An IKKα–E2F1–BMI1 cascade activated by infiltrating B cells controls prostate regeneration and tumor recurrence

Abstract

Tissue injury triggers an inflammatory response that promotes clearance of dead cells and activates a regenerative wound healing process (Velnar et al. 2009). While the role of damage-associated molecular patterns (DAMPs) in the activation of injury-induced inflammation is well established (Zitvogel et al. 2010), the molecular mechanisms through which the consequent inflammatory response promotes regeneration remain nebulous. In addition to its role in tissue regeneration, the inflammatory response triggered by tissue injury is likely to play a key role in tumorigenesis, and several malignancies are induced as a consequence of chronic tissue damage (Kuraishy et al. 2011). Indeed, tumors were pointed out to be analogous to wounds that do not heal (Dvorak 1986). Furthermore, signaling pathways that promote tissue regeneration and stem cell renewal or expansion of transient amplifying cells, the Wnt and Hedgehog pathways, were also found to be key players in tumorigenesis (Beachy et al. 2004). However, it is unknown which regenerative pathways are activated in response to inflammatory signals generated by tissue injury. It is also not clear how tissue injury leads to stem cell activation or expansion of transient amplifying cells in both normal tissues and malignant tumors.

The NF-κB signaling pathway is involved in inflammation, tissue repair, and cancer (Ben-Neriah and Karin 2011). To study roles of NF-κB in cancer, we focused on the IκB kinase α (IKKα) and IKKβ catalytic subunits (Karin 2009). While both kinases can activate NF-κB-mediated transcription, IKKα also has NF-κB-independent functions in development (Hu et al. 2001) and tumorigenesis (Luo et al. 2007; Ammirante et al. 2010). The NF-κB-independent activities of IKKα are evident in prostate cancer (PCa), the most common nonskin cancer in men. PCa accounts for >32,000 deaths each year in the United States, making it the second leading cause of cancer deaths in men (Siegel et al. 2012). This is due to the fact that while androgen ablation or chemical castration is effective in the short term, the cancer will inevitably switch to a more aggressive and metastatic form, termed castration-resistant PCa (CRPC) (Gulley et al. 2003). The genesis of CRPC depends on the emergence of PCa stem/progenitor cells that either do not require androgen signaling for growth and survival or are highly sensitive to castrate levels of androgens (Maitland and Collins 2008). Using mouse models, we found that androgen ablation causes infiltration of the regressing tumors with lymphocytes that produce cytokines, which activate IKKα in surviving PCa cells, thereby accelerating CRPC emergence (Ammirante et al. 2010). Among the infiltrating cells, B cells are of particular importance, as they produce lymphotoxin (LT), a heterotrimeric member of the TNF family that activates IKKα (Ammirante et al. 2010). Ablation of B cells or inhibition of their recruitment into the regressing tumors prevents IKKα activation and delays CRPC regrowth. LT signaling leads to nuclear accumulation of IKKα and stimulates metastatic spread in a mouse model of PCa with neuroendocrine characteristics (Luo et al. 2007). However, the molecular mechanisms by which nuclear IKKα enhances survival and proliferation of cells that give rise to metastases and CRPC are not clear. Previous studies suggested that nuclear IKKα acts as a histone H3 kinase (Yamamoto et al. 2003), whereas other reports have described an interaction between IKKα and the histone acetylase (HAT) CBP/p300 that results in phosphorylation of the latter (Huang et al. 2007). IKKα was also reported to regulate the expression and activity of transcription factor E2F1 (Tu et al. 2006) and counteract SMRT repressor activity leading to acetylation of p65/RelA by p300 (Hoberg et al. 2006).

Here we asked whether IKKα activation by inflammatory signals is also of importance in androgen-induced regeneration of the normal, nonneoplastic prostate. Androgen-induced prostate regeneration is thought to be mediated through activation of adult tissue progenitor/stem cells that remain after involution (English et al. 1987). Presumably, these progenitors are induced to proliferate in a cell-autonomous manner upon androgen receptor (AR) activation (English et al. 1987). We now show that androgen-induced prostate regeneration is also dependent on a non-cell-autonomous mechanism mediated by B lymphocytes that activate IKKα within tissue progenitors. The same mechanism may account for the recurrence of PCa after androgen ablation. In both cases, IKKα operates by inducing the phosphorylation and nuclear translocation of E2F1 to promote its recruitment to the promoters/regulatory regions of the Bmi1 and Ccne genes, which encode positive regulators of progenitor cell proliferation.

Results and Discussion

To study the mechanisms underlying androgen-induced prostate regeneration, we used castration to deprive mice of androgen and reduce prostate weight by >90% within 1 wk (Supplemental Fig. S1A). As shown before (Karhadkar et al. 2004), subcutaneous injection of testosterone (described hereafter as androgen supplementation) every 4 d fully restored prostate weight and structure (Supplemental Fig. S1B,C). To delineate the role of inflammatory signaling in androgen supplementation-induced prostate regeneration, we performed the same experiment in Ikkα^AA/AA mutant mice in which activation loop serines, whose phosphorylation is needed for IKKα activation, were replaced with alanines (Cao et al. 2001). IKKα responds to extracellular inflammatory signals by translocating to the nucleus, and this process is critical for CRPC emergence (Ammirante et al. 2010). Ikkα^+/AA and Ikkα^AA/AA mice exhibited similar prostate weight and structure before castration (Supplemental Fig. S1D). However, unlike Ikkα^+/AA mice, Ikkα^AA/AA mice failed to regenerate their prostates after androgen supplementation (Fig. 1A). Interestingly, immunohistochemistry (IHC) revealed a more than fourfold increase in nuclear IKKα after castration and androgen supplementation in wild-type mice (Fig. 1B). In PCa cells, IKKα activation and nuclear translocation are controlled by LTα:β heterotrimers produced by infiltrating B cells (Supplemental Fig S2) that are recruited upon castration-induced tumor shrinkage (Ammirante et al. 2010). Indeed, immunofluorescence (IF) revealed infiltration of B cells surrounding the regenerating prostate lobe after castration and androgen supplementation (Fig. 1C), and quantitative RT–PCR (qRT–PCR) revealed increased expression of genes encoding LTα and LTβ subunits in prostates of castrated and androgen-supplemented mice (Fig. 1D). B-cell infiltration, LT induction, and IKKα dependence of prostate regeneration were also observed upon daily supplementation of castrated mice with a different androgen-dihydrotestosterone (DHT), indicating that this phenomenon is not procedure-specific (Supplemental Fig. S3A–C). As seen in the previous experiments, the prostate weight in the Ikkα^AA/AA mutant was significantly lower than in heterozygous control mice (Supplemental Fig. S3B). Furthermore LTα and LTβ production was increased in regenerated prostates (Supplemental Fig. S3C). The presence of B cells within the regenerated prostate was specific, as it was not detected in Rag1^−/− mice (which lack both B and T cells) unless they were reconstituted with splenic B cells (Supplemental Fig. S3D). These results indicate that, as seen in PCa, after castration and androgen supplementation, the infiltration of inflammatory cells into the involuting prostate results in IKKα nuclear translocation. This process may be important for androgen-induced regeneration.

Open in a separate window

Figure 1.

B cells and IKKα control prostate regeneration. (A) Prostates of Ikkα^+/AA and Ikkα^AA/AA mice were weighed 17 d after castration and androgen replacement. (B) Prostates of sham-operated or castrated wild-type (WT) mice were collected and analyzed by IHC for IKKα nuclear translocation. Magnification, 400×. Percentages of cells displaying nuclear IKKα are indicated below each panel. n = 3. (C) Prostates of wild-type mice were collected after sham operation or castration and regeneration and analyzed by IF for B220⁺ cell infiltration. Magnification, 400×. Percentages of B220⁺ cells are indicated below each panel. n = 3. (D) RNAs extracted from prostates of wild-type mice 17 d after sham operation or castration and subsequent androgen replacement were analyzed for LTα and LTβ expression. (E) Prostates of wild-type, J_H^−/−, or Cd4^−/− mice were weighed 17 d after castration and androgen replacement. In all panels, the values represent mean ± SD; n = 3.

To examine the role of infiltrating lymphocytes in androgen-induced prostate regeneration, we subjected J_H^−/− and Cd4^−/− mice, which lack either B or CD4⁺ T cells, respectively (Rahemtulla et al. 1991; Chen et al. 1993), to the castration regeneration experiment described above. The B-cell, but not the CD4⁺ T-cell, deficiency abrogated androgen-induced prostate regrowth (Fig. 1E). Furthermore, Rag1^−/− mice, which lack both B and T cells, were also defective in androgen-induced prostate regeneration, but their reconstitution with splenic B cells derived from either wild-type or Ikkα^AA/AA mice restored regeneration (Supplemental Fig. S4A) and expression of LTβ (Supplemental Fig. S4B) as well as the presence of B cells within the regenerating prostate (Supplemental Fig. S3D), thus confirming that B cells rather than T cells are needed for androgen-induced prostate regeneration. J_H^−/− male mice exhibited fully normal prostates prior to castration (Supplemental Fig. S1D) and did not exhibit any reproductive abnormalities, indicating that this phenotype is not due to a defect in initial prostate development,. Prostate epithelial cells expressed LTβR but hardly any receptor activator of NF-κB (RANK) (Supplemental Fig. S5A,B), thus explaining why the RANK ligand, which can also activate IKKα (Luo et al. 2007) and whose expression is also induced upon castration but is not affected by B-cell depletion (data not shown), cannot substitute for LTα:β. Notably, LTα, whose expression is not reduced after B-cell ablation (Supplemental Fig. S4B; Ammirante et al. 2010), cannot bind to LTβR in the absence of LTβ. We therefore postulate that the effect of B cells on the regenerating normal prostate may also be mediated via LTα:β heterotrimers, as demonstrated for CRPC development (Ammirante et al. 2010).

We hypothesized that defective prostate regrowth in Ikkα^AA/AA mice correlated with diminished expansion of prostate epithelial progenitors that are responsible for androgen-induced regeneration and can form protospheres in culture (English et al. 1987; Karhadkar et al. 2004). Indeed, the prostate rudiments of castrated androgen-supplemented Ikkα^AA/AA mice contained much lower numbers of protosphere-forming units (Fig. 2A). To further support our hypothesis, we examined expression of two markers of prostate progenitor cells: p63 and BMI1 (Lukacs et al. 2010). IHC of regenerating wild-type prostates revealed a substantial increase in cells expressing p63 and BMI1 (Fig. 2B,C). Furthermore, nuclear translocation of IKKα was observed in both p63⁺ and BMI1⁺ cells in the regenerating prostates (Supplemental Fig. S6A,B), and IF analysis revealed colocalization of IKKα and BMI1 in nuclei within the regenerating prostate (Fig. 2D). Bmi1 mRNA also increased after castration and androgen supplementation in an IKKα-dependent and B-cell-dependent manner (Fig. 2E,F). Furthermore, B-cell reconstitution of Rag1^−/− mice restored Bmi1 mRNA induction in the regenerating prostates (Supplemental Fig. S4B). Expression of two other putative prostate progenitor cell markers, Sox2 and Cd133 (Kasper 2008), was also higher in Ikkα^+/AA mice than in Ikkα^AA/AA mice (Supplemental Fig. S6C). These results implicate IKKα in the proliferation of prostate epithelial progenitors in response to castration and androgen supplementation. We also found that IKKα and B cells were required for induction of CD45⁻CD49f⁺Sca1⁺ cells within the regenerating prostate (Supplemental Fig. S6D,E). CD49f and Sca1 are thought to be selective markers of murine prostate stem cells (Lawson et al. 2007).

Open in a separate window

Figure 2.

IKKα and B cells control prostate and PCa epithelial progenitor proliferation. (A) Single-cell suspensions of Ikkα^+/AA and Ikkα^AA/AA prostates 17 d after castration and androgen replacement were analyzed for sphere-forming ability. After 10 d, the spheres were dissociated, and equal numbers of cells were replated for secondary sphere formation. The number of sphere-forming cells per prostate was calculated by normalizing the observed number of primary and secondary spheres to the total amount of epithelial cells harvested from each prostate. (B,C) Prostates of wild-type (WT) mice collected 17 d after sham operation or castration and subsequent androgen replacement were analyzed by IHC for p63 (B) or BMI1 (C) expression. Percentages of p63⁺ and BMI1⁺ cells were determined. n = 3. (D) Sham-operated and regenerated prostates of wild-type mice were analyzed by IF for IKKα (green) and BMI1 (red). Magnification, 200×. (E,F) RNAs extracted from prostates of the indicated genotypes prepared as above were analyzed for Bmi1 mRNA by qRT–PCR. In all panels, the values represent mean ± SD; n = 3.

If IKKα controls expansion of prostate epithelial progenitors, it may have a similar effect on PCa progenitors, which presumably account for CRPC development (Maitland and Collins 2008). Indeed, analysis of subcutaneous tumors formed by androgen-dependent myc-CaP cells at 2 wk after castration revealed that castration increased the tumoral content of p63⁺ cells and Sox2, Nanog, and Cd133 mRNAs and that all of these increases were IKKα-dependent (Supplemental Fig. S7A–C). CD133⁺ cells from myc-CaP tumors displayed nuclear IKKα (Supplemental Fig. S7D), and the castration-induced increase in their frequency was inhibited by IKKα silencing but rescued by ectopic BMI1 expression at 1 wk after castration (when tumors formed by IKKα-positive and IKKα-silenced cells were similar in size) (Supplemental Fig. S7E,F). TRAMP mice in which PCa development is induced by SV40 T antigen also develop CRPC after castration (Greenberg et al. 1995). Homozygosity for the Ikkα^AA allele modestly retarded primary tumor growth (Supplemental Fig. S8A) but completely prevented CRPC emergence in castrated TRAMP mice (Fig. 3A; Supplemental Fig. S8B). Carcinoma cells of CRPC in Ikkα^+/AA/TRAMP mice exhibited IKKα nuclear translocation (Supplemental Fig. S8C), and tumoral expression of BMI1 was IKKα-dependent (Fig. 3B; Supplemental Fig. S8D,E). A strong correlation between nuclear IKKα and high BMI1 expression was also seen in human PCa (Fig. 3C). These data suggest that nuclear IKKα controls CRPC development through expansion of BMI1⁺ progenitors and that it may have a similar role in human PCa.

Open in a separate window

Figure 3.

IKKα stimulates E2F1-mediated Bmi1 gene induction. (A) Histological analysis (H&E staining of paraffin-embedded sections; magnification, 100×) of prostates from 24-wk-old Ikkα^+/AA/TRAMP and Ikkα^AA/AA/TRAMP mice that were castrated or sham-operated at 12 wk of age. (B) Prostates of 24-wk-old mice of the indicated genotypes castrated at 12 wk of age were analyzed for Bmi1 mRNA expression by qRT–PCR. (C) Serial sections of paraffin-embedded human PCa tissues were analyzed for nuclear IKKα and BMI1 by IHC. Nuclear staining indices were quantitated, and the correlation coefficient is indicated at the left. Magnification, 400×. (D) ChIP analysis of IKKα recruitment to the Bmi1 promoter. Cross-linked and sheared chromatin from myc-CaP cells stably infected with GFP or IKKα(EE) lentiviruses was immunoprecipitated with IKKα or HA antibodies and analyzed by qPCR with primers to the Bmi1 promoter or an intergenic region of chromosome 8. The percentage input was calculated by determining the ratio of immunoprecipitated DNA with each antibody to a 10% input sample. (E) Chromatin was isolated from myc-CaP tumors established by mock-silenced (Scr.) or IKKα-silenced cells 2 wk after sham operation or castration. After cross-linking and shearing, chromatin was immunoprecipitated with antibodies to E2F1 or control IgG, and the content of Bmi1 promoter DNA was determined as above. (F) Nuclear extracts were prepared from the myc-CaP tumors described above, and E2F1 DNA-binding activity was analyzed by EMSA with the Bmi1 E2F1-binding site. For competition experiments, 100-fold excess of nonradioactive competitor oligonucleotides representing the Bmi1 E2F1 site or consensus binding sites of the indicated transcription factors were added to the reactions 10 min before probe addition. In B, D, and E the values represent mean ± SD; n = 3.

As IKKα nuclear translocation correlated with BMI1 expression in both normal and transformed prostate epithelial cells, we asked whether nuclear IKKα directly regulated Bmi1 transcription. Bmi1 mRNA expression increased in myc-CaP cells during CRPC formation in an IKKα-dependent manner (Supplemental Fig. S9A–C). LT sequestration with an LTβR-Fc fusion protein inhibited Bmi1 mRNA expression in myc-CaP tumors of castrated mice (Supplemental Fig. S9D). Transduction of human and mouse PCa cell lines with lentiviruses encoding constitutively active HA-tagged IKKα(EE), but not with inactivatable IKKα(AA) or nuclear transport-defective IKKα(EE-NLS) (Sil et al. 2004), enhanced Bmi1 expression (Supplemental Fig. S10A,B), supporting a direct involvement of IKKα in control of Bmi1 transcription. Chromatin immunoprecipitation (ChIP) analysis of these cells revealed an association between IKKα(EE) and the Bmi1 promoter (Fig. 3D). Similarly, ChIP analysis of CRPC tumor samples revealed an association between endogenous IKKα and the Bmi1 promoter (Supplemental Fig. S10C,D). Binding of IKKα to the Bmi1 promoter was specific, as it was not detected in cells with silenced IKKα expression (Supplemental Fig. S10D). As IKKα has no recognizable DNA-binding domain, we searched for a sequence-specific transcription factor that may recruit the kinase to the Bmi1 promoter. While ChIP experiments ruled out involvement of NF-κB RelA or RelB proteins (Supplemental Fig. S11A), similar experiments demonstrated IKKα-dependent recruitment of E2F1 to the Bmi1 promoter (Fig. 3E). Direct control of Bmi1 transcription by E2F1 has been reported previously (Tu et al. 2006). IKKα-dependent E2F1 binding to the Bmi1 promoter was not due to effects of IKKα on E2F1 expression or c-Myc recruitment to the Bmi1 promoter (Supplemental Fig. S11B,C; Guo et al. 2007). Electrophoretic mobility shift assays (EMSAs) confirmed IKKα-dependent binding of E2F1 to a DNA sequence from the Bmi1 promoter (Fig. 3F), and UV cross-linking experiments revealed that E2F1 bound this sequence much stronger than IKKα (Supplemental Fig. S11D).

We examined how IKKα may control E2F1 chromatin recruitment. In vitro, IKKα phosphorylated E2F1 at Ser403 (Supplemental Fig. S11E), a previously reported E2F1 phosphorylation site (Ivanova et al. 2009). A phosphomimetic Ser403-to-Glu403 substitution promoted E2F1 nuclear localization (Fig. 4A), and after castration, E2F1 translocated to nuclei of prostate epithelial and PCa cells in an IKKα-dependent manner (Fig. 4B,C), thus explaining how IKKα enhances E2F1 recruitment to the Bmi1 promoter. Castration also increased the nuclear abundance of phosphorylated (phospho-T433) E2F1 in PCa cells (Supplemental Fig. S11F). Activated IKKα also supported an interaction between E2F1 and CBP (Supplemental Fig. S11G), while it interacted with E2F1 (Supplemental Fig. S11H). CBP is a histone H3 acetyltransferase (Iyer et al. 2004), and castration of tumor-bearing mice led to IKKα-dependent H3 acetylation and CBP recruitment (Fig. 4D,E) to the Bmi1 promoter in Myc-CaP cells. This suggests that IKKα may further regulate E2F1 activity via CBP, which can acetylate E2F1 to increase its stability and DNA binding (Trouche and Kouzarides 1996; Ianari et al. 2004) and is recruited to certain promoters in an IKKα-dependent manner (Hoberg et al. 2006; Huang et al. 2007). In support of this interpretation, ectopic E2F1 expression in Myc-CaP cells increased Bmi1 expression in an IKKα-dependent manner (Supplemental Fig. S12A). Functional cooperation between IKKα and E2F1 was also observed at the Ccne (cyclin E) gene, a well-documented E2F1 target (Supplemental Fig. S12B,C; Ohtani et al. 1995). Taken together, these results suggest that E2F1 is phosphorylated by IKKα in order to enhance E2F1 nuclear localization. This results in E2F1-mediated recruitment of CBP to the Bmi1 promoter, thus enhancing Bmi1 transcription.

Open in a separate window

Figure 4.

IKKα phosphorylates BM1 and controls its nuclear translocation. (A) HEK293T cells were transfected with vectors expressing wild-type (WT) E2F1 or mutants in which the IKKα phosphorylation site was replaced with alanine (A) or glutamate (E) residues. After 48 h, nuclear extracts were prepared and analyzed by immunoblotting. (B) Prostates of wild-type mice were collected after sham operation or castration and androgen replacement and analyzed by IHC for E2F1α. Magnification, 400×. (C) Subcutaneous myc-CaP bearing mock-silenced (Scr.) or IKKα-silenced tumors were collected 2 wk after castration or sham operation and analyzed by IHC for E2F1. Magnification, 400×. (D,E) Chromatin isolated from subcutaneous myc-CaP tumors formed by mock-silenced (Scr.) or IKKα-silenced cells 2 wk after sham operation or castration was subjected to ChIP analysis with antibodies to CBP (D) or Ac-H3 (E). Relative percentage input was determined as above. In D and E, the values represent mean ± SD; n = 3.

We next examined the functional role of IKKα-induced Bmi1 expression in PCa. BMI1 depletion with siRNA underscored its importance for growth of myc-CaP tumors, as BMI1 silencing in Myc-CaP cells reduced tumor growth, and the few tumors that grew after 4 wk displayed BMI1 expression, suggesting that they were derived from cells that escaped BMI1 silencing (Supplemental Fig. S13A,B). Conversely, ectopic BMI1 expression in myc-CaP cells whose IKKα was siRNA-silenced strongly enhanced tumor regrowth in castrated mice (Fig. 5A; Supplemental Fig. S13C). Curiously, 3 wk after castration, IKKα-silenced myc-CaP cells had regained BMI1 expression (Fig. 5B), thus explaining the temporary nature of delayed tumor regrowth caused by IKKα silencing in these very aggressive cells. BMI1 was proposed to promote cell proliferation and inhibit senescence through epigenetic silencing of the p16^Ink4a/p19^Arf locus (Jacobs et al. 1999). Congruently, tumors formed by IKKα-silenced Myc-CaP cells or those that grew in the absence of LT signaling exhibited elevated p16 and p19 expression, which was reduced by ectopic BMI1 (Fig. 5C; Supplemental Fig. S13D). ChIP experiments indicated that castration increased the recruitment of BMI1 to multiple sites at the Ink4a/Arf locus (Fig. 5D; Supplemental Fig. S13E). BMI1 and other PRC1 components suppress Ink4a/Arf transcription through direct binding and H2A ubiquitination (Wang et al. 2004), a chromatin mark whose presence at the Ink4a/Arf locus increased after castration in an IKKα-dependent manner (Supplemental Fig. S13F). There was also a significant, IKKα-dependent increase in global ubiquitinated H2A (ubi-H2A) within PCa cells after castration in both myc-CaP and TRAMP tumors (Fig. 5E,F). We also found significantly elevated BMI1 and ubi-H2A staining in human PCa specimens relative to normal prostatic tissue or benign prostatic hyperplasia (Fig. 5G,H) as well as a strong correlation coefficient of 0.94 between BMI1 and ubi-H2A staining of these samples. These results suggest that IKKα-induced Bmi1 expression is a critical factor in promoting PCa proliferation, possibly through suppression of the Ink4a/Arf locus.

Open in a separate window

Figure 5.

BMI1 controls castration-resistant tumor growth. (A) myc-CaP tumors were established in wild-type (WT) mice with cells subjected to mock silencing (Scr.), IKKα silencing, or IKKα silencing + ectopic BMI1 expression. When tumors reached 500 mm³, mice were castrated, and tumor volume was measured. (B) RNA from subcutaneous myc-CaP tumors formed by mock-silenced (Scr.) or IKKα-silenced cells was collected 2 or 3 wk after castration and analyzed for Bmi1 expression. (C) RNAs from subcutaneous myc-CaP tumors formed by the cells in A were analyzed for p16 and p19 expression 2 wk after castration. (D) Chromatin from subcutaneous myc-CaP tumors formed by mock-silenced (Scr.) or IKKα-silenced cells collected 2 wk after castration or sham operation was subjected to ChIP analysis using antibodies to BMI1 or control Ig. The presence of Ink4a/Arf sequences was examined by PCR using the primers in Supplemental Figure S13E or the intergenic region from chromosome 8. (E) Subcutaneous myc-CaP tumors formed as in A were analyzed by IHC for ubi-H2A content. (F) Prostates of Ikkα^+/AA/TRAMP and Ikkα^AA/AA/TRAMP mice were collected 12 wk after sham operation or castration and stained for ubi-H2A. (G,H) Paraffin-embedded human prostate sections representing nonmalignant tissue (n = 17) and PCa (n = 13) were stained for BMI1 (G) or ubi-H2A (H). The staining indices are indicated at the bottom. P < 0.01 for the malignant samples. Magnification, 400× for all images. In A–D, the values represent mean ± SD; n = 3.

Tissue injury and subsequent inflammation can promote tumorigenesis by activating a regenerative response (Beachy et al. 2004; Kuraishy et al. 2011). However, how injury-triggered inflammation controls regeneration has not been clear. We now show that androgen-induced regrowth of the involuted prostate is not an entirely cell-autonomous process as previously assumed, as it is highly dependent on activation of IKKα in prostate tissue progenitors in response to a signal provided by B cells that infiltrate the androgen-deprived prostate. This signal is likely to be LTα:β heterotrimers, which bind to LTβR (Ammirante et al. 2010). Most likely, the trigger for B-cell infiltration in both cases is tissue injury or cancer cell death, which generates signals that lead to induction of CXCL13, a B-cell chemoattractant (Ammirante et al. 2010). The one difference between androgen-induced prostate regeneration and recurrence of CRPC is their dependence on androgens. Whereas regeneration of the normal prostate does not occur in the absence of androgen supplementation, the residual amount of androgens present in castrated mice is sufficient for driving the growth of CRPC as long as IKKα is activated. Our results demonstrate that the IKKα––BMI1 axis is also activated in human PCa and suggest that its pharmacological inhibition in combination with androgen ablation can prevent CRPC emergence. Notably, neither IKKα inhibition nor androgen ablation alone can completely block prostate tumor growth in TRAMP mice. However, mice that have been subjected to both castration and IKKα inhibition (at least genetically) do not display any recurrent tumors for at least several months after androgen ablation. Thus, IKKα inhibition appears to exert a synthetic lethal effect when combined with androgen ablation.

Materials and methods

In brief, mice were handled according to institutional and National Institutes of Health (NIH) guidelines. For prostate regeneration experiments, all mice were on the FVB genetic background. Ikkα^AA and TRAMP mice were intercrossed to create Ikkα^AA/AA/TRAMP and Ikkα^+/AA/TRAMP littermates of nearly identical genetic background. subcutaneous myc-CaP tumors were grown in FVB mice. Human material was obtained from Cooperative Human Tissue Network (CHTN). Additional methods are available in the Supplemental Material.

Acknowledgments

Footnotes

References