Gadd45a Functions as a Promoter or Suppressor of Breast Cancer Dependent on the Oncogenic Stress

Abstract

Gadd45a plays a pivotal role as a stress sensor that modulates cellular responses to various stress stimuli, including oncogenic stress. We reported that the stress sensor Gadd45a gene functions as a tumor suppressor in Ras-driven breast tumorigenesis via increasing JNK-mediated apoptosis and p38-mediated senescence. In contrast, here, we show that Gadd45a promotes Myc-driven breast cancer by negatively regulating MMP10 via GSK3 β/β-catenin signaling, resulting in increased tumor vascularization and growth. These novel findings indicate that Gadd45a function as either tumor promoter or suppressor is dependent on the oncogenic stress and is mediated via distinct signaling pathways. Collectively, these novel findings highlight the significance of the type of oncogenic alteration on how stress response genes function during initiation and progression of tumorigenesis. Since gadd45a is a target for BRCA1 and p53, these finding have implications regarding BRCA1/p53 tumor suppressor functions.

INTRODUCTION

Gadd45a plays a pivotal role as a stress sensor that modulates the cellular response to a variety of stress conditions, including genotoxic and oncogenic stress^1,2,3. Gadd45a mediates stress responses via interacting with and modulating the function of partner proteins that play key roles in cell cycle control^4,5,6, DNA repair^7,8 and cell survival^9,10,11,12.

We have recently reported that loss of Gadd45a significantly accelerated the onset of Ras-driven breast tumorigenesis, providing a novel model for Gadd45a in the suppression of breast tumor development¹³. Gadd45a suppressed Ras-driven tumor growth by increasing both JNK-mediated tumor cell apoptosis and p38-mediated tumor cell senescence.

C-Myc is a key regulator of cell proliferation, where deregulation of Myc expression contributes to the initiation and progression of tumorigenesis¹⁴. For nearly all cases of solid human tumors, the relative amount of Myc protein is increased in the tumor tissue compared to the surrounding normal tissue, which suggests that deregulated and/or elevated levels of Myc contributes to the tumorigenic phenotype¹⁴. Nearly 50% of all breast cancer tumors display significantly elevated levels of Myc protein^14,15,16. In light of these observations, we sought to extend our investigation of the role Gadd45a plays in modulating breast carcinogenesis to ask whether Gadd45a also modulates Myc-driven breast tumorigenesis. To this end, we took advantage of the established breast cancer prone MMTV-c-Myc transgenic mouse model, where deregulated Myc is under the control of the mouse mammary tumor virus (MMTV) promoter¹⁷. Gadd45a deficient mice and MMTV-c-Myc transgenic mice were interbred to generate MMTV-c-Myc transgenic mice that are either wildtype or null for Gadd45a. Using these mice, in conjunction with MMTV-Ras mice that are either wildtype or null for Gadd45a, we asked whether Gadd45a modulates Myc-driven tumorigenesis and how this compares to its involvement in Ras-driven breast carcinogenesis.

Intriguingly, in contrast to what was observed with oncogenic Ras, we demonstrate that Gadd45a promoted Myc-driven breast tumorigenesis. Furthermore, it is shown that the mechanism by which Gadd45a promotes Myc driven breast tumorigenesis differs radically from the mechanism employed to suppress Ras-driven breast carcinogenicity. Collectively, these novel observations provide a novel paradigm indicating that, depending on the nature of the oncogenic stress, Gadd45a may function as an either tumor suppressor or tumor promoter by employing different effector pathways.

EXPERIMENTAL PROCEDURES

Mice

MMTV-Myc transgenic mice (FVB genetic background) were obtained from Charles River Laboratories (Wilmington, MA, USA). Gadd45a−/− mice (C57Bl/6 x 129Sv background) were graciously provided by Albert Fornace. Offspring from interbreeding Gadd45a−/− and MMTV-Myc mice were generated as littermates from common matings so all animals were maintained in a mixed genetic background. Offspring from crosses between MMTV-Myc and Gadd45a−/− mice were screened by PCR for their Myc and Gadd45a status. At the same time, MMTV-Ras mice that differed in their Gadd45a status were also generated as published in Tront et al.¹³. At the time of weaning, genomic DNA was isolated from a tail clipping by standard procedures for PCR analysis. Primers for the detection of MMTV-Myc were 5′-CCCAGGTGATAGTCCCTTCACATC-3′ (5′ sense) and 5′-GAAAAGTGCCACCTGACGTCTAAGA -3′ (3′ antisense). To assess Gadd45a status, PCR reactions using three primers allowed for simultaneous detection of the wildtype and mutant Gadd45a allele. These primers consisted of a 5′ upstream primer (5′-CACCTCTGCTTACCTCTGCACAAC -3′), a common 3′ antisense primer (5′-CCAGAAGACCTAGACAGCACGGTT -3′) and a neo specific primer (5′-AAGCGCATGCTCCAGACTGCCTT -3′). Reactions were run for 37 cycles of 94°C for 1 min, 63°C for 14 sec, and 72° for 12 sec.

Tumor Formation and Onset

Four-week old female mice from all genotypes were observed twice weekly for the formation of visible tumor masses. Upon detection of a mass, the tumor growth properties where monitored every other day for 13 days or until the general health of the animal was compromised, at which time the mouse was sacrificed in accordance with Temple University and NIH guidelines. Tumor measurements were taken with hand calipers to evaluate tumor volume (calculated Tumor Volume (mm³) = (W² X L), where W is width and L is length). Tumor growth curves were generated by plotting the average daily tumor growth against time. Tumor onset was plotted using a Kaplan-Meier Survival curve. Differences between Kaplan-Meier Survival cures were determined using a Mantel-Cox Log Rank statistical test. Differences in tumor incidence were determined by the chi-square test.

Western Blotting

75ug of protein extract was resolved on an SDS-Polyacrylamide gel, followed by transfer to an Immobilon-P membrane. Following blocking, the membrane was incubated with primary antibody overnight and developed using chemiluminescence. (Gadd45a antibody, Santa Cruz Biotechnology).

Histological Evaluation of Mouse tumors

Tumor samples were fixed in 10% buffered formalin then embedded in paraffin. The sections were then stained with hematoxylin and eosin to examine histological differences (University of Pennsylvania Core Histology Facility, Philadelphia, PA). Ten tumors from each genotype were examined for histological characteristics.

Apoptosis

Tumor samples were fixed in 10% buffered formalin then embedded in paraffin for sectioning. Tumor tissue sections were In situ labeled for apoptotic cells using the Apo Alert DNA fragmentation Assay Kit (BD Biosciences, Franklin Lakes, NJ). Cells were analyzed using light microscopy. Necrotic regions of the tumor were avoided. Using a 10X450 field range, the number of TUNEL-positive stained cells and the total number of PI stained cells was determined with Image J photo program. Percent apoptosis was calculated by dividing the total number of positive cells by the total number of cells. A minimum of 5 samples per genotype were analyzed. Differences in percent apoptosis between genotypes were evaluated using the Student t test.

Immunohistochemistry

Paraffin embedded tissue slides were deparaffinized, rehydrated and subjected to antigen unmasking by sodium citrate (pH 6.0) for 30 minutes at a sub-boiling temperature. Endogenous peroxidase activity was blocked by incubation in 3% hydrogen peroxide for 10 minutes. Sections were blocked with 5% serum for one hour at room temperature, followed by incubation with primary antibody overnight at 4°C. (Phospho-Jnk (9251), Phospho-p38 (4631), GSK3β (9315)- Cell Signaling Technology; Beta-Galactosidase (ab616), MMP10 (ab4045), CD31 (28365), CD105 (27422), Gadd45a (ab33173) – Abcam; β-catenin (610154) – BD Biosciences). Sections were incubated with a peroxidase-conjugated secondary antibody for 30min at room temperature, followed by treatment with ABC reagent (Vector Laboratories) for 30 min. Sections were stained with 3,3′-diaminobenzidine substrate and counterstained with hematoxylin. For CD31 & CD105 double stating, equal amounts of both antibodies were mix and used for the overnight staining. For all immunohistochemical analysis, a minimum of 5 samples from each genotype were analyzed blindly for each analysis. Samples were analyzed in triplicate. The number of positive stained cells and the total number cells was determined with Image J photo software, using a 10X450 field range. Differences between genotypes were evaluated using the Student t test.

Microarray Analysis

Five micrograms of total RNA from tumor samples from the various genotypes were used as a template for cDNA synthesis. cDNA was labeled with biotin-dUTP using AmpoLabeling-LRP kit (SuperArray Bioscience). The cDNA probe was applied to prehybridized Mouse Signal Transduction in Cancer Gene Array (MM-044) membranes. The hybridization was done at 60° for 12 hours. After washing, the membranes were blocked and treated with alkaline phosphatase-conjugated streptavidin and exposed to alkaline phosphatase chemiluminescent substrate. The membranes were exposed to X-ray film and the spots were analyzed using the GEArray Expression Analysis Suite Software.

EMSA

Searching GenBank, we identified a potential TCF binding site in the promoter of MMP10, which matched the consensus sequence 5′A/T A/T CAAG-3′. The following double stranded oligonucleotide was used as a probe: TCF: 5′-ATA TAT TCA AAG GAC CCA GGT; TCF-m: 5′-ATA TAG CCA AAG GAC CCA GGT. The probe was end-labeled and used in a reaction with nuclear extracts from SW480 colon cancer cells. Samples were incubated for 20 minutes at room temperature, antibody was added and the samples were incubated for an additional 20 minutes (Anti- β-catenin, Transduction Laboratories; Anti-Myc, Santa Cruz Biotechnologies). For competition, 100X of unlabeled probe was used.

Tumor Transplantation & siRNA Treatment

Myc+Gadd45a−/− mammary tumors were excised 14 days after first visualization. The tumor was washed in PBS. The tumor was minced manually and then incubated with dissociation medium (DMEM, Hepes, BSA, Insulin, Hydrocortisone and Collagenase) for 2 hours at 37°. Red Blood cells were lysed, followed by filter sterilization and counting. 5×10⁶ cells were mixed with matrigel and injected into the number 4 mammary fat pad of Gadd45a null mice. Starting at day 0, mice were intratumorally injected with siRNA to MMP10 or control prepared using Invivofectamine (Invitrogen) every four days. Mice were monitored every four days for tumor growth.

RESULTS

Gadd45a is Up-regulated During Myc-Driven Breast Carcinogenesis, Resulting in an Acceleration of Myc-Driven Mammary Tumorigenesis

Our working hypothesis was that Gadd45a is a stress sensor protein, which is up-regulated by oncogenic stress during breast carcinogenesis and functions to modulate tumor development. To assess the validity of our hypothesis, Gadd45a expression was examined in normal and tumor mammary tissue obtained from both MMTV-Myc and MMTV-Ras mice that are either wildtype or null for Gadd45a expression. For comparison, Gadd45a expression was also assessed in non-mammary tissue (i.e., spleen tissue).

Gadd45a expression was undetectable in normal mammary and spleen tissue obtained from all genotypes (Figure 1), whereas Gadd45a expression was observed in comparable breast tumor tissue obtained from MMTV-Myc (Figure 1A) and MMTV-Ras (Figure 1B) mice. Taken together, these data, in conjunction with our previous findings¹³, support the hypothesis that Gadd45a expression is up-regulated during breast carcinogenesis driven by either of the two oncogenes.

Figure 1. Gadd45a in oncogene induced breast tumorigenesis.

(A, B). Western blot showing Gadd45a expression in normal spleen and mammary tissue and mammary tumor tissue. Actin probe was used for control. (C). Kaplan-Meier curve showing the proportion of tumor free mice as a function of time for Myc/gadd45a−/− (n=25), Myc/gadd45a+/− (n=25), Myc/gadd45a+/+ (n=24) (P<0.05). (D). Average tumor growth rate for tumors arising in myc/gadd45a−/− (n=8), Myc/gadd45a+/− (n=8), Myc/gadd45a+/+ (n=6) mice. Tumor growth was monitored every other day for 13 days by caliper measurement (P<0.05). (E). Comparison between Myc and Ras-mediated tumorigenesis.

To assess the effect of Gadd45a deficiency on Myc-driven breast carcinogenesis compared to the Ras-driven carcinogenesis, female animals from MMTV-Myc mice that are either wildtype or null for Gadd45a expression were monitored twice weekly for the formation of tumors. Tumorigenesis was decelerated in Myc/Gadd45a−/− mice compared to Myc/Gadd45a+/+ control animals (P<0.05; Figure 1C). The median tumor onset, measured as the time in which 50% of animals develop tumors, was 6 months for Myc/Gadd45a+/+ mice, whereas the loss of Gadd45a delayed the median tumor onset to 10 months. Myc expressing Gadd45a heterozygous mice display an intermediate phenotype with a median tumor onset of 8 months. The loss of Gadd45a resulted in an overall decrease in tumor incidence, where only 67% of mice developed tumors within 12 months compared to 88% for Myc/Gadd45a+/+ mice. This was in sharp contrast to the Ras breast cancer model, where the loss of Gadd45a resulted in an increase in tumor incidence (Figure 1E).

To determine the effect of loss of Gadd45a on the overall rate of tumor growth, tumor volume was monitored every two days by caliper measurements, starting upon first tumor visualization and continuing for approximately two weeks or until the general well being of the animal was compromised. Mammary tumors arising from Myc expressing Gadd45a deficient mice displayed significantly decreased rates of growth compared to those mice that express Gadd45a (P<0.05; Figure 1D). For example, at 14days, the average Myc/Gadd45a−/− tumor had a volume of 3,166±881mm³ (n=8), whereas the Gadd45a expressing tumors had an average tumor volume of 4,476±996mm³ (n=6). These observations are in stark contrast to results for MMTV-Ras mice, where the loss of gadd45a resulted in increased tumor size (Figure 1E). Myc/Gadd45a+/− tumors displayed an intermediate phenotype.

Taken together, these data indicate that Gadd45a functions to promote Myc-driven tumor growth, which is in contrast to its tumor suppressive function in the context of Ras-driven tumorigenesis.

Histological Examination Reveals Massive Cell Death in Myc-Expressing Gadd45a Deficient Tumors

To elaborate on the differences in the rate of formation and growth between Myc and Ras driven breast carcinogenesis, the histopathology of tumor tissue sections from the different genotypes was determined. Higher histological tumor grades are associated with loss of cellular shape and size uniformity, increased nuclear size, hyperchromatic nuclei and presence of multinucleated cells. As seen in Figure 2 and Supplemental Figure 1, the presence of Gadd45a in MMTV-Myc mice is associated with a higher histological grade than Gadd45a deficient tumors. Myc expressing Gadd45a wildtype tumors displayed a loss of cellular shape and size uniformity, increased nuclear size and hyperchromatic nuclei. Importantly, Myc expressing Gadd45a deficient tumor tissue samples displayed massive cellular death, consisting of cellular debris and nuclei fragments (Figure 2A). Myc/Gadd45a+/− tumors displayed less aggressive phenotypes than wildtype tumors, which is reflected in the tumor growth rates. It should be noted that only in the complete absence of Gadd45a do we find massive levels of cell death. In sharp contrast, in MMTV-Ras mice, the Gadd45a-deficient tumors had a more aggressive histological phenotype than Gadd45a-expressing tumors, without massive cell death (Figure 2B)¹³.

Figure 2. Histopathology of tumors from different genotypes.

(A). Histological characteristics of formalin-fixed, paraffin-embedded tumor tissue arising from Myc expressing Gadd45a+/+, Gadd45a+/−, Gadd45a−/− (40X) mice and Myc+Gadd45a−/− sample displaying areas of massive cell death at 10X. (B). Ten tumors from each genotype were examined for histological characteristics. “+” was assigned if all ten samples displayed the individual characteristic; “++” was assigned if all ten samples displayed multiple occurrences of the characteristic.

These observations suggest that for Myc-driven breast cancer Gadd45a promotes the development of aggressive tumors, whereas in Ras-driven breast cancer Gadd45 suppresses the development of aggressive tumors.

Gadd45a Has Opposing Effects on Apoptosis and Senescence in Myc-driven compared to Ras-driven Breast Carcinogenesis

Since histological examination detected massive cell death in Myc expressing Gadd45a deficient tumors, the level of apoptosis for the different tumors was ascertained, using TUNEL analysis. In the presence of Myc, loss of Gadd45a resulted in a significant increase in apoptosis compared to wildtype controls (P<0.05) (Figure 3A). The opposite effect was observed for Ras-driven breast carcinogenesis, where loss of Gadd45a resulted in decreased apoptosis (Figure 3E).

Figure 3. Loss of Gadd45a is associated with increased apoptosis and senescence, but has no effect on activated JNK or p38 in Myc-driven tumorigenesis.

(A). Percent apoptotic cells of Myc expressing tumors determined by TUNEL analysis. Positive cells were visualized by light microscopy and quantified as described in Materials and Methods (Myc+Gadd45a+/+ n=9, Myc+Gadd45a−/− n=12) (* represents P<0.05). (B). Average percentage of Beta-Galactosidase positive cells from Myc+Gadd45a+/+ (n= 5) and Myc+Gadd45a−/− (n=6) (* represents P<0.05). (C). Average percentage of Phospho-JNK positive cells on tissue sections. (Myc+Gadd45a+/+ n=7, Myc+Gadd45a−/− n=8). (D) Average percentage of Phospho-p38 positive cells for Myc+Gadd45a+/+ (n=7) and Myc+Gadd45a−/− (n=6) tumors. For A-D, all IHC quantitation was performed blindly using Image J software. (E). Comparison of levels of TUNEL, Beta-galactosidase, Phospho-Jnk and Phospho-p38 expression between Myc and Ras-mediated tumors.

Oncogene-induced senescence has been reported to play a role in retarding tumor progression^18,19,20. Therefore, we investigated the levels of senescence by immunohistochemistry using anti-Beta-Galactosidase antibody. It was observed that the loss of Gadd45a significantly increased the level of cellular senescence in the presence of Myc (P<0.05) (Figure 3B). In contrast, loss of Gadd45a significantly decreased cellular senescence in the context of Ras-driven breast tumorigenesis (Figure 3E)

In the case of Ras-driven breast tumors, Gadd45a mediated activation of JNK and p38 were correlated with Ras-induced apoptosis and senescence, respectively (Figure 3E)¹³. It was of interest, therefore, to analyze the levels of both activated JNK and p38 in Myc-driven breast tumors. Intriguingly, it was observed that in breast tumors from MMTV-Myc mice there was no significant difference between the levels of either activated JNK or p38 in the presence or absence of Gadd45a (Figure 3C, D & E).

Taken together, these findings show that Gadd45a employs an alternate mechanism to promote Myc-driven breast carcinogenesis as opposed to suppressing Ras-mediated breast cancer.

Gadd45a suppresses MMP10 Expression resulting in increased Capillary Vascularization in Myc-driven Breast Tumors

To explore possible mechanisms by which Gadd45a promotes Myc-driven breast carcinogenesis, a gene expression profiling study was conducted using DNA microarray membranes designed to screen for signal transduction pathways in cancer (GEArray Q Series Mouse Signal Transduction in Cancer Gene Array, SuperArray) (Figure 4A). The expression profiles of Myc-generated breast tumors from Gadd45a wildtype, heterozygous and null mice were evaluated using the appropriate software (Materials & Methods). A list of genes that displayed significant changes in expression between Myc+/Gadd45a+/+ and Myc+/Gadd45a−/− tumors are provided in Supplemental Figure 2. The Matrix Metalloproteinase 10 (MMP10) gene was observed to be expressed more than 6-fold higher in Myc expressing Gadd45a deficient breast tumors compared to Gadd45a wildtype tumors. The significance of this finding became the focus of additional experiments.

Figure 4. Myc expressing Gadd45a deficient tumors have elevated levels of MMP10 and decreased tumor vascularization.

(A) Myc+Gadd45a+/+, Myc+Gadd45a+/− and Myc+Gadd45a−/− expression profiles. Total RNA was extracted using whole tumors from the three genotypes with Trizol. Following RNA extraction, membrane microarray analysis was performed using GEArray Q-Series Mouse Signal Transduction in Cancer Gene Array as described in Materials and Methods. Levels of gene transcripts were normalized using housekeeping genes. □ indicates MMP10 expression, which was up regulated more than 6-fold in the Gadd45a deficient sample. (B.) Average percentage of MMP10 positive cells from Myc+Gadd45a+/+ (n=6) and Myc+Gadd45a−/− (n=6) using immunohistochemistry (* represents P<0.05). (C.) Average percentage of CD31 and CD105 positive cells from Myc+Gadd45a+/+ (n=6) and Myc+Gadd45a−/− (n=6) using immunohistochemistry (* represents P<0.05). For B-C, all IHC quantitation was performed blindly using Image J software. (D.) Comparison of levels of MMP10 and CD31 & CD105 expression between Myc and Ras-mediated tumors.

Immunohistochemistry employing MMP10 antibody confirmed the array induction data; there was increased MMP10 expression in Myc expressing Gadd45a deficient tumor samples compared to wildtype tumor samples (P<0.05) (Figure 4B). Levels of MMP10 were examined in Ras-driven breast tumors, either wildtype or null for Gadd45a, to determine if MMP10 expression is altered in Ras-driven breast tumors, however, no significant difference was detected (Figure 4D & Supplemental Figure 3A). These results indicate that Gadd45a elicits distinct signaling profiles depending on the nature of the activated oncogene.

MMP10 (also known as Stromelysin-2) plays a role in capillary tubular network collapse and regression due to its ability to degrade various components of the basement membrane matrix, such as collagen type IV, laminin and proteoglycans²¹. Thus, we examined the capillary tubular network in breast tumors from the various genotypes to determine how the level of MMP10 correlated with the status of the capillary networks. Immunohistochemistry was performed with antibodies to CD31 and CD105, marker proteins known to play a role in capillary network maintenance, angiogenesis and vascularization^22,23. It can be seen that loss of Gadd45a was associated with significantly decreased levels of vascularization in Myc-driven breast tumors (P<0.05) (Figure 4C), whereas no significant differences in vascularization were detected between Gadd45a wildtype and Gadd45a deficient tumors in Ras-driven breast tumors (Figure 4D & Supplemental Figure 3B).

Collectively, the data suggest a novel mechanism for Gadd45a-mediated promotion of Myc-driven tumorigenesis via suppression of MMP10 expression.

MMP10 expression is regulated by GSK3-β/β-catenin signaling in Myc-driven breast tumors

Evidence has implicated the transcriptional activator β-catenin in regulating expression of MMPs²⁴. While the majority of β-catenin is associated with cell–cell adhesion complexes, a small fraction is present in the cytoplasm, its level regulated by glycogen synthase kinase 3β (GSK3β) phosphorylation and subsequent proteasome degradation²⁵. Phosphorylated GSK3β is inactive, whereas dephosphorylation renders it active contributing to degradation of β-catenin. Hence, inactivation of GSK3β via phosphorylation results in accumulation of hypo-phosphorylated β-catenin, which in turn leads to its nuclear translocation, and binding to and activation of the TCF family of transcription factors²⁴.

While direct β-catenin transcriptional regulation of several MMPs has been identified, the role of β-catenin in MMP10 regulation had not been established^24,26. Therefore, it was first determined if β-catenin can bind to MMP10 promoter sequences. Searching GenBank, a potential TCF binding element was identified in the promoter of MMP10, which matched the consensus sequence, 5′-A/T A/T CAAAG-3′. Electromobility shift assays (EMSA) were performed, using the identified binding sequence in the MMP10 promoter as probe and nuclear extracts from SW480 colon cancer cells, known to express high levels of β-catenin²⁴. β-catenin bound to the probe, whereas excess cold probe competed off the binding (Figure 5A). The addition of β-catenin antibody resulted in a specific supershift of the band, indicating that β-catenin was bound to the probe. Taken together, these results implicate β-catenin in transcriptional activation of the MMP10 gene.

Figure 5. Gadd45a regulates Myc-driven breast tumorigenesis via the GSK3β/β-catenin pathway.

(A.) Electrophoresis mobility shift assay (EMSA) using MMP10 promoter sequence oligonucleotide probe incubated with SW480 nuclear extracts. Supershift resulted by addition of Anti-β-catenin antibody. Anti-Myc antibody was used as a negative control. (B.) Average percentage of phospho-GSK3β positive cells from Myc+Gadd45a+/+ (n=6) and Myc+Gadd45a−/− (n=6) using immunohistochemistry (* represents P<0.05). (C.) Average percentage of β-catenin positive cells from Myc+Gadd45a+/+ (n=6) and Myc+Gadd45a−/− (n=6) using immunohistochemistry (* represents P<0.05). For B-C, all IHC quantitation was performed blindly using Image J software. (D.) Comparison of levels of Phospho-GSK3β and β-catenin expression between Myc and Ras-mediated tumors

Next, we asked if Gadd45a regulation of MMP10 expression was mediated by the GSK3-β/β-catenin signaling pathway. To this end, we determined the phosphorylation status of GSK3-β and the level of β-catenin in our breast tumor tissue samples by immunohistochemistry. As shown in Figure 5B, there was a high level of phosphorylated GSK3-β in Myc expressing Gadd45a deficient tumor samples compared to wildtype Gadd45a samples (P<0.05). Consistent with this observation, loss of Gadd45a also was correlated with significantly increased levels of nuclear β-catenin (P<0.05) (Figure 5C). In contrast, loss of Gadd45a had no effect on phospho-GSK3-β or β-catenin in Ras-driven breast tumors (Figure 5D & Supplemental Figure 4A and B).

While it is known that Gadd45a promotes GSK3-β dephosphorylation in keratinocytes, we chose to further our investigation of the role of Gadd45a on regulation of the GSK3-β pathway in the mammary tissue²⁶. Using total cell lysates from Gadd45a null mammary tissue, we performed a phosphatase assay with increasing amounts of purified Gadd45a protein. Our results demonstrate that under this in vitro experimental setting, Gadd45a can promote the dephosphorylation of GSK3-β in a dose dependent manner (Supplemental Figure 5). It should be noted, however, that under physiological in vivo setting, the effect of Gadd45a on the phosphorylation status of GSK3-β appears to be manifested in the presence of deregulated Myc, but not in the presence of oncogenic Ras, since Gadd45a loss had no effect on the phosphorylation status of GSK3-β or β-catenin in Ras-driven breast tumors.

Taken together, these novel findings implicate Gadd45a as a negative regulator of MMP10 expression via GSK3-β/β-catenin signaling, which promotes Myc-driven breast tumor vascularization and tumor development.

RNAi mediated suppression of MMP10 promotes tumor vasculature and growth

To further explore the role of MMP10 in the suppression of breast tumor development, in vivo RNA interference (RNAi) was employed. Chemically modified siRNA molecules designed against MMP10 were injected intratumorally every 4 days to knock down expression levels of MMP10 in Myc+Gadd45a−/− transplanted tumors implanted into Gadd45a null mice. Tumor volume was monitored every 4 days. Forty eight days after transplantation, animals were sacrificed and the tumors were excised.

As shown in Figure 6A, RNAi mediated suppression of MMP10 resulted in significantly increased tumor volume compared to control siRNA (P<0.05). Western blotting demonstrated that MMP10 was knocked down in tumors which exhibited increased growth (Figure 6B). As noted in Figure 2, Myc expressing Gadd45a deficient tumor tissue samples displayed massive cellular death, consisting of cellular debris and nuclei fragments. As shown in Figure 6C, tumors treated with MMP10 siRNA, but not control siRNA, lacked massive cellular death confirming that blocking MMP10 resulted in a rescue from the phenotype. To further investigate the role of MMP10 in degradation of tumor vasculature, we assessed CD31 and CD105 levels as a means to measure tumor vascularization. As shown in Supplemental Figure 6, a significant increase in the levels of tumor vascularization was observed in tumors where MMP10 expression was knocked down (P<0.05).

Figure 6. siRNA Mediated Suppression of MMP10 expression promotes tumor growth.

(A.) Representative samples from multiple experiments showing tumor growth rates for tumors treated with siRNA to MMP10 (3) and control (3) every 4 days (P<0.05). (B.) Western blot for MMP10 demonstrating efficient knock down of MMP10 active protein. (C.) Histological characteristics of formalin-fixed, paraffin-embedded tumor tissue sections stained with H&E.

Taken together, these results indicate that MMP10 suppresses Myc-driven breast tumorigenesis by promoting degradation of tumor vascularization.

DISCUSSION

Generation and side by side analysis of MMTV-Myc versus MMTV-Ras mice strains highlight a unique role for Gadd45a as either a suppressor or promoter of breast cancer development, employing distinct signaling pathways in response to distinct oncogenic stress stimuli (Figure 7 and Supplemental Figure 7). Our data indicates that the Gadd45a tumor suppressor function, mediated through activation of JNK and p38 stress kinases, contributes to Ras-induced apoptosis and senescence respectively, and is a unique response to Ras oncogenic stress. In contrast, the tumor promoter function of Gadd45a, mediated through negative regulation of MMP10 expression via the GSK3β/β-catenin signaling cascade, resulting in increased tumor vascularization, is a unique response to oncogenic Myc. These novel results indicate that Gadd45a can function to either promote or suppress breast tumor development through engagement of different signaling pathways, depending on the molecular nature of the activated oncogene.

Figure 7. Schematic diagram demonstrating how gadd45a modulates mammary tumor development in Myc-driven tumors compared to Ras-driven tumors.

Gadd45a acts as a tumor promoter in the presences of Myc. It blocks MMP10 expression, accounting for a decrease in both apoptosis and tumor cell senescence. Gadd45a acts as a tumor suppressor in the presence of Ras by promoting Jnk-mediated apoptosis and p38-mediated senescence.

In the case of Myc-driven breast carcinogenesis, we have shown that the loss of Gadd45a results in increased levels of MMP10 through inactivation of the GSK3β/β-catenin signaling pathway (Figure 7). MMP10 was identified as a direct target of β-catenin. Loss of Gadd45a increased phosphorylation of the GSK3-β kinase, rendering it inactive, and dramatically increased the levels of β-catenin. Inactivation of GSK3-β by phosphorylation is known to result in accumulation of hypo-phosphorylated β-catenin, which in turn translocates to the nucleus where it binds to TCF and functions as a transcriptional activator²⁴. It is noteworthy that while our work was in progress evidence was obtained that in keratinocytes, UV induced Gadd45a directly associates with GSK3β to promote GSK3β dephosphorylation and activation²⁶. These findings are supported by our data providing direct evidence that in normal mammary tissue, Gadd45a promotes dephosphorylation of GSK3-β

MMP10 is known to play a role in capillary tubular network collapse and regression due to its ability to degrade various components of basement membrane matrix, such as collagen type IV, laminin and proteoglycans²¹. Thus, it is logical to assume that the decrease in vascularization that is observed in Myc expressing tumors deficient for Gadd45a is a direct consequence of MMP10 activation. In turn, the massive levels of cellular death, associated with accumulation of cellular debris and increased apoptosis observed in Myc expressing tumors deficient for Gadd45, is likely to be the consequence of the decrease in vascularization. It is notable that the massive cell death, as evident by large areas of cellular debris and nuclei fragments, observed upon histological analysis of Myc/Gadd45a−/− tumor sections (Figure 2A), by far exceeds the percentage of apoptotic cell death detected by the TUNEL assay (Figure 3A). Keeping in mind that the TUNEL immunohistochemical data in essence is a snapshot of cells undergoing apoptosis at the time of tumor fixation, the histological data reflects accumulative apoptotic cell death during tumor development. However, it is also possible that in addition to apoptosis, necrosis contributes to Myc/Gadd45a−/− tumor cell death. The increase in cell senescence in Myc expressing tumor tissue deficient for Gadd45a may also be due at least partially to the decrease in vascularization. Alternatively or in addition, it may also be the consequence of the role implicated for MMP10 as a regulator of cellular senescence²⁷. Whatever may be the case, the in vivo RNAi mediated suppression of MMP10 expression correlating with increases in Myc-driven tumor growth, provides direct evidence for a novel role of MMP10 in suppression of breast carcinogenesis.

The notion that the response of Gadd45a to oncogenic stress signals depends on the nature of the activated oncogenes is novel and intriguing. These results stress the significance of the type of oncogenic alterations found in the target cell on how stress response genes, such as Gadd45a, influence the initiation and progression of tumors. Since Gadd45a is a transcriptional target for both BRCA1 and p53, our observations regarding the unique role played by Gadd45a as either a tumor suppressor or tumor promoter, depending on the activated oncogene, also has important implications regarding the role of BRCA1 and p53 as tumor suppressors in the context of distinct oncogenic stressors. It is noteworthy that deficiency of the CDKI p21 was observed to differentially affect Ras- versus Myc-driven mammary tumor properties, promoting either growth arrest or proliferation, depending on the specific cellular context, although an exact mechanism was not identified²⁸.

Finally, we would like to point out that our findings raise several interesting questions, which warrant further investigation: (1) How does Gadd45a interface with different signaling cascades in response to distinct oncogenic stress; (2) How does Gadd45a influence breast carcinogenesis driven by oncogenes other than Ras and Myc; (3) How does Gadd45a modulate development of other tumor types driven by distinct oncogenic stress; (4) What role does Gadd45a expression or the lack of it play in human breast carcinogenicity and finally, (5) Do other Gadd45 proteins, i.e. Gadd45b and Gadd45g, either separately or in combination with Gadd45a, modulate breast tumor development. Current research is targeted at addressing these interesting issues.