Clusterin Silencing in Prostate Cancer Induces Matrix Metalloproteinases by an NF-κB-Dependent Mechanism

Clusterin (CLU) is a stress-activated glycoprotein, whose expression is altered both in inflammation and cancer. Previously, we showed that abrogation of CLU expression in cancer-prone mice (TRAMP) results in the enhancement of tumor spreading and homing, concomitant with an enhanced expression of NF-κB. In the present paper, we carried out an extensive experimental work by utilizing microarray gene expression data, as well as in vitro and in vivo models of prostate cancer (PCa). Our results demonstrated that (i) CLU expression is significantly downregulated in human PCa and inversely correlates with the expression of p65 in metastases; (ii) CLU overexpression in PCa cells reduces the Ser536 phosphorylation of p65, inhibits NF-κB nuclear translocation, and reduces the transcription of matrix metalloproteinase-9 and metalloproteinase-2 (MMP-9 and MMP-2). Conversely, CLU silencing promotes NF-κB activation and transcriptional upregulation of MMP-9; and (iii) expression and activity of MMP-2 and MMP-9 are increased in CLU−/− mice (CLUKO) and in TRAMP/CLUKO mice in comparison to their relative Clu+/+ littermates. Taken together, our data support the hypothesis that CLU downregulation, an early and relevant event in PCa onset, may inhibit NF-κB activation and limit the execution of a transcriptional program that favor the disease progression towards a metastatic stage.


Introduction
Chronic infection and inflammation have been recognized to start cell transformation and promote cancer development [1]. e transcription factor NF-κB is the main trigger of proinflammatory processes and key molecular link of inflammation to tumor initiation and progression [2,3]. NF-κB is also required to sustain a proinflammatory microenvironment that facilitates extracellular matrix (ECM) degradation and tumor cell dissemination in prostate cancer (PCa) [4,5]. Clusterin (CLU) is a secreted glycoprotein known to be involved both in inflammation and cancer [6][7][8]. CLU may be retained inside the cell as a consequence of various stress-activated mechanisms, which include retrotranslocation from the endoplasmic reticulum, alternative splicing, alternative transcription initiation, and internalization from the extracellular compartment [9,10]. Intracellular CLU interferes with relevant cell signaling pathways, including NF-κB [11,12]. In vivo CLU has antiinflammatory functions; indeed, in the experimental model of induced autoimmune myocarditis and pancreatitis, CLU knockout mice (CLUKO) show signs of more severe inflammation and cellular pathology than CLU-expressing wild-type controls (WT) [13,14]. CLU expression is altered in many tumors including PCa, although conflicting data about its tumor suppressive or tumor permissive role have been published [8]. We and other authors have observed that CLU is downregulated in human PCa progression [15,16] and in tumors arising in the TRansgenic Adenocarcinoma of the Mouse Prostate (TRAMP) model [17,18]. Moreover, CLUKO mice are more susceptible than WT to chemically induced skin tumorigenesis, suggesting that CLU might negatively modulate epithelial cell transformation [19]. When CLUKO mice were crossed with TRAMP to obtain TRAMP/CLUKO mice, we found that tumor spreading and metastases occurred earlier in animals lacking CLU expression [20]. Cancerous lesions of TRAMP prostates are positive for NF-κB and Ki67 staining in comparison to WT. Of note, an even stronger expression of these proteins was detected in the cancerous lesions of age-matched TRAMP/ CLUKO [20], suggesting the hypothesis that one possible mechanism by which CLU slows down tumor spreading in TRAMP mice might also involve limiting NF-κB activity. e aim of the present work is to challenge the aforementioned hypothesis by an articulated experimental approach that started from examining the expression of p65 and CLU in large microarray dataset retrieved in GEO (Gene Expression Omnibus) which included expression profiling of normal, prostate primary tumors, and metastatic tissue. en, we evaluated the changes of NF-κB expression and activity in human PCa cells following either CLU overexpression or silencing. Finally, we measured in prostate tissue obtained from WT, CLUKO, TRAMP, and TRAMP/ CLUKO, the expression and activity of ECM metalloproteinases (MMP-2 and MMP-9) that are known to be involved in tumor dissemination being regulated by the NF-κB pathway.

Public Domain
Data. mRNA expression levels of the human prostate sample from microarray dataset GSE6919 were analyzed with GEO2R web tool (http://www.ncbi. nlm.nih.gov/geo/geo2r/). GSE6919 was performed on the Affymetrix Human U95 Version 2 Array platform (GPL8300), which comprises the expression profiles of the following human specimens: 18 normal prostates, 63 normal prostates adjacent to the tumor, 65 primary prostate tumors, and 25 prostate metastases. Expression levels of CLU and p65 were compared between primary tumors and metastatic tissues. e Benjamini and Hochberg false discovery rate method was used to correct for multiple comparisons.

Cell
Culture. PC3 cells were purchased from the American Tissue Culture Collection and were routinely grown in Ham's F12 : MEM medium (1 : 1). e culture media were supplemented with 10% foetal bovine serum (Lonza, Basel, CH), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. e cells were incubated at 37°C under a 5% CO 2 atmosphere and harvested by trypsin/ EDTA (Sigma-Aldrich, Steinheim, DE). e full-length human CLU coding sequence was previously cloned into the bicistronic expression vector pIREShyg1 (U89672, Clontech, Oxford, UK) [21]. PC3 CLU and PC3 mock were generated by clonal selection and maintained as previously described [21].  [22]. CLU-deficient mice (CLUKO) backcrossed to the C57Bl/6 strain for more than 10 generations were obtained breeding heterozygous parents and genotyping the offspring for CLU expression [13]. Female TRAMP mice were mated with homozygous (CLUKO) male mice to obtain TRAMP/CLU +/− mice. en, TRAMP/CLU +/− females were mated with heterozygous (CLU +/− ) male mice to obtain TRAMP/CLUKO mice. Mice were housed in a standard animal facility under controlled environmental conditions (22 ± 2°C, 12 hours light/dark cycle) and were allowed free access to food and water. For the immunohistochemical detection of p65, a total of 12 mice, 3 for each experimental group, were sacrificed at 12 and 24 weeks of age, the prostates were excised, fixed in 10% (vol/vol) neutral-buffered formalin at 4°C overnight, and processed to paraffin using standard procedures for immunohistochemical analysis. For the zymography assay, a total of 12 male mice, 3 for each experimental group (WT, CLUKO, TRAMP, and TRAMP/ CLUKO) were euthanized at 36 weeks of age. Prostates were excised, snap-frozen in liquid nitrogen, and stored at − 80°C until use. All the experimental procedures involving transgenic mice were approved and conducted in accordance with the Italian law (D.lgs 26/2014).

CLU Transient
manufacturer's instructions [23]. Primers information is reported in Table 1

Immunohistochemical Analysis.
Paraffin-embedded tissue sections (5 mm) mounted onto glass slides were hydrated in xylene, graded alcohol. Antigen retrieval was performed by microwave (5 min at 700 W) using a sodium citrate buffer (10 mM, pH 6.0). Endogenous peroxidase activity was quenched with 3% H 2 O 2 . Nonspecific binding was blocked with swine serum diluted 1 : 10 in BSA 1%. Immunostaining was performed using the mouse monoclonal anti-p65 antibody dilution 1 : 50 in BSA 1% (Santa Cruz Biotechnology, Dallas, TX) and incubated for 1 h at room temperature. From the secondary antibody to the chromogen reaction, a Universal LSAb2 HRP kit (Dako-Cytomation, Glostrup, DK) was used according to the manufacturer's instruction as previously described [20]. 2.14. Immunoprecipitation Assay. 500 μg of intracellular protein from PC3 mock and PC3 CLU cells was precleared with 50 μL of protein G-agarose beads (Roche, Basel, CH) for 30 min at 4°C with gentle rotation. en, the lysates were incubated for 18 hours at 4°C with 5 μg of the following antibodies: mouse monoclonal anti-CLU (05-354, Millipore, Billerica, MA), mouse monoclonal anti-p65 (#6956, Cell Signaling Technology, Denver, MA), and normal mouse IgG (NI03, Millipore, Billerica, MA). 50 μL of protein G-Agarose beads (Roche, Basel, CH) was added and incubated for 4 hours at 4°C with gentle rotation. After centrifugation, immune-precipitated pellets were washed four times with icecold RIPA buffer, dissolved in the Laemmli buffer added with 100 nM dithiothreitol, and heated at 100°C for 3 min for reverse cross-link.

Statistical Analyses.
e IBM SPSS statistical package (International Business Machines Corporation, Armonk, NY, USA, version 24) was used. Normal distribution of variables was checked by means of the Kolmogorov-Smirnov test. Statistics of variables included mean-± standard deviation (SD), the unpaired Student's t-test (luciferase assay and qPCR), and the Mann-Whitney U test (qPCR data). Statistical significance was set at p < 0.05. Pearson's correlation test on microarray data (n � 80) was performed with GraphPad Prism 6 software. Details on the specific test used are reported in the figure legend of each experiment.

CLU and p65 Are Significantly Inversely Correlated in Human Samples of PCa Metastases.
e GEO2R web tool was used to compare the mRNA expression of CLU and p65 in the microarray dataset GSE6919, which comprises primary and metastatic prostate tumors, normal prostate tissue adjacent to tumor, and normal prostate tissue from healthy donors (Figure 1(a)). e highest expression of CLU was found in normal prostate tissue and gradually decreased in normal tissue adjacent to the tumor and primary prostate tumor, reaching the lowest value in metastatic samples. A statistically significant decrease in CLU along with a significant increase in p65 expression was observed when tissues from primary and metastatic PCa were compared (Figure 1(a)). Moreover, in primary tumor and metastases, the expression of CLU and p65 is inversely correlated (Figure 1(b)).

NF-κB Transcriptional Activity Is Inhibited by CLU
Overexpression. At least five CLU (PC3 CLU ) and five mock clones (PC3 mock ) were analyzed for CLU content, cell morphology, cell proliferation, and percentage of distribution of cells in different phases of the cell cycle. CLU mRNA is significantly more expressed in PC3 CLU than in PC3 mock (Figure 2(a)). Both the precursor form of 64 kDa (psCLU) and the mature form of 40 kDa (sCLU) were expressed at a higher level in cell lysates and culture media of PC3 CLU compared to PC3 mock , although the intensity of the WB blot bands shows some differences depending on the single clone analyzed (Figure 2(c)-S1). ese results have also been confirmed in a polyclonal population obtained by pooling together several PC3 CLU clones different from those analyzed individually (Figure 2(c)). e immunofluorescence analyses confirmed that CLU is significantly more expressed in PC3 CLU than in PC3 mock and localizes to the cytoplasm (Figure 2(b)). PC3 CLU cells are more round shaped than PC3 mock ( Figure S2(a)) and grow slower than PC3 mock ( Figure S2(b)). We also measured an impairment in the cell cycle distribution, which is consistent with a slowdown of the progression between the G0/ G1 and the S phase in PC3 CLU compared to PC3 mock ( Figure S2(c)). When, in the same clones, we measured the expression of NF-κB (p65 total) and active NF-κB (p-p65 S536 ), we detected a reduction of p-p65 S536 in PC3 CLU compared to controls (Figure 2(c)-S1), also in the polyclonal samples that ideally represent the averaged trend of many cell clones (Figure 2(c)). e immunocytochemical (IC) analysis showed that p-p65 S536 is reduced in PC3 CLU compared to PC3 mock , Table 1: Sequences of the primers used in qPCR.

CLU Silencing Increases NF-κB Transcriptional Activity.
CLU expression was almost completely abrogated in PC3 cells, starting from 24 and up to 48 hours after CLU siRNA transfection in comparison to negative control (NC) transduced cells (Figure 5(a)). In the same samples, p-p65 S536 increased after 24 hours and much more significantly after 48 hours in CLU siRNA ( Figure 5(b), lanes CLU 24 and CLU 48) compared to NC transduced cells ( Figure 5(b) lanes NC 24 and NC 48). IκBα and IKKβ did not change, while a significant increase of Akt was detected in CLU 48 in comparison to NC 48 ( Figure 5(b)). By the Luciferase assay, we found that NF-κB transcriptional activity increased in PC3 cells transduced with CLU siRNA in comparison to NC (Figure 5(c)). Consistently, the expression of MMP-9 mRNA is significantly increased 24 and 48 hours after CLU siRNA transfection in comparison to NC ( Figure 5(d)).

CLU Does Not Prevent NF-κB Activation by means of a
Direct Interaction with the p65 Subunit. We investigated whether CLU may physically hide the accessibility of S 536 to specific kinases such as IKKβ by direct binding with p65. erefore, we immune-precipitated (IP) CLU and p65 from PC3 CLU and PC3 mock cell lysates. en, we searched for CLU and p65 physical interaction by WB analysis of the IP fractions. CLU was successfully pulled down when the specific anti-CLU antibody was used for immunoprecipitation (IP positive control), as demonstrated by the presence of a band at 64 kDa in the IP fraction ( Figure 6(a), upper panel). e result of the immunoprecipitation reaction is specific because no CLU band is detectable in the mouse IgG immunoprecipitated sample (negative control). No bands were detected, instead, when the same membrane was probed with an anti-p65 antibody, indicating that no direct interaction took place between CLU and p65 in PC3 CLU compared to PC3 mock (Figure 6(a), lower panel). Similarly, when the intracellular lysates were immunoprecipitated with an anti-p65 antibody, we were able to detect p65 in the IP fraction (positive control), while no p65 was detected in the mouse IgG immunoprecipitated sample (negative control) ( Figure 6(b), upper panel). No bands were detected, instead, when the same membrane was probed with an anti-CLU antibody ( Figure 6(b), lower panel).

MMP-2 and MMP-9 Gelatinolitic Activity Increased in TRAMP/CLUKO Mice Compared to TRAMP Littermates.
By immunohistochemical analysis, we observed that p65 is more expressed in prostate tissues of TRAMP/CLUKO mice than age-matched TRAMP animals expressing CLU ( Figure S3). en, we proceeded measuring the expression level and the enzymatic activity of MMP-2 and MMP-9 in prostate samples obtained from WT, CLUKO, TRAMP, and TRAMP/CLUKO littermates. e lowest median values of normalized CTfor MMP-2 were measured in the CLUKO and TRAMP/CLUKO prostates by qPCR analysis of mRNA expression. e increase in MMP-2 expression is statistically significant when TRAMP/CLUKO is compared with TRAMP (Figure 7(a)). e lowest median values of normalized CT for MMP-9 were measured in CLUKO mice, in which MMP-9 expression is therefore significantly upregulated compared to WT littermates (Figure 7(b)). e enzymatic activity of MMP-2 and MMP-9 was assessed by gelatine zymography. A strong increase in MMP-9 and MMP-2 gelatinolitic activity was observed in both CLUKO and TRAMP/CLUKO mice in comparison to WT and TRAMP mice, respectively. e strongest difference was observed for the pro-MMP-9 band, corresponding to the uncleaved precursor protein. Additional bands at lower molecular weight, corresponding to the cleaved active enzymes (aMMP-9 and a-MMP-2), are more expressed in the prostate tissues of animals that lack CLU expression.

Discussion
Although the adaptive immune system mediates antitumor effects through immune surveillance, many tumors acquire the ability to subvert inflammatory signals to their benefit [4]. NF-κB is aberrantly activated in the majority of PCa cases, and it is considered the major transcription factor contributing to sustain a proinflammatory, tumor-permissive microenvironment [24]. Although many molecules act as NF-κB activators, very few proteins are known to negatively interfere with NF-κB signaling, a part from IκBs proteins, which are the major inhibitors of the pathway.  CLU is a heavy glycosylated secreted heterodimeric protein that is involved in many processes characterized by cellular stress, like ageing, chronic tissue inflammation, metabolic diseases, and cancer [8,25]. CLU mRNA is downregulated in the vast majority of naïve cancers according to Oncomine ® database, and PCa does not make exception [6]. e analyses performed on the microarray dataset GSE6919 with GEO2R web tool showed that CLU expression is reduced in PCa in comparison to normal prostate tissue, following an ideal decreasing gradient from normal healthy prostate to epithelial tissues adjacent to primary tumors and then primary tumors up to metastases. Interestingly, the expression of NF-κB is significantly increased in metastases, where it inversely correlates with CLU expression. On the one hand, these results support early reports from our group [15,16,26] and agree with the finding that histone tails modification, including H3 pan-deacetylation and H3K27 trimethylation as well as CpG island hypermethylation at the CLU promoter contribute to the epigenetic repression of this gene both in human PCa cell lines and in the TRAMP model [17,27]. On the other hand, the same results highlight that the inverse relation between CLU and NF-κB expression, previously noticed in CLUKO and TRAMP/CLUKO mice [20], can be extended from the animal model to the human disease. Previously, some in vitro results have suggested that the relation between NF-κB and CLU is complex, in that CLU is transcriptionally upregulated by NF-κB [28] and, at the same time, intracellular CLU can inhibit NF-κB activation by stabilizing the inhibitor IκB [7,11,12,29]. In addition to the IκB-mediated regulation of NF-κB nuclear translocation, several studies have shown that the NF-κB proteins are posttranslationally modified. ese changes may directly control not only dimer interaction with other factors, their stability, and turnover but also may contribute to the selective regulation of NF-κB transcriptional activity in a gene-specific manner. Altogether, these events contribute to a multifaceted NF-κB signaling control that is enormously more sophisticated than a simple on/off switch [30]. By gain-offunction and loss-of-function experiments in androgenindependent PCa cell line PC3, we demonstrated that CLU expression is required to reduce the amount of constitutively active NF-κB, by decreasing p-p65 S536 , without significantly affecting IκBα expression. Indeed, keeping in mind that IκBα is a direct transcriptional target of NF-κB, the slight reduction of IκBα measured in PC3 cells overexpressing CLU 48 hours after transfection is a direct consequence of the inhibition of NF-κB transcriptional activity measured in these cells [31,32]. Indeed, following ectopic expression of CLU c-DNA, the expression of p-p65 S536 is reduced along with p65 fraction translocated in the cell nuclei in comparison to empty vector transfected cells. e S 536 phosphorylation site is located in the transactivation domain of p65 and is conserved in both human and mouse [33]. It has been observed that the constitutive phosphorylation of p65 at serine 536 (p-p65 S536 ) affects the p65 dimer composition as well as the selective recruitment of phospho-p65 to specific promoters [34]. Of note, p-p65 S536 is increased in patients carrying TMPrSS2/ ERG (T/E) fusion, the most common gene rearrangement in PCa, where it is suggested to play a critical role in PCa tumorigenesis by enhancing the p65 transcriptional activity of both proinflammatory chemokines and ECM-modifying enzymes, thus producing a tumor-permissive microenvironment [35]. Under this experimental condition, the luciferase assay confirms a reduction of NF-κB transcriptional activity. e expression of NF-κB transcribed genes involved in ECM degradation, such as MMP-9 and MMP-2, is consistently reduced. In order to exclude any possible bias coming from the unphysiological condition of protein overexpression, we integrated our experimental design with a complementary approach of gene function modulation by silencing CLU through siRNA oligonucleotides. After silencing CLU in PC3 cells, we observed opposite results compared to those detected in CLU overexpression models, i.e., an increase of p-p65 S536 and an increase of Nf-κB activity accompanied by an increase of MMP-9 expression. Our  Journal of Oncology results are in good agreement with previous findings from other authors in different models [11,12,29]. A unique study showed that NF-κB is activated by CLU-mediated stabilization of COMMD1, a protein that facilitates IκB ubiquitination and proteasomal degradation [36]. We observed that CLU regulates NF-κB activity by reducing p65 phosphorylation at S 536 . Because p-p65 S536 does not bind IκBα nor is proteasome degraded, we suggest that CLU might inhibit NF-κB constitutive activation by an IκB-independent pathway [34]. e mechanism is supposed to be particularly relevant in androgen-independent cells such as PC3 and DU145 that constitutively express high levels of active Nf-κB, being almost unresponsive to TNFα stimulation [37]. Based on the results of the IP analysis, we excluded a direct binding between CLU and p65 and suggest that CLU regulates the amount of intracellular p-p65 S536 by regulating the activity of upstream kinases as IKKβ or Akt by a mechanism similar to that observed in cancer cells under growth-constraining condition [38]. is hypothesis is corroborated by the fact that Akt and IKKβ expression are reduced in PC3 cells overexpressing CLU for 48 hours in comparison to mock transfected controls. On the contrary, Akt expression is increased in PC3 cells silenced for CLU for 48 hours compared to NC-transduced control cells. Although the present study was conducted in a single cell line, which is in fact a limitation of the experimental design, we previously showed that CLU stable overexpression in prostate epithelial immortalized cells (PNT1a) inhibits NF-κB activity by reducing Akt expression [20]. e finding that CLU reduces MMP-9 and MMP-2 expression by inhibiting NF-κB is biologically relevant for PCa because these enzymes play a role in tumor dissemination by degrading ECM and basement membrane. Conflicting data have been reported about CLU effects on MMP expression in cultured cells. Administration of highly supraphysiological amounts of exogenous purified CLU in a murine macrophage cell line increased MMP-9 expression and activity. Disappointingly, no effects were observed following CLU silencing, questioning the specificity of the observed phenomena [39]. CLU increased MMP-2 expression in hepatocellular carcinoma cells [40], while CLU overexpression inhibited MMP-9 expression by reducing  NF-κB activation in vascular smooth muscle cells [41]. CLU prevented stress-induced MMP-9 aggregation and activation of various MMP, both inside and outside human epithelial cells, by direct binding with the catalytic MMP domain [42]. To the best of our knowledge, our study is the first one to report the effects of CLU abrogation on MMP expression and activity in vivo, in a preclinical model of PCa. We found that MMP-2 and MMP-9 expression is higher in prostate samples from CLUKO and TRAMP/CLUKO mice in comparison to WT and TRAMP littermates. Moreover, the results of zymography assay show that MMP-2 and MMP-9 activity is higher in the prostate of CLU-null mice. ese data are consistent with the observation that tumors in TRAMP/CLUKO mice showed increased tendency to metastasize when comparing with tumors in TRAMP mice expressing normal levels of CLU [20].

Conclusions
Overall, the presented data support the hypothesis that in normal physiological condition CLU participates to a negative loop in which transcriptional activation of CLU by NF-κB is evoked to dampen the excessive constitutive activation of NF-κB itself. When CLU expression is reduced in PCa by epigenetic mechanisms, the physiological brake on NF-κB might be relieved.
is effect is more important in the late PCa stage, when Nf-κB activity and expression reach the maximum. In fact, the increased NF-κB activity may cause an increased expression of proteases involved in ECM degradation that favors PCa progression towards a metastatic disease. Considering the important involvement of inflammatory processes in PCa onset and progression, the pharmacological strategies aimed at counteract epigenetic CLU downregulation in early stages of PCa could be crucial for effectively controlling the signaling pathways that lead to proliferation and invasion in advanced cancers.

Data Availability
No data were used to support this study.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper. Figure S1: p65 expression and phosphorylation in additional PC3mock, PC3CLU, and p-p65S536 clones. Quantification of CLU, p65, and p-p65S536 protein by WB analysis in PC3CLU (namely, clones #C5, #C6, and #C7) and PC3mock (namely, clone #M4). Red Ponceau staining was used for loading and transfer control. icCLU, intracellular CLU; psCLU, uncleaved CLU precursor, 64 kDa; sCLU, cleaved mature CLU, 40 kDa. Figure S2: morphology, proliferation, and cell cycle analysis of PC3mock and PC3CLU clones. (a) Phase-contrast images of PC3 cells, PC3mock, and PC3CLU. Magnification: 20×. (b) Cellular proliferation was assessed in PC3mock (black square) and in PC3CLU (white square) clones at the indicated time after seeding by crystal violet assay. Error bars represent SD of the mean of three PC3mock and three PC3CLU clones. * p < 0.05 (the unpaired t-test vs. mock). (c) Cell cycle analysis was carried out in PC3mock and PC3CLU by FACS analysis. e graphs are representative of all the clones analyzed. e percentage of distribution of the cells in the cell cycle phases is reported in the table. Figure S3: expression of p65 in TRAMP and TRAMP/ CLUKO prostate tissues. Immunohistochemical staining of p65 in prostate tissues of 12-and 24-week-old (left and right panels, respectively) TRAMP and TRAMP/CLUKO mice. ree animals for each experimental group were examined. Images magnification 40×. (Supplementary Materials)