World Journal of Urology

, Volume 25, Issue 1, pp 19–30

Molecular staging of prostate cancer in the year 2007

Authors

    • Martini-Clinic, Prostate Cancer CenterUniversity Medical Center Hamburg–Eppendorf
  • Andreas Erbersdobler
    • Department of PathologyUniversity Medical Center Hamburg–Eppendorf
  • Martina Mirlacher
    • Department of PathologyUniversity Medical Center Hamburg–Eppendorf
  • Guido Sauter
    • Department of PathologyUniversity Medical Center Hamburg–Eppendorf
Topic Paper

DOI: 10.1007/s00345-007-0153-z

Cite this article as:
Schlomm, T., Erbersdobler, A., Mirlacher, M. et al. World J Urol (2007) 25: 19. doi:10.1007/s00345-007-0153-z
  • 81 Views

Abstract

Numerous attempts towards improving patient management by molecular staging have been fruitless so far. No single molecular parameter is routinely analyzed in prostate cancer tissue. This may be partly due to genuine properties of prostate cancer that may make this tumor a difficult target. Furthermore, inherent logistical problems result in a shortage of prostate cancer tissue for research purposes. For the future, it can be hoped that the availability of more powerful molecular techniques in combination with better tissue archives will allow more rapid progress. Powerful DNA array and proteomics methods allow the systematic analysis of virtually all genes of a cancer on the DNA, RNA, and protein level. Although such approaches are sometimes labeled as “fishing expeditions,” it cannot be totally disregarded that the simultaneous analysis of all genes has a high likelihood of identifying significant new information. In future, one of the major scientific challenges will be the validation of several potential biomarkers in large enough and clinically well-characterized patient cohorts. In particular, studies on needle core biopsies and hormone refractory cancers are imperatively needed for investigating the natural history of the disease or to discover potential predictive markers for radiation therapy and new therapeutic target genes to answer the clinically most important questions for optimal clinical decision making in prostate cancer patients: which patients will not require local therapy? If local therapy is needed, what is the treatment of choice? What medications should be given if metastases are present?

Keywords

Prostate cancerGenesPrognosisPredictionTargeted therapyMolecular stagingMicroarray

Introduction

Prostate cancer (PCA) is a relatively simple disease. Most prostate cancers have a rather benign natural history. It has even been suggested that many patients may not need treatment. As long as there are no metastases, most cancers can locally be cured either by surgery or radiation therapy. Only if metastases are present, cure is no longer possible. In these patients, a combination of endocrine and cytotoxic therapies can substantially improve quality of life and potentially also prolong survival [1]. For clinicians, three questions are of paramount importance for optimal clinical decision making in prostate cancer patients. These are: which patients will not require local therapy? If local therapy is needed, what is the treatment of choice? What medications should be given if metastases are present?

It is obvious that our current diagnostic procedures must be specifically improved to address these questions. Given the advances in molecular medicine, it is our hope and assumption that molecular cancer staging may become critically helpful in this respect. To optimize initial therapy, better information on the local and distant tumor extension would be most useful, both of which could be either achieved through better molecular characterization or imaging strategies. Increasing evidence suggests that these two approaches could potentially be combined in a “molecular imaging” strategy. Applying prognostic markers could help to identify the “pussycat” tumors that will be unlikely to require immediate treatment. And finally, a better prediction of response to specific drugs could be achieved through either DNA array analyses of “molecular tumor signatures” or by specific analysis of therapeutic target genes.

Almost 10,000 scientific articles have been published on molecular features in prostate cancer between 1996 and 2006, according to Medline. Many of these studies suffered from small patient groups. Most of these studies do not link their data to clinical endpoints. A limited Medline search for the keywords “prostate + cancer + immunohistochemistry + prognosis” and a review of these abstracts identified only 267 studies comparing specific molecular features with prognostic relevant clinical endpoints (Fig. 1). The target genes of these studies are summarized in Table 1. Most of these studies included fewer than 100 patients, typically with heterogeneous treatment and diagnostic procedures of involved subjects. Remarkably, many studies that generated stunning results have surprisingly little impact. Most biomarkers that were shown to be significantly associated or were of independent prognostic importance in small pilot studies were subsequently not followed up further neither by the original authors nor by other investigators. The number of biomarkers that have been intensively evaluated by several different groups is more limited. Some published data will be summarized below.
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig1_HTML.gif
Fig. 1

Medline analysis (“prostate” + “cancer” + “immunohistochemistry” + “prognosis” + “year1996–2006”): a From 3,408 listed articles, only 267 provide sufficient correlation to IHC results and PCA prognosis. b Size of study cohorts (inset) (red with prognostic relevance; blue without prognostic relevance)

Table 1

Prognostic relevance of all investigated proteins in 267 studies (see Fig. 1)

Protein

Studies

Prognostic relevant

Not prognostic relevant

p53

29

5

Ki-67

22

5

Bcl-2

18

8

AR

15

1

CD44

9

 

Her-2

9

5

p27

8

3

CgA

7

5

PSA

7

2

alpha-catenin

5

 

E-cadherin

5

1

NSE

5

2

Bax

4

3

Beta-catenin

4

 

p21

4

1

Caveolin-1

3

 

COX-2

3

 

NF kappa B

3

 

p16

3

1

PCNA

3

1

serotonin (5HT)

3

1

survivin

3

2

FGF8

2

 

hMSH2

2

 

IGFbp-3

2

 

MYC

2

 

Pin-1

2

 

PLK1

2

 

pRB

2

 

PSMA

2

 

TGF-beta1

2

 

VEGF

2

 

actin

1

 

Adrenomedullin

1

 

AKT/PKB

1

 

Akt-1, Akt-2, Akt-3

1

 

AMACR

1

 

AP-2

1

 

AZGP1

1

 

BAG-1

1

 

BARK1

1

 

beta-MSP

1

 

BMP

1

 

BMP2

1

 

cathepsin D

1

 

CD24

1

 

CD31

1

 

CD34

1

 

clusterin

1

 

cyclin D1

1

1

desmin

1

 

EGFR

1

2

Endoglin

1

 

Ep-CAM

1

1

FAS

1

1

FGF17

1

 

FHIT

1

 

Her-3

1

 

hsp27

1

1

hsp60

1

 

hsp70

1

 

Human protectin (CD59)

1

 

Hyaluronan

1

 

Id proteins (Id-1, 2, 3, 4)

1

 

IGF-1

1

 

IGF-2

1

 

Mab

1

 

MDM2

1

1

MIC-1

1

 

MIF

1

 

MMPs

1

 

MUC1

1

 

Mucin1

1

 

NEP

1

 

NPY

1

 

Oncoprotein 18

1

 

Osteopontin

1

 

p120

1

 

P-Akt-1

1

 

Pepsinogen C

1

 

PPP1CA

1

 

S-100 protein

1

 

sFRP4

1

 

Smad4

1

 

Smad8

1

 

Stat5

1

 

Syndecan-1

1

 

TEF3

1

 

Tiam1

1

 

TSP-1

1

 

VCP

1

 

vimentin

1

 

Wnt-1

1

 

ZAG

1

 

ADCP

 

1

beta1C

 

1

bFGF

 

1

caspase-3

  

CDK1

 

1

CDK2

 

1

CDK6

 

1

CK 18

 

1

Cyclin D3

 

1

delta-Catenin

 

1

DSPP

 

1

GRK2

 

1

hK2

 

1

iNOS

 

1

NM23–H1

 

1

NSF

 

1

p62

 

1

PPP2CB

 

1

VWf

 

1

It is possible that some of the discrepancies in the results of different groups are caused by the absence of standardized laboratory methods for the analysis of these proteins.

The examples of other tumors, perhaps most impressively in the case of breast cancer, have shown that molecular staging can be pivotal for cancer patient management. In breast cancer, estrogen receptor, progesterone receptor and HER2 analyses are routinely done to assess the receptiveness to anti-hormonal or gene-specific therapy and molecular profiling has been shown to be substantially predictive for the prognosis and response to cytotoxic therapy [2]. It is remarkable that comparable, generally accepted, clinically relevant molecular markers are not available for prostate cancer, even though this tumor has a comparable frequency and clinical importance. This “research delay” may be partly due to the genuine properties of prostate cancer that may make this tumor a difficult target. However, some of our lack of research success appears to be also caused by inherent logistical problems that result in a massive shortage of prostate cancer tissue for research purposes. Very much in contrast to breast cancer, unfixed/frozen tissue is only rarely collected from prostate cancer, which practically disables molecular profiling attempts and many other kinds of analyses. Moreover, tissues from patients undergoing watchful waiting or having advanced androgen insensitive cancers are hardly available.

Independent of the causes for the misery of our prostate cancer research, we have to accept the current lack of generally accepted clinically helpful molecular markers. For the future, it can be hoped that the availability of more powerful molecular techniques in combination with better tissue archives will allow more rapid progress.

Potential prognostic markers

p53

p53 is a tumor suppressor gene involved in the maintenance of genomic integrity and apoptosis control. p53 inactivation occurs mostly through mutation of one allele together with physical loss of the second copy of the gene. As most p53 DNA mutations result in an inactivated gene product with increased half-life, p53 inactivation often results in a nuclear accumulation of p53 protein. Thus, p53 inactivated cancers are typically “p53 positive” by immunohistochemistry (IHC) (Fig. 2a). More than 800 studies have analyzed p53 in prostate cancer. The range of p53 altered cancers ranges from 4 to 61% in these studies [3, 4]. This broad range of different results reflects the inherent methodological problems in measuring p53 protein levels by IHC. As the protein is physiologically expressed in all cells, an increased assay sensitivity will lead to “false positive” results. It is conspicuous, however, that most studies with a low rate of p53 altered cancers, typically between 2.5 and 20%, found associations between p53 alterations and poor patient prognosis [58]. The results of studies that included p53 sequencing do also suggest that the true rate of p53-mutated cancers is not much higher than 10% [911]. Two studies suggesting a role of p53 alterations for radiation resistance involved only 53 and 54 patients, respectively [12, 13]. In several other cancers, initial reports of a predictive role of p53 alterations for radiation therapy success were subsequently not confirmed [1416].
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig2_HTML.jpg
Fig. 2

Examples of immunostainings using the tissue microarray (TMA) technique. a p53 positive. b Chromogranin-A, weak positive, c PSMA strong positive, d HER2 negative, e HER2 strong positive, f EGFR strong positive

bcl-2

bcl-2 is an apoptosis inhibiting protein. bcl-2 was found to be expressed in 1–69% of prostate cancers [4, 17]. Based on the outcome differences between bcl-2 positive and bcl-2 negative prostate cancers after antiandrogen therapy [1820] or irradiation [21], it was suggested by several authors that bcl-2 might predict clinical response to these treatments. However, similar outcome differences are also observed between bcl-2 negative and bcl-2 positive cancers that were treated by radical prostatectomy alone [2225]. This observation strongly argues for the possibility that elevated bcl-2 expression may be a marker of poor prognosis, independent of treatment. In our opinion, the bcl-2 expression status should therefore—based on the existing data—not be used as a parameter for therapeutic decisions. This is all the more true, as no generally accepted assays exist that could be used for standardized bcl-2 analysis.

Ki67 labeling index

Tumor cell proliferation is an obvious parameter of prognosis in prostate cancer and most other tumors. Quantitation of Ki67 protein, expressed in all nuclei of cells in G2, S-, and M-phase of the cell cycle has been shown to be a highly efficient way for proliferation measurement. The percentage of Ki67 positive cells (Ki67 labeling index; Ki67 LI) typically ranges between 2 and 15% in prostate cancers. This is lower than in most other cancer types, which reflects the long natural history of prostate cancer. Several studies have shown that the Ki67 LI is prognostically relevant [2628]. This also includes studies that analyzed only small cancer portions in core needle biopsies. That small tissue cores in the range of what can be collected in biopsies can be sufficient for obtaining prognostically relevant information is also demonstrated in an own tissue microarray study. In an analysis of one sample measuring only 0.6 mm in diameter from 2,385 prostate cancers, a clear association between Ki67 LI and PSA recurrence could be found (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig3_HTML.gif
Fig. 3

Influence of the cell proliferation marker Ki67 (Ki67-labeling index) on PSA recurrence

The Ki67 LI is also strongly related to the Gleason score. It thus remains unclear, how much additional clinically significant information is provided by Ki67 LI analysis.

Androgen receptor and PSA expression

Several studies have suggested a poor prognosis in patients with reduced PSA or androgen receptor (AR) expression in their tumors. This observation may be related to a general dedifferentiation of these cancers. Accordingly, none of these studies have shown an independent prognostic relevance of PSA expression changes. The potential predicitive role of AR alterations is discussed below.

E-Catherin and α/β-catenin

E-Catherin and catenin proteins are intercellular adhesion molecules. The role of E-Catherin in the morphogenesis of lobular breast carcinoma and signet-ring cell carcinoma of the stomach posed the question whether adhesion molecules could also be important in prostate cancer. Indeed, a reduced expression of these markers could be observed mainly in tumors with a higher Gleason score indicating a discohesive growth pattern of tumor cells [29, 30]. A downregulation could also be demonstrated in prostate cancers treated by androgen deprivation [31]. An independent prognostic value of α-catenin in local and locally advanced prostate cancers could be found in a large study with 215 cases and a mean follow-up period of 13 years [32].

P27Kip1

P27Kip1 is an inhibitor of cyclin-dependent kinases and a negative regulator of cell proliferation [33]. However, genetic aberrations of the P27Kip1 gene are rare and the expression is mainly regulated on a post-transcriptional level [34, 35]. In the normal prostate, P27Kip1 is expressed in secretory cells and partly also in basal cells [34, 36]. A reduced expression was found in 19.5–84% of prostate cancers and was associated with a higher T-catgory, higher Gleason scores and worse prognosis [34, 3640].

On the other hand, a reduced expression of P27Kip1 could already be observed in benign hyperplastic glands from the transition zone by some investigators [34, 41].

Neuroendocrine differentiation

Neuroendocrine differentiation, typically detected with antibodies directed against neuron specific enolase (NSE) or chromogranin A (Fig. 2b) occurs in 24–98.5% of prostate cancers [42, 43]. Neuroendocrine differentiation belongs to the most frequently discussed potential prognostic prostate cancer markers in textbooks and review articles [44] and has also been suggested to predict poor response to radiation and antiandrogen therapy [4547]. Some centers therefore analyze neuroendocrine differentiation in their prostate biopsies on a routine basis. A review of the existing study data shows, however, that the scientific evidence for utilizing neuroendocrine differentiation on a routine basis is rather weak. Only 7 of 16 studies investigating the prognostic impact of neuroendocrine differentiation showed positive associations with poor prognosis [42]. Similarly, as already outlined for bcl2, we feel that it is incorrect to assume a predictive relevance of neuroendocrine differentiation for response to anti-androgen and radiation therapy and to base therapeutic decisions on this parameter. This is based on three reasons: (1) There are only seven studies suggesting a predictive role for neuroendocrine differentiation and these studies only involve 25, 33, 71, 72, 90, 104 and 105 patients, respectively. (2) Similarly, as discussed for bcl2, there are also studies that may suggest a prognostic role of neuroendocrine differentiation in patients treated by surgery alone. This again raises the possibility that neuroendocrine differentiation is a prognostic and not a predictive parameter. (3) There are no standardized methods for measuring neuroendocrine differentiation.

Prostate specific membrane antigen (PSMA)

Prostate specific membrane antigen is strongly expressed on the membranes of most prostate cancers. As compared to normal prostate tissue, it is highly upregulated in about 90% of cancers (Fig. 2c). Prostate cancers with high levels of PSMA were found to have a particularly poor prognosis in some studies [4851]. As cell surface proteins are optimal targets for targeted cancer drugs, PSMA could also be a suitable drug target. Clinical studies are currently underway to test this interesting opportunity.

Potential predictive markers

An increasing number of cancer drugs target specific molecules (targets) expressed on cancer cells. Typical examples include trastuzumab targeting HER2 positive breast cancer, imatinib targeting KIT positive GIST or various anti-EGFR drugs targeting epidermal growth factor receptor (EGFR) altered lung or colon cancers. Evidence is accumulating that such drugs that are currently only approved for few indications may also be efficient in other cancer types carrying similar aberrations. Alterations of target genes of established cancer drugs in prostate cancer is therefore of high topical interest.

HER2

HER2 is a remarkable example of an error influenced scientific discussion in prostate cancer. HER2 is a transmembranous thyrosine kinase receptor, which is the target for trastuzumab (herceptin) in breast cancer. Most HER2 amplified breast cancers respond strikingly well to trastuzumab treatment both in metastatic and adjuvant indications [2]. Frequent HER2 amplification had also been described in prostate cancer in early studies using non-stringent criteria for gene amplification [52]. Based on these studies, trastuzumab has been extensively discussed in prostate cancer and the drug was applied—without success—in clinical trials. Subsequent studies on thousands of prostate cancers using FDA approved reagents and standardized criteria for gene amplifications had then shown a virtual absence of gene amplification in prostate cancer [5356]. It appears, however, that low level HER2 expression (in the absence of gene amplification) occurs in 16–36% of prostate cancers and that these tumors may have a deteriorated prognosis [4, 53, 56] (Fig. 2e).

Epidermal growth factor receptor

Epidermal growth factor receptor is another member of the HER2 gene family and the target for several drugs including antibodies and small inhibitory molecules [57]. Expression of EGFR was detected by IHC in 1–100% of prostate cancers [4, 5860] (Fig. 2f). High EGFR expression may be associated with poor prognosis [61, 62]. However, studies utilizing anti EGFR drugs in prostate cancer patients showed little success [63, 64]. This observation parallels findings in lung cancer where a high-level overexpression of EGFR was insufficient to predict significant clinical response. From the findings in lung cancer, it appears, that only cancers with a DNA aberration (mutation or amplification) will respond to therapy. EGFR gene amplification occurs in less than 1% of prostate cancer and EGFR exon 18–21 have not been sufficiently analyzed as yet [60].

Biomarkers for predicting response to radiation and antiandrogen therapy

Recent data support the concept of clinical radiosensitivity as a complex genetically controlled event. It is reasonable to assume that potential predictive markers could be found among proteins involved in DNA repair [6567], cellular signaling [68, 69], cell proliferation [7073] and apoptosis [13, 21, 7479]. However, clinical data supporting this concept have not been collected. As outlined above for bcl-2 and p53, it is important to understand that biomarkers found to be linked to clinical outcome in irradiated or hormonally treated patients are not necessarily related to response to therapy, but could also represent prognostic markers independent of therapy.

Known mechanisms for resistance to hormone therapy are DNA-based alterations in the AR gene, such as amplifications or mutations [80, 81]. Neither AR amplification nor mutations have been found in untreated primary tumors. Great hope has been placed on a possible role of these alterations for defining second line treatment strategies in hormone resistant cancers. For example, it was expected, that total hormone blockage should be the therapy of choice in AR amplified cancers in which highly upregulated AR should be able to utilize very low (rest) blood levels of testosterone for growth stimulation. However, clinical trials investigating patients after antiandrogen therapy could not find a clinical utility of detailed analysis of the type of AR alterations [82, 83].

Other potential mechanisms for hormone resistance include ligand-independent activation of the AR (e.g. H-Ras, growth factors and interleukins) [8487] and alternative signaling pathways, bypassing the AR [81, 88]. Neuroendocrine differentiation is also discussed as a potential mechanism for hormone independent tumor growth. Much of this is based on the observation that neuroendocrine cells in normal prostate do not express androgen receptor. Prospective clinical studies demonstrating a predicitive role of neuroendocrine differentiation are lacking, however. The same applies to other suggested predictive biomarkers such as bcl-2 [1820], p53 [89], p21 [90, 91], fibronectin [92] and fatty acid synthase (FAS) [93, 94].

Future strategies

It is obvious that our strategies to search for prostate cancer markers should be better adjusted to aggressively search for answers to the initially described clinical key questions: which patients will not require local therapy? If local therapy is needed, what is the treatment of choice? What medications should be given if metastases are present? The key prerequisite for successful research is the availability of appropriate tissue resources. Given the fact that radical prostatectomy is one of the most frequent oncological surgeries, it is unfortunate that so many scientific studies have been executed on small patient sets (Fig. 1). The tissue microarray (TMA) approach enables the simultaneous analysis of more than 600 cancer samples on one tissue slide. This method solves many of the previous logistical obstacles for large-scale research. Our own prostate TMA includes samples from more than 3,000 patients that underwent prostatectomy at our clinics (Fig. 4). Several studies using this TMA have shown that prognostic factors can be clearly identified by this approach in prostate cancer. However, many of the necessary studies cannot be done on tissues from radical prostatectomies. Tissues from needle core biopsies are imperatively needed for investigating the natural history of the disease or potential predictive markers for radiation therapy. Tissues from hormone refractory cancers are required to investigate predictive markers for new therapies, as the character of the tumor changes massively under antiandrogen therapy, and the selected hormone refractory clone is unlikely to be identifiable in the pretherapeutic biopsy.
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig4_HTML.jpg
Fig. 4

a Prostate cancer tissue microarray (TMA) containing samples from 3,200 patients that underwent radical prostatectomy at the University Medical Center Hamburg–Eppendorf. b Representative tissue spot (PSA immunostaining)

Our modified approach of TMA manufacturing from core needle biopsies (Fig. 5) is optimally suited for analyzing large numbers of biopsies from untreated patients. In this approach, fragments from previously collected needle core biopsies are perpendicularly re-embedded in an empty paraffin block similar to the classical TMA approach. Such needle core TMAs may enable finding associations of molecular features with clinical outcome in patients that underwent radiation or opted for a watchful waiting strategy.
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig5_HTML.jpg
Fig. 5

Model of prostate needle core biopsy TMA. a Needle core biopsy with marked cancer. b Corresponding cancer containing tissue spot on needle core biopsy TMA

Large-scale resources of unfixed tissues are required for identifying new and better prognostic and predictive biomarkers in the future. Powerful DNA array and proteomics methods allow the systematic analysis of virtually all genes of a cancer on the DNA, RNA and protein level (Fig. 6) [95]. Although such approaches are sometimes labeled as “fishing expeditions” by traditional scientists, it cannot be totally disregarded that the simultaneous analysis of all genes has a high likelihood of identifying significant new information. It is noteworthy that AMACR, one of the most established diagnostic prostate cancer markers in histolopathology has been discovered in cDNA microarray experiments [96].
https://static-content.springer.com/image/art%3A10.1007%2Fs00345-007-0153-z/MediaObjects/345_2007_153_Fig6_HTML.jpg
Fig. 6

cDNA Microarray (GeneChip) analysis of laser microdissected prostate tissue: Tumor versus adjacent normal tissue (35 patients). Shown are the 45 most significantly deregulated genes out of 37,531 analyzed cDNA closes (ANOVA, P < 0.0005); red upregulated, blue downregulated

The standardization of more generally available techniques will be another major challenge in our campaign towards a molecular staging of prostate cancer. IHC, for example, is done in almost every laboratory involved in cancer research. However, the accumulated literature data suggest significant deficiencies in our IHC work. There is no other explanation for the dramatically high variability in the percentage of “positive cases” between different studies analyzing the same target protein. Extended education in IHC would be extremely important to improve the reproducibility of IHC studies.

In summary, prostate cancer is an important and frequent disease. Despite this, our attempts toward improving patient management by molecular staging have been fruitless so far. Not one single molecular parameter is routinely analyzed in prostate cancer tissue. Modern screening technologies are now available to easily identify candidate biomarkers. Once we achieve to validate such candidates in large enough and clinically well-characterized patient cohorts, a much better understanding of individual prostate cancers based on their molecular signature will soon become possible.

Copyright information

© Springer-Verlag 2007