Upregulation of lactate dehydrogenase a by 14-3-3ζ leads to increased glycolysis critical for breast cancer initiation and progression

Metabolic reprogramming is a hallmark of cancer. Elevated glycolysis in cancer cells switches the cellular metabolic flux to produce more biological building blocks, thereby sustaining rapid proliferation. Recently, new evidence has emerged that metabolic dysregulation may occur at early-stages of neoplasia and critically contribute to cancer initiation. Here, our bioinformatics analysis of microarray data from early-stages breast neoplastic lesions revealed that 14-3-3ζ expression is strongly correlated with the expression of canonical glycolytic genes, particularly lactate dehydrogenase A (LDHA). Experimentally, increasing 14-3-3ζ expression in human mammary epithelial cells (hMECs) up-regulated LDHA expression, elevated glycolytic activity, and promoted early transformation. Knockdown of LDHA in the 14-3-3ζ-overexpressing hMECs significantly reduced glycolytic activity and inhibited transformation. Mechanistically, 14-3-3ζ overexpression activates the MEK-ERK-CREB axis, which subsequently up-regulates LDHA. In vivo, inhibiting the activated the MEK/ERK pathway in 14-3-3ζ-overexpressing hMEC-derived MCF10DCIS.COM lesions led to effective inhibition of tumor growth. Therefore, targeting the MEK/ERK pathway could be an effective strategy for intervention of 14-3-3ζ-overexpressing early breast lesions. Together, our data demonstrate that overexpression of 14-3-3ζ in early stage pre-cancerous breast epithelial cells may trigger an elevated glycolysis and transcriptionally up-regulating LDHA, thereby contributes to human breast cancer initiation.


INTRODUCTION
One in five breast cancer patients presents earlystage disease in the clinic, such as atypical ductal hyperplasia (ADH) and/or ductal carcinoma in situ (DCIS) [1]. Although advances in targeted therapies have substantially reduced the mortality rate for breast cancer patients [2], a better mechanistic understanding of cancer initiation will undoubtedly lead to significant improvement in disease prevention and further reduce breast cancerspecific mortality.
Breast cancer initiation and progression involve multiple cellular alterations, including metabolic dysregulation [3][4][5]. Recently, new evidence has emerged to suggest that metabolic perturbation, including glycolytic shift, starts at early-stage diseases that contribute to the early transformation of normal epithelial cells and the initiation of cancer [6,7]. The Warburg effect is a metabolic abnormality of cancer cells characteristic of elevated aerobic glycolysis, which promotes production of biological building blocks for cell proliferation [8]. Critical mediators of the Warburg effect include transcription factors MYC and HIF-1α [9]. However, it remains unclear what other mediators may promote Warburg effect, especially during the early-stage breast cancer initiation and progression.
Given the concomitant 14-3-3ζ overexpression and abnormal metabolic alterations during the early-stage breast disease, we postulated that overexpression of 14-3-3ζ may lead to increased glycolysis and contribute to early transformation of mammary epithelial cells and subsequent breast cancer initiation/progression. To this end, we examined gene expression profiling data of early-stage neoplastic breast lesions and found a strong correlation between the expressions of 14-3-3ζ and genes of the canonical glycolytic pathway, particularly lactate dehydrogenase A (LDHA). By exogenously overexpressing 14-3-3ζ in nontransformed human mammary epithelial cells (hMECs), we identified a direct mechanistic link between 14-3-3ζ overexpression and LDHA upregulation. We revealed that 14-3-3ζ-mediated LDHA upregulation is critical to early transformation of hMECs. Our data provide direct evidence that 14-3-3ζ overexpression in earlystage breast disease contributes critically to the metabolic dysregulation of hMECs, and that 14-3-3ζ confers a metabolic advantage to initiate neoplastic transformation. Importantly, our data also demonstrate that targeting 14-3-3ζ-induced metabolic dysregulation could be an efficacious strategy for prevention and early intervention of early-stage breast cancer.

14-3-3ζ overexpression increases glycolysis in human breast epithelial cells
Since metabolic dysregulations has recently been implicated to take place during the early-stage of neoplastic transformation [6,7], and abnormal 14-3-3ζ overexpression was also observed in pre-cancerous breast lesions [14], we hypothesized that 14-3-3ζ overexpression contributes to metabolic dysregulations in early-stage breast cancer. To gain insights on this conjecture, we first examined the relationship between 14-3-3ζ expression and genes involved in metabolic functions, using a microarray dataset (GSE16873) generated from histologically normal epithelia, simple ductal hyperplasia (SDH), atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS) [15]. Remarkably, the 14-3-3ζ expression level is most strongly correlated with the expressions of genes involved in glycolysis (Gene Ontology, GO:0006096) in these pre-cancerous and early-stage breast lesions ( Figure 1A). In addition, this strong correlation between 14-3-3ζ expression level and glycolytic genes persisted in breast cancer patients [16] (Supplementary Figure S1, GSE2109).
To test whether 14-3-3ζ functionally modulates cellular glycolytic activity in pre-cancerous mammary epithelial cells, we measured glycolytic activities in widely used and validated models of pre-cancerous mammary epithelial cells, i.e., MCF10A and MCF12A hMECs. We compared three glycolytic indices (i.e., glucose uptake, lactate production, and oxygen consumption) in MCF10A and MCF12A hMECs that had either exogenous 14-3-3ζ overexpression (10A.ζ and 12A.ζ cells) or endogenous 14-3-3ζ knockdown (10A.shζ, and 12A.shζ cells) ( Figure  1B) [11]. Indeed, glucose uptake and lactate production significantly increased in both 10A.ζ and 12A.ζ cells compared to this control cells, but significantly reduced in both 10A.shζ, and 12A.shζ cells compared to their control cells ( Figure 1C and 1D). Cells with higher glycolysis activity tend to reduce the rates of oxidative phosphorylation and oxygen consumption and shift the metabolic flux from the ATP-generating TCA cycle to the biomass-producing glycolysis. Consistently, the oxygen consumption rates of the 14-3-3ζ-overexpressing 10A.ζ and 12A.ζ cells were significantly reduced than that of the control cells ( Figure 1E). However, no significant difference in oxygen consumption was detected between the 10A.shζ and 12A.shζ cells and their control cells ( Figure 1E), suggesting that 14-3-3ζ does not modulate basal oxygen consumption in the 14-3-3ζ low-expressing MCF10A and MCF12A cells. Next, we evaluated the overall impact of 14-3-3ζ on cellular glycolytic activity by calculating the glycolytic index, G x L / O, where G is for glucose uptake rate, L is for lactate generation production, and O is for oxygen consumption rate [17]. 14-3-3ζ-overexpressing 10A.ζ and 12A.ζ cells had 4 to 5 fold increases, whereas 14-3-3ζ knockdown 10A.shζ and 12A.shζ cells had approximately 70% decreases, in their glycolytic index compared to their respective control cells (Supplementary Table S1 and S2). Collectively, these data demonstrate that 14-3-3ζ positively modulates glycolytic activity in nontransformed MCF10A and MCF12A cells.

14-3-3ζ overexpression upregulates LDHA leading to increased aerobic glycolysis
To investigate the molecular mechanisms of the 14-3-3ζ-mediated increase of glycolysis, we focused on LDHA because its expression, compared with that of other glycolytic genes in human pre-cancerous lesions, is more strongly associated with 14-3-3ζ expression level (Figure www.impactjournals.com/oncotarget 1A). Furthermore, 14-3-3ζ-overexpressing 10A.ζ and 12A.ζ cells had significantly increased levels of LDHA mRNA and protein expression compared to their vector control cells; whereas knockdown of 14-3-3ζ in 10A.shζ and 12A.shζ cells led to significantly decreased LDHA mRNA and protein levels compared to the control shCtrl cells (Figure 2A and 2B).
To identify the cis-regulatory element in the -2000bp LDHA promoter and 5'-UTR region that is responsible for 14-3-3ζ-induced transcriptional upregulation, we made a series of deletion constructs in the -2000 to +272 region, subcloned them into the pGL3-Basic vector, and transfected them into the 10A.ζ, 10A.Vec, 10A.shζ, and 10A.shCtrl cells ( Figure 4A, left). Remarkably, a specific 65bp region (+85 to +150) in the 5'-UTR of LDHA gene was necessary and sufficient to induce the luciferase gene expression ( Figure 4A). Transfection of this 65bp 5'-UTR region resulted in higher luciferase activity in 10A.ζ than in vector control cells ( Figure 4A, middle). In contrast, the luciferase activity driven by the 65bp 5'-UTR was lower in the 10A.shζ cells compared to the 10A.shCtrl cells ( Figure 4A, right).
Next, to identify the transcription factors that bind to the 65bp 5'-UTR region of LDHA and are responsible for transcriptional upregulation by 14-3-3ζ, we analyzed this DNA sequence for putative transcription factor binding sites using the Transcription Element Search System (TESS) and University of California-Santa Cruz (UCSC) genome browser online analysis tools [21,22]. We identified potential binding sites for five transcription factors USF1, CREB, MYC, SP1, and ATF-1 (Supplementary Figure S4A), knocked down them individually in the 10A.ζ cells and compared their luciferase activities. Knocking down each of the five genes reduced the LDHA 65bp 5'-UTR driven luciferase activity in 10A.ζ cells to various degrees and knocking down CREB led to the most significant reduction ( Figure 4B). Furthermore, we examined LDHA mRNA and protein levels in the 10A.Vec and 10A.ζ cells that had the five transcription factors knocked down individually. Among these five transcription factors, CREB and MYC knock down in the 10A.ζ cells (Supplementary Figure S4B and S4C), significantly reduced LDHA protein and mRNA compared to shCtrl cells ( Figure  4C and Supplementary Figure S4D). However, CREB knockdown had less effect on the LDHA protein level in the control 10A.Vec cells compared to MYC knockdown cells ( Figure 4C), suggesting that CREB has a more specific role in regulating LDHA expression in 14-3-3ζ overexpressing cells. Therefore, we focused on how 14-3-3ζ overexpression leads to CREB transactivation of LDHA that contributes to early transformation in 14-3-3ζ overexpressing hMECs.

Targeting MEK/ERK/CREB pathway effectively inhibits LDHA expression and tumor outgrowth in a 14-3-3ζ overexpressing DCIS model
As the above data demonstrated that the MEK/ ERK/CREB pathway is critical for 14-3-3ζ-induced LDHA upregulation which contributes to hMEC early transformation. Conceivably, targeting this pathway to inhibit metabolic adaptation of early-stage breast cancer cells towards glycolysis may be an effective strategy to intervene cancer progression. Therefore, we next investigated whether targeting the 14-3-3ζ downstream MEK/ERK pathway may effectively prevent or intervene the early-stage breast cancer further progression in vivo.
To this end, we exogenously overexpressed 14-3-3ζ in early-stage DCIS model of MCF10DCIS.COM cells. The MCF10DCIS.COM line is a MCF10A cells-derived model that forms DCIS-like mammary lesions and ultimately progresses to invasive mammary tumors in nude mice [28]. Recent studies revealed almost identical genomic profiles between the MCF10A and MCF10DCIS. COM cells, supporting our efforts to extend the above studies of MCF10A to in vivo investigations using the MCF10DCIS.COM line [29]. Consistent with the MCF10A and MCF12A cells, exogenous overexpression of 14-3-3ζ in the MCF10DCIS.COM cells (DCIS.COM.ζ) led to activation of the ERK/CREB pathway and LDHA upregulation in vivo; whereas 14-3-3ζ knockdown in MCF10DCIS.COM led to decreased ERK/CREB activity and LDHA downregulation (Supplementary Figure S5 and S6).
To determine the impact of MEK/ERK inhibitor treatment on 14-3-3ζ/ERK/CREB/ LDHA, as well as on tumor cell proliferation and apoptosis, we collected tumors from control and treatment groups. IHC analysis showed that 14-3-3ζ overexpressing DCIS.COM.ζ tumors had increased phospho-ERK and phospho-CREB levels that correlated with higher LDHA expression compared with DCIS.COM.Vec tumors ( Figure 5C) but had no significant effect on total ERK, CREB or 14-3-3ζ expression levels (Supplementary Figure S6). AZD6244 treatment significantly reduced phospho-ERK, phospho-CREB and LDHA levels in DCIS.COM.ζ and DCIS. COM.Vec tumors compared to vehicle treatment ( Figure  5C and Supplementary Table S4). Compared to the DCIS. COM.Vec.Vehicle tumors, the DCIS.COM.ζ.Vehicle tumors showed no significant difference in apoptosis, but a singnificantly increased Ki67 positive proliferating cells ( Figure 5D), which were inhibited by AZD6244 ( Figure 5C and 5D). These data indicate that AZD6244 effectively inhibit the MEK/ERK/CREB/LDHA axis and proliferation of 14-3-3ζ overexpressiong tumor cells, thereby suppressing DCIS-like tumor outgrowth.

The 14-3-3ζ-LDHA axis as potential biomarkers for predicting clinical outcome
Having established that the 14-3-3ζ/MEK/ERK/ CREB/LDHA axis is potently active in nontransformed hMECs and highly correlated in early neoplastic breast lesions (R 2 >0.8) ( Figure 1A). Next, we extended our examination by bioinformatics analysis of datasets generated from breast cancers. We found that the correlative relationship between the expression levels of 14-3-3ζ (YWHAZ) and LDHA reached R 2 of 0.32 and 0.31 ( Figure 6A and Supplementary Figure S1) in RNAseq-derived TCGA dataset [30], and microarrayderived dataset GSE2109[16] respectively. The strength of such positive correlation is much weaker than that in early-stage diseases ( Figure 1A), suggesting that with cancer progression into advanced stages, neoplastic cells may have metabolitically adapted to a complex tumor environment. Furthermore, when we examined LDHA levels in 14-3-3ζ overexpressing established breast cancer cell lines such as, HCC1954 HER2+ and MDA-MB-231 TNBC, we did not detect a significant up-regulation of LDHA or increase of glycolysis (data not shown). These data suggest that the 14-3-3ζ/CREB/LDHA pathway may be more critical in the early-stage breast cancer initiation.
As the above bioinformatics analyses were performed on RNA data, we further evaluated the clinical relevance of 14-3-3ζ/CREB/LDHA axis at protein level using tissue microarray (TMA) of mixed stages of breast cancers. Indeed, IHC staining revealed that the 14-3-3ζ protein levels were significantly correlated with the LDHA protein levels in these breast cancer specimens ( Figure  6B and Supplementary Table S5). Importantly, the 14-3-3ζ expression levels were also significantly correlated with CREB phosphorylation in the consecutive TMA slides from the same group of patients ( Figure 6B and Supplementary Table S5).
We then evaluated whether 14-3-3ζ and LDHA levels hold prognostic values for breast cancer. As there is no early-lesion dataset with follow-up on clinical outcome, we instead used a breast cancer gene expression dataset (GSE3494) with disease-specific overall survival data. We found that concomitant high expression of 14-3-3ζ and LDHA predicts worse survival of breast cancer patients compared to high expression of either gene alone (P=0.00257 vs. P=0.0127 and P=0.0322) ( Figure 6C). Furthermore, the 5-year survival rate for patients with concurrent high expression of both 14-3-3ζ and LDHA genes dropped almost 10% compared to high expression of either gene alone ( Figure 6C). These data suggest that when combined together, the expression levels of 14-3-3ζ and LDHA have a more power in predicting the clinical outcome of breast cancer patients. It is possible that 14-3-3ζ and LDHA expression levels together may have an even stronger power in predicting the clinical outcome of early-stage disease progression in patients, which should be investigated in future studies.

DISCUSSION
Increased glycolysis in cancer cells is a general phenomenon to satisfy the demanding needs of rapid cell proliferation. However, recent studies suggests that such metabolic dysregulation may start at very early stage of epithelial cell transformation [6,7] and can be used for early cancer detection and diagnosis [31]. In this study, we identified that 14-3-3ζ overexpression-mediated LDHA upregulation contributed to the metabolic switch toward glycolysis and cell transformation in early-stage breast cancer. In addition to our previous studies showing that exogenous 14-3-3ζ overexpression cooperates with ErbB2 to promote DCIS to IDC transition [32]. Here, we demonstrated a novel role of 14-3-3ζ in mediating earlystage breast cancer progression by upregulating LDHA expression and cellular glycolysis (Figure 7).
In early-stage neoplastic lesions, the expression of 14-3-3ζ strongly and positively associated with the expression of LDHA. However, when the disease progressed into advanced stages, the correlation between 14-3-3ζ and LDHA is not very strong. It suggests that 14-3-3ζ-induced LDHA upregulation is critical for hMEC early transformation and breast cancer early initiation but may be less critical as cancer progresses and adapts to a more complex metabolic environment. Nevertheless, 14-3-3ζ/LDHA axis still provides prognostic values for overall survival of breast cancer patients ( Figure 6C). LDHA is known to be regulated by other oncogenes such as HIF-1a during cancer progression, thus cancer cells have expanded regulatory mechanism to confer metabolic advantages in the advanced stages of diseases. Previous studies showed that 14-3-3ζ is involved in multiple functions including chemoresistance, epigenetics regulation, adipogenesis and metastasis [11,[33][34][35]. Here, we demonstrated a novel role of 14-3-3ζ in cancer cell metabolism and our findings bring new insights into the mechanistic understanding of 14-3-3ζ overexpression-mediated breast cancer initiation and progression of early-stage breast diseases.
LDHA is a CREB target gene because the LDHA promoter has c-AMP response element (CRE) [36]. A genome-wide promoter analysis of CREB target genes showed that many of the conserved CREs (TGACGTCA) exist either within 1000-bp of the ATG initiation codon or within 250-bp of the 5'-UTR sequence in mouse and human genome [36]. In our study, 10A.ζ transfected with the 65bp region of the 5'-UTR showed higher luciferase activity than those with the 2000-bp upstream promoter region ( Figure 4A), suggesting that the CRE binding site in the 5'-UTR is critical for CREB binding and transactivation of LDHA mRNA in hMECs. CREB activation is mediated through phosphorylation by multiple upstream pathways [37]. Here, we demonstrated that 14-3-3ζ overexpression upregulates LDHA expression and increases glycolysis through the MEK/ERK/CREB signaling axis in early breast cancer, delineating the important signaling pathway for an essential cellular phenotype. Although previous studies showed that c-Myc is critical to regulate LDHA expression and promote cancer progression [38,39], however, in our models, CREB is more important for LDHA upregulation in 14-3-3ζ-overexpressing hMECs ( Figure 4C). Additionally, we demonstrated strong correlation between 14-3-3ζ, CREB phosphorylation and LDHA expression levels in human breast cancers by IHC staining of patient-derived tissue microarray (TMA), which extended our in vitro and preclinical findings to human breast cancer specimens and validated the clinical relevance of our studies. Because metabolic dysregulation is a hallmark of cancer [40,41], targeting abnormally altered metabolic pathways in cancer cells has emerged as an attractive therapeutic approach that is being actively investigated [42,43]. In highly proliferating cells and cancer cells, LDHA is a key enzyme in the glycolytic cascade that converts pyruvate to lactate and cycles NADH back to NAD + to sustain rapid proliferation [44,45]. After identifying the important roles of the 14-3-3ζ/ERK/ CREB/LDHA signaling axis in mediating cellular glycolysis, we showed that targeting MEK/ERK pathway in 14-3-3ζ overexpressing DCIS.COM tumors (DCIS.COM.ζ) with AZD6244 significantly inhibited the mammary tumor growth by inhibiting proliferation. Although MEK/ERK could potentially modulate multiple downstream targets, our data clearly demonstrated that inhibiting MEK/ERK leads to reduced LDHA and tumor inhibition. These data suggest that 14-3-3ζ-mediated LDHA upregulation and metabolic dysregulation could be intervened concurrent by targeting for MEK/ERK and early intervention of cancer [46].
In summary, our studies defined a novel role of 14-3-3ζ during early transformation and early-stage breast cancer by transcriptionally upregulating LDHA and functionally increasing glycolysis, which ultimately facilitate breast cancer initiation and progression. Furthermore, our data in a preclinical DCIS model indicated that inhibiting the MEK/ERK pathway could be an effective strategy for intervention of the earlystage breast cancer, and this strategy may potentially be applicable in the clinic.

Cell lines and cell culture
MCF10A and MCF12A cells were obtained from ATCC and were maintained in hMECs medium [11]. MCF10DCIS.COM cells, provided by Dr. Fariba Behbod [28], were maintained in DMEM/F-12 media supplemented with 5% horse serum. All cell lines were authenticated and validated by MD Anderson Cancer Center's Characterized Cell Line Core.

Bioinformatics
Glycolysis-related genes were extracted from the Gene Ontology database under the term "glycolytic process" (accession #GO:0006096). The corresponding patientderived gene expression values were retrieved from the Gene Expression Omnibus data repository microarray datasets for early-stage breast neoplasia (GSE16873) and advanced breast cancers (GSE2109), as well as the RNA-seq datasets from the TCGA data portal. The clustering analysis of gene expression and heatmap visualization of the correlation matrix were performed in the corrplot package (v0.73) on the statistical computation platform R (v3.1.2). Breast cancer patient-derived gene expression dataset GSE3494 had clinical follow-up information and was used to investigate the relationship between 14-3-3ζ and/or LDHA expression and patient prognosis. Patients were stratified as having either "low" or "high" expression of the gene by median expression values of 14-3-3ζ and LDHA in the data cohort, respectively. Patients with "low" or "high" expression of both 14-3-3ζ and LDHA genes were further extracted to investigate whether there is synergistic predictive effect to combine the two biomarkers together. The Kaplan-Meier plots and survival analysis of breast cancer patients (GSE3494) were performed using R package survival (v2.38).

Metabolic assays
For glucose uptake assays, cells were starved for 3 hours before 2-deoxy-D-glucose-[1,2-3 H(N)] (Moravek Biochemicals) was added in. Tritium signal was measured by liquid scintillation counting. For lactate production assay, cells were plated for 24 hours, followed by freshmedium incubation for 1 hour. Lactate production was measured according to the manufacturer's instructions (BioVision Inc). For oxygen consumption assays, cells were plated in a 96-well oxygen biosensor system (BD Biosciences) and according to the manufacturer's instructions [47]. The overall cellular glycolytic activity was evaluated as previously described [17].

Soft agar colony formation assay
Anchorage-independent cell growth was analyzed by the methods previously describe [14].

Luciferase reporter assay
The LDHA promoter fragments were amplified from a bacterial artificial chromosome (#RP11-107C21) and cloned into a pGL3-basic vector. The promoter constructs were co-transfected with a Renilla luciferase control reporter vector into cells using the Amaxa HMEC Nucleofector kit, program T-24 (Lonza). Bioluminescence activity was assessed by methods previously described [11].

siRNAs and chemical inhibitors
siRNA control and siRNAs were purchased from Sigma for LDHA, CREB, MYC, ATF1, USF1, and SP1 knockdown in vitro. Cells were transfected with siRNAs using Pepmute siRNA Transfection Reagent (SignaGen Laboratories) according to manufacturer's protocol. The final siRNA concentration was 25nM for each transfection. AZD6244 (Selumetinib) was obtained from Selleck Chemicals and AdooQ Bioscience (#A10257). AZD6244 (10μM) was used to treat cells in vitro for at least 5 hours prior to lysate collection from cell culture.

Three-dimensional culture and immunofluorescence staining
All of the three-dimensional cultureand immunofluorescence procedures were using methods described previously [49].

Tumor xenograft studies
All the mouse experiments were carried out in accordance with protocols approved by MD Anderson's Institutional Animal Care Committee. We established breast cancer xenografts by injecting 5x10 5 MCFDCIS. COM.Vec or MCFDCIS.COM.ζ cells orthotopically into the mammary fat pads of 6-week-old female SWISS nu/nu mice. We then divided the mice randomly into two groups, AZD6244 treatment or vehicle control group. Each group contained 5 mice. AZD6244 was suspended in sterile HPMC solution (0.5 % HPMC, 0.1 % Tween 80 in 25 mM citrate) and given to mice through oral gavage daily at a dose of 16 mg/kg body weight or with vehicle control. Tumor growth was monitored with caliper measurements every 4 days for 3 weeks. Tumor volume was calculated using the formula volume = length x (width) 2 /2 [50]. Mice were sacrificed 20 days after treatment initiation and tumors were collected and embedded in paraffin following a routine pathological procedure.

Statistical analyses
Between-group differences were assessed using the Student's t test or ANOVA. P values < 0.05 were considered statistically significant.