Mitochondrial DNA in serum of patients with prostate cancer: a predictor of biochemical recurrence after prostatectomy


Jörg Ellinger, Klinik und Poliklinik für Urologie, Universitätsklinikum Bonn, Sigmund–Freud–Strasse 25, 53105 Bonn, Germany. e-mail:



To investigate the role of circulating mitochondrial DNA (mtDNA) in patients with localized prostate cancer, as recent reports show that patients with advanced cancer have increased levels of mtDNA.


DNA was isolated from the serum of 100 patients with prostate cancer and 18 with benign prostate hyperplasia (BPH). A quantitative real-time polymerase chain reaction was used to amplify 79 bp and 230 bp fragments of the mitochondrial 16s-RNA gene, the short fragment representing total mtDNA, including mtDNA truncated by apoptosis, and the long fragment representing mostly mtDNA from other cell death entities. mtDNA integrity was defined as the ratio of long to short mtDNA fragments.


The short and long mtDNA levels, and mtDNA integrity, were similar in patients with BPH and cancer (= 0.940, 0.211 and 0.441, respectively), and were not correlated with clinical or pathological variables, e.g. age, prostate-specific antigen (PSA) level, cT stage, pT stage, seminal vesicle infiltration, lymph node invasion, or Gleason score (P = 0.075 to 0.961). However, patients with high levels of short mtDNA (>75th percentile) had a greater risk of PSA progression and this variable was the strongest predictor of PSA recurrence in a multivariate Cox analysis (= 0.023; hazard ratio 0.31; 95% confidence interval 0.113–0.851).


Circulating mtDNA levels did not distinguish between patients with prostate cancer or BPH. However, there was a significant increase in short mtDNA fragments in patients with early PSA recurrence after radical prostatectomy.


mitochondrial DNA


radical prostatectomy


quantitative PCR.


Prostate cancer is one of the most common malignancies in the world [1]. The early diagnosis of prostate cancer was improved decisively in the 1980s with the advent of PSA testing. Nevertheless, identifying patients with significant cancer remains a challenge for urologists. The development of additional noninvasive prognostic biomarkers should therefore be helpful for decision making in patients with localized prostate cancer.

Previously it was reported that circulating DNA levels are increased in the serum/plasma of patients with various malignancies [2]. Although circulating DNA was already detected in the plasma of patients with cancer in the late 1970s [3], knowledge about these nucleic acids remains largely enigmatic. Only a small percentage of circulating DNA is derived from tumour cells [4,5] or tumour-infiltrating lymphocytes [4]. It seems that most circulating DNA originates from healthy ‘bystander’ and peripheral cells, and it was reported that circulating DNA levels increase on stimulation with apoptosis- or necrosis-inducing drugs [4]. Recently, it was shown that either short DNA fragments (<200 bp, derived from apoptotic cells in prostate cancer) [6] or long fragments (>200 bp, derived from necrotic cells in breast and colon cancer) [7,8] were increased in the serum of patients with cancer, depending on the tumour entity. This indicates that pathophysiological mechanisms leading to an increase in circulating DNA can differ in different tumour entities. Nevertheless, circulating DNA seems to be a convenient cancer marker, and increased DNA concentrations and the fragmentation pattern might be useful for distinguishing between patients with cancer or no cancer in various malignancies, including prostate [2,6–10]. Furthermore, high concentrations and increased fragmentation might also be associated with cancer progression, as in prostate cancer [6,11].

Previous studies focused mainly on the detection of free circulating genomic DNA in patients with cancer [4,6,8,10,11]. In comparison, mitochondrial DNA (mtDNA) not only circulates in a cell-free form, but also in a particle-associated form [12]. In this context, Mehra et al.[13] showed that circulating concentrations of genomic DNA and mtDNA were not correlated with each other in plasma, which might be caused by the different compartmentalization and degradability of mtDNA and genomic DNA. In oncology, mtDNA levels have been shown to be of diagnostic and prognostic value, and their quantification in bodily fluids is highly accurate [14]. For instance, patients with advanced prostate cancer have significantly greater mtDNA levels than those with benign prostatic disease. Also, mtDNA levels were high in patients with a poor outcome [13]. In patients with head and neck cancer, the saliva contains higher levels of mtDNA than in those with no cancer [15], and these levels decreased after radiotherapy [16]. Thus, the aim of the present study was to determine if mtDNA levels could also provide valuable information for managing patients with localized prostate cancer, a group in which mtDNA has so far not been examined. We analysed mtDNA levels and its fragmentation in the serum of patients and compared these to men with BPH.


We included 100 randomly chosen serum samples of patients with prostate cancer treated with radical prostatectomy (RP), and of 18 with BPH who had TURP at the Department of Urology, University Hospital Bonn, Germany; Table 1 shows the clinicopathological variables. Follow-up information was available for 92 patients with cancer, of whom 17 had biochemical recurrence after RP (defined as one serum PSA level of >0.2 ng/mL). The median (range) follow-up was 23 (1–72) months.

Table 1.  The clinicopathological variables of patients with prostate cancer or BPH
Mean, median (range) or nProstate cancer (100)BPH (18)
Age, years65.9, 66.0 (53–79)69.1, 69.5 (55–80)
Prostate weight, g55.8, 52.0 (21–167)55.8, 50.0 (18–135)
Preop PSA, ng/mL 11.0, 27.6 (0.19–55.8) 5.93, 3.72 (0.69–21.62)
Not available 2 1
 Not suspicious60 
 Suspicious for cT237 
 Not available 3 
Pathological stage
 pT4 2 
Surgical margins +ve36 
Capsular penetration33 
Seminal vesicle infiltration13 
Lymph node invasion 3 
Gleason score
 3 + 4 = 713 
 4 + 3 = 7 8 

DNA levels are lower in plasma than in serum and it was assumed that DNA release during clotting of fragile cells is the reason for this difference. A recent report showed clearly that DNA concentrations in serum are six times those in plasma, but the contribution of extraneous DNA from ruptured cells is negligible [17]. Thus, we used serum for the present study. All blood samples were collected before surgery in serum S-Monovette Gel tubes (Sarstedt, Nürnbrecht, Germany) containing a clot-activation additive and barrier gel. Clotting of serum samples was allowed for ≥60 min before centrifugation (1800g, 10 min) and the supernatants were stored at −80 °C. All patients had given written informed consent, according to the institutional guidelines, before inclusion into the study. The QIAamp Ultrasens Virus Kit (Qiagen, Hilden, Germany) was used to isolate circulating DNA from 2 mL of serum. DNA was isolated according to the manufacturer’s protocol (elution volume 60 µL AVE buffer).

We designed two primer sets for the mt-specific 16s-RNA. The first primer pair (mtDNA79) amplified a 79-bp fragment that corresponds to total mtDNA and includes DNA truncated by apoptosis. The second primer pair (mtDNA230) amplified a 230-bp fragment that correlates with mtDNA mainly from non-apoptotic origin (i.e. necrosis and other types of cell death, producing larger DNA fragments) [4]. The sequence for the forward primer (both, mtDNA79 and mtDNA230) was 5′-CAG-CCG-CTA-TTA- AAG-GTT-CG-3′, and the reverse primers were 5′-CCT-GGA-TTA-CTC-CGG-TCT-GA-3′ (mtDNA79) and 5′-GGG-CTC-TGC-CAT-CTT-AAC-AA-3′ (mtDNA230), respectively. To date, no mutations have been reported in these selected regions in patients with prostate cancer. The degree of fragmentation was defined as mtDNA integrity and represents the ratio of mtDNA230 to mtDNA79 copies.

Quantitative real-time PCR (qPCR) experiments were performed in triplicate on an ABIPrism 7900HT (Applied Biosystems, Foster City, CA, USA). Each 10 µL reaction consisted of 1 × SYBRGreenER Mix (Invitrogen, Paisley, Scotland), 400 nm forward/reverse primer and 0.7 µL of DNA sample. qPCR conditions were 90 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 60 s. Melting-curve analysis was used to confirm the specificity of the PCR products. Each run included serial dilutions of an external standard and water blanks. Samples were analysed with no previous knowledge of the specimen identity.

To create a standard curve, we amplified a 490-bp DNA fragment that covered both mtDNA fragments (primer sequences: forward 5′-GGG-ATA-ACA-GCG-CAA-TCC-TA-3′ and reverse 5′-ATG-TTG-GGA-CCT-TTG-CGT-AG-3′). The PCR product was purified (Qiagen PCR purification Kit) and the DNA concentration determined using the PicoGreen dsDNA Quantification Kit (Molecular Probes, Eugen, OR, USA). The copy number of the 490 bp mtDNA fragment was calculated according to the protocol ‘Creating Standard Curves with Genomic DNA or Plasmid DNA Templates for Use in Quantitative PCR’ (Applied Biosystems).

Differences in mtDNA levels in the serum from patients with BPH or prostate cancer were analysed using the Mann–Whitney test. Prostate weight was correlated with mtDNA levels using the Spearman test. Correlations between clinicopathological variables were assessed using the Mann–Whitney and Kruskal–Wallis tests, as appropriate. Kaplan-Meier analysis and the Cox proportional hazard model were used to correlate the period of PSA-free survival with mtDNA levels.


Using the short mtDNA79 and long mtDNA230 primer sets, the mtDNA fragment levels were at higher levels in men with prostate cancer than in those with BPH (mean mtDNA79, 2.25 × 106 vs 1.61 × 106 copies/mL; and mean mtDNA230, 1.06 × 106 vs 0.62 × 106 copies/mL, respectively). However, these differences were not statistically significant (= 0.211 and 0.441, respectively; Mann–Whitney test). Furthermore, the mtDNA integrity (defined as the ratio of long to short mtDNA fragments), an indicator of the underlying cell death entity (i.e. apoptosis vs necrosis), was comparable in both groups (mean 0.314 vs 0.306; P = 0.940), showing that mostly short mtDNA fragments are formed (Fig. 1).

Figure 1.

Circulating mtDNA fragment levels in serum, determined by real-time qPCR using primer sets for a 79-bp (A) and 230 bp (B) gene fragment of the mitochondrial 16 s-RNA as described. mtDNA integrity (C) was defined as the ratio of long to short mtDNA levels.

We then analysed whether mtDNA fragment levels or mtDNA integrity were correlated with clinicopathological variables (i.e. age, preoperative PSA level, prostate weight, pT stage, capsular penetration, seminal vesicle infiltration, extraprostatic extension, surgical margin status, Gleason score). Both mtDNA fragment levels and mtDNA integrity were not significantly associated with these variables (= 0.120 to 0.903).

Follow-up information was available for 92 patients with prostate cancer, of whom 17 had biochemical recurrence after RP. A univariate Cox proportional hazard model was used to analyse whether mtDNA fragments in serum, or clinicopathological variables, were correlated with biochemical recurrence. Patients with short mtDNA fragment levels >75th percentile had a significantly greater risk of biochemical recurrence after RP (= 0.044; Fig. 2). Also, the pT stage (= 0.016) and extraprostatic extension (= 0.009) were significantly correlated with PSA progression. Interestingly, some established variables associated with poor outcome had no statistically significant correlation with biochemical recurrence (i.e. seminal vesicle infiltration, P = 0.053; Gleason score, P = 0.069; Table 2). Finally, using a multivariate Cox regression model including the pT stage and extraprostatic extension, high levels of short mtDNA (i.e. levels >75th percentile) significantly correlated with PSA recurrence/progression (= 0.023; hazard ratio 0.312; Table 2), whereas the other variables did not. mtDNA levels as a continuous variable did not correlate with PSA recurrence (= 0.750). In summary, these findings suggest that circulating mtDNA levels in serum are an independent prognostic variable in patients with prostate cancer.

Figure 2.

Patients with 79 bp mtDNA levels of >75th percentile had a greater risk of biochemical recurrence after RP (log rank P = 0.036). Among 17 patients with PSA recurrence, seven had 79 bp DNA levels of >75th percentile, vs 16 (21%) with no PSA recurrence. (A) Kaplan-Meier analysis and (B) boxplots for patients with (left) and without (right) PSA recurrence.

Table 2. 
Univariate and multivariate Cox proportional hazard models for biochemical recurrence in 92 patients with clinically localized prostate cancer
VariablePHazard ratio (95% CI)
  • *

    Invalid analysis; large variance.

 mtDNA > 75th percentile0.0440.368 (0.139–0.975)
 mtDNA 79 bp0.7501.000 (1.000–1.000)
 mtDNA 230 bp0.5831.000 (1.000–1.000)
 mtDNA integrity0.8101.229 (0.228–6.622)
 Preoperative PSA0.9541.001 (0.963–1.041)
 Suspicious DRE0.2271.835 (0.685–4.939)
 Age0.2120.949 (0.874–1.030)
 Capsular penetration0.1710.514 (0.198–1.332)
 Lymph node invasion0.1494.570 (0.582–35.899)
 Surgical margins0.0120.472 (0.181–1.230)
 Gleason score0.0691.286 (0.981–1.686)
 Seminal vesicle invasion0.0533.089 (0.984–9.699)
 pT-stage0.0161.575 (1.089–2.277)
 Extraprostatic extension0.0090.277 (0.105–0.728)
 mtDNA > 75th percentile0.0230.312 (0.114–0.852)
 pT stage0.914*
 Extraprostatic extension0.935*


Circulating DNA levels are greater in patients with malignancies than in healthy individuals [2], but importantly also than in patients with nonmalignant disease, e.g. infections or cardiovascular, autoimmune and musculoskeletal disease [9]. Although the pathophysiological mechanisms that cause the increase in circulating DNA remain largely unknown, many studies support the idea that circulating DNA is a useful tumour marker, with excellent sensitivity in various tumour entities [6–11]. Previous studies focused on the detection of genomic DNA in serum/plasma of patients with cancer and prostate cancer [6,10,18], and a more recent study showed that mtDNA levels are also useful for predicting the outcome in patients with advanced malignancies, including of the prostate [13]. In this study we evaluated the diagnostic and prognostic information of mtDNA levels in patients with localized prostate cancer.

The present results show that mtDNA levels were not correlated with any clinicopathological variables (i.e. age, serum PSA, prostate weight, DRE results, pT stage, lymph node metastasis, seminal vesicle invasion, capsular penetration). We chose the biochemical recurrence rate (or the PSA-free survival time) as a reasonable factor to assess the prognostic value of mtDNA fragments, because a sufficient follow-up to assess cancer-specific mortality had not yet been reached. In the 92 patients with a follow-up of up to 72 months, high levels of short mtDNA fragments were significantly correlated with early biochemical recurrence after RP. Importantly, the multivariate Cox proportional hazard model showed that a high mtDNA level was the strongest predictor of PSA recurrence. Although the follow-up was relatively short we assume that this finding has important prognostic implications, because the risk of cancer-specific mortality is especially greater in patients with PSA recurrence within the first 3 years after RP [19]. Nevertheless, our data might over- or underestimate the prognostic value of mtDNA, as a strong endpoint such as cancer-specific mortality had not been reached. Additional studies including patients with a longer follow-up might provide a stronger rationale for prognostic relevance. We assume that the analysis of mtDNA in serum might help to identify patients who benefit from early multimodal therapy. Similarly, Mehra et al.[13] reported an association of high mtDNA levels and cancer-specific survival in patients with advanced prostate cancer. In addition, mtDNA in serum/plasma could serve as a postoperative surveillance marker, especially in patients with PSA-negative primary cancer. It was shown in various malignancies that cell-free DNA levels decreased after initial therapy and increased in those patients with recurrent disease [20,21].

In the present study short (= 0.211) and long (= 0.441) mtDNA levels were not significantly different in patients with BPH and localized prostate cancer. This is surprising, because previous studies showed that genomic DNA levels are up to seven times higher in patients with prostate cancer [6,10]. Mehra et al.[13] reported a three-fold difference between patients with advanced cancer and those with non-malignant urological disease (i.e. prostatitis, BPH, benign bladder disease). However, they did not investigate mtDNA levels in patients with early prostate cancer. The varying proportions and different forms (i.e. cell-free and particle-associated) of genomic DNA and mtDNA in the circulation suggest a different origin of these nucleic acids. In a previous study we reported a predominantly apoptotic origin of genomic DNA in prostate cancer compared to patients with BPH [6]. By contrast, mtDNA fragmentation patterns were similar in BPH and prostate cancer in the present study. Therefore the release mechanism of mtDNA fragments in patients with BPH and localized prostate cancer seem to be alike. Patients with more aggressive cancer (i.e. early PSA recurrence after RP) had significantly higher mtDNA levels, suggesting that a different pathophysiological mechanism leads to the release of mtDNA in aggressive tumours. Possibly the enhancement of anti-apoptotic mechanisms in advanced malignancies protect mitochondrial destruction and mtDNA, thus causing higher mtDNA levels. Also the cell death of other tissue types and bystander cells in the sequel of aggressive prostate cancer might induce higher mtDNA levels in serum.

There were relatively many patients with locally advanced prostate cancer in the present study (extraprostatic in 35% of the patients). Because there is no PSA-based screening for prostate cancer in Germany, more patients are diagnosed with locally advanced disease than in countries with PSA screening. Thus, there could be differences if our results are transferred to other populations, especially in a screened population.

We did not use an internal control for the isolation of DNA; in preliminary tests the DNA isolation was highly reproducible (data not shown). Furthermore, the manufacturer of the Ultrasens Virus Kit showed that linear DNA isolation is possible over a large range of DNA concentrations (i.e. 107−102 DNA molecules). We therefore abandoned the use of an internal control. Future studies should use an internal control to allow a better control for isolation biases.

Circulating mtDNA levels cannot be used to distinguish between patients with localized prostate cancer and BPH; nevertheless, patients with early biochemical recurrence after RP have significantly greater mtDNA levels. Thus, the quantification of mtDNA levels in the sera of patients with prostate cancer might be useful in therapeutic decision making. Prospective, large-scaled multicentre studies are necessary to confirm the value of circulating mtDNA in clinical practice.


The study was supported by BONFOR grants to Jörg Ellinger and Patrick J. Bastian.


See Acknowledgements.