PSA is a 33 kDa serine protease produced by both normal and neoplastic prostate epithelial cells [1,2]. PSA is a member of the kallikrein gene family and the PSA gene is now termed KLK3, encoding for hK3 (Fig. 1) . The kallikrein genes (KLK1 to KLK15 encoding hK1 to hK15) all cluster within a 300-kb region on human chromosome 19q13.4  (Fig. 1). All genes have five coding exons and significant sequence homologies at the DNA and amino acid levels (40–80%) . At least 12 of the kallikreins are known to be regulated by steroid hormones [3,4]. Among all kallikreins, at least eight (KLK2–4, KLK10–13 and KLK15) are expressed in relatively high amounts in prostatic tissue . However, hK2 and hK3 expression is highly restricted to the prostate in males and for this reason these proteins are useful as biomarkers .
All of the kallikreins are serine proteases that are produced as ‘pre-pro’ enzymes, and all have a conserved position of the catalytic triad . Of the kallikreins that have been characterized, most have trypsin-like proteolytic activity because of the presence of aspartate in the substrate-binding pocket [3,4]. However, PSA is unique among the family in that it has chymotrypsin-like substrate specificity [6–8]. The major proteolytic substrates for PSA are the gel-forming proteins in freshly ejaculated semen, semenogelin I (SgI) and semenogelin II (SgII), which are synthesized and secreted by the seminal vesicles [1,7,9]. The ejaculatory mixing of the secretions from the seminal vesicles and the prostate results in the immediate formation of a loosely connected gel structure that entraps spermatozoa. SgI and SgII are the major structural proteins in the gel and enzymatically active PSA in the seminal fluid proteolytically cleaves preferentially tyrosyl and glutaminyl peptide bonds to generate many soluble fragments of SgI and SgII . These SgI and SgII cleavage sites have been used to generate a specific substrate for measuring PSA enzymatic activity . Other substrates for PSA have been described and include IGF-binding protein-3 , parathyroid hormone-related protein , fibronectin, collagen and laminin .
PSA is synthesized as a zymogen or prepro-peptide form consisting of a 17-amino acid pre-sequence that is removed intracellularly by signal peptidases and a seven amino-acid pro-sequence that is removed extracellularly (Fig. 2) [13,14]. For PSA, the amino acid sequence of this pro-peptide is Ala-Pro-Leu-Ile-Leu-Ser-Arg (APLILSR) . PSA must be correctly processed to become enzymatically active [15–17]. Under-processing and over-processing of the pro-peptide have been detected and in both cases result in the production of enzymatically inactive PSA (Fig. 2) . Incorrect processing affects the folding of the N-terminal portion of PSA and prevents hydrogen-bond formation between the N-terminal amino acid (i.e. Ile) and an amino acid near the catalytic centre (i.e. D194) .
The physiological protease responsible for the correct processing of pro-PSA to active PSA is unknown. The presence of arginine at the cleavage site suggests that pro-PSA is activated by an enzyme with trypsin-like substrate specificity. It has been shown in biochemical assays that inactive pro-PSA can be converted into active PSA by hK2, hK4 (prostase) and hK15 (prostin) [14–16,19,20]. Further studies are needed to determine whether one of these or other protease(s) are responsible for processing pro-PSA in vivo. This protease may itself be a useful biomarker. Identifying the PSA-processing protease may also yield insights into why under-processed, over-processed and complexed PSA proteins are present in human serum in varying ratios in patients with normal prostates compared to those with BPH or prostate cancer.
PSA AS A BIOMARKER OF NEOPLASTIC PROSTATIC DISEASE
Historically, high serum PSA levels have been associated with a diagnosis of prostate cancer. Lower PSA values of 4–10 ng/mL, are associated with lower rates of positive biopsy [21,22]. Overall, PSA testing is associated with an average lead time of 5–6 years for prostate cancer detection when a PSA level of 4.0 ng/mL is used as a threshold for biopsy . There is some degree of age-dependency for this threshold. Lower PSA levels of 3.5–4.0 ng/mL for men 50–70 years old, and 2.0–2.5 ng/mL for men 40–50 years old, are generally considered abnormal .
In 1995, Gann et al. reported that men with PSA levels of 2–3 ng/mL were five to six times more likely to be diagnosed with prostate cancer over the next 10 years than were men with PSA levels of < 1.0 ng/mL. The mean age of the 1464 men in that study at the time of baseline PSA measurement was 63 years. This result led Fang et al. to assess the long-term risk of prostate cancer as a function of PSA in younger men followed for longer period. This group reported an almost fourfold increase in prostate cancer risk for men with PSA values of > 0.6 ng/mL for men aged 40–49 years, and > 0.7 ng/mL for men aged 50–59 years than for men with PSA levels below these levels, respectively.
Although normal PSA levels increase with age, in clinical practice a PSA level of 4 ng/mL in the blood may be used, often when there are two consecutive rises in PSA level, as a threshold for further diagnostic studies, e.g. prostatic biopsy. However, PSA levels of 4–10 ng/mL also occur in men with BPH, and prostate cancer is present in only 25% of patients with PSA levels in this range. Based on this overlap and the realization that PSA is organ- rather than disease-specific, several investigators began to evaluate forms of PSA in serum in an attempt to improve the specificity of the PSA assay.
Shortly after PSA serine protease activity was discovered, two reports [26,27] showed that serum contained two distinct forms of PSA. The predominant form consisted of PSA covalently attached to the serum protease inhibitor α1-antichymotrypsin (ACT), i.e. PSA-ACT. A second form was also present that consisted of PSA that was not complexed to serum protease inhibitors (free PSA). These investigators and others also determined that PSA could form complexes with abundant α2-macroglobulin (A2M) present in the serum, but these PSA-A2M complexes are not detectable using standard immunoassays [26–29].
Subsequent to these early discoveries antibodies were developed that can specifically measure the free form of PSA and the total PSA (free plus ACT-complexed). Several studies were then reported using antibodies to determine the amounts of free and total PSA in serum . In one of the larger of these, Catalona et al. prospectively evaluated levels in 773 men with serum PSA levels of 4–10 ng/mL and biopsy-confirmed prostate cancer or BPH. They found that using a threshold of 25% for free PSA they would have been able to detect 95% of cancers, while avoiding 20% of unnecessary biopsies. The ultimate impact that measuring free/total PSA will have in reducing the number of unnecessary biopsies needs to be evaluated in future studies.
MOLECULAR FORMS OF PSA IN HUMAN SERUM
PSA is present in serum in several different forms, all of which are enzymatically inactive . These forms can be classified into two general categories: complexed PSA (i.e. bound to serum protease inhibitors) and free PSA (i.e. unbound, inactive PSA) [26,27]; 65–95% of PSA in the serum is PSA-ACT and this is the predominant form of PSA in the serum in men with normal prostates, BPH or prostate cancer . The remaining PSA consists of free PSA forms (5–35%) and PSA-A2M . A small amount of PSA is also found in complex with α1-protease inhibitor (PSA-API) . PSA-A2M and PSA-API are not routinely measured, although methods for determining the levels of these complexes are now available. Therefore, although in current clinical tests total PSA measures free PSA + PSA-ACT, the true amount of total PSA in the serum is the sum of all forms of PSA (i.e. free PSA + PSA-ACT + PSA-A2M + PSA-API) (Fig. 3).
PSA binds covalently to ACT and in vitro, this binding takes several hours at 37 °C . Once formed, the PSA-ACT complex is very stable and does not dissociate in vivo. The half-life of PSA-ACT has been difficult to calculate accurately because of non-exponential elimination, but available data suggest that PSA-ACT is cleared slowly and linearly over hours to days . In contrast, PSA binds very rapidly (in minutes) to A2M in human plasma [6,28,29]. PSA cleaves a peptide sequence in the bait region of the large macroglobulin protein causing a conformation change in the protein. The catalytic centre of PSA is not covalently bound to A2M, but PSA does become cross-linked to A2M at regions outside the PSA catalytic centre. PSA-A2M therefore can still hydrolyse small peptide substrates, but is unable to cleave larger protein substrates. Once the A2M undergoes conformational change after PSA cleavage of the bait region, it becomes a ligand for A2M-receptor/low density lipoprotein receptor-related protein present in the liver, and the PSA-A2M complex is subsequently rapidly cleared from the plasma with a half-life of a few minutes .
While PSA-ACT is the predominant complex found in human serum, the in vitro kinetics of complex formation suggest that PSA should preferentially bind to A2M. When enzymatically active PSA is added to human serum not containing PSA almost all of the PSA rapidly forms the PSA-A2M complex (Fig. 4). This result suggests that perhaps several orders of magnitude more enzymatically active PSA may be entering the blood and forming complexes with A2M, which are then rapidly cleared and not measured. The exact magnitude of this complex formation and clearance is unknown. Varying amounts of PSA-A2M complex could be produced in patients with BPH than with prostate cancer. The ability to accurately measure these PSA-A2M complexes (or the flux of PSA-A2M through the liver) therefore may help better discriminate BPH from prostate cancer [35,36].
The kinetics of the binding of PSA to ACT and A2M also raise questions as to the origin of the PSA-ACT complex in the blood (Fig. 4). ACT appears to be produced in small amounts in normal and BPH tissue, and in much higher amounts by prostate cancer cells [37–40]. In addition, Miller et al., using antibody microarray profiling, showed that the ACT protein is present in higher amounts in patients with prostate cancer than in controls. This would suggest that PSA-ACT complexes are pre-formed in the extracellular fluid of prostate cancers and BPH, and then these complexes leak into the blood. This question was addressed by Jung et al. who used immunoassays and gel filtration to define the molecular forms of PSA in malignant and benign prostate tissue. In that study most of the PSA within prostate cancers was free PSA; < 2% of the total detectable PSA in these assays was PSA-ACT.
These assays relied on extracting PSA forms using different antibodies and some of the results could be explained by differences in the efficiency of antibody extraction. However, they suggest that the amount of PSA-ACT in prostate cancers is lower than expected and thus again raise questions about the origin of the PSA-ACT complex.
In addition to studies evaluating the importance of PSA complexes as biomarkers, more recent studies have further characterized the uncomplexed free PSA forms in the blood. Free PSA in serum is composed of three distinct forms, all of which are enzymatically inactive. The first form consists of precursor forms of PSA in which the propeptide sequence has either not been processed, been under-processed, or over-processed (Fig. 3) . These forms appear to be more highly associated with prostate cancer. A second form of PSA, termed BPSA, is a degraded form of PSA that has been shown to be associated with prostate BPH transition zone tissue (Fig. 3) . BPSA contains two internal cleavage sites but remains intact in the serum because of disulphide bonds present in PSA . As yet unvalidated BPSA assays have been developed that show this isoform to be significantly higher in the serum of biopsy-negative men with elevated PSA levels . A third form of PSA appears to be composed of intact PSA that may have structural or conformational changes that cause it to be enzymatically inactive.
There is accumulating evidence that precursor forms of PSA are present in the serum of men with prostate cancer [46,47]. Normally, the conversion of proPSA to PSA is very efficient, and no proPSA forms have been found in the seminal plasma . In contrast, several proPSA forms have been found in the serum of men with prostate cancer, including the form in which all seven amino acids of the pro sequence are present (i.e. [-7]pPSA) as well as additional truncated forms in which some portion, but not all of the proPSA sequence, has been removed.
Mikolajczyk et al. developed monoclonal antibodies to [-7]pPSA, [-4]pPSA and [-2]pPSA and used them to examine serum of men with PSA levels of ≈ 10 ng/mL. In that study, in five men with prostate cancer over half of the free PSA in the serum was [-2]pPSA, while in three men who were biopsy-negative the [-2]pPSA levels were, on average, fivefold lower . The other forms were also detected although less consistently than the [-2]pPSA. This apparent enrichment of the [-2]pPSA may be associated with the observation that this form was more resistant to further complete processing to active PSA than the intact [-7]pPSA or [-5]pPSA. Other studies show that the [-7] and [-5] forms can be readily converted to mature PSA in vitro by hK2 or trypsin, while the [-2]pPSA form was completely resistant to activation by either hK2 or trypsin [14–16,47]. These results suggest that some disruption of the normal secretory pathway may be occurring in prostate cancer, possibly because of the lower levels of correct PSA-processing protease or the aberrant presence of a protease that incorrectly processes PSA to these enzymatically inactive, truncated forms. Further studies are needed to define the role of these truncated forms of PSA in the pathophysiology of prostate cancer, and to evaluate their utility as biomarkers for the diagnosis of prostate cancer.
QUESTIONS AND FUTURE DIRECTIONS
Much has been learned about the clinical utility of molecular forms of PSA as biomarkers over the last decade. It is now well recognized that PSA is organ-specific rather than disease-specific and thus may have potential not only as a diagnostic tool for prostate cancer but also for BPH. Newer forms, including the proPSA isoforms and BPSA, have been recently characterized and immunoassays to accurately measure these forms in serum are under development. Once these immunoassays have been validated, large clinical studies will be required to define the potential of these new PSA forms in differentiating BPH from prostate cancer.
The kinetics of PSA binding to serum protease inhibitors suggests that large amounts of PSA-A2M must be formed and rapidly cleared from the circulation. Previous studies have shown that measuring PSA-A2M complexes can increase the specificity of the free PSA to total PSA ratio  and predict bone metastases . PSA-A2M complexes are not easily measurable by standard immunoassays because PSA becomes engulfed upon binding to A2M, such that no PSA epitopes are accessible for antibody detection. New, high throughput assays need to be developed to measure PSA-A2M complexes, to evaluate this complex in clinical trials as a potential additional biomarker. Further studies are also needed to quantify amounts of PSA-A2M complex being formed in normal tissue, BPH and cancer. These studies could reveal that much higher amounts of PSA are produced than previously suspected, and the absolute total production of PSA may be of diagnostic and/or prognostic significance.
Studies to date show clearly that there are different forms of PSA present in differing amounts in normal tissue and prostate cancer. The mechanisms underlying these differences are not yet completely understood. Normal prostate epithelial cells secrete proPSA into the lumen of prostatic ducts where it is correctly converted to enzymatically active PSA by an as yet not fully characterized protease, which may be hK2. Even though the normal 20 g prostate produces ‘mg/mL’ quantities of PSA in the seminal plasma, only a very small amount of this PSA leaks into the systemic circulation to produce levels of PSA of < 1 ng/mL. In BPH, levels of PSA increase and a new form of PSA, BPSA, is detectable. The mechanisms underlying the production of increased total PSA and BPSA in BPH are also not understood. Internal cleavage of the PSA protein to produce BPSA suggests that PSA is ‘seeing’ a protease in the extracellular fluid of BPH that may not be present in normal tissue or prostate cancers.
The mechanisms whereby large amounts of PSA-ACT are produced by relatively small amounts of prostate cancer are also not understood. The assumption has been that, because the normal tissue architecture is disrupted, PSA rapidly leaks into the circulation where it complexes with ACT.
Since PSA was discovered in the 1980s much has been learned about its function as a serine protease and putative substrates identified. The role played by PSA in the physiology of the normal prostate and the pathophysiology of BPH and prostate cancer remains largely unknown. Several substrates for PSA have been defined in vitro, but it is unclear which of them are relevant in vivo. PSA is a protease and therefore could be involved in the initiation and growth of prostate cancer via the abnormal release of growth factors or proteolysis of growth factor binding proteins. It may also be involved in invasion and metastases pathways through degradation of collagen and laminin. Alternatively, the PSA molecule or one of its isoforms may itself be a signalling molecule.
The precise role of PSA in the pathophysiology of prostatic disease has been difficult to define, partly because there are no appropriate in vivo models. Although the value of PSA as a diagnostic tool for prostate cancer has been increasingly questioned, as described in this supplement, the utility in stratifying patients with BPH has been reinforced.