Main

Genomic rearrangements that create cell-type-specific fusion oncogenes are hallmarks of cancer1,2. The crucial step for development of such genomic rearrangements is the generation of DNA DSBs and subsequent illegitimate repair of these breaks3,4. However, the factors that are involved in DSB genesis and the mechanisms that produce cell-type-specific genomic rearrangements are not well understood for most malignancies. In human prostate cancer, a number of recurrent rearrangements involving various androgen receptor target genes fused to oncogenic transcription factors of the ETS family have been described5. The most common of these rearrangements, occurring in nearly 50% of prostate cancers, involves fusion of TMPRSS2, an androgen regulated gene, with ERG, an ETS transcription factor gene, resulting in activation of the TMPRSS2-ERG fusion oncogene through subversion of androgen signaling5,6. Given the high prevalence of prostate cancer in the population7 and the high fraction of prostate cancers harboring this fusion oncogene5, the TMPRSS2-ERG gene fusion is the most common fusion gene observed in any human cancer. However, the mechanisms involved in generating the DSB that leads to these TMPRSS2-ERG rearrangements are not completely understood.

Dysfunction of topoisomerase II (TOP2), an enzyme that catalyzes a transient DSB to resolve DNA topological constraints, can produce DSBs that have been implicated in the pathogenesis of rearrangements in treatment-related acute myeloid leukemia (t-AML) and childhood leukemias8,9,10,11. Estrogen receptor signaling has been reported to recruit the TOP2 isoform TOP2B to regulatory regions of estrogen receptor target genes, leading to TOP2B-mediated DSBs12. We hypothesized that androgen receptor signaling might lead to TOP2B-mediated DSBs and that such breaks could be involved in the generation of recurrent TMPRSS2-ERG rearrangements in prostate cancer.

We show that androgen signaling leads to co-recruitment of androgen receptor and TOP2B and to TOP2B-mediated DSBs at regulatory regions of androgen receptor target genes, as well as at regions of the TMPRSS2 and ERG loci that were highly aligned with genomic breakpoints of TMPRSS2-ERG fusion genes in prostate cancer. These androgen-induced TOP2B-mediated DSBs were highly recombinogenic and could lead to de novo production of TMPRSS2-ERG fusion transcripts in a TOP2B-dependent manner. In human prostate tissue, androgen receptor and TOP2B, which were rarely both present in normal TMPRSS2-ERG fusion-negative prostate cells at high levels, were both robustly coexpressed in TMPRSS2-ERG fusion-positive neoplastic cells comprising prostatic intraepithelial neoplasia (PIN) lesions, known precursors to prostate cancers. These findings implicate androgen receptor signaling and TOP2B action in the generation of TMPRSS2-ERG rearrangements. The occurrence of TOP2B-mediated DSBs during regulated transcription represents a new paradigm for the development of DSBs involved in generating gene rearrangements in cancer.

Results

TOP2B is required for androgen-mediated gene expression

We used the LNCaP and LAPC4 prostate cancer cell lines, which have intact androgen signaling (Supplementary Fig. 1) but do not harbor TMPRSS2-ERG rearrangements, to investigate whether TOP2B is required for efficient activation of androgen receptor target genes in androgen-deprived cells stimulated with the prototypical androgen dihydrotestosterone (DHT). By genome-wide gene expression microarray analysis, we found that pharmacological inhibition (with etoposide or merbarone) or depletion of TOP2B by RNA interference (RNAi) (sh-TOP2B) coordinately attenuated the upregulation of DHT-induced genes. This finding was confirmed by quantitative RT-PCR of candidate genes in both LNCaP and LAPC4 cells (Fig. 1 and Supplementary Figs. 1–3). Gene set enrichment analysis13,14 revealed that androgen-induced transcriptional programs were enriched for downregulation by TOP2 inhibition (P < 10−14) or sh-TOP2B (P < 10−2) compared to all other genes (Fig. 1a,b and Supplementary Figs. 2 and 3). These analyses show that both pharmacological and genetic modulations of TOP2B lead to concordant downregulation of androgen receptor-regulated genes.

Figure 1: TOP2B is required for efficient induction of androgen receptor target gene expression after androgen stimulation.
figure 1

(a) Inhibition of TOP2B with etoposide (ET) or merbarone (Mer) before DHT stimulation of LNCaP cells attenuated expression of an androgen-induced geneset (P < 10−14), defined as the 45 genes upregulated >2-fold in DHT- compared with control-treated LNCaP cells. (b) Geneset enrichment analysis shows highly statistically significant enrichment for downregulation of three independently derived but overlapping androgen-responsive genesets in ET/DHT vs. DHT, Mer/DHT vs. DHT, and sh-TOP2B/DHT vs. DHT treatment groups (see also Supplementary Figs. 2 and 3). y axis shows the degree of confidence for geneset enrichment as −log10(P-value). Significance threshold corresponding to P = 0.05 is indicated. (c,d) Real-time RT-PCR analysis of relative expression of selected genes with respect to GAPDH expression confirms downregulation of representative androgen-responsive genes in ET/DHT, Mer/DHT and sh-TOP2B/DHT-treated cells. Data are shown as mean ± s.e.m. of two to three replicates.

Full size image

TOP2B catalytic activity at androgen-receptor target genes

We next investigated whether TOP2B was directly involved in DHT-induction of androgen receptor target genes. Chromatin immunoprecipitation (ChIP) experiments revealed coordinate recruitment of both androgen receptor and TOP2B to the enhancer and promoter regions of KLK3 (which encodes the prostate specific antigen, PSA) and TMPRSS2 in DHT-treated cells. This coordinate recruitment at regulatory regions seemed to be fairly specific as our ChIP assay did not show recruitment to the 'middle' region of KLK3 or to regulatory regions of other arbitrarily selected genes (Fig. 2a,b and Supplementary Fig. 4). Androgen receptor and TOP2B are recruited to these regulatory regions as part of a complex, as stimulation with DHT resulted in coimmunoprecipitation of TOP2B and androgen receptor (Fig. 2c) and co-recruitment of TOP2B and androgen receptor to androgen-responsive elements (AREs) of KLK3 and TMPRSS2 in ChIP-re-ChIP experiments (Fig. 2d). This androgen receptor-TOP2B complex formation did not depend on the presence of DNA, as pre-treatment of lysates with DNAseI or excesses of ethidium bromide did not significantly disrupt complex formation (Supplementary Fig. 4). To ascertain whether TOP2B was catalytically cleaving genomic DNA at regulatory regions of KLK3 and TMPRSS2, we conducted a potassium SDS (KSDS) assay15, which allows etoposide-stabilized TOP2 catalytic cleavage complexes to be isolated. This showed robust TOP2 cleavage at the KLK3 enhancer and promoter and the TMPRSS2 enhancer but not at regions that did not contain androgen receptor or TOP2B binding sites (Fig. 2e and Supplementary Fig. 5). This activity was abolished by short interfering (si)-TOP2B treatment before DHT stimulation (Fig. 2f and Supplementary Fig. 6), indicating that the catalytic activity was mediated by TOP2B.

Figure 2: Androgen stimulation induced recruitment of androgen receptor–TOP2B and TOP2B catalytic cleavage at known regulatory regions of androgen receptor target genes.
figure 2

(a) Schematic of AREs in the promoter and enhancer regions of KLK3 and TMPRSS2. Bars indicate positions (relative to transcriptional start site) of amplicons analyzed. (b) ChIP analysis of androgen-deprived LAPC4 cells that were stimulated with 100 nM DHT for varying time points. DHT induced recruitment of androgen receptor and TOP2B to the enhancers and promoters of KLK3 and TMPRSS2 but not to an intervening region (middle) between the enhancer and promoter of KLK3. (c) Stimulation of androgen-depleted LAPC4 cells with 100 nM DHT for 6 h allows coimmunoprecipitation of TOP2B with androgen receptor antibodies and vice versa. IP, immunoprecipitating antibody; IB, immunoblotting antibody; TCL, total cell lysate. (d) ChIP-re-ChIP experiments show that stimulation of androgen-deprived LAPC4 cells with 100 nM DHT induces co-recruitment of androgen receptor and TOP2B to AREs of KLK3 and TMPRSS2. The first round of ChIP was performed using anti-TOP2B antibodies and the resulting immunoprecipitates were subjected to a second round of ChIP with anti-androgen receptor antibodies. Relative enrichment was determined by quantitative (q)PCR and is shown as percentage of input DNA. (e,f) KSDS method shows TOP2 catalytic cleavage of androgen receptor target sites in LAPC4 cells in a DHT- and TOP2B-dependent fashion. Results are presented as mean percentage of input with ± s.e.m. of two to three replicates.

Full size image

TOP2B catalytic cleavage near TMPRSS2-ERG breakpoints

As TOP2 cleavage can produce fragile sites for development of DSBs leading to genomic rearrangements in other malignancies8,9,10,11, we hypothesized that androgen-mediated TOP2 catalytic cleavage at key sites within TMPRSS2 and/or ERG might create DSBs and TMPRSS2-ERG genomic rearrangements in prostate cancer. Examination of previously published datasets revealed that genomic breakpoints in TMPRSS2 and ERG occur most frequently at the ∼25-kb region between exon 1 and exon 4 of TMPRSS2 and at a ∼75-kb stretch in intron 3 of ERG16,17,18 (Supplementary Table 1). We identified breakpoints at high resolution in eight cases described in these previous reports, and we sequenced the TMPRSS2 and ERG genomic breakpoints to nucleotide resolution in three of these cases (Fig. 3a, Supplementary Fig. 7 and Supplementary Table 1). We used the KSDS assay and tiled 71 real-time PCR amplicons across the ∼100 kb surrounding the TMPRSS2 and ERG genomic segments that showed hotspots for breakpoints, and found regions with high levels of DHT-induced TOP2 cleavage that were significantly enriched for proximity to TMPRSS2 (P = 0.010) and ERG (P = 0.013) breakpoints (Fig. 3a).

Figure 3: Androgen stimulation induces androgen receptor–TOP2B recruitment and TOP2B catalytic cleavage at genomic breakpoints of TMPRSS2 and ERG observed in human prostate cancer.
figure 3

(a) Sites closest to TMPRSS2-ERG breakpoints from eight prostate cancer cases determined from various studies (arrows) were significantly enriched for high KSDS enrichment of TOP2 catalytic cleavage (P = 0.010 and 0.013 for TMPRSS2 and ERG breakpoints, respectively) in LAPC4 cells16,17. Labeled sites (for example, T8, T23, E5, E13, etc.) are analyzed in subsequent experiments. (b) DHT- and TOP2B-dependent TOP2 catalytic cleavage in LAPC4 cells around the case 24 breakpoint aligning with region T8 (upper panel). The lack of KSDS enrichment at region E47 at ERG was re-confirmed in an independent KSDS experiment (lower panel). (c) SLOT assay showed that DHT-induced TOP2 catalytic activity in LAPC4 cells was significantly higher at an NspI fragment most proximal to the TMPRSS2 breakpoint observed in case 24 than to the adjacent, more distal NspI fragment (see Supplementary Figs. 9 and 10 for overview of SLOT). (d) ChIP enrichment of androgen receptor (AR) and TOP2B at representative TOP2 catalytic cleavage sites in DHT-stimulated LAPC4 cells relative to untreated controls. (e,f) 3C analysis reveals DHT-dependent spatial chromatin interaction of TMPRSS2 enhancer and promoter with region T8 (see first lane vs. fourth lane). Omission of NspI restriction enzyme and/or DNA ligase served as assay negative controls. Inhibition with Mer prevented these DHT-induced interactions. Error bars indicate ± s.e.m. of two to three experiments.

Full size image

When we examined the breakpoints of two of the three subjects (cases 24 and 30) for whom we sequenced rearrangement junctions, we found that the TMPRSS2 and ERG breakpoints seen in these cases were aligned with DHT-induced, TOP2B-mediated cleavage sites in LAPC4 cells (Fig. 3b and Supplementary Fig. 8). To verify this association, we developed a novel assay for selective labeling of TOP2 cleavage sites (SLOT), which allows specific labeling of the etoposide-stabilized TOP2-DNA cleavage complex (Supplementary Figs. 9 and 10). Using SLOT, we observed strong enrichment for DHT-dependent TOP2 cleavage at the NspI restriction fragment closest to the precise TMPRSS2-ERG genomic breakpoint observed in cases 24 and 30 (Fig. 3c and Supplementary Fig. 8).

Androgen receptor binds specifically to thousands of genomic sites, even if those sites lack canonical AREs19. As the sites that showed high DHT-induced TOP2B cleavage within the TMPRSS2 and ERG genes did not contain known AREs, we carried out ChIP experiments to understand whether TOP2 cleavage at these sites was a consequence of such non-canonical DHT-induced androgen receptor recruitment. As expected, all of the sites that showed high TOP2 cleavage (for example, T8, T6, E5, E13, E28; Fig. 3a) showed DHT-induced co-recruitment of both TOP2B and androgen receptor (1.5–15-fold over control; Fig. 3d), suggesting direct androgen-induced androgen receptor-coupled TOP2B recruitment to these sites, whereas sites that did not show strong TOP2B cleavage (for example, T23, E47, E35) showed weak or absent recruitment of both TOP2B and androgen receptor (Supplementary Fig. 11).

In further support of the idea that these intragenic androgen receptor-TOP2B target sites have a DHT-induced regulatory role, a representative site, T8, showed a specific DHT-induced chromatin interaction with both the enhancer and promoter of TMPRSS2 in chromosome conformation capture (3C) experiments (Fig. 3e,f and Supplementary Fig. 12)20. Consistent with the hypothesis that recruitment and activity of TOP2B at these sites is required for assembly of these enhancer- and promoter-T8 chromatin complexes, treatment of androgen-deprived LAPC4 cells with merbarone before stimulation with DHT abolished the formation of these complexes. We speculate that TOP2B activity allows resolution of topological entanglements created during the formation of these transcription-associated chromatin loops.

TOP2B-mediated DSBs are recognized by the repair machinery

We next tested whether sites that showed DHT-induced androgen receptor–TOP2B recruitment and TOP2B catalytic cleavage could be fragile sites for persistent DSB formation. We found that stimulation with DHT could induce widespread formation of γH2A.X foci (Fig. 4a), which form around DSBs21. We also observed pronounced ATM recruitment, a central feature of DSB repair pathways22,23, to sites showing strong androgen receptor–TOP2B recruitment and TOP2B catalytic activity (KLK3 enhancer and promoter, regions T6, T8, T20, E5, E13 and E28), but much less so to a region in ERG (E35) that did not show significant TOP2 catalytic activity (Fig. 4b and Supplementary Fig. 13).

Figure 4: Androgen stimulation results in TOP2B-dependent DSB formation.
figure 4

(a) Androgen stimulation induced formation of γH2A.X foci in LAPC4 cells. (b) Androgen stimulation led to recruitment of ATM to representative sites of high TOP2B activity but much less so to a region showing low TOP2B catalytic cleavage (E35). (c,d) Stimulation of LAPC4 cells with DHT induced DSBs that could be end-labeled with biotin at representative TOP2 cleavage sites, but not at regions showing low TOP2 cleavage (KLK3 middle, E35). (e) DSBs largely resolve by 24 h after DHT stimulation of LAPC4 cells as shown by reduced biotin labeling and ATM recruitment. The time course of DSB formation/resolution and ATM recruitment are highly parallel. (f) DHT-induced DSBs at a representative TOP2B enriched site (T8) depend on TOP2B. A site showing low TOP2B enrichment (T23) served as a negative control. (g) DHT-induced strand breaks at representative TMPRSS2 sites are enriched for TOP2B and ATM binding as seen by ChIP-re-ChIP experiments. Results are presented as mean percentage of input with ± s.e.m. of two to three replicates. (h) DHT induces chromosomal breaks at TMPRSS2 in LAPC4 cells, monitored as fraction of cells (n > 200) showing split-apart of FISH probes 5′ (green) and 3′ (red) of TMPRSS2. (i) Treatment of LAPC4 cells with ET ± DHT increased split-aparts at ERG compared to control or DHT treatment. Error bars indicate 95% confidence intervals.

Full size image

To detect DSBs at relevant sites directly, we labeled these breaks with biotin-conjugated nucleotides in situ using terminal DNA transferase12 (Supplementary Fig. 13). In parallel with ATM recruitment, DHT induced DSBs at sites of TOP2B recruitment/catalytic activity and androgen receptor binding within the TMPRSS2 and ERG genes (Fig. 4c,d). Conversely, regions in KLK3 and ERG that did not show significant TOP2B catalytic activity (middle, E35) did not show significant DSB formation (Fig. 4d). DSBs and ATM recruitment appeared within the first 6 h after DHT stimulation and largely resolved after 24 h (Fig. 4e and Supplementary Fig. 13). ChIP-re-ChIP experiments showed recruitment of ATM to the biotin-labeled strand breaks, indicating that these breaks were truly DSBs and not just single-strand nicks (Fig. 4f,g). A ligation-mediated PCR (LM-PCR) method allowed single nucleotide mapping of breaks that occurred at a representative region with high DHT-induced TOP2B enrichment/catalytic cleavage, ATM recruitment and close alignment to a TMPRSS2 genomic breakpoint in a clinical prostate cancer case (region T6; Supplementary Fig. 14). This provided further direct evidence of DSBs at these sites. TOP2B was also present at these biotin-labeled DSBs in ChIP-re-ChIP experiments, pointing to its involvement in generating these breaks (Fig. 4g and Supplementary Fig. 13). In support of this, RNAi-mediated depletion of TOP2B or androgen receptor (sh-AR) significantly reduced formation of biotin-labeled strand breaks (Fig. 4f and Supplementary Fig. 13).

To investigate whether these DSBs could result in genomic breaks and/or chromosomal rearrangements at an individual cell level, we performed dual-color fluorescence in situ hybridization (FISH) assays with a pair of probes flanking the TMPRSS2 or ERG genes (Supplementary Fig. 15). A significant separation, or 'split-apart', of these probes would indicate the presence of a genomic break or rearrangement in the intervening genomic sequence. As a positive control, we treated LAPC4 cells with etoposide, a known TOP2-dependent genomic-break-inducing agent. This resulted in generation of split-aparts at the TMPRSS2 and ERG loci in ∼16% and ∼8% of cells, respectively (Fig. 4h,i). Stimulation of LAPC4 and LNCaP cells with DHT alone resulted in a significant increase in genomic breaks in TMPRSS2 compared to unstimulated cells (Fig. 4h and Supplementary Fig. 16). Similarly, stimulation with DHT followed by treatment with etoposide showed increased split-aparts at both the TMPRSS2 and ERG loci compared to control LNCaP and LAPC4 cells or cells treated with DHT alone or etoposide alone (Fig. 4h,i and Supplementary Fig. 16). This DHT-induced increase in split-aparts was significantly attenuated by si-TOP2B (Supplementary Fig. 16). These data suggest that DHT stimulation could cooperate with TOP2B action to induce genomic breaks and/or chromosomal rearrangements at these loci.

TOP2B-mediated double strand breaks are recombinogenic

We devised an androgen-induced genomic recombination assay to assess whether these androgen-induced TOP2B-mediated DSBs were competent recombination foci, and whether these breaks could significantly increase genomic recombination events at these sites. For this assay, we cloned a region showing high DHT-induced TOP2B catalytic activity (T8 in intron 1 of TMPRSS2) and one showing low TOP2B catalytic activity (TMPRSS2 exon 6; Fig. 5a) into a plasmid vector containing a blasticidin-resistance gene to create the pcDNA6.2-IN-1 (IN-1) and pcDNA6.2-EX-6 (EX-6) constructs, respectively. Transfection of these constructs into LAPC4 cells followed by stimulation with DHT led to increased formation of biotin-labeled DSBs in the IN-1 insert, but not the EX-6 insert (Fig. 5b). To assess whether these DHT-induced breaks increased recombination efficiency, we transfected the IN-1 or EX-6 constructs along with control or sh-TOP2B constructs into LAPC4 cells, depleted the media of androgens, stimulated the cells with DHT for 24 h to induce DSB formation, split the cells to clonogenic density, and subjected them to blasticidin selection in androgen-containing media for 2 weeks. The number of colonies formed after 14 days would indicate the number of successful recombination events that allowed incorporation of blasticidin-resistance vectors into genomic DNA (Fig. 5c). Consistent with our hypothesis, the IN-1 vector produced significantly more colonies than the EX-6 construct, and this increased frequency was abolished by sh-TOP2B (Fig. 5d,e). Transfection of LAPC4 cells with sh-TOP2B alone without blasticidin selection did not decrease clonogenic survival compared to control transfected cells (Supplementary Fig. 17). Furthermore, both plasmids were equally competent in producing recombination events if they were linearized with a restriction enzyme before transfection (Supplementary Fig. 18). In addition, a ligation-mediated PCR strategy showed that the increased frequency of recombination in the IN-1 construct was due to recombination events occurring in the insert and not non-specifically at the vector backbone (Fig. 5f). These data provide evidence that DHT-induced TOP2B-mediated DSBs at target sites can promote genomic recombination of these target sites.

Figure 5: Androgen-induced TOP2B-mediated DSBs are recombinogenic and promote de novo production of TMPRSS2-ERG fusion genes.
figure 5

(a) Selection of TMPRSS2 regions showing high (intron 1-T8) and low (TMPRSS2-exon 6) KSDS enrichment in response to DHT stimulation of LAPC4 cells. (b) DHT-induced breaks can be detected in plasmids containing sequences surrounding region T8 of TMPRSS2 intron 1 (pcDNA6.2-IN-1) as shown by increased biotin labeling in DHT-stimulated LAPC4 cells transfected with this plasmid. Plasmids containing TMPRSS2 exon 6 (pcDNA6.2-EX-6) served as a negative control. (c) Schematic of androgen-induced genomic recombination assay in LAPC4 cells transfected with pcDNA6.2-IN-1 or pcDNA6.2-EX-6, which contain a blasticidin-resistance gene. Number of colonies represents the number of recombination events allowing integration of the blasticidin resistance vectors into the LAPC4 genome. (d) Representative results of genomic recombination assays. (e) pcDNA6.2-IN-1–transfected LAPC4 cells produced significantly more androgen-induced recombination events than pcDNA6.2-EX-6–transfected cells. Treatment with sh-TOP2B abolished this effect. (f) Although both showed similar recombination frequency at the vector backbone, recombination frequency in the IN-1 insert was significantly higher than in the EX-6 insert in pooled colonies as determined using the strategy shown on the right. Data are shown as mean ± s.e.m. of two to three replicates. (g) DHT stimulation of LAPC4 cells leads to increased TMPRSS2-ERG fusion transcripts compared to background levels in LAPC4 cells grown in androgen-containing (steady-state) or androgen-deprived (control) media. Pharmacological or genetic modulation of TOP2B (Mer, sh-TOP2B), PARP1 (3-AB, PJ-34, si-PARP1) or DNA-PKCS (Wort, si-DNA-PKCS) reduces TMPRSS2-ERG fusion transcripts without significantly altering GAPDH or TBP expression. (h) DHT stimulation leads to de novo formation of TMPRSS2-ERG fusion transcripts in LNCaP cells.

Full size image

We next assessed whether DHT-induced TOP2B-mediated DSBs can result in production of de novo TMPRSS2-ERG fusion genes in LNCaP and LAPC4 cells. We developed a highly sensitive RT-PCR assay that can detect fusion transcripts juxtaposing exon 1 of TMPRSS2 with exon 4 of ERG, at a frequency of ∼1 in 108 cells (Supplementary Fig. 19). As LAPC4 and LNCaP cells do not harbor TMPRSS2-ERG genomic rearrangements, the low level of TMPRSS2-ERG fusion transcripts seen when these cells were maintained in their fully supplemented, androgen-containing media (steady state) or when grown in androgen-depleted media (control) represents the background level for this sensitive assay. In four independent experiments involving two different cell lines, DHT stimulation of LAPC4 or LNCaP cells led to increased de novo production of TMPRSS2-ERG fusion transcripts (Fig. 5g,h and Supplementary Fig. 19). Pharmacological inhibition or RNAi-mediated depletion of TOP2B, or of the DSB repair pathway components PARP1 or DNA-PKCS, before stimulation of LAPC4 cells with DHT resulted in complete reduction of de novo TMPRSS2-ERG fusion transcript formation back to background levels (Fig. 5g).

TOP2B and androgen receptor are highly coexpressed in PIN

The TMPRSS2-ERG rearrangements are known to occur as early as the PIN prostate cancer precursor lesions24 and are probably selected for during the formation of invasive adenocarcinoma5,25. We carried out FISH and coimmunostaining of androgen receptor and TOP2B in PIN and matched normal prostatic tissue. In agreement with previous studies24,25, 10 of 23 subjects with PIN and none of the matched benign prostate tissues showed TMPRSS2-ERG rearrangement as determined by TMPRSS2 FISH (Fig. 6a,b and Supplementary Tables 2 and 3). Basal cells in the normal prostate glands showed much stronger staining for TOP2B than the luminal cells (Fig. 6c). By contrast, the luminal cells in normal glands showed much stronger staining for androgen receptor than the basal cells (Fig. 6c). However, TOP2B staining became more promiscuous in PIN, with luminal cells expressing nearly as much TOP2B as basal cells in these lesions. Androgen receptor continued to be highly expressed in PIN luminal cells, creating a more pronounced coexpression of both TOP2B and androgen receptor in these cells (Fig. 6d and Supplementary Table 3). This increased coexpression of TOP2B and androgen receptor might predispose PIN luminal cells to develop the androgen receptor- and TOP2B-mediated DSBs that are found in DHT-stimulated LNCaP and LAPC4 cells.

Figure 6: TMPRSS2-ERG rearrangements are observed in PIN prostate cancer precursor lesions and are associated with changes in TOP2B expression.
figure 6

(a) H&E stain and corresponding TMPRSS2 FISH in a representative PIN lesion. FISH image of red boxed area shows interphase nuclei of PIN lesion with split-aparts of 5′ (green arrow) and 3′ (red arrow) TMPRSS2 FISH probes. (b) H&E stain and corresponding TMPRSS2 FISH (green boxed area) of normal prostate epithelium. Orange arrows indicate normal configuration of 5′ and 3′ TMPRSS2 FISH probes. Scale bars, 10 μm. (c) Normal prostate epithelium shows strong expression of TOP2B in basal cells (red arrow) and low expression in luminal cells, whereas androgen receptor is predominantly expressed in luminal cells (green arrow). Lumen of prostate gland is marked with an asterisk. (d) PIN lesion shows high TOP2B in basal and luminal cells (purple arrow). Note the accentuated nucleolar localization of TOP2B (white arrow).

Full size image

Discussion

These data suggest a model in which intrinsic androgen signaling in neoplastic prostate cells induces TOP2B-mediated DSBs at many genomic loci, including breakpoints of TMPRSS2-ERG rearrangements in prostate cancer, even without extrinsic DSB-inducing genotoxic stress. These androgen-induced TOP2B-dependent DSBs can participate in genomic recombination events and can generate rare TMPRSS2-ERG gene rearrangements (Fig. 7). Another report has suggested that in the presence of high levels of exogenous genotoxic stress, other enzymes such as activation-induced cytidine deaminase (AID) and LINE-1 repeat-encoded ORF2 endonuclease can cooperate with androgen receptor signaling to establish DSBs and, ultimately, prostate cancer gene rearrangements26. Together, the data from the current study and this previous report indicate that multiple enzymatic activities, including that of TOP2B in the case of intrinsic androgen signaling (current study) and that of other nucleases in the case of exogenous genotoxic stress26, can cooperate with androgen signaling to generate recombinogenic DSBs and TMPRSS2-ERG rearrangements.

Figure 7: Proposed model for androgen-induced TOP2B-mediated double strand breaks and TOP2B instability (TIN) for the formation of TMPRSS2-ERG gene fusions.
figure 7

Androgen stimulation leads to co-recruitment of ligand-bound androgen receptor and TOP2B to regulatory regions of androgen-responsive genes as well as to regions of TMPRSS2 and ERG genes that participate in genomic rearrangements. TOP2B catalytic activity and the associated DSB formation are required for efficient activation of androgen-responsive genes, and these DSBs are usually resolved shortly after induction. In some circumstances, stabilization of these TOP2B-mediated DSBs can lead to illegitimate recombination and rare rearrangements between TMPRSS2 or other androgen-responsive genes and ERG that are then subject to selection during neoplastic outgrowth.

Full size image

Although the biochemical mechanisms by which TOP2B can mediate generation of frank DSBs are not well understood, a number of hypotheses can be drawn from previous studies. First, reactive oxygen species, such as those elaborated by androgen signaling in prostate cells in vitro27, or the inflammatory milieu often seen surrounding prostate preneoplastic lesions in vivo28, can stabilize TOP2-DNA cleavable complexes29. Dietary constituents, such as various bioflavonoids, can also stabilize TOP2-DNA cleavable complexes11. These stabilized cleavable complexes can be converted to overt DSBs by proteasomal degradation of TOP2B30 or by collision with replication or transcriptional machinery31. These hypotheses might help to explain the epidemiological evidence that links pro-oxidant stress, inflammation and dietary factors in the etiology of prostate cancer28,32. Indeed, the relatively high rate at which we observed genomic breaks in PIN cells and in androgen-stimulated LAPC4 and LNCaP cells might reflect somewhat dysregulated androgen signaling, which is known to promote both caricaturized differentiation and proliferation/survival pathways in these cells33,34. Owing to androgen-induced increases in both proliferation and transcription in these cells, such dysregulation of androgen signaling could cause increased rates of collision between transcription or replication forks and androgen-induced TOP2B cleavable complexes, with resultant DSB formation. By contrast, androgen signaling in normal cells induces terminal differentiation with suppression of proliferation, a state that might be largely resistant to the generation of genomic breaks and rearrangements35. Furthermore, as DNA damage-sensing pathways are likely to be functional in normal cells, even if rare breaks occur in such cells, they are likely to undergo senescence, cell death or appropriate repair36. However, if a rare genomic error leading to rearrangement of the TMPRSS2 and ERG loci does occur, then this event is likely to become subject to strong selection5,37,38,39. That these illegitimate repair events might be extremely rare and subject to selection helps to explain the observation that fusions of TMPRSS2 with ERG or other ETS family transcription factors are mutually exclusive5.

If our model of androgen-induced TOP2B instability (TIN) (Fig. 7) during regulated transcription holds, we would predict that several different androgen receptor target genes should participate in genomic rearrangements in prostate cancer. In support of this, recent studies have shown rearrangements involving different androgen-induced genes, including TMPRSS2, KLK2, CANT1, SLC45A3, AX747630 and HERV-K_22q11.23 (refs. 5,40). Several of these androgen-induced fusion targets were modulated by TOP2B inhibition or depletion (Fig.1 and Supplementary Figs. 2–4). These data suggest that androgen-induced TOP2B-mediated DSBs contribute to the prostate cancer specificity of TMPRSS2-ERG fusions. We speculate that androgen-induced TOP2B-mediated DSBs might cooperate with androgen-induced proximity of TMPRSS2 and ERG genes, as suggested previously26,41, to confer this lineage specificity.

In addition, given that other transcriptional processes, including ER-, RAR-, and AP1-mediated transcriptional programs, also seem to involve TOP2B activity and associated DSBs12, the TIN model might be generalizable to lineage-specific genomic rearrangements in other malignancies. Finally, we speculate that the TIN model might underlie some of the genomic instability and copy number alterations observed during cancer progression42.

Methods

Cell culture, reagents and antibodies.

LAPC4 cells were grown in ISCOVE's Modified Dulbecco's Medium (Invitrogen) containing 10% FBS and 1 nM R1881 (PerkinElmer). LNCaP cells were propagated in RPMI medium supplemented with 10% FBS. To deplete cells of basal androgens, cells were washed with serum-free medium three times for 1 h and incubated in 5% charcoal-stripped FBS containing medium for 48 h before experiments. Dihydrotestosterone (DHT), merbarone (Mer), etoposide (ET), PJ-34, 3-aminobenzamide (3-AB) and wortmannin (WORT) were purchased from Sigma-Aldrich. Antibodies were anti-androgen receptor (AR) rabbit polyclonal (N20), anti-TOP2β rabbit polyclonal (H286; Santa Cruz Biotechnology), anti-ATM (Ab-3; Calbiochem), anti-H2A.X p139 (clone JBW301; Millipore), anti-PARP1 mouse monoclonal (F2; Santa Cruz), anti-DNA-PKCS (Ab-2; Calbiochem), anti-Biotin (clone BN-34) and anti-β-Actin (clone AC-15; Sigma-Aldrich). Anti-TOP2B (IHC-00166) for immunohistochemistry was from Bethyl Laboratories. Rabbit Gamma Globulin (IgG) was from Jackson ImmunoResearch. Anti-AR mouse monoclonal antibody (N441) was kindly provided by D.P. Edwards43. AR (ON-TARGETplus SMARTpool L-003400), TOP2β (ON-TARGETplus SMARTpool L-004240) and control non-targeting pool siRNA oligonucleotides were from Dharmacon. siPARP-1 (cat. no. 1027400) was from Qiagen; siDNA-PKcs (sc-35200) was from Santa Cruz Biotechnology. shRNA constructs against TOP2B in pGFP-V-RS vectors (cat. no. TG308698) were from Origene.

Fusion transcript detection assay.

Total RNA was extracted from 108 cells using RNeasy Maxi Kits (Qiagen). First strand synthesis was carried out using Superscript III (Invitrogen) with primers specific to exons 5–11 of ERG (Supplementary Table 4). Reactions were cleaned up using PCR purification kit (Qiagen) and resulting cDNA libraries were split into 192 reactions and subjected to PCR amplification using primers and Taq-man probes specific to exon 1 of TMPRSS2 and exon 4 of ERG. The presence of an amplification product was monitored by real-time PCR. Products were also separated in an agarose gel and bands were sequence verified.

Chromatin immunoprecipitation (ChIP) and ChIP-re-ChIP.

Chromatin immunoprecipitation was performed as described44 with some modifications. Briefly, protein G Dynabeads (Invitrogen) were used to bind antibody-protein-DNA complexes. Protein-DNA complexes were eluted and crosslinks were reversed by heating at 65 °C for 12 h. Recovered DNA was purified using a PCR purification kit (Qiagen) and analyzed by real time PCR using iQ SYBR Green Supermix (Bio-Rad) on a Bio-Rad iCycler. For ChIP-Re-ChIP, after the first round of immunoprecipitataion, DNA protein complexes were eluted with 10 mM dithiothreitol in TE buffer for 45 min at 37 °C. The eluate was diluted 40 × in dilution buffer (1% Triton X-100, 2 mM EDTA, 150 mM NaCl, 20 mM Tris-HCl pH 8 and protease inhibitors) and subjected to another round of immunoprecipitation with either specific antibodies or control rabbit gamma globulin (Jackson ImmunoResearch Laboratories).

Potassium–SDS assay (KSDS).

KSDS assay was performed as described15, with modifications. Briefly, cells were cultured in charcoal-stripped FBS and treated with 100 μM ET, 100 nM DHT or control vehicle. Cells were then washed with ice cold PBS and collected by scraping in KSDS-lysis buffer (2% SDS, 10 mM EDTA and 100 μg/ml salmon sperm DNA in 10 mM Tris-HCl pH 8). Lysates were then subjected to sonication. A fraction of the samples was used as total input and the remainder was heated for 5 min at 65 °C. After heating, covalent complexes were precipitated by adding one volume of KCl buffer (120 mM KCl and 10 mM Tris-HCl pH 8). Samples were cooled on ice to allow precipitation of complexes and centrifuged for 5 min at 1,500 RPM. Pellets containing covalent complexes were washed with wash buffer (100 mM KCl, 1 mM EDTA, 20 μg/ml BSA, 100 μg/ml salmon sperm DNA and 10 mM Tris -HCl pH 8). Samples were then heated for 10 min at 65 °C and cooled to 4 °C for 5 min and centrifuged. This washing procedure was repeated three times. To elute DNA from the protein complexes, pellets were incubated in wash buffer supplemented with Proteinase K (Invitrogen) for 6 h at 65 °C. BSA was added to precipitate SDS and samples were centrifuged for 10 min at 12,000 RPM. Supernatants were purified using the Qiagen PCR purification Kit (Qiagen) and recovered DNA was analyzed by real time PCR.

Chromosome conformation capture (3C).

3C was basically performed as described45. Briefly, cells were incubated with 2% formaldehyde for 10 min at room temperature. Cell nuclei were extracted and resuspended in NEB2 restriction enzyme buffer (New England Biolabs). Intact nuclei were digested for 12 h at 37 °C by addition of 400 U of NspI (New England Biolabs). The digestion was stopped by addition of SDS and heating to 65 °C for 20 min. Samples were then diluted to a final volume of 8 ml with ligation buffer and 2000 U of T4 ligase (New England Biolabs) were added. DNA fragments were ligated for 5 h at 16 °C and for 30 min at room temperature. Crosslinks were reversed by incubating samples at 65 °C in the presence of proteinase K (Invitrogen). Finally RNase (Invitrogen) was added and DNA was extracted by phenol extraction followed by ethanol precipitation. Samples were further purified using the Qiagen PCR purification Kit and total DNA concentration quantified using Nanodrop. 100 ng of recovered DNA was used for each PCR reaction.

SLOT assay.

Cells were lysed in SLOT lysis buffer (2% SDS, 10 mM EDTA in 10 mM Tris-HCl pH 8) and DNA was extracted after proteinase K digestion using the Qiagen Blood and Tissue Kit. Tyrosine residues covalently linked to DNA were labeled using EDC and amine-PEG3-biotin (Thermo Scientific). Biotin-labeled DNA was precipitated using strepavidin-coated magnetic beads (NEB), digested with NspI (NEB), and unbound fragments were removed by washing with wash buffer (10 mM Tris HCl, 1mM EDTA, 2 M NaCl). Precipitated DNA was eluted in 95% formamide, 10 mM EDTA for 5 min at 65 °C, purified using the Qiagen PCR Purification Kit and analyzed by real-time PCR (see Supplementary Fig. 9 for an overview of the SLOT assay). To test the specificity of the SLOT assay, the pRYG plasmid, which contains a TOP2 cleavage and recognition site46, was incubated with recombinant TOP2A (TopoGen) for 30 min at 37 °C in the absence or presence of etoposide (Sigma). Reactions were stopped by addition of SDS and were digested by proteinase K (Invitrogen). DNA was extracted by phenol-chloroform and recovered DNAs were subjected to the SLOT assay.

DNA break labeling assay.

DNA strand breaks (DSBs) were labeled in intact nuclei with biotin-11-dUTP by terminal transferase. Biotin-labeled DNA fragments were then subjected to ChIP analysis using anti-biotin antibodies as described12.

Fluorescence in-situ hybridization (FISH).

BAC clones RP11-35C4 (TMPRSS2-5′), RP11-825A8 (TMPRSS2-3′), RP1195I2I (ERG-5′) and RP11-476D17 (ERG-3′) were purchased from BACPAC (http://bacpac.chori.org). Probes were labeled with SpectrumGreen (all 5′ probes) and SpectrumOrange (all 3′ probes) conjugated dUTP (Abbott). Slides were pretreated using Paraffin Pretreatment reagent kits (Abbott). Samples were analyzed under a 100× oil immersion objective using a Nikon E400 fluorescence microscope (Nikon). Samples were scored as positive for TMPRSS2 rearrangements if >10% of cells showed either split-apart or deletion FISH configurations.

Clonogenic survival assay.

PCR amplicons encompassing the breakpoint site of case 1 in intron 1 (IN-1) of TMPRSS2 and a site in exon 6 (EX-6) of TMPRSS2 were cloned into pcDNA6.2N-YFP TOPO vectors (Invitrogen). LAPC4 cells were deprived of androgens and transfected with pcDNA6.2 vectors together with sh-TOP2β or shRNA-control vectors using FuGENE6 (Roche). Twenty-four hours after transfection, cells were stimulated with 100 nM DHT and split the following day to clonogenic density in medium containing 5 μg/ml Blasticidin (Invitrogen). Cells were cultured for 14 d and medium was replaced every third day. To assess clonogenic survival, cultures were fixed in 25% methanol and stained with crystal violet as described47. Colonies containing at least 50 cells were counted using the Versa Doc Imaging system 3000 and the Quantity One software package (Biorad).

SNP-microarray analysis and cloning of genomic breakpoints.

Primers spanning the putative TMPRSS2-ERG deletion breakpoints identified in a previously published Affymetrix SNP 6.0 metastatic prostate cancer dataset16 were used to amplify breakpoints from the corresponding cases from that dataset. Products were cloned into pCR2.1TOPO vectors (Invitrogen) and analyzed by Sanger sequencing.

Prostate cancer and PIN tissue in tissue microarray (TMA).

TMAs from formalin-fixed paraffin-embedded tissue blocks containing PIN and matched normal tissue from 23 men were constructed as described48.

Expression array analysis.

RNA from inhibitor and knockdown experiments was processed, labeled and hybridized to HG-U133 2.0 Plus whole genome gene expression microarrays (Affymetrix) according to the manufacturer's protocols at the Johns Hopkins Microarray facility. The microarray gene expression data were pre-processed with RMA14,49. Statistical significance was assessed by Wilcoxon rank-sum P-values after ranking genes by absolute log fold change of expression between conditions. For genes represented by more than one probe set, the one with the highest fold change was chosen. All analysis was carried out using R 2.9 (R Development Core Team, 2009) and Bioconductor 2.4 (ref. 50).

Coimmunoprecipitation.

Coimmunoprecipitation was performed as described51. To test whether the interaction between TOP2B and AR depends on the presence of DNA, cell lysates were incubated with 50 μg/ml ethidium bromide (Sigma) or 200 U/ml DNase (NEB) for 12 h at 4 °C52. Lysates were then subjected to coimmunoprecipitation assays using AR-specific antibodies. Precipitated complexes were separated by SDS-PAGE and immunoreactive bands were detected by western blotting.

shRNA knockdown of TOP2B and cell sorting.

LNCaP cells were transfected with pGFP-V-RS shRNA-control or pGFP-V-RS shRNA-TOP2b vectors. Sixteen hours after DHT treatment, cells were harvested by trypsinization and resuspended in 2 mM EDTA in PBS. Fifty-thousand GFP positive singlet cells were sorted by FACS and RNA and protein was extracted using the AllPrep Kit from Qiagen.

Ligation-mediated linker PCR for identification of DHT-induced DNA strand breaks at nucleotide resolution.

DNA was extracted from intact cells by resuspending cells in lysis buffer (2% SDS, 10 mM EDTA, 10 mM TRIS-HCl ph 7.5) and incubating lysates for at least 6 h in the presence of 0.2 mg/ml Proteinase K (Invitrogen) at 37 °C. Proteinase K was inactivated by heating the lysate for 10 min at 70 °C and SDS was sequestered by adding Triton X-100 to a final concentraion of 1% and incubating for 1 h at 37 °C. RNase (Invitrogen) was added to a final concentration of 40 μg/ml and incubated for 1 h at 37 °C. DNA was then extracted in 1 volume of ultrapure phenol:chloroform:isoamyl alcohol (25:24:1, v/v; Invitrogen) and precipitated with ethanol. DNA ends were repaired using the End-It DNA End-Repair Kit (Epicenter Biotechnologies). DNA was recovered by phenol-chloroform extraction followed by ethanol precipitation. Double stranded Linkers (Linker 1; Linker 2) were prepared by annealing 20 pmol/μl of each oligonucleotide in 250 mM Tris-HCl (pH 7.8) by heating at 95 °C for 3 min, quickly transferring to 70 °C and gradually cooling to 4 °C over a period of 3 h. Linkers were then ligated to 1 μg of processed DNA using T4 ligase (NEB). Gene- and linker-specific primers used in amplification reactions are listed in Supplementary Table 4. Products of PCR amplification reactions were separated in acrylamide Gels (Invitrogen) and sequenced by Sanger sequencing.

Determination of integration frequencies.

To determine the integration frequency of the plasmid insert sequences of stable clones in the genome, DNA from pooled clones selected for blasticidin resistance was digested with NheI (New England Biolabs), which does not have cutting sites in the plasmid sequence. Blunt DNA ends were generated by using the End-It DNA End-Repair Kit (Epicenter Biotechnologies) following manufacturer's recommendations. A double-stranded linker (Linker 3; Linker 4) was annealed and ligated to DNA molecules as described above. After DNA clean-up using the DNeasy kit (Qiagen), DNA was analyzed by quantitative real-time PCR using primers. Relative integration frequencies were calculated for pcDNA6.2-IN-1 and pcDNA6.2-EX-6 according to the formula: 2^{(ctClone_DHT_insert − ctClone_DHT_blast) − (ctClone_control_insert − ctClone_control_blast)} where ctClone_DHT_insert is the ct value of clones that were androgen deprived and then DHT-treated before clonogenic survival using a forward primer flanking the plasmid (pcDNA6.2_insert_A) insert and reverse linker specific primers (Linker primer); ctClone DHT blast is the ct value of the same DNA using forward and reverse primers specific to the blasticidin resistance gene (pcDNA6.2_blast); and ctClone_control_insert/blast is the ct value of clones that were not androgen deprived before clonogenic survival using a forward primer flanking the plasmid (pcDNA6.2_insert_A) insert and reverse specific to the linker sequence (Linker primer) or the pcDNA6.2_blast primers.

Accession codes.

DNA microarray data reported in the manuscript have been deposited in the Gene Expression Omnibus (accession number GSE19445).