Detection of tumor-specific DNA in blood and bone marrow plasma from patients with prostate cancer



Tumor tissues, blood plasma and bone marrow (BM) aspirates of 57 prostate cancer patients (PCa) without clinical signs of overt metastases were assessed for LOH (loss of heterozygosity) by a PCR-based fluorescence microsatellite analysis, using a panel of 15 markers. Additionally, micrometastatic tumor cells in BM were monitored by an immunocytological cytokeratin assay. In total, 25 (44%), 32 (56%) and 41 (72%) of the patients had at least 1 LOH in their blood, BM and tumor samples, respectively. Among the informative cases, the frequency of LOH was highest in blood plasma for the markers D8S360 (18%) and D10S1765 (15%), and in BM plasma for THRB (24%) and D8S137 (22%). Comparison of blood plasma and BM with tumors showed discrepant results in 35% and 45% of patients, respectively. Whereas all LOHs at THRB in BM plasma were also detected in the autologous tumor tissues, LOHs at D6S474 and D11S898 in BM were not retrieved in the tumors. The comparison with established risk factors showed a correlation of borderline significance for LOH at D9S1748 in the BM aspirates (p = 0.055) and a significant correlation in the tumor samples (p = 0.004) with increasing pathologic Gleason scores. Interestingly, 22% of the PCa patients harbored tumor cells in their BM and tended (p = 0.065) to have more frequent LOH (16%) in BM plasma compared to patients without tumor cells (9%). These data demonstrate, for the first time, the presence of free tumor-specific DNA in blood and BM of PCa patients and suggest a possible relationship to BM micrometastasis. © 2006 Wiley-Liss, Inc.

Approximately 25% of patients with clinically localized prostate cancer (PCa) will eventually experience biochemical evidence of tumor recurrence after surgical resection of the primary tumor.1 A possible explanation for this clinical observation may be an early onset of tumor cell dissemination to the bone marrow (BM) in these men.2, 3 PCa preferentially metastasize into the skeleton system, and this spread is responsible for the considerable morbidity and mortality of this disease.1 Thus, several groups (including ours) have developed sensitive methods and were able to detect disseminated tumor cells (DTCs) in BM of 20–30% of PCa patients at the time of primary surgery years before the onset of overt metastases.4, 5, 6, 7, 8 In particular, enrichment techniques and immunocytochemical assays based on monoclonal antibodies specific for epithelial cytokeratins (CK) allow now the detection of single DTCs in BM of patients with solid epithelial tumors including PCa.2

Initiation and progression of PCa is accompanied by a multistep process with numerous genetic aberrations.9, 10, 11 The most consistent alteration in PCa is probably loss of heterozygosity (LOH). Allele typing and comparative genomic hybridization analyses have detected DNA deletions at diverse chromosomal loci in PCa.12 Cell biological studies on the introduction of deleted chromosomal DNA into a cell line have demonstrated that the deleted regions contain genes playing a substantial role in tumor growth and disease progression of PCa.13 The systematic use of microcell hybrid clones has identified a number of chromosomal regions which are important for the proliferation of prostate tumor cells.13 Significant stage-specific differences in the frequency of allelic losses were shown to specify genetic loci that may play key roles in PCa progression.14

LOH on free DNA has been identified in serum or plasma of peripheral blood from patients with carcinomas of breast, lung, gastrointestinal tract, kidney, head and neck, and melanomas.15, 16, 17, 18, 19, 20, 21, 22, 23 Despite the use of similar PCR-based techniques, the published studies showed a broad range of detection rates of LOH with contradictory results in lung, colorectal and breast cancer patients.19, 20, 24, 25, 26 Increased concentrations of extracellular DNA have been detected in the plasma from numerous cancer patients15, 27 including PCa patients28, 29, 30, 31 compared with the low amounts detected in the blood from healthy individuals. The origin of free DNA in blood and BM is still discussed, and it is supposed that free extracellular DNA, which is early released into the blood circulation during the formation of primary tumors, is derived from necrotic and apoptotic cells.32, 33 Since in contrast to the rapidly circulating blood, BM is a stationary system and an important site of tumor cell dissemination, BM plasma from PCa patients could be a crucial source of tumor-specific DNA. To date, findings of tumor-related genetic markers in BM plasma of breast cancer patients have previously been published.34

So far, investigations have only engaged the incidence of LOH in tumor tissues14, 35, 36, 37 but no study has focused on the detection of LOH on free plasma DNA from PCa patients. In view of the lack of tumor-specific molecular markers applicable to the detection of early micrometastatic spread, we selected a PCa-relevant panel of microsatellite markers and searched for tumor-derived DNA in cell-free blood and BM plasma of patients with apparently localized primary tumors. Besides the development of a new cell-free detection system for monitoring systemic tumor cell spread, the molecular analysis of DNA in BM and blood plasma samples may also lead to a better understanding of metastatic progression in PCa patients.


BL-LOH, LOH detected in blood plasma; BM, bone marrow; BM-LOH, LOH detected in bone marrow plasma; CK, cytokeratin; DTCs, disseminated tumor cells; LOH, loss of heterozygosity; MSI, microsatellite instability; PCa, prostate cancer; PSA, prostate specific antigen; Tu-LOH, LOH detected in tumor tissues; % fPSA, percent free prostate specific antigen.

Material and methods

Clinical and pathological data of PCa patients

Between October 2003 and December 2004, 57 consecutive men were included in this study which had approval from the local Hamburg research ethics committee. All patients were referred to the Department of Urology at the University Medical Center of Hamburg-Eppendorf for a prostatic evaluation due to either an abnormal digital rectal examination (DRE) and/or an elevated PSA (prostate-specific antigen) level. Clinical stage was assigned by the attending urologist according to the TNM classification 2002. All patients with a biopsy proven clinically localized prostate cancer patients (PCa) underwent radical retropubic prostatectomy. Prostatectomy specimens were processed according to the Stanford protocol and graded according to the Gleason score.

Of all 57 PCa patients, the majority (n = 49; 86%) showed an unsuspicious clinical stage T1c. Median and mean pre-treatment serum PSA values were 6.3 and 8.6 ng/ml, ranging from 1.7 to 36.2 ng/ml. Median and mean pre-treatment %fPSA values were 12.2 and 13.8%, ranging from 0.1 to 62.9%. Biopsy-based Gleason scores ≤(3 + 4) were recorded in 51 (89.5%) men. Of all PCa men, the majority (n = 47; 82.5%) demonstrated a localized stage at final pathology and pathological Gleason scores ≤3 + 4 (n = 52; 91.2%).

Laser microdissection and pressure catapultation of paraffin-embedded tissues

Sections of 5 μm thickness were cut from formalin-fixed, paraffin-embedded prostate tumor blocks and mounted on microscope slides. For the microscopic evaluation the paraffin slices were stained with hematoxylin and eosin (H/E, Merck, Darmstadt, Germany). The tumor tissue of paraffin slices was isolated in 2 manners. The relatively solid tumor tissues were scraped off fromH/E-stained slides without surrounding fibroblasts under a microscope. For multi-focal tumor tissues, scraping off was not possible without substantial contamination by healthy prostate tissue. Tumor foci were computer-assisted microdissected from unstained slides via laser pressure catapultation. An H/E-stained slide served as guideline assistance for distinction of the microscopic structures of tumor and normal tissue. Usually, 3 microscope slides from each patient were used for isolation of the tumor tissues.

Preparation of plasma and leukocytes

Twenty milliliters whole blood of each patient was collected in routine EDTA-containing tubes (EDTA Monovetten, Sarstedt GmbH, Nürmbrecht, Germany) and centrifuged at 2,500g for 10 min. The upper phase contained the blood plasma, from which 3–4 ml was removed for the extraction and analysis of the circulating DNA. The remaining 16–17 ml blood were supplemented up to 50 ml with lysis buffer containing 0.3 M sucrose, 10 mM Tris-HCl pH 7.5, 5 mM MgCl2 and 1 % Triton X100 (Sigma, Taufkirchen, Germany). Following incubation for 15 min on ice, the isolation and purification of leukocytes were carried out by 2 centrifugation steps at 2,500g, 4°C for 20 min.

Preparation of bone marrow specimens

Bone marrow (BM) specimens obtained in parallel with the blood samples were aspirated directly after surgery under general anesthesia from both iliac crests and screened for the presence of CK-positive cells and circulating tumor DNA. Ten milliliters of BM aspirates collected from the left and right anterior iliac crest and stored in heparinized tubes with DMEM (Gibco, Eggenstein, Germany) were processed immediately, according to our previous work.38 After centrifugation at 400g for 5 min the supernatant contained the cell-free plasma for further analysis of circulating DNA. The lower phase contained cellular components of the aspirate.

Enrichment and detection of disseminated tumor cells in bone marrow

The components of the lower phase of the BM aspirates were separated by density centrifugation with 20 ml Ficoll Hypaque (Amersham Biosciences, Freiburg, Germany) at 1,200g, 4°C, for 40 min. The interface layer containing mononuclear cells was collected. A concentration of 5 × 105 cells was cytocentrifuged (Hettich, Tuttlingen, Germany) onto glass slides (Menzel, Braunschweig, Germany) at 200g for 3 min and dried overnight for staining. Cells of the cancer cell line MCF7 (ATCC) served as positive controls. Following fixation of the slides with acetone for 10 min and washing with TBS, the slides were blocked with 10% of the blocking solution AB/PBS (Biotest, Dreieich, Germany) for 20 min and incubated with 2 μg/ml of the primary monoclonal antibody (Ab) anti-CK mAb A45-B/B3 (Micromet, Munich, Germany) or 2 μg/ml of the murine monoclonal antibody MOPC-21 (Sigma, Deisenhofen, Germany) for 45 min. The mAb A45-B/B3 specifically recognizes an epitope of CK-polypeptides, including the CK-heterodimers 8-18 and 8-19, and MOPC-21 served as IgG1 isotype negative control. After incubation of the secondary polyvalent rabbit antimouse immunoglobulin antiserum Z259 (Dako, Hamburg, Germany) for 30 min the visualization of antibody binding was performed by incubation with alkaline phosphatase anti-alkaline phosphatase complexes (Dako, Hamburg, Germany) for 30 min and Neufuchsin stain (Merck, Darmstadt, Germany) for 20 min.3 Finally, the slides were counterstained with hematoxylin (Merck, Darmstadt, Germany) for 20 sec. The evaluation of the slides for the presence of tumor cells was carried out on an automated cellular imaging system (ACIS, ChromaVision Medical System, San Juan, Capistrano, CA).

DNA extraction and fluorescence-labeled PCR

Genomic DNA was extracted from microdissected tumor tissues, blood plasma, BM plasma and leukocytes of each patient using the QIAamp DNA Mini Kit and a vacuum chamber (Qiagen QIAvac24) according to the manufacturer×s instructions (Qiagen, Hilden, Germany). The corresponding leukocyte DNA served as reference DNA. Quantifications and qualities of the isolated DNA were spectrophotometrically determined at 260 and 280 nm, respectively (Eppendorf, Hamburg, Germany). Dilution experiments to determine the lowest portion of tumor-specific DNA which can be detected were performed. For this study we mixed and amplified known quantities and proportions of normal (leukocyte) DNA and tumor or plasma DNA.

Ten nanograms of DNA was amplified in a 10 μl reaction volume containing PCR Gold buffer (150 mM Tris-HCl, pH 8.0 and 500 mM KCl), 2.5 mM MgCl2 (Applied Biosystems, Mannheim, Germany), 20 nM dNTPs (Roche, Mannheim, Germany), 0.4 pM of primer sets (Sigma, Taufkirchen, Germany) and 0.2 units of AmpliTaq Gold DNA-Polymerase (Applied Biosystems, Mannheim, Germany). The following microsatellite markers were used: D3S3703 (3q13.3), THRB (3p24), D6S474 (6q21-22), D6S1631 (6q16), D7S522 (7q31.1), D8S87 (8p12), D8S137 (8p21.1), D8S286 (8q21.3), D8S360 (8p21), D9S171 (9p21-22), D9S1748 (9p21), D10S1765 (10q23.3), D11S898 (11q22), D11S1313 (11q11-p11) and TP53.6 (17p13.1). The sense primer was fluorescence-labeled (HEX, FAM or TAMRA) at the 5′ end. The reaction was started with activation of the DNA polymerase at 95°C for 5 min followed by 40 cycles of PCR amplification. To verify the microsatellite alterations, each PCR was repeated at least twice.

Analysis of microsatellite alterations

The fluorescence-labeled PCR products were separated by capillary gel electrophoresis and detected by a fluorescence laser on the Genetic Analyzer 310 (Applied Biosystems, Mannheim, Germany). The evaluation of the peaks was performed by the Gene Scan Analysis program. The 500-ROX size marker served as an internal standard. The incidence of LOH was calculated by division of the quotient of the peak intensity derived from blood, BM plasma or tumor DNA by the quotient of the peak intensity derived from the corresponding leukocyte DNA. LOH was interpreted if the final quotient was <0.6 or >1.67. MSI was defined by the occurrence of a number of additional peaks. Tumors were assessed to have MSI, when it occurred at 1 marker. Homozygous and non-analyzable peaks as well as the non-accessibility of clinical material were designated as non-informative cases.

Statistical analysis

Statistical analysis was performed using the SPSS software package, version 11.0 (SPSS, Chicago, IL) and StatView, version 5.0 (SAS Institute Cary, NC). Fischer's exact test was employed to identify possible associations of the patterns of LOH or MSI with the clinical, histopathological and serum parameters (clinical stage, total tumor volume, histopathological tumor stage, resection margin status, pathological Gleason score, concentration of PSA in blood and percentage of free PSA). A value of p < 0.05 was considered to be statistically significant.


Frequencies of LOH in tumor, blood plasma and bone marrow plasma samples

Genomic DNA extracted from leukocytes, tumor tissues, blood plasma and BM plasma of 57 PCa patients was amplified by PCR using 15 different polymorphic microsatellite markers. Figure 1 shows a representative example of LOH in tumor tissue and BM plasma from a PCa patient. Table I outlines the results of the microsatellite analysis which include the PCa patients listed according to increasing Gleason scores, the number of LOHs detected in tumor (Tu-LOH), blood plasma (BL-LOH) and BM plasma (BM-LOH), and the LOH profiles at the different markers of each patient.

Figure 1.

Example of LOH detected in tumor tissue and BM plasma from a PCa patient. The fluorescence-labeled PCR products of leukocytes, tumor and BM plasma DNA were separated by capillary gel electrophoresis on a Genetic Analyzer and evaluated with the Gene Scan Analysis program. The abscissa indicates the length of the PCR product, while the ordinate gives information on the fluorescence intensity represented as peaks. The upper diagram shows the leukocytes DNA (reference) and the lower diagrams show tumor and BM plasma DNA amplified with the primer as indicated in the upper diagram. In the lower diagrams the arrows refer to LOH.

Table 1. Summary of Gleason Scores, Number of LOH and MSI, and the Incidence of LOH and MSI at 15 Different Polymorphic Markers in Tumor, Blood Plasma and BM Plasma of 57 PCa Patients
  1. Abbreviations: Tu tumor, BL, blood plasma; BM, bone marrow; neg, no CK-positive cells; pos, number of CK-positive cells detected in the BM aspirates; Total, number of all LOHs found in tumor, blood and BM plasma samples; equation image, LOH; equation image, MSI; equation image, retention of heterozygosity; equation image non-informative case.

original image

Using the marker panel 52 (91%) men of the examined 57 PCa patients had at least one LOH or MSI in any of their samples (Table I). The LOH index (number of LOH of each patient divided by the informative cases) of the respective patients was significantly higher in primary tumors (72%) than in BM (56%) and blood plasma (44%) samples (p < 0.0001) and higher in BM than in blood plasma samples (p < 0.0005). The frequency of LOH was 5% and 11% in all analysis of the paired blood and BM plasma samples, respectively, performed with the 15 different microsatellite markers. Concordance of LOH identified in the tumor and blood samples was observed in 65% of the informative cases, whereas in 35% of the analyses LOHs were detected in the blood plasma samples but not in the paired tumor samples. LOHs which we identified in the BM plasma samples could be retrieved in 55% of the paired tumor samples. Surprisingly, only 8% of the BM-LOHs could be restored in the matched blood plasma samples.

The bar chart in Figure 2 represents the frequencies of detected Tu-, BL- and BM-LOH aberrations at each microsatellite marker. In addition, Table II complements the data of Figure 2 and lists the number of Tu-, BL- and BM-LOH and informative cases for each marker. Among the informative cases, the frequency of BL-LOH was highest for the chromosomal loci 8p21 (D8S360, 18%) and 10q23.3 (D10S1765, 15%), and in BM plasma 3p24 (THRB, 24%) and 8p21.1 (D8S137, 22%) were predominantly affected regions. In the tumor tissues LOH at 8p21.1 (D8S137, 50%) and 8p12 (D8S87, 44%) were most frequent (Fig. 2, Table II). As shown in Table I, all LOHs at THRB recorded in the BM aspirates were also detected in the tumor tissues of the same patient, whereas the BM-LOHs at D6S474 and D11S898, and to lesser extent the BM-LOHs at D9SS1748 could not be retrieved in the matched tumor tissues. In contrast to the marker D6S1631 which displayed a somewhat heterogeneous spreading of LOH, the other markers (D7S522, D8S87, D8S137, D8S286, D8S360, D9S171 D10S1765 D11S1313, TP53.6) showed a homogeneous profile (Table I). At the markers D7S522 and D8S286 LOH was found neither in blood nor in BM plasma. With the exception of the marker D8S137, the loci at which no LOH was found in the blood plasma (THRB, D7S522, D8S286, D11S898) or BM plasma (D7S522, D8S286, D11S1313) also had a low rate of these LOHs in the analogous tumors (<20% of each locus, Fig. 2).

Figure 2.

Comparison of the frequency of LOH at 15 different chromosomal markers detected in tumor tissues, blood and BM samples. The frequency of LOH was calculated by division of the number of LOHs with the informative cases for each locus, as shown in Table II.

Table II. Number of LOH and MSI Detected in Tumor, Blood and BM Plasma at the Different Microsatellite Markers
  1. Inform, informative samples.


In contrast to the LOH data, the frequent MSIs, which were associated with the loci 11q11-p11 (D11S1313, 51.5%) and 9p21-22 (D9S171, 30%) in the tumor tissues, could not be recovered in the blood plasma and BM aspirates (Table II). Figure 3 shows a representative example of such a tumor-associated MSI, which was defined by the occurrence of additional peaks with a size different from that seen in the normal leukocyte DNA.

Figure 3.

Example of a tumor-associated MSI. The upper diagram shows the leukocytes DNA (reference) and the lower diagram tumor DNA amplified with the primer as indicated in the upper diagram. In the lower diagram, the occurrence of additional peaks symbolizes MSI.

Correlation to clinicopathological risk factors

The LOH profiles detected at the microsatellite markers in BM plasma samples as well as in primary tumor tissues were correlated with the following risk parameters: clinical stage, total tumor volume, histopathological tumor stage (pT), resection margin status, pathological Gleason score, concentration of PSA in blood and the percentage of free PSA that is not bound to proteins. The Gleason score indicates the differentiation grades of the prostate tumors. The statistical evaluation of Tu-LOH showed a significant correlation between increasing pathologic Gleason scores and LOH at D9S1748 (p = 0.004). In our collective 40 PCa patients were informative at this locus. As shown in Figure 4 and Table I, all 3 patients with high Gleason scores ((4 + 3) and (4 + 4)) also harbored Tu-LOHs whereas the other 37 patients with lower Gleason scores ((3 + 2), (3 + 3) and (3 + 4)) had predominantly no Tu-LOH. A borderline significance (p = 0.055) could be calculated for the positive correlation between increasing Gleason scores and the detection of LOH at this marker on free DNA in BM plasma (Fig. 4). The statistical analyses of the LOH aberrations detected in tumor and BM samples with the other aforementioned parameters did not reach statistical significance.

Figure 4.

Relationship of the frequency of Tu- and BM-LOH at the marker D9S1748 with increasing pathologic Gleason scores.

Because of the low frequency of LOH detected in the blood plasma, a statistical assessment of these samples with the clinical parameters would be inconclusive and was therefore not performed.

Relationship of the detected LOHs in BM plasma and tumor tissues with the presence of disseminated tumor cells in bone marrow

To determine whether a correlation exists between BM-LOH and dissemination of tumor cells into the BM, we additionally monitored BM for CK-positive cells. Cytokeratins form a part of the cytoskeleton of epithelial cells and are at present the most reliable markers for detection of DTCs in BM.2, 38 BM samples from the PCa patients were assessed by Ficoll density gradient centrifugation and an immunocytological CK assay with the anti-CK antibody A45-B/B3. The antibody is well evaluated and has already been applied for BM analysis to obtain clinical relevant information on patients with different solid tumor types including PCa.3, 4, 5, 6, 7, 8 Out of 57 PCa patients, we were able to investigate BM from 55 patients, and detected DTCs in 12 (22%) samples. A representative example of a CK-positive cell in the BM of a PCa patient is shown in Figure 5.

Figure 5.

Staining and visualization of CK-positive cells in BM with the antibody A45-B/B3 and the alkaline phosphatase anti-alkaline phosphatase technique. Presence of a CK-positive cell in BM of a PCa patient. [Color figure can be viewed in the online issue, which is available at]

As shown in Figure 6a, the association of the frequency of BM-LOH with the occurrence of DTCs demonstrated that PCa patients with DTCs in their BM tended to have more frequently BM-LOH than patients who harbored no tumor cells in their BM (Table I). Eight of 12 (67%) patients with DTCs in BM and 24 of 43 (56%) patients with no DTCs had at least one LOH on free DNA in BM plasma (Table I). In patients with DTCs in BM, 86 LOH-analyses of free DNA in BM plasma were informative and 14 (16%) LOH aberrations were detected. In contrast, patients with no DTCs in BM showed LOHs on free DNA in only 36 (9.5%) of 373 informative analysis. This difference was of borderline significance (p = 0.065).

Figure 6.

Relationship of the frequency of BM-LOH (a) and Tu-LOH (b) with the presence of DTCs in BM. (a), The incidence of BM-LOH at the microsatellite markers THRB, D8S87, D8S137, D9S171, D9S1748, D11S898 and TP53.6 was higher in DTCs-positive (pos.) BM than in DTCs-negative (neg.) BM. (b), The incidence of Tu-LOH at the microsatellite markers THRB and TP53.6 was higher in DTCs-positive (pos.) BM than in DTCs-negative (neg.) BM.

Next, we evaluated whether specific genetic changes are determinants of micrometastatic spread to BM. The association of the frequency of Tu-LOH with the occurrence of DTCs in BM did not show a striking overall pattern (Fig. 6b). However, a higher incidence of Tu-LOH at the marker THRB could be observed when BM was DTCs-positive (30%) in comparison to DTCs-negative BM (10%). Conversely, at the marker D10S1765, the frequency of Tu-LOH was higher when BM was DTCs negative (35%) than positive (11%; Fig. 6b). It is of interest that for BM-positive cases LOH at the marker D11S898 could only be found in BM plasma and not in tumor tissues (Table I, Fig. 6).


Microsatellite analysis of blood plasma and BM aspirates could be a particularly attractive, new approach for detection of tumor-specific DNA, because blood is easily accessible by simple venous puncture and BM is the most frequent site of metastatic relapse in PCa. To the best of our knowledge, no study to date has assayed extracellular tumor-specific DNA in blood and BM from PCa patients. For our assay, we defined a specific panel of 15 microsatellite markers which was suitable for detecting Tu-LOH, BL-LOH and BM-LOH aberrations in 72, 44 and 56% of the analyzed PCa patients, respectively. These data show that the LOH index of the respective patients was significantly higher in primary tumors than in BM and blood plasma samples (p < 0.0001) and higher in BM than in blood plasma samples (p < 0.0005).

In contrast to the common occurrence of LOH in tumor, blood and BM, MSI was only detected in the tumor tissues. Using our marker set, the most affected regions were represented by the loci 11q11-p11 (D11S1313) and 9p21-22 (D9S171). Our results are consistent with the data from Dahiya et al., who used a larger microsatellite marker panel and showed that ∼45% of primary tumors harbor MSI.39 The reason for the absence of MSI in blood and BM revealed in our study remains unclear.

The concordance of LOH aberrations identified in 65% of the blood and 55% of the BM plasma samples with the analogous primary tumors and the fact that the most frequent BM-LOHs at the chromosomal locus 3p24 (THRB) were all found in the corresponding tumor tissues suggest that at least part of the free DNA in blood and BM may originate from the primary tumor. Our investigation was performed using DNA derived from an arbitrary area of the tumor. The subsets of BL-LOH and BM-LOH, which were not concordant with the detected tumor alterations, might be therefore due to the known heterogeneity of the PCa tumors. Prostate tumors arise often multifocally and each focal area may harbor different genetic alterations.40 In addition, PCa tumor samples are characterized by sections of stromal cells containing normal intact DNA. To avoid contamination with normal DNA from these cells, we have microdissected the tumor tissues.

The statistical evaluation of our LOH data revealed a correlation between increasing pathologic Gleason scores and LOH at the tumor marker D9S1748 in BM and primary tumor. This chromosomal region encodes the tumor suppressor product CDKN2/p16, an inhibitor of the cyclin dependent kinase (CDKN2). Immunohistological data showed a high expression of p16 in normal epithelial cells of the prostate, while the expression decreased in prostate tumors.41

Our findings have shown that LOH on cell-free DNA from PCa patients was less frequently detected in the paired blood plasma (5%) than in BM plasma (11%) which we obtained in parallel, suggesting that BL-LOH might be masked by the prevalence of normal DNA in blood42 and that free tumor-specific DNA might accumulate in BM of PCa patients. Taback et al.,34 who reported the occurrence of tumor-specific DNA in BM plasma from 11 (23%) of 48 patients with breast cancer, also detected tumor-specific DNA with a higher frequency of LOH in BM than in blood plasma of these patients. While 65% of LOHs that we identified in the blood plasma samples could be retrieved in the paired tumor samples, only 8% of LOHs that we detected in the BM samples could be restored in the matched blood plasma samples. In contrast to the BM-LOHs at marker THRB, which were all found in the paired tumor samples, the LOHs at D6S474 and D11S898 in BM plasma, and partly the BM-LOHs at D9SS1748, were not detected in the tumor samples. The reasons for these profiles of LOH are unknown. Apart from the dilution of the tumor-specific plasma DNA by normal DNA and the heterogeneity of the prostate tumors, it is possible that the patterns of LOH, particularly the profile of BM-LOH, may not only reflect the profile of the primary tumors but also that of the disseminated tumor cells.

Since BM is a major homing site for metastatic cells in PCa,2 it was of interest to evaluate whether there is a correlation between the detected BM-LOH profiles and the presence of DTCs in BM. There is mounting evidence that the presence of DTCs identified by CK immunostaining is associated with an unfavorable outcome in patients with primary breast cancer.43, 44 The specificity of the CK test has been shown in a previous study on almost 200 noncancer control patients.38 In BM from PCa patients without overt metastases, the presence of CK-positive DTCs has been shown to be related to an unfavorable outcome.6 However, cytokeratins are no tumor-specific markers, and human epithelial antigen, prostate-specific transcripts (e.g., PSA mRNA) or tumor-associated MAGE transcripts have therefore also been used as markers for DTCs.4, 45, 46 Although these transcripts are usually not amplified from normal BM cells, they are not expressed by all prostate cancer cells and can occur in non-tumor cells.47, 48

In the present investigation, 22% of the BM samples from PCa patients contained DTCs identified by CK-immunostaining. Furthermore, our data show that the subset of PCa patients with DTCs in their BM had a trend to have more BM-LOH than patients who harbored no tumor cells in their BM, indicating that an increase in BM-LOH might be associated with the presence of DTCs in BM. It is conceivable that the free tumor-specific BM DNA could be related to the rate of turnover of the DTCs and that the free DNA might originate from both the DTCs and the primary tumor. To substantiate this hypothesis, a reliable technology platform for isolation of single DTCs from BM is required. Analysis of the DNA of single DTCs derived from BM and blood samples from patients with various types of cancers by comparative genomic hybridization of their DNA has shown a high genetic divergence of the DTCs and the paired tumors,49 and the DTCs in breast cancer patients harbored fewer genomic aberrations than the corresponding primary tumors.50 This could explain at least in part the divergent results observed between BM and primary tumors in our study. From recent microarray studies, there is increasing evidence that the metastatic capacity of solid tumors is already determined early in the primary tumor.2 We therefore analyzed the relationship between LOH aberrations in the primary tumor and the presence of DTCs in BM. Although for most markers we failed to observe consistent differences between primary tumors from patients with or without DTCs in BM, two genomic regions (THRB and D10S1765) appeared to be associated with the BM status. While THRB showed a positive correlation to the presence of DTCs in BM, LOH at the locus D10S1765 was less frequent in primary tumors of BM-positive patients, suggesting opposite roles of genes at these genomic areas in metastatic spread.

In conclusion, the genomic analysis of cell free BM plasma might lead to a new approach to detect micrometastatic spread in cancer patients. However, large clinical studies with long term follow up in relation to defined endpoints (e.g. metastatic relapse) are necessary to substantiate this claim. The future investigation of the origin and role of the free DNA in BM may provide additional insights into metastatic progression. Free DNA may also play an active role in this process because spontaneous passages of DNA from prokaryotic and eukaryotic cells to eukaryotic cells have been described.24


We thank Ms. Antje Andreas, Hannelore Suck, Andrea Speckmann and Jessica Gädicke for their excellent technical assistance. We are indebted to Ms. Hannelore Ellinghausen and Karin Beutel for performing the microdissection of the tumor tissues. We also thank Dr. Thomas Rau for support in the statistical evaluation of the data. Finally, we are grateful to Prof. B. Brandt for many useful discussions.