Prostate cancer is the second most commonly diagnosed cancer in males worldwide and the fifth most frequent cause of cancer related death in men worldwide. It is a heterogenous condition ranging from relatively indolent to aggressive disease. Androgen deprivation therapy is the mainstay of treatment in metastatic prostate cancer (mPC), specifically using luteinising hormone releasing hormone agonists or antiandrogens. Unfortunately, prostate cancer cells become resistant to this treatment (termed castration-resistant), usually after 18–24 months.

Circulating serum biomarkers are molecules produced by normal and abnormal cells. They are attractive for clinical use as they are readily available and easily obtained through minimally invasive methods. Prostate specific antigen (PSA) is an example of a serum biomarker used widely in the diagnosis and prognostication of prostate cancer and in assessing response to treatment. It is a kallikrein protease produced by the luminal cells in the prostate gland. PSA has its limitations as it is organ- but not cancer-specific. It can be elevated in benign conditions, including prostatitis and benign prostatic hyperplasia (BPH). At present, PSA in addition to the Gleason score and radiological staging are used in the diagnosis and prognostication of prostate cancer. Novel biomarkers are needed in prostate cancer to improve the prognostic and predictive accuracy of PSA, imaging and biopsy alone.

There is an expanding choice of treatments for mPC; predictive factors would help clinicians to tailor the type and sequence of treatment to each patient. Despite extensive research, very few biomarkers have been implemented in routine clinical practice to date.

The aim of this review is to summarise the current published data focussing on a selection of promising biomarkers with the potential to improve the prognostic and predictive accuracy of PSA in mPC. The categories of biomarkers investigated include microRNA, androgen receptor variants (AR-Vs), bone metabolism, neuroendocrine and metabolite biomarkers.

At present, there is significant interest in the use of miRNAs as biomarkers to improve the diagnosis and prognostication of prostate cancer. miRNAs are endogenous short non-coding RNAs (19–26 nucleotides in length) that bind to complementary messenger RNA to suppress gene expression post transcriptionally. Approximately 1000 human miRNAs have been identified to date. Alterations in miRNA expression can affect important cellular processes, including the cell cycle, proliferation, apoptosis and epithelial to mesenchymal transition (EMT). They are relatively stable biomarkers and can be found free in the serum and within exosomes (extracellular vesicles in the serum, semen and urine). miRNAs are found in greater concentration in exosomes, as they are protected against RNAases. Reliable methods to extract miRNAs from serum and exosomes have been developed and documented. Reverse transcriptase polymerase chain reaction is then used to quantify miRNAs.

miRNAs can be divided into oncogenes or tumour suppressors dependent on their targets. Oncogenic miRNAs in prostate cancer can enhance proliferation by down regulating cell-cycle-dependent kinase inhibitors and transcription factors. High levels of miRNA-21 have been observed in multiple different cancers. In prostate cancer, it contributes to pathogenesis and castration resistance.

Studies have been performed to determine whether miRNAs can determine which patients with CRPC will have a favourable response to treatment with docetaxel chemotherapy. This would be of great clinical benefit, as there is only a 50% response rate in all patients with CRPC treated with docetaxel and the treatment can cause significant side effects. Serum miRNAs in the miRNA-200 and miRNA-17 families were associated with a PSA response and improved overall survival (OS) in CRPC receiving treatment with docetaxel. They have the potential to be used as early treatment response and prognostic biomarkers. The miRNA-200 family members are involved in the regulation of EMT, which is a mechanism of drug resistance and metastasis, whereas the miRNA-17 family has immune regulatory functions. These mi-RNA families may be involved in the mechanism of docetaxel resistance.

Exosomal-miRNA-1246 is a further potential miRNA biomarker with predictive and prognostic potential. Higher exosomal-miRNA-1246 expression was found in prostate cancers with higher stage, grade, presence of positive lymph node, and distant metastases. However, the serum concentration of miRNA-1246 is lower than healthy normal. An elevated level of miRNA-141 correlates with an increasing number of bone metastases. Further evidence suggests that miRNA-218-5p could also be a biomarker for bone metastases. It suppresses the NF-κB signaling pathway, which suppresses the invasion and migration ability of prostate cancer cells. MiRNA-218-5p expression was reduced in the serum of patients with bone metastatic prostate cancer compared to those with non-bone metastatic prostate cancer. Serum levels were significantly correlated with PSA levels, Gleason grade and bone metastasis status, and low levels of miRNA-218-5p predicted poor bone metastasis-free survival.

CRPC is not androgen independent; however, 20–40% of patients have a primary resistance to novel antiandrogens (enzalutamide and abiraterone) and nearly all develop a secondary resistance. AR-Vs are a possible cause of this resistance. They are truncated androgen receptor (AR) proteins without the AR-ligand binding domain. This results in constitutive AR signaling in the absence of androgens. The AR Splice Variant 7 (AR-V7) has been studied in detail to date. It can be detected in circulating tumour cells, exosomes and whole blood samples. Patients with metastatic castrate-resistant prostate cancer (mCRPC), who are receiving enzalutamide or abiraterone and are positive for AR-V7, have a worse progression free survival (PFS) and OS than patients who are negative for AR-V7. However, being positive for AR-V7 is not associated with a significant resistance to taxanes. AR-V7 could be a useful biomarker in predicting patients, who will benefit from treatment with novel androgen receptor blocking agents.

Prostate cancer frequently metastasizes to the bone. Over 90% of patients with mPC have bone metastases, which can cause considerable morbidity, including severe bone pain, hypercalcaemia, spinal cord compression and pathological fracture. Being able to predict patients who are at higher risk of developing these skeletal related events (SRE) enables optimisation of their clinical management to improve their quality of life and possibly survival. Currently, bone metastases are evaluated by nuclear medicine bone scans. These scans have a low specificity for detecting early bone metastases.

Bone metastases from prostate cancer are typically sclerotic in nature. Tumour cells secrete factors, which stimulate osteoblasts and therefore bone formation. Biochemical markers of bone metabolism include bone alkaline phosphatase (BALP), N-terminal propeptide of type 1 collagen (P1NP), beta-isomer of carboxyterminal telopeptide of collagen 1 (β-CTX), bone sialoprotein (BSP) and osteopontin (OPN). BALP and P1NP are markers of bone formation, whereas β-CTX is a marker of bone resorption. OPN and BSP are markers of both bone formation and resorption. These markers are all present in the serum and enable the monitoring of osteoclast and osteoblast activity.

BALP is an enzyme secreted by osteoblasts that promotes bone mineralization. Increased levels in the serum are a sign of increased osteoblast activity. It can also be elevated due to hepatic disease, as monoclonal antibodies against BALP exhibit 15% cross-reactivity with a hepatic isoenzyme. Type 1 collagen is a significant component of the bone matrix, and forms from type 1 procollagen. P1NP is a propeptide cleaved from the N-terminal of procollagen and can be detected in the serum. Levels of P1NP are associated with bone formation. Type 1 collagen releases carboxy-telopeptides (CTX) during bone resorption into the serum and urine. β-CTX is an isomerised form of CTX and has a circadian rhythm depending on eating. It should therefore be tested early in the morning in patients who have fasted. BSP and OPN are protein components of the non-collagenous bone matrix. BSP is expressed in osteoclasts and osteoblasts. OPN is expressed in osteoblasts. They are involved in the regulation of bone resorption and formation as well as in the process of forming metastases.

Several studies have demonstrated that markers of bone metabolism are useful in predicting SRE and prognostication in patients with prostate cancer and bone metastases. 

In TUGAMO prospective multicentre study, the serum levels of BALP, P1NP and β-CTX were monitored at baseline and then three-monthly in 98 patients with prostate cancer and bone metastases over an 18-month time period. All patients received ZA 4 mg every four weeks. Once ZA was established, no patients received hormone therapy, which can cause a rise in the bone metabolism biomarkers. Alternative standard systemic anti-cancer therapies were allowed during this study. All patients had raised markers at baseline, and a significant positive correlation was demonstrated between the level of the markers and the burden of bone metastases. No association was detected between the bone biomarkers and disease progression. A decrease in BALP and P1NP were strongly associated with the development of SRE. A decrease in BALP of <87% between baseline and 3 months indicated a 4-fold increase risk of SRE. A lack of normalisation of P1NP after treatment increased the risk of SRE 3.8 times. A decrease in β-CTX was associated with a lower OS (a decrease of <40% between baseline and 3 months indicated a six-fold increase in mortality).

BSP and OPN have demonstrated use in the prognostication in prostate cancer. Higher BSP levels are related to a shorter time to develop bone metastases in patients with prostate cancer. OPN is not a marker of tumour burden but may be of use in assessing treatment response post chemotherapy in patients with CRPC.

Neuroendocrine differentiation (NED) represents a poor prognostic feature and is generally seen in higher-grade and higher-stage prostate cancers. It is reported that 10 to 100% of conventional prostate adenocarcinomas display evidence of NED immunohistochemically, with a higher incidence detected in castrate resistant disease. Neuroendocrine prostate cancer (NEPC) is uncommon and is associated with low PSA secretion and loss of androgen receptor expression. De novo small cell prostate cancer is an aggressive histological variant and is very rare in untreated patients (<1%).

Neuroendocrine cells store neuropeptides in cytoplasmic granules, including chromogranin A (CgA) and neurone-specific enolase (NSE); the most widely evaluated neuropeptides in prostate cancer. Neuropeptides may stimulate growth, differentiation and secretory processes, and can be detected in the serum [38]. Serum levels of CgA and NSE are associated with the degree of NED in prostate cancer cells. Serum levels of CgA are significantly higher in metastatic than in non-metastatic prostate cancer, and are adversely associated with survival in patients with CRPC and a Gleason score ≥8.

No relationship has been observed between baseline CgA and PSA response. CgA may be expressed by a subclone of prostate cancer cells in which NED affects clinical outcome without affecting the PSA response, i.e., NED represents an alternative AR independent mechanism of resistance.

Metabolic changes are affected by environmental, genetic and epigenetic factors. Metabolomic profiling of prostate cancer has been a promising area in the development of novel biomarkers. The prostate produces extremely high amounts of citrate, and citrate metabolism differs in prostate cancer compared to BPH.

The metabolite sarcosine (an N-methyl derivative of the amino acid glycine) is found in significantly higher concentrations in mPC than in non-mPC. It is thought to promote prostate cancer growth and progression, as in vitro studies demonstrated that administering exogenous sarcosine to benign prostate cells induced an invasive phenotype. Sarcosine has proven to be an independent prognostic factor in terms of PFS and OS in prostate cancer.

Serum glutamate are positively correlated to Gleason grade and prostate cancer aggressiveness. Higher serum levels of glutamate in patients with prostate cancer can be explained by an increased rate of glutaminolysis in proliferating prostate cancer cells or carboxypeptidase function of prostate specific membrane antigen (PSMA) overexpressed by prostate cancer cells.

The current risk stratification tools used in prostate cancer are inadequate. Additional serum biomarkers can potentially identify patients with more aggressive disease and enable the most effective treatments to be tailored to specific patients in the optimum sequence. This review highlights five different classes of promising biomarkers: miRNA, AR-Vs, bone metabolism, neuroendocrine and metabolic markers. They are all detected in the serum and are practical and minimally invasive to obtain. These biomarkers have been extensively investigated over the last decade, however none are routinely used in clinical practice. Further prospective trials are required for clinical validation.

In addition to the aforementioned serum biomarkers, genomic profiling of prostate cancer is an exciting emerging field. Evidence is accumulating associating specific genomic alterations with castrate resistance and metastatic progression. Genomic biomarkers that have been investigated in clinical trials include BRCA1/BRCA2/ATM for PARP inhibitors, PTEN/AKT for AKT inhibitors and PIK3CB for (PI3K)-β inhibitors. Genomic profiling could add further important predictive and prognostic information when managing patients with prostate cancer. Machine learning is a field that has developed with advances in computing and artificial intelligence. Machine learning models can rapidly analyse large data sets, and their use in identifying biomarkers from genomic data in prostate cancer is promising and attracting much research.