Functional Interaction of Histone Deacetylase 5 (HDAC5) and Lysine-specific Demethylase 1 (LSD1) Promotes Breast Cancer Progression

Abstract

We have previously demonstrated that crosstalk between lysine-specific demethylase 1 (LSD1) and histone deacetylases (HDACs) facilitates breast cancer proliferation. However, the underlying mechanisms are largely unknown. Here we report that expression of HDAC5 and LSD1 proteins were positively correlated in human breast cancer cell lines and tissue specimens of primary breast tumors. Protein expression of HDAC5 and LSD1 was significantly increased in primary breast cancer specimens in comparison with matched normal adjacent tissues. Using HDAC5 deletion mutants and co-immunoprecipitation studies, we showed that HDAC5 physically interacted with LSD1 complex through its domain containing nuclear localization sequence and phosphorylation sites. While the in vitro acetylation assays revealed that HDAC5 decreased LSD1 protein acetylation, siRNA-mediated HDAC5 knockdown did not alter the acetylation level of LSD1 in MDA-MB-231 cells. Overexpression of HDAC5 stabilized LSD1 protein and decreased the nuclear level of H3K4me1/me2 in MDA-MB-231 cells, whereas loss of HDAC5 by siRNA diminished LSD1 protein stability and demethylation activity. We further demonstrated that HDAC5 promoted the protein stability of USP28, a bona fide deubiquitinase of LSD1. Overexpression of USP28 largely reversed HDAC5-KD induced LSD1 protein degradation, suggesting a role of HDAC5 as a positive regulator of LSD1 through upregulation of USP28 protein. Depletion of HDAC5 by shRNA hindered cellular proliferation, induced G1 cell cycle arrest, and attenuated migration and colony formation of breast cancer cells. A rescue study showed that increased growth of MDA-MB-231 cells by HDAC5 overexpression was reversed by concurrent LSD1 depletion, indicating that tumor-promoting activity of HDAC5 is an LSD1 dependent function. Moreover, overexpression of HDAC5 accelerated cellular proliferation and promoted acridine mutagen ICR191 induced transformation of MCF10A cells. Taken together, these results suggest that HDAC5 is critical in regulating LSD1 protein stability through posttranslational modification, and the HDAC5-LSD1 axis plays an important role in promoting breast cancer development and progression.

Introduction

LSD1 is the first identified FAD-dependent histone demethylase that has been typically found in association with a transcriptional repressor complex that includes CoREST, HDAC1/2, BHC80, and others (1–4). A role for elevated expression of LSD1 has been implicated in tumorigenesis in various cancers including breast cancer (3, 5–9). Studies from our and other laboratories consistently showed that inhibition of LSD1 hindered proliferation of breast cancer cells (6, 8, 10). Lim et al. reported that LSD1 is highly expressed in ER-negative breast cancers (6). A recent study found that LSD1 is significantly overexpressed in high grade DCIS or IDC versus low/intermediate DCIS (11). These studies point to a tumor promoting role for LSD1 in breast cancer. We were among the first to report the use of small molecule compounds and preclinical treatment strategies that have promise to work through this target in cancer (8, 9, 12). The development of novel LSD1 inhibitors is progressing rapidly. For example, a new generation of (bis)urea/(bis)thiourea LSD1 inhibitors displayed improved potency against LSD1 in cancer cells (13). A newly reported GSK-LSD1 inhibitor exhibited interesting cell type specific inhibition against small cell lung cancer cells in preclinical models. (14).

However, how LSD1 is upregulated in breast cancer and the precise role of LSD1 in breast cancer development are still unclear. Our most recent work showed that siRNA-mediated inhibition of HDAC5 led to a significant increase of H3K4me2, a known substrate of LSD1, suggesting a potential role of HDAC5 in regulating LSD1 activity (10). However, little is known about the precise role of HDAC5 and mechanisms underlying its regulation on LSD1 activity in breast cancer. HDAC5 is an important member of class IIa HDAC isozymes with important functions in transcriptional regulation, cell proliferation, cell cycle progression, and cellular developmental activities (15, 16). HDAC5 has been shown to play important roles in many diseases including cancer (17, 18). In this study, we addressed the following clinically relevant issues that have been understudied: (1) Is elevation of LSD1 expression associated with HDAC5 overexpression during breast cancer development? (2) How is LSD1 regulated by HDAC5 in breast cancer? (3) What is the role of the HDAC5-LSD1 axis in breast cancer initiation, proliferation and metastasis? To answer these questions, we delineated the mechanisms underlying the functional link between LSD1 and HDAC5 in chromatin remodeling and demonstrated that these two important chromatin modifiers closely cooperate to mediate proliferation, cell cycle and metastasis of breast cancer cells.

Results

1. HDAC5 and LSD1 proteins are coordinately expressed in human breast cancer

To study the potential association of HDAC5 and LSD1 in breast cancer, we first examined mRNA levels of HDAC5 and LSD1 in human immortalized normal mammary epithelial MCF10A cells, fully malignant MCF10A–CA1a cells transformed from MCF10A cells with transfection of HRAS (19), and several human breast cancer cell lines. qPCR studies showed that there was no clear association of mRNA expression between HDAC5 and LSD1 in breast cancer cell lines (Figure 1a). The Oncomine-TCGA database showed moderate change of the mRNA level of LSD1 and HDAC5 in IBC (Supplementary Figure 1a and 1b). mRNA levels of both HDAC5 and LSD1 are altered in approximately 6% of breast cancer patients (www.cbioportal.org) without an apparent association with specific subtypes (Supplementary Figure 1c and 1d). However, protein expression of both HDAC5 and LSD1 was significantly elevated in malignant breast cell lines compared with MCF10A (Figure 1b), and protein levels of HDAC5 and LSD1 were positively correlated (Figure 1c). The correlation of HDAC5 and LSD1 protein expression was further validated in 50 primary breast cancers using immunohistochemical staining with validated antibodies (Supplementary Figure 2a and 2b). Chi-square analysis showed a statistically significant correlation between HDAC5 and LSD1 protein expression in these tumors (Figure 1d). Furthermore, IHC analysis showed that breast cancer tissues (n=18) expressed significantly higher level of HDAC5 and LSD1 than matched normal adjacent tissues (n=18) (Figure 1e). The mean H-score for HDAC5 staining in stage 3 breast tumors (n=25) was statistically significantly higher than stage 2 counterparts (n=25). The mean H-score of LSD1 staining for stage 3 tumors was also higher than that of stage 2 tumors with a p-value of 0.07 (Figure 1f). These results were further validated with independent manual H score evaluations by two breast cancer pathologists with moderate interobserver concordance (Supplementary Figure 3a and 3b). Taken together, these findings suggest that HDAC5 and LSD1 proteins are coordinately overexpressed in breast cancer cell lines and tissue specimens.

Figure 1.

Correlated overexpression of HDAC5 and LSD1 protein in breast cancer. (a) The levels of mRNA expression of HDAC5 and LSD1 in breast cancer cell lines versus MCF10A cells (set as fold 1) using real-time qPCR with β-actin as an internal control. (b) Immunoblots with anti-HDAC5 and LSD1 antibodies in indicated cell lines. β-actin protein was blotted as a loading control. (c) Histograms represent the mean protein levels of HDAC5 or LSD1 in three determinations relative to β-actin ± s.d. as determined by quantitative immunoblots. (d) 50 primary human invasive breast tumor samples were immunostained with antibodies against HDAC5 or LSD1. Chi-square study was performed by using median H-scores as the cutoff for high vs low protein expression. (e) Representative HDAC5 and LSD1 staining (200x) in invasive breast carcinoma and adjacent normal tissue specimens from one representative patient. H-scores represent average staining intensity in breast tumors (n=18) versus adjacent normal breast tissue (n=18). (f) Representative HDAC5 and LSD1 staining (200x) in stage 2 and 3 invasive breast carcinoma specimens. H-scores represent average staining intensity in stage 3 breast tumors (n=25) versus stage 2 breast tumors (n=25). * p<0.05, ** p<0.01, *** p<0.001, Student’s t-test.

2. Physical interaction of LSD1 and HDAC5 in breast cancer cells

To address whether LSD1 and HDAC5 physically interact, a co-immunoprecipitation study was carried out in MDA-MB-231 and MCF10A–CA1a cells transiently transfected with pcDNA3.1 or pcDNA3.1-FLAG-HDAC5 plasmids. After immunoprecipitation (IP) with LSD1 antibody, we found that both endogenous and exogenous HDAC5 proteins were co-immunoprecipitated with LSD1 protein (Figure 2a). The interaction between native LSD1 and HDAC5 was further validated in additional breast cancer cell lines (Figure 2b). A similar result was obtained in the reciprocal immunoprecipitation using anti-FLAG antibody to confirm that LSD1 was co-immunoprecipitated with FLAG-HDAC5 (Figure 2c). To precisely map the HDAC5 domain(s) responsible for interaction with LSD1, we expressed a series of HDAC5 deletion mutants engineered in pcDNA3.1-FLAG plasmids in MDA-MB-231 cells (Figure 2d). Immunoprecipitation assays of cells transfected with full length HDAC5 cDNA confirmed the HDAC5-LSD1 interaction. Deletion of an N-terminal myocyte enhancer factor-2 (MEF2) binding domain (HDAC5-Δ1) alone had no impact on HDAC5-LSD1 interaction. However, removal of both the MEF2 domain and nuclear localization sequence (NLS) (HDAC5-Δ2) completely abolished HDAC5-LSD1 interaction. Further deletion of an N-terminal HDAC and nuclear export sequence (NES) (HDAC5-Δ3) and MEF2 domain (HDAC5-Δ4) did not adversely alter LSD1 binding with HDAC5 fragments (Figure 2e). Immunofluorescence studies showed nuclear localization of full length HDAC5, HDAC5-Δ1, HDAC5-Δ3 and HDAC5-Δ4. Depletion of the NLS-containing domain (HDAC5-Δ2) completely blocked HDAC5 nuclear translocation (Figure 2f). In vitro pull-down assays by using His-tag recombinant LSD1 protein incubating with HDAC5 full length or deletion mutants validated that HDAC5 domain containing NLS element is essential for interaction with LSD1 (Supplementary Figure 4).

Figure 2.

HDAC5 and LSD1 physically interact in breast cancer cells. (a) MDA-MB-231 or MCF10A–CA1a cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5 plasmids. Immunoprecipitation (IP) was performed with anti-LSD1 antibody followed by immunoblotting (IB) with anti-LSD1, anti-FLAG or anti-HDAC5 antibodies, respectively. (b) Whole cell lysates were immunoprecipitated with anti-LSD1 antibody followed by IB with anti-HDAC5 and LSD1 antibodies in indicated breast cancer cell lines. IgG was used as negative control. (c) MDA-MB-231 cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5-FLAG plasmids, and IP was performed with anti-FLAG followed by IB with anti-LSD1 and anti-FLAG antibodies, respectively. (d) Schematic representation of full length and deletion mutants of HDAC5-FLAG constructs. (e) FLAG-tagged full-length or deletion mutants of HDAC5 were expressed in MDA-MB-231 cells. Extracts were immunoprecipitated with anti-FLAG antibody, and bound LSD1 was examined by IB using anti-LSD1 antibody. IB with anti-FLAG was used to detect the levels of FLAG-tagged HDAC5 full-length or deletion mutants in IP and input samples (10%). (f) MDA-MB-231 cells were transfected with plasmids expressing FLAG-tagged full-length or deletion mutants of HDAC5 proteins. Immunofluorescence study was performed using anti-FLAG antibody. DAPI was used as a control for nuclear staining. All the experiments were performed three times with similar results.

3. HDAC5 promotes LSD1 protein stability and activity

Next, we examined whether the mRNA or protein levels of HDAC5 and LSD1 were affected by their interaction with each other. Overexpression of HDAC5 in MDA-MB-231 cells failed to alter LSD1 mRNA expression, but led to a significant increase of LSD1 protein expression (Figure 3a and 3b). HDAC5 knockdown by siRNA attenuated LSD1 protein expression without affecting its mRNA level (Figure 3c and 3d). The effect of LSD1 on HDAC5 expression was subsequently assessed using our previously established MDA-MB-231-LSD1-KD cells (10). Depletion of LSD1 exerted no effect on HDAC5 mRNA or protein levels (Figure 3e and 3f). Simultaneous overexpression of pcDNA3.1-HDAC5 with HDAC5 siRNA significantly reversed the decrease of LSD1 (Supplementary Figure 5a). These results suggest that HDAC5 functions as an upstream regulator that governs LSD1 protein stability via posttranslational regulation. Quantitative immunoblots showed that levels of H3K4me1/2 and AcH3K9, the substrates for LSD1 and HDAC5 respectively, were downregulated by HDAC5 overexpression, whereas loss of HDAC5 exerted the opposite effect (Figure 3g; Supplementary Figure 5b), suggesting a critical role of HDAC5 in governing chromatin modifying activity of LSD1. Cycloheximide (CHX) chase assay showed that overexpression of HDAC5 significantly extended LSD1 protein half-life, whereas depletion of HDAC5 by siRNA decreased LSD1 protein half-life in MDA-MB-231 cells (Figure 3h and 3i; Supplementary Figure 5c). To determine whether other recognized LSD1 cofactors or HDACs exert similar effects on LSD1 protein stability, MDA-MB-231 cells were treated with siRNA against several LSD1 complex co-factors (CoREST, HDAC1 and HDAC2) or other class II HDAC isozymes (HDAC 4, 6, 7, 9, 10) respectively. Transfection with siRNA probes effectively knocked down mRNA expression of target genes without affecting their protein levels (Figure 3j; Supplementary Figure 6a). To confirm the qPCR results, quantitative immunoblotting was performed and showed depletion of CoREST led to insignificant change of LSD1 protein stability (Supplementary Figure 6b and 6c). Together, these results strengthen the conclusion that HDAC5 functions as a positive regulator of LSD1 protein in breast cancer cells.

Figure 3.

HDAC5 stabilizes LSD1 protein in breast cancer cells. (a) MDA-MB-231 cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5 for 48 h. mRNA expression of HDAC5 and LSD1 was measured by quantitative real-time PCR with β-actin as an internal control. (b) MDA-MB-231 cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5 plasmids for 48 h. Effect of HDAC5 overexpression on LSD1 protein expression in MDA-MB-231 cells was evaluated by immunoblots with anti-LSD1 and anti-HDAC5 antibodies. (c) MDA-MB-231 cells were transfected with scramble siRNA or HDAC5 siRNA for 48 h. Effect of HDAC5 knockdown on LSD1 mRNA expression was examined by quantitative real-time PCR with β-actin as internal control. (d) Effect of HDAC5 siRNA on LSD1 protein expression in MDA-MB-231 cells. (e) Effect of depletion of LSD1 on mRNA expression of HDAC5 in MDA-MB-231-Scramble or MDA-MB-231-LSD1-KD cells. (f) Effect of LSD1-KD on protein expression of HDAC5 in MDA-MB-231-scramble or MDA-MB-231-LSD1-KD cells. (g) MDA-MB-231 cells were transfected with control vector pcDNA3.1, pcDNA3.1-HDAC5, scramble siRNA or HDAC5 siRNA for 48 h and analyzed by immunoblots for nuclear expression of indicated histone marks. PCNA was used as loading control. (h) Effect of HDAC5 overexpression or siRNA on LSD1 protein half-life in cycloheximide chase study. (i) Measurement of LSD1 half-life using Calcusyn program. (j) Effect of siRNA knockdown of LSD1 cofactors or class II HDACs on LSD1 protein level. All the experiments were performed three times. Bars represent the mean of three independent experiments ± s.d. * p<0.05, ** p<0.01, *** p<0.001, Student’s t-test.

4. HDAC5 regulates LSD1 protein stability through modulation of the LSD1 associated ubiquitination system

Protein ubiquitination assays indicated that HDAC5 overexpression significantly attenuated LSD1 polyubiquitination (Figure 4a), whereas depletion of HDAC5 by siRNA facilitated LSD1 polyubiquitination (Supplementary Figure 7a). Recently, Jade-2 and USP28 were identified as specific E3 ubiquitin ligase and deubiquitinase for LSD1 respectively (20, 21). Our study showing that increase of LSD1 protein expression by Jade-2 siRNA and decrease of LSD1 protein expression by USP28 siRNA in MDA-MB-231 cells confirmed the roles of Jade-2/USP28 as LSD1 ubiquitin ligase/deubiquitinase in breast cancer cells (Figure 4b; Supplementary Figure 7b). qPCR studies demonstrated that mRNA level of either Jade-2 or USP28 was not altered by HDAC5 knockdown or overexpression (Figure 4c). The regulation of HDAC5 on protein expression of Jade-2 or USP28 was subsequently assessed. Due to the lack of highly specific antibody against Jade-2, plasmids expressing Jade-2-FLAG fusion protein were transfected into cells as an alternative approach. MDA-MB-231 and MCF10A–CA1a cells expressing Jade-2-FLAG protein were simultaneously treated with HDAC5 siRNA to evaluate the effect of HDAC5 on Jade-2 protein expression. Immunoblot showed that depletion of HDAC5 did not change the protein level of Jade-2 (Figure 4d). However, overexpression of HDAC5 led to significant increase of USP28 protein expression in both cell lines (Figure 4e). In vitro pull-down assay using His-tag recombinant LSD1 protein incubated with USP28-FLAG protein indicated a direct interaction of HDAC5 and USP28 (Supplementary Figure 4), and HDAC5 overexpression significantly attenuated USP28 polyubiquitination (Supplementary Figure 7c). To understand whether HDAC5 may stabilize LSD1 protein through upregulation of USP28 protein stability, a rescue study was carried out in MDA-MB-231 and MCF10A–CA1a cells using concurrent transfection of HDAC5 siRNA and USP28 expression plasmids, and showed that overexpression of USP28 completely blocked the destabilization of LSD1 by HDAC5 depletion (Figure 4f, Supplementary Figure 7d). In an additional rescue experiment, overexpression of HDAC5 failed to promote LSD1 protein expression when cells were simultaneously treated with USP28 by siRNA (Supplementary Figure 7e). All these data support the notion that HDAC5 stabilizes LSD1 protein by enhancing protein expression of its deubiquitinase.

Figure 4.

HDAC5 regulates LSD1 by altering USP28 stability. (a) MDA-MB-231 cells transfected with pcDNA3.1-FLAG, pcDNA3.1-FLAG-HDAC5 or pcDNA3-HA-ubiquitin plasmids were treated with or without proteasome inhibitor 10µM MG132 for 10 h followed by immunoprecipitation (IP) using LSD1 antibody and immunoblots with anti-HA, LSD1 or HDAC5 antibodies. (b) Effect of siRNA of Jade-2, USP28 and HDAC5 on LSD1 protein expression in MDA-MB-231 cells. Results represent the mean of three independent experiments ± s.d. *** p<0.001, Student’s t-test. (c) MDA-MB-231 cells were transfected with scramble siRNA, HDAC5-siRNA, control vector pcDNA3.1, or pcDNA3.1-HDAC5 plasmids for 48 h. mRNA expression of Jade-2 and USP28 was measured by quantitative PCR. β-actin was used as an internal control. (d) MDA-MB-231 or MCF10A–CA1a cells were simultaneously transfected with pcDNA3.1-FLAG-Jade-2 and HDAC5 siRNA for 48 h and subjected to immunoblots with anti-HDAC5 or Jade-2 antibodies. β-actin was used as loading control to normalize target protein levels. (e) After MDA-MB-231 or MCF10A–CA1a cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5 plasmids for 48 h, immunoblotting was performed for expression of HDAC5 and USP28. (f) MDA-MB-231 or MCF10A–CA1a cells were transfected with scramble or HDAC5 siRNA alone, or in combination with pDZ-USP28 for 48 h. Whole cell lysates were analyzed for protein levels of HDAC5, USP28 and LSD1. β-actin was used as loading control to normalize target protein levels. The experiments were performed three times with similar results.

To examine whether interaction of HDAC5 with LSD1/USP28 complex deacetylates LSD1 or USP28, in vitro protein acetylation assays was first carried out by incubating GST-tagged recombinant HDAC5 protein with cellular pull-down of LSD1-FLAG or USP28-FLAG by IP, and immunoprecipitates of IgG was incubated with recombinant HDAC5 protein as negative control of assays (Figure 5a). Bulk histone was used as control substrate (Supplementary Figure 8). Quantitative immunoblots using antibody against pan-acetylated lysine showed that HDAC5 reduced acetylation level of LSD1 without altering the acetylation status of USP28 (Figure 5a and 5b). Next, the in vivo effect of HDAC5 depletion on LSD1 acetylation was investigated in MDA-MB-231 cells transfected with scramble or HDAC5 siRNAs. After immunoprecipitation with LSD1 antibody or IgG (negative control), immunoblotting was performed and the results showed that expression levels of both total LSD1 protein and acetylated LSD1 protein were decreased by HDAC5 depletion (Figure 5c). Quantitative immunoblots indicated that the relative acetylation level of LSD1 was not statistically altered by HDAC5 siRNA in MDA-MB-231 cells (Figure 5d). AcetylH3K9 was used as control of substrate and its expression was increased by HDAC5 siRNA (Figure 5c). These results suggest that inhibition of HDAC5 alone is not sufficient enough to increase LSD1 acetylation in breast cancer cells.

Figure 5.

Effect of HDAC5 on protein acetylation of LSD1/USP28 and transcription of LSD1 target genes. (a) The immunoprecipitates of FLAG-M2 agarose from MDA-MB-231 cells overexpressing FLAG-tagged USP28 or FLAG-tagged LSD1 were used as substrates for protein deacetylation assay. IgG was used as negative control. Active or heat inactivated recombinant human GST-tagged HDAC5 protein were mixed with immunoprecipitates and incubated at 37°C for 6 h as described in “Materials and Methods”. The reactions were then subjected to immunoblots with anti-acetyl lysine antibody. FLAG-tagged USP28 or LSD1 proteins were probed with anti-FLAG antibody. HDAC5-GST protein was probed with anti-HDAC5 antibody. (b) Histograms represent the means of levels of acetyl-LSD1, acetyl-USP28 and acetyl-histone determined by quantitative immunoblotting using infrared immunoblotting detection and analysis. (c) MDA-MB-231 cell transfected with scramble or HDAC5 siRNAs for 48 h. LSD1 or IgG antibodies were added to cell lysate. Immunoprecipitation (IP) was performed with anti-LSD1 antibody followed by immunoblotting with anti-acetyl lysine and anti-LSD1 antibodies, respectively. Effect of HDAC5 siRNA on Acetyl-H3K9 protein expression in MDA-MB-231 cells was examined by immunoblotting with anti-acetyl-H3K9 antibody. (d) Histograms represent the means of relative levels of acetyl-LSD1 determined by quantitative immunoblotting using infrared immunoblotting detection and analysis. (e) mRNA expression of indicated genes in MDA-MB-231 cells transfected with scramble siRNA or HDAC5 siRNA. Data are means ± s.d. of three independent experiments. (f) Quantitative ChIP analysis was used to determine the occupancy by acetyl-H3K9, H3K4me2, LSD1, and HDAC5 at promoters of p21 or CLDN7 in MDA-MB-231 cells transfected with scramble or HDAC5 siRNA. *p<0.05, **p<0.01, *** p<0.001, Student’s t-test.

5. Inhibition of HDAC5 reactivates expression of LSD1 target genes

In cancer cells, amplified LSD1 expression is frequently associated with abnormal suppression of key tumor suppressor genes (TSGs) (3, 22). We next examined whether expression of LSD1 target TSGs could be reactivated following HDAC5 inhibition. Loss of expression of CDK inhibitor p21 and epithelial marker claudin-7 (CLDN7) has been reported to be associated with an aggressive phenotype of breast cancer (23, 24). The transcription activity of p21 and CLDN7 has been found to be suppressed by enhanced activity of LSD1 in breast cancer (6, 25). Transfection of HDAC5 siRNA resulted in significantly increased mRNA expression of p21 and CLDN7 in MDA-MB-231 cells (Figure 5e). Quantitative chromatin immunoprecipitation (qChIP) assays revealed that depletion of HDAC5 decreased occupancy of both HDAC5 and LSD1, and increased enrichment of H3K4me2 and acetylH3K9 at the promoters of both genes (Figure 5f). These data suggest that transcriptional de-repression of these genes lies largely in the cooperation between HDAC5 and LSD1 at key active histone marks.

6. Inhibition of HDAC5-LSD1 axis hinders breast cancer proliferation and metastasis

To explore the functional role of the HDAC5-LSD1 axis in regulating breast cancer development, stable knockdown of HDAC5 mRNA (HDAC5-KD) was generated in MDA-MB-231 and MCF10A–CA1a cells by infection with shRNA lentiviral particles. Similar to the effect of transient inhibition of HDAC5 by siRNA, stable knockdown of HDAC5 expression significantly reduced LSD1 protein expression in two independent HDAC5-KD clones (Figure 6a). Loss of HDAC5 in both clones hindered cell proliferation and colony formation in soft agar (Figure 6b and 6c). Flow cytometry analysis showed that inhibition of HDAC5 resulted in a greater fraction of cells accumulated at G1 phase and reduction of the S-phase cell fraction (Figure 6d; Supplementary Figure 9). Moreover, loss of HDAC5 attenuated motility and invasion of MDA-MB-231 cells in a Boyden chamber assay (Figure 6e). A rescue experiment indicated that HDAC5 overexpression promoted growth of MDA-MB-231-Scramble cells, but failed to alter the growth of MDA-MB-231-LSD1-KD cells (Figure 6f). An additional rescue study revealed that LSD1 overexpression rescued growth inhibition by HDAC5 depletion in MDA-MB-231-HDAC5-KD cells (Figure 6g). Taken together, these results demonstrate that tumor promoting activity of HDAC5 is dependent on LSD1 activity in breast cancer cells.

Figure 6.

HDAC5-LSD1 axis is implicated in breast cancer progression. (a) Depletion of HDAC5 by shRNA lentivirus infection downregulated LSD1 protein expression in MDA-MB-231 and MCF10A–CA1a cells. (b) Scramble shRNA and HDAC5-KD cells were analyzed for growth and viability by crystal violet assays. (c) Soft agar colony formation for HDAC5-KD and scramble control of MDA-MB-231 and MCF10A–CA1a cells. (d) Scramble shRNA and HDAC5-KD cells were harvested and stained for DNA with propidium iodide for flow cytometric analysis. The fractions corresponding to G1, S and G2/M phases of the cell cycle are indicated. (e) Boyden Chamber transwell migration assays for cell invasion for MDA-MB-231-Scramble or MDA-MB-231-HDAC5-KD-1 cells. (f) MDA-MB-231-Scramble or MDA-MB-231-LSD1-KD cells were transfected with control vector pcDNA3.1 or pcDNA3.1-HDAC5 for 5 days and crystal violet assays for growth were carried out. (g) MDA-MB-231-Scramble or MDA-MB-231-HDAC5-KD cells were transfected with empty or pReceiver-LSD1 expression plasmids for 5 days and crystal violet assays for growth were carried out. Bars represent the means of three independent experiments ± s.d. * p<0.05, ** p<0.01, *** p<0.001, Student’s t-test.

7. Overexpression of HDAC5 promotes mutagen-induced tumorigenic development in MCF10A cells

To address whether enhanced interaction between HDAC5 and LSD1 is a critical epigenetic alteration driving tumorigenic transformation of breast cancer, we generated two MCF-10A cell lines overexpressing HDAC5 (MCF10A–HDAC5). Stable overexpression of HDAC5 in MCF10A cells increased LSD1 protein level and promoted cell proliferation of both clones (Figure 7a and 7b), indicating a growth-promoting role for HDAC5 in MCF10A cells. Inhibition of LSD1 by shRNA significantly hindered MCF10A growth and reversed the growth promotion mediated by HDAC5 overexpression, suggesting that HDAC5 promotes MCF10A growth in an LSD1 dependent manner (Figure 7c; Supplementary Figure 10). To evaluate if MCF10A–HDAC5 cells have altered susceptibility to tumorigenesis, MCF10A–Vector and MCF-10A–HDAC5 cells were cultured for 7 months in medium containing 500ng/ml ICR191. ICR191 generates genomic instability and genetic variability, and has been successfully used to induce epithelial cell transformation in several models including MCF-10A (26, 27). MCF10A–HDAC5 cells were subsequently tested for the capacity of anchorage-independent growth in soft agar for 4 weeks. Soft agar colony formation study demonstrated that ICR191 treatment improved the ability of MCF10A cells to form growing colonies, and overexpression of HDAC5 significantly promoted ICR191-induced colony formation in MCF10A cells (Figure 7d). To determine the role of LSD1 in HDAC5 enhanced tumorigenic transformation induced by ICR191, scramble control and LSD1 shRNA lentivirus particles were infected into MCF10A–Vector or MCD10A–HDAC5 cells which had been treated with ICR191 for 7 months, and soft agar growth assays showed that loss of LSD1 in MCF10A–HDAC5 cells significantly abolished cellular ability in colony formation (Figure 7e). A model illustrating the role of HDAC5-LSD1 axis in breast cancer development is proposed based on the above findings (Figure 7f).

Figure 7.

Effect of HDAC5 on growth and mutagen-induced tumorigenic transformation in MCF10A cells. (a) pcDNA3.1 or pcDNA3.1-HDAC5 transfected MCF10A cells (clone 1 and 2) were analyzed for protein levels of HDAC5 and LSD1 by immunoblots with anti-HDAC5 and anti-LSD1 antibodies. (b) Crystal violet assay for growth of MCF10-A stably transfected with control vector or pcDNA3.1-HDAC5 plasmids. (c) MCF10A–Vector-1 or MCF10A–HDAC5-1 cells were infected with scramble or LSD1 shRNA lentivirus particles for 5 days followed by crystal violet assays for growth. (d) MCF10A cells transfected with pcDNA3.1 or pcDNA3.1-HDAC5 plasmids were treated with DMSO or 500 ng/ml ICR191 for 7 months followed by soft agar colony formation assays. (e) After treatment with 500 ng/ml ICR191 for 7 months, MCF10A–HDAC5 cells were infected with scramble control or LSD1 shRNA lentivirus particles and soft agar colony formation assay was carried out. (f) Proposed model of the role of HDAC5-LSD1 axis in breast cancer development. Bars represent the means of three independent experiments ± s.d. ** p<0.01, *** p<0.001, Student’s t-test.

DISCUSSION

High levels of HDAC5 have been found to be associated with poor survival in multiple cancer types (28, 29). LSD1 overexpression has been reported to be a poor prognostic factor in basal-like breast cancer, a subtype with aggressive clinical characteristics (6, 30). In this study, IHC analysis showed that breast cancers expressed higher levels of HDAC5 compared to the matched normal adjacent breast tissue. Importantly, our study found a positive correlation between HDAC5 and LSD1 proteins in breast tumor cell lines and patient tissue specimens. Increased expression of HDAC5 and LSD1 is correlated with higher stage of breast cancer in our exploratory study. These findings suggest that the coordinated overexpression of HDAC5 and LSD1 may serve as potential novel prognostic markers as well as possible therapeutic targets for breast cancer. More robust studies will be necessary to understand the precise role of elevated protein expression levels of HDAC5 and LSD1 in the risk stratification of breast cancer patients.

LSD1 protein stability is controlled by several posttranslational modifications such as ubiquitination and methylation (20, 21, 31). However, the precise mechanism of how LSD1 protein stability is regulated is still not understood. A previous study reported that stable depletion of CoREST facilitated LSD1 degradation in HeLa cells (32). However, siRNA-mediated knockdown of CoREST alone in breast cancer cells failed to destabilize LSD1 protein, suggesting additional layers of control of LSD1 protein stability are required in breast cancer. In this study, we observed for the first time that LSD1 protein stability is promoted by HDAC5. We further found that the HDAC5 domain containing NLS is essential for LSD1-HDAC5 interaction. The NLS element provides docking sites for 14–3-3 chaperone binding and has been shown to be critical for HDAC5 import into the nucleus and the regulation of its repressor activity (17, 33). Although an in vitro assay demonstrated that HDAC5 reduced LSD1 acetylation, HDAC5 siRNA treatment in breast cancer cells failed to alter acetylation of LSD1 protein. Our in vivo results suggest that LSD1 acetylation is likely regulated by a large complex that may involve additional protein deacetylases or cofactors. Further studies are needed to identify the regulatory complex and clarify the precise role of HDAC5 in regulation of LSD1 acetylation in breast cancer cells.

Our studies revealed that HDAC5 regulates LSD1 via enhancement of the protein stability of deubiquitinase USP28. High expression of USP28 has been found to promote the progression of breast and colon cancers (20, 34). Importantly, USP28 has been reported to deubiquitinate important tumor growth regulators such as c-Myc and TP53BP1 that are involved in MYC proto-oncogene stability and DNA damage response checkpoint regulation respectively (35, 36). Our pilot microarray study revealed that inhibition of the HDAC5-LSD1 axis down-regulates c-Myc expression (data not shown). Sen et al. recently reported that HDAC5 is a key component in the temporal regulation of p53-mediated transactivation (37). All of these findings imply an interaction of HDAC5/LSD1 axis and USP28-associated ubiquitin proteasome system in regulating downstream targets involved in tumor development. USP28 has been well-characterized for its role in promoting tumorigenesis, and thus is a potential candidate target in cancer therapy. Given the current inability to use drugs to directly target USP28–driven cancer proliferation, our study suggests a novel alternative approach of targeting USP28 stability by development of HDAC5-specific inhibitors in cancer.

Our findings provide supportive evidence showing that HDAC5 control of cell proliferation is largely dependent on LSD1 stabilization. Furthermore, in this study, we showed that non-transformed MCF10A cells overexpressing HDAC5 significantly promoted ICR191-induced transformation of MCF10A cells. The overexpressed HDAC5 is consistently associated with upregulated LSD1 protein expression over the entire course of transformation induction. These data indicate that enhanced crosstalk between HDAC5 and LSD1 may represent a critical mechanism contributing to breast tumorigenesis. HDAC inhibitors (HDACi) hold great promise for cancer therapy. Despite the promising clinical results produced by HDACi in treatment of hematological cancers such as T cell lymphoma, no apparent clinical evidence indicates that HDAC inhibitors work effectively as a monotherapy against solid tumors including breast tumors (38–41). From a clinical perspective, our novel findings have significance for design and development of novel combination strategies targeting HDAC5-LSD1 axis as an alternative approach for improvement of therapeutic efficacy of HDAC inhibitors in breast cancer.

As summarized in Figure 7f, we show for the first time that LSD1 protein stability is promoted by HDAC5 through the LSD1 associated ubiquitin-proteasome system, confirming that the regulation of LSD1 by HDAC5 is a posttranslational event. Our novel findings also provide supportive evidence that an orchestrated interaction between HDAC5 and LSD1 is a critical epigenetic mechanism to suppress transcriptional activities of important tumor suppressor genes that may contribute to breast cancer development.

Materials and methods

Reagents and cell culture conditions

MDA-MB-231, MDA-MB-468, MCF-7, T47D, HCC-202 and SK-BR-3 cell lines were obtained from the ATCC/NCI Breast Cancer SPORE program. MCF10A–parental and MCF10A–CA1a cells were gifts from Dr. Saraswati Sukumar (Johns Hopkins University). Cells were cultured in growth medium as described previously (10, 42).

Tissue Microarrays (TMAs) and immunohistochemistry

TMAs (US Biomax, Rockville, MD) were stained using LSD1 or HDAC5 antibodies. Standard staining procedure for paraffin sections was used for IHC according to manufacturer’s recommendations (Vector Labs, Inc., Burlingame, CA). Monoclonal antibodies were used for detection of LSD1 (1:800; Cell Signaling, Danvers, MA) and HDAC5 (1:100; Santa Cruz). The staining was visualized using diaminobenzidine, and quantitated using IHC Profiler, an ImageJ plugin (43). H-scores were calculated as previously described (44). The manual scoring of H-scores was also carried out by two breast cancer pathologists.

Plasmid construction and stable transfection

Plasmids pcDNA3.1(+)-FLAG, pcDNA3.1(+)-FLAG-HDAC5, and pDZ-FLAG-USP28 were purchased from Addgene (Cambridge, MA). pReceiver-FLAG-LSD1 was obtained from Gene Copoeia (Rockville, MD). A FLAG-tagged ORF cDNA clone for Jade-2 was purchased from GenScript (Piscataway, NJ). pcDNA3-HA-ubiquitin was obtained from Dr. Yong Wan (University of Pittsburgh). HDAC5 deletion mutants were engineered into pcDNA3.1(+)-FLAG-HDAC5 by PCR with primers shown in Table S1. HDAC5-Δ2 was constructed by digesting full length plasmids with SacII from amino acid 61 to 489. Stable transfection was carried out using Lipofectamine 3000™ transfection reagent (Life Technologies, Grand Island, NY), and colonies were selected with 800 µg/ml G418.

siRNA and shRNA treatment and stable cell line generation

Pre-designed siRNA and non-targeting scramble siRNA (Santa Cruz) were transfected into cells following the manufacturer’s protocol. Cells were harvested 48 h post-transfection for further analysis. Scramble control, LSD1-specific or HDAC5-specific shRNA lentiviral particles (Santa Cruz) were infected into cells according to manufacturer’s protocol. Cells were treated with 10µg/ml puromycin 72 h after infection. Single colonies were analyzed for expression of LSD1 or HDAC5 via immunoblots.

RNA extraction and qPCR

Total RNA extraction and cDNA synthesis used the methods described previously (10). Quantitative real-time PCR was performed on the StepOne real-time PCR system (Life Technologies). All of the TaqMan® Gene Expression Assays were predesigned and obtained from Life Technologies.

Western blotting

Western blotting was performed as previously described (12, 45, 46). Antibodies used in this study were shown in Table S2. Membranes were scanned with Li-Cor BioScience Odyssey Infrared Imaging System (Lincoln, NE).

Crystal violet, and cell invasion assays

Crystal violet proliferation assays were performed as described in our previous study (47). The invasive capability of breast carcinoma cells was tested with Millipore QCM™ 24-Well invasion assay Kit (Merck KGaA, Germany) according to manufacturer’s protocol.

Soft agar colony formation assay

1.2 % Bacto-agar (BD Biosciences, Franklin Lakes, NJ) was autoclaved and mixed with growth medium to produce 0.6 % agar. The mixture was quickly plated and solidified for 45 min. Cells were suspended in 0.6 ml 2x growth medium and mixed gently with 0.6 ml 0.8 % agar /medium. 1 ml of cells with 0.4% agar/medium mixture was added onto plate for solidification. Colony formation was examined using stereo microscopy and analyzed (CellSens Dimension, Olympus).

Flow cytometry analysis

Cells were harvested and fixed with 70% ethanol. The cell pellet was then treated with 1% TritonX-100. Cells were subsequently resuspended in 50 µg/ml propidium iodide (Sigma) containing RNaseI (Roche, Indianapolis, IN) followed by analysis on the LSR II XW4400 workstation (BD Biosciences).

Immunofluorescence

48 h after transfection, cells were fixed with 4% paraformaldehyde and incubated with primary antibodies (1:250) overnight at 4 °C. After washing, cells were incubated with fluorescence-labeled secondary antibody (1:100). After washing, coverslips were placed on a glass slide using UltraCruz™ mounting medium (Santa Cruz) before fluorescence microscope examination.

Immunoprecipitation, ubiquitination and protein half-life assays

The cell lysate was obtained by using immunoprecipitation lysis buffer as described previously (48). LSD1 or IgG antibodies were added to cell lysate. Protein G-plus agarose beads (Santa Cruz) or Flag-M2 affinity gel were collected and subjected to immunoblotting. HA-Ubiquitin, pcDNA3.1-Flag-HDAC5 or empty vector plasmids were co-transfected into cells for 38 h. Cells were then treated with 10 µM MG-132 for 10 h and harvested for immunoprecipitation assay with protein G-plus agarose beads. For half-life studies 48 h after transfection with pcDNA3.1-HDAC5 or HDAC5-siRNA, cells were treated with 100 µg/ml cycloheximide and then harvested at indicated times for immunoblotting.

Protein acetylation assay

The immunoprecipitates of FLAG-M2 agarose from MDA-MB-231 cells overexpressing FLAG-tag USP28 or FLAG-tag LSD1 were used as substrates for protein deacetylation assay. Pull-down of IgG was used as negative control. 0.25 µg of recombinant human GST-tagged HDAC5 protein (Creative BioMart, NY, NY) was mixed with 30 µl immunoprecipitates or 1.5 µg bulk histone at 37°C for 6h in a buffer containing 40 mM Tris-HCl (pH 8.0), 2.5 mM MgCl2, 50 mM NaCl, 2 mM KCl, 0.5mM DTT, 1mM EDTA and protease inhibitor. The reactions were then subjected to immunoblots with anti-acetyl lysine antibody (EMD Millipore, Billerica, MA). FLAG-tagged USP28 or LSD1 and bulk histone were probed with anti-FLAG antibody or H3 antibody as loading control. Inactive HDAC5-GST protein was used as negative control by heating recombinant protein at 95°C for 5 min. In vivo protein acetylation assay was performed using cell lysate of MDA-MB-231 cell transfected with scramble and HDAC5 siRNAs. LSD1 or IgG antibodies were added to cell lysate. Protein G-plus agarose beads (Santa Cruz) were collected and subjected to immunoblotting with anti-acetyl lysine or LSD1 antibodies.

Chromatin immunoprecipitation

Chromatin immunoprecipitation assay was performed as described previously (12). Primary antibodies against HDAC5, LSD1, H3K4me2 and acetyl-H3K9 were used as indicated for immunoprecipitation of protein–DNA complexes. PCR primer sets used for amplification of precipitated fragments were shown in Table S1. Input DNA was used for normalization.

Statistical analysis

Data were represented as the mean ± standard deviation of the mean (s.d.) of three independent experiments. The quantitative variables were analyzed by two tailed Student's t-test. Chi-square study was used to assess the correlation between HDAC5 and LSD1 protein expression by using median H-scores as the cutoff for high vs low protein expression. P-value<0.05 was considered statistically significant for all tests. Statistical analyses were performed using GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA).