Primer sequences are listed in Supplementary Table S2.

For western blotting, proteins were extracted in lysis buffer containing protease and phosphatase inhibitors, separated by SDS-PAGE, and transferred to PVDF membranes. After incubation with the indicated primary and secondary antibodies, signals were detected by enhanced chemiluminescence. Antibody information is provided in Supplementary Table S3.

Immunofluorescence

Cells were fixed, permeabilized, and blocked, then incubated overnight at 4 °C with antibodies against γ-H2AX or α-SMA. After incubation with Alexa Fluor 594-conjugated secondary antibody, nuclei were counterstained with DAPI. Images were acquired using a confocal microscope.

Quantification of HBsAg and HBeAg

Secreted HBsAg and HBeAg in culture supernatants were quantified using commercial ELISA kits (Enzyme-linked Biotechnology Co., Ltd, China) according to the manufacturer’s instructions. Measurements were performed in triplicate and recorded as absorbance at 450/630 nm.

HBV DNA assay

Extracellular virion-associated HBV DNA was enriched from 4 mL of culture supernatant by PEG8000 precipitation, followed by TURBO DNase digestion to remove free nucleic acids. DNA was extracted using a Viral Genome DNA/RNA Rapid Extraction Kit (Tiangen, China) and quantified by absolute qPCR. Primer sequences were as follows: forward, 5′-CCTAGTAGTCAGTTATGTCAAC-3′; reverse, 5′-TCTATAAGCTGGAGGAGTGCGA-3′.

Determination of intracellular ROS levels

Intracellular reactive oxygen species (ROS) levels were measured using a commercial ROS detection kit. HepG2.2.15 cells were seeded in 12-well plates at 1 × 10^5 cells/well and treated with ETV for 72 h. Cells were then incubated with 5 µM DCFH-DA for 30 min, and fluorescence signals were visualized under a fluorescence microscope. Data were normalized to the control and are presented as relative ROS intensity. Fluorescence intensity was quantified using ImageJ software.

Statistical analysis

Statistical analyses were performed in R (version 4.4.2). Normally distributed continuous variables are presented as mean ± standard deviation (SD) and were compared using the independent-samples t test. Non-normally distributed continuous variables are presented as median (interquartile range [IQR]) and were compared using the Mann–Whitney U test. Categorical variables are presented as n (%) and were compared using the chi-square test or Fisher’s exact test, as appropriate.

Longitudinal repeated-measures data were analyzed using generalized estimating equation (GEE) models to compare trajectories of HBV-related parameters, biochemical indices, and non-invasive fibrosis markers between groups from baseline to week 48. A time-by-group interaction term was included to assess between-group differences in temporal trends, and models were adjusted for age and sex. All tests were two-sided, and P < 0.05 was considered statistically significant.

Results

Baseline characteristics

A total of 109 pediatric patients with CHB who initiated ETV therapy were included in the analysis (Table 1). Of these, 30 were classified into the LLV group and 79 into the MVR group, corresponding to an LLV incidence of 27.5%. No significant between-group differences were observed in most baseline characteristics, including age, sex, baseline HBV DNA, HBcrAg, HBeAg positivity, ALT, AST, APRI, FIB-4, GPR, and LSM. In contrast, baseline HBV pgRNA levels were significantly higher in the LLV group than in the MVR group [5.89 (4.12–7.42) vs. 3.85 (2.66–6.70) log10 IU/mL, P = 0.0114].

Table 1.

Baseline characteristics of patients with LLV and MVR

Characteristics	MVR (n = 79)	LLV (n = 30)	P
Age (years)	6.50 (4.57, 9.57)	5.92 (4.25, 10.09)	0.9243
Male, n (%)	51 (64.6%)	21 (70.0%)	0.7569
HBV DNA (log10 IU/mL)	7.69 (6.98, 8.42)	7.56 (7.14, 8.17)	0.6014
HBcrAg (Log10 IU/mL)	4.80 (3.55, 7.25)	6.25 (4.12, 8.10)	0.1531
HBV pgRNA (Log10 IU/mL)	3.85 (2.66, 6.70)	5.89 (4.12, 7.42)	0.0114
HBeAg positive, n (%)	70 (88.6%)	26 (86.7%)	0.7498
WBC (×10^9/L)	6.50 (5.67, 7.90)	5.88 (5.24, 8.03)	0.3121
PLT (×10^9/L)	252.72 ± 72.64	266.17 ± 57.28	0.3148
HGB (g/L)	127.00 (121.50, 137.00)	126.00 (122.00, 137.50)	0.8016
TBil (µmol/L)	8.00 (4.95, 10.00)	7.80 (6.45, 11.15)	0.2913
DBil (µmol/L)	5.20 (3.20, 6.95)	4.95 (3.32, 8.18)	0.4657
IBil (µmol/L)	2.40 (0.00, 3.65)	2.50 (1.80, 3.45)	0.6316
ALP (U/L)	267.00 (226.00, 302.00)	243.00 (192.35, 291.35)	0.0842
ALT (U/L)	133.70 (71.05, 218.15)	86.20 (51.98, 280.58)	0.2658
AST (U/L)	94.00 (59.80, 168.50)	69.60 (48.03, 187.88)	0.2961
APRI	0.92 (0.58, 1.85)	0.63 (0.52, 1.71)	0.2853
FIB-4	0.23 (0.13, 0.39)	0.24 (0.11, 0.39)	0.7887
GPR	0.08 (0.05, 0.13)	0.08 (0.04, 0.11)	0.3687
GGT (U/L)	19.00 (13.00, 28.50)	17.00 (10.25, 31.25)	0.6249
LSM (kPa)	7.10 (5.85, 9.25)	6.75 (6.23, 8.67)	0.8387
Fibrosis stage (S), n (%)			0.6122
0	11 (13.9%)	6 (20.0%)
1	41 (51.9%)	12 (40.0%)
2	21 (26.6%)	8 (26.7%)
3	6 (7.6%)	4 (13.3%)
Inflammation grade (G), n (%)			0.3018
0	0 (0.0%)	0 (0.0%)
1	36 (45.6%)	9 (30.0%)
2	39 (49.4%)	20 (66.7%)
3	4 (5.1%)	1 (3.3%)

Data are shown as mean ± SD or median (IQR). APRI = (AST/ULN) × 100 / PLT; ;; LLV, low-level viremia; MVR, maintained virological response; HBV, hepatitis B virus; HBcrAg, hepatitis B core-related antigen; pgRNA, pregenomic RNA; WBC, white blood cell count; PLT, platelet count; ALT, alanine aminotransferase; AST, aspartate aminotransferase; ALB, albumin; GGT, gamma-glutamyl transferase; ALP, alkaline phosphatase; LSM, liver stiffness measurement; APRI, aspartate aminotransferase-to-platelet ratio index

Longitudinal trends of serum biomarkers during initial ETV therapy

Further longitudinal analyses showed (see Table 2 and Fig. 1) that serum HBV DNA levels declined in both groups during ETV therapy, but the reduction was consistently greater in the MVR group. In both unadjusted and age- and sex-adjusted GEE models, the group-by-time interaction for HBV DNA was significant (both P < 0.001). In the adjusted model, HBV DNA remained significantly higher in the LLV group at weeks 12, 24, and 48, with between-group differences of 1.09, 0.97, and 2.02 log₁₀ IU/mL, respectively (all P < 0.01). Importantly, this separation emerged as early as week 12 and widened further by week 48, indicating persistent between-group differences in virological suppression during treatment.

Table 2.

Longitudinal changes during ETV therapy in children with CHB

	MVR (n = 79)	LLV (n = 30)	GEE unadjusted Effect		GEE adjusted Effect
			β (95%CI)	P	Β (95%CI)	P
HBV DNA (log10 IU/mL)				< 0.001		< 0.001
Baseline	7.41 ± 1.36	7.44 ± 1.06
12 Week	4.09 ± 1.41	5.21 ± 1.82	1.09 (0.38, 1.80)	0.003	1.09 (0.38, 1.80)	0.003
24 Week	2.87 ± 1.33	3.87 ± 1.37	0.97 (0.24, 1.69)	0.009	0.97 (0.24, 1.69)	0.009
48 Week	0.90 ± 0.10	2.95 ± 0.98	2.02 (1.43, 2.61)	< 0.001	2.02 (1.43, 2.61)	< 0.001
ALT(U/L)				0.024		0.024
Baseline	181.43 ± 160.14	170.92 ± 162.21
12 Week	68.92 ± 42.50	134.67 ± 169.18	76.26 (23.02, 129.49)	0.005	76.25 (23.02, 129.48)	0.005
24 Week	42.1 ± 19.08	55.16 ± 51.67	23.57 (-39.51, 86.64)	0.464	23.43 (-39.64, 86.50)	0.467
48 Week	28.44 ± 12.99	31.31 ± 13.31	13.38 (-54.72, 81.48)	0.700	13.23 (-54.95, 81.41)	0.704
AST(U/L)				0.089		0.088
Baseline	125.66 ± 88.04	132.23 ± 122.64
12 Week	57.81 ± 24.47	108.54 ± 130.44	44.17 (2.34, 85.99)	0.038	44.15 (2.33, 85.97)	0.039
24 Week	42.81 ± 11.18	52.46 ± 29.71	3.08 (-41.49, 47.65)	0.892	2.80 (-41.73, 47.33)	0.902
48 Week	34.71 ± 9.47	37.68 ± 8.73	-3.59 (-51.43, 44.24)	0.883	-3.90 (-51.72, 43.93)	0.873
APRI				0.084		0.085
Baseline	1.38 ± 1.13	1.29 ± 1.20
12 Week	0.66 ± 0.58	1.21 ± 1.49	0.63 (0.13, 1.14)	0.014	0.63 (0.13, 1.14)	0.014
24 Week	0.43 ± 0.16	0.46 ± 0.25	0.11 (-0.36, 0.57)	0.648	0.11 (-0.35, 0.57)	0.643
48 Week	0.36 ± 0.16	0.35 ± 0.10	0.08 (-0.42, 0.58)	0.751	0.08 (-0.42, 0.58)	0.746
FIB-4				0.973		0.955
Baseline	0.29 ± 0.23	0.27 ± 0.19
12 Week	0.27 ± 0.29	0.26 ± 0.20	0.01 (-0.05, 0.07)	0.775	0.01 (-0.05, 0.07)	0.768
24 Week	0.21 ± 0.12	0.2 ± 0.12	0.00 (-0.06, 0.06)	0.961	0.01 (-0.05, 0.07)	0.780
48 Week	0.23 ± 0.13	0.21 ± 0.13	0.01 (-0.06, 0.07)	0.817	0.01 (-0.05, 0.08)	0.647
GPR				0.190		0.181
Baseline	0.12 ± 0.13	0.09 ± 0.08
12 Week	0.14 ± 0.28	0.13 ± 0.17	0.02 (-0.06, 0.10)	0.602	0.02 (-0.06, 0.10)	0.600
24 Week	0.08 ± 0.09	0.06 ± 0.04	0.00 (-0.04, 0.04)	0.958	0.00 (-0.03, 0.04)	0.885
48 Week	0.05 ± 0.03	0.05 ± 0.02	0.02 (-0.01, 0.06)	0.220	0.03 (-0.01, 0.06)	0.191
LSM(kPa)				0.583		0.571
Baseline	7.79 ± 3.12	8.16 ± 4.12
12 Week	6.53 ± 1.68	6.79 ± 2.16	-0.08 (-1.35, 1.18)	0.899	-0.08 (-1.35, 1.18)	0.896
24 Week	5.18 ± 1.22	6.01 ± 1.51	0.46 (-1.01, 1.94)	0.538	0.47 (-1.00, 1.94)	0.533
48 Week	4.68 ± 1.27	5.41 ± 1.43	0.37 (-1.22, 1.95)	0.651	0.37 (-1.21, 1.94)	0.646

Data are presented as n (%) or n/N (%), or mean ± SD

Fig. 1.

Longitudinal trends in serum biomarker levels during antiviral therapy in the LLV and MVR groups. (A) HBV DNA, (B) ALT, (C) AST, (D) APRI, (E) FIB-4, (F) GPR, and (G) LSM. Line plots represent mean values at each time point, with SD indicated by error bars

*P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant

Serum ALT and AST levels also decreased in both groups, although between-group.

differences were largely restricted to the early treatment phase. In the adjusted GEE model, a significant group-by-time interaction was observed for ALT (P = 0.024), whereas the interaction for AST was borderline (P = 0.088). At week 12, both ALT and AST remained significantly higher in the LLV group, with adjusted between-group differences of 76.25 U/L (95% CI, 23.02-129.48; P = 0.005) and 44.15 U/L (95% CI, 2.33–85.97; P = 0.039), respectively. These differences were no longer evident at weeks 24 or 48, suggesting that biochemical divergence was most pronounced early during treatment and diminished over time.

Among the noninvasive fibrosis-related indices, only APRI differed significantly between groups at week 12, being higher in the LLV group (adjusted β = 0.63, 95% CI, 0.13–1.14; P = 0.014), whereas the overall group-by-time interaction was not statistically significant (adjusted P = 0.085). No significant overall or time-specific differences were observed for FIB-4, GPR, or LSM. Taken together, these findings suggest that week 12 may represent the earliest clinically informative time point at which LLV and MVR become distinguishable.

Integrated transcriptomic and machine-learning analyses identify RRM2B as a CHB-associated host candidate

Because liver transcriptomic datasets directly stratified by LLV and MVR are currently unavailable, four GEO datasets of CHB liver tissues (GSE65359, GSE61260, GSE84044, and GSE230397) were integrated to identify host candidates associated with chronic HBV-related liver injury and stress adaptation. After normalization and batch correction, 276 CHB liver samples and 143 normal liver samples were included in the integrated analysis, and principal component analysis supported acceptable cross-dataset comparability (Fig. 2A). Differential expression analysis identified 5,080 differentially expressed genes between CHB and normal liver tissues (Fig. 2B and D), which were mainly enriched in GSH metabolism, oxidative stress, and DNA damage-related pathways (Fig. 2C). Intersection with a GSH metabolism-related gene set yielded 26 candidate genes (Fig. 2D). To assess whether these candidates reflected nonspecific hepatic inflammation, we analyzed GSE206364 as a pediatric non-HBV inflammatory liver disease reference. Among 836 AIH-associated DEGs, including 355 upregulated and 481 downregulated genes, none overlapped with the GSH metabolism-related gene set (Supplementary Fig. 1A, B). Subsequent LASSO regression reduced the candidate genes to nine feature genes (Fig. 2E, F). Among the evaluated machine-learning models, AdaBoost achieved the best discriminative performance in the validation set, with an area under the curve (AUC) of 0.937 (Supplementary Table S2; Fig. 2G). SHAP analysis ranked RRM2B as one of the top contributing features among the selected genes, and it was therefore prioritized for subsequent validation in pediatric LLV (Fig. 2H).

Fig. 2.

Integrated transcriptomic and machine-learning analyses prioritize RRM2B in CHB. (A) PCA of the integrated CHB liver transcriptomic datasets before (left) and after (right) ComBat-based batch correction. (B) Volcano plot showing the DEGs between CHB and normal liver tissues; genes on the right side of the volcano plot indicate genes upregulated in the CHB group. (C) Functional enrichment analyses of the DEGs, including representative GO terms and KEGG pathways. (D) Venn diagram showing the overlap between CHB-related DEGs and the KEGG-derived glutathione metabolism-related gene set, yielding 26 candidate genes. (E) LASSO regression analysis, including the cross-validation curve (left) and coefficient profiles (right), used for feature selection. (F) Model-based importance ranking of the nine selected feature genes. (G) ROC curves showing the discriminative performance of the evaluated machine-learning models in the validation set (left) and decision curve analysis of model performance (right). AdaBoost showed the best predictive performance. (H) SHAP analysis showing the contribution of the selected feature genes to model output, with RRM2B ranked as the top feature. CHB, chronic hepatitis B; DEG, differentially expressed gene; ROC, receiver operating characteristic; AUC, area under the curve; SHAP, Shapley additive explanations

RRM2B is elevated in progressive HBV-related liver disease and pediatric LLV

To further assess the biological relevance of RRM2B in CHB-related disease progression, its expression was examined in liver fibrosis and HBV-related HCC datasets. RRM2B expression was significantly higher in fibrotic liver tissues than in non-fibrotic controls and was likewise elevated in HBV-related HCC tissues relative to non-tumorous controls (Fig. 3A). These findings suggest that RRM2B upregulation is associated not only with chronic infection but also with more advanced HBV-related liver pathology.

Fig. 3.

RRM2B is clinically associated with pediatric LLV. (A) RRM2B expression levels in independent liver fibrosis datasets and an HBV-related hepatocellular carcinoma dataset. (B) Representative images of immunohistochemical staining for RRM2B in liver biopsy tissues from pediatric patients with MVR and LLV. The quantitative analysis of the RRM2B-positive area is shown on the right. (C) Serum RRM2B levels in the MVR and LLV groups at baseline and during treatment. (D) Correlation analysis between RRM2B and HBV pgRNA levels. (E) ROC curves showing the discrimination performance of RRM2B, HBV pgRNA, and the combined nomogram for LLV. (F) The nomogram constructed using RRM2B and HBV pgRNA, together with the calibration curve of the original model. (G) Internal calibration curve of the nomogram after bootstrap validation. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant; MVR, maintained virological response; LLV, low-level viremia; ROC, receiver operating characteristic; AUC, area under the curve

RRM2B was then evaluated in liver biopsy tissues and peripheral blood samples from pediatric patients with CHB. Immunohistochemical staining was performed using baseline liver biopsy tissues obtained before antiviral therapy initiation from 11 patients in the LLV group and 11 patients in the MVR group. Pediatric CHB liver biopsy tissues showed stronger hepatic RRM2B staining in the LLV group than in the MVR group, which was confirmed by quantitative analysis of the positive staining area (Fig. 3B). Consistently, circulating RRM2B levels were higher in the LLV group than in the MVR group at baseline and week 12, whereas no significant between-group difference was observed at weeks 24 or 48 (Fig. 3C). Together, these findings support a clinically relevant association between RRM2B upregulation and the LLV phenotype in pediatric CHB.

RRM2B correlates with HBV pgRNA and supports exploratory LLV risk estimation

To examine whether RRM2B was linked to persistent viral transcription, correlation analysis was performed between RRM2B and HBV pgRNA. RRM2B levels were positively correlated with HBV pgRNA levels (Fig. 3D), supporting an association between RRM2B and sustained HBV transcriptional activity.

An exploratory nomogram integrating RRM2B and HBV pgRNA was constructed to estimate the probability of LLV in pediatric CHB. Receiver operating characteristic (ROC).

analysis yielded AUC values of 0.845 for RRM2B, 0.940 for HBV pgRNA, and 0.960 for the combined nomogram model (Fig. 3E). The calibration curve of the original model suggested good agreement between predicted and observed probabilities (Fig. 3F), and bootstrap-based internal validation showed that the corrected model retained acceptable calibration performance (Fig. 3G). Nevertheless, some deviation remained in specific probability intervals, underscoring the need for external validation in an independent cohort.

An LLV-like state is accompanied by RRM2B induction and stress activation

To characterize host responses accompanying persistent low-level viral replication, an in vitro LLV-like model of graded virological suppression was established in HepG2.2.15 cells. Relative to untreated high-replication controls, low-dose ETV treatment (1 µg/ml) markedly reduced supernatant HBV DNA levels while maintaining detectable viral DNA, whereas high-dose ETV treatment (5 µg/ml) further suppressed HBV DNA to near the lower limit of detection (Fig. 4A). HBsAg and HBeAg levels showed no marked change in the low-dose ETV group, whereas HBV pgRNA remained relatively preserved compared with the high-dose ETV group, consistent with the establishment of an LLV-like state characterized by incomplete virological suppression (Fig. 4A), supporting the use of this condition as an in vitro LLV-like model.

Fig. 4.

An LLV-like HBV state induces RRM2B expression and cellular stress responses. (A) Relative HBV DNA, HBV pgRNA, HBsAg, and HBeAg levels in HepG2.2.15 cells treated with 0, 1, or 5 µg/ml ETV. (B) Representative western blots showing RRM2B and γ-H2AX expression under graded ETV treatment. The quantitative analysis is shown on the right. (C) Representative DCFH-DA fluorescence images showing intracellular ROS levels. The quantitative analysis is shown on the right. (D) Representative immunofluorescence images of nuclear γ-H2AX foci. (E) Representative γ-H2AX staining and quantitative analysis in baseline liver biopsy specimens from pediatric patients in the LLV and MVR groups. The quantitative analysis is shown on the right. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant; ETV, entecavir; ROS, reactive oxygen species

γ-H2AX, a widely used marker of DNA damage, particularly DNA double-strand breaks, was evaluated to assess DNA damage responses under LLV-like conditions. Under these conditions, RRM2B and γ-H2AX protein levels were both increased in the low-dose ETV group, whereas high-dose ETV reduced their expression relative to the LLV-like condition (Fig. 4B). The LLV-like condition was accompanied by increased intracellular ROS accumulation and increased nuclear γ-H2AX foci formation (Fig. 4C, D), consistent with increased oxidative stress and DNA damage signaling under LLV-like conditions.

To validate these in vitro findings in clinical samples, γ-H2AX staining was additionally performed in baseline liver biopsy specimens from pediatric patients in the LLV and MVR groups. Compared with the MVR group, liver tissues from the LLV group showed more pronounced nuclear γ-H2AX staining, and quantitative analysis confirmed a higher γ-H2AX-positive nuclear signal in LLV liver tissues (Fig. 4E).

RRM2B knockdown suppresses HBV replication and stress responses under LLV-like conditions

To assess the functional role of RRM2B in LLV maintenance, RRM2B was silenced in the HepG2.2.15-based LLV model using siRNA (20 nM, 24 h transfection), followed by ETV treatment (1 µg/ml for 72 h). Western blotting and RT-qPCR confirmed effective suppression of RRM2B expression in the LLV + si-RRM2B group, which was accompanied by a significant reduction in supernatant HBV DNA levels (Fig. 5A, B).

Fig. 5.

RRM2B knockdown suppresses HBV replication and attenuates stress responses under LLV-like conditions. (A) RRM2B knockdown efficiency in HepG2.2.15 cells was confirmed by western blotting and RT-qPCR. (B) Relative HBV DNA levels after RRM2B silencing under LLV-like conditions. (C) Representative western blots showing TGF-β1, RRM2B, and γ-H2AX expression in the indicated groups. (D) Quantitative analysis of TGF-β1, RRM2B, and γ-H2AX protein levels. (E) Representative DCFH-DA fluorescence images showing intracellular ROS accumulation. The quantitative analysis is shown on the right. (F) Representative immunofluorescence images of γ-H2AX foci. The quantitative analysis is shown on the right. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant

TGF-β1 expression was evaluated as a representative profibrotic mediator to assess potential changes in hepatocyte-mediated profibrotic signaling under LLV-like conditions. Western blot analysis showed that RRM2B knockdown reduced TGF-β1 expression and decreased γ-H2AX levels relative to the LLV + negative control SiRNA (SiNC) group (Fig. 5C, D). This pattern suggests that RRM2B inhibition, together with reduced HBV DNA levels, is associated with attenuation of DNA damage-related and profibrotic marker expression under LLV-like conditions.

Consistently, immunofluorescence analysis demonstrated that the LLV + Si-RRM2B group exhibited fewer nuclear γ-H2AX foci and reduced ROS accumulation compared with the LLV + SiNC group (Fig. 5E, F). Together, these findings indicate that RRM2B downregulation suppresses HBV replication and alleviates oxidative and DNA damage-associated stress under LLV-like conditions, supporting an association between RRM2B and stress-related as well as profibrotic cellular responses under LLV-like conditions.

PI3K/AKT signaling is associated with RRM2B induction in an in vitro LLV-like model

Previous studies have suggested that PI3K/AKT signaling participates in HBV-related cellular responses, including oxidative stress and DNA damage-associated signaling, while RRM2B has been implicated in cellular stress responses [20, 21]. Compared with the high-replication condition, the LLV-like state was characterized by increased phosphorylation of PI3K and AKT, whereas total PI3K and AKT levels remained largely unchanged, indicating activation of the PI3K/AKT pathway (Fig. 6A).

Fig. 6.

PI3K/AKT signaling contributes to RRM2B induction under LLV-like conditions. (A) Representative western blots showing p-PI3K/PI3K and p-AKT/AKT expression in HepG2.2.15 cells treated with different concentrations of ETV. The quantitative analysis is shown on the right. Phosphorylated PI3K and AKT were normalized to total PI3K and AKT, respectively. Total protein levels were normalized to GAPDH. (B) Effects of PI3K inhibition on RRM2B and γ-H2AX expression under LLV-like conditions, shown by representative western blots and quantitative analysis, and normalized with GAPDH. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant; ETV, entecavir

RRM2B expression was repressed by the PI3K inhibitor LY294002 (20 µM, 24 h), accompanied by reduced γ-H2AX expression (Fig. 6B). These observations suggest a possible association between PI3K/AKT signaling and RRM2B induction under LLV-like conditions.

RRM2B-upregulated hepatocytes promote LX-2 activation through paracrine signaling

Given the critical role of hepatocyte–hepatic stellate cell crosstalk in liver fibrogenesis, and the predominant expression of RRM2B in hepatocytes, we next explored whether RRM2B upregulation in LLV-like hepatocytes altered the secretion of profibrotic mediators. ELISA analysis of HepG2.2.15 culture supernatants showed that TGF-β1 levels were significantly increased under LLV-like conditions and remained elevated in the SiNC group, whereas RRM2B silencing markedly reduced TGF-β1 levels in the corresponding supernatants (Fig. 7A). These findings suggested that hepatocellular RRM2B upregulation may be associated with enhanced release of profibrotic signals.

Fig. 7.

RRM2B-upregulated hepatocytes promote LX-2 activation through paracrine signaling. (A) ELISA analysis of TGF-β1 levels in culture supernatants collected from HepG2.2.15 cells under control, LLV-like, LLV + SiNC, and LLV + SiRRM2B conditions. (B) Representative western blots showing COL1A1, p-Smad3, Smad3, and α-SMA expression in LX-2 cells after stimulation with hepatocyte-derived conditioned media. (C) Quantitative analysis of p-Smad3/Smad3, α-SMA, and COL1A1 protein levels in LX-2 cells. GAPDH was used as an internal control (left). Protein expression was normalized to that of GAPDH (right). (D) Representative immunofluorescence images of α-SMA in LX-2 cells. The quantitative analysis of α-SMA intensity is shown on the right. *P < 0.05, **P < 0.01, ***P < 0.001

To determine whether these hepatocyte-derived factors could influence HSCs activation, culture supernatants from control, LLV-like, LLV + SiNC, and LLV + SiRRM2B HepG2.2.15 cells were collected and used to prepare conditioned media (CM) for LX-2 stimulation (Fig. 7B, C). Compared with Ctrl-CM, both ETV-CM and SiNC-CM increased α-SMA and COL1A1 expression in LX-2 cells, whereas SiRRM2B-CM substantially attenuated this profibrotic activation pattern. Consistently, p-Smad3 levels were increased in LX-2 cells exposed to ETV-CM and SiNC-CM, but this effect was markedly blunted in the SiRRM2B-CM group. Immunofluorescence analysis of α-SMA further supported these findings (Fig. 7D). Collectively, these data suggest that hepatocytes with elevated RRM2B under LLV-like conditions may promote HSCs activation, at least in part, associated with increased TGF-β1 secretion and activation of downstream Smad3 signaling.

Discussion

LLV is increasingly recognized as a clinically meaningful virological state during NAs therapy [18], yet its relevance in pediatric CHB remains insufficiently defined. To our knowledge, few studies have jointly interrogated the clinical trajectory and host biological correlates of LLV specifically in children. In the present study, children who developed LLV exhibited higher baseline HBV pgRNA levels and comparatively attenuated virological and biochemical responses during ETV treatment. Notably, divergence between the LLV and MVR groups was most evident at week 12, particularly for HBV DNA, aminotransferases, and APRI. It should be noted that the observed APRI difference at this time point may primarily reflect variations in aminotransferase levels. These findings suggest that pediatric LLV is not merely a delayed endpoint phenomenon, but may instead reflect an early divergence in treatment-response kinetics.

From a clinical standpoint, this observation highlights the potential value of early on-treatment monitoring for identifying children at risk of suboptimal virological suppression.

The biological basis of LLV is unlikely to be explained solely by persistent cccDNA activity. Although cccDNA remains the central transcriptional reservoir of HBV, increasing evidence indicates that viral persistence under antiviral pressure is also shaped by host metabolic adaptation and stress-response networks [13, 15]. When comparing CHB liver tissues with normal liver tissues, our transcriptomic analysis identified prominent enrichment of GSH metabolism, oxidative stress, and DNA damage-associated pathways in persistent CHB, pointing to substantial disruption of redox homeostasis in chronically infected liver tissue. Within this framework, RRM2B emerged as a prioritized candidate gene following integrated prioritization by differential expression analysis, LASSO regression, multiple machine-learning models, and SHAP interpretation. Importantly, RRM2B upregulation was observed not only in CHB-related fibrosis and HBV-associated HCC datasets but also in liver tissues and peripheral blood from pediatric patients with LLV. Its positive correlation with HBV pgRNA further supports a link between RRM2B and ongoing viral transcription rather than a nonspecific association with liver injury alone [22]. Collectively, these findings support an association between RRM2B and stress-related host responses accompanying the LLV phenotype.

The longitudinal serum profile further contextualized the association between RRM2B and the LLV phenotype. In patients with LLV, serum RRM2B levels peaked at week 12 and declined thereafter, and the early between-group differences were no longer evident at weeks 24 and 48. These findings suggest that circulating RRM2B may reflect a dynamic host-response state during the early phase of antiviral therapy rather than a stable indicator of residual viremia. Week 12 represents a clinically informative window for evaluating early virological response in CHB, during which incomplete viral suppression may coincide with enhanced HBV-related cellular stress [23]. Given the established involvement of RRM2B in DNA damage and replication-stress responses, its early elevation in LLV patients may indicate a stress-adaptive response linked to residual viral activity. As treatment continued, the decline in HBV DNA and liver injury markers, as shown in Fig. 1A, B, was accompanied by a gradual reduction in serum RRM2B levels in the LLV group toward the relatively stable profile observed in the MVR group.

The functional data provide mechanistic context for this association. RRM2B encodes the p53R2 small subunit of ribonucleotide reductase and participates in deoxyribonucleotide supply, DNA repair, oxidative stress responses, and mitochondrial homeostasis [24, 25]. These functions place RRM2B at the intersection of cellular stress adaptation and viral persistence-related host responses. In the present study, low-dose ETV induced an LLV-like state characterized by increased RRM2B expression, ROS accumulation, and γ-H2AX activation. Notably, RRM2B silencing reduced HBV DNA levels and was accompanied by attenuation of oxidative stress, DNA damage signaling, and profibrotic mediator expression. These findings suggest that RRM2B upregulation is associated with stress-related cellular responses under LLV-like conditions.

Our results additionally suggest a potential association between PI3K/AKT activation and RRM2B upregulation under LLV-like conditions. Activation of PI3K/AKT was observed in the LLV-like model, whereas pharmacological PI3K inhibition reduced RRM2B expression together with γ-H2AX activation and ROS accumulation. Although the mechanistic relationships within this pathway remain to be clarified, these findings support a possible association between PI3K/AKT signaling and stress-associated host responses under incomplete virological suppression.

Another important aspect of the present study is the potential link between LLV-associated host adaptation and fibrogenic signaling. Supernatants from hepatocytes under LLV-like conditions contained higher levels of TGF-β1 and promoted α-SMA, COL1A1, and p-Smad3 expression in LX-2 cells, whereas these effects were attenuated following RRM2B silencing. These findings suggest that the biological consequences of RRM2B upregulation are unlikely to be confined to the infected hepatocyte itself. Rather, RRM2B-associated changes in hepatocyte signaling may contribute to enhanced profibrotic communication with HSCs. This observation raises the possibility that persistent LLV may be associated with fibrosis-related signaling changes, thereby extending the significance of LLV beyond incomplete viral suppression alone.

Several limitations should be acknowledged. First, this was a single-center retrospective study with a relatively modest sample size, which may limit the generalizability of the findings. Second, the predictive model underwent internal validation only and requires further external validation. In addition, quantitative serum HBsAg and hepatic HBsAg immunohistochemistry may provide additional information for distinguishing LLV from MVR; however, incomplete quantitative HBsAg data and the predominantly qualitative nature of HBsAg immunohistochemistry records limited further analyses. Finally, although our in vitro experiments support an association among RRM2B, PI3K/AKT signaling, and hepatocyte-mediated profibrotic responses, the precise molecular mechanisms underlying this regulatory network require further investigation.

Conclusion

In conclusion, the present study identifies RRM2B as a host factor associated with LLV during ETV therapy in pediatric CHB. Our findings suggest that LLV may reflect not only incomplete virological suppression but also stress-related host responses. In vitro findings further indicate that RRM2B may contribute to hepatocyte-mediated profibrotic signaling. These results highlight the potential value of RRM2B as a biomarker candidate and support further investigation of host-directed mechanisms in LLV-associated CHB.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

LLV

Low-level viremia

CHB

Chronic hepatitis B

MVR

Maintained virological response

pgRNA

Pregenomic RNA

HCC

Hepatocellular carcinoma

HBsAg

Hepatitis B surface antigen

NAs

Nucleos(t)ide analogues

ETV

Entecavir

HBV-HCC

HBV-related hepatocellular carcinoma

cccDNA

Covalently closed circular DNA

GSH

Glutathione

Author contributions

YZ designed the study. AT and TW planned and supervised the data collection. YZ, YC, and XP performed the data analyses. YZ, KZ, and QT collected tissue specimens and clinical data. YZ wrote the original manuscript, and HX revised it. All authors read and approved the final manuscript.

Funding

This study was funded by 2023 National Key R&D Program of China (2023YFC2308100).

Data availability

The data that support the findings of this study are available on request from the corresponding author.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Children’s Hospital of Chongqing Medical University (No. 2025154) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from the parents or legal guardians of all pediatric participants. Analyses of publicly available GEO datasets did not require additional ethical approval.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.