Development and Validation of a HAUS-Related Prognostic Signature in Liver Hepatocellular Carcinoma: A Multi-Omics Integration from Pan-Cancer Analysis to Single-Cell Resolution

Abstract

Background

The HAUS family proteins (HAUS1–HAUS8) are essential for mitotic spindle microtubule nucleation. Although HAUS dysregulation has been linked to tumor progression, whether these oncogenic functions are conserved or tissue-specific remains unclear. Therefore, a pan-cancer analysis is needed to identify universal HAUS drivers and context-dependent members for precision oncology.

Methods

We systematically evaluated the expression and prognostic significance of all eight HAUS genes across 33 cancer types using TCGA data. Focusing on liver hepatocellular carcinoma (LIHC), we identified molecular subtypes based on HAUS co-expression patterns and characterized their associations with clinicopathological features, the tumor microenvironment (TME), immune checkpoint expression, and therapeutic response. A HAUS-related prognostic signature was developed using LASSO and Cox regression analyses and validated in independent ICGC cohorts. Regulatory relationships among HAUS members and key signature genes were experimentally validated using immunohistochemistry and Western blotting.

Results

Most HAUS members were overexpressed across multiple cancers and correlated with poor clinical outcomes. In LIHC, two HAUS-based subtypes differed significantly in survival, clinicopathological profiles, immune features, mutational burden, stemness indices, and predicted therapeutic response. A prognostic signature comprising DTYMK and SPP1 effectively stratified LIHC patients into distinct risk groups, with the high-risk group showing significantly worse survival. A nomogram integrating the HAUS-related risk score with clinicopathological variables demonstrated strong predictive performance. Recombinant osteopontin induced HAUS1 and DTYMK expression in LIHC cell lines, supporting a functional SPP1–HAUS1/DTYMK axis.

Conclusion

This study establishes the broad oncogenic relevance of HAUS genes across cancers and demonstrates that a HAUS-based signature enables prognostic stratification in LIHC.

Introduction

The American Cancer Society’s 2025 report shows a continuous decline in cancer-related deaths through 2022, preventing nearly 4.5 million deaths since 1991. This progress is mainly due to decreased tobacco use, earlier diagnosis of some cancers, and improved therapies. However, ongoing racial disparities and increasing cancer rates among middle-aged and younger populations pose challenges to future improvements.¹ Therefore, sustained investment in translational and clinical research is crucial to advance cancer prevention and treatment.

Precision oncology provides the most effective treatments tailored to individual cancer patients through comprehensive genomic and molecular analyses. Discoveries of key driver mutations, like BRCA1/2, have advanced precision diagnostics and treatment.² The clinical application of immune checkpoint inhibitors (ICIs) targeting CTLA-4, PD-1, and PD-L1 has revolutionized cancer treatment, significantly improving outcomes for patients with advanced-stage cancer, as recognized by the 2018 Nobel Prize in Physiology or Medicine.^3,4 Together, these advances highlight the importance of identifying additional oncogenic drivers and signaling pathways that can be therapeutically exploited across cancer types.

The HAUS complex, the human homolog of the Drosophila Augmin complex, is an octameric assembly (HAUS1–HAUS8) essential for microtubule (MT) nucleation and mitotic spindle formation.^5,6 Structurally, it comprises two tetrameric subcomplexes: a microtubule-binding module (HAUS2/6/7/8) and a γ-tubulin ring complex (γ-TuRC)-recruiting module (HAUS1/3/4/5). Together, these modules promote MT-dependent MT nucleation during mitosis.^7–9 Beyond this canonical function, individual HAUS members also exert context-specific, non-mitotic roles. For example, HAUS8 may be implicated in the ubiquitination of RIG-I, VISA, and TBK1, thereby activating interferon responses through the RLR–VISA antiviral signaling pathway.¹⁰ In contrast, HAUS7 interacts with DOCK3 in retinal ganglion cells to promote axon regeneration.¹¹

Emerging evidence also implicates the HAUS family in tumor progression across multiple cancers. In oral squamous cell carcinoma (OSCC), HAUS6, HAUS8, and HAUS3 are required for centrosome clustering and bipolar spindle formation in cells with supernumerary centrosomes, thereby preventing multipolar mitoses and apoptosis driven by chromosomal instability.¹² Outside mitosis, HAUS genes appear to display diverse oncogenic roles. For example, HAUS1 promotes LIHC progression through CDK4 activation,¹³ drives colorectal cancer tumorigenesis via the HAUS1–EZH2–E2F1 axis,¹⁴ and is recurrently fused to NPM1 in acute myeloid leukemia (AML).¹⁵ Similarly, HAUS5 contributes to tumor progression in glioblastoma,¹⁶ breast cancer,¹⁷ and LIHC.¹⁸ Collectively, these findings suggest that the HAUS family has multifaceted roles in cancer biology and may represent important candidates for precision oncology. However, three key questions remain unresolved: whether HAUS-associated oncogenic functions are conserved or tissue-specific; how HAUS expression relates to the TME, treatment response, and patient outcomes across cancers; and which HAUS members are most promising as prognostic biomarkers or therapeutic targets.

A pan-cancer analytical framework may help address these gaps by integrating multi-omics datasets to distinguish pan-cancer relevant HAUS members from context-dependent ones, identify immunotherapy-related biomarkers, and prioritize HAUS members for experimental validation. This approach aligns with precision oncology efforts to identify convergent molecular hubs across cancers.

We hypothesized that HAUS members exhibit cancer-specific oncogenic potential and co-expression patterns that may serve as prognostic and immunotherapy-related biomarkers in specific cancers. To test this hypothesis, we conducted a pan-cancer analysis of all eight HAUS genes. Focusing on LIHC, we identified HAUS-based molecular subtypes based on co-expression patterns. Here, we characterized HAUS expression and prognostic significance across 33 cancer types, defined HAUS subtypes in LIHC, and evaluated their associations with the TME, immune checkpoint expression, and therapeutic response. We also developed and validated a HAUS-related prognostic model for LIHC. Finally, we experimentally verified the expression and regulatory relationships of HAUS-associated genes, including HAUS1, DTYMK, and SPP1, in LIHC. Collectively, this study provides a comprehensive pan-cancer landscape of the HAUS family and establishes a foundation for HAUS-based patient stratification and personalized therapeutic strategies in liver cancer.

Material and Methods

Ethics Statement

This study was approved by the Second Affiliated Hospital of Guangzhou Medical University Clinical Research and Application Institutional Review Board (approval No. LYZX-2026-083-01). Informed consent was waived in accordance with Item 1 of Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects, dated February 18, 2023, China. This provision permits waiver of ethical review for research using legally obtained, publicly available biological samples or data, provided the research poses no risk to subjects, involves no sensitive personal information or commercial interests, and does not require interference with public behavior. Additionally, the liver cancer tissue microarray (array No. HLivH150CS03) was purchased from Shanghai Outdo Biotech Company (Shanghai, China), and the Shanghai Outdo Biotech Company Ethics Committee granted ethical approval for use of the samples (approval No. SHYJS-CP-1507007).

Data Source and Preprocessing

RNAseq-HTSeq-FPKM data, along with corresponding clinical and single-nucleotide variation (SNV) data from 33 cancer types, were sourced from the UCSC Xena platform (xenabrowser.net). Single-cell sequencing (ScRNA-seq) data, including 10 primary LIHC tumors (GSM4505944, GSM4505945, GSM4505947, GSM4505949, GSM4505951, GSM4505953, GSM4505956, GSM4505959, GSM4505961, GSM4505964) and 8 non-tumor tissues (GSM4505946, GSM4505948, GSM4505950, GSM4505952, GSM4505954, GSM4505957, GSM4505960, GSM4505963), were downloaded from the Gene Expression Omnibus (GEO) database (GSE149614, available at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149614).¹⁹ RNA-Seq data of hepatocellular carcinoma (LIHC) and adjacent normal liver samples from TCGA (https://portal.gdc.cancer.gov) and ICGC (https://dcc.icgc.org) were also downloaded.

For preprocessing, samples with missing survival information or follow-up time less than 30 days were excluded to avoid immortal time bias. To normalize the transcriptomic data, a log2(Transcripts Per Million [TPM] + 1) transformation was applied. For TCGA data, TPM values were calculated from FPKM using the formula: TPM = FPKM × 10⁶/Σ(FPKM).

Differential Analysis and Prognostic Significance of HAUS Family Genes

The differential expression of HAUS family genes between normal and tumor tissues was assessed using the Wilcoxon rank-sum test. Kaplan-Meier survival analysis and univariate Cox regression were performed using the survminer and survival R packages to evaluate the impact of HAUS family genes on overall survival, progression-free survival, and disease-free survival. The Log rank test was used to assess statistical significance.

Consensus Clustering

To identify HAUS-related molecular subtypes, unsupervised consensus clustering was performed using the ConsensusClusterPlus R package. This method used 50 iterations (reps), 80% sample resampling (pItem = 0.8), and 100% gene resampling (pFeature = 1) to identify clusters from 2 to 9 (k = 2:9) with the k-means algorithm (clusterAlg = “km”). The distance metric was set to Euclidean (distance = “euclidean”). The optimal number of clusters was determined by examining the cumulative distribution function (CDF) of the consensus matrix, the relative change in area under the CDF curve (delta area), and the consensus matrix heatmap at each k.

Analysis of HAUS Molecular Subgroups and Immune Landscape in LIHC

To investigate the relationship between the HAUS molecular subgroups and tumor-infiltrating immune cells in LIHC, single-sample gene set enrichment analysis (ssGSEA) scores were used to assess the enrichment levels of immune signatures within tumor tissues. Additionally, to explore the relationship between the HAUS molecular subgroups and the expression levels of immune checkpoint inhibitor (ICI)-related genes, such as CD274, PDCD1, CTLA4, TIGIT, and LAG3, we used the ggpubr package in R to generate boxplots. Furthermore, the responsiveness to PD-1/PD-L1 or CTLA-4 inhibitors was predicted using the Immunophenotype Score (IPS) algorithm, which integrates immune features such as MHC molecule expression and effector cell abundance. Lastly, tumor immune evasion was quantified using the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm, in which higher TIDE scores indicate greater resistance to immune checkpoint blockade (ICB).

Function Enrichment Analysis

To clarify the biological implications of different HAUS molecular patterns in LIHC, we performed Gene Set Variation Analysis (GSVA) using the GSVA R package with default parameters (min.sz = 10, max.sz = 500, verbose = TRUE, parallel.sz = 1).²⁰ Transcriptomic profiles of each patient from the TCGA dataset were scored against the Hallmark (h.all.v7.5.1.symbols.gmt) and KEGG (c2.cp.kegg.v7.4.symbols.gmt) gene sets. Additionally, differential pathway analysis between the HAUS-high and HAUS-low groups was conducted using the R package limma, and significant pathways (adjusted p-value < 0.05) were visualized using the pheatmap and ggpubr packages. Furthermore, mutation counts and profiles for the two HAUS molecular subgroups were estimated using the maftools, reshape2, and ggpubr packages.

Assessment of Stem-Cell Characteristics and Drug Sensitivity

To investigate the association between HAUS molecular subtypes and cancer stem cell (CSC)-like properties, stemness scores for LIHC samples were calculated using a previously validated mRNA-based stemness index (mRNAsi).

Chemotherapeutic response predictions were conducted using the pRRophetic R package, which leverages Ridge regression models trained on pharmacogenomic data from the Cancer Genome Project (CGP) cell lines. These models were applied to LIHC patient-specific gene expression profiles to estimate half-maximal inhibitory concentrations (IC50) for various chemotherapeutic and targeted agents. Boxplots illustrated the IC50 distributions across the two HAUS molecular subgroups, with lower predicted IC50 values indicating higher sensitivity to the therapies.

Single-Cell RNA Sequencing Analysis

Single-cell RNA-seq analyses were performed in Seurat v3.0. Quality control was applied before downstream analysis. Cells with fewer than 200 detected genes, more than 5,000 detected genes, or mitochondrial gene content greater than 5% were excluded. Mitochondrial gene percentages were calculated with PercentageFeatureSet. The remaining cells were normalized with NormalizeData, and the 4,000 most variable genes were identified with FindVariableFeatures. The datasets were integrated with FindIntegrationAnchors and IntegrateData to reduce batch effects. Principal component–based neighbor graph construction and clustering were performed with FindNeighbors and FindClusters, using a resolution of 0.4. Clusters were visualized with UMAP and annotated using canonical marker genes.

Construction and Validation of the Prognostic Signature for HAUS Molecular Patterns-Related Genes

To develop a prognostic model based on HAUS molecular pattern-related genes, least absolute shrinkage and selection operator (LASSO) regression analysis was performed in R using the glmnet package to identify the most significant survival-associated HAUS molecular pattern-related genes in LIHC. Subsequently, multivariate Cox regression analysis was employed to calculate the coefficients (β) for each gene ultimately included in the prognostic model. The risk score for each patient in the TCGA cohort was computed using the formula: risk score = (β1 × expression of mRNA₁) + (β2 × expression of mRNA₂).

Moreover, using the median risk score, patients in the TCGA-LIHC cohort were stratified into low- and high-risk clusters. To assess survival differences between the two clusters, Kaplan-Meier survival curves were generated and Log rank tests were performed using the survival and survminer packages. The predictive accuracy of the model for 1-, 3-, and 5-year overall survival was further evaluated using time-dependent receiver operating characteristic (ROC) curves and area under the curve (AUC) values calculated with the “timeROC” package. To validate the robustness and effectiveness of the prognostic model, another independent LIHC cohort, ICGC-LIRI (https://dcc.icgc.org/), was evaluated.

Both univariate and multivariate Cox proportional hazards models were used to examine whether the prognostic signature remained an independent predictor of survival after adjusting for clinicopathological variables, including age, gender, histologic grade, pathologic stage, and T, M, and N stages. A nomogram was then constructed by integrating significant independent prognostic factors identified from the univariate analysis. Calibration plots were used to assess the agreement between nomogram-predicted and observed survival outcomes.

Immunohistochemical Analysis of HAUS1, DTYMK, and SPP1 Expression in LIHC

A tissue microarray comprising 71 paired LIHC specimens and matched adjacent non-tumor tissues (HLivH150CS03, Shanghai Outdo Biotech Co., Ltd.) was utilized for IHC analysis. Following deparaffinization and rehydration, antigen retrieval was conducted by boiling the sections for 10 minutes in 10 mM sodium citrate buffer, pH 6.0 (16H10A24, BOSTER). Endogenous peroxidase activity was quenched with Reagent 1 (MXB Biotechnologies, KIT-9720) for 10 minutes, followed by blocking of non-specific binding with Reagent 2 (same kit) for 15 minutes at room temperature. The sections were subsequently incubated overnight at 4 °C with primary antibodies targeting HAUS1 (rabbit polyclonal, 1:200, #11094-2-AP, Proteintech), DTYMK (rabbit polyclonal, 1:200, #15360-1-AP, Proteintech), or SPP1 (rabbit polyclonal, 1:200, #22952-1-AP, Proteintech). After three washes in phosphate-buffered saline (PBS), the slides were sequentially incubated with biotinylated goat anti-rabbit IgG (Reagent 3, 10 minutes) and streptavidin-peroxidase (Reagent 4, 10 minutes) from the same MXB kit. Immunoreactivity was visualized using 3,3’-diaminobenzidine (DAB-0031, MXB Biotechnologies), followed by counterstaining with hematoxylin (absin, abs9217), dehydration, and mounting. The IHC staining results were imaged with a Leica LMD6 laser microdissection microscope and analyzed using ImageJ-IHC Profiler. Staining intensity was graded on a scale of 0 to 3 (negative, weak, moderate, strong), and the percentage of positive tumor cells was scored as 0 (< 0%), 1 (11–50%), 2 (51–80%), 3 (81–100%), or 4 (>80%). The final IHC score for each sample was obtained by multiplying the intensity score by the proportion score.

Cell Culture and Osteopontin Stimulation

The LIHC cell lines Huh7 (YC-D001, Ubigene Biosciences) and HepG2 (YC-C001, Ubigene Biosciences) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37 °C in a humidified 5% CO₂ atmosphere. For osteopontin stimulation, cells were seeded at ~40% confluence and, after a 12-hour incubation, recombinant human OPN (rOPN, 200 ng/mL; #HY-P70499, MedChemExpress) was added. Cultures were then incubated for an additional 72 h before analysis.

Western Blotting

On the third day after rOPN treatment, total protein was extracted from control and treated cells using RIPA buffer (P0013, Beyotime) supplemented with PMSF (ST506, Beyotime) and a protease/phosphatase inhibitor cocktail (P1046, Beyotime). Protein concentration was quantified using the BCA kit (P0010, Beyotime). Lysates were prepared with InstantView™ SEMS-PAGE protein staining and loading buffer (P0280, Beyotime), then separated on 10% SDS-PAGE gels and electrotransferred onto polyvinylidene difluoride (PVDF) membranes (#1620177, Bio-Rad). After blocking with 5% skim milk, membranes were incubated overnight at 4 °C with primary antibodies against HAUS1 (rabbit polyclonal, 1:200; #11094-2-AP, Proteintech), DTYMK (rabbit polyclonal, 1:200; #15360-1-AP, Proteintech), and GAPDH (1:1,000; #8884, Cell Signaling Technology). Membranes were then washed and incubated for 1 hour with HRP-conjugated horse anti-rabbit IgG (1:10,000; #7076, Cell Signaling Technology). Finally, chemiluminescent signals were visualized using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (WB322159, Thermo) and captured with a ChemiScope 6100 Touch system (Clinx, China).

Statistical Analysis

All statistical analyses were conducted in R (version 3.6.1) using relevant packages. Survival curves were generated using the Kaplan–Meier method, and the Log rank test was applied to compare differences among HAUS clusters. Univariate and multivariate Cox proportional hazards models were used to evaluate the prognostic significance of clinical features and the HAUS signature. Additionally, the Chi-square or Fisher’s exact test was used to assess associations between HAUS clusters and clinical characteristics. Meanwhile, the Mann–Whitney U-test (Wilcoxon rank-sum test) was used to compare HAUS clusters. Two-tailed p-values < 0.05 were considered significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Results

Pan-Cancer Profiling of the HAUS Family Gene Expression

To elucidate the role of HAUS family genes in tumorigenesis, we comprehensively analyzed the expression patterns of all eight HAUS members (HAUS1, HAUS2, HAUS3, HAUS4, HAUS5, HAUS6, HAUS7, and HAUS8) across 33 cancer types using RNA-sequencing data from the UCSC Xena and GTEx databases. Our analysis revealed that most HAUS members were significantly upregulated in several solid tumors, including cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney renal clear cell carcinoma (KIRC), brain lower-grade glioma (LGG), and LIHC (Figure 1). Conversely, downregulation of these genes was observed primarily in acute myeloid leukemia (LAML), ovarian serous cystadenocarcinoma (OV), and prostate adenocarcinoma (PRAD) (Figure 1).

Figure 1.

Pan-cancer expression of HAUS genes. (A–H) Comparison of HAUS genes mRNA expression between cancer and adjacent normal tissues. The HAUS family genes include (A) HAUS1, (B) HAUS2, (C) HAUS3, (D) HAUS4, (E) HAUS5, (F) HAUS6, (G) HAUS7, (H) HAUS8. *P < 0.05, **P < 0.01, ***P < 0.001.

Pan-Cancer Prognostic Significance of the HAUS Family Genes

To assess the prognostic significance of HAUS family genes, we conducted Cox regression analyses across various cancer types. The results indicated that, across multiple tumor types, including adrenocortical carcinoma (ACC), kidney chromophobe (KICH), LGG, LIHC, mesothelioma (MESO), and PRAD, the majority of HAUS family genes served as risk factors for overall survival (OS) (Figure 2A). Regarding disease-specific survival (DSS), they were identified as risk factors in ACC, KICH, LGG, LIHC, MESO, and PRAD (Figure 2B). Additionally, for progression-free survival (PFS), these genes also emerged as risk factors in ACC, KICH, LGG, LIHC, and PRAD (Figure 2C). Conversely, most HAUS family genes exhibited protective effects on OS in rectum adenocarcinoma (READ) and thymoma (THYM) (Figure 2A).

Figure 2.

Pan-cancer survival landscape of HAUS genes. (A–C) Association analyses of HAUS genes with OS (A), DSS (B), and PFS (C) across 33 tumor types. *P < 0.05, **P < 0.01, ***P < 0.001.

Abbreviations: OS, overall survival; DSS, disease-specific survival; PFS, progression-free survival.

By integrating pan-cancer expression profiles with survival data, we found that higher expression of HAUS family genes is significantly associated with worse survival outcomes, particularly in LGG and LIHC. These findings underscore the potential of the HAUS family genes as prognostic biomarkers and therapeutic targets in specific cancer types.

Elevated Expression of HAUS Family Genes and Their Association with Progression and Poor Prognosis in LIHC Patients

To further explore the specific role of HAUS family genes in LIHC, we validated their mRNA expression levels utilizing data from The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC). As illustrated in Figure 3A, all members of the HAUS family (HAUS1–HAUS8) were significantly upregulated in LIHC tissues compared to adjacent non-tumorous tissues. This dysregulation was further corroborated by ICGC transcriptomic data, which consistently exhibited elevated HAUS expression in LIHC samples (Figure 3B).

Figure 3.

Expression, survival, and clinical correlation analyses of HAUS genes in LIHC. (A) Comparison of HAUS expression between tumor and paired normal samples in the TCGA-LIHC dataset. (B) Comparison of HAUS expression between tumor and paired normal samples in the ICGC-LIRI dataset. ****p < 0.0001. (C) Univariate Cox analysis of the association between HAUS genes and overall survival. The values in bold were statistically significant. (D) Association between HAUS genes and tumor stage. *p < 0.05, **p < 0.01, ***p < 0.001.

Next, we examined the protein expression of HAUS family members using data from the Human Protein Atlas database. Representative immunohistochemistry (IHC) images of HAUS1–HAUS8 from normal liver tissue and LIHC tissues were presented in Supplementary Figure 1A–G, with clinicopathological parameters (patient ID, gender, age, staining level, intensity, and quantity) indicated for each sample. HAUS7 protein expression data were unavailable due to the lack of specific antibodies. Quantitative analysis of IHC staining intensity revealed that HAUS2 and HAUS4 were significantly overexpressed in LIHC tissues (Normal, n=3; LIHC, n=7). Although HAUS1, HAUS3, HAUS5, HAUS6, and HAUS8 also showed elevated expression in LIHC tissues, these differences did not reach statistical significance, likely due to the limited sample size (Supplementary Figure 1H).

Furthermore, univariate Cox regression analyses identified HAUS1, HAUS2, HAUS5, HAUS6, and HAUS8 as potential prognostic risk factors, with their high expression correlating with shorter overall survival (Figure 3C). In addition, we examined the relationships between HAUS expression and clinicopathological parameters. Notably, HAUS1, HAUS2, HAUS6, and HAUS8 exhibited strong correlations with advanced pathological stages (Figure 3D), with their expression levels increasing markedly as LIHC progresses. Collectively, our findings suggest that the dysregulated overexpression of HAUS family genes may play a substantial role in the pathogenesis and progression of LIHC.

Identification of Two LIHC Subtypes via Consensus Clustering of the HAUS Family Genes

To clarify the functional significance of HAUS family genes in LIHC, we analyzed their co-expression patterns with the PerformanceAnalytics R package. Pearson correlation analysis revealed significant positive correlations (r > 0.5) among HAUS1, HAUS2, HAUS3, HAUS5, HAUS6, and HAUS8 (Figure 4A). Using these six HAUS members, we performed consensus clustering with the ConsensusClusterPlus R package, stratifying 370 LIHC patients into two molecular subclusters: a HAUS-high cluster (C1) and a HAUS-low cluster (C2) (Figure 4B). Furthermore, principal component analysis (PCA) confirmed the distinct separation between these two subtypes (Figure 4C).

Figure 4.

HAUS expression patterns mediated by 6 HAUS genes. (A) Correlation analysis of 8 HAUS genes. *p < 0.05, ***p < 0.001. (B) Heatmap of 6 HAUS genes using unsupervised clustering. Clustering heatmap with consensus k=2. (C) PCA analysis of the 2 HAUS clusters. (D) Kaplan-Meier curve of the prognostic relationship between the two clusters in the TCGA-LIHC cohort. (E–H) Association of the 2 HAUS clusters with grade (E), stage (F), AFP (G), and age (H) in LIHC.

Survival analysis showed that patients in the C1 cluster had significantly shorter median overall survival than those in the C2 cluster (Figure 4D). Clinically, the HAUS-high cluster was strongly associated with more aggressive tumor features, including higher histological grades (G3: 43.23% versus 25.84%; G4: 4.52% versus 2.39%) (Figure 4E), advanced clinical stage (Stage III: 31.33% versus 18.97%) (Figure 4F), higher serum AFP levels (>400 ng/mL: 36.36% versus 14.46%) (Figure 4G), younger age at diagnosis (Figure 4H), and greater fibrosis severity (Ishak score 3–4: 44.55% vs. 22.82%) (Table 1). In contrast, no statistically significant differences were observed concerning gender, Child–Pugh class, or hepatitis virus etiology (HBV/HCV) (all p > 0.05; Table 1).

Table 1.

Correlations Between Different HAUS Expression Patterns and Clinical Characteristics of LIHC Patients

Variable	Total (n = 369)	Cluster		Statistic	P
		C1 (HAUS High, n = 158)	C2 (HAUS Low, n = 211)
Grade, n (%)				χ2=17.379	<0.001
G1	55 (15.11)	14 (9.03)	41 (19.62)
G2	176 (48.35)	67 (43.23)	109 (52.15)
G3	121 (33.24)	67 (43.23)	54 (25.84)
G4	12 (3.3)	7 (4.52)	5 (2.39)
Stage, n (%)				–	0.018
I	171 (49.57)	63 (42.00)	108 (55.38)
II	85 (24.64)	39 (26.00)	46 (23.59)
III	84 (24.35)	47 (31.33)	37 (18.97)
IV	5 (1.45)	1 (0.67)	4 (2.05)
Tumor status, n (%)				χ2=3.989	0.046
Tumor free	201 (57.43)	77 (51.33)	124 (62.00)
With tumor	149 (42.57)	73 (48.67)	76 (38.00)
Gender, n (%)				χ2=2.596	0.107
Female	121 (32.79)	59 (37.34)	62 (29.38)
Male	248 (67.21)	99 (62.66)	149 (70.62)
Age, n (%)				χ2=8.254	0.004
<=60	176 (47.7)	89 (56.33)	87 (41.23)
>60	193 (52.3)	69 (43.67)	124 (58.77)
Vascular invasion, n (%)				χ2=1.460	0.227
No	206 (65.81)	76 (61.79)	130 (68.42)
Yes	107 (34.19)	47 (38.21)	60 (31.58)
AFP ng mL, n (%)				χ2=17.824	<0.001
<=400	212 (76.81)	70 (63.64)	142 (85.54)
>400	64 (23.19)	40 (36.36)	24 (14.46)
Fibrosis ishak score, n (%)				χ2=14.933	0.005
0	70 (28)	20 (19.80)	50 (33.56)
1/2	28 (11.2)	12 (11.88)	16 (10.74)
3/4	79 (31.6)	45 (44.55)	34 (22.82)
5	8 (3.2)	3 (2.97)	5 (3.36)
6	65 (26)	21 (20.79)	44 (29.53)
Child pugh classification grade, n (%)				–	0.520
A	216 (90.76)	78 (88.64)	138 (92.00)
B	21 (8.82)	10 (11.36)	11 (7.33)
C	1 (0.42)	0 (0.00)	1 (0.67)
HBV serology, n (%)				χ2=0.582	0.446
HBV	141 (86.5)	70 (88.61)	71 (84.52)
Non-HBV	22 (13.5)	9 (11.39)	13 (15.48)
HCV serology, n (%)				χ2=0.123	0.726
HCV	103 (63.19)	51 (64.56)	52 (61.90)
Non-HCV	60 (36.81)	28 (35.44)	32 (38.10)

Notes: The values in bold were statistically significant.

Correlation of HAUS Molecular Patterns with TME in LIHC Patients

Previous research has linked HAUS family genes to immune modulation,²¹ prompting us to examine the relationship between HAUS molecular patterns and the TME in LIHC patients. ssGSEA revealed significant differences in immune cell infiltration between the C1 and C2 clusters. Notably, the C1 cluster showed reduced infiltration by activated CD8+ T cells and increased proportions of total CD4+ T cells and immunosuppressive Th2 cells (Figure 5A). Additionally, the C1 cluster displayed marked upregulation of immune checkpoint genes, including CD274 (PD-L1), PDCD1LG2 (PD-L2), PDCD1 (PD-1), CTLA4, LAG3, TIGIT, ICOS, LGALS9, CSF1R, IL10/IL10RB, VEGFA, TGFB1-3, TGFBR1, and members of the TNFSF/TNFRSF family (TNFSF4/18, TNFRSF4/9/18) (Figure 5B).

Figure 5.

Correlation analysis of immune regulation and the HAUS expression patterns. (A) Differences in T cell abundance between the two HAUS clusters. ns, not significant; *p < 0.05, ***p < 0.001. (B) Expression of 30 immune checkpoints in HAUS clusters C1 and C2. ***p < 0.001, ****p < 0.0001. (C–E) Comparison of IPS (C), IPS-PD1/PD-L1/PD-L2 blocker (D), and IPS-CTLA4 blocker scores (E) between HAUS clusters C1 and C2 in LIHC patients from TCGA-LIHC. Mean ± SEM; Kruskal–Wallis test. (F–N) Correlation of HAUS clusters C1 and C2 with TIDE (F), exclusion (G), MDSC (H), TAM M2 (I), IFNG (J), CD274 (K), Merck18 (L), Dysfunction (M), and CD8 (N) in TCGA. **p < 0.01, ***p < 0.001, ****p < 0.0001.

Given the strong association between HAUS molecular patterns and immune checkpoint gene expression, we compared the therapeutic responses of C1 and C2 to ICB therapy. Using the TCIA-recommended Immunophenoscore (IPS) and Tumor Immune Dysfunction and Exclusion (TIDE) algorithms, we found that the C1 cluster had significantly lower IPS scores than the C2 cluster (Figure 5C), including lower scores for both the IPS-PD1/PD-L1/PD-L2 blocker (Figure 5D) and the IPS-CTLA4 blocker (Figure 5E). This suggests reduced immunogenicity and a lower likelihood of benefiting from ICB in patients with high HAUS expression. Consistent with this, TIDE scores were markedly elevated in the C1 cluster (Figure 5F), indicating a higher probability of resistance to ICB. Mechanistically, TIDE analysis identified potential drivers of ICB resistance in tumors with elevated HAUS expression, including increased T-cell exclusion (Figure 5G) and enrichment of immunosuppressive cell populations, notably myeloid-derived suppressor cells (MDSCs) (Figure 5H) and M2-polarized tumor-associated macrophages (TAMs) (Figure 5I). Additionally, immune checkpoints (IFNG, CD274) (Figure 5J and K) and the Merck18 signature (Figure 5L) were upregulated, while T cell dysfunction (Figure 5M) and CD8+ T cell activity (Figure 5N) were attenuated in this group. Collectively, these findings suggest that HAUS family genes may facilitate immune evasion by promoting T cell exclusion and enriching immunosuppressive cell populations, including MDSCs, M2 macrophages, and Th2 cells.

Functional Analyses of the HAUS Molecular Patterns

To investigate the biological significance of distinct HAUS molecular patterns, we performed GSVA using the R package GSVA. GSVA-HALLMARK enrichment analysis showed that cell cycle-associated GSVA scores, including the G2M checkpoint, mitotic spindle, DNA repair, E2F targets, and MYC targets, were significantly higher in the C1 cluster than in the C2 cluster (Figure 6A). Moreover, GSVA-KEGG enrichment analysis highlighted DNA repair and cell cycle pathways in the C1 cluster, including homologous end joining, base excision repair, mismatch repair, nucleotide excision repair, and DNA replication (Figure 6B). Additionally, oncogenic pathways such as PI3K/AKT/mTOR, mTORC1, Wnt/β-catenin, NOTCH, and TGF-β signaling were significantly enriched in the C1 cluster (Figure 6A and B), suggesting a potential activation profile associated with carcinogenesis.

Figure 6.

Functional enrichment and genomic alterations associated with the HAUS expression pattern. (A and B) HALLMARK gene set (A) and KEGG pathway (B) enrichment analysis between the HAUS clusters C1 and C2 in the TCGA-LIHC dataset; mean ± SEM, Wilcox test, ***p < 0.001. (C and D) Genomic profiling of the top 20 most frequently altered genes in the HAUS clusters C1 (C) and C2 (D). (E and F) Boxplots of the estimated IC50 for nutlin 3a (p53 stabilizer) (E), roscovitine (Cdc2/CDK2/CDK5 inhibitor), and PD0332991 (CDK4/6 inhibitor) (F) between the HAUS clusters C1 and C2. Mean ± SEM, Wilcox test; p < 0.05 was considered statistically significant.

Furthermore, we analyzed the genomic mutation profile associated with HAUS molecular patterns using the R package Maftools. LIHC patients in the C1 cluster exhibited higher mutation frequencies for TP53 (40% versus 15%) and OBSCN (12% versus 4%), and lower frequencies for CTNNB1 (19% versus 31%) and ALB (6%versus 13%) than those in the C2 cluster (Figure 6C and D). Notably, the predominant TP53 mutation in the C1 cluster was missense (Figure 6C), which impaired tumor-suppressive function and conferred oncogenic gain-of-function effects independent of wild-type p53 activity. This provided insight into the underlying mechanism of the hyperactive cell cycle observed in the C1 cluster. Consistently, patients in the C1 cluster appeared to be more resistant to Nutlin 3a (Figure 6E), a small molecule that stabilizes wild-type p53 by inhibiting MDM2 and induces cell cycle arrest. Concurrently, patients in the C1 cluster demonstrated significantly higher IC50 values for other cell cycle inhibitors, including roscovitine (Cdc2/CDK2/CDK5 inhibitor) and PD0332991 (CDK4/6 inhibitor) (Figure 6F). Overall, these findings demonstrate a strong correlation between HAUS molecular patterns and abnormal cell-cycle regulation, with critical implications for therapeutic strategies and mechanisms of oncogenesis.

Association Analyses of HAUS Molecular Patterns with Stem Cell Characteristics and Therapeutic Response

Recent studies increasingly highlight that a specific subset of tumor cells, referred to as tumor-initiating cells (TICs) or cancer stem cells (CSCs), exhibits stem cell-like properties. These characteristics enable them to evade immune surveillance and resist current therapeutic interventions.²² Consequently, we examined the correlation between HAUS molecular patterns and stemness scores. The boxplot showed a significantly higher stemness score in the C1 cluster than in the C2 cluster (Figure 7A). Furthermore, markers associated with cancer stem cells, including EPCAM, ANPEP, CD24, CD44, and CD47, were elevated in the C1 cluster (Figure 7B).

Figure 7.

Associations of the HAUS expression pattern with stemness score and clinical drug treatment responses. (A and B) Differences in stemness score (A) and liver cancer stem cell markers (B) between HAUS clusters C1 and C2. **p < 0.01, ****p < 0.0001. (C–H) Boxplots of estimated IC50 values for chemotherapeutic and targeted agents in HAUS clusters C1 and C2, including PD0325901 (mirdametinib, MEK inhibitor), RDEA119 (refametinib, MEK inhibitor), AZD6244 (selumetinib, MEK inhibitor), CI-1040 (MEK inhibitor) (C); Erlotinib (EGFR inhibitor), AMG 706 (VEGFR inhibitor) (D); Imatinib (BCR/ABL/v-Abl/PDGFR/c-kit inhibitor), AZD0530 (Src inhibitor) (E); cisplatin, gemcitabine (F); doxorubicin (G); and methotrexate (H). Mean ± SEM; Wilcoxon test; p < 0.05 was considered statistically significant.

Furthermore, cell line data from the Cancer Genome Project (CGP) database were used to predict IC50 values of commonly used chemotherapeutic agents for LIHC patients across risk clusters. The C2 cluster showed significantly lower estimated IC50 values for PD0325901 (mirdametinib, a MEK inhibitor), RDEA119 (refametinib, a MEK inhibitor), AZD6244 (selumetinib, a MEK inhibitor), CI.1040 (PD 184352, a MEK inhibitor) (Figure 7C), Erlotinib (a EGFR inhibitor), AMG.706 (a VEGFR1/2/3 inhibitor) (Figure 7D), Imatinib (an inhibitor targeting BCR/ABL, v-Abl, PDGFR, and c-kit), and AZD0530 (Saracatinib, a Src inhibitor) (Figure 7E). These results suggest potential therapeutic benefits of these agents for patients in the C2 cluster. Conversely, the C1 cluster showed lower IC50 values for cisplatin, gemcitabine, doxorubicin, and methotrexate (Figure 7F–H), indicating that patients in the C1 cluster were more sensitive to these chemotherapeutic agents.

HAUS Family Genes May Be Novel Biomarkers for Liver Cancer Stem-Like Cells

To further elucidate the distinctive role of the HAUS family in LIHC progression, we performed single-cell RNA sequencing (scRNA-seq) on 18 samples (GSE149614) to investigate HAUS family expression within the hepatic TME. After quality control, 30,489 cells from ten primary liver tumors and 28,258 cells from eight non-tumor liver tissues were retained. These cells were merged, clustered, and annotated, then mapped to cell types, including T cells, macrophages, endothelial cells (Endothelial), liver cancer stem-like cells (LCSC-like cells), natural killer (NK) cells, fibroblasts, plasma cells, B cells, hepatocytes, and mast cells, based on cell-type-specific marker genes (Figure 8A and B). We then mapped the expression profiles of six HAUS members, namely HAUS1, HAUS2, HAUS3, HAUS5, HAUS6, and HAUS8. As illustrated in Figure 8C, these genes were broadly expressed across nearly all clusters, indicating their essential involvement in cellular viability. Notably, HAUS1, HAUS5, HAUS6, and HAUS8 showed specific expression in LCSC-like cells (Figure 8C), and their expression was positively correlated with the presence of LCSC-like cells (Figure 8D).

Figure 8.

Expression patterns of the HAUS members in the liver TME. (A) Dot plot depicting the typical cell-type-specific markers in liver cancer (GSE149614). (B) UMAP plot of cell types in liver cancer (indicated by colors). Cells were annotated as T cells, macrophages, endothelial cells (Endothelial), liver cancer stem-like cells (LCSC-like cells), NK cells (NK), fibroblasts, plasma cells, B cells, hepatocytes, and mast cells. (C and D) Expression distributions of the HAUS members (HAUS1, HAUS2, HAUS3, HAUS5, HAUS6, and HAUS8). (E) Dot plot of known LCSC-specific markers. (F) Heatmap of HALLMARK gene sets in subsets of LCSC-like cells.

Given the high expression of HAUS1, HAUS5, HAUS6, and HAUS8 in LCSC-like cells, we sought to elucidate their functional roles in these cells. Using dimensionality reduction, LCSC-like cells were classified into sixteen distinct subpopulations. Notably, HAUS1, HAUS5, HAUS6, and HAUS8 were highly expressed in LCSC(6), LCSC(7), LCSC(8), LCSC(15), and LCSC(16) (Figure 8E). Functional enrichment analysis of genes specific to LCSC(6), LCSC(7), and LCSC(15) showed significant enrichment in pathways related to the cell cycle, G2/M checkpoint, and E2F targets. Furthermore, LCSC(15) exhibited gene signatures primarily associated with DNA repair, oxidative phosphorylation, epithelial-mesenchymal transition (EMT), ROS pathway, MTORC1 signaling, MYC targets, and glycolysis. Significantly, LCSC(8) showed prominent enrichment in DNA repair, oxidative phosphorylation, and EMT pathways, whereas genes upregulated in LCSC(16) predominantly mediated processes including hypoxia, the p53 pathway, and PI3K/AKT/MTOR signaling (Figure 8F). Consistent with previous GSVA analyses, these findings further support the potential carcinogenic activation role of the HAUS family.

Construction of a Prognostic Model Related to HAUS Molecular Patterns

To identify genes associated with HAUS modification patterns, we used the limma package to detect differentially expressed genes (DEGs) between HAUS clusters C1 and C2. The volcano plot showed 6440 up-regulated and 362 down-regulated DEGs (Supplementary Figure 2). These DEGs were then subjected to LASSO-penalized Cox regression analysis, yielding a 2-gene prognostic signature (DTYMK and SPP1) at the optimal λ cutoff (Figure 9A and B). Both signature genes were significantly overexpressed in LIHC tissues compared with matched adjacent non-tumor tissues across both the TCGA-LIHC (Supplementary Figure 3A) and ICGC-LIRI cohorts (Supplementary Figure 3B).

Figure 9.

Establishment and validation of a HAUS-related signature. (A) LASSO coefficient profile for 6802 HAUS-related DEGs in LIHC. (B) Tenfold cross-validation of LASSO regression to estimate the optimal regularization parameter. (C) Heatmap of mRNA expression for 2 key HAUS-related DEGs in LIHC patients from the TCGA-LIHC dataset (n = 370). (D) Risk score distribution, survival status, and survival time of LIHC patients in the TCGA-LIHC dataset. (E) Kaplan–Meier curves for the high- and low-risk groups in the TCGA-LIHC cohort; Log-rank (Mantel–Cox) test. (F) ROC analysis of overall survival at 1, 3, and 5 years in the TCGA-LIHC cohort. (G) Risk score distribution, survival status, and survival time of LIHC patients in the ICGC-LIRI dataset (n = 240). (H) Kaplan–Meier curves for the high- and low-risk groups in the ICGC-LIRI cohort. (I) ROC analysis of overall survival at 1, 3, and 5 years in the ICGC-LIRI cohort. (J and K) Univariate (J) and multivariate (K) regression analyses of the risk score in the TCGA-LIHC cohort; Cox proportional hazards regression. p < 0.05 was considered statistically significant.

Next, we aimed to develop a prognostic model using DEGs associated with HAUS molecular patterns. The mRNA expression levels of DTYMK and SPP1 were multiplied by their respective coefficients derived from multivariate Cox regression analysis, and these values were subsequently summed: the risk score = (0.55 × DTYMK mRNA expression level) + (0.11 × SPP1 mRNA expression level). This process yielded the weighted prognostic risk score for each LIHC patient in both the TCGA-LIHC and ICGC-LIRI cohorts. Correspondingly, heatmap visualization confirmed higher expression levels of DTYMK and SPP1 in the high-risk cluster (Figure 9C). Furthermore, scatter plots showed a robust positive correlation between elevated risk scores and adverse survival outcomes, including shorter survival times and higher mortality rates (Figure 9D). Kaplan-Meier analysis also demonstrated a statistically significant decrease in OS within the high-risk cohort (Log rank test, p < 0.0001) (Figure 9E). Furthermore, time-dependent receiver operating characteristic (ROC) curves demonstrated the high sensitivity and specificity of the HAUS prognostic model. The area under the ROC curve (AUC) values were 0.755, 0.677, and 0.698 for 1-, 3-, and 5-year survival rates, respectively (Figure 9F).

Furthermore, the 2-gene HAUS prognostic signature was validated using an independent dataset from the International Cancer Genome Consortium (ICGC) Liver Cancer - RIKEN (LIRI) project. A total of 240 LIHC patients were stratified into high-risk (n = 127) and low-risk (n = 113) groups using the risk score cutoff derived from the TCGA cohort. Consistent with findings from the TCGA dataset, LIHC patients in the high-risk group had significantly higher mortality and worse survival outcomes (p < 0.001) than those in the low-risk group (Figure 9G and H). The AUC scores for predicting OS at 1 and 3 years were 0.697 and 0.753, respectively (Figure 9I). In conclusion, the 2-gene HAUS signature demonstrated robust prognostic performance for survival in hepatocellular carcinoma.

For clinical application, the mRNA expression levels of DTYMK and SPP1 in the analyzed LIHC tissues were quantified by RNA sequencing (RNA-seq). Subsequently, a risk score was calculated for each patient using the aforementioned formula. Patients with risk scores below 0.9672 were deemed to have a favorable prognosis (low risk), whereas those with scores of 0.9672 or higher were considered to have an unfavorable prognosis (high risk).

The HAUS-Related Prognostic Signature Was an Independent Prognostic Factor for LIHC

To identify independent prognostic factors for LIHC patients, we incorporated the HAUS-related risk score and clinical characteristics into a univariate Cox proportional hazards regression analysis. The univariate Cox regression analysis revealed that pathological stage (HR = 1.879, 95% CI 1.466–2.408), T stage (HR = 1.816, 95% CI 1.443–2.287), M stage (HR = 3.924, 95% CI 1.230–12.519), and risk score (HR = 1.701, 95% CI 1.421–2.037) were significant prognostic factors in the TCGA cohort (Figure 9J). Subsequently, these four variables were included in the multivariate Cox regression analysis, which indicated that the risk score (HR = 1.564, 95% CI 1.291–1.895) was an independent prognostic factor for OS in LIHC patients (Figure 9K).

Furthermore, we developed a prognostic nomogram to predict the 1-year, 3-year, and 5-year OS for LIHC patients by integrating pathological stage, T stage, M stage, and risk score. For instance, a patient with stage 2 (52 points), T2 (54 points), M0 (50 points), and a high-risk score (51 points) would have a total score of 207 points. The corresponding 1-year, 3-year, and 5-year survival probabilities were 74.2%, 54.2%, and 42.5%, respectively (Supplementary Figure 4A). Subsequently, we evaluated the predictive accuracy of the nomogram-based model using calibration curves. The calibration curves closely approximated the 45-degree diagonal line, indicating a strong concordance between the survival probabilities predicted by the nomogram and those observed in the TCGA-LIHC cohort (Supplementary Figure 4B).

SPP1 Positively Regulated HAUS1 and DTYMK Expression in LIHC

To further clarify the relationships among HAUS genes, DTYMK, and SPP1, we evaluated their expression in tissue microarrays from 69 LIHC patients. HAUS1 was predominantly localized around the nucleus, and its expression was upregulated in LIHC tumors compared with non-tumor tissues (Figure 10A–C). DTYMK was detected in both the cytoplasm and the nucleus, with significantly higher expression in tumor cell nuclei (Figure 10D–G). Although SPP1 expression was significantly elevated in a subset of the 69 LIHC patients, no statistically significant difference was observed between tumor and adjacent non-tumor tissues, likely due to inter-patient heterogeneity (Figure 10H–J).

Figure 10.

Correlation analysis among HAUS1, SPP1, and DTYMK. (A) IHC analysis of HAUS1 in a tissue microarray of 69 paired LIHC specimens (scale bar: 7 mm). (B) Representative IHC images show higher HAUS1 levels in LIHC tumor tissues than in adjacent non-tumor liver tissues (scale bar: 200 μm). (C) IHC images of HAUS1 were quantified with an IHC score using ImageJ (n = 69). (D) IHC analysis of DTYMK in a pair of 69 LIHC patients (scale bar: 7 mm). (E) Representative IHC images show higher DTYMK levels in LIHC tumor tissues than in adjacent non-tumor liver tissues (scale bar: 200 μm). (F and G) IHC images of DTYMK were quantified with an IHC score using ImageJ (n = 69). (H) IHC analysis of SPP1 in a pair of 69 LIHC patients (scale bar: 7 mm). (I) Representative IHC images show higher SPP1 levels in LIHC tumor tissues than in adjacent non-tumor liver tissues (scale bar: 200 μm). (J) IHC images of SPP1 were quantified with an IHC score using ImageJ (n = 69). (K) Correlation analysis of HAUS1 with SPP1 using IHC data. (L) Correlation analysis of SPP1 with DTYMK cytoplasm using IHC data. (M) Correlation analysis of SPP1 with DTYMK nucleus using IHC data. (N) Correlation analysis of HAUS1 with DTYMK cytoplasm using IHC data. (O) Correlation analysis of HAUS1 with DTYMK nucleus using IHC data. (P) Correlation analysis of DTYMK cytoplasm with DTYMK nucleus using IHC data. (Q) SPP1 may positively regulate HAUS1 and DTYMK expression in liver cancer cells.

Furthermore, we observed positive correlations between SPP1 and HAUS1, as well as between SPP1 and cytoplasmic/nuclear DTYMK (Figure 10K–M). Similarly, HAUS1 expression correlated positively with both cytoplasmic and nuclear DTYMK (Figure 10N and O). Additionally, cytoplasmic and nuclear DTYMK expression levels were positively correlated (Figure 10P). Functionally, treatment of HepG2 and Huh7 cells with recombinant OPN (200 ng/mL) significantly induced HAUS1 and DTYMK expression (Figure 10Q and Supplementary Figure 5). Collectively, these results indicate that SPP1 may act as a positive regulator of HAUS1 and DTYMK in LIHC.

Discussion

The HAUS complex is an evolutionarily conserved, eight-subunit regulator of microtubule nucleation, essential for mitotic spindle assembly and chromosomal fidelity.²³ Increasing evidence suggests that HAUS members may have diagnostic, prognostic, and therapeutic relevance across a range of malignancies, supporting their potential utility in precision oncology.^17,18,24,25 However, a systematic analysis of their collective expression patterns across cancers has been lacking.

In this study, we show that coordinated HAUS member expression may serve as a prognostic indicator across diverse cancer types and identify LIHC as a setting where HAUS-defined subtypes intersect with immune evasion, stemness, and therapeutic resistance. Across 33 TCGA cancer types, most HAUS members were overexpressed in tumors relative to normal tissues, with particularly pronounced changes in CHOL, COAD, GBM, HNSC, KIRC, LGG, and LIHC. Survival analyses further indicated that HAUS dysregulation is associated with adverse outcomes in multiple malignancies, especially ACC, KICH, LGG, LIHC, MESO, and PRAD. Together, these findings suggest broad oncogenic relevance for HAUS members and highlight particularly strong associations in LGG and LIHC, consistent with prior reports.^18,21,26

Liver cancer continues to be a significant global health concern, with over one million new cases per year worldwide.²⁷ LIHC, the most prevalent form of liver cancer, ranks as the second leading cause of cancer-related mortality globally.²⁸ Although the molecular mechanisms underlying hepatocarcinogenesis and tumor progression remain incompletely elucidated, several signaling pathways, including receptor tyrosine kinase (RTK), immune evasion, and the p53 pathway, have been recognized as pivotal in the pathophysiology of LIHC. Targeted therapies against these pathways now constitute the standard of care for advanced disease. Novel combination strategies, including immune checkpoint inhibitors (ICIs) with tyrosine kinase inhibitors (TKIs) or anti-VEGF agents, as well as dual immunotherapy regimens, are under clinical investigation and hold promise for revolutionizing LIHC management across all stages.²⁹ Despite advances in systemic therapies, outcomes for LIHC patients remain heterogeneous and generally unfavorable. Against this background, our data suggest that HAUS expression may identify a biologically aggressive subset of LIHC, although prospective validation will be needed to establish clinical utility.

Unsupervised clustering based on HAUS expression separated LIHC into two molecular subgroups with distinct clinicopathological features. The HAUS-high cluster was associated with more aggressive histopathologic characteristics and worse survival. Functional enrichment linked this subgroup to cell-cycle and mitotic spindle programs, while mutational profiling showed a strong association with TP53 alterations. Notably, the HAUS-high group also showed reduced predicted sensitivity to p53 stabilizers and cell-cycle inhibitors, suggesting that HAUS overexpression may mark a therapy-resistant state driven by impaired checkpoint control and proliferative stress.³⁰ Alternative explanations, including differences in drug metabolism or uptake, cannot be excluded.

TME plays a crucial role in oncogenesis and malignant progression.^31,32 We observed notable inter-cluster differences within the TME, with patients characterized by HAUS high expression demonstrating elevated TIDE immune scores and reduced immunogenicity, consistent with prior research.³³ Furthermore, HAUS high-expression patients had decreased infiltration of CD8⁺ T cells and increased proportions of CD4⁺ T cells, type 2 T helper cells, and M2-type tumor-associated macrophages, all of which suppress antitumor immunity and contribute to tumor progression. CD8⁺ T cells are essential effector cells in antitumor immunity; however, in the TME, persistent antigen exposure can drive CD8⁺ T cells into a dysfunctional state known as T-cell exhaustion, a significant impediment to effective antitumor responses.³⁴ Notably, the HAUS expression pattern showed robust correlations with T-cell exhaustion markers, suggesting its potential utility for predicting an immunosuppressive TME in patients with LIHC. These associations are correlative and do not establish whether HAUS expression actively shapes immune composition or whether both are driven by a common upstream factor such as oncogenic signaling or tumor mutational burden. Even so, the observed correlation between HAUS expression and exhaustion-related markers raises the possibility that HAUS-high tumors are characterized by attenuated immune surveillance. One speculative model is that HAUS-driven mitotic stress generates immunostimulatory signals, potentially through cGAS-STING activation by micronuclei, which then lead to compensatory checkpoint upregulation and immune exclusion;³⁵ this mechanism remains to be tested experimentally.

Recently, ICB has profoundly transformed cancer treatment, offering significant clinical benefits for patients with refractory malignancies.³⁶ Comparative transcriptomic profiling of therapeutically relevant immune checkpoint genes revealed significant upregulation of multiple key regulators, including CD274 (PD-L1), PDCD1LG2 (PD-L2), PDCD1 (PD-1), CTLA4, LAG3, TIGIT, ICOS, LGALS9, CSF1R, IL10/IL10RB, VEGFA, TGFB1–3, TGFBR1, and members of the TNFSF/TNFRSF families, in patients with elevated HAUS expression. This expression pattern suggests that HAUS-high LIHC may be characterized by active immune checkpoint signaling and could therefore represent a distinct immunotherapy-relevant subgroup.

Cancer stem cells (CSCs) contribute to therapy resistance through diverse mechanisms, including ABC transporter-mediated drug efflux, impaired DNA repair, enhanced anti-apoptotic signaling, metabolic adaptation, epigenetic alterations, and activation of stemness pathways such as Notch, WNT, and Hedgehog.^37,38 Our analysis also suggests a role for HAUS members in cancer stem cell biology. HAUS-high tumors showed elevated stemness scores and higher expression of CSC-associated markers, including EPCAM, ANPEP, CD24, and CD44. Single-cell RNA-seq further demonstrated enrichment of HAUS family genes in a liver cancer stem-like subpopulation. In glioblastoma stem-like cells, HAUS5 knockdown has been reported to disrupt HAUS localization on the spindle, leading to centrosome fragmentation, spindle multipolarity, and reduced self-renewal and tumorigenic capacity. By contrast, depletion in neural progenitor cells has minimal effects on spindle or centrosome integrity.¹⁶ Together, these findings support the idea that HAUS dysregulation may sustain proliferation and stem-like properties, thereby contributing to treatment resistance and disease relapse. Functional validation with tumor-initiating assays and lineage tracing will be needed to test this hypothesis.

To explore the regulatory mechanisms underlying HAUS family gene expression, we compared high- and low-HAUS expression groups to identify differentially expressed genes. Using LASSO regression and multivariate Cox analysis, we identified two key HAUS-related prognostic genes, DTYMK and SPP1, and developed a risk model. This risk score independently predicted LIHC patient outcomes, with ROC analysis showing strong sensitivity and specificity. Additionally, we created a comprehensive nomogram combining the HAUS risk score with T stage, M stage, and pathological stage, providing clinicians with a useful tool to accurately estimate patient prognosis. DTYMK has been reported to catalyze the phosphorylation of deoxythymidine monophosphate (dTMP) to form deoxythymidine diphosphate (dTDP), an essential material in DNA synthesis. DTYMK expression was upregulated and involved in LIHC development, suggesting its potential as a biomarker and therapeutic target for LIHC.³⁹ OPN, encoded by the SPP1 gene, is a sialic acid-rich, glycosylated phosphoprotein that functions as a chemokine-like molecule and binds integrins. It is widely expressed across cell types and contributes to cancer progression through receptor-mediated signaling pathways in LIHC and other malignancies.^40,41 The co-expression of HAUS, DTYMK, and SPP1, together with our observation that recombinant OPN induces HAUS1 and DTYMK protein expression, supports a potential functional relationship among these factors. However, the directionality of regulation remains unresolved: whether SPP1 acts upstream of HAUS, whether feedback loops exist, and whether these interactions occur in vivo are all questions for future work.

These findings have several translational implications. First, the SPP1–HAUS1/DTYMK axis may represent a therapeutically relevant network, offering opportunities to disrupt upstream SPP1 signaling or target DTYMK-dependent nucleotide metabolism in HAUS-high tumors. Second, HAUS expression may help refine immunotherapy stratification, as HAUS-high tumors are enriched for immune checkpoint and exhaustion-related features and may be less likely to respond to single-agent immune checkpoint blockade. Third, the association between HAUS-high status and predicted resistance to cell-cycle inhibitors suggests that rational combination strategies may be required to overcome this phenotype. Finally, developing small molecules that interfere with HAUS members or function is attractive, although the essential role of HAUS in normal mitosis raises important concerns about selectivity and toxicity.

Several limitations should be noted. Our bioinformatics analyses relied on retrospective TCGA data and are therefore subject to confounding by treatment heterogeneity, sample quality, and demographic imbalance. Most findings are associative rather than causal. Although key results were supported by IHC and Western blotting, additional in vitro and in vivo studies are needed to clarify the mechanistic links among SPP1, HAUS, and DTYMK. Finally, validation in independent cohorts and prospective clinical studies will be required before clinical translation can be considered.

Conclusion

In this study, we characterized HAUS gene expression patterns in LIHC and developed a HAUS-related prognostic risk model to assess its clinical relevance. ROC and survival analyses indicated that both the HAUS expression signature and the corresponding risk score have prognostic value in LIHC. Bioinformatics analyses further showed that HAUS expression was associated with immune checkpoint expression, T-cell exhaustion markers, and immunosuppressive cell populations in the TME. Immunohistochemical staining and Western blot analysis provided preliminary experimental support for the differential expression patterns identified by transcriptomic analysis. However, the clinical value of the HAUS-related signature requires validation in larger, independent cohorts before translational use can be considered. In addition, because the present results are mainly associative, functional studies are needed to clarify whether HAUS dysregulation directly contributes to LIHC progression. Future work should focus on elucidating the mechanisms linking HAUS to immune evasion and on assessing its potential as a therapeutic target in LIHC.

Funding Statement

This work was supported by grants from the Basic and Applied Basic Research Project of the Guangzhou Science and Technology Bureau (202201011003) and the Plan to Enhance Scientific Research at GMU (2024).

Disclosure

The authors report no conflicts of interest in this work.