Sage Journals: Discover world-class research

Abstract

Background

Malignant neoplasms of bone and articular cartilage represent a rare group of highly invasive tumours with a poor prognosis, whose molecular mechanisms remain incompletely understood. Systematic identification of potential pathogenic proteins holds significance for early diagnosis and targeted therapy.

Methods

A genetic tool was constructed using nine large-scale cis-protein quantitative trait loci (pQTL) datasets. Dual-sample Mendelian randomisation analysis was performed with plasma proteins as exposure and malignant neoplasms of bone and articular cartilage risk as the outcome. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses were performed on proteins showing suggestive causal associations (p < 0.05). Sensitivity analyses, Steiger’s directionality tests and co-localisation studies were conducted on Bonferroni-corrected significant proteins. Finally, candidate small-molecule drugs were predicted using the Enrichr database and molecular docking analysis.

Results

Preliminary Mendelian randomisation analysis identified 171 suggestive causal proteins, with 80 positively and 91 negatively correlated with the risk of malignant neoplasms of bone and articular cartilage. Enrichment analysis revealed that these proteins primarily participated in immune regulation, extracellular matrix degradation and proteolysis-related pathways. Following Bonferroni correction, seven proteins—GNLY, PCSK7, ADAMTS5, PDCD1LG2, SCG3, CXCL16 and CNTN1—retained significant causal associations. Further co-localisation analysis revealed that ADAMTS5, GNLY and PCSK7 shared genetic variants associated with the risk of malignant neoplasms of bone and articular cartilage. Molecular docking analysis indicated that compounds such as aspirin and vitamin E exhibited low binding energies with GNLY, PCSK7 and ADAMTS5, suggesting potential therapeutic intervention opportunities.

Conclusion

This study identified three proteins (GNLY, PCSK7 and ADAMTS5) associated with a high risk of malignant neoplasms of bone and articular cartilage through Mendelian randomisation and co-localisation analyses, providing novel molecular evidence for early diagnosis, risk assessment and potential targeted therapies for malignant neoplasms of bone and articular cartilage.

Keywords

Malignant bone and cartilage tumours Mendelian randomisation plasma proteins cis-protein quantitative trait locus co-localisation analysis GNLY PCSK7 ADAMTS5

Introduction

Malignant neoplasms of bone and articular cartilage (MNBAC) are a rare and severe tumour category. The term ‘bone tumour’ refers to all cancers that originate in the skeleton or other bone tissue, including primary, secondary and metastatic tumours.¹ Primary MNBAC includes osteosarcoma, chondrosarcoma and Ewing’s sarcoma.² The primary clinical manifestations of MNBAC are pain, swelling and impaired function. Although MNBAC has a low overall incidence, it predominantly affects children, adolescents and young adults.³ Characterised by high invasiveness, a significant metastasis rate and poor prognosis, MNBAC severely impacts patients’ quality of life and survival rates. Despite recent advances in integrated therapeutic approaches, including surgical resection, chemotherapy and radiotherapy, there have been only limited improvements in the 5-year survival rates of MNBAC patients.^4–7 Consequently, identifying novel targets for the screening, prevention and treatment of MNBAC is of paramount importance.

At the molecular level, the development of bone and cartilage tumours involves several biological processes, such as cell proliferation, differentiation, apoptosis, angiogenesis and the immune microenvironment. Previous studies have identified several signalling pathways that are closely associated with MNBAC, including the phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT), Wnt/β-catenin and transforming growth factor-beta (TGF-β) pathways.^8–11 However, as these studies are largely based on candidate gene analysis or small-scale transcriptome sequencing, it is difficult to elucidate causal relationships at the global level. Concurrently, the field of drug discovery faces high failure rates during target identification, partly due to insufficient target validation and the lack of effective means to identify disease-driving factors at a causal level. Consequently, the critical scientific question of how to identify potential therapeutic targets using large-scale population data and systems biology approaches requires urgent attention.

Advancements in human genomics and proteomics have enabled the study of protein quantitative trait loci (pQTLs), providing valuable insights into the genetic links between protein-level variation and complex diseases.^12,13 When combined with causal inference methods such as Mendelian randomisation (MR), genetic variation can act as an instrumental variable, enabling the exploration of causal relationships between proteins and diseases.¹³ This approach overcomes the limitations of traditional observational studies, which are susceptible to confounding and reverse causality effects.¹⁴ In the field of bone and cartilage tumours, however, such applications are still in their infancy, providing an opportunity for researchers to use integrated genomics and proteomics strategies to identify new pathogenic proteins and potential therapeutic targets.

Against this backdrop, this study used a two-sample MR approach to systematically evaluate the causal relationships between plasma proteins and MNBAC, using large-scale pQTL data and MNBAC genome-wide association study (GWAS) data from the Finnish cohort (Figure 1). First, we conducted a global MR analysis of candidate proteins to identify those showing significant and suggestive causal associations. Next, we explored potential biological mechanisms through Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses, and we validated the robustness of the results via sensitivity analyses, including tests for heterogeneity and pleiotropy, Steiger’s test and co-localisation analysis. Finally, we used the Enrichr drug database to screen for small-molecule compounds targeting these proteins, with the aim of identifying potential therapeutic agents. We believe that our results will provide novel insights and evidence to support the elucidation of the molecular mechanisms of MNBAC and the development of targeted therapeutic strategies.

Figure 1.

A flowchart of the Mendelian randomisation analysis framework for evaluating the impact of plasma proteomics on MNBAC. MNBAC: malignant neoplasms of bone and articular cartilage.

Materials and methods

Study exposures

The exposures in this study were genetically predicted circulating protein levels, instrumented exclusively by cis-pQTLs. Cis-pQTLs were defined as variants located within ±500 kb of the encoding gene, and only these variants were retained to maximise biological plausibility and minimise horizontal pleiotropy. We constructed MR instruments for plasma proteins based on nine large-scale GWAS, each with a sample size greater than 500 and profiling more than 50 proteins. Plasma pQTLs were used as candidate instrumental variables, and the selection process was as follows:

Genome-wide significance. We included single nucleotide polymorphisms (SNPs) associated with protein levels according to the p-value thresholds recommended by the original GWAS.

Major histocompatibility complex (MHC) exclusion. SNPs located within the MHC region (chromosome 6: 26–34 Mb) were excluded due to the complex linkage disequilibrium (LD) structure.

LD pruning. To ensure independence, LD clumping was performed with a threshold of r² <0.01 and a physical distance of ±5 Mb.

Pleiotropy control. Variants associated with five or more proteins were removed to minimise potential horizontal pleiotropy.

Study outcomes

Summary-level GWAS data for MNBAC were obtained from the FinnGen consortium (release R12; https://r12.finngen.fi/). The dataset comprised data from 379,397 participants of European ancestry, including 648 MNBAC cases and 378,749 controls. MNBAC cases were defined based on ICD-10 diagnostic codes C40–C41, covering MNBAC. The control population consisted of all other FinnGen participants who did not have any MNBAC diagnoses, i.e. individuals with no ICD-10 codes C40–C41 recorded at any time. Controls included participants with unrelated medical conditions or healthy individuals; however, any individual with a diagnosis of malignant bone or cartilage tumours was excluded from the control group. Detailed information regarding phenotyping, genotyping and quality control procedures is available on the FinnGen website.

MR analysis

In this study, genetically predicted protein levels served as the exposure variable, while MNBAC acted as the outcome variable in a two-sample MR analysis. For proteins with only one cis-pQTL, the Wald ratio method was used to estimate causal effects;^15,16 whereas for proteins associated with two or more cis-pQTLs, an inverse-variance weighted (IVW) method was used for integrated analysis.¹⁷ The magnitude of causal effects was expressed as the change in the MNBAC risk corresponding to each standard deviation increase in protein expression levels, presented as odds ratios (ORs) with 95% confidence intervals (CIs).

To mitigate the risk of false positives from multiple hypothesis testing, a dual-threshold strategy was employed. When the p-value was <0.05, a protein was considered to show suggestive causal association with MNBAC and was subjected to enrichment analyses to explore potential biological pathways; when the p-value met the Bonferroni-corrected threshold (p < 0.05/number of proteins), the protein was deemed to have a significant causal association with MNBAC and was subjected to further sensitivity analysis.¹⁶

Sensitivity analysis encompassed the following aspects. First, the Steiger directionality test was performed to determine whether the MR results were distorted by reverse causality, thereby excluding the influence of counterfactual causation.¹⁸ A result of true with p <0.05 supported a positive causal effect of exposure (protein) on outcome (MNBAC). Second, MR–Egger regression was employed to assess potential horizontal multivariability. A significant deviation of the intercept term from zero indicated uncontrolled multivariable bias.¹⁹ Concurrently, the weighted median method was used to estimate causal effects, which provides robust estimates even when over 50% of the weighted instrumental variables are invalid. Furthermore, Cochran’s Q statistic was used to assess heterogeneity among instrumental variables.²⁰ Significant heterogeneity suggests potential systematic discrepancies between estimates derived from different instrumental variables. This integrated assessment enhances the reliability and interpretability of MR results.

Co-localisation analysis

Following multiple testing threshold screening, we further performed Bayesian co-localisation analysis to validate MR results to assess whether each site was influenced by the same genetic variation that affected protein levels and MNBAC risk simultaneously. This approach reduces false associations arising from LD. The co-localisation analysis was performed using the R package coloc²¹ (http://cran.r-project.org/web-packages/coloc), which calculates posterior probabilities for five hypotheses: posterior probability (PPH0), functional association only (PPH1), GWAS association only (PPH2), independent functional association (PPH3) and co-localised functional/GWAS association (PPH4),²² where PPH4 indicates shared causal variation between traits. We performed co-localisation analysis on all SNPs within a 1-megabase window of significant pQTLs derived from the MR analysis. When the posterior probability for H4 exceeded 0.8, this was interpreted as strong co-localisation evidence.²²

Functional enrichment analysis

To explore the potential biological functions and associated signalling pathways of proteins exhibiting suggestive causal relationships, this study conducted functional annotation and enrichment analysis on the genes corresponding to the screened proteins. GO analysis encompassed three major categories: biological process (BP), molecular function (MF) and cellular component (CC). KEGG pathway analysis was performed to reveal key signalling pathways associated with disease onset and progression. The analysis used the clusterProfiler package in the R programming language. Significance was calculated using the hypergeometric test, with p-values corrected for multiple hypothesis testing using the Benjamini–Hochberg method. A corrected p-value <0.05 was adopted as the significance threshold. Results were visualised using bubble plots.

Candidate drug identification and molecular docking

For the causative proteins identified with strong evidence in the aforementioned analysis, we first used the Enrichr online analysis tool to identify small-molecule candidate compounds capable of interacting with these proteins. Subsequently, we retrieved the three-dimensional structures of the proteins from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB) database and the three-dimensional structures of the small-molecule compounds from the PubChem database.²³ When selecting protein structures, priority was given to complete, low-resolution crystal structures to ensure the reliability of the docking results. Molecular docking was performed using AutoDock Vina software to calculate the binding energy between the protein and ligand across different conformations. The conformation yielding the lowest binding energy in the docking results was deemed the most probable binding mode. This was further visualised using PyMOL software to illustrate the binding site and interactions within the protein–ligand complex.²⁴

Results

Construction of plasma protein genetic tools

This study screened pQTLs based on nine large-scale GWAS to construct genetic tools for plasma proteins. Table 1 summarises key information from the nine studies, with complete details provided in Supplementary Table 1. Following consolidation and quality control screening, 8285 pQTLs corresponding to 4421 proteins were obtained. Given that this study exclusively utilised cis-pQTLs, 3811 cis-pQTLs across 2958 proteins were ultimately selected for subsequent MR analysis. Detailed information is provided in Supplementary Table 2. All selected tool variables were strong tools, with a minimum F-value of 28.52, ensuring the robustness of the analysis.

Table 1.

Source information of the instrumental variables.

PMID	Author	Sample size	Number of proteins measured	p-value threshold of pQTLs selection
28240269	Karsten Suhre et al.	997	1124	8.72E−11
31320639	Robert F. Hillary et al.	750	92	5.40E−10
33328453	Maik Pietzner et al.	10,708	179	5.00E−08
29875488	Benjamin B. Sun et al.	3301	2994	1.50E−11
30111768	Chen Yao et al.	6861	71	1.25E−07
34857953	Egil Ferkingstad et al.	35,559	4719	1.80E−09
28369058	Lasse Folkersen, et al.	3394	83	5.00E−08
33303764	Arthur Gilly et al.	1407	257	7.45E−11
35078996	Alexander Gudjonsson et al.	5368	4135	1.05E−11

Proteins exhibiting suggestive causal associations with MNBAC and their enrichment patterns

Given that cis-pQTLs possess a higher a priori probability of biological effect compared with trans-pQTLs, this study used cis-pQTLs as genetic instruments for MR analysis to assess the causal effects of plasma proteins on MNBAC (Supplementary Table 3). Results revealed that when p-values were below 0.05, the Wald ratio and IVW methods jointly identified 171 proteins exhibiting suggestive causal associations with MNBAC (Figure 2(a)). Among these, 80 proteins showed positive correlations with MNBAC, indicating that elevated expression levels were associated with increased disease risk; conversely, 91 proteins demonstrated negative correlations with MNBAC, suggesting that increased expression levels reduce disease risk.

Figure 2.

Proteins exhibiting suggestive causal associations with MNBAC and their enrichment results. (a) Volcano plot of MR analysis results. (b) Bubble plot of GO enrichment. (c) Bubble plot of KEGG enrichment. MNBAC: malignant neoplasms of bone and articular cartilage; MR: Mendelian randomisation; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes.

Further GO and KEGG enrichment analyses were conducted on proteins with suggestive causal associations. GO enrichment revealed that these proteins primarily participate in immune response, inflammatory response, cytokine-mediated signalling pathways, cell adhesion and bone metabolism-related processes. Their MFs predominantly focused on receptor binding, chemokine activity and cytokine activity, while CCs showed enrichment in the extracellular space, plasma membrane and immune-related complexes (Figure 2(b)). KEGG pathway analysis revealed significantly enriched pathways including cytokine–receptor interactions, chemokine signalling, interleukin (IL)-17 signalling, Janus kinase (JAK)–signal transducer and activator of transcription (STAT) signalling and tumour necrosis factor (TNF) signalling—classic immune and inflammatory pathways. This suggests that immune regulation and inflammatory responses play crucial roles in MNBAC pathogenesis. Furthermore, significant enrichment was observed in pathways associated with rheumatoid arthritis and osteoclast differentiation linked to bone resorption, suggesting that MNBAC is closely related to abnormal bone metabolism and joint inflammatory processes. Concurrently, enrichment in glycosaminoglycan biosynthesis, sphingolipid metabolism and lysosomal pathways suggests that alterations in extracellular matrix (ECM) remodelling, lipid metabolism and cellular clearance mechanisms collectively contribute to the disease pathology (Figure 2(c)). These findings indicate that the pathogenic mechanisms of MNBAC-associated proteins are closely linked to immune-inflammatory responses, cytokine signalling regulation and bone remodelling metabolic processes.

Proteins and sensitivity analysis of significant causal relationships following Bonferroni correction

Following Bonferroni correction, seven proteins were identified as significantly associated with MNBAC (Figure 3(a)). These specifically included granulysin (GNLY, OR = 1.206), contactin-1 (CNTN1, OR = 0.562), proprotein convertase subtilisin/kexin type 7 (PCSK7, OR = 0.866), secretogranin-3 (SCG3, OR = 0.705), programmed cell death 1 ligand 2 (PDCD1LG2, OR = 1.299), C-X-C motif chemokine 16 (CXCL16, OR = 0.533) and A disintegrin and metalloproteinase with thrombospondin motifs 5 (ADAMTS5, OR = 0.805). Among these, GNLY and PDCD1LG2 exhibited OR values >1, indicating that their elevated expression levels were significantly associated with increased MNBAC risk. Conversely, PCSK7, SCG3, CXCL16 and ADAMTS5 displayed OR values <1, suggesting that their higher expression levels confer a potential protective effect against MNBAC (Figure 3(b)).

Figure 3.

Proteins exhibiting significant causal associations with MNBAC. (a) Volcano plot of MR analysis results and (b) forest plot of detailed results for the seven proteins. MNBAC: malignant neoplasms of bone and articular cartilage; MR: Mendelian randomisation.

In further sensitivity analyses, we systematically assessed potential pleiotropy, robustness of instrumental variables and the possibility of reverse causality (Table 2 and Supplementary Table 4). Egger regression intercept tests revealed pleiotropy p-values >0.05 for all candidate proteins, indicating no significant pleiotropy. Cochran’s Q test results further revealed heterogeneity p-values >0.5 for all proteins, indicating good consistency among instrumental variables. Additional Steiger directionality tests showed p-values well below 0.05 for all seven proteins, with effect directions consistent with the hypothesis, supporting unidirectional causality.

Table 2.

Sensitivity analysis of seven significant causal proteins.

Exposure	Outcome	Pleiotropy			Heterogeneity			Steiger pval
Exposure	Outcome	Egger_intercept	SE	p-value	Q	Q_df	Q_pval	Steiger pval
GNLY	Malignant neoplasm of bone and articular cartilage	−0.036	0.159	0.829	0.581	5	0.989	0
CNTN1		−0.113	0.203	0.615	0.330	4	0.988	2.10E−195
PCSK7		0.085	0.094	0.459	1.451	3	0.694	0
SCG3		−0.425	0.948	0.698	0.484	3	0.922
PDCD1LG2		−0.014	0.147	0.928	1.099	6	0.982
CXCL16		0.033	0.183	0.867	1.580	4	0.812	1.11E−122
ADAMTS5		−0.200	0.130	0.264	2.365	3	0.500	0

Co-localisation of the pQTLs with MNBAC risk loci

To further mitigate the potential confounding effects of LD on MR-based association results, we conducted co-localisation analyses for seven significant causal-associated proteins to determine whether genetic associations at the protein level with disease risk stemmed from the same causal variants. This analysis was performed only for proteins possessing a single instrumental variable and complete GWAS pooled statistics. Results revealed significant co-localisation evidence in three of the seven candidate proteins (Table 3): ADAMTS5 (PPH4 abf = 0.874), GNLY (PPH4 abf = 0.861) and PCSK7 (PPH4 abf = 0.827). These findings suggest that these three proteins share pathogenic SNP loci with MNBAC risk, thereby further supporting their potential role in disease pathogenesis.

Table 3.

Co-localisation analysis of MR-identified proteins with MNBAC risk loci.

Exposure	nsnps	PP.H0.abf	PP.H1.abf	PP.H2.abf	PP.H3.abf	PP.H4.abf
ADAMTS5	209885	0	0.119992236	0	0.016	0.874
CNTN1	196866	8.40E−119	0.273459118	7.85E−119	0.256	0.471
CXCL16	226970	7.10E−98	0.345968151	6.34E−98	0.309	0.345
GNLY	338925	0	0.124778077	0	0.014	0.861
PCSK7	660986	0	0.166927193	0	0.007	0.827
PDCD1LG2	224288	0	0.275188818	0	0.236	0.489
SCG3	216474	0	0.210942493	0	0.215	0.574

MR: Mendelian randomisation; MNBAC: malignant neoplasms of bone and articular cartilage; nsnps: number of single nucleotide polymorphisms.

Co-localisation of pQTLs with MNBAC risk loci

Using the Enrichr platform, we identified small-molecule compounds predicted to target the three causal proteins (ADAMTS5, GNLY and PCSK7). The screening results are summarised in Supplementary Table 5. After excluding compounds with p.adjust ≥ 0.05, we obtained four compounds targeting ADAMTS5, seven compounds targeting GNLY, two compounds targeting PCSK7, two compounds simultaneously targeting ADAMTS5 and PCSK7 and one compound targeting both GNLY and PCSK7.

To further explore potential therapeutic agents, we focused on compounds capable of simultaneously targeting two causal proteins and conducted protein–ligand docking analyses. For each protein–compound pair, we calculated the binding mode and binding energy and visualised the most favourable binding conformations to better characterise their structural interactions (Figure 4). Each docking analysis generated five binding modes, from which the lowest-energy conformation was selected for interpretation. Of these, aspirin exhibited a minimum binding energy of −4.9 kJ/mol for GNLY (Figure 4(a)) and −5.8 kJ/mol for PCSK7 (Figure 4(b)). Vitamin E showed the strongest binding affinity, with a minimum binding energy of −7.5 kJ/mol for both ADAMTS5 (Figure 4(c)) and PCSK7 (Figure 4(d)). Although no universally accepted threshold exists for docking energy, it is generally recognised that lower binding energies indicate stronger and more favourable protein–ligand interactions.

Figure 4.

Molecular docking analysis of causal proteins and small-molecule compounds. (a) Docking conformation of aspirin with GNLY. (b) docking conformation of aspirin with PCSK7. (c) docking conformation of vitamin E with ADAMTS5 and (d) docking conformation of vitamin E with PCSK7.

Discussion

In this study, we integrated nine large-scale cis-acting pQTL datasets and performed MR to systematically investigate causal relationships between circulating proteins and bone and cartilage tumours. We identified 171 candidate proteins with suggestive causal effects, among which ADAMTS5, CNTN1, CXCL16, GNLY, PCSK7, PDCD1LG2 and SCG3 retained significant causal effects on MNBAC risk after multiple testing correction. Further co-localisation analysis revealed strong causal evidence linking ADAMTS5, GNLY and PCSK7 to MNBAC development. These findings provide novel insights into the molecular mechanisms of MNBAC and suggest therapeutic intervention targets.

Previous research on MNBAC has predominantly focused on driver gene mutations, signalling pathway abnormalities and alterations in the tumour microenvironment.^25–27 However, translating these findings into effective molecular targeted therapies remains a considerable challenge. Unlike traditional observational studies, our MR analysis utilises genetic variants as instrumental variables, substantially reducing the influence of confounding factors and reverse causality. This approach enables more reliable inference of causal relationships between protein levels and disease risk.²⁸ Notably, this study exclusively used cis-pQTLs as instrumental variables. These genetic variants reside proximate to target genes, making them more likely to directly regulate corresponding protein expression levels while substantially mitigating the potential risk of pleiotropy.²⁹ This strategy enhances the robustness and biological interpretability of the findings.

First, among the three key proteins identified in this study, we found that only GNLY expression was positively correlated with the risk of developing MNBAC. GNLY, a cytotoxic protein secreted by activated cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells, belongs to the saposin-like family. Its primary function is mediating anti-tumour and anti-infective effects by disrupting target cell membrane integrity and inducing apoptosis.^30,31 In various solid tumours and haematological malignancies, GNLY expression is recognised as being closely associated with the anti-tumour activity of immune cells.³²

However, GNLY may play dual roles in tumourigenesis and progression. It enhances the killing capacity of NK cells and CTLs, aiding tumour cell clearance, and its expression is correlated with favourable prognosis in certain tumour types.³³ In contrast, persistently elevated GNLY expression may induce chronic inflammation and abnormalities in the immune microenvironment. Research indicates that GNLY can induce the release of cytokines such as interferon (IFN)-γ and TNF-α within inflammatory environments, thereby driving the immune microenvironment towards a pro-inflammatory state.³⁴ Persistence of this process may promote cellular genomic instability and abnormal proliferation, creating conditions conducive to tumourigenesis. In studies of bone and cartilage tumours, although direct reports on GNLY are limited, immunohistochemical and transcriptomic analyses have demonstrated significant NK cell and CTL infiltration in osteosarcoma tissues.^35–37 This immune cell infiltration reflects the body’s immune surveillance function against tumours; however, it may also lead to excessive immune-related inflammatory responses, causing tissue damage and remodelling of the tumour microenvironment. Combined with our findings, the positive correlation between GNLY upregulation and increased MNBAC risk suggests that GNLY does not merely function as a straightforward anti-tumour molecule. Instead, under certain circumstances, it may promote the initiation and progression of bone and cartilage tumours by modulating the immune microenvironment and inducing chronic inflammation. Consequently, GNLY may function as both a key effector molecule in immune responses and a potential immune-related risk factor. This apparent contradiction warrants further elucidation through functional experiments and validation using clinical samples.

We found that compared with GNLY, the upregulation of PCSK7 and ADAMTS5 expression was negatively correlated with the risk of developing MNBAC, suggesting that they exert a protective effect against the disease. PCSK7, belonging to the proprotein convertase family, is a calcium-dependent serine endopeptidase primarily functioning to cleave and activate various precursor proteins (such as growth factors, cytokines and receptors).^38,39 By regulating the maturation of multiple signalling molecules, PCSK7 plays a crucial role in cellular growth, differentiation and immune responses.

Existing research indicates that the role of PCSK7 in various tumours is inconsistent. In certain solid tumours (such as colorectal cancer, breast cancer and hepatocellular carcinoma), high PCSK7 expression is correlated with tumour progression, metastasis and poor prognosis^40–42; however, other studies suggest that PCSK7 suppresses inflammatory responses and abnormal cell proliferation by regulating lipid metabolism and immune responses.^43–45 This apparent contradiction may be closely linked to the tumour type and tissue microenvironment. Direct studies on PCSK7 and MNBAC remain relatively scarce. Nevertheless, given that PCSK7 regulates multiple molecules closely associated with bone tumour development—including matrix metalloproteinases (MMPs) and TNF receptors^46–48—its downregulation may lead to increased matrix degradation and enhanced pro-inflammatory signalling, thereby promoting tumour progression. Conversely, high PCSK7 expression may exert tumour-suppressive effects by maintaining ECM homeostasis, reducing pro-inflammatory factor activation and modulating the tumour immune microenvironment. This aligns with our MR findings indicating a protective causal relationship for PCSK7. Consequently, it may be inferred that PCSK7’s role in MNBAC extends beyond protein processing, potentially counteracting the risk effects of pro-inflammatory molecules such as GNLY by suppressing inflammatory responses and maintaining tissue homeostasis. However, this hypothesis requires further validation through functional experiments and clinical sample analysis.

ADAMTS5 is a metalloproteinase primarily degrading aggregated proteoglycans, belonging to the ADAMTS protease family. It is extensively involved in the metabolism, remodelling and maintenance of ECM homeostasis within the cartilage matrix.⁴⁹ In degenerative diseases such as osteoarthritis, ADAMTS5 is recognised as a core driver of cartilage degradation and disease progression, with its excessive activation leading to articular cartilage damage.⁵⁰ However, the role of ADAMTS5 in tumours exhibits tissue specificity and complex bidirectional regulatory effects, suggesting that its function extends beyond ECM degradation. Previous studies have indicated that ADAMTS5 and other ADAMTS family members not only participate in ECM remodelling but also influence tumour growth and invasion by regulating tumour-associated angiogenesis, cell migration, apoptosis and signalling pathways such as Wnt/β-catenin and TGF-β.^51,52 In multiple solid tumours, reduced ADAMTS5 expression is correlated with increased tumour invasiveness, active angiogenesis and poor prognosis.^53–55 Potential mechanisms may include the maintenance of ECM integrity to restrict tumour cell migration, inhibition of angiogenesis by regulating matrix growth factor activity and improvement of the local immune microenvironment to indirectly suppress tumour progression.⁵⁶

Additionally, we performed small-molecule compound screening using the Enrichr database for three key proteins to explore their translational potential. The screening results revealed that small-molecule compounds such as aspirin and vitamin E exhibit favourable binding affinity with these proteins. Aspirin, a widely used nonsteroidal anti-inflammatory drug, has been demonstrated to reduce the risk of multiple cancers by inhibiting cyclooxygenase activity. Research has shown that aspirin may play a role in suppressing cancer metastasis by modulating immune system function.⁵⁷ Vitamin E is a fat-soluble antioxidant possessing diverse biological activities. Different forms of vitamin E, particularly γ- and δ-tocotrienols, demonstrate potential in cancer prevention.⁵⁸ However, binding energy values serve only as preliminary screening indicators; further validation of their biological activity and clinical utility requires in vitro and in vivo experimentation.

Conclusion

Using MR analysis, this study identified three proteins causally associated with MNBAC. This discovery not only provides novel molecular markers for the early diagnosis and risk prediction of MNBAC but also reveals potential therapeutic targets, laying the groundwork for the development of protein-based intervention strategies and targeted therapies.

Supplemental Material

sj-xlsx-1-imr-10.1177_03000605261416727 - Supplemental material for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage

Supplemental material, sj-xlsx-1-imr-10.1177_03000605261416727 for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage by Yun Wang, Ruhao Zhou, Zhuangzhuang Wu and Zhi Lv in Journal of International Medical Research

Supplemental Material

sj-xlsx-2-imr-10.1177_03000605261416727 - Supplemental material for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage

Supplemental material, sj-xlsx-2-imr-10.1177_03000605261416727 for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage by Yun Wang, Ruhao Zhou, Zhuangzhuang Wu and Zhi Lv in Journal of International Medical Research

Supplemental Material

sj-xlsx-3-imr-10.1177_03000605261416727 - Supplemental material for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage

Supplemental material, sj-xlsx-3-imr-10.1177_03000605261416727 for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage by Yun Wang, Ruhao Zhou, Zhuangzhuang Wu and Zhi Lv in Journal of International Medical Research

Supplemental Material

sj-xlsx-4-imr-10.1177_03000605261416727 - Supplemental material for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage

Supplemental material, sj-xlsx-4-imr-10.1177_03000605261416727 for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage by Yun Wang, Ruhao Zhou, Zhuangzhuang Wu and Zhi Lv in Journal of International Medical Research

Supplemental Material

sj-xlsx-5-imr-10.1177_03000605261416727 - Supplemental material for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage

Supplemental material, sj-xlsx-5-imr-10.1177_03000605261416727 for Evaluating the causal effects of circulating plasma proteins on the risk of malignant neoplasms of bone and articular cartilage by Yun Wang, Ruhao Zhou, Zhuangzhuang Wu and Zhi Lv in Journal of International Medical Research

Footnotes

Acknowledgements

We are grateful to the nine proteomics GWAS and FinnGen research teams for providing aggregated data for this study.

Author contributions

Yun Wang: Concepts, study design, data acquisition, data analysis, statistical analysis, manuscript preparation, manuscript editing and manuscript review

RuHao Zhou: Data acquisition, data analysis, manuscript editing and manuscript review

Zhuangzhuang Wu: Statistical analysis, manuscript preparation and manuscript review

Zhi Lv: Concept, study design, manuscript editing, manuscript review and guarantor

Data availability

Data are available in a public, open access repository.

Declaration of conflicting interests

All authors of this study declare that there were no conflicts of interest during the research and writing of this paper.

Ethics approval

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

Funding

Not applicable.

Supplemental material

Supplemental material for this article is available online.

ORCID iD

Zhi Lv

References

Yang

, et al. Rheumatoid arthritis increases the risk of malignant neoplasm of bone and articular cartilage: a two-sample bidirectional Mendelian randomization study. Arthritis Res Ther 2023; 25: 219.

Choi

JY.

The 2020 WHO classification of tumors of bone: an updated review. Adv Anat Pathol 2021; 28: 119–138.

Marturano-Kruik

Villasante

Yaeger

, et al. Biomechanical regulation of drug sensitivity in an engineered model of human tumor. Biomaterials 2018; 150: 150–161.

Zhang

, et al. Pelvic-girdle reconstruction with three-dimensional-printed endoprostheses after limb-salvage surgery for pelvic sarcomas: current landscape. Br J Surg 2023; 110: 1712–1722.

Wang

, et al. Three-dimensional multimodality image reconstruction as teaching tool for case-based learning among medical postgraduates: a focus on primary pelvic bone tumour education. BMC Med Educ 2023; 23: 944.

Jones

DTW

Kieran

Bouffet

, et al. Pediatric low-grade gliomas: next biologically driven steps. Neuro Oncol 2018; 20: 160–173.

Wang

Hwang

Guha

, et al. High-throughput chemical screening identifies focal adhesion kinase and aurora kinase B inhibition as a synergistic treatment combination in Ewing sarcoma. Clin Cancer Res 2019; 25: 4552–4566.

Czarnecka

Synoradzki

Firlej

, et al. Molecular biology of osteosarcoma. Cancers (Basel) 2020; 12: 2130.

Verrecchia

Rédini

Transforming growth factor-β signaling plays a pivotal role in the interplay between osteosarcoma cells and their microenvironment. Front Oncol 2018; 8: 133.

10.

Lamora

Talbot

Mullard

, et al. TGF-β signaling in bone remodeling and osteosarcoma progression. J Clin Med 2016; 5: 96.

11.

Liu

Wang

, et al. The role of programmed cell death in osteosarcoma: from pathogenesis to therapy. Cancer Med 2024; 13: e7303.

12.

Yazdanpanah

Wang

, et al. Clinically relevant circulating protein biomarkers for type 1 diabetes: evidence from a two-sample Mendelian randomization study. Diabetes Care 2022; 45: 169–177.

13.

Ferkingstad

Sulem

Atlason

, et al. Large-scale integration of the plasma proteome with genetics and disease. Nat Genet 2021; 53: 1712–1721.

14.

Sheehan

Didelez

Burton

, et al. Mendelian randomisation and causal inference in observational epidemiology. PLoS Med 2008; 5: e177.

15.

Cheng

Yang

, et al. Mendelian randomisation analysis of causal association between lifestyle, health factors, and keratoconus. Bioengineering (Basel) 2024; 11: 221.

16.

Zhang

Xie

Wen

, et al. Evaluating the causal effect of circulating proteome on the risk of osteoarthritis-related traits. Ann Rheum Dis 2023; 82: 1606–1617.

17.

Burgess

Butterworth

Thompson

SG.

Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 2013; 37: 658–665.

18.

Hemani

Tilling

Davey Smith

Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet 2017; 13: e1007081.

19.

Schmidt

Dudbridge

Mendelian randomization with Egger pleiotropy correction and weakly informative Bayesian priors. Int J Epidemiol 2018; 47: 1217–1228.

20.

Bowden

Davey Smith

Haycock

, et al. Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016; 40: 304–314.

21.

Cullell

Gallego-Fábrega

Cárcel-Márquez

, et al. ICA1L is associated with small vessel disease: a proteome-wide association study in small vessel stroke and intracerebral haemorrhage. Int J Mol Sci 2022; 23: 3161.

22.

Giambartolomei

Vukcevic

Schadt

, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 2014; 10: e1004383.

23.

Burley

Bhikadiya

, et al. RCSB Protein Data Bank: powerful new tools for exploring 3D structures of biological macromolecules for basic and applied research and education in fundamental biology, biomedicine, biotechnology, bioengineering and energy sciences . Nucleic Acids Res 2021; 49: D437–D451.

24.

Chen

, et al. Mendelian randomization analysis to identify potential drug targets for osteoarthritis. PLoS One 2025; 20: e0316824.

25.

Xie

Bian

, et al. A comprehensive understanding of the genomic bone tumor landscape: a multicenter prospective study. Front Oncol 2022; 12: 835004.

26.

Sun

Zhang

Fang

, et al. Abnormal signal pathways and tumor heterogeneity in osteosarcoma. J Transl Med 2023; 21: 99.

27.

Chen

Cui

, et al. Single-cell transcriptomics reveals metabolic remodeling and functional specialization in the immune microenvironment of bone tumors. J Transl Med 2025; 23: 554.

28.

Davey Smith

Hemani

Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet 2014; 23: R89–R98.

29.

Montgomery

Dermitzakis

ET.

From expression QTLs to personalized transcriptomics. Nat Rev Genet 2011; 12: 277–282.

30.

PEñA

Krensky

AM.

Granulysin, a new human cytolytic granule-associated protein with possible involvement in cell-mediated cytotoxicity. Semin Immunol 1997; 9: 117–125.

31.

Clayberger

Krensky

AM.

Granulysin. Curr Opin Immunol 2003; 15: 560–565.

32.

Stenger

Hanson

Teitelbaum

, et al. An antimicrobial activity of cytolytic T cells mediated by granulysin. Science 1998; 282: 121–125.

33.

Tewary

Yang

De la Rosa

, et al. Granulysin activates antigen-presenting cells through TLR4 and acts as an immune alarmin. Blood 2010; 116: 3465–3474.

34.

Kaspar

Okada

Kumar

, et al. A distinct pathway of cell-mediated apoptosis initiated by granulysin. J Immunol 2001; 167: 350–356.

35.

Buddingh

Kuijjer

Duim

, et al. Tumor-infiltrating macrophages are associated with metastasis suppression in high-grade osteosarcoma: a rationale for treatment with macrophage activating agents. Clin Cancer Res 2011; 17: 2110–2119.

36.

Gomez-Brouchet

Illac

Gilhodes

, et al. CD163-positive tumor-associated macrophages and CD8-positive cytotoxic lymphocytes are powerful diagnostic markers for the therapeutic stratification of osteosarcoma patients: an immunohistochemical analysis of the biopsies fromthe French OS2006 phase 3 trial. Oncoimmunology 2017; 6: e1331193.

37.

Zhang

Huang

, et al. Characterization of the tumour microenvironment phenotypes in malignant tissues and pleural effusion from advanced osteoblastic osteosarcoma patients. Clin Transl Med 2022; 12: e1072.

38.

Artenstein

Opal

SM.

Proprotein convertases in health and disease. N Engl J Med 2011; 365: 2507–2518.

39.

Thomas

Furin at the cutting edge: from protein traffic to embryogenesis and disease. Nat Rev Mol Cell Biol 2002; 3: 753–766.

40.

Scamuffa

Calvo

Chrétien

, et al. Proprotein convertases: lessons from knockouts. FASEB J 2006; 20: 1954–1963.

41.

Prabhala

Duval

Evagelidis

, et al. Shedding of GPP130 by PC7 and furin: potential implication in lung cancer progression. Int J Mol Sci 2025; 26: 6164.

42.

Bassi

Mahloogi

Klein-Szanto

AJ.

The proprotein convertases furin and PACE4 play a significant role in tumor progression. Mol Carcinog 2000; 28: 63–69.

43.

Sachan

Le Dévéhat

Roubtsova

, et al. PCSK7: A novel regulator of apolipoprotein B and a potential target against non-alcoholic fatty liver disease. Metabolism 2024; 150: 155736.

44.

Turpeinen

Oksanen

Kivinen

, et al. Proprotein convertase subtilisin/kexin type 7 (PCSK7) is essential for the zebrafish development and bioavailability of transforming growth factor β1a (TGFβ1a). J Biol Chem 2013; 288: 36610–36623.

45.

Rose

Duhamel

Rodet

, et al. The role of proprotein convertases in the regulation of the function of immune cells in the oncoimmune response. Front Immunol 2021; 12: 667850.

46.

Kessenbrock

Plaks

Werb

Matrix metalloproteinases: regulators of the tumor microenvironment. Cell 2010; 141: 52–67.

47.

Łukaszewicz-Zając

Szmitkowski

Litman-Zawadzka

, et al. Matrix metalloproteinases and their tissue inhibitors in comparison to other inflammatory proteins in gastric cancer (GC). Cancer Invest 2016; 34: 305–312.

48.

Jang

Kim

Chung

YH.

Clinical aspects of tumor necrosis factor-α signaling in hepatocellular carcinoma. Curr Pharm Des 2014; 20: 2799–2808.

49.

Stanton

Rogerson

East

, et al. ADAMTS5 is the major aggrecanase in mouse cartilage in vivo and in vitro. Nature 2005; 434: 648–652.

50.

Verma

Dalal

ADAMTS-4 and ADAMTS-5: key enzymes in osteoarthritis. J Cell Biochem 2011; 112: 3507–3514.

51.

Rocks

Paulissen

El Hour

, et al. Emerging roles of ADAM and ADAMTS metalloproteinases in cancer. Biochimie 2008; 90: 369–379.

52.

Wang

Xing

, et al. Development and validation of a basement membrane-associated immune prognostic model for hepatocellular carcinoma. Transl Gastroenterol Hepatol 2025; 10: 28.

53.

Huang

Sun

Chen

, et al. ADAMTS5 acts as a tumor suppressor by inhibiting migration, invasion and angiogenesis in human gastric cancer. Gastric Cancer 2019; 22: 287–301.

54.

Lian

Zhou

Liu

Identification and validation of the TRHDE-AS1/hsa-miR-449a/ADAMTS5 axis as a novel prognostic biomarker in prostate cancer. Biofactors 2024; 50: 1251–1267.

55.

Xiong

Yang

, et al. Lost expression of ADAMTS5 protein associates with progression and poor prognosis of hepatocellular carcinoma. Drug Des Devel Ther 2015; 9: 1773–1783.

56.

Bacchetti

Yuan

Rainero

ADAMTS proteases: their multifaceted role in the regulation of cancer metastasis. Dis Res 2024; 4: 40–52.

57.

Petrelli

Ghidini

Antista

, et al. Effect of aspirin use on cancer incidence and mortality: an updated systematic review and meta-analysis. Public Health 2025; 248: 105924.

58.

De Oliveira

Oliveira

IKF

Pereira

, et al. Consumption and supplementation of vitamin E in breast cancer risk, treatment, and outcomes: a systematic review with meta-analysis. Clin Nutr ESPEN 2023; 54: 215–226.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.55 MB

0.57 MB

0.28 MB

0.01 MB