Phase II study of cytarabine in men with docetaxel-refractory, castration-resistant prostate cancer with evaluation of TMPRSS2-ERG and SPINK1 as serum biomarkers


Anthony Joshua, Medical Oncology Department, Princess Margaret Hospital, 610 University Avenue, Toronto, ON, M5G2M9, Canada. e-mail:


Study Type – Therapy (cohort)

Level of Evidence 2b

What's known on the subject? and What does the study add?

To date, there has been limited impetus to examine the use of cytarabine in prostate cancer. We presented preliminary laboratory data to suggest its utility in the castration refractory prostate cancer (CRPC) population which, combined with a previous case report, suggested it may have hitherto unrecognized utility in this setting. Embedded in this study was peripheral blood sampling for TMPRSS2-ERG and SPINK1, two genes that are believed to define prostate cancer genotypes, to assess their utility as biomarkers

This study suggests that at the delivered doses, cytarabine has limited efficacy and significant myelotoxicity suggesting, it does not have a role in the treatment of docetaxel-refractory CRPC. The presence of serum TMPRSS2-ERG and SPINK1 mRNA biomarkers recovered from blood suggest that their analysis is worthy of further study.


  • • To run a phase II clinical trial of cytarabine in men with docetaxel-refractory, castration-resistant prostate cancer (CRPC), based on evidence that cytarabine might be effective in men with abnormalities of ERG oncogenes.
  • • To measure mRNA levels of prostate cancer-related genes in blood as biomarkers.


  • • Ten of a planned maximum of 30 men received i.v. cytarabine at doses of 0.25–1g/m2 at 21-day intervals. The primary endpoint was prostate-specific antigen (PSA) response.
  • • Archival tumour samples were assessed by fluorescence in-situ hybridization for TMPRSS2:ERG translocation, and by immunohistochemistry for serine peptidase inhibitor Kazal type 1 (SPINK1).
  • • Blood was processed for mRNA quantification of TMPRSS2:ERG (exon1:exon4), SPINK1 and PSA.


  • • A PSA response was not observed in any patient. The trial was stopped at the end of stage 1 of a modified Flemming design.
  • • The median number of cycles administered was 3. Grade 3–4 haematological toxicity was common. Five patients were subsequently excluded from the study for toxicity, and five for disease progression.
  • • Analysis of whole blood mRNA for T1:E4 translocation in TMPRSS2:ERG was consistent with that in the tumour in 8/9 evaluable cases (one was concordantly positive, seven were concordantly negative), SPINK1 results were concordant in 9/10 cases (two were concordantly positive, seven were concordantly negative [P= 0.047 for the predictive value]).
  • • There was no correlation between PSA or SPINK protein and their respective mRNA copy levels in blood.


  • • Cytarabine at the doses used is ineffective for men with CRPC.
  • • Blood mRNA levels of prostate cancer genes may represent a novel aspect of monitoring prostate cancer and have implications for the understanding of tumour-derived mRNA.

castration-resistant prostate cancer


ETS transcription factor


serine peptidase inhibitor, Kazal type 1


upper limit normal


present pain intensity


Functional Assessment of Cancer Therapy–Prostate


Response Evaluation Criteria in Solid Tumours


progression-free survival


adverse event


fluorescence in-situ hybridization




Prostate cancer is the most common cancer diagnosis, and the second most common cause of cancer death in men in North America [1]. Until the recent approval of abiraterone, cabazitaxel and sipuleucel-T, patients could expect a median survival of ≈18 months after the diagnosis of castration-resistant prostate cancer (CRPC) with docetaxel-based chemotherapy as the standard treatment [2–4].

There is now an increased understanding of the molecular biology of prostate cancer. The fusion gene TMPRSS2-ERG (present in up to 70% of prostate cancers [5–7]) brings together an androgen-responsive transmembrane serine protease, with members of the ETS family of transcription factors (ETS transcription factors [ETFs]). The resulting transcript leads to the overexpression of ETFs, which have been implicated in stem cell development, cell senescence, proliferation, invasion/migration and resistance to apoptosis in the prostate [6,8]. TMPRSS2-ERG-negative prostate cancers are thought to be hetergeneous but include subgroups that include serine peptidase inhibitor Kazal type 1 (SPINK1)-positive cancers (∼10% of all prostate cancers) [9]

Before their discovery in prostate cancer, translocations of the ETF gene family had been recognized as relevant to the pathogenesis of Ewing's sarcoma; however, it is difficult to develop drugs to inhibit transcription factors. Stegmaier et al. [10], therefore, developed a high-throughput assay to screen US Federal Drug Administration-approved drugs to identify agents that led to the down-regulation of a 14-gene EWS/FLI gene signature, which is the ETF fusion gene implicated in Ewing's sarcoma. Cytarabine (ara-C) was identified by this mechanism and led to decreases in EWS/FLI protein expression consistent with decreased target protein translation or increased degradation. These studies led us to hypothesize that cytarabine might be efficacious in other cancers driven by similar molecular pathways, such as prostate cancer. Pre-clinical studies also suggested activity of cytarabine in prostate cell lines that do not have TMPRSS2-ERG translocations (see Supporting Information).

A case report of a patient with osteoblastic bony metastases from prostate cancer, who was diagnosed with acute leukaemia, supports our hypothesis. While receiving multi-agent (hyper-CVAD) chemotherapy for his leukaemia (which included high doses of steroids and cytarabine at 4 doses of 3g/m2) he experienced a rapid decline in serum PSA concentration with improvement in the radiographic appearance of bony disease. He subsequently suffered a relapse of his leukaemia and died from sepsis, but at autopsy there was no evidence of metastatic prostate cancer [11].



Eligible patients were men with docetaxel-refractory CRPC, diagnosed histologically or by a serum PSA concentration >20 ng/mL with a clinical presentation consistent with metastatic prostate cancer. Castration independence was documented by disease progression despite orchidectomy or continuous treatment with a LHRH agonist.

To be included in the study, patients had to have European Cooperative Oncology Group (ECOG) Performance Status ≤2 with an estimated life expectancy >3 months, and normal haematological and end-organ function (absolute neutrophil count >1 500/uL, platelets >100 000/uL, total bilirubin <1.5 × upper limit normal [ULN], AST [SGOT]/ALT [SGPT] <2 × ULN and creatinine <1.5 × ULN). The number of lines of previous therapy was not specified, and previous mitoxantrone and radiotherapy was allowed, provided treatment had been completed >4 weeks before enrolment in the study. Pain control had to be optimized (>1 week) on a stable dose of narcotic analgaesics before enrolment.


All patients had a baseline evaluation consisting of history and physical examination, blood evaluation and CT imaging of chest/abdomen/pelvis and bone scan. Patients also completed the present pain intensity (PPI) scale from the McGill-Melzack questionnaire and the Functional Assessment of Cancer Therapy–Prostate (FACT-P) questionnaire to assess pain and quality of life.

All patients were admitted to the hospital for treatment; they received single agent cytarabine at 1.0 g/m2 twice daily for 2 days on a 21-day cycle, with protocol-specified reductions to 0.75, 0.5 and 0.25 g/m2 in case of grade 3/4 toxicities. After the third patient, the starting dose was reduced from 1g/m2 to 0.75 g/m2 because of grade 3–4 haematological toxicity seen in the first three patients. Patients were encouraged to use oral indomethacin (25 mg thrice daily) from day –1 to +3 and prednisolone eye drops (1%) for 5–7 days after cytarabine treatment as prophylaxis against cytarabine inflammatory syndrome. Treatment with colony-stimulating factors was allowed at the discretion of the treating physician.

The response evaluation included a physical examination and serum PSA assessment with each 21-day cycle. Imaging (CT and bone scans) was repeated after cycles 3 and 6. Serum haematology and biochemistry were monitored weekly for cycles 1–3 then every 3 weeks. Pain and quality-of-life assessments were undertaken at each cycle and at regular intervals after completion of therapy.

The primary endpoint was serum PSA response based on PCWG2 Criteria [12], with secondary endpoints of response of measurable disease (using Response Evaluation Criteria in Solid Tumours [RECIST]) [13], pain and quality-of-life indices, PSA progression-free survival (PFS), and safety (incidence of all adverse events (AEs). PSA response was defined as at least a 50% decline in PSA level confirmed by a second measurement >3 (not 4) weeks later. Early increases in PSA level were used to indicate PSA progression only if they continued to increase beyond 12 weeks [14].

Pain response was defined as at least a 2-point reduction in PPI on the McGill-Melzack scale or a 50% reduction in analgesic use from baseline. Quality-of-life response required a 16-point (i.e. 10%) improvement in FACT-P from baseline.

Initially, a two-stage Phase 2 Fleming design was to be used and a maximum of 30 patients were to be recruited. This regimen was assumed to be inactive if the PSA response rate was at most 5% and potentially active if it was at least 20%. After initial experience with high treatment-associated toxicity and slow enrolment, a modified Simon's two-stage design was used with the same P0 and P1 criteria with an initial accrual of 10 patients planned for stage 1, and continuation only if a PSA response was observed in one or more patients.


Fluorescence in-situ hybridization (FISH) to evaluate TMPRSS2-ERG and immunohistochemistry (IHC) to evaluate SPINK1 were carried out as described previously [15–18]. The criteria for TMPRSS2/ERG rearrangement were visualization of separate green 5′ ERG and red 3′ ERG signals, and an intact blue (telomeric 5′ ERG) signal. Generally accepted criteria for ETS translocations were used (ratio of 5′ ERG signal over 3′ ERG <0.80) where break-apart FISH indicates the presence of a fusion genomic rearrangement.

Two whole-blood samples per patient were collected in PaxGene tubes to stabilize intracellular mRNA. Samples were collected at baseline, on Cycle 1, Day 1, after each cycle and within 2 weeks of study termination. Analysis for gene translocations was conducted by Gen-Probe Inc. (San Diego, CA, USA) using transcript-mediated amplification technology [19,20]. This assay uses target capture to eliminate interfering nucleic acids, followed by transcription-mediated amplification of target sequences, which are then detected by chemiluminescent DNA probes. The amplification primers used were specific for the TMPRSS2 exon 1: ERG exon 4 (T1:E4) translocation [21]. In order to determine the relevant level of SPINK1 positivity (as unlike TMPRSS2-ERG it is constitutively expressed), we collated all the serum SPINK1 levels of the men who did not have evidence of SPINK1-positive tumours. We determined the mean copies/mL (147) and sd (134.69), and therefore used a threshold of positivity of mean (2) sd (416) in order to define significant levels of serum SPINK1 mRNA. If the mean number of copies/mL was below the limit of detection of the assay, the result was considered to be negative.

The primary molecular endpoint was of the association between PSA response and the presence of the ERG fusion gene as detected by FISH on tumour samples.


Ten men were recruited to the study from June 2007 to January 2010. Their median (range) age was 66 (53–74) years, eight of the 10 men had performance status ECOG 0/1 and their median (range) Gleason score was 8 (6–9). Their median (range) serum PSA concentration at baseline was 275 (61–4940) ng/mL. All patients had previously undergone radiotherapy; three had previously received two regimens of systemic therapy (excluding hormonal manipulations) and one patient was chemotherapy naive. A total of 26 cycles of cytarabine were administered, with a median (range) of 3 per patient (1–4).


Nine patients experienced grade 3–4 haematological toxicity: leuko/neutropenia (nine patients each), thrombocytopenia (three patients), anaemia (three patients). Haematological toxicity led to dose reduction of cytarabine. Five men stopped treatment because of toxicity. The toxicity results are shown in Table 1.

Table 1.  All grade and serious AEs (Grade 3–4)
ToxicityPatients affected (N= 10), n
Any gradeGrade 3–4
 Febrile neutropenia22
 Wound infection11
Laboratory aberrations  
 Increased AST40
 Increased ALP21
 Decreased HCO320
 Increased INR11
 Decreased PO411


Six men had progressive disease, while four were inevaluable for response. This result met the pre-specified endpoint for limited efficacy and the study was terminated. The median PFS (composite for PSA/RECIST/bone scan) was 2.5 months (95% CI: 0.4 – not reached [Fig. 1]). The median (range) overall survival time was 9.2 (4.4–16.1) months (Fig. 2). No patient experienced a pain response but two patients experienced improvement in quality-of-life, as assessed by FACT-P.

Figure 1.

PFS (PSA/tumour/bone progression). The median PFS was 2.5 months (95% CI: 1.1–8.0). The 6-month PFS rate was 30% (95% CI: 7–58%). N = 10.

Figure 2.

Overall survival. Eight deaths were observed for this analysis and others were censored at last follow-up. The median overall survival time was 9.2 months (95% CI: 4.4–16.1). The 6-month overall survival rate was estimated to be 78% (95% CI: 36–94%).


Using three-colour FISH, we found that TMPRSS2-ERG (T1:E4) translocation was present in archival tumour samples from two patients (Table 2). Translocation-positive disease was felt to be the dominant lesion in only one of these two patients, on the basis of size (patient #5). The other positive case had both one small positive TMPR2SS2-ERG focus and one positive SPINK1 focus, and no detectable TMPRSS2-ERG mRNA in the blood. An additional case was indeterminate on FISH analysis, and negative on blood mRNA analysis and by SPINK1 IHC. Taken together, blood mRNA analysis for the T1:E4 translocation was predictive of tumour results (presence or absence) in 8/9 evaluable cases (one was concordantly positive, seven were concordantly negative; Fisher's exact test, two-tailed P= 0.222 with nine evaluable cases; Table 2).

Table 2.  Results of fusion gene analysis by three-colour FISH in tumour and mRNA hybridization in plasma
Patient no.Baseline PSAGleason scoreSerum TMPRSS-ERGTumour TMPRSS-ERGSPINK IHCSerum SPINKMean SPINK, copies/mL
  1. *Probably complex, non-canonical rearrangement on basis of FISH result. †Below limit of detection of assay. ‡Mean copies/mL = 471.

 3859NegativePositive (1 focus)Positive (1 focus)Negative105
104378NegativeNegativePositivePositive17 718

SPINK1 analyses revealed three positive tissue samples, of which one also had a TMPRSS2-ERG focus. The two uniquely positive samples had >416 copies/mL (our predefined benchmark), but the mixed sample did not. Seven samples that were negative according to IHC were also negative in blood. In total, there was a concordance of 9/10 cases with serum SPINK1. (Fisher's exact test, two-tailed P= 0.06; Table 2).

If both analyses are considered together in the nine evaluable patients with both tissue markers available, then statistical significance is achieved for the process of serum-tissue correlation, albeit limited by the small numbers involved (Fisher's exact test, two-tailed P= 0.047).

There was no relationship between serum PSA and mRNA copies of PSA or PCA3 (copies/mL) in whole blood (data not shown). In the one patient with TMPRSS2-ERG quantification available (five values over the course of the trial), there was a high correlation between PSA mRNA copies/mL and TMPRRS2-ERG mRNA copies/mL (r2= 0.911, two-tailed P= 0.03), and a similar correlation between serum PSA concentration and TMPRSS2-ERG copies/mL (r2= 0.84, two-tailed P= 0.076), probably owing to the common androgen dependency.


Based on our cohort of 10 patients with CRPC, cytarabine at the dose used has no significant anti-tumour activity and even at low doses the drug is tolerated poorly in this population. Exploratory analyses aimed at identifying the TRMPRSS2-ERG fusion gene and SPINK1 positivity in tumours by FISH and IHC, respectively, concorded well with the transcription-mediated amplification assays of whole blood.

There are several potential explanations for the demonstrated lack of benefit of cytarabine, despite encouraging preclinical data. Severe haematological toxicity led to dose reductions, despite modifying the protocol to create a reduced starting dose (0.75 mg/m2) after the first two patients. Even for Ewing's sarcoma, where a higher dose could be delivered, there was a similar lack of efficacy and associated toxicities [22].

In addition, our intermittent scheduling may not be effective in maintaining the cytarabine tumour levels necessary to down-regulate ERG or FLI1 expression as originally hypothesized. A novel EWS-FLI inhibitor that has been shown to be effective in preclinical Ewing's sarcoma models has also been evaluated in in vitro models of human prostate cancer [23]. These studies showed an inhibition of transcriptional activity of ETS1 and ERG with a concurrent reduction in invasive and metastatic potential, thus showing a biological relevance to ETS family inhibition in prostate cancer.

It is also unclear whether translocations found in primary tumours are relevant to subsequent disease progression. Inhibition of ERG-related pathways may not be efficacious if alternate pathways are integral to driving tumour progression and growth in late-stage disease. Indeed, whilst ERG overexpression is believed to contribute to invasive/metastatic behaviour [23], after the present trial was begun, pre-clinical data showed that cytarabine was not effective as an inhibitor of ERG-mediated invasion [6]. Recent correlative clinical data has also suggested that there is little prognostic importance of the TMPRSS2-ERG translocations in prostate cancer [24]. On the other hand, SPINK1 overexpression has, albeit in limited studies, been linked to more aggressive disease [18].

An important finding from this study is the ability to correlate the status of the tumours in the prostate with circulating mRNA from the patient. Putative sources include circulating tumour cells or circulating RNA from necrotic tumour debris. The relationship of either of these to tumour dynamics and prognosis requires further study [25] as there is currently little clinical information available on this topic [26]. Initially these assays were designed to track quantitative changes in the amount of relevant transcripts, assuming that cytarabine was effective. They might provide useful biomarkers for evaluating more effective treatments.

Other reports have documented the presence of the TMPRSS2-ERG translocation in circulating tumour cells, involving costly immune-magnetic capture systems, or proprietary equipment and reagents [27–30]. The present report suggests that transcription-mediated amplification technology to assess for prostate cancer-related genes in the blood might provide a less expensive alternative strategy. We are aware of only one similar study that applied FISH on blood samples to detect the presence of the TMPRSS-ERG fusion gene in circulating tumour cells in almost 50% patients examined [26]. The assay used in the present study was specific to the T1-E4 gene fusion and other variants would not have been detected by this method.

In conclusion, cytarabine at the dose used shows no activity but considerable haematological toxicity in docetaxel-refractory CRPC. The evaluation of serum TMPRRS2-ERG/SPINK1 is worthy of further study and these have potential as novel biomarkers.


We wish to thank Prostate Cancer Canada for funding and Hospira Inc., for support in-kind.


John R. Day and Jack Groskopf are Employees of Gen-Probe Inc.