Abstract

   Bone metastasis in hepatocellular carcinoma (HCC) poses a significant
   clinical challenge, characterized by poor prognosis and severe skeletal
   complications. This study identifies the HMGB1/LCN2/JAK1/STAT3 axis as
   the central mechanism driving HCC bone metastasis through
   tumor-osteoclast crosstalk. High-mobility group box 1 (HMGB1) induces
   osteoclast activation and differentiation, promoting lipocalin-2 (LCN2)
   secretion by osteoclasts, which activates the JAK1/STAT3 pathway in HCC
   cells, forming a feedback loop that enhances osteolytic bone resorption
   and tumor dissemination. Integrated single-cell and bulk RNA sequencing
   reveal enriched osteoclast-related and pro-metastatic pathways in the
   tumor-bone microenvironment, while functional assays involving
   knockdown and overexpression demonstrate that modulating the
   HMGB1/LCN2/JAK1/STAT3 axis regulates osteoclast activity, tumor growth,
   and bone destruction in vitro and in vivo. These results suggest the
   HMGB1/LCN2/JAK1/STAT3 axis as a potential therapeutic target, offering
   a strategy to reduce skeletal damage and systemic tumor progression,
   thereby contributing to improved management of advanced HCC.

   graphic file with name 41419_2025_8037_Figa_HTML.jpg

   Subject terms: Bone metastases, Bone cancer

Introduction

   Hepatocellular carcinoma (HCC) ranks as the third leading cause of
   cancer-related mortality worldwide, with a high propensity for distant
   metastasis [[32]1]. Among metastatic sites, the skeleton represents the
   second most common target, with bone metastasis affecting 25.5–38.5% of
   patients [[33]2, [34]3]. These patients experience a poor prognosis,
   with a median survival of only 4.6 months following diagnosis,
   accompanied by significant complications such as severe pain,
   fractures, and neurological compression [[35]4, [36]5]. Thus, a deeper
   understanding of the mechanisms governing the tumor-bone
   microenvironment is essential for developing effective therapies
   targeting metastasis with clinical relevance.

   The bone-tumor microenvironment is a complex and dynamic system
   comprising tumor cells, osteoclasts, stromal cells, and immune
   components that interact through intricate signaling networks to drive
   tumor progression [[37]6]. In HCC, osteolytic bone destruction is the
   predominant phenotype of bone metastasis, primarily mediated by
   excessive osteoclast activation [[38]7]. Tumor-derived factors such as
   parathyroid hormone-related protein (PTHrP), interleukin-6 (IL-6), and
   tumor necrosis factor-alpha (TNF-α), have been shown to promote
   osteoclast differentiation and activity, thereby accelerating bone
   resorption [[39]7, [40]8]. This resorptive process, in turn, releases
   matrix-embedded growth factors, including transforming growth
   factor-beta (TGF-β), insulin-like growth factor (IGF), and calcium,
   which further enhance tumor cell proliferation, survival, and invasion
   [[41]8, [42]9]. Notably, the crosstalk between osteoclasts and tumor
   cells establishes a vicious cycle that sustains bone destruction and
   tumor growth [[43]10]. Emerging evidence from other malignancies, such
   as breast and lung cancers, underscores the multifaceted roles of
   osteoclasts beyond bone resorption [[44]11, [45]12]. For instance,
   osteoclasts have been shown to suppress T cell-mediated immunity and
   secrete pro-tumorigenic factors, actively contributing to tumor
   progression within the bone microenvironment [[46]13]. Together, the
   osteolytic phenotype and tumor-osteoclast interactions play pivotal
   roles in HCC bone metastasis, emphasizing the need for targeted
   strategies to interrupt this pathological cycle and improve clinical
   outcomes.

   High-mobility group box 1 (HMGB1), a widely expressed
   chromatin-associated protein, regulates nuclear functions such as gene
   transcription, DNA repair, and chromatin remodeling [[47]14],
   modulating processes such as cell differentiation, migration,
   inflammation, and tumor progression [[48]15, [49]16]. HMGB1’s oncogenic
   roles are well-documented in cancers such as colorectal, oral, and
   breast cancers [[50]17–[51]19], with growing evidence of its importance
   in HCC. In HCC, HMGB1 promotes tumor growth through immune modulation
   [[52]20] and enhances metastatic potential under hypoxic conditions via
   mechanisms involving mitochondrial dynamics [[53]21]. Within the bone
   microenvironment, HMGB1 interacts with toll-like receptor 4 (TLR4) to
   regulate osteoclast activation and modulate macrophage responses,
   processes implicated in osteolytic bone destruction [[54]22–[55]24].
   However, its specific role in the HCC tumor-bone microenvironment,
   particularly in orchestrating tumor-osteoclast interactions to promote
   osteolytic bone metastasis, remains underexplored. This gap highlights
   the need for detailed investigation to clarify HMGB1’s contribution to
   HCC bone metastasis and to identify potential therapeutic targets.

   This study aims to investigate the role of HMGB1 in facilitating tumor
   progression, with a particular emphasis on its interactions with
   osteoclast activation and the JAK1/STAT3 signaling pathway within the
   tumor-bone microenvironment, providing a foundation for understanding
   the mechanisms of HCC bone metastasis.

Results

HMGB1 is an unfavorable prognostic factor and driver of bone metastasis in
HCC

   To explore the role of HMGB1 in HCC metastasis, a metastatic HCC
   single-cell RNA sequencing (scRNA-seq) dataset was analyzed. As
   described in our previous studies [[56]25], UMAP-based dimensionality
   reduction identified three HCC subpopulations: the primary tumor group
   (HCC^PT), the transformation group with metastatic potential
   (HCC^Trans), and the metastatic lesion group (HCC^PVTT) (Fig. [57]1A).
   HMGB1 expression was significantly upregulated in HCC cells compared to
   normal hepatocytes, with the highest expression observed in the
   HCC^Trans subpopulation (Fig. [58]1B). To assess clinical relevance,
   HMGB1 protein expression was assessed in tissue microarrays from a
   cohort of 270 HCC patients at Shanghai East Hospital using
   immunohistochemistry. Staining intensities, quantified with QuPath
   software, enabled stratification of patients into high and low HMGB1
   expression groups based on median values. Statistical analysis revealed
   significant associations between high HMGB1 expression and increased
   bone metastasis (p = 2.7e−6), larger tumor size (p = 7.0e−7), advanced
   BCLC stage (p = 2.5e−10), and vascular invasion (p = 2.5e−4) (Table
   [59]1). Survival analysis further indicated that patients with high
   HMGB1 expression experienced reduced overall survival (OS, Fig. [60]1C)
   and bone metastasis-free survival (BM-FS, Fig. [61]1D).
   Immunofluorescence analysis confirmed elevated HMGB1 protein expression
   in tumor tissues from bone metastasis patients compared to
   non-metastasis controls (Fig. [62]1E).

Fig. 1. HMGB1 expression and its role in HCC bone metastasis progression.

   [63]Fig. 1
   [64]Open in a new tab

   A UMAP of single-cell RNA-seq data identifying three HCC
   subpopulations: primary tumor group (HCC^PT), transformation group with
   metastatic potential (HCC^Trans), and metastatic lesion group
   (HCC^PVTT). B Violin plots showing HMGB1 upregulation in HCC cells
   (p < 0.001), particularly in the HCC^Trans subpopulation (p < 0.001).
   C, D Kaplan–Meier curve depicting OS and BM-FS of HCC patients
   stratified by high and low HMGB1 expression. E Representative
   immunofluorescence images of HMGB1 expression in HCC tissues with or
   without bone metastasis. F Schematic of the HCC bone metastasis mouse
   model and bone marrow RNA-seq analysis. G Volcano plot of
   differentially expressed genes in bone marrow (Vector vs. Normal)
   showing upregulation of osteoclast-related genes and downregulation of
   epithelial and immune markers. H Volcano plot of differentially
   expressed genes in bone marrow (HMGB1 vs. Vector) highlighting
   osteoclast-associated genes and osteogenesis-related markers.

Table 1.

   Demographic and clinical characteristics of patients included in this
   study.
                                    HMGB1 expression group    p value
                                  High (n = 135) Low (n = 135)
   Age (years)                                                6.9e−4
                        Mean ± SD 62.05 ± 7.96   65.43 ± 8.20
   Gender                                                     0.87
                        Male      113 (41.85%)   111 (41.11%)
                        Female    22 (8.15%)     24 (8.89%)
   Bone metastasis                                            2.7e−6
                        Yes       38 (14.07%)    8 (2.96%)
                        No        97 (35.93%)    127 (47.04%)
   HBsAg                                                      0.11
                        Positive  102 (37.78%)   89 (32.96%)
                        Negative  33 (12.22%)    46 (17.04%)
   AFP (ng/ml)                                                0.30
                        ≤20       87 (32.22%)    96 (35.56%)
                        >20       48 (17.78%)    39 (14.44%)
   BCLC stage                                                 2.5e−10
                        0–A       42 (15.56%)    95 (35.19%)
                        B–C       93 (34.44%)    40 (14.81%)
   Tumor size (cm)                                            7.0e−7
                        ≤5        34 (12.59%)    75 (27.78%)
                        >5        101 (37.41%)   60 (22.22%)
   vascular invasion                                          2.5e−4
                        Yes       79 (29.26%)    48 (17.78%)
                        No        56 (20.74%)    87 (32.22%)
   Intrahepatic control                                       1.5e−12
                        Yes       43 (15.93%)    102 (37.78%)
                        No        92 (34.07%)    33 (12.22%)
   [65]Open in a new tab

   NA Not applicable.

   To further investigate the functional role of HMGB1, a mouse model of
   HCC bone metastasis was established (Fig. [66]1F). RNA-seq of bone
   marrow from this model identified significant changes in gene
   expression. Compared to the normal group, 122 genes were upregulated in
   the Vector group, including osteoclast-related Fgf23, the poor
   prognostic marker Mcam, and angiogenesis-related Plvap, alongside
   downregulation of 246 genes, such as the epithelial marker Cdh1 and
   immune-related genes Slamf7 and Csf3r, were downregulated (Fig.
   [67]1G). Gene Set Enrichment Analysis (GSEA) identified enrichment of
   the “GOBP_lymphocyte activation involved in immune response” pathway in
   the normal group, indicating enhanced anti-tumor immunity
   (Supplementary Fig. [68]S1A), while the “HALLMARK_epithelial
   mesenchymal transition” (EMT) pathway was prominent in the Vector
   group, indicating enhanced metastatic potential (Supplementary Fig.
   [69]S1B).Further comparison between the HMGB1 group and the Vector
   group revealed 258 upregulated genes in the HMGB1 group, including
   osteoclast-associated genes Ctsk, Mmp14, and Mmp12, and 229
   downregulated genes, such as osteogenesis-related markers Col24a1 and
   Bglap (Fig. [70]1H). GSEA shows that pathways such as “MP_increased
   metastatic potential” (Supplementary Fig. [71]S1C) and the
   “HALLMARK_epithelial mesenchymal transition” (Supplementary Fig.
   [72]S1D) were significantly enriched in the HMGB1 group, reinforcing
   the pro-metastatic role of HMGB1. Additionally, GSEA analysis
   demonstrated significant enrichment of the “GOBP bone_resorption” and
   “GOBP bone_remodeling” pathways in the HMGB1 group (Supplementary Fig.
   [73]S1E), consistent with increased osteoclast activity.

HMGB1 drives HCC bone metastasis and induces skeletal complications

   To assess the functional role of HMGB1 in promoting HCC bone
   metastasis, stable cell lines with HMGB1 overexpression (Fig. [74]2A)
   and knockdown (Supplementary Fig. [75]S2A) were established in HuH-7
   and HCCLM3 cells. Intratibial injection of these cells into nude mice
   was performed, followed by weekly monitoring of bone metastasis
   progression using bioluminescence imaging (BLI).

Fig. 2. HMGB1 promotes HCC bone metastasis and skeletal complications in
vivo.

   [76]Fig. 2
   [77]Open in a new tab

   A Western blot analysis validating HMGB1 overexpression in HuH-7 and
   HCCLM3 cells. B Kaplan–Meier curve showing earlier onset of bone
   metastasis in nude mice injected with HMGB1-overexpressing HCCLM3
   cells. C Representative images from BLI, SPECT, and μCT demonstrating
   increased systemic bone metastasis and severe osteolytic lesions in the
   HMGB1-overexpression group compared to the control. D μCT
   quantification of bone parameters, showing increased Tb.sp and BS/BV,
   alongside decreased BV/TV and Tb.th in the HMGB1-overexpression group.
   E In vivo BLI at multiple time points, quantifying tumor burden based
   on signal intensity in mice. F Ex vivo BLI of hind limbs at the
   experimental endpoint, assessing tumor burden based on signal intensity
   in mice. G HE staining depicting osteolytic bone destruction and TRAP
   positive osteoclast areas following HMGB1 overexpression. H
   Quantitative analysis of TRAP-positive osteoclast areas, indicating
   increased osteoclast activation in the HMGB1-overexpression group.

   Mice injected with HMGB1-overexpressing cells showed significantly
   earlier onset of systemic bone metastasis (Fig. [78]2B), BLI,
   single-photon emission computed tomography (SPECT), and micro-computed
   tomography (μCT) analyses indicated an increased tumor burden and
   skeletal-related events (SREs) in the HMGB1 overexpression group
   compared to controls (Fig. [79]2C). Quantitative μCT analysis revealed
   severe osteolytic lesions and a higher incidence of skeletal-related
   events (SREs), characterized by increased trabecular separation (Tb.sp)
   and bone surface-to-volume ratio (BS/BV), alongside decreased bone
   volume fraction (BV/TV) and trabecular thickness (Tb.th). (Fig.
   [80]2D). In vivo BLI at multiple time points (Fig. [81]2E) and Ex vivo
   BLI of hind limbs demonstrated an elevated tumor burden in the HMGB1
   overexpression group (Fig. [82]2F). Histological analysis further
   confirmed larger osteolytic areas and a notable increase in
   tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts at the
   bone-tumor interface (Fig. [83]2G, H), suggesting enhanced osteoclast
   activation.

   In contrast, HMGB1 knockdown in HCCLM3 cells significantly suppressed
   bone metastasis progression. Mice injected with HMGB1-knockdown cells
   exhibited reduced tumor burden and fewer SREs compared to controls
   (Supplementary Fig. [84]S2B). μCT analysis indicated altered bone
   structural parameters (Fig. [85]S2C), BLI demonstrated HMGB1 knockdown
   reduce tumor burden (Supplementary Fig. [86]S2D, E), while TRAP
   staining revealed decreased osteoclast activity at the bone-tumor
   interface (Supplementary Fig. [87]S2F, G). Together, these results
   indicate that HMGB1 is associated with HCC bone metastasis and skeletal
   complications, linked to enhanced osteoclast activation and osteolytic
   bone destruction.

HMGB1-mediated osteoclastogenesis and tumor-macrophage interactions in
tumor-bone microenvironment

   To investigate the role of HMGB1 in osteoclastogenesis, conditioned
   medium (CM, [88]Supplementary File, Materials and Methods for further
   details) from HMGB1-overexpressing HCC cells was applied to RAW264.7
   cells. TRAP staining revealed a significant increase in the number of
   multinucleated TRAP^+ osteoclasts treated with CM from
   HMGB1-overexpressing cells compared to controls (Fig. [89]3A, B).
   Enzyme-linked immunosorbent assay (ELISA) confirmed higher HMGB1
   concentrations in the culture supernatants of HMGB1-overexpressing HCC
   cells (Fig. [90]3C). To isolate the direct effect of HMGB1 on
   osteoclast differentiation, RAW264.7 cells treated with recombinant
   human HMGB1 (rhHMGB1) exhibited enhanced osteoclast differentiation
   (Fig. [91]3D, E), consistent with the results from CM treatment of
   HMGB1-overexpressing HCC cells. A bone resorption assay demonstrated
   severe surface erosion and increased formation of resorption pits in
   bone slices treated with CM from HMGB1-overexpressing cells (Fig.
   [92]3F, G). Moreover, quantitative real-time PCR (qRT-PCR) analysis
   demonstrated that osteoclasts induced by CM from HMGB1-overexpressing
   HCC cells exhibited significantly increased expression levels of
   osteoclast differentiation and activation markers, including TRAP,
   RANKL, Atp6v0d2, Itgb3, Ctsk, and Dcstamp. In contrast, the expression
   of Opg, a soluble decoy receptor that inhibits osteoclastogenesis by
   neutralizing RANKL, was markedly decreased (Fig. [93]3H). These
   findings demonstrate that HMGB1 secreted by HCC cells, with elevated
   levels in CM, contributes to the differentiation of mononuclear
   progenitor osteoclasts into multinucleated mature osteoclasts.

Fig. 3. HMGB1 enhances osteoclastogenesis and facilitates interaction of HCC
cells and macrophages in the tumor-bone microenvironment.

   [94]Fig. 3
   [95]Open in a new tab

   A TRAP staining of RAW264.7 cells treated with CM from
   HMGB1-overexpressing HCC cells. B Quantification of TRAP-positive
   osteoclasts, confirming a significant increase in multinucleated cells
   treated with CM from HMGB1-overexpressing HCC cells. C ELISA analysis
   demonstrating elevated HMGB1 concentrations in the culture supernatants
   of HMGB1-overexpressing HCC cells. D TRAP staining of RAW264.7 cells
   treated with recombinant rhHMGB1. E Quantification of TRAP-positive
   osteoclasts, verifying enhanced osteoclast differentiation in RAW264.7
   cells treated with rhHMGB1. F The scanning electron microscopy image of
   bone resorption assay showing increased resorption pit formation on
   bone slices treated with CM from HMGB1-overexpressing HCC cells. G
   Quantification of resorption pits, confirming severe surface erosion on
   bone slices treated with CM from HMGB1-overexpressing HCC cells. H
   qRT-PCR analysis showing upregulation of osteoclast differentiation and
   activation markers (TRAP, RANKL, Atp6v0d2, Itgb3, Ctsk, Opg, Dcstamp)
   in RAW264.7 cells treated with CM from HMGB1-overexpressing HCC cells.
   I mIHC of HCC bone metastasis tissues from the animal model showing
   HMGB1 (red) co-localized with macrophages (F4/80, yellow) near HCC
   cells (GFP, green), with DAPI (blue) staining nuclei. J Magnified mIHC
   image highlighting the co-localization of HMGB1 with GFP-labeled HCC
   cells and F4/80-labeled macrophages in the tumor-bone interface. K
   Quantitative fluorescence intensity analysis along selected lanes in
   the mIHC overlay images, showing co-localization patterns. L
   Correlation analysis presenting high Pearson’s correlation coefficients
   that confirm significant co-localization of HMGB1 with GFP-labeled HCC
   cells (R = 0.38) and F4/80-labeled macrophages (R = 0.81).

   To investigate cellular interactions within the HCC bone
   microenvironment, multiplex immunohistochemistry (mIHC) was performed
   on tissue sections from the HCC bone metastasis mouse model. Figure
   [96]3I illustrates the spatial distribution of HMGB1 (red),
   F4/80-labeled macrophages (yellow), and GFP-labeled HCC cells (green)
   at the tumor-bone microenvironment. The magnified region of the mIHC
   images highlighted the complex tumor-bone microenvironment of HCC bone
   metastasis, showing co-localization of HMGB1 with GFP-labeled HCC cells
   infiltrating the bone marrow cavity and with F4/80-labeled macrophages
   in certain areas (Fig. [97]3J). Quantitative fluorescence intensity
   analysis along selected lanes in the overlay images revealed
   co-localization patterns (Fig. [98]3K). Correlation analysis with the
   Coloc2 plugin confirmed significant co-localization of HMGB1 with
   GFP-labeled HCC cells and F4/80-labeled macrophages, as evidenced by
   high Pearson’s correlation coefficients (Fig. [99]3L).

HMGB1-LCN2 mediates crosstalk among HCC cells, macrophages, and osteoclasts
in the tumor-bone microenvironment

   To explore the role of HMGB1 in tumor-bone microenvironmental
   interactions, RNA-seq was performed on RAW264.7-derived osteoclasts
   under two conditions: a control group (osteoclasts induced under
   standard conditions) and an rhHMGB1-treated group (osteoclasts induced
   with 100 ng/mL rhHMGB1 in addition to standard conditions).
   Differential expression analysis identified 2009 upregulated and 2112
   downregulated genes in the rhHMGB1-treated group (Fig. [100]4A).
   Upregulated genes included the secretory protein Lcn2, the recognized
   macrophage activation marker Tlr4, and osteoclast-related genes such as
   Ctsk, Acp5, and Atp6v0d2, whereas downregulated genes mainly involved
   immune and macrophage markers, including Adgre1, Adgre5, and Cd24a
   (Fig. [101]4B). KEGG pathway enrichment analysis revealed significant
   enrichment of the osteoclast differentiation pathway (Fig. [102]4C).
   GSEA further confirmed enrichment of the
   “GOBP_osteoclast_differentiation” gene set in the rhHMGB1 group
   (Supplementary Fig. [103]S3A), alongside the
   “BROWN_myeloid_cell_development_up” gene set (Supplementary Fig.
   [104]S3B). Conversely, the “BROWN_myeloid_cell_development_dn” gene set
   was enriched in the control group (Supplementary Fig. [105]S3C).
   Additionally, Immunofluorescence staining demonstrated increased
   expression of TLR4 in RAW264.7 cells following treatment with rhHMGB1
   (Supplementary Fig. [106]S4A), supporting the presence of an HMGB1-TLR4
   molecular interaction in macrophages within the tumor-bone
   microenvironment.

Fig. 4. HMGB1 promotes osteoclast differentiation and LCN2 secretion,
enhancing tumor invasion.

   [107]Fig. 4
   [108]Open in a new tab

   A Volcano plot showing significantly upregulated (red) and
   downregulated (blue) genes in RAW264.7-derived osteoclasts treated with
   rhHMGB1 compared to controls. B Heatmap highlighting representative
   genes altered by rhHMGB1 treatment, including upregulation of Lcn2,
   Tlr4, and osteoclast-related genes and downregulation of
   immune/macrophage markers. C KEGG enrichment analysis plot illustrating
   the significant enrichment of the osteoclast differentiation pathway
   among differentially expressed genes. D ELISA indicating elevated LCN2
   secretion by osteoclasts treated with CM from HMGB1-overexpressing HCC
   cells. E Transwell invasion assay images demonstrating enhanced
   invasive capacity of HMGB1-overexpressing HCC cells treated with OC CM
   or rhLCN2. F Quantification of Transwell invasion assay, confirming the
   increased invasive ability of HMGB1-overexpressing HCC cells. G Wound
   healing assay images showing increased migratory capacity of
   HMGB1-overexpressing HCC cells treated with OC CM or rhLCN2. H
   Quantification of wound healing assay, verifying the enhanced migration
   of HMGB1-overexpressing HCC cells.

   Then, ELISA demonstrated significantly elevated LCN2 levels in the
   supernatants of osteoclasts treated with CM from HMGB1-overexpressing
   HCC cells (Fig. [109]4D). To evaluate the impact of osteoclast-derived
   LCN2 on tumor progression, we co-cultured HCC cells with
   osteoclast-conditioned medium (OC CM, [110]Supplementary File,
   Materials and Methods for further details) or recombinant LCN2 (rhLCN2)
   protein. Transwell invasion assays showed that both OC CM and rhLCN2
   significantly enhanced the invasive ability of HMGB1-overexpressing HCC
   cells (Fig. [111]4E, F). Similarly, wound healing assays showed
   increased migration of HMGB1-high HCC cells upon treatment with OC CM
   or rhLCN2 (Fig. [112]4G, H). Notably, immunofluorescence analysis
   confirmed a dose-dependent increase in TLR4 expression in RAW264.7
   cells treated with rhLCN2 (Supplementary Fig. [113]S4B).
   Co-immunoprecipitation (Co-IP) experiments further confirmed the
   interaction between HMGB1 and TLR4, with rhLCN2 enhancing this
   interaction in a concentration-dependent manner (Supplementary Fig.
   [114]S4C, D).

   In summary, these findings establish a HMGB1-LCN2 feedback loop within
   the tumor-bone microenvironment: HMGB1 derived from HCC cells promotes
   osteoclast differentiation and LCN2 secretion, while
   osteoclast-secreted LCN2 enhances HCC cell invasion and migration and
   strengthens the HMGB1-TLR4 interaction in macrophages. This crosstalk
   reinforces osteoclastogenesis and drives tumor progression.

Osteoclast-derived LCN2 activates the JAK1/STAT3 pathway in
HMGB1-overexpressing HCC cells

   Given the established role of HMGB1 in activating the JAK1/STAT3
   signaling pathway, which contributes to disease progression [[115]26].
   We explored whether osteoclast-derived LCN2 influences this pathway in
   HMGB1-overexpressing HCC cells. Incubation of HCCLM3 cells with OC CM
   significantly increased JAK1 and STAT3 phosphorylation in
   HMGB1-overexpressing cells, but not in controls. Knockdown of HMGB1
   abolished the OC CM-induced activation of JAK1 and STAT3 (Fig.
   [116]5A). Similar results were observed in HuH-7 cells (Fig. [117]5B).
   To assess the specific contribution of LCN2, treatment with rhLCN2
   similarly increased JAK1/STAT3 phosphorylation in both HCCLM3 and HuH-7
   cells overexpressing HMGB1, while HMGB1 knockdown diminished this
   effect (Fig. [118]5C, D). Consistent with these findings,
   immunohistochemical staining of clinical HCC tissues revealed a
   positive correlation between HMGB1 expression and STAT3 phosphorylation
   (Fig. [119]5E, F). Together, these results demonstrate that
   osteoclast-derived LCN2 activates the JAK1/STAT3 signaling pathway
   specifically in HMGB1-overexpressing HCC cells.

Fig. 5. Osteoclast-derived LCN2 activates the JAK1/STAT3 signaling pathway in
HMGB1-overexpressing HCC cells.

   [120]Fig. 5
   [121]Open in a new tab

   Western blot analysis of JAK1, STAT3, and their phosphorylated forms
   (p-JAK1 and p-STAT3) in HCCLM3 (A) and HuH-7 (B) cells treated with OC
   CM for 24 h. JAK1/STAT3 activation was observed in HMGB1-overexpressing
   cells but was abolished upon HMGB1 knockdown. Western blot analysis
   showing that rhLCN2 treatment (75 ng/mL for 15 min) promotes JAK1 and
   STAT3 phosphorylation in HMGB1-overexpressing HCCLM3 (C) and HuH-7 (D)
   cells, while HMGB1 knockdown diminishes this effect. E
   Immunohistochemical staining of HMGB1 and p-STAT3 in clinical HCC
   tissues from two representative patients. where HMGB1-high samples
   exhibit increased STAT3 phosphorylation. F Scatter plot illustrating a
   positive correlation between HMGB1 expression and p-STAT3 levels in
   clinical HCC samples.

STAT3 is essential for HMGB1-mediated bone metastasis

   To assess whether STAT3 functions downstream of HMGB1 in promoting bone
   metastasis, STAT3 was knocked down in both control and
   HMGB1-overexpressing HCCLM3 cells, as validated by Western blot
   analysis (Fig. [122]6A). Histological analysis further revealed that
   STAT3 knockdown markedly decreased tumor infiltration and osteolytic
   destruction at the tumor-bone interface (Fig. [123]6B). In mouse bone
   metastasis model, STAT3 knockdown significantly decreased
   HMGB1-enhanced tumor burden and osteolytic lesions. This was evidenced
   by BLI and μCT in Fig. [124]6C, μCT indicating decreased osteolytic
   lesion volume, increased BV/TV and Tb.th, and reduced BS/BV and Tb.Sp
   (Fig. [125]6D). In vivo BLI confirmed a significant tumor burden
   reduction (Fig. [126]6E), while ex vivo BLI showed decreased tumor
   signal intensity (Fig. [127]6F). Mechanistically, STAT3 knockdown
   inhibited the phosphorylation of STAT3 induced by OC CM and rhLCN2 in
   tumor cells (Fig. [128]6G).

Fig. 6. STAT3 Knockdown Inhibits HMGB1-Mediated Bone Metastasis.

   [129]Fig. 6
   [130]Open in a new tab

   A Western blot analysis confirming effective STAT3 knockdown in control
   and HMGB1-overexpressing HCCLM3 cells. B HE staining of bone sections
   showing reduced tumor infiltration and osteolytic destruction at the
   bone-tumor interface following STAT3 knockdown. C BLI and μCT imaging
   showing that STAT3 knockdown reduces HMGB1-enhanced bone metastasis and
   osteolytic lesions. D Quantification of μCT revealing reduced
   osteolytic lesions in mice with STAT3-knockdown HMGB1-overexpressing
   HCCLM3 cells. E In vivo BLI quantifying reduced tumor burden in mice
   with STAT3-knockdown HMGB1-overexpressing cells. F Ex vivo BLI of hind
   limbs assessing decreased tumor burden in the STAT3-knockdown group. G
   Western blot analysis showing that STAT3 knockdown abolishes the
   phosphorylation of STAT3 induced by OC CM and rhLCN2 in
   HMGB1-overexpressing HCCLM3 cells.

   To evaluate the therapeutic potential of STAT3 inhibition, mice
   inoculated with HMGB1-overexpressing HCCLM3 cells were treated with
   intraperitoneal injections of the STAT3 inhibitor Napabucasin (Napa).
   Napa effectively suppressed tumor growth and osteolysis, as evidenced
   by reduced tumor burden and SREs (Supplementary Fig. [131]S5A, B). BLI
   results confirmed decreased tumor burden (Supplementary Fig. [132]S5C,
   D), and Napa treatment inhibited STAT3 activation induced by OC CM and
   rhLCN2 in HMGB1-overexpressing cells (Supplementary Fig. [133]S5E).
   These results collectively demonstrate that STAT3 plays a pivotal
   regulatory role in the HMGB1-LCN2 axis, contributing to tumor
   progression and osteolysis, and highlight its potential as a
   therapeutic target in HCC bone metastasis.

HMGB1 antagonist dipotassium glycyrrhizinate inhibits bone metastasis and
JAK1/STAT3 activation

   To evaluate the therapeutic efficacy of targeting HMGB1 in bone
   metastasis, the HMGB1 antagonist Dipotassium Glycyrrhizinate (Dg) was
   used. Mice were intratibial injected with HCCLM3 cells overexpressing
   HMGB1 or Vctor controls, and Dg treatment (150 mg/kg,
   intraperitoneally) was initiated one week post-injection and
   administered every other day.

   BLI and μCT revealed that Dg treatment significantly reduced tumor
   burden and SREs in mice with HMGB1-overexpressing HCCLM3 cells compared
   to the IgG control group (Fig. [134]7A). Quantitative analysis of μCT
   revealed that Dg markedly inhibited bone destruction, as evidenced by
   decreased Tb.sp and BS/BV, and increased BV/TV and Tb.th (Fig.
   [135]7B). BLI results further confirmed the reduction in tumor burden,
   with ex vivo BLI showing a clear decrease in metastatic spread across
   hind limb tissues (Fig. [136]7C), while multi-time point in vivo BLI
   depicted a consistent decline in tumor growth (Fig. [137]7D).

Fig. 7. Dg inhibits HMGB1-driven bone metastasis and suppresses JAK1/STAT3
activation.

   [138]Fig. 7
   [139]Open in a new tab

   A Representative BLI and μCT showing reduced tumor burden and inhibited
   bone destruction in mice injected intratibially with
   HMGB1-overexpressing HCCLM3 cells following Dg treatment. B
   Quantitative analysis of μCT bone parameters, demonstrating the
   inhibitory effects of Dg on HMGB1-driven bone destruction. C Ex vivo
   BLI quantifying a significant reduction in tumor burden with tumor
   signal intensity decreasing. D Multi-time point in vivo BLI showing a
   progressive reduction in tumor burden, with photon flux intensity
   declining. E, F Western blot analysis showing that Dg treatment
   (100 μM, 24 h) suppresses JAK1 and STAT3 phosphorylation in
   HMGB1-overexpressing HCCLM3 cells stimulated with OC CM (24 h) or
   rhLCN2 (75 ng/mL for 15 min). G Schematic model of this study. Drawn
   and adapted by Figdraw ([140]https://www.figdraw.com). HMGB1 from HCC
   cells induces osteoclast formation from macrophages, while LCN2 from
   osteoclasts fuels the loop via HMGB1-TLR4 interactions and enhances HCC
   malignancy via HMGB1-TLR4 and JAK1-STAT3 signaling, with therapeutic
   implications for advanced HCC.

   Mechanistically, Western blot analysis showed that Dg treatment
   effectively inhibited JAK1 and STAT3 phosphorylation induced by OC CM
   in HMGB1-overexpressing HCCLM3 cells (Fig. [141]7E). Similar inhibition
   of STAT3 activation was observed with rhLCN2 treatment (Fig. [142]7F).
   Together, these findings confirm that the HMGB1 antagonist Dg
   effectively suppresses HMGB1-driven bone metastasis by inhibiting the
   JAK1/STAT3 signaling pathway, highlighting its therapeutic potential in
   HCC bone metastasis.

   Collectively, this study elucidates a novel tumor-osteoclast crosstalk
   mechanism within the tumor-bone microenvironment of HCC bone
   metastasis, introducing a novel HMGB1-LCN2-JAK1-STAT3 axis. Here, HMGB1
   from metastatic HCC cells triggers a vicious cycle by driving
   macrophage-to-osteoclast transdifferentiation, amplifying osteolytic
   spread, while osteoclast-derived LCN2 fuels this loop via HMGB1-TLR4
   interactions and JAK1-STAT3 activation in HCC cells, intensifying
   malignancy. This comprehensive experimental evidence identifies a novel
   therapeutic target and indicates potential applications for improving
   the management of advanced HCC (Fig. [143]7G).

Discussion

   Bone metastasis in HCC presents a critical clinical challenge, marked
   by severe skeletal complications and its role as a reservoir for
   disseminating cancer cells, fueling systemic disease progression
   [[144]27]. Unlike tumor self-seeding or cross-seeding, the bone
   microenvironment provides a unique niche that sustains secondary
   metastases by enhancing tumor cell survival and proliferation [[145]8,
   [146]27]. The hallmark of osteolytic phenotype, primarily driven by
   excessive osteoclast activation, exacerbates bone destruction while
   creating conditions that further support tumor progression [[147]12].
   However, the molecular and cellular interactions within the bone-tumor
   microenvironment in HCC remain poorly understood. This study addresses
   this gap by unveiling the central role of HMGB1 in HCC bone metastasis.
   Previously, HMGB1 has been implicated in HCC progression through
   dendritic cell-mediated CD8^+ T cell responses [[148]20], metastasis
   via the KLF7-TLR4/PTK2 signaling axis under inflammation [[149]28], and
   enhanced migration and invasion through mitochondrial transfer under
   hypoxic conditions [[150]21]. Our research distinctively reveals
   HMGB1’s novel function in orchestrating tumor-osteoclast crosstalk,
   identifying the HMGB1/LCN2/JAK1/STAT3 axis as a critical driver of
   osteolytic bone metastasis.

   This study reveals a novel mechanism by which HMGB1 drives HCC bone
   metastasis through orchestrating a dynamic tumor-osteoclast crosstalk.
   Traditionally, the bone microenvironment has been regarded as a passive
   site of tumor colonization, where osteoclasts mediate bone resorption
   and release growth factors to sustain tumor progression [[151]9]. Our
   findings indicate that HMGB1 contributes to osteoclast activity,
   leading to LCN2 secretion, which is required for JAK1/STAT3 pathway
   activation in HCC cells. This establishes a self-sustaining feedback
   loop that amplifies osteolysis and tumor dissemination, highlighting a
   unique dependency on LCN2 that distinguishes the HCC bone-tumor
   microenvironment from other cancer contexts where STAT3 activation
   relies on diverse mediators. The critical role of LCN2 is substantiated
   by its established influence on STAT3 signaling in fibrotic [[152]29]
   and inflammatory microenvironments [[153]30], providing a strong
   scientific basis for our mechanistic insights. The JAK1/STAT3 pathway,
   well-established for its oncogenic roles in cellular proliferation,
   immune evasion, and metastatic dissemination [[154]11, [155]26], is
   unveiled in this study as a pivotal mediator within the
   osteoclast-tumor signaling axis. LCN2, previously implicated in immune
   modulation and metastasis in other cancers [[156]31, [157]32], is
   identified as a key effector linking osteoclast activity to tumor cell
   aggressiveness in HCC. By integrating osteoclast-driven bone resorption
   with tumor cell signaling, HMGB1 modifies the bone microenvironment
   into a dynamic and active participant in metastatic progression. These
   findings illuminate the potential of targeting the
   HMGB1/LCN2/JAK1/STAT3 signaling axis as a therapeutic approach, with
   the capacity to disrupt the self-reinforcing pathological feedback loop
   that exacerbates tumor growth and osteolytic destruction in HCC bone
   metastasis.

   Despite the widespread use of bisphosphonates and denosumab for
   managing bone metastases, their therapeutic efficacy remains largely
   restricted to mitigating SREs without effectively halting the systemic
   progression of metastatic cancer [[158]33–[159]35]. Recent evidence
   suggests that combination therapies targeting the crosstalk between
   tumor cells and the bone microenvironment may provide superior outcomes
   by disrupting the feedback loops that sustain tumor growth and bone
   resorption [[160]36]. Previous studies have demonstrated that
   HMGB1-targeted approaches, including the application of Dg, can
   attenuate inflammation and tumor progression [[161]37, [162]38], their
   potential in bone metastasis has remained unexplored until now.
   Additionally, STAT3 inhibitors like Napa have exhibited promising
   results in dismantling tumor-driven immunosuppressive environments and
   amplifying chemotherapy efficacy [[163]39–[164]41]. Our study
   integrates these insights, proposing that targeting the
   HMGB1/LCN2/JAK1/STAT3 axis can effectively disrupt tumor-osteoclast
   crosstalk, suppress osteolytic bone resorption, and potentially curtail
   systemic tumor dissemination. This therapeutic approach not only holds
   promise for improving skeletal outcomes but also offers a systemic
   benefit by addressing the molecular drivers of metastasis. In this
   study, Dg significantly reduced osteoclast activation and tumor burden
   in vivo, offering a potential upstream intervention to disrupt the
   tumor-osteoclast interaction. By blocking HMGB1 release, Dg prevents
   the initiation of the molecular cascade that sustains the osteolytic
   microenvironment. Additionally, Napa effectively inhibited JAK1/STAT3
   signaling in HMGB1-overexpressing tumor cells exposed to OC-CM or
   rhLCN2. Future investigations into the synergistic potential of
   combining HMGB1 and STAT3 inhibitors with current anti-osteoclast
   therapies could offer a transformative strategy for managing HCC bone
   metastasis, reducing both skeletal complications and systemic tumor
   progression.

   This study advances the understanding of tumor-osteoclast crosstalk in
   HCC bone metastasis, yet several dimensions warrant further exploration
   to refine these insights. The upstream regulation of HMGB1
   secretion—whether governed by intrinsic tumor factors,
   microenvironmental signals, or systemic influences—remains an
   unresolved question necessitating detailed investigation. Moreover, the
   applicability of this axis beyond HCC, particularly in osteolytic
   metastases such as those in breast or prostate cancer, requires
   comparative studies to assess its broader relevance. Furthermore, the
   prognostic value of serum HMGB1, as suggested by its association with
   adverse outcomes in advanced HCC post-treatment [[165]42], merits
   further study as a non-invasive biomarker. Future research will conduct
   inhibition experiments, such as using RANKL inhibitors or
   osteoclast-specific genetic knockdowns, to elucidate the causal
   interactions between HMGB1-driven osteoclast activation and HCC bone
   metastasis, thereby refining the theoretical framework of the axis.
   Additionally, preclinical studies employing nanotechnology-based
   delivery systems could enhance therapeutic precision within the
   bone-tumor microenvironment, reducing systemic effects and improving
   clinical outcomes. These efforts promise to deepen the understanding
   and management of HCC bone metastasis.

   In conclusion, this study establishes a critical molecular framework by
   identifying the HMGB1/LCN2/JAK1/STAT3 axis as a central mechanism in
   the development of osteolytic bone metastasis in HCC. Our comprehensive
   analysis shows that HMGB1 orchestrates osteoclast activation, induces
   LCN2 signaling, and activates the JAK1/STAT3 pathway, creating a
   feedback loop that enhances bone destruction and tumor dissemination
   within the bone-tumor microenvironment. This finding highlights a
   promising therapeutic target and contributes to the understanding of
   HCC metastasis, providing a critical step toward translating molecular
   discoveries into clinically effective strategies for managing bone
   metastasis in HCC.

Supplementary information

   [166]Supplementary Figure legend^ (14.1KB, docx)
   [167]Supplementary Materials and Methods^ (37.2KB, docx)
   [168]Full uncropped Gels and Blots^ (490.2KB, pdf)
   [169]Figure S1^ (1.4MB, png)
   [170]Figure S2^ (5.3MB, png)
   [171]Figure S3^ (652.8KB, png)
   [172]Figure S4^ (7.7MB, png)
   [173]Figure S5^ (2.7MB, png)
   [174]Table S1^ (8.9KB, xlsx)
   [175]Table S2^ (9.3KB, xlsx)

Author contributions

   YC and ZX conceived the study. YC, DX, and YJ performed the
   experiments. YC and JM. performed bioinformatics analysis. JM performed
   clinical data collection. DX collected clinical samples. YC wrote the
   manuscript. ZX supervised the study. All authors read and approved the
   final manuscript.

Funding

   This work was supported by the National Natural Science Foundation of
   China (Grant No. 82160591), Key Specialty Construction Project of
   Shanghai Pudong New Area Health Commission (Grant No. PWZzk2022–02),
   Outstanding Leaders Training Program of Pudong Health Bureau of
   Shanghai (Grant No. [176]PWR12023–02), Shanghai Science and Technology
   Innovation Action Plan (Grant No. 23Y11909000), and Key Project of
   Clinical Research of Shanghai East Hospital, Tongji University (Grant
   No. DFLC2022012).

Data availability

   Bulk RNA-seq data generated in this study have been deposited in the
   NCBI’s Sequence Read Archive (SRA) under BioProject accession number
   PRJNA1208261
   ([177]https://dataview.ncbi.nlm.nih.gov/object/PRJNA1208261?reviewer=br
   qfoq1nq6d4uhqlden4o6ane7). Publicly available scRNA-seq datasets
   included in the analysis are from the GEO database, all data are
   accessible through their respective GEO entries.

Competing interests

   The authors declare no competing interests.

Ethics approval and consent to participate

   This study was approved by the Ethics Committee of Shanghai East
   Hospital (No.2022250). All procedures involving human participants
   adhered to the principles of the Declaration of Helsinki and informed
   consent was obtained from all patients, the animal procedures in this
   study were approved by the Ethical Committee for Animal Research of
   Tongji University (Shanghai, China).

Footnotes

   Edited by Professor Anastasis Stephanou

   Publisher’s note Springer Nature remains neutral with regard to
   jurisdictional claims in published maps and institutional affiliations.

Supplementary information

   The online version contains supplementary material available at
   10.1038/s41419-025-08037-6.

References