Silencing hsp25/hsp27 Gene Expression Augments Proteasome Activity and Increases CD8⁺ T-Cell–Mediated Tumor Killing and Memory Responses Free

The 25-kDa Hsp25 belongs to the family of small HSPs and is the murine homologue of human Hsp27, which was originally identified as an estrogen-responsive gene in breast cancer cells (1). Unlike the large HSPs, which function through ATP-dependent mechanisms, Hsp25/27 operates through ATP-independent mechanisms (2, 3). Importantly, elevated Hsp27 levels have been found in various tumors, including breast, prostate, gastric, uterine, ovarian, head and neck, and tumors, arising from the nervous system and urinary system (4). In estrogen receptor (ER)-α–positive benign neoplasia, elevated levels of Hsp27 have been shown to promote the progression to more malignant phenotypes (5). These studies were supported by findings that show that enhanced Hsp27 protein in breast cancer cells correlated well with increased anchorage-independent tumor growth (6), increased resistance to chemotherapeutic drugs (including cisplatin and doxorubicin), and increased metastatic potential in vitro (7–9). Together, these studies predict that elevated Hsp27 in breast cancer will give rise to aggressive disease that is refractory to treatment and so has poor prognosis (4). Indeed, elevated Hsp27 expression in tumors correlates with shorter disease-free survival and recurrence in node-negative breast cancer (10, 11), whereas the induction of Hsp27 following chemotherapy predicts poor prognosis and shorter disease-free survival (12). Currently, several selective Hsp27 inhibitors have reached clinical trials, including the Hsp27 inhibitor, OGX-427, which has completed phase I trials (clinicaltrials.gov - NCT00487786) and is now in phase II trials of castrate-resistant prostate cancer (clinicaltrials.gov - NCT01120470) and bladder cancer (clinicaltrials.gov - NCT00959868).

The inability of CD8⁺ T cells to recognize tumor-associated antigenic (TAA) peptides presented on MHC class I molecules remains a formidable barrier, limiting the success of immunotherapy (13). In normal cells, the proteasome system efficiently generates peptides from intracellular antigens, which are loaded onto MHC class I molecules for presentation to T cells (14). Within the proteasome system, the proteasome activator 28 (PA28) subunit, is a modulator of the proteasome-catalyzed generation of peptides presented via MHC class I molecules, and the selective increase in cellular levels of PA28α results in improved antigen presentation (15, 16). In addition, PA28 is essential for the recognition of epitopes on melanoma cells by specific cytotoxic T lymphocytes (CTL; ref. 17) and may alter the quality of products generated by proteasome cleavage (18, 19). The overexpression of the PA28α/β subunit enhanced MHC class I–restricted presentation of 2 viral epitopes and purified PA28α and PA28β subunit–accelerated T-cell epitope generation by the 20S proteasome in vitro (15). Taken together, these studies suggest that an efficient, well-functioning proteasome system is beneficial for specific CD8⁺ CTL recognition of tumors and ultimately cytolysis (for review, see ref. 20).

In this study, we showed that short-term silencing of Hsp25 or Hsp27 using siRNA or permanent silencing of Hsp25 using lentivirus RNA interference technology enhanced proteasome activity via increased PA28α subunit expression, abrogated metastatic potential, induced the regression of established breast cancer cells via tumor-specific CD8⁺ T cells, and stimulated long-lasting memory responses.

4T1 cells are a highly metastatic breast cancer cell line derived from a spontaneously arising BALB/c mammary tumor. BNL 1MEA.7R.1 (BNL) cells are a mouse-transformed hepatocellular carcinoma (HCC) cell line derived from BALB/c mice. MCF7 cells are a nonaggressive human breast cancer cell line. MDA-MB-232 cells are a highly aggressive human breast cancer cell line. All breast cancer cells were purchased directly from the American Type Cell Culture (ATCC), which routinely conducts cell line characterization. All breast cancer cells were passaged in our laboratory for not more than 6 months after receiving them from ATCC. 4T1 cells were maintained in monolayer cultures in Dulbecco's Modified Eagle's Medium (DMEM; Cellgro) supplemented with 10% FBS and antibiotics/antimycotics (Invitrogen Life Technologies). BNL cells were maintained in DMEM supplemented with 10% heat-inactivated FBS, antibiotics, and antimycotics (Gibco BRL/Life Technologies, Inc.). MCF7 cells were maintained in minimum essential medium (Eagle) with 2 mmol/L l-glutamine and Earle's balanced salt solution adjusted to contain 1.5 g/L sodium bicarbonate, 0.1 mmol/L nonessential amino acids, and 1 mmol/L sodium pyruvate and supplemented with 0.01 mg/mL bovine insulin and 10% FBS. MDA-MB-231 cells were maintained in ATCC-formulated Leibovitz L-15 medium supplemented with 10% heat-inactivated FBS, antibiotics, and antimycotics (Gibco BRL/Life Technologies, Inc.). All breast cancer cells were maintained in an incubator adjusted to 37°C with humidified atmosphere and 5% CO₂.

An HIV-derived 3-plasmid system was kindly provided by D. Trono (Department of Microbiology and Molecular Medicine, University of Geneva, Geneva, Switzerland). The packaging plasmid psPAX2 encodes HIV-1 gag and pol genes. The envelope plasmid pMD2G encodes vesicular stomatitis virus (VSV) G envelope protein. The transfer vector pLVTHM encodes elongation factor-1α (EF-1α) and H1 promoters, GFP as fluorescent marker, and woodchuck hepatitis posttranscriptional regulatory element (WPRE). The siRNAs for hsp25 were designed using the standard web-based program (Invitrogen). The siRNAs were then converted into small hairpin RNA (shRNA) according to the web-based program from Promega. The shRNAs contained restriction overhangs such as MluI and ClaI and standard hairpin loop structure TTCAAGAGA and a Pol III termination signal which consists of a run of at least 4Ts (TTTTTT). Oligos were synthesized at the minimal synthesis and purification scales. Then, the complimentary oligos were annealed according to the manufacturer's instruction (Invitrogen). To construct vector plasmids, the plasmid pLVTHM was digested with MluI and ClaI and ligated to an oligonucleotide pair containing Hsp25shRNA or control shRNA carrying MluI and ClaI restriction overhangs and transformed into Max Stbl2 competent cells. The positive clones were identified by digesting the control pLVTHM vector and the vector containing Hsp25shRNA inserts using MluI and XbaI enzymes. Positive clones were also identified by DNA sequencing.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

See supplementary information section.

For comparisons between groups, Dunn multiple comparison tests and the Student t test and one-way ANOVA were used in this study (P < 0.001 was considered significant).

We used a lentivirus-based vector (pLVTHM) that expresses RNA interference inducing the 25-kDa Hsp25 shRNA (Hsp25shRNA) under the control of the H1 promoter (Fig. 1A). This bicistronic vector was engineered to coexpress enhanced GFP (eGFP) as a reporter gene under the tight control of the EF-1α) promoter, permitting transduced/infected target cells to be tracked using in vivo imaging. Stable silencing of hsp25 gene expression in 4T1 tumor cells was achieved by subcloning the Hsp25shRNA cassette into pLVTHM, a self-inactivating (SIN) lentiviral vector using MluI and ClaI restriction sites (4T1-Hsp25shRNA hairpin loop sequence; Fig. 1A). We also constructed control/scrambled shRNA-containing lentiviral vector which does not have sequence homology to the mouse genome (4T1-controlshRNA hairpin loop sequence; Fig. 1A). These constructs were introduced into 293FT viral packaging cells to make lentivirus. The concentrated lentivirus preparation was used to infect target 4T1 breast adenocarcinoma cells. The resulting GFP expression was assessed 4 days postinfection by flow cytometry and further enriched for only highly expressing GFP-positive cells. The resulting sorted 4T1-Hsp25shRNA cells were 96.7% positive for GFP (Fig. 1B). The high GFP expression exhibited by both 4T1-controlshRNA and Hsp25shRNA stably transfected cells remained high even after 6 weeks of culture (Fig. 1C). We confirmed that high GFP expression in 4T1-Hsp25shRNA cells corresponded to efficient silencing of Hsp25 protein expression consistently by more than 98% after 6 to 8 weeks in vitro cell culture, as compared with the expression of Hsp25 in 4T1-controlshRNA cells (Fig. 1D). To negate the possibility that the stable transfection of only one cell line with Hsp25shRNA and selection after 6 to 8 weeks of culture might lead to the selection of a particular phenotype, additional experiments after short-term treatment with siRNA were also carried out. We showed that transient transfection of human breast cancer cells MCF7 and MDA-MB-231 with Hsp27-siRNA resulted in effective suppression of Hsp27 protein expression as compared with cells transfected with control-siRNA (Ctrl-siRNA; Fig. 1D). Gain-of-function experiments using Hsp25-plasmids in 4T1 cells and Hsp27-plasmids in MCF7 cells and MDA-MB-231 cells resulted in the significant increase in Hsp25 and Hsp27 expression, as compared with control-plasmid (Ctrl-plasmid; Fig. 1D).

The uncontrollable growth of tumors and their ability to metastasize and invade distant organs is the hallmark of aggressive forms of cancer. We showed that silencing Hsp25 protein expression dramatically increased cell death in tumor cells and the ability of the tumor to migrate in vitro. There were consistently lesser cells recovered from culture plates of 4T1-Hsp25shRNA cells as than 4T1-controlshRNA or wild-type 4T1 (4T1-wt) cells (Supplementary Fig. S1A, left). Results of cell death measurements suggest the low cell count is due to a concomitant increase in the percentage of cell death in 4T1-Hsp25shRNA cells as compared with 4T1-controlshRNA or 4T1-wt (Supplementary Fig. S1A, right). We showed that Hsp25shRNA treatment adversely affects the directional cell migration of 4T1 cells in vitro, approximately to the same extent as serum starvation, as judged by the wound-healing experiment (Supplementary Fig. S1B). These results correlated well with the inability of 4T1-Hsp25shRNA cells to invade extracellular matrix in vitro as compared with 4T1-controlshRNA cells (Supplementary Fig. S1C).

We validated these data in experiments in which mouse and human breast cancer cells were transiently transfected with siRNA directed against Hsp25 or Hsp27, respectively. We showed that silencing Hsp25/27 effectively suppressed the expression of proteins known to be important in cancer functions including; ABCB4 (growth, survival, proliferation), ENO1 (growth, differentiation, colony formation), HSP90AB1 (survival, migration, negative regulation of proteasomal ubiquitin-dependent protein catabolic process), HSPA8 (apoptosis, ubiquitination, cell-cycle progression, survival, cell viability), lactate dehydrogenase (LDH; growth, transformation, proliferation), MCM3 (growth, migration), and MIF (proliferation, activation, migration). In addition, there was a concomitant decrease in the expression of proteins associated with cell movement [e.g., CAP (morphogenesis, migration, organization, morphology, depolymerization, elongation), EMTH (proliferation, quantity, development, activation), FLNA (cell spreading, motility, formation, migration), FLNB (migration, binding, anchoring), HSPD1 (chemotaxis, polarity, migration), MYH9 (adhesion, morphogenesis, movement, retraction, migration, formation), S100A6 (proliferation, survival, invasion, focus formation), and TWIST1 (survival, migration, motility, adhesion, proliferation]. On the other hand, there was an increase in the expression of proteins known to be associated with cell death [e.g., EEF2 (autophagy, cell death, apoptosis), FASN (apoptosis, G₁ phase, G₂ phase, senescence), GAPDH (apoptosis, cell death, caspase-independent cell death), HNRNPA1 (apoptosis, nuclear export, alternative splicing, stress response), KRT18 (apoptosis, cell death, collapse, fragmentation), YWHAE (apoptosis, cell-cycle progression, mitosis, transmembrane potential)] and antigen presentation [e.g., PSMB5 (degradation, proteolysis), TAP1 (autophosphorylation, antigen presentation), TAPBP (substrate selection, ubiquitination, selection, immune response); Supplementary Fig. S1D]. Together, these results indicate that silencing the expression of Hsp25/27 in breast adenocarcinoma tumors significantly interferes with its ability to survive and migrate in vitro.

To obtain an integrative understanding of the effect of Hsp25 silencing on the global protein profile of 4T1 breast adenocarcinoma cells, we used 2-dimensional (2D) SDS-PAGE combined with liquid chromatography/tandem mass spectrometry (LC/MS-MS) techniques to compare the protein profiles between controlshRNA and Hsp25shRNA stably transfected 4T1 cells. Three unique spots were selected from 4T1-Hsp25shRNA cells (Fig. 2A, right) which were absent in 4T1-controlshRNA cells (Fig. 2A, left). Further characterization using LC/MS-MS and bioinformatics revealed that the unique proteins were Ng, Ng-dimethylarginine dimethylaminohydrolase-2 and prohibitin (Table 1, square), PA28α, PA28γ subunits, and mitochondrial ribosomal protein L46 (Table 1, circle). Proteins expressed within the triangle could not be identified, possibly due to the highly glycosidic nature of the proteins (Table 1, triangle). Because of the obvious relevance to tumor growth and metastasis, we chose to validate prohibitin and PA28α by real-time PCR and Western blot analysis. We showed that silencing the hsp25 gene increased prohibitin mRNA expression by 3-fold (Fig. 2B). The mRNA expression levels correlated well with a 2.5-fold increase in prohibitin protein expression as judged by Western blot analysis (Fig. 2C). Similar increases were observed for PA28α mRNA expression which was upregulated by 1.5-fold, as judged by real-time PCR (Fig. 3A) and by 2-fold as judged by Western blot analysis (Fig. 3B, left), as compared with respective controls. There was no significant alteration in the PA28γ subunit by measuring protein and RNA levels (data not shown). Transient transfection experiments showed that treatment of 4T1 cells transfected with Hsp25-siRNA dose dependently increased PA28α expression (Fig. 3B). Gain-of-function experiments to increase Hsp25 expression in 4T1 cells (using Hsp25-plasmids) showed that the overexpression of Hsp25 dramatically suppresses PA28α expression (Fig. 3B). To further validate the finding that silencing Hsp25 protein expression increases proteasome activity, we measured the chymotrypsin-like activity of 20S proteasome in 4T1-controlshRNA and 4T1-Hsp25shRNA cell extracts. We showed that 4T1-Hsp25shRNA cells showed 50% greater proteasome activity than 4T1-controlshRNA tumor cells (Fig. 3C). Because chymotrypsin-like activity of 20S proteasome uses Suc-LLVY-AMC as the substrate for proteasome activity measurement, we also measured trypsin-like and caspase-like proteasome activities in mouse and human breast cancer cells. We showed that transient transfection using Hsp25-siRNA and Hsp27-siRNA effectively increases chymotrypsin-like, caspase-like, and trypsin-like proteasome activities in 4T1, MCF7, and MDA-MD-231 cells, as compared with respective control-siRNA's (Ctrl-siRNA; Table 2). The overexpression of Hsp25 and Hsp27 (using Hsp25-plasmid and Hsp27-plasmid, respectively) suppressed chymotrypsin-like, caspase-like, and trypsin-like proteasome activities in 4T1, MCF7, and MDA-MD-231 cells, as compared with respective control plasmids (Ctrl-plasmids; Table 2). Taken together, these results indicate that silencing of Hsp25 or Hsp27 enhances proteasome function via PA28α.

To determine the consequence of lentivirus-mediated hsp25 gene silencing in vivo, 4T1-controlshRNA and 4T1-Hsp25shRNA cells were injected subcutaneously into the mammary pad of female BALB/c mice. As early as 7 days after tumor cells injection (TCI), tumors could be visualized growing in the mammary pad of all mice using the Maestro In Vivo Imaging System (CRI). We showed that 13 days post-TCI, the GFP signal from 4T1-controlshRNA tumors were significantly higher than GFP signal from 4T1-Hsp25shRNA tumors (Fig. 4A). Starting 7 days post-TCI, there was a steady regression in GFP signal from 4T1-Hsp25shRNA tumors. By day 25 post-TCI, there was no detectable GFP signal in any mouse bearing 4T1-Hsp25shRNA tumors (Fig. 4A). Efficient Hsp25 silencing (>95%) could still be shown in 4T1-Hsp25shRNA tumors before they completely disappeared (data not shown). To negate the possibility that antitumor responses were directed against the GFP protein instead of unknown “tumor-associated antigens” that are better processed as a consequence of silencing Hsp25 expression in tumor cells, tumor growth experiments using eGFP positive(+) and negative(−) 4T1-Hsp25shRNA and 4T1-controlshRNA and wild-type 4T1 cells were carried out. We showed that eGFP did not significantly alter tumor growth curves of BALB/c wild-type mice injected with eGFP positive(+) or negative(−) 4T1-controlshRNA cells (Fig. 4B, left). Experiments carried out in BALB/c nude mice further revealed that the tumor volume of 4T1-Hsp25shRNA cells (with or without GFP) is significantly smaller than 4T1-controlshRNA (with or without GFP) or 4T1-wt cells (Fig. 4B, right). An additional observation observed in the BALB/c nude mice experiments was that whereas the injected 4T1-controlshRNA and 4T1-wt cells rapidly metastasize to distant organs including lungs, liver, and brain, injected 4T1-Hsp25shRNA cells did not metastasize to these organs (data not shown), suggesting that although a competent immune system (possibly CD8⁺ CTL) may be one of the mechanisms associated with downregulation of Hsp25, there might be other mechanisms. At the end of the experiment (day 34 post-TCI), gross pathology of multiple organs, including lungs, brain, bone and liver, showed an absence of tumor metastasis in mice injected with 4T1-Hsp25shRNA but not 4T1-controlshRNA mice. Hematoxylin and eosin staining of lungs from mice injected with 4T1-controlshRNA revealed micrometastasis in lung tissues (Fig. 4C, left). In contrast, lungs of mice injected with 4T1-Hsp25shRNA had no visible micrometastasis (Fig. 4C, right). To confirm that micrometastasis undetectable by light microscopy did not exist in 4T1-Hsp25shRNA–injected mice, we conducted clonogenicity assays on lung tissues in complete media containing 6-thioguanine. 4T1 breast adenocarcinoma cells are resistant to 6-thioguanine; however, all other contaminating cells will be destroyed. Mice injected with the 4T1-controlshRNA cells exhibited large numbers of colonies at all dilutions, reflecting robust metastasis of tumors to the lungs (Fig. 4D). In contrast, no colonies were observed in dishes plated with lung tissue harvested from mice injected with 4T1-Hsp25shRNA cells (Fig. 4D). Taken together, these data suggest that permanent silencing of Hsp25 results in tumor regression and inhibition of metastasis in vivo.

To determine the nature of the cells responsible for tumor regression following silencing of Hsp25 expression in 4T1 breast adenocarcinoma cells, prior to TCI, we conducted in vivo depletion of cells known to play an important role in tumor regression. Here, we showed that in vivo depletion of CD8⁺ CTL prior to injection of 4T1-controlshRNA cells drastically increased tumor growth rate, and by day 34 post-TCI, the size of the tumors were approximately 10 times larger than mice injected with PBS only (Fig. 5A, left). The in vivo depletion of CD4⁺ T cells did not significantly alter tumor growth rate or tumor volume in mice injected with 4T1-controlshRNA cells (Fig. 5A, left). Unexpectedly, using similar mice, the in vivo depletion of natural killer (NK) cells using the 5E6 monoclonal antibody induced complete tumor regression (Fig. 5A, left). In mice injected with 4T1-Hsp25shRNA cells, no tumor growth was seen in any of the mice by the end of the experiment (Fig. 5A, right). As expected, the in vivo depletion of CD8⁺ T cells and NK cells prior to injection with 4T1-Hsp25shRNA cells resulted in tumor growth. Similar depletion of CD4⁺ T cells initially resulted in increased tumor growth, followed by tumor regression (Fig. 5A, right). Interestingly, the in vivo depletion of CD8⁺ T cells prior to injection with 4T1-Hsp25shRNA cells resulted in increased tumor growth (Fig. 5A, right). Gross pathology of lung, brain, and bone did not reveal any signs of metastasis to the lungs (data not shown). Similarly, injection of 4T1-Hsp25shRNA cells into the breast pad of BALB/c nude mice resulted in tumor growth without metastasis (data not shown).

To confirm that CD8⁺ T cells mediated the enhanced cytolytic effects after silencing Hsp25, reactive CD8⁺ T cells were harvested from the spleen of mice which had been injected with 4T1-Hsp25shRNA cells and were tumor free (days 21–28 post-TCI) and the specific T-cell cytotoxicity measured against 4T1-controlRNA target cells ex vivo. Extracted splenic CD8⁺ T cells were enriched using negative selection by magnetic beads and consistently exhibited greater than 95% purity, as judged by flow cytometry (Fig. 5B). Experiments were next carried out to negate the possibility that the tumor-associated response was directed against GFP protein. We showed that reactive CD8⁺ T, but not CD8⁻ T cells (non-CD8⁺ T cells), effector cells harvested from the spleen of mice injected with 4T1-Hsp25shRNA cells exhibited potent-specific lysis against 4T1-controlshRNA eGFP-positive and eGFP-negative targets with similar activity (Fig. 5C). CD8⁺ cells did not exhibit significant lytic activity against BNL, which served as an irrelevant target (Fig. 5C). As expected, both CD8⁺ and CD8⁻ T cells from mice injected with 4T1-controlshRNA cells did not mediate significant lysis above baseline levels against 4T1-controlshRNA targets.

To determine whether 4T1-Hsp25shRNA–reactive CD8⁺ T cells can rescue mice injected with 4T1-controlshRNA cells, 4T1-Hsp25shRNA–reactive CD8⁺ T cells were adoptively transferred into 4T1-controlshRNA tumor–bearing mice. As predicted, the adoptive transfer of 4T1-Hsp25shRNA–reactive CD8⁺ T cells into 4T1-controlshRNA tumor–bearing mice induced significant tumor regression starting by day 17 post-TCI, and by day 28, there was no detectable tumor growth (Fig. 5D). In contrast, 4T1-controlshRNA tumor–bearing mice adoptively transferred with CD8⁻ T-cell fraction were not protected and mice rapidly developed tumors (Fig. 5D) and metastasis (data not shown).

To prove that the improvement in antigen presentation is due, in part, to silencing of Hsp25 expression, we used the in vitro cross-presentation assay. Bone marrow-derived dendritic cells (BMDC) were recovered from female C57BL/6 (H2^b) and BALB/c (H2^d) mice and treated with ovalbumin (OVA) during the culture process. BMDCs were then transfected with either Hsp25-siRNA or negative control-siRNA and fixed with paraformaldehyde and later admixed with S8L peptide–specific T-cell hybridoma, B3Z cells. We showed that B3Z cells released significantly more IFN-γ when admixed with C57BL/6 (H2^b)-derived BMDCs in which Hsp25 has been silenced (Hsp25-siRNA) than control-siRNA–treated BMDCs (Fig. 5E; left). In addition, we showed that pretreatment of both Hsp25-siRNA–and control-siRNA–treated BMDCs with the specific proteasome inhibitor, MG-132, significantly reduced the concentration of released IFN-γ (Fig. 5E, left). Finally, we showed that BMDCs recovered from BALB/c mice which express H2^d did not release significant quantities of IFN-γ under similar conditions (Fig. 5E; right). To prove that 4T1-Hsp25shRNA generates memory responses, tumor-free immunocompetent female BALB/c mice were rechallenged with wild-type 4T1 (4T1-wt) or an irrelevant tumor, murine transformed HCC cells, BNL, 60 days after initial challenge with 4T1-Hsp25shRNA. We showed that rechallenge of 4T1-wt cells does not result in tumor growth, which is similar to mice injected with 4T1-Hsp25shRNA alone (Fig. 5F). However, rechallenge with BNL (after 4T1-Hsp25shRNA) resulted in tumor growth in a similar fashion to mice injected with BNL alone (Fig. 5F).

Mouse and human breast cancer cells express elevated levels of Hsp25 or Hsp27, respectively, which is effectively suppressed by permanently silencing Hsp25 or short-term silencing of Hsp25 and Hsp27 protein expression (Fig. 1D). This is important not only because elevated levels of Hsp27 in breast cancer give rise to aggressive disease and poor prognosis (4) but also because elevated Hsp27 levels have been reported to confer tumor protection against bortezomib-induced cell death (21). Bortezomib is characterized as a reversible proteasome inhibitor, with potent anticancer effects against multiple myeloma (22). Bortezomib was shown to effectively induce apoptotic cell death in DHL6 lymphoma cells (do not express significant Hsp27) but not in DHL4 lymphoma cells (expressing high basal levels of Hsp27). Blocking the elevated Hsp27 expression in DHL4 lymphoma cells using antisense against Hsp27 restored sensitivity to bortezomib. These authors concluded that combining agents that suppress Hsp27 expression might provide a therapeutic advantage to overcome tumors that might be resistant to bortezomib treatment (21).

Our study shows that silencing Hsp25/27 effectively suppresses proteins known to be important in cancer functions in cells (Supplementary Fig. S1D). Because the working hypothesis for this study is that high Hsp25/27 expression represses proteasome activity, therefore gain-of-function studies were also conducted. Our results showed that increasing Hsp25 or Hsp27 expression using Hsp25- or Hsp27-plasmids effectively increases Hsp25 or Hsp27 expression, respectively (Fig. 1D). The biologic significance of increased Hsp25 or Hsp27 expression was a significant increase in dramatic inhibition of PA28α protein expression (Fig. 3B). We further showed that silencing Hsp25 or Hsp27 concomitantly increases proteasome chymotrypsin-like activity (Fig. 3C). Because chymotrypsin-like activity uses Suc-LLVY-AMC as the substrate for proteasome activity measurement and because the same substrate can be digested by calpains, we also measured trypsin-like and caspase-like proteasome activities. We showed that knockdown of Hsp25 or Hsp27 expression enhanced chymotrypsin-like proteasome activity, caspase-like proteasome activity, and trypsin-like proteasome activity in mouse and human breast cancer cell lines, respectively (Table 2). This is significant because Groettrup and colleagues reported that increased expression of PA28α results in marked enhancement of recognition by virus-specific cytotoxic T cells (15). In addition, an essential role for PA28 was described in the melanoma cell line, Mel-18a. These authors showed that recognition of TRP2-expressing melanoma cells by TRP2_360–368–specific CTLs directly correlated with the presence of PA28, and impaired epitope presentation on Mel-18a cells could be rescued by transfection of PA28-encoding plasmids (17). Here, we showed that IFN-γ release from B3Z (an S8L peptide–specific T-cell hybridoma) cells is greatly enhanced in OVA-treated BMDCs transfected with Hsp25-siRNA and recovered from female C57BL/6 (H2^b) mice but not BALB/c (H2^d) mice (Fig. 5E). The role of the proteasome was further shown in experiments in which pretreatment with proteasome inhibitor, MG-132, dramatically inhibited IFN-γ release (Fig. 5E). We further showed that Hsp25-plasmid bearing 4T1 tumors grow approximately 3 times larger than Ctrl-plasmid bearing 4T1 tumors or 4T1-wt tumors or 4T1-controlshRNA tumors (Fig. 5F). Taken together, our data suggest that Hsp25/Hsp27 decreases ubiquitination and proteasomal degradation. This is in agreement with recent findings by other scientists (23–27). However, there are other authors who have previously shown that Hsp27 enhances ubiquitination and proteasomal degradation (28–31). The reason for this discrepancy is currently unknown. However, studies using antibodies to pull-down various components of the proteasome complex followed by mass spectrometry and bioinformatics analysis in response to high and low Hsp27 expression are currently underway in our laboratory (Nagaraja et al., manuscript in preparation) and should shed more light on this controversy.

Our additional working hypothesis that high Hsp25 expression represses proteasome activity in turn downregulates peptide loading onto MHC class I molecules for effective recognition by CD8⁺ T cells is substantiated by data presented in this study. The first suggestion for a role of CD8⁺ T cells was obtained in experiments in which BALB/c nude mice (which have defective CD8⁺ and CD4⁺ T cells) were injected with 4T1-controlshRNA cells that grew approximately 3 times larger than similar cells injected into BALB/c wild-type mice (Fig. 4B). The role of CD8⁺ T cells was further substantiated in experiments in which the in vivo depletion of CD8⁺ T cells using anti-CD8 (Ly-2) resulted in larger tumors in mice injected with 4T1-controlshRNA and 4T1-Hsp25shRNA cells than in animals injected with isotype control (Fig. 5A). Interestingly, although tumors grew significantly larger in the breast pad of 4T1-Hsp25shRNA–bearing mice in the absence of CD8⁺ T cells, there were no lung metastases in these mice, as compared with 4T1-controlshRNA–bearing mice (data not shown). Although the possibility exists that the reason for the lack of pulmonary metastasis in 4T1-Hsp25shRNA–bearing mice in the absence of CD8⁺ T cells is only due to a decrease in primary tumor growth due to enhanced cell death (Supplementary Fig. S1A) and a decrease in proteins associated with cell death (Supplementary Fig. S1D), our in vitro data showing that silencing Hsp25 also inhibits migration (Supplementary Fig. S1B), invasion (Supplementary Fig. S1C), and downregulated proteins involved in cell movement (Supplementary Fig. S1D), as well as published data showing that Hsp25 is critical for maintaining the integrity of cytoskeleton (32), suggest that in the absence of specific CD8⁺ CTL–mediated killing, the tumor is still incapable of leaving the primary tumor foci. Data showing that depletion of NK cells in vivo resulted in enhanced antitumor killing and complete tumor regression were initially confounding (Fig. 5A, green lines). However, further examination of the monoclonal antibody used to deplete NK cells in vivo revealed it to be the 5E6 Mab F(ab′)₂ fragment, which reacts with Ly49C, an NK-inhibitory receptor expressed on subsets of NK cells and NK1.1⁺ or DX5⁺ T cells in BALB/c mice (33). Studies by Koh and colleagues showed that NK cell–mediated antitumor effector functions are increased against syngeneic tumors in vitro and in vivo by blockade of the Ly49C and Ly49I inhibitory receptors using the 5E6 monoclonal antibody (34). In addition, the ability of adoptively transferred 4T1-Hsp25shRNA–reactive CD8⁺ cells, but not CD8⁻ T cells (non-CD8⁺ T cells), to rescue mice injected with 4T1-controlshRNA tumors (Fig. 5D), which has been shown to succumb from the tumor burden (Figs. 4A and 4B and 5A), suggest that silencing Hsp25 improved the quality and/or quantity of peptides recognized by CD8⁺ T cells via a mechanism dependent on enhanced proteasome activity. Conclusive proof that silencing Hsp25 improves the quantity and/or quality of peptides presented onto MHC class I for specific CD8⁺ T-lymphocyte recognition was obtained indirectly using the in vitro cross-presentation assay. Here, we showed that silencing Hsp25 enhances recognition of B3Z (an S8L peptide–specific T-cell hybridoma) cells for OVA-treated BMDCs which have been transfected with Hsp25-siRNA and recovered from female C57BL/6 (H2^b) mice but not BALB/c (H2^d) mice (Fig. 5E).

The central role of the proteasome is shown in experiments in which IFN-γ release in these cells was drastically blunted by pretreatment with proteasome inhibitor, MG-132 (Fig. 5E). The possibility that silencing Hsp25/27 improves antigen presentation is further substantiated by data showing that Hsp25-siRNA or Hsp27-siRNA treatment of breast cancer cells significantly increased genes important in antigen presentation (Supplementary Fig. S1D). Our studies further showed that silencing Hsp25/27 generates T-cell memory responses. We showed that immunocompetent female BALB/c mice rendered tumor-free by injection with 4T1-Hsp25shRNA can be rechallenged with 4T1-wt or 4T1controlshRNA or 4T1-Ctrl-plasmid–containing cells 60 days after initial challenge without tumor growth (Fig. 5F). However, rechallenge with an irrelevant tumor, BNL cells (a murine-transformed HCC cell line), resulted in tumor growth. The potent efficacious antitumor activities observed by silencing Hsp25 strongly suggest that other antitumor mechanisms have also been activated. This is supported by reports showing that the human homologue of the mouse Hsp25, Hsp27 plays an essential role in (i) stabilizing actin filaments, a structural protein important for maintaining the integrity of cytoskeleton (32), (ii) cell-cycle progression and proliferation (35), and (iii) apoptosis via caspase-3 activation (36, 37). Data from the manuscript suggest that enhanced tumor recognition by CD8⁺ CTLs may be one of the mechanisms associated with downregulation of Hsp25/27 but does not rule out other pathways. In fact, several studies raise the possibility that the down regulation of Hsp25/27 might activate CD8⁺ CTL-independent mechanisms associated with HSPs, including findings that the surface expression of Hsp70 in metastatic melanoma (38), acute myeloid leukemia (39), and head and neck cancer (40) stimulates specific NK cell-mediated cytolytic functions are some recent examples. In addition, the development of a Hsp70 peptide which stimulates NK cell-mediated killing of leukemic blast cells (41, 42), and the demonstration that NK cell-mediated targeting of membrane Hsp70 on tumors can be greatly enhanced after treatment with the cmHsp70.1 monoclonal antibody (43, 44), support the possibility that the down regulation of Hsp25/27 might activate these CD8⁺ CTL-independent mechanisms.

Taken together, these studies show that increasing proteasome activity of 4T1 breast adenocarcinoma cells by lentivirus-mediated hsp25 gene silencing or by the transient transfection of siRNA directed against Hsp25 in mice or Hsp27 in human breast cancer cell lines increased specific CD8⁺ T-lymphocyte tumor killing and enhanced memory responses. The hypothetical model of findings is summarized schematically in Fig. 6. Our data have obvious clinical applications in light of recent successful clinical trials with the proteasome inhibitors. These clinical trials do not contradict our findings because bortezomib resistance is observed in tumors with elevated Hsp27 expression (21). This would suggest that screening for patients on the basis of Hsp27 expression and proteasome activity might add benefit. In addition, this would suggest that combining agents that suppress Hsp27 expression might provide a therapeutic advantage for patients with bortezomib resistance. Clinical trials on the application of siRNA for the treatment of various diseases are underway and early results show it to be safe, with no patients experiencing any serious adverse events (45–47). However, there are current limitations to RNA interference technology primarily in terms of possible “off-target” or nonspecific effects. Off-target effects occur when an siRNA is processed by the RNA-induced silencing complex (RISC) and downregulates an unintended target(s) with similar sequence. An in-depth understanding of how siRNA is metabolized and determining whether intermediate metabolites are toxic or harmful needs to be addressed. Efforts to recognize and circumvent off-target effects for enhanced target identification and to improve the therapeutic application of RNA interference therapy in the clinic are now underway (reviewed in ref. 48).

No potential conflicts of interest were disclosed.

The authors thank Dr. Dider Trono (Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland) for providing the lentivirus plasmids, Dr. Jahangir Kabir for helpful discussions, Dr. Hongying Zheng for expert flow cytometric assistance, and the Scott & White Proteomics Core Facility.

This work was supported in part by Scott & White Memorial Hospital and Clinic Research Advancement Awards (to G.M. Nagaraja and P. Kaur); and the U.S. NIH (RO1CA91889), Scott & White Memorial Hospital and Clinic, the Texas A&M Health Science Center College of Medicine, the Central Texas Veterans Health Administration, and an Endowment from the Cain Foundation (to A. Asea).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.