Elucidating the Causal Impact of Plasma Proteins on Osteoporosis Risk and the Mediating Role of Immune Cells through Integrated Multi-omics Data Analysis

doi:10.21203/rs.3.rs-4940543/v1

Download PDF

Article

Elucidating the Causal Impact of Plasma Proteins on Osteoporosis Risk and the Mediating Role of Immune Cells through Integrated Multi-omics Data Analysis

https://doi.org/10.21203/rs.3.rs-4940543/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Recent evidence increasingly suggests that plasma proteins and immune factors play crucial roles in bone metabolism. However, the specific mechanisms of how plasma proteins impact osteoporosis (OP) and whether immune factors mediate the process remain largely unknown. In this study, we employed a two sample mendelian randomization (TSMR) model and bidirectional analysis to identify plasma proteins linked to OP, explore whether immune phenotypes mediate OP risk associated with these proteins. Our findings revealed that plasma MGP and TMEM38B proteins were negatively correlated with OP (p < 0.001), while CD14, RUFY1, and IL6ST were positively associated with the risk of OP (p < 0.001). Further validation using whole blood eQTL data revealed a positive association between CD14 and OP risk (OR = 1.254, 95% CI: 1.016–1.547, p < 0.05). The analysis of the GEO dataset further corroborated this finding. Analysis revealed 62 immune phenotypes related to OP. Bidirectional TSMR analysis showed a positive correlation between plasma protein CD14 and IgD + CD38-%B cells (Ivw beta = 0.349, OR = 1.418, 95% CI: 1.068–1.882, p < 0.05). In addition, mediation analysis results showed that IgD + CD38-%B cells mediates 12% of the effect of CD14 on OP risk. Furthermore, Bayesian colocalization analysis reinforced our findings. Finally, in the drug target MR study and Phe-MR analysis, anti-CD14 exhibited a protective effect against OP (OR = 0.798, 95%CI: 0.646–0.984, p < 0.05). Our study provides evidence that elevated plasma CD14 is linked to OP risk, partially mediated by IgD + CD38-%B cells. Blocking CD14 shows potential in preventing or delaying OP. Further research is warranted to elucidate the underlying physiological and pathological mechanisms.

Health sciences/Endocrinology/Endocrine system and metabolic diseases

Health sciences/Risk factors

osteoporosis

plasma proteins

immune phenotypes

mendelian randomization

causality

Osteoporosis, a complex metabolic and endocrine disorder, is characterized by diminished bone mineral density (BMD), posing significant health challenges and socioeconomic implications¹. The reduction in bone mass increases vulnerability to fractures ². Globally, forecasts predict that by 2040, the number of individuals aged 50 and above suffering from osteoporosis will double, rising from 158 million in 2010 ³. In the USA alone, the financial burden of OP is projected to reach $25.3 billion by 2025 ⁴. However, the pathogenesis of OP remains a highly intricate and largely unexplored field.

Plasma proteins are essential for human health and disease management, involving critical physiological functions such as immune responses against pathogens (interferons, chemokines, and complement factors⁵), blood coagulation⁶, hormone transport⁷, and energy metabolism regulation⁸. Plasma proteins are widely used as markers of individual physiological status, accounting for 42% of all blood test applications⁹. So far, the US Food and Drug Administration (FDA) has approved 235 plasma proteins as diagnostic, prognostic, risk predictive or therapeutic response biomarkers¹⁰, which are used in a variety of diseases such as lung defects¹¹, cancer¹², autoimmune diseases¹³, and metabolic diseases¹⁴. Besides the relationship with clinical outcomes, the natural heterogeneity of plasma protein levels in general population has also been widely reported, but has not been taken into consideration in clinical application. Proteomic research enhances molecular comprehension and uncovers therapeutic targets, as numerous proteins in circulation often serve as key regulators of molecular pathways¹⁵. Prior observational studies have linked protein levels to BMD^16,17, yet their conclusions may be skewed by confounders and reverse effects, resulting in an unclear causal relationship.

Besides, recent evidence increasingly recognizes the close correlation between the immune and skeletal systems, with immune dysregulation playing a pivotal role in OP pathogenesis, though the precise immune factors remain elusive. Bone homeostasis hinges on the dynamic cellular balance during bone remodeling, revealing a far broader array of molecular and cellular factors involved in bone mass maintenance and augmentation than initially assumed¹⁸. "Osteoimmunology" encapsulates advancements in the complex interplay between the immune and skeletal systems. Recent studies have gained significant insights into this interaction, especially the immune system's impact on osteoclasts¹⁹. Osteoclasts and adaptive immune cells share a common ancestor, multipotent progenitors (MPPs), derived from hematopoietic stem cells (HSCs) and residing in bones ²⁰. Progress in OP research has led to a new branch, "immunoporosis" ²¹, highlighting the role of abnormal immune activation in both primary (aging, estrogen deficiency) and secondary (autoimmune diseases, infections) cases. These conditions are often accompanied by a pro-inflammatory state and RANKL production. RANKL, expressed by osteoblasts, osteocytes, and activated T/B cells, is a key cytokine that bridges the two systems by regulating osteoclast differentiation and proliferation through RANK receptors ^22,23. Recent studies have shown that immune factors play a crucial role in the development of OP. Elevated levels of osteoclastogenic factors, including TNF-α and RANKL, coupled with heightened GM-CSF secretion from B lymphocytes or depleted osteoclastogenic inhibitors like IFN-γ, all elevate the risk of OP^24,25. The aforementioned research findings have solidified the close correlation between OP and the immune system, yet the precise immune factors driving this association remain largely elusive.

Mendelian randomization (MR) is a cutting-edge analytical approach that leverages genetic variants as instrumental variables to evaluate causal relationships between variables and outcomes. It effectively tackles unresolved research questions, especially when causal links between exposures and outcomes are difficult to observe directly ²⁶. In MR, genetic variants serve as mediators to assess the causal effect of exposures on outcomes, validated by studies like Zhao et al²⁷. Compared to other observational methods, MR offers significant advantages. Since genetic variants are randomly distributed during gametogenesis from parents to offspring, the relationship between genotype and phenotype is preserved, thus eliminating biases from confounding factors and reverse causality ²⁸. This makes MR a powerful and efficient tool for assessing causal relationships in complex environments.

In this study, we employed a TSMR model and bidirectional analysis to identify plasma proteins linked to OP, explore whether immune phenotypes mediate OP risk associated with these proteins, and quantify the mediating role of immune phenotypes in this mechanism. Furthermore, Bayesian colocalization analysis reinforced our findings. Ultimately, we evaluated CD14 inhibitor reuse potential for OP treatment via drug target MR design.

Research design

In summary, as described in the chart abstract (Fig. 1), our model employed a six-step progressive process. Firstly, TSMR was utilized to identify plasma proteins as exposure variables and OP as outcome variables, thus obtaining plasma proteins associated with OP risk. Secondly, OP was taken as the exposure variable, and the proteins obtained in the first step were treated as the outcome variables for reverse MR analysis to exclude plasma proteins affected by OP. Thirdly, whole-blood eQTL data was used to verify the OP-related proteins screened in the first step. Fourthly, 731 immune phenotypes were used as exposure variables and OP as the outcome variable to screen OP-related immune phenotypes. Fifthly, the plasma protein screened in the first three steps was taken as exposure variables, and the immune phenotypes obtained in the fourth step were taken as outcome variables to explore the immune phenotypes related to plasma proteins and calculate the mediation proportion of immune phenotypes. Bayesian colocalization analysis further consolidates the analysis results. Besides, we evaluated the reutilization potential of CD14 inhibitors in treating OP using drug target MR design. Finally, the potential side effects of targeting CD14 as a drug target are evaluated using phe-MR analysis.

Source of exposure and outcome datasets

The exposure factor for MR was plasma circulating proteins, with consequences on OP risk. The largest and most comprehensive GWAS summary dataset for plasma proteins came from the deCODE Genetics Consortium (https://www.decode.com/summarydata/)²⁹. The original report provides a comprehensive overview of the dataset. In essence, the pQTL dataset encapsulates the associations between genome-wide genetic variations and 4719 plasma proteins, identified by 4907 aptamers, in a cohort of 35,559 Icelanders. These associations were adjusted for age, sex, and sample age. The SomaScan version 4 assay (SomaLogic) was utilized to quantify plasma protein levels.

The GWAS data of OP was derived from the BioBank Japan Project (http://jenger.riken.jp/en/)³⁰, which represents one of the largest non-European biobanks, comprising approximately 200,000 individuals. This data encompassed the analysis of 8,885,805 SNPs across 7,788 OP cases and 212,453 control participants.

To derive a more comprehensive and reliable understanding of the causal role of immune markers in the association between plasma proteins and OP risk, we selected the largest GWAS on peripheral blood immune profiling published so far (https://gwas.mrcieu.ac.uk/), encompassing 118 absolute cell counts, 389 median fluorescence intensities reflecting surface antigen levels, 32 morphological parameters, and 192 relative cell counts. GWAS summary statistics for 731 immune traits are publicly available in the GWAS Catalog (accession numbers from GCST0001391 to GCST0002121) ³¹. This GWAS analysis was based on 3,757 Sardinian samples (57% female), genotyping approximately 22 million single-nucleotide polymorphisms (SNPs) with high genetic significance. These SNPs were corrected for sex, age, and age squared using a density array. The SNPs were estimated using a Sardinian sequence-based reference panel ³².

In addition, we utilized a whole blood dataset from GTEx v8 (https://www.gtexportal.org/home/aboutAdultGtex), comprising 670 samples, to validate the impact of CD14 on OP as well as for drug target gene validation.

The GEO dataset GSE230665 was utilized to validate the expression differences of CD14 between healthy women and osteoporosis patients. It involved the application of a human whole-genome oligonucleotide microarray to examine the gene expression differences between two patient groups utilizing bone tissue samples from 15 patients. The patients were divided into 12 post-menopausal osteoporosis patients and 3 post-menopausal healthy women as the control group³³.

We collected single-cell transcriptomic data from GEO database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi, acc = GSE169396) ³⁴. This dataset contains bone tissue single-cells from femoral heads of two patients with osteoporosis and one patient with osteopenia. For more details about this dataset, please refer to the originally published paper ³⁴.

Instrumental variable selection

In MR analysis, single-nucleotide polymorphisms (SNPs) are typically chosen as instrumental variables (IVs) to estimate exposure-outcome causality, but they must satisfy three key assumptions ³⁵. Firstly, the relevance assumption requires that IVs must be strongly associated with the exposure to avoid bias from weak IVs ³⁶. Second, the exclusion restriction assumption ensures that the outcome is only influenced by the exposure and not by any other factors, implying no multiple causal pathways ³⁷. Third, the independence assumption requires that IVs are unaffected by confounding factors in the exposure-outcome association ³⁸.

In our previous MR analysis, we identified that plasma CD4 is a risk factor for OP. Therefore, in our drug target study, the exposure of interest is characterized as pharmacological inhibition of CD14. We first selected SNPs located within or near the genomic region of CD14 (± 1 Mb), which encodes CD14 (specifically, chromosome 5 and base pairs 140011317–140013286 based on GRCh37).

Statistical analysis

TSMR analyses

First, the impact of plasma proteins on OP was measured using TSMR. Subsequently, the first step of a TSMR approach was used to estimate the effect of immune phenotypes on OP, while the second step measured the impact of plasma proteins on immune phenotypes and conducted a reverse causality test to exclude bidirectional causality. We have conducted an MR analysis using the TwoSampleMR and Mendelian Randomization packages in R software (version 4.3.3).

Sensitivity analyses

In the TSMR analysis, we confirmed the robustness of Ivw results through rigorous methods including the MR Egger and MR-PRESSO. Under the InSIDE assumption, the MR-Egger method assesses whether genetic variants exert a non-zero, directional pleiotropic effect on the outcome ³⁹. The MR-PRESSO method detects outlier SNPs with horizontal pleiotropy, assuming the largest candidate set with similar estimates represents valid instruments, and evaluates the impact of their exclusion on causal estimates ⁴⁰. Our pleiotropy test hinges on the MR Egger intercept, suggesting a potential breach in TSMR's instrumental variable assumptions. Additionally, we evaluated instrument heterogeneity using Cochran's Q statistic.

To refine our findings, we applied the coloc R package (version 5.1.0; https://chr1swallace.github.io/coloc/) to conduct Bayesian colocalization analysis on MR outcomes, estimating the posterior probability of shared variation ⁴¹. For each leading SNP investigated in the GWAS databases, we retrieved all SNPs within a 50 kb region both upstream and downstream of the leading SNP for colocalization analysis, specifically to analyze the posterior probability of hypothesis H4 (PP.H4). A PP.H4 > 0.8 indicates a high probability of shared genetic causation between exposure and outcome, suggesting horizontal pleiotropy that could bias inferences and conclusions ⁴².

Calculating the Mediation Effect Proportion

After excluding SNPs used in the MR analysis between immune phenotypes and OP, we applied TSMR to obtain β1 coefficient representing the relationship between plasma proteins and immune phenotypes. Similarly, after excluding SNPs in the MR analysis between plasma CD4 and immune phenotypes, we got β2 coefficient representing the relationship between immune phenotypes and OP, and β12 = β1 * β2 reflecting the relationship between plasma proteins and OP. The total effect of plasma proteins on OP can be decomposed into: 1) direct effect (β3) and 2) indirect effect mediated by the intermediate (β12). We calculated the percentage of mediation effect by dividing the indirect effect by the total effect. Additionally, the 95% confidence interval was derived using the delta method⁴³.

MR‑PheWAS in FinnGen

The FinnGen study stands out for successfully integrating genomic information with the digital health data of participants over the age of 18 living in Finland. The study's resources encompass prospective epidemiological cohorts, disease-specific cohorts, and hospital biobanks. Comprehensive details about this study have been elaborated in other literature ⁴⁴ and can be accessed through the official website (https://www.finngen.fi/fi). FinnGen's disease summary data comes from the MR-Base platform. Out of a total of 2408 traits, we filtered out phenotypes with fewer than 200 cases, potential duplicates, and diseases with unclear definitions. After this series of selection processes, we ultimately chose 1,640 phenotypes for in-depth study. It is worth noting that, due to FinnGen not providing corresponding ICD-10 disease codes, we did not associate them with phecodes but directly adopted the provided disease descriptions.

Identifying the putative causal proteins on OP

An overview of the study design is shown in Fig. 1. Based on the GWAS study with the largest scale and the most comprehensive coverage, we conducted a TSMR analysis to identify the putative causal proteins that influencing the risk of OP. According to the selection criteria for instrumental variables (IVs), 11,980 proxies representing 4,253 unique proteins possessed sufficient IVs (n ≥ 3) for the MR analysis, with the number of IVs ranging from 3 to 156 (Supplementary Table S1, S2). The F-statistic of all IVs was greater than 10 (Supplementary Table S1), supporting that these cis-pQTLs are strong instruments. We employed 5 different methods (MR Egger, Weighted median, Inverse variance weighted (Ivw), Simple mode and Weighted mode) for MR analysis, and through the Ivw method, we identified 5 plasma proteins that are associated with the risk of OP (Supplementary Table S3). Given that there was only one positive result after Bonferroni correction with p < 1^10 − 5, we set the p-value threshold at p < 0.001 (Fig. 2a). As depicted in Fig. 2b, proteins CD14, LI6ST, and RUFY1 originated from chromosome 5, while TMEM38B stemmed from chromosome 9, and MGP was derived from chromosome 12. Of these 5 proteins, CD14, RUFY1, and IL6ST contribute to an increased risk of OP (with OR values of 1.249, 1.545, and 1.177, respectively, p < 0.001), while MGP and TMEM38B exert inhibitory effects on OP risk (with OR values of 0.754 and 0.767, respectively, p < 0.001) (Table 1, Fig. 2c).

Table 1

The results of TSMR and drug-target MR.
Exposure	Outcome	Method	β	β 95% CI	OR	OR 95% CI	P value
MGP	OP	Ivw	-0.283	-0.406 --0.16	0.754	0.667–0.852	p<0.001
TMEM38B	OP	Ivw	-0.265	-0.413 --0.117	0.767	0.662–0.889	p<0.001
IL6ST	OP	Ivw	0.163	0.067–0.26	1.177	1.069–1.297	p<0.001
CD14	OP	Ivw	0.222	0.11–0.335	1.249	1.116–1.398	p<0.001
RUFY1	OP	Ivw	0.435	0.188–0.682	1.545	1.207–1.979	0.001
OP	MGP	Ivw	-0.083	-0.225 -0.058	0.920	0.799–1.06	0.248
OP	TMEM38B	Ivw	-0.092	-0.23 -0.045	0.912	0.795–1.046	0.188
OP	IL6ST	Ivw	-0.024	-0.168 -0.12	0.976	0.845–1.128	0.744
OP	CD14	Ivw	-0.058	-0.198 -0.083	0.944	0.82–1.086	0.420
OP	RUFY1	Ivw	-0.097	-0.234 -0.041	0.908	0.791–1.041	0.167
CD14 eQTL	OP	Ivw	0.226	0.016–0.436	1.254	1.016–1.547	0.035
CD14	IgD + CD38- B cell	Ivw	0.349	0.349 − 0.145	1.418	1.068–1.882	0.016
CD14	BCD38_on_IgD	Ivw	-0.268	-0.533 --0.003	0.765	0.587–0.997	0.047
IgD + CD38- B cell	CD14	Ivw	0.002	-0.002 -0.007	1.002	0.998–1.007	0.278
anti-CD14	OP	Ivw	0.226	0.016–0.436	0.798	0.646–0.984	0.035
OP indicates osteoporosis; Ivw, inverse variance weighted; MGP, matrix Gla protein; TMEM38B, transmembrane protein 38B; IL6ST, interleukin 6 cytokine family signal transducer; RUFY1, RUN and FYVE domain containing 1

Due to the potential influence of pleiotropy in MR analysis, several sensitivity analyses were conducted. Both Ivw MR and MR-Egger regression indicated a similar direction of effect on OP risk, as well as comparable estimates for all identified proteins. Cochrane's Q test revealed low heterogeneity in the associations between plasma proteins and OP risk (Supplementary Table S4, Table 2). MR-Egger intercept analysis provided little evidence of directional pleiotropy for all five proteins (p > 0.05) (Supplementary Table S5, Table 2).

Table 2

Summary results of pleiotropy test
Outcome	Exposure	Egger intercept	Estimate	P value
OP	RUFY1	-0.005	0.035	0.89
OP	MGP	-0.014	0.009	0.159
OP	TMEM38B	0.005	0.015	0.761
OP	IL6ST	-0.002	0.007	0.771
OP	CD14	-0.016	0.036	0.672
OP	CD14 eQTL	0.1	0.148	0.623
IgD + CD38- B cell	CD14	0.035	0.053	0.553
BCD38 on IgD	CD14	0.023	0.034	0.537
OP indicates osteoporosis; Ivw, inverse variance weighted; MGP, matrix Gla protein; TMEM38B, transmembrane protein 38B; IL6ST, interleukin 6 cytokine family signal transducer; RUFY1, RUN and FYVE domain containing 1

To exclude the influence of OP on the causal relationship of plasma proteins, we conducted a reverse MR analysis next. Based on the criteria for selecting IVs, 15 SNPs were finally included in the MR analysis (Supplementary Table S11). The Ivw method was utilized for this analysis. The results showed no significant association between OP and these five plasma proteins (p > 0.05) (Table 1). In summary, these findings indicate that OP did not cause changes in the levels of these five proteins in circulating plasma, suggesting that abnormalities in these proteins in plasma may be risk factors for OP.

To validate the reliability of the results, we further conducted a validation using whole blood eQTL data. Eventually, we selected 3 SNPs representing CD14, 4 SNPs representing RUFY1, and 2 SNPs representing TMEM38B for the MR analysis (Supplementary Table S6). However, there were no valid SNPs for MGP and IL6ST to conduct the MR analysis. The results showed that CD14 significantly increased the risk of OP (OR = 1.254, 95%CI 1.016–1.547, p = 0.035) (Table 1, Fig. 2c). We also analyzed the GEO dataset GSE230665 and similarly discovered a significantly elevated expression of CD14 in the bone tissue of osteoporotic patients compared to that of healthy individuals (P < 0.05) (Fig. 2d). Therefore, CD14 was chosen for further investigation.

Examination of the causal relation of immunophenotypes on OP

Since many studies have shown that immune factors play an important role in the occurrence and development of OP, we next explored the impact of immune phenotypes on the risk of OP. According to the mentioned criteria for selecting instrumental variables, a total of 728 immune phenotypes were included in the MR analysis, with IVs ranging from 3 to 237 (Supplementary Table S7). We employed five methods (MR Egger, Weighted median, Ivw, Simple mode, and Weighted mode) for the MR analysis. Generally, the standard error of the Ivw method is smaller than that of the MR-Egger method⁴⁵. Therefore, in the absence of heterogeneity and horizontal pleiotropy, the Ivw results, serving as the gold standard for MR analysis, are prioritized. According to the IVW results, 62 immune phenotypes were found to be associated with the risk of OP. (Supplementary Table S8). Among them, 42 immune phenotypes might increase the risk of OP, while 20 immune phenotypes had a protective effect against OP (Fig. 3). Furthermore, for all four associations, the MR-Egger intercept tests dismissed the likelihood of horizontal pleiotropy (Supplementary Table S9). Sensitivity analyses offered robust evidence to substantiate the robustness of the identified causal relationships (Supplementary Table S10). The stability of the data was further validated through scatterplots (Supplementary Figure S1).

Exploring the Impact of Plasma CD14 Protein on Immunophenotypes

Since plasma proteins are closely related to various immune responses in the body, we further investigated the impact of CD14 on the aforementioned 62 immunophenotypes through TSMR analysis. Similarly, 5 methods were employed for the MR analysis (Supplementary Table S12). The findings indicated that plasma CD14 protein exerted an influence on two distinct immune phenotypes of B cells: IgD + CD38- %B cells and CD38 + IgD + B cells (Table 1). Specifically, CD14 stimulated the proliferation of IgD + CD38- %B cells, whereas it exhibited an inhibitory effect on CD38 + IgD + B cells (Fig. 4). Specifically, using the Ivw method, the ratio of CD14 to IgD + CD38- %B cells (OR) was found to be 1.418 (95% CI: 1.068–1.882, p = 0.016; Table 1). In contrast, the ratio of CD14 to CD38 + IgD + B cells (OR) was 0.765 (95% CI: 0.587–0.997, p = 0.047; Table 1). Since the Egger intercept of MR-Egger showed no statistical significance when compared to 0 (p > 0.05, Table 2), we can infer the absence of horizontal pleiotropy. Our sensitivity analysis further strengthened the credibility of the causal relationships derived from this analysis (Table 3).

Table 3

Summary results of sensitivity and heterogeneity analyses
Exposure	Outcome	Method	Estimate	P value
RUFY1	OP	MR Egger	5.572	0.35
		Ivw	5.595	0.47
MGP	OP	MR Egger	13.324	0.501
		Ivw	15.542	0.413
TMEM38B	OP	MR Egger	7.199	0.515
		Ivw	7.298	0.606
IL6ST	OP	MR Egger	21.128	0.513
		Ivw	21.215	0.568
CD14	OP	MR Egger	11.731	0.816
		Ivw	13.761	0.745
		MR-PRESSO (raw)	15.555	0.778
CD14 eQTL	OP	MR Egger	0.059	0.808
		Ivw	0.512	0.774
Anti-CD14	OP	MR-PRESSO (raw)	8.292	0.399
CD14	IgD + CD38- B cell	MR Egger	2.472	0.48
		Ivw	2.915	0.572
CD14	BCD38 on IgD	MR Egger	3.483	0.323
		Ivw	4.043	0.400
OP indicates osteoporosis; Ivw, inverse variance weighted; MGP, matrix Gla protein; TMEM38B, transmembrane protein 38B; IL6ST, interleukin 6 cytokine family signal transducer; RUFY1, RUN and FYVE domain containing 1

Besides, upon a further analysis of single-cell sequencing data in bone tissue, it was revealed that the content of B cells in bone tissue derived from osteoporosis patients (16.23%) was significantly higher than that in non-osteoporosis bone tissue (4.71%), which indirectly reaffirmed B cells as one of the risk factors for osteoporosis (Fig. 5a).

Previous studies have shown that CD14 can promote the growth and differentiation of B cells^46–48, thus the promotional effect of CD14 on IgD + CD38- %B cells is more in line with previous research results. Since CD14 protein is also an immune factor, we further investigated the impact of IgD + CD38- %B cells on plasma CD14 protein using reverse MR analysis. Based on previous criteria, 5 IVs representing IgD + CD38- %B cells were used for the MR analysis. Through 5 analytical methods, the results showed that IgD + CD38- %B cells had no significant impact on plasma CD14 protein levels (p > 0.05) (Table 1 and Supplementary Table S14). In summary, plasma CD14 protein exerts a distinct unidirectional promoting effect on IgD + CD38- %B cells.

Mediating effect of IgD + CD38- %B cells in the association between CD14 and OP

A harmful causal link was identified between CD14 and OP risk, utilizing the Ivw model as the standard method. Specifically, the Ivw β coefficient stood at 0.222, with a 95% confidence interval ranging from 0.110 to 0.335, indicating statistical significance (P < 0.05), β3 = 0.222. A positive causal link between plasma CD14 protein and IgD + CD38- %B cells was identified, utilizing Ivw analysis as the primary method (Ivw β = 0.349, 95% CI: 0.066–0.633, P < 0.05), β1 = 0.349. In addition, after excluding SNPs in the MR analysis between plasma CD4 and immune phenotypes, the OR analysis indicated a significant positive causal relationship between IgD + CD38- %B cells and OP risk (Ivw β = 0.074, 95% CI: 0.020–0.129, P < 0.05), β2 = 0.074 (Fig. 5b). Then, the indirect effect mediated by the intermediate of plasma CD14 and OP risk was 0.026. This suggests that IgD + CD38- %B cells mediates 12% of the effect (Supplementary Table S14). In addition, Bayesian colocalization analysis revealed that plasma CD14 had an effect on OP risk with PP.H4 < 0.8 (Supplementary Table S15). Similarly, plasma CD14 influenced IgD + CD38- %B cell phenotypes, and IgD + CD38- %B cells also had an impact on OP risk (Fig. 5c).

The causal effect of CD14 inhibition on OP risk

IC14 (atibuclimab) is a monoclonal antibody against CD14, which was once used to treat sepsis and has been adopted to treat COVID-19 and ALS in recent years^49–51. We intend to explore the potential of using IC14 targeting CD14 for the treatment of OP. Screening based on CD14's chromosomal location information and IV criteria revealed that only rs5744454 in the GWAS data of plasma CD14 from the deCODE database met the criteria but was not suitable for MR analysis. Therefore, we further screened the whole blood eQTL data in the GTEx database and ultimately selected rs1862176, rs9688094, and rs778587 as the three IVs for MR analysis. The results of the Ivw method showed that inhibiting CD14 has a significant protective effect against OP (OR = 0.798, 95%CI 0.646–0.984, p = 0.035) (Fig. 5d). This indicates that IC14 could serve as a potential drug candidate for the treatment of OP.

Phe-MR analysis for the associations between CD14 and 1640 phenotypes

We conducted a MR-PheWAS scanning to comprehensively assess the influence of CD14 on the risk of 1640 non-osteoporosis phenotypes, aiming to uncover its potential side effect profiles. However, in the Phe-MR analysis using the IVW method, seven associations (Acute appendicitis, Dysplastic lesion of cervix uteri, vagina or vulva, Persistent delusional disorders, Third [oculomotor] nerve palsy, Bell's palsy, Haemorrhagic disorder due to circulating anticoagulants, Arterial embolism and thrombosis) were less than the FDR-corrected significance threshold P = 0.05 (Fig. 5e and Table S15). Among these seven associations, CD14 was also a risk factor for five of the diseases (Acute appendicitis, Dysplastic lesion of cervix uteri, vagina or vulva, Persistent delusional disorders, Haemorrhagic disorder due to circulating anticoagulants, Arterial embolism and thrombosis).

Through the genetic causality analysis in this study, we discovered a positive genetic correlation between plasma CD14 and OP among East Asians. Follow-up investigations on TSMR confirmed this association and verified the positive causal relationship between plasma CD14 and OP. In addition, a two-step MR analysis showed that IgD + CD38- %B cells might be a potential mediator between plasma CD14 exposure and OP, with a mediation rate of 12%. Additionally, the co-localization analysis demonstrated the robustness of the MR analysis results. In the subsequent drug target MR investigation, we studied the causal impact of CD14 monoclonal antibody (IC14) on OP, confirming that reducing CD14 could reduce the likelihood of OP occurrence.

Plasma proteins play an important role in bone health, but the exact biological mechanisms of how they affect OP are largely unknown. Soluble CD14 (sCD14), a core receptor of bacterial lipopolysaccharides and a pro-inflammatory cytokine, may have a more direct link with bone health⁵². In the human body, it is produced by proteolytic cleavage of membrane-bound CD14 on hepatocytes and monocytes and macrophages (precursors of mature osteoclasts)⁵³. Existing data on sCD14 and bone metabolism are quite limited. Previous studies have shown that mice with inactivating mutations in the CD14 gene had a phenotype of increased bone mineral content ⁵⁴. In humans, CD14 mRNA expression was upregulated in osteoporotic bones compared to normal bones⁵⁵. In a large cohort of elderly people in the community, the higher the level of sCD14, the higher the risk of hip fracture in white person⁵⁶. Our results confirm that elevated plasma CD14 was a risk factor for OP in Asians, consistent with previous studies in white populations, and the use of GWAS data analysis further strengthened the reliability of our findings.

B lymphocytes are the most important component of humoral adaptive immune system. Mature B lymphocytes are developed from hematopoietic stem cells (HSC) precursors in the bone marrow, whose main function is to secrete antibodies, neutralize pathogens, and enhance the effector function of other immune cells. Increasing data indicate that there exists a complex relationship between B cells and bone cell differentiation. Under physiological conditions, the bone protectin (OPG) produced by B cells accounts for about 40–60% of the total OPG derived from bone marrow, thereby inhibiting the differentiation of osteoclasts ⁵⁷. Activated B cells under inflammatory conditions can also secrete RANKL, which activates osteoclast formation and leads to OP ⁵⁸. In postmenopausal women who do not use estrogen replacement therapy, B lymphocytes in the bone marrow secrete more RANKL (41). Further studies have shown that reducing RANKL secretion by B cells partially protects ovariectomized mice from trabecular bone loss (42). Although there are indications that B cells may inhibit osteoblast differentiation through CCL3 and TNF signals, there are few studies on the effects of B cells on osteoblasts ⁵⁹. B-cell-derived TGF-β1 inhibits osteogenesis and contributes to bone loss ⁶⁰. This research findings also indicate that an increase in IgD + CD38- %B cells is one of the risk factors for OP. In conclusion, B cells seem to play a significant role in regulating osteoclastogenesis through the RANKL/OPG signaling system. During postmenopausal OP, B cells are activated under estrogen deficiency and proinflammatory conditions, increasing bone resorption by secreting higher levels of G-CSF and RANKL ⁶¹.

Previous studies have indicated that soluble CD14 activated by pathogens and tissue injury factors may initiate signal transduction in various cells ⁶². However, there are few studies on the role of sCD14 in the growth and development of B lymphocytes. Existing research has shown that soluble CD14 can promote the growth and development of B cells ^46–48. This study also found a positive correlation between plasma protein CD14 and IgD + CD38- % B cells, suggesting that plasma CD14 can promote the growth of IgD + CD38- %B cells. Nevertheless, the specific mechanism of CD14 promoting B cell growth and development still needs further experimental validation.

This study boasts several strengths, including the utilization of bidirectional TSMR design and mediation MR design to enhance causal inference, the exploration of drug target validation, the verification using whole-blood eQTL data, and the consistency of results derived from sensitivity analysis. However, in interpreting our findings, it is imperative to discuss some limitations. Firstly, our dataset solely comprises Asians, limiting the generalizability of our results. Further research is required to validate whether these findings apply to other ethnic groups. Secondly, the GWAS summary datasets are derived from published studies, and the population is not stratified based on specific factors such as age and gender. Additionally, given the complexity of OP pathogenesis, we cannot definitively determine the specific role of plasma proteins and immune phenotypes in OP development. The results from LDSC and MR studies only indicate potential genetic correlations and causal relationships at the genetic level, and further mechanism-based experiments are needed to substantiate the biological plausibility and elucidate the biological mechanisms underlying the involvement of plasma protein CD14 and B cells in OP pathogenesis.

Our research indicated that circulating plasma protein CD14 was associated with risk factors for OP, with the upregulation of IgD + CD38- %B cells being a partial cause of the association between plasma CD14 protein and OP risk. We have also demonstrated that inhibiting CD14 can prevent or delay the onset of OP. Our findings further substantiate the interaction between plasma CD14 protein, immunophenotypes, and the risk of OP, thus confirming that CD14 is a potential target for preventing OP in the general population. This holds significant clinical and public health implications for OP intervention measures.

Ethics approval and consent to participate:

Not applicable.

Declaration of interests:

The authors state that they have no conflicts of interest.

Acknowledgments:

Not applicable.

Data availability:

All data analyzed in this study have been described in detail in the "Source of exposure and outcome datasets" section of this article and accompanied with download addresses.

Contributors:

Guangjun Jiao and Yunhao You planned the study and interpreted the analyses. Zhongjie ji and Xiang Li performed the analyses, produced the figures and tables, and drafted the manuscript. Jinlong Ma, Zhenqian Sun, Yunhao You: Methodology, Software, Visualization. Hongliang Wang and Wenliang Wu: Formal analysis, Methodology, Data curation, Visualization. Yunzhen Chen: Conceptualization, Writing review & editing, Validation. All authors contributed to the interpretation of the results and revision of the paper and approved the final manuscript.

Funding

This study was supported by the Natural Science Foundation of Shandong Province (Grant number ZR2021MH293).

Dong, Y. et al. Global, Regional, and National Burden of Low Bone Mineral Density From 1990 to 2019: Results From the Global Burden of Disease Study 2019. Front Endocrinol (Lausanne) 13, 870905, doi:10.3389/fendo.2022.870905 (2022).
Riggs, B. L. & Melton, L. J., 3rd. Evidence for two distinct syndromes of involutional osteoporosis. Am J Med 75, 899-901, doi:10.1016/0002-9343(83)90860-4 (1983).
Cooper C, F. S. IOF compendium ofosteoporosis. (2017). 1st ed (2017). 2017).
Burge, R. et al. Incidence and economic burden of osteoporosis-related fractures in the United States, 2005-2025. J Bone Miner Res 22, 465-475, doi:10.1359/jbmr.061113 (2007).
Ray, S., Patel, S. K., Kumar, V., Damahe, J. & Srivastava, S. Differential expression of serum/plasma proteins in various infectious diseases: specific or nonspecific signatures. Proteomics Clin Appl 8, 53-72, doi:10.1002/prca.201300074 (2014).
Davie, E. W. Introduction to the blood coagulation cascade and cloning of blood coagulation factors.
Pardridge, W. M. Plasma protein-mediated transport of steroid and thyroid hormones. Am J Physiol 252, E157-164, doi:10.1152/ajpendo.1987.252.2.E157 (1987).
Qaid, M. M. & Abdelrahman, M. M. Role of insulin and other related hormones in energy metabolism—A review. Cogent Food & Agriculture 2(1), 1-18 (2016).
Geyer, P. E., Holdt, L. M., Teupser, D. & Mann, M. Revisiting biomarker discovery by plasma proteomics. Mol Syst Biol 13, 942, doi:10.15252/msb.20156297 (2017).
Amur, S., LaVange, L., Zineh, I., Buckman-Garner, S. & Woodcock, J. Biomarker Qualification: Toward a Multiple Stakeholder Framework for Biomarker Development, Regulatory Acceptance, and Utilization. Clin Pharmacol Ther 98, 34-46, doi:10.1002/cpt.136 (2015).
Wu, A. C. et al. Current Status and Future Opportunities in Lung Precision Medicine Research with a Focus on Biomarkers. An American Thoracic Society/National Heart, Lung, and Blood Institute Research Statement. Am J Respir Crit Care Med 198, e116-e136, doi:10.1164/rccm.201810-1895ST (2018).
Enroth, S. et al. High throughput proteomics identifies a high-accuracy 11 plasma protein biomarker signature for ovarian cancer. Commun Biol 2, 221, doi:10.1038/s42003-019-0464-9 (2019).
Goudy, K. et al. Human IL2RA null mutation mediates immunodeficiency with lymphoproliferation and autoimmunity. Clin Immunol 146, 248-261, doi:10.1016/j.clim.2013.01.004 (2013).
Ojha, A., Ojha, U., Mohammed, R., Chandrashekar, A. & Ojha, H. Current perspective on the role of insulin and glucagon in the pathogenesis and treatment of type 2 diabetes mellitus. Clin Pharmacol 11, 57-65, doi:10.2147/CPAA.S202614 (2019).
Finan, C. et al. The druggable genome and support for target identification and validation in drug development. Sci Transl Med 9, doi:10.1126/scitranslmed.aag1166 (2017).
Al-Ansari, M. M. et al. Proteomics Profiling of Osteoporosis and Osteopenia Patients and Associated Network Analysis. Int J Mol Sci 23, doi:10.3390/ijms231710200 (2022).
Huang, D. et al. Proteomic profiling analysis of postmenopausal osteoporosis and osteopenia identifies potential proteins associated with low bone mineral density. PeerJ 8, e9009, doi:10.7717/peerj.9009 (2020).
Cao, R. R. et al. The immune factors have complex causal regulation effects on bone mineral density. Front Immunol 13, 959417, doi:10.3389/fimmu.2022.959417 (2022).
Ponzetti, M. & Rucci, N. Updates on Osteoimmunology: What's New on the Cross-Talk Between Bone and Immune System. Front Endocrinol (Lausanne) 10, 236, doi:10.3389/fendo.2019.00236 (2019).
Walker, D. G. Bone resorption restored in osteopetrotic mice by transplants of normal bone marrow and spleen cells. Science 190, 784-785, doi:10.1126/science.1105786 (1975).
Srivastava, R. K., Dar, H. Y. & Mishra, P. K. Immunoporosis: Immunology of Osteoporosis-Role of T Cells. Front Immunol 9, 657, doi:10.3389/fimmu.2018.00657 (2018).
Kong, Y. Y. et al. Activated T cells regulate bone loss and joint destruction in adjuvant arthritis through osteoprotegerin ligand. Nature 402, 304-309, doi:10.1038/46303 (1999).
Graves, D. T., Oates, T. & Garlet, G. P. Review of osteoimmunology and the host response in endodontic and periodontal lesions. J Oral Microbiol 3, doi:10.3402/jom.v3i0.5304 (2011).
Charatcharoenwitthaya, N., Khosla, S., Atkinson, E. J., McCready, L. K. & Riggs, B. L. Effect of blockade of TNF-alpha and interleukin-1 action on bone resorption in early postmenopausal women. J Bone Miner Res 22, 724-729, doi:10.1359/jbmr.070207 (2007).
Pineda, B. et al. Gene expression profile induced by ovariectomy in bone marrow of mice: a functional approach to identify new candidate genes associated to osteoporosis risk in women. Bone 65, 33-41, doi:10.1016/j.bone.2014.05.001 (2014).
Zheng, J. et al. Phenome-wide Mendelian randomization mapping the influence of the plasma proteome on complex diseases. Nat Genet 52, 1122-1131, doi:10.1038/s41588-020-0682-6 (2020).
Zhao, S. S., Mackie, S. L. & Zheng, J. Why clinicians should know about Mendelian randomization. Rheumatology (Oxford) 60, 1577-1579, doi:10.1093/rheumatology/keab007 (2021).
Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 37, 658-665, doi:10.1002/gepi.21758 (2013).
Ferkingstad, E. et al. Large-scale integration of the plasma proteome with genetics and disease. Nat Genet 53, 1712-1721, doi:10.1038/s41588-021-00978-w (2021).
Ishigaki, K. et al. Large-scale genome-wide association study in a Japanese population identifies novel susceptibility loci across different diseases. Nat Genet 52, 669-679, doi:10.1038/s41588-020-0640-3 (2020).
Orru, V. et al. Complex genetic signatures in immune cells underlie autoimmunity and inform therapy. Nat Genet 52, 1036-1045, doi:10.1038/s41588-020-0684-4 (2020).
Sidore, C. et al. Genome sequencing elucidates Sardinian genetic architecture and augments association analyses for lipid and blood inflammatory markers. Nat Genet 47, 1272-1281, doi:10.1038/ng.3368 (2015).
Xie, L. et al. Comparisons of gene expression between peripheral blood mononuclear cells and bone tissue in osteoporosis. Medicine (Baltimore) 102, e33829, doi:10.1097/MD.0000000000033829 (2023).
Qiu, X. et al. Single-cell RNA sequencing of human femoral head in vivo. Aging (Albany NY) 13, 15595-15619, doi:10.18632/aging.203124 (2021).
Birney, E. Mendelian Randomization. Cold Spring Harb Perspect Med 12, doi:10.1101/cshperspect.a041302 (2022).
Brion, M. J., Shakhbazov, K. & Visscher, P. M. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42, 1497-1501, doi:10.1093/ije/dyt179 (2013).
Ding, P., VanderWeele, T. J. & Robins, J. M. Instrumental variables as bias amplifiers with general outcome and confounding. Biometrika 104, 291-302, doi:10.1093/biomet/asx009 (2017).
Border, R. et al. Assortative mating biases marker-based heritability estimators. Nat Commun 13, 660, doi:10.1038/s41467-022-28294-9 (2022).
Burgess, S. & Thompson, S. G. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 32, 377-389, doi:10.1007/s10654-017-0255-x (2017).
Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693-698, doi:10.1038/s41588-018-0099-7 (2018).
Giambartolomei, C. et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 10, e1004383, doi:10.1371/journal.pgen.1004383 (2014).
Yin, Q. & Zhu, L. Does co-localization analysis reinforce the results of Mendelian randomization? Brain 147, e7-e8, doi:10.1093/brain/awad295 (2024).
Yuan, J. et al. Genetically predicted C-reactive protein mediates the association between rheumatoid arthritis and atlantoaxial subluxation. Front Endocrinol (Lausanne) 13, 1054206, doi:10.3389/fendo.2022.1054206 (2022).
Mars, N. et al. The role of polygenic risk and susceptibility genes in breast cancer over the course of life. Nat Commun 11, 6383, doi:10.1038/s41467-020-19966-5 (2020).
Fei, Y., Yu, H., Wu, Y. & Gong, S. The causal relationship between immune cells and ankylosing spondylitis: a bidirectional Mendelian randomization study. Arthritis Res Ther 26, 24, doi:10.1186/s13075-024-03266-0 (2024).
Filipp, D. et al. Soluble CD14 enriched in colostrum and milk induces B cell growth and differentiation. Proc Natl Acad Sci U S A 98, 603-608, doi:10.1073/pnas.98.2.603 (2001).
Ziegler-Heitbrock, H. W. et al. CD14 is expressed and functional in human B cells. Eur J Immunol 24, 1937-1940, doi:10.1002/eji.1830240835 (1994).
Kimura, S. et al. CD14-dependent and independent pathways in lipopolysaccharide-induced activation of a murine B-cell line, CH12. LX. Scand J Immunol 51, 392-399, doi:10.1046/j.1365-3083.2000.00696.x (2000).
Olszyna, D. P. et al. Effect of IC14, an anti-CD14 antibody, on plasma and cell-associated chemokines during human endotoxemia. Eur Cytokine Netw 14, 158-162 (2003).
Mabrey, F. L. et al. Phase 2, randomized, double-blind, placebo-controlled multi-center trial of the clinical and biological effects of anti-CD14 treatment in hospitalized patients with COVID-19 pneumonia. EBioMedicine 93, 104667, doi:10.1016/j.ebiom.2023.104667 (2023).
Henderson, R. D. et al. Phase 1b dose-escalation, safety, and pharmacokinetic study of IC14, a monoclonal antibody against CD14, for the treatment of amyotrophic lateral sclerosis. Medicine (Baltimore) 100, e27421, doi:10.1097/MD.0000000000027421 (2021).
Meuleman, P. et al. Human hepatocytes secrete soluble CD14, a process not directly influenced by HBV and HCV infection. Clin Chim Acta 366, 156-162, doi:10.1016/j.cca.2005.09.022 (2006).
Bazil, V. et al. Structural relationship between the soluble and membrane-bound forms of human monocyte surface glycoprotein CD14. Mol Immunol 26, 657-662, doi:10.1016/0161-5890(89)90048-5 (1989).
Johnson, G. B., Riggs, B. L. & Platt, J. L. A genetic basis for the "Adonis" phenotype of low adiposity and strong bones. FASEB J 18, 1282-1284, doi:10.1096/fj.04-1572fje (2004).
Hopwood, B., Tsykin, A., Findlay, D. M. & Fazzalari, N. L. Gene expression profile of the bone microenvironment in human fragility fracture bone. Bone 44, 87-101, doi:10.1016/j.bone.2008.08.120 (2009).
Bethel, M. et al. Soluble CD14 and fracture risk. Osteoporos Int 27, 1755-1763, doi:10.1007/s00198-015-3439-9 (2016).
Walsh, M. C. & Choi, Y. Biology of the RANKL-RANK-OPG System in Immunity, Bone, and Beyond. Front Immunol 5, 511, doi:10.3389/fimmu.2014.00511 (2014).
Weitzmann, M. N. The Role of Inflammatory Cytokines, the RANKL/OPG Axis, and the Immunoskeletal Interface in Physiological Bone Turnover and Osteoporosis. Scientifica (Cairo) 2013, 125705, doi:10.1155/2013/125705 (2013).
Sun, W. et al. B cells inhibit bone formation in rheumatoid arthritis by suppressing osteoblast differentiation. Nat Commun 9, 5127, doi:10.1038/s41467-018-07626-8 (2018).
Chen, Y. et al. B-Cell-Derived TGF-beta1 Inhibits Osteogenesis and Contributes to Bone Loss in Periodontitis. J Dent Res 102, 767-776, doi:10.1177/00220345231161005 (2023).
Anargyrou, K. et al. Low Bone Mineral Density and High Bone Turnover in Patients With Non-Hodgkin's Lymphoma (NHL) Who Receive Frontline Therapy: Results of a Multicenter Prospective Study. Hemasphere 3, e303, doi:10.1097/HS9.0000000000000303 (2019).
Sharygin, D., Koniaris, L. G., Wells, C., Zimmers, T. A. & Hamidi, T. Role of CD14 in human disease. Immunology 169, 260-270, doi:10.1111/imm.13634 (2023).

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Elucidating the Causal Impact of Plasma Proteins on Osteoporosis Risk and the Mediating Role of Immune Cells through Integrated Multi-omics Data Analysis

Status:

Version 1

Abstract

Figures

Introduction

Materials and methods

Research design

Source of exposure and outcome datasets

Instrumental variable selection

Statistical analysis

TSMR analyses

Sensitivity analyses

Calculating the Mediation Effect Proportion

MR‑PheWAS in FinnGen

Results

Identifying the putative causal proteins on OP

Examination of the causal relation of immunophenotypes on OP

Exploring the Impact of Plasma CD14 Protein on Immunophenotypes

Mediating effect of IgD + CD38- %B cells in the association between CD14 and OP

The causal effect of CD14 inhibition on OP risk

Phe-MR analysis for the associations between CD14 and 1640 phenotypes

Discussion

Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1