To the best of our knowledge, this is the first study to examine the causal associations between plasma proteins and PCa using MR and Bayesian colocalisation. Combining the results of external validation, Bayesian co-localization, and PPI, the six identified potential target proteins were categorised into three tiers. Specifically, KDELC2, MSMB, GSTP1, and TNFRSF10B were designated as tier 1 target candidates, implying a higher likelihood of them being potential therapeutic targets for PCa. SPINT2 was confirmed to be a Tier 2 target candidate, suggesting its moderate potential as an effective therapeutic target. The IGF2R was classified as a Tier 3 target candidate, indicating a relatively lower likelihood.
Despite great advances in PCa chemotherapy using cytotoxic drugs and ADT, patients often eventually progress to CRPC and develop resistance to traditional therapeutic drugs, leading to treatment failure. Tumours can still progress and deteriorate even under extremely low levels of hormone, rendering first-line hormonal treatments less effective [29, 30]. The emergence of CRPC highlights the need for innovative therapeutic targets that mediate alternative pathways, inhibit androgen receptor (AR) mutations, or disrupt critical PPIs that drive tumour growth in this advanced stage of PCa. Given the advantages of MR in inferring causal relationships and enriching GWAS and pQTL databases, the development of MR for drug target prediction has gained increased momentum.
In the present study, we used a multitude of methods to identify novel drug targets among plasma proteins in PCa. Bidirectional MR and Steiger Filtering analyses were performed to confirm the accurate direction between the proteins and PCa (Table 2). To minimise the effect of the horizontal pleiotropy of the instrumental variant, only cis-pQTLs were used in this study. Moreover, Bayesian co-localization analysis was used to further eliminate bias, and showed that MSMB and KDELC2 shared the same genetic instrument as PCa (Table 2). Furthermore, we used PPI to explore the relationship between the identified drug targets and established drug targets approved by the FDA for the treatment of PCa to enhance the credibility of the identified proteins as potential novel targets.
Beta-microseminoprotein (MSMB) is an auxiliary diagnostic biomarker for PCa [31]. Previous studies have shown that MSMB was the target gene of rs10993994, which affected the secretion of β-MSP (a prostatic secretion secreted by multiple mucosal tissues) by regulating the expression of MSMB and further impairs the normal physiological activities of the prostate [32–34]. From the results of the co-localization analysis in this study, rs10993994 served as the leader SNP for MSMB (Bayesian analysis result of PPH4 > 0.8), suggesting that the target protein and PCa shared the same loci. These results provide supportive evidence regarding the direct influence of rs10993994 on MSMB, demonstrating that downregulation of MSMB increases the risk of PCa [35–37]. Hence, MSMB is a protein that causes PCa and activating its function of MSMB nay be an effective strategy for PCa treatment.
In our study, the Bayesian analysis highlighted the causal effects of KDELC2 on PCa. Protein O-glucosyltransferase (POGLUT, formerly KDELC2), a pivotal enzyme in the biosynthesis of protein O-glycosylation, plays a crucial role in the post-translational modification of various proteins by catalysing the attachment of glucose moieties to specific serine or threonine residues [38]. This enzymatic activity is central to the regulation of protein structure, stability, localisation, and function. Its involvement in Notch signalling pathway modulation highlights its significance in controlling cell differentiation and proliferation, which are dysregulated in diseases, including cancer [39]. In recent years, the Notch signalling pathway has been extensively studied in PCa. Aberrant activation of the Notch signalling pathway results from dysregulated expression or mutations in Notch receptors and ligands that significantly influence PCa initiation, progression, invasion, metastasis, and the development of resistance to CRPC [40–42]. Considering the vital role of KDELC2 in the Notch signalling pathway, the development of novel therapeutic interventions that target KDELC2 provides valuable insights for the treatment of PCa. In our study, we demonstrated a causal relationship between KDELC2 and PCa via MR imaging. However, we failed to reveal any interactions between KDELC2 and the approved PCa medication targets in the PPI network. In future studies, further experimental validation should be performed to elucidate the function of KDELC2 in PCa treatment.
GSTP1 has been previously reported as a biomarker of PCa [43]. GSTP1 belongs to the pi class of enzymes of the glutathione-S transferase family. Glutathione S-transferase Pi 1 (GSTP1) is a ubiquitously expressed phase II metabolic enzyme present in various tissues, notably the liver. Its principal functions encompass detoxification, antioxidant defence, and modulation of signal transduction pathways [43, 44]. Specifically, GSTP1 catalyses the conjugation of glutathione (GSH) with electrophilic compounds, including reactive oxygen species, xenobiotics, and metabolic byproducts, facilitating their excretion, thereby mitigating cellular damage. As an integral component of the antioxidant system, it protects cells against oxidative stress, which is a pivotal mechanism in preventing disease pathogenesis, including carcinogenesis [45]. Mahon KL et al. [46] verified that methylated GSTP1 could serve as a stable prognostic marker for PCa treated with docetaxel and as a prospective drug target for the treatment of PCa. In our PPI results, GSTP1 displayed confirmed interactions with established targets of several drugs for PCa, such as abiraterone (an inhibitor of CYP17A1) and enzalutamide (an inhibitor of the androgen receptor). This interaction suggests that drugs targeting GSTP1 may inhibit androgen synthesis, thereby exerting a therapeutic effect on PCa.
Tumour necrosis factor receptor superfamily member 10 B (TNFRSF10B) is also known as tumour necrosis factor-related apoptosis-inducing ligand 2 (TRAIL-2) or death receptor 5 (DR5). It serves as a critical receptor for the tumour necrosis factor-related apoptosis-inducing ligand (TRAIL), facilitating the assembly of a complex that activates caspase cascades and ultimately leads to programmed cell death. Dysregulation of TNFRSF10B expression has been implicated in cancer development and progression, and is involved in modulating immune responses and other complex and multifaceted roles in tumour biology, warranting extensive investigation as a promising therapeutic target. Various studies have revealed that therapies targeted TNFRSF10B selectively induces apoptosis in cancer cells [47, 48]. Although Bayesian analysis demonstrated that TNFRSF10B and PCa did not share the same SNP, PPI networks revealed an interaction between TNFRSF10B and the established drug targets BCL-2 (a target of docetaxel) and RARA (a target of flutamide). As expected, Seki et al. showed that regulating the TRAIL/DR5 signalling pathway could reduce the risk of androgen-dependent PCa [49]. Shishodia also reported that tetrandrine, a bioactive anticancer compound, sensitises PCa cells to TRAIL-induced apoptosis [50]. These previous studies, as well as our own, support the targeted efficacy of TNFRSF10B.
Collectively, our study identified four promising therapeutic target proteins for PCa. These proteins have been implicated in crucial biological processes underlying the initiation and progression of PCa. Therefore, reinforcing target-derived drug design involving these four proteins is a reliable strategy for PCa treatment. These findings offer promising leads for more effective PCa treatments, potentially reducing drug development costs and advancing personalised medicine approaches [51].
4.1. Limitation
Although this study identified several protein targets with therapeutic potential in PCa, some unavoidable limitations persisted. First, a limitation lies in the selection of the study cohort was limited. While the pQTLs were deliberately chosen to encompass diverse ethnicities, including non-European races, the GWAS data for PCa were exclusively derived from European races. This could potentially introduce bias into the MR analysis given the disparities in genetic backgrounds and linkage disequilibrium patterns. To mitigate this problem, we used aptamer-based plasma proteins. Furthermore, this study exclusively utilised cis-pQTLs as genetic variables, omitting trans-pQTLs. This decision was made to minimise the effects of heterogeneity; however, other regulatory mechanisms that contribute to the complexity of PCa may have been overlooked. In addition, the limited number of cis-pQTLs precludes the adoption of a more comprehensive MR approach to predict the sensitivity of well-defined targets. Notably, that all SNPs exhibited F-statistic values exceeding 10, indicating a low risk of weak instrumental variables. Moreover, the accuracy of PPI networks and protein complexes is heavily dependent on the quality of protein structural data. Although these methods offer a visual representation of protein interactions and complex binding, their specific binding sites require further investigation. Consequently, further experimental and clinical validation is necessary to confirm the therapeutic potential of the predicted drug targets.