Potential protective function and causal association of LEPR in osteosarcoma

Abstract

The highly heterogeneous properties and dynamic immunogenic features of osteosarcoma (OS) lead to limited therapeutic efficacy. The oncogenetic mechanisms of OS await further explanation, and new insights into therapeutic targets for OS remain urgent. We used single-cell RNA-sequencing data to identify cell type-specific differentially expressed genes between normal human osteocytes and OS cells. Subsequently, we analyzed the differential expression of leptin receptor (LEPR) in OS cell lines through in vitro experiments. We assessed the prognostic potential of LEPR and immune cell infiltration status associated with differential LEPR expression in OS patients. By integrating protein quantitative trait loci and methylation quantitative trait loci datasets with OS genome-wide association studies summary statistics, we performed Mendelian randomization analysis to elucidate the causal effect of LEPR on OS risk and identify cytosine-phosphate-guanine sites that may influence LEPR expression and OS risk. We observed lower LEPR expression in OS cell clusters compared to normal osteocyte clusters. RT-qPCR results indicated that the LEPR gene showed consistent differential expression in OS cells. Survival analysis revealed that lower LEPR expression was associated with worse survival outcomes in OS patients. OS patients with high LEPR expression tended to exhibit more immune cell infiltration. Mendelian randomization analysis indicated that genetically-determined LEPR was inversely associated with OS risk. We identified 8 methylation cytosine-phosphate-guanine sites within the LEPR promoter region causally associated with both lower LEPR protein expression and higher OS risk. Our integrated multi-omics analyses supported a causal association between LEPR expression and OS risk, suggesting that LEPR may offer protective benefits for OS patients. The relationship between LEPR and OS risk highlights the potential of LEPR as an early prognostic biomarker and may provide new insights into the etiology of OS.

1. Introduction

Osteosarcoma (OS) is the most common bone sarcoma among adolescents and young adults, accounting for 1% of malignant neoplasms globally.^[1] It has a high propensity for local invasion, metastasis, and recurrence after therapy, posing considerable challenges in terms of OS therapy and prognosis.^[2] With the current combination of surgery with neoadjuvant chemotherapy, immunotherapy, and targeted therapy for OS treatment, the 5-year survival rate, which remains approximately 60 to 70%, has not notably improved for decades.^[3] The existence of tumor histological heterogeneity and the complexity of the molecular and genetic aberrations associated with tumorigenesis hinder breakthroughs in OS treatment. More recently, the contribution of molecular profiling together with large, well-annotated tissue banks to OS biology has led to the exploration of potential therapeutic targeting.^[4] To improve the survival of OS patients, novel biomarker-based clinical trials designed to identify efficacious therapies are necessary.

Leptin receptor (LEPR), encoded by the db gene, belongs to the class I cytokine receptor family.^[5] Many previous studies have focused on LEPR in energy homeostasis and appetite regulation, the abnormal regulation of which has been implicated in obesity.^[6] Recently, the role of LEPR expressed in diverse malignant cancer types in tumorigenesis has also attracted attention for potential clinical applications.^[7] Results indicating the negative correlation of LEPR expression with tumor metastasis and aggressiveness have been reported.^[8,9] cytosine-phosphate-guanine (CpG) hypermethylation in LEPR is prone to lower recurrence-free survival rates and worse prognosis in renal cell carcinoma patients.^[10] LEPR also marks bone marrow mesenchymal stromal cells, which are the major source of osteoblasts.^[11] LEPR antagonizes osteogenesis by regulating mesenchymal stromal cells differentiation.^[12] Hence, further exploration to elucidate the prognostic significance of LEPR in OS and the association between LEPR and OS is warranted.

Single-cell RNA-sequencing (scRNA-seq) can accurately characterize and classify each cell from OS, unraveling the functional diversity of heterogeneous cell populations.^[13] Mendelian randomization (MR) serves as a statistical approach based on genome-wide association studies (GWAS) summary data to provide genetic support for causation, where genetic variation should satisfy 3 key assumptions: relevance, independence, and exclusion restriction. MR has also been implemented to systematically integrate multiple quantitative trait loci (QTL) data with GWAS data to identify candidate biomarkers and therapeutic targets for various tumors.^[14–16]

In this study, we mapped scRNA-seq of normal human subchondral bone tissue and OS at single-cell resolution to profile OS-dependent cell cluster specificity, in which low expression of LEPR was observed in OS cell clusters and validated in OS cell lines by RT-qPCR. Subsequently, we validated the predictive potential of LEPR for survival and explored its association with immune infiltration landscape of OS within TARGET database. Lastly, we employed MR to deduce the potential causality between LEPR and OS risk by integrating human QTL data with GWAS summary data.

2. Material and methods

2.1. Bioinformatics analysis and validation

2.1.1. scRNA-seq data processing

Considering that OS cells originate from bone mesenchymal stem cells with osteoblastic lineage commitment,^[17] public scRNA-seq data from GSE196678 (4 samples from human subchondral bone tissues, GPL24676) and GSE162454 (6 samples from human OS tissues, GPL24676) were downloaded from the Gene Expression Omnibus database. The Seurat and Harmony packages were used to analyze and integrate these scRNA-seq data. Batch effects were corrected across both datasets using the harmony function (max.iter.harmony = 20). Ineligible cells and low-quality cells (min.cells < 3; min.features < 200; and percent.mt > 15%) were filtered out. After the remaining cells were normalized, the top 2000 genes with highly variable characteristics were selected to perform a principal component analysis. On the basis of the top 30 principal components, the t-distributed stochastic neighbor embedding (t-SNE) algorithm was utilized for dimension reduction. According to marker genes from the COSine similarity-based marker Gene identification package,^[18] the cell clusters were manually annotated as different cell types. Annotation of the cell clusters to cell types relied on their known lineage markers: endothelial cells, von Willebrand factor, plasma vesicle-associated protein, and carcinoembryonic antigen-related cell adhesion molecule-1; myeloid cells, CD14, CD68 and CD83; myogenic cells, myogenin and myogenic differentiation 1; lymphoid cells, CD3, CD8A, killer cell lectin-like receptor subfamily B member 1, and B-cell maturation antigen; bone-related cells, lumican, collagen type 1 alpha 1 and alpha 2 chains, SRY-Box Transcription Factor 9, RUNX Family Transcription Factor 2, and alkaline phosphatase. As a gene set search engine containing rich synthesized information about mammalian gene sets, Enrichr (https://maayanlab.cloud/Enrichr/) allows for gene set enrichment analyses with interactive visualizations.

2.1.2. TARGET data processing and prognostic analysis

Expression profiles and phenotype data of OS were obtained from the TARGET database (https://ocg.cancer.gov/programs/target), which contained 89 OS patients. Survival and clinical information of the OS patients, including survival status, survival time, gender, age (≤ 14 and > 14), metastasis, surgery, and chemotherapy (received or not received), was selected. In order to more stably reflect the central tendency of LEPR expression profiles, the median LEPR expression was used to divide OS patients into high- and low-expression groups. Kaplan–Meier analysis was performed to evaluate the correlation of LEPR expression with overall survival and event-free survival of OS patients using the survival package. Additionally, differences in clinical characteristics were assessed between the LEPR high- and low-expression groups. Univariate and multivariate Cox regression analyses were conducted to identify the independent prognostic factors associated with overall survival (P < .05). Based on the above filtering prognostic factors, the nomogram for survival prediction was developed and presented.

2.1.3. Gene set enrichment analysis (GSEA) and tumor microenvironment (TME) estimation

To identify signaling pathways that were enriched in OS patients with different LEPR expression, the differentially expressed genes (DEGs) between the LEPR high- and low-expression groups were selected using the limma R package, and GSEA was conducted using clusterProfiler package to detect annotated enrichment pathways. Based on the ESTIMATE package, the stromal scores, immune scores and ESTIMATE scores in TME were computed between the LEPR high- and low-expression OS patients. The 22 infiltrating immune cells abundances in OS were estimated using the CIBERSORT algorithm. Comparisons of immune cell infiltration abundances were performed between OS patients with differential LEPR expressions based on Spearman correlation analysis. Considering that immune checkpoint expressions may predict immune checkpoint blockade therapy, the correlations between LEPR expression and immune checkpoint expressions related to cancer were also analyzed with Spearman correlation.

2.1.4. Sensitivity analysis to chemotherapy

Given the introduction of adjuvant chemotherapy to improve the overall survival of OS patients, chemotherapy drug sensitivity was assessed between the LEPR high- and low-expression groups. The half-maximal inhibitory concentration (IC50) of several well-known chemotherapeutic drugs was calculated to predict the sensitivity of OS patients with differential LEPR expression to chemotherapy.

2.1.5. Cell cultures

Human OS cell lines (U2OS and HOS) and normal osteoblast cell line (hFOB 1.19) were provided by the Sigma-Aldrich LLC. U2OS cells were cultured in DMEM medium (Gibco) supplemented with 10% fetal bovine serum, L-glutamine and 1% penicillin/streptomycin at 37°C with 5% CO₂, while HOS cells were cultured in MEM (Gibco) supplemented with 10% fetal bovine serum, L-glutamine, sodium pyruvate and 1% penicillin/streptomycin. hFOB 1.19 cells were cultured in DMEM/F-12 medium (Gibco) supplemented with 10% fetal bovine serum, L-glutamine, and 1% penicillin/streptomycin at 35°C with 5% CO₂.

2.1.6. RT-qPCR assays

Total RNA from cell lines was extracted using TRIzol reagent (Invitrogen), followed by reverse transcription using a cDNA synthesis kit (Takara). Subsequently, TB Green Fast qPCR Mix (Takara) was used for RT-qPCR on the StepOnePlus Real-Time PCR System (ABI7500). RT-qPCR reactions were run in triplicate, and transcript level of mRNA was normalized to GAPDH using the 2^−∆∆Ct method. RT-qPCR result for LEPR was recorded from 3 or more independent cell cultures. The following LEPR gene primers were used: forward 5’-GTTTGAAAATCACATCTGGTGGAGTA-3’ and reverse 5’-CATATTTATGGGCTGAACTGACATTAG-3’.

2.2. MR analysis

2.2.1. Study population and data source

GWAS summary statistics for OS outcome were derived from FinnGen cohorts with 64 cases and 314,193 controls of Finnish individuals (https://r10.finngen.fi). Protein QTL (pQTL) data for LEPR were obtained from deCODE database (https://www.decode.com/summarydata/), which consisted of 18,084 single nucleotide polymorphisms (SNPs) associated with 4907 aptamer-measured whole blood circulating protein levels in 35,559 Icelanders. Methylation QTL (mQTL) data for LEPR were collected from the Genetics of DNA Methylation Consortium (GoDMC; http://mqtldb.godmc.org.uk/downloads), which presents a database of > 270,000 independent SNPs derived from the quantification of whole blood 420,509 DNA methylation (DNAm) sites in 27,750 European individuals.

2.2.2. Proteome-wide MR analysis

The platform ID for LEPR was mapped to its encoding protein based on annotations from the deCODE database. For LEPR as exposure, instrumental variables (IVs) selection was acquired from cis-pQTLs (SNPs < 1 Mb from LEPR) of deCODE database. To ensure the accuracy of causal inferences, SNPs associated with LEPR were selected (P-value threshold < 5 × 10⁻⁸), and then independent (r2 < 0.001) SNPs for LEPR were identified with PLINK v1.90 on the basis of genotypes from the 1000 Genomes Project. To minimize potential weak IVs bias, an F-statistic > 10 was considered to be sufficient for MR analysis. Finally, 7 IVs were kept for causal effect of LEPR (exposure) on OS (outcome) by TwoSampleMR package. A Manhattan plot was generated for independent SNPs using the CMplot package. The inverse-variance weighted (IVW) method was performed to obtain overall MR estimates by combining the Wald ratios of cis-pQTLs. Sensitivity analyses using pleiotropy-robust methods, including MR-Egger, simple mode, weighted median, and weighted mode, were conducted. To assess heterogeneity among IVs and horizontal pleiotropy of the cis-pQTLs used as IVs, the Cochran Q statistic from IVW and MR-Egger, and the intercept term in MR-Egger were deployed, with a P-value < .05, respectively. Additionally, the MR-Steiger directionality test was applied to infer the direction of causality between LEPR plasma protein and OS risk.

2.2.3. Methylome-wide MR analysis

Due to genetic variants associated with DNAm levels often incorporating with complex traits and disease susceptibility,^[19] this type of methodology can be extended to assess the potential influence of DNAm on LEPR protein expression and OS outcome. Using GoDMC summary data, cis-mQTLs (SNPs < 1 Mb from CpG sites) associated with each LEPR-related CpG site were retrieved as IVs (P-value threshold < 5 × 10⁻⁸). Multiple MR methods, including IVW, MR-Egger, simple mode, weighted median, and weighted mode, were employed to evaluate associations of LEPR-related CpG sites with LEPR cis-pQTL and OS outcome. For CpG sites with only 1 IV, the Wald ratio method was applied. Associations with Bonferroni-corrected P < .05/47 = .00106 were considered significant.

2.2.4. Bayesian colocalization analysis

Bayesian colocalization analysis evaluates the presence of local shared genetic variants for 2 traits, aiding in identifying MR associations potentially influenced by linkage disequilibrium (LD). Based on the coloc package, colocalization analyses of LEPR protein, LEPR-related CpG sites, and OS outcome were conducted on the genomic region within ± 250 kb from LEPR gene region. Default priors were applied, with p1 = p2 = 1 × 10⁻⁴, and p12 = 1 × 10⁻⁵. The posterior probabilities (PP.H1, PP.H2, PP.H3, and PP.H4) were quantified using approximate Bayesian factorization. Results with PP.H3 + PP.H4 or PP.H4/(PP.H3 + PP.H4) > 0.8 within the LEPR gene locus represented the probability of colocalization conditional on the presence of a causal variant for the outcome.

2.3. Statistical analysis

All data processing and statistical analyses were conducted in R Studio (version 4.3.3). A threshold of P < .05 and log2FoldChange > 1 was used to identify DEGs. Throughout, P < .05 was considered statistically significant.

2.4. Ethical statement

Ethical approval was not necessary, because this study did not include any experiments involving humans or animals.

3. Results

3.1. Profiling of scRNA-seq and screening of DEGs

Integrated scRNA-seq data were collected from human subchondral bone tissue (GSE196678) and OS (GSE162454). Following quality filtering and batch correction (Fig. S1A and B, Supplemental Digital Content 1), 53,964 cells including 14,175 bone tissue cells and 39,789 OS tissue cells were obtained for subsequent transcriptomic analysis. After dimensionality reduction and t-SNE algorithm-based clustering, cell clusters were annotated as endothelial cells, myeloid cells, myogenic cells, lymphoid cells, and bone-related cells based on their lineage markers (Fig. S1C and D, Supplemental Digital Content 1). Due to similar tissue origin, OS cells and normal osteocytes were distributed together as shown in Figure 1A. The FindMarkers embedded in Seurat was applied to identify DEGs between OS cells and normal osteocytes. A total of 317 DEGs were obtained (|log2FC| > 1, P < .01), with 147 elevated and 170 decreased in the OS cells. These DEGs were input into the MGI Mammalian Phenotype database for functional enrichment analysis, which indicated enrichment of abnormal bone remodeling, decreased bone mineralization, and abnormal osteoblast morphology (Fig. 1B). Notably, the reduction in LEPR was observed in OS cells compared to normal osteocytes (Fig. 1C). RT-qPCR results also revealed that downregulation of LEPR was observed in U2OS and HOS cells compared to hFOB1.19 cells (Fig. 1D).

Figure 1.

Single-cell transcriptomic analysis of osteosarcoma (OS). (A) The t-SNE plots of 5 main cell types in normal subchondral bone tissue (GSE196678) and OS (GSE162454) were represented by different colors. (B) Enrichment analysis of significant different genes in the MGI Mammalian Phenotype databases. (C) Volcano plot of significant different genes between normal osteocytes and OS cells. The circle indicates the low LEPR expression in OS. (D) The result of RT-qPCR. *P < .05; **P < .01. LEPR = leptin receptor.

3.2. Survival and prognostic analysis of LEPR in OS

Through data mining in TARGET database, the Kaplan–Meier curve indicated that OS patients with higher LEPR expression had significantly better overall survival and event-free survival than those with lower LEPR expression (Fig. 2A, B). The clinical characteristics of LEPR high- and low-expression patients in TARGET database were probed, grouped by gender, age, metastasis, surgery, and chemotherapy (Fig. 2C). LEPR expression revealed significant differences in different metastasis (P = .0014), and chemotherapy statuses (P = .00049; Fig. 2D, E). Univariate and multivariate Cox regression analyses, with 15 outcome events, demonstrated that LEPR expression (uni-Cox: HR = 0.199, P = .011; multi-Cox: HR = 0.187, P = .027), metastasis (uni-Cox: HR = 13.745, P < .001; multi-Cox: HR = 5.872, P = .011), and chemotherapy (uni-Cox: HR = 0.044, P = .004; multi-Cox: HR = 0.029, P = .011) may serve as potential prognostic factors for OS patients (Table 1). Thus, according to these 3 prognostic factors, a predictive nomogram was generated to predict the 1-, 3-, and 5-year survival rates for OS patients, by adding up the corresponding score for each factor (Fig. 2F).

Figure 2.

Association of LEPR expression with clinical prognosis in OS. (A, B) overall survival and progression-free survival Kaplan–Meier survival curves in OS patients. (C) Heatmap of clinical characteristics in OS patients with differential LEPR expression. (D, E) Differential LEPR expression with different metastasis and chemotherapy features. (F) Nomogram for prediction of the 1-, 3-, or 5-year survival rates for OS patients. LEPR = leptin receptor.

Table 1.

Univariate and multivariate Cox analyses of clinical characteristics for overall survival in osteosarcoma (OS) patients.

Characteristics	Univariate Cox analysis			Multivariate Cox analysis
	HR	95% CI	P value	HR	95% CI	P alue
LEPR	0.198	0.057–0.689	.011	0.187	0.042–0.826	.026
Gender	1.302	0.444–3.815	.629	2.167	0.605–7.751	.234
Age	0.995	0.892–1.111	.938	1.017	0.913–1.133	.751
Metastasis	13.744	3.904–48.379	4.471E−05	5.872	1.504–22.927	.011
Surgery	0.426	0.055–3.279	.412	0.498	0.034–7.138	.607
Chemotherapy	0.044	0.005–0.366	.003	0.029	0.002–0.448	.011

LEPR = leptin receptor.

3.3. Enrichment analysis and immunological properties of LEPR in OS

To further elucidate the potential biological function of LEPR in OS, GSEA analysis based on LEPR differential expression was conducted. The top 5 LEPR-related signaling pathways were mainly enriched in immune response-related processes, including cell adhesion molecules, antigen processing and presentation, leukocyte transendothelial migration, primary immunodeficiency, and hematopoietic cell lineage (Fig. 3A). Then, tumor immune status of OS patients was assessed to understand the immune characteristics associated with LEPR in TME. ESTIMATE analysis confirmed that OS patients with high LEPR expression obtained higher StromalScore, ImmuneScore, and ESTIMATEScore (Fig. 3B). CIBERSORT algorithm demonstrated that the infiltration of NK cells activated and T cells CD4 memory activated was positively correlated with LEPR expression, while the infiltration of mast cells activated, T cells regulatory, monocytes, and B cells memory was negatively correlated with LEPR expression (Fig. 3C). In addition, correlation analysis revealed that positive associations existed between LEPR expression and 9 immune checkpoint expressions, except for ICOSLG and TNFRSF18 (Fig. 3D).

Figure 3.

Estimation of immune infiltration and comparison of chemotherapy outcomes. (A) Visualization of immune response-related pathways. (B) TME scores for differential LEPR expression. (C) Correlation analysis chart of immune cell infiltration. (D) Heatmap for correlation of LEPR expressions with immune checkpoints. (E–H) Boxplots for chemotherapy sensitivity with differential LEPR expression. LEPR = leptin receptor, TME = tumor microenvironment.

3.4. Efficacy analyses of chemotherapy

To identify the preclinical rational for targeting LEPR to enhance chemotherapy sensitivity, the pRRophetic algorithm was employed to investigate the correlation between differential LEPR expression and IC50 values for common antitumor chemotherapeutic drugs. Differential sensitivity analysis indicated that OS patients with low LEPR expression exhibited higher sensitivity in response to docetaxel, paclitaxel, crizotinib, and mitoxantrone, as reflected by lower IC50 values (Fig. 3E–H).

3.5. Proteome-wide MR analysis of LEPR in association with OS

Using the deCODE database, after performing LD clumping in LEPR cis-pQTLs, 7 independent SNPs (rs80086931, rs7535099, rs138267513, rs115480944, rs75379640, rs10889540, and rs72675459) were used as IVs in MR analysis, which were located on chromosome 1, as shown in the Manhattan plot (Fig. 4A). Findings from MR-proteomics revealed LEPR protein expression was negatively associated with OS risk for IVW method (OR = 0.391, 95% CI = 0.216–0.707, P-value = .002), indicating that genetically predicted lower level of LEPR plasma protein was associated with a higher risk of OS. This MR association was directionally consistent for sensitivity analyses, including weighted median (OR = 0.415, 95% CI = 0.208–0.829, P-value = .015), weighted mode (OR = 0.401, 95% CI = 0.199–0.806, P-value = .026), although not for MR Egger and simple mode. There was insufficient evidence for heterogeneity (IVW, Q 2.095, P = .911; MR-Egger, Q 1.257, P = .939) and horizontal pleiotropy (MR-Egger intercept, −0.181, P = .402) in the IVs (Table 2). The risk of OS was heightened by the absence of these mutant loci that were associated with LEPR protein expression, which was present on visual inspection in the scatter plot (Fig. 4B). The forest plot summarizing the effect size for each mutant locus in LEPR on OS risk was shown in Figure 4C. The leave-one-out analysis illustrated that the causal correlation between LEPR protein expression and OS risk remained stable, attributed to collective action of these mutant loci (Fig. 4D). The funnel plot illustrated symmetry generated for these mutant loci, indicating less possibility of bias on causal correlation (Fig. 4E). Additionally, the causal direction of LEPR plasma protein on OS risk was scrutinized for a robust connection by MR-Steiger directionality test (P < .001; Table 2).

Figure 4.

Visualization of the MR analysis of LEPR on OS risk. (A) The Manhattan plot for independent cis-pQTLs of LEPR after LD − clumping. (B) Scatterplot of the MR analysis of the effect of LEPR on OS risk. (C) Forest plots of causal effects of LEPR-associated cis-pQTLs on OS risk. (D) Leave-one-out sensitivity analysis of the effect of LEPR on OS risk. (E) Funnel plot of genetically predicted LEPR on OS risk. LD = linkage disequilibrium, LEPR = leptin receptor, MR = Mendelian randomization, OS = osteosarcoma, pQTL = protein QTL, QTL = quantitative trait loci, SNP = single nucleotide polymorphism.

Table 2.

Summary results from MR analyses of LEPR cis-pQTL on osteosarcoma (OS).

Exposure-outcome	Method	OR (95%CI)	P	IVW		MR-Egger				MR-Steiger
				Cochran Q	P	Intercept	P	Cochran Q	P	Correct Causal direction	P
LEPR-OS	IVW MR Egger Simple mode Weighted median Weighted mode	0.391 (0.216–0.707) 0.621 (0.195–1.975) 0.429 (0.209–0.878) 0.415 (0.208–0.829) 0.401 (0.199–0.806)	.002 .457 .109 .015 .026	2.095	.911	−0.181	.402	1.257	.939	TRUE	0

IVW = inverse-variance weighted, LEPR = leptin receptor, MR = Mendelian randomization, pQTL = protein QTL.

3.6. Methylome-wide MR analysis of LEPR in association with OS

To test the causal effect of methylation status at CpG sites in LEPR gene promoter on OS carcinogenesis, 6896 independent cis-mQTLs (P < 5e−8) associated with 47 LEPR-related CpG sites were extracted from the GoDMC database, which were integrated with LEPR cis-pQTL data and OS GWAS summary data for MR analysis. Following Bonferroni-corrected multiple testing, evidence of causality between LEPR DNAm level and LEPR plasma protein were observed at 40 CpG sites (P < .05/47 = .00106), including 21 positive and 19 negative correlation pairs. Similarly, causal associations with OS risk were detected at 29 CpG sites, including 9 positive and 20 negative correlation pairs. Finally, for associations at 8 common CpG sites (cg12075498, cg15985891, cg16009734, cg21683284, cg22733910, cg23013958, cg24049468, and cg26615127), both lower LEPR protein level and higher OS risk were consistently observed in MR analyses (Table 3).

Table 3.

Summary results from MR analyses of LEPR-related CpG sites on LEPR cis-pQTL and osteosarcoma (OS).

Exposure (CpG Sites)	NSNPs (mQTL)	Outcome	P	OR	95%CI
cg12075498 cg15985891 cg16009734 cg21683284 cg22733910 cg23013958 cg24049468 cg26615127	42 42 2 2 213 216 190 193 173 176 16 16 128 130 21 23	LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS LEPR (pQTL) OS	4.043E−221 3.513E−04 0 2.453E−04 8.210E−235 5.181E−66 1.780E−252 2.361E−60 2.393E−90 2.127E−57 0 1.248E−07 4.140E−128 3.878E−49 0 1.123E−09	0.027 3.946 2.270E−05 8301.492 0.099 5.636 0.097 6.199 0.120 7.152 0.104 17.800 0.142 6.921 0.041 45.669	0.022 to 0.034 1.858 to 8.377 1.470E−05 to 3.503E−05 66.749 to 1,032,443.287 0.086 to 0.113 4.626 to 6.867 0.085 to 0.112 4.984 to 7.711 0.097 to 0.147 5.617 to 9.105 0.098 to 0.109 6.120 to 51.768 0.121 to 0.166 5.351 to 8.953 0.038 to 0.044 13.352 to 156.201

CpG = cytosine-phosphate-guanine, LEPR = leptin receptor, mQTL = methylation QTL, MR = Mendelian randomization, pQTL = protein QTL.

3.7. Colocalization analyses in the LEPR region

Colocalization analyses were assumed to assess the probability of sharing genetic determinants between LEPR-related cis-pQTL, cis-mQTL, and OS GWAS data by integrating their summary statistics. The posterior probability of colocalization was > 80% conditional on the presence of a causal variant for the associations between LEPR-related cis-pQTL and OS, LEPR-related cis-mQTL and OS, and LEPR-related cis-mQTL and LEPR-related cis-pQTL, which provided complementary evidence for MR inference by reducing the likelihood of LD confusion (Table 4).

Table 4.

Colocalization Results among LEPR cis-pQTL, cis-mQTL and osteosarcoma (OS).

Trait1	Trait2	H1	H2	H3	H4	H3 + H4	H4/(H3 + H4)
LEPR (pQTL)	OS	8.37E−01	4.36E−135	6.32E−04	1.63E−01	1.63E−01	9.96E−01
LEPR (mQTL)	OS	7.21E−01	7.53E−54	5.37E−02	2.26E−01	2.79E−01	8.10E−01
LEPR (mQTL)	LEPR (pQTL)	0.00E + 00	0.00E + 00	9.99E−01	8.51E−07	1.00E−01	8.51E−07

LEPR = leptin receptor, mQTL = methylation QTL, pQTL = protein QTL.

4. Discussion

The heterogeneity of OS at the intra- or inter-tumoral level is attributed to the complexity and instability of OS somatic genome, which makes current therapeutic strategies devoted exclusively to OS cells difficult.^[20] Given this therapeutic dilemma, by capturing multi-layered pathobiological connections based on multi-omics profiles, it is worth considering developing potential therapeutic targets to block tumor progression in OS.^[21] Here, we identified 5 cell clusters with t-SNE clustering in combination, among which bone-related cells contribute to OS. We observed down-regulation of LEPR expression in OS cells based on scRNA-seq, which accorded with RT-qPCR results. Using bioinformatics analysis, our study provided evidence for the favorable prognostic value of LEPR in OS and explored its impact as a protective functional biomarker on immune status in TME. Based on integrated pQTLs, mQTLs, and GWAS data, we conducted MR and colocalization analyses to uncover the putative causal effect of LEPR in OS, which may guide future clinical treatment of OS.

The potential pathological involvement and prognostic significance of LEPR in multiple cancer types have attracted attention for its anticancer properties in recent years. A significant reduction in LEPR expression was observed in bladder cancer patients.^[22,23] Down-regulated LEPR expression was correlated with higher pathologic phenotype and increased recurrence risk in renal cell carcinoma patients.^[10] A cohort study enrolling anaplastic thyroid carcinoma patients indicated a significant correlation of positive LEPR expression with longer disease-free survival.^[24] In the present study, results obtained from Kaplan–Meier analysis of overall survival and event-free survival indicated that LEPR down-regulation was related to poor clinical outcomes in OS, insinuating the protective function of LEPR in OS progression. Furthermore, Cox regression analyses revealed that the potency of higher LEPR expression to predict better OS prognosis was internally validated in the TARGET database. The survival nomogram based on LEPR, metastasis, and chemotherapy performed effectively in terms of prognosis prediction for OS patients. Thus, These findings imply that LEPR expression may act as a predictor of OS prognosis.

In order to explore the protective mechanism of LEPR in OS, GSEA analysis was conducted to predict the LEPR-related enriched pathways. According to GSEA results, LEPR was mainly involved in immune response-related processes, which implied that LEPR might be involved in the altered immune status of OS. Immune score decreased in OS patients with low LEPR expression, suggesting that OS patients with low LEPR expression may be in an immunosuppressed TME state. TME is a complex and dynamic ecosystem that is crucial for tumor generation and metastasis.^[25,26] Immune cells embedded in the TME can exert anti-tumorigenic function by triggering innate or adaptive immunity.^[27–29] Our study demonstrated more significant infiltration of NK cells activated and T cells CD4 memory activated in OS patients with high LEPR expression, which indicated that increased LEPR expression might result in the antitumor immune response by recruiting innate and adaptive immune cells. Tumor-associated immune checkpoints are considered to mediate immune evasion by restricting T cell activity.^[30] In the present study, positive associations between LEPR expression and most immune checkpoints were detected, which might provide clues for immunotherapy of LEPR in OS. Furthermore, OS patients with low LEPR expression were more sensitive to chemotherapy by comparing the IC50 values in 4 chemotherapeutic agents. Thus, targeting LEPR in OS may help guide individualized and precise treatment.

Due to advances in large-scale multi-omics data acquisition, combining MR analyses with bioinformatics facilitates extracting causality from gene expression and complex traits.^[31] Our MR analyses strengthened existing bioinformatics evidence, pointing to the negative genetic association of LEPR protein expression with OS risk. After removing these SNPs associated with confounders under the criteria of IVs assumption, we leveraged genetic association MR analysis to reassess the causal association between LEPR and OS risk. Further sensitivity analyses indicated no evidence of heterogeneity and horizontal pleiotropy confounding the causal inference. The directionality test conducted by MR Steiger confirmed the accurate causal direction in which LEPR protein expression correlated with OS risk. Additionally, given the increasing attention paid to epigenetic alterations mediating the influence of gene expression on phenotype, we identified 8 methylation CpG sites mapping to LEPR for epigenetic associations with both lower LEPR protein expression and higher OS risk. DNAm at CpG sites has been widely elaborated to trigger carcinogenesis from an epigenetic perspective.^[32] Aberrant DNAm has been reported to be spread throughout the OS genome, contributing to OS tumorigenesis and aiding in identifying clinically relevant tumor subtypes in OS.^[33,34] Therefore, these candidate CpG sites in LEPR promoter may be potential DNAm targets for OS therapy.

However, this study has certain limitations. First, our bioinformatics analysis results regarding the prognostic potential of LEPR in OS and its association with immune infiltration were obtained from public databases. The data volume is limited, so the efficacy of LEPR as a potential prognostic factor for OS patients needs to be further verified by long-term prospective studies. Although Harmony was used to eliminate batch effects caused by the integration of 2 independent scRNA seq datasets, this does not fully remove differences caused by tissue source, disease context, and cellular composition. The lack of external validation for the constructed nomogram model exacerbates the overfitting effect caused by the relatively limited sample size of OS patients. In addition, although subchondral bone originates from bone mesenchymal stem cells, public scRNA-seq data derived from subchondral bone tissue (GSE196678) may not fully reflect osteoblastic lineage characteristics. Second, relatively small sample size of OS GWAS also limits our ability to determine causality between exposure (LEPR) and outcome (OS). Further, our MR analyses were conducted in participants of European ancestry, which may limit the generalizability of these results to other ancestries due to differences in local LD structure across ancestries. Third, our proteome and methylation data were derived from peripheral blood samples. There is a proportion of SNPs that are not shared across tissues, which may attenuate the genetic associations linking the target gene to related phenotypes.

In conclusion, our study suggested that LEPR might be a favorable prognostic biomarker for OS and mediated immune cell infiltration in the TME. We also provided supportive genetic and epigenetic evidence for a causal effect of LEPR on OS risk, which was expected to contribute to potential therapeutic interventions in future clinical guidelines. Further experimental research and longitudinal clinical trials are required to ascertain the current findings.

Acknowledgments

We acknowledge GEO, TARGET, deCODE, GoDMC, and FinnGen databases for supplying summary results data used in these analyses.

Author contributions

Conceptualization: Qinwen Li.

Data curation: Qinwen Li.

Formal analysis: Qinwen Li.

Funding acquisition: Hua Zhang.

Investigation: Man Qiu.

Methodology: Man Qiu.

Project administration: Hua Zhang.

Resources: Qinwen Li.

Software: Man Qiu.

Supervision: Hua Zhang.

Validation: Qinwen Li.

Visualization: Man Qiu.

Writing – original draft: Qinwen Li.

Writing – review & editing: Hua Zhang.