Knockdown of KDM1A suppresses tumour migration and invasion by epigenetically regulating the TIMP1/MMP9 pathway in papillary thyroid cancer

Abstract Epigenetic dysregulation plays an important role in cancer. Histone demethylation is a well‐known mechanism of epigenetic regulation that promotes or inhibits tumourigenesis in various malignant tumours. However, the pathogenic role of histone demethylation modifiers in papillary thyroid cancer (PTC), which has a high incidence of early lymphatic metastasis, is largely unknown. Here, we detected the expression of common histone demethylation modifiers and found that the histone H3 lysine 4 (H3K4) and H3 lysine 9 (H3K9) demethylase KDM1A (or lysine demethylase 1A) is frequently overexpressed in PTC tissues and cell lines. High KDM1A expression correlated positively with age <55 years and lymph node metastasis in patients with PTC. Moreover, KDM1A was required for PTC cell migration and invasion. KDM1A knockdown inhibited the migration and invasive abilities of PTC cells both in vitro and in vivo. We also identified tissue inhibitor of metalloproteinase 1 (TIMP1) as a key KDM1A target gene. KDM1A activated matrix metalloproteinase 9 (MMP9) through epigenetic repression of TIMP1 expression by demethylating H3K4me2 at the TIMP1 promoter region. Rescue experiments clarified these findings. Altogether, we have uncovered a new mechanism of KDM1A repression of TIMP1 in PTC and suggest that KDM1A may be a promising therapeutic target in PTC.

mainly acts on lysine residues and regulates either the repression or activation of gene expression. Histone lysine methylation is reversibly modulated by specific lysine methyltransferases (KMTs) and lysine demethylases (KDMs) and is involved in various biological processes and diseases in humans. [4][5][6][7] As the first demethylase discovered, KDM1A is the best characterized KDM. KDM1A is also known as LSD1 and AOF2, 8 and can interact with various protein complexes (eg CoREST, NuRD, RCOR2), receptors (eg oestrogen, androgen, TLX), non-histone proteins (eg p53, E2F1, DNMT1), and transcription factors (eg TLA and SNAIL). 9 KDM1A promotes carcinogenesis progression by removing the methyl groups from methylated histone H3 lysine 4 (H3K4) and H3 lysine 9 (H3K9). Recent studies have shown that KDM1A is overexpressed in multiple malignant tumours, including lung cancer, cervical cancer, oesophageal cancer and ovarian cancer. Elevated KDM1A levels are also associated with tumour stage, histological grade and lymph node metastasis. [10][11][12][13] Furthermore, these studies have suggested that KDM1A might function as an oncogene by regulating the degree of histone methylation of its target gene.
Tissue inhibitors of metalloproteinases (TIMPs) comprise a four-member family: TIMP1, TIMP2, TIMP3 and TIMP4, which bind matrix metallopeptidases (MMPs) in a 1:1 ratio to inhibit their proteolytic activity. 14 These endogenous secreted proteins inhibit all MMPs, and each TIMP targets multiple enzymes. TIMP1 was discovered in the 1980s, and contains 184-194 amino acids and has a molecular weight of 21 kDa. As an important regulator of MMP9, it is functionally relevant, as demonstrated by the association of the compromised TIMP1/MMP9 ratio in prostate cancer and liver cancer. 15,16 However, the function of TIMP1 in PTC is unknown.
Compared with other epigenetic modifications such as DNA methylation and non-coding RNA abnormality, little is known about the role of histone demethylation in PTC. Here, we investigated which KDMs are abnormally expressed in PTC tissues. We found that KDM1A was overexpressed in PTC tissues compared to the adjacent non-cancerous tissues. We then evaluated the relationship between KDM1A and the clinical features of PTC, and detected the migration and invasive ability of PTC cells after KDM1A down-regulation. We report, for the first time, that TIMP1 is a target gene of KDM1A and that KDM1A may regulate TIMP1 via its demethylase activity. Our results could partly clarify the molecular mechanisms of PTC metastasis.

| Patient samples and tissue microarray
We obtained fresh tumour tissue and the corresponding non-cancerous tissue from 60 patients with PTC. After resection, the samples were snap-frozen in liquid nitrogen and stored at −80°C until RNA and protein extraction. For the tissue microarray (TMA), we obtained 61 paired samples of PTC tissue and adjacent non-cancerous tissue and 94 samples of PTC tissue alone from 155 patients.

| Cell culture
The human PTC cell line BCPAP was obtained from DSMZ FBS. All cells were cultured at 37°C in a humidified atmosphere with 5% CO2.     supplemented with 10% FBS. Following 24 hours incubation, the cells were fixed with 4% formaldehyde and stained with 0.1% crystal violet. The cells on the underside of the membrane were counted using Image J software. The experiments were carried out independently in triplicate.

| Western blotting
Total proteins were extracted using a Total Protein Extraction Kit

| Chromatin immunoprecipitation
Chromatin immunoprecipitation (ChIP) assays were performed as indicated in an enzymatic ChIP kit (CST, #9002). Briefly, 3 × 10 7 IHH-4 cells or 1.5 × 10 7 TPC1 cells were crosslinked using 1% formaldehyde. Micrococcal nuclease (0.5 µL) was added per IP and sonicated into 200-500 bp fragments. Approximately 5 μg antibody was added per IP and incubated overnight at 4°C. Rabbit immunoglobulin G (IgG)was used as a control. Protein G agarose beads were added and incubated at 4°C for 2 hours, and then the chromatin and antibodies were eluted at 65°C for 4 hours in IP elution buffer.
The cross-links were reversed, and the DNA was purified using a

| Gelatin zymography
The cells were cultured in high-glucose DMEM supplemented with 10% FBS until approximately 70%-80% confluent, washed twice, and the culture was continued in serum-free DMEM for 24 hours.
The conditioned medium was collected and centrifuged. The conditioned media from all samples were adjusted to contain the same protein concentrations. An 8% acrylamide gel containing gelatin was prepared. After electrophoresis, the gel was incubated in incubation buffer for 24 hours at 37°C. The gel was stained with 0.5% Coomassie brilliant blue R-250 and photographed after elution.

| Metastasis assay in nude mice
Briefly, 1 × 10 7 TPC1 cells of sh-KDM1A transfection and sh-NC transfection were injected intravenously through the tail vein into 4-5-week-old severe combined immunodeficient nude mice. After 6 weeks, the number of tumour nodules formed on the lung surfaces was counted.

| Statistical analysis
All statistical analyses were performed using SPSS 22.0 (IBM, Chicago, IL). The correlation between KDM1A expression and the patients' clinical characteristics were examined using Pearson's chisquare test. The correlation between KDM1A and TIMP1 expression were evaluated using Pearson correlation analysis. Student's t test was used to analyse the comparison of cell migration and invasion, qRT-PCR, and ChIP experiments. A two-sided test was considered statistically significant at P < 0.05.

| KDM1A expression was elevated in PTC and correlated with lymph node metastasis
Initially, to identify histone demethylation modifiers with oncogenic properties in PTC, we assessed the histone demethylation modifiers that may be highly expressed in 16 pairs of PTC tissue and the adjacent non-cancerous tissue using qRT-PCR. KDM1A, KDM5A and KDM7A were up-regulated in the PTC tissues as compared to the non-cancerous tissue ( Figure 1A). Then, we expanded the sample size to 60 pairs of PTC tissue and non-cancerous tissue, and found no difference in KDM5A expression between the paired tissues. RNA interference indicated that KDM7A may not be important for migration and invasion in PTC. These results led to our selection of KDM1A as a primary candidate for subsequent functional analyses.
PTC tissues had increased KDM1A mRNA expression compared to the paired adjacent non-cancerous tissues ( Figure 1B,C). KDM1A protein expression levels were detected from TMA via immunohistochemistry (IHC). As shown in (Figure 1D,E), PTC tissues had significantly elevated KDM1A protein levels (Table 1), especially in tissue from patients with lymph node metastasis. The IHC score was used to determine whether KDM1A expression level was associated with the clinicopathological features of the patients with PTC. As shown in (Table 2), KDM1A positive expression was significantly related to age <55 years (P = 0.019) and lymph node metastasis (P = 0.035). The high KDM1A expression in PTC tissues was confirmed by western blotting using a small amount of fresh tissues ( Figure 1F). As the PTC tissue had higher KDM1A expression than the non-cancerous tissue at both mRNA and protein level, we assessed whether PTC cell lines had up-regulated KDM1A expression. qRT-PCR and western blotting showed that the IHH-4, TPC1 and BCPAP PTC cells expressed higher levels of KDM1A, whereas its expression was lower in the K1 PTC cell line than in the Nthy-ori 3-1 human normal thyroid follicular epithelial cell line ( Figure 1G,H). Taken together, these findings suggest that KDM1A may play an oncogenic role in PTC development.

| KDM1A knockdown inhibited PTC cell migration and invasion
To further understand the function of KDM1A in PTC, we knocked down KDM1A in the IHH-4 and TPC1 KDM1A-overexpressing cell lines using two siRNA sequences, both of which could downregulate KDM1A effectively (Figure 2A

| The TIMP1 gene is a key target gene of KDM1A
To further understand the mechanisms of KDM1A promotion of metastasis in PTC, we detected the expression of MMP2 and MMP9. invasiveness. 18,19 Interestingly, the si-KDM1A groups did not have obviously altered MMP9 mRNA expression, but had significantly decreased MMP9 protein levels. In the si-KDM1A groups, MMP2 was unaltered at both mRNA and protein level as compared with the si-NC group ( Figure 3A,B). Gelatin zymography indicated that the si-KDM1A groups had also significantly repressed MMP9 activation compared with the si-NC groups ( Figure 3C). These results suggest that MMP9 might be regulated post-transcriptionally, at both expression and activation level. So we speculated that TIMPs, the natural inhibitors of the MMPs, may play important roles in the KDM1A-induced effects in PTC cells. We detected the mRNA expression of all TIMPs (ie TIMP1-TIMP4); The results showed that TIMP1 was significantly up-regulated in the si-KDM1A groups in both the IHH-4 and TPC1 cell lines ( Figure 3D,E). In agreement with this finding, KDM1A knockdown also increased TIMP1 protein levels ( Figure 3F). TIMPs not only form tight complexes with MMPs, they also bind to pro-MMPs, 14 suggesting that TIMP1 can repress both the expression and activation of MMP9. Moreover, KDM1A expression was negatively correlated with TIMP1 expression in 60 PTC tissues ( Figure 3G). Therefore, we presume that TIMP1 may be a key target gene of KDM1A and that KDM1A may promote PTC cell migration and invasion by regulating TIMP1 negatively.

| TIMP1 down-regulation partly impaired KDM1A depletion-mediated cell migration and invasion
To confirm whether TIMP1 repression is necessary for KDM1A-induced PTC cell migration and invasion, we co-transfected IHH-4 and

| KDM1A demethylated H3K4me2 at the TIMP1 promoter
The above data indicate that TIMP1 is a potential KDM1A target gene. However, the mechanism of how KDM1A regulates TIMP1 remains to be clarified. KDM1A is a transcriptional repressor because it demethylates the gene activation marker H3K4me2. 20 To understand the mechanism of how KDM1A down-regulation up-regulates TIMP1 expression in PTC cells, we focused on whether KDM1A could act as a TIMP1 repressor by demethylating H3K4me2.

ChIP was performed to determine whether KDM1A bound to TIMP1
and what regions KDM1A occupied. Three primers were designed at +1051(a), +1626(b), +1868(c) of the promoter region. We found that KDM1A was localized at the proximal promoter region (region a) but not the distal promoter region (region b and c) of the TIMP1 gene ( Figure 5B,C). Supporting the finding of KDM1A localization at the proximal promoter region of the TIMP1 gene, KDM1A knockdown decreased KDM1A levels and increased H3K4me2 levels at the same region. Additional ChIP data demonstrated that KDM1A knockdown did not alter H3K9me2 levels at these regions (ie region a, b and c) ( Figure 5D,E). These results suggest that KDM1A directly represses the TIMP1 gene by binding to its promoter and specifically demethylating H3K4me2 at the proximal promoter region. This is consistent with a previous report indicating that the KDM1A binding sites are located at the transcription start regions. 21  node metastases can be found in approximately 60% of cases. 22 Moreover, long-term follow-up has shown that patients with lymph node metastases have higher recurrence rates and lower survival rates. [23][24][25] In recent years, various genomic and epigenetic aberrations related to metastasis have been reported in PTC. 26,27 However, the specific mechanism of how these abnormalities affect PTC remains poorly understood.

| D ISCUSS I ON
Histone modification regulates gene expression through different ways, such as through acetylation, methylation, ubiquitination and phosphorylation. Methylation was the first histone post-translational cer, breast cancer, colon cancer and non-small cell lung cancer. [32][33][34][35][36] In particular, KDM1A is overexpressed in PTC. Kong   can also activate several latent proteinases and angiogenic factors or cytokine receptors, which enhance invasion and metastasis. 39 MMP9 is a cancer promoter in PTC, for example, MMP9 is up-regulated and correlates with tumour diameter, lymph node metastasis, degree of PTC infiltration and clinical stage. 40,41 We found that si-KDM1A reduced MMP9 expression and activation, suggesting that it may promote tumour progression by regulating MMP9.
Intriguingly, we found that while MMP9 mRNA expression did not decrease, MMP9 protein expression was decreased significantly.
This led us to speculate that MMP9 is regulated post-transcriptionally. Accordingly, we detected the expression level of the TIMPs, way. 45 However, TIMP1 has also been reported as an oncogene and is increased in more advanced tumours, and predicts a shorter time to relapse and worse prognosis in endometrial, breast and brain cancer. [46][47][48] These discrepancies may be due to the complicated functions of TIMP1 in different cancers. TIMP1 can not only inhibit cancer by repressing MMP expression and activation, but also promotes cancer via angiogenesis, cell growth promotion and tumour inflammation. 49 These studies all suggest the complicated role TIMP1 plays in cancer development.
Histone H3 demethylation is both positively and negatively associated with tumour metastasis. In fact, KDM1A affects tumour characteristics by regulating H3K4me2 or H3K9me2. 50,51 Here, we found that the global protein levels of H3K4me2 and H3K9me2 were upregulated after si-KDM1A transfection. However, ChIP indicated that only H3K4me2 expression levels were decreased in the si-KDM1A groups. These data demonstrate that KDM1A can repress TIMP1 expression by demethylating H3K4me2 at its promoter regions.
In conclusion, we have identified KDM1A as a metastasis promoter in PTC and that it is associated with lymph node metastasis.
KDM1A enhances MMP9 expression and activation by binding to the promoter region of TIMP1 and demethylating H3K4me2, which represses TIMP1 expression. Our work uncovers a new invasive mechanism in PTC and suggests that KDM1A can be used as a potential therapeutic target in PTC ( Figure 5F).

CO N FLI C T O F I NTE R E S T
Authors declare no conflicts of interest for this article.