Open Access

Promoter methylation of TRIM9 as a marker for detection of circulating tumor DNA in breast cancer patients

  • Chieko Mishima1,
  • Naofumi Kagara1Email author,
  • Saki Matsui1,
  • Tomonori Tanei1,
  • Yasuto Naoi1,
  • Masafumi Shimoda1,
  • Atsushi Shimomura1,
  • Kenzo Shimazu1,
  • Seung Jin Kim1 and
  • Shinzaburo Noguchi1
SpringerPlus20154:635

https://doi.org/10.1186/s40064-015-1423-7

Received: 8 October 2015

Accepted: 9 October 2015

Published: 22 October 2015

Abstract

The aim of the present study was to investigate the promoter methylation status of TRIM9 in breast cancer and to determine the presence of TRIM9-methylated circulating tumor DNA (ctDNA) in plasma. Bisulfite sequencing with a next generation sequencer showed TRIM9 promoter methylation in 92 % (11/12) of breast cancer cell lines (BCCs) and 68 % (13/19) of breast tumor tissues but not in any normal breast tissues (0/19). Methylation ratio of TRIM9 was significantly lower in basal type (9 %, n = 23) than luminal A (69 %, n = 29, P = 0.0003). Quantitative RT-PCR of BCCs disclosed an inverse correlation between TRIM9 mRNA expression and methylation ratio. TRIM9 methylated ctDNA in plasma was detected in 18 % (10/56) of metastatic breast cancer patients but not in any of 60 healthy controls. These results indicate that TRIM9 promoter hypermethylation, which suppresses TRIM9 mRNA expression, occurs in a significant proportion of breast tumors, and that TRIM9-methylated ctDNA thus may serve as a tumor marker for breast cancer.

Keywords

Breast cancer TRIM9 Methylation Biomarker

Background

The promoter methylation of tumor suppressor genes is one of the most common events in carcinogenesis and has been detected in various malignant diseases including breast cancer. Recent studies have also revealed that tumor-specific gene methylation can be detected in the circulating tumor DNA (ctDNA) of cancer patients and methylated ctDNA is considered to be a promising biomarker. Several genes including GSTP1 (Glutathione S-transferase P1), RASSF1A (Ras association domain family 1A), and RARβ2 (Retinoic acid receptor β2) have been identified as methylated genes in breast cancer (Yamamoto et al. 2012; Arai et al. 2006; Miyake et al. 2012) but each of these markers is not always specific to breast cancer and several markers have been used in various combinations. Therefore, the need has arisen for methylation markers which are more specific to breast cancer.

TRIM9 belongs to the TRIM (tripartite motif-containing protein) family which has been identified as an ubiquitin ligase (E3) and plays important roles in various cellular processes (Berti et al. 2002). The TRIM family consists of over 70 members, several of which, i.e., TRIM8, 13, 19, 24, 25, 27, 28, 29, 31, 32, 33, 40 and 69, are known to be involved in oncogenesis or tumor progression by affecting specific signal pathways such as RARα and p53 (Hatakeyama 2011). Specifically for breast cancers, TRIM 24, 25 and 27 have been shown to be significant for breast cancer prognosis, such as facilitation of the ubiquitination of estrogen receptors or HER2 gene amplification (Hatakeyama 2011; Tsai et al. 2010; Chambon et al. 2011; Suzuki et al. 2005; Cao et al. 1996). TRIM9 protein is known as a brain-specific E3 ligase expressed in the human brain neurons and associated with neurological disorders such as Parkinson’s disease, Alzheimer’s disease, epilepsy and stroke (Tanji et al. 2010; Winkle et al. 2014; Shi et al. 2014). However, there have been no reports on possible correlation between TRIM9 and carcinogenesis.

Using the Illumina Human Methylation 450 database, we found TRIM9 is specifically methylated in breast cancer tissues. The aim of the present study was therefore first to investigate whether methylation of TRIM9 promoter is associated with its gene expression in breast cancer cells, and second to clarify the clinicopathological characteristics of TRIM9 methylated breast tumors. We analyzed the methylation of TRIM9 promoter by means of next generation sequencing (NGS), which yields a quantitative methylation ratio within a broad CpG area. Lastly, we examined whether TRIM9 methylated ctDNA can be detected in plasma of breast cancer patients and explored its utility as a novel blood biomarker for breast cancer diagnosis.

Methods

Extraction of targeted gene

We used a common methylation database, Illumina Human Methylation 450, provided by the Cancer Genome Atlas (TCGA) Data Portal, National Cancer Institute, Washington, D.C., USA (http://cancergenome.nih.gov/) to find 90 cases which included methylation data of both primary breast carcinoma and normal breast tissue (https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix.htm?mode=ApplyFilter&showMatrix=true&diseaseType=BRCA&tumorNormal=TN&tumorNormal=T&tumorNormal=NT&platformType=2&platformType=42). We downloaded the β-score calculated from about 485,000 CpG sites of 90 paired cancerous and non-cancerous breast tissues, and 547 probes met all of three criteria for inclusion in our study, that is, methylation ratio in cancer tissues >45 %, methylation ratio in normal tissues <5 %, and area under the ROC curve >0.85. Next, we used the t test to compare the methylation status of breast cancer and other cancers. Finally, ten probes showing the highest methylation ratio specific to breast cancers qualified as candidates, and among these we decided to target TRIM9, since this was the only probe as yet not known to be associated with breast cancers.

Patients and breast tumor samples

Study I

Nineteen pairs of tumor tissues and normal tissues were obtained at surgery between 2001 and 2004 from primary breast cancer patients who had received no preoperative chemotherapy or hormonal therapy. The clinicopathological characteristics of these patients are summarized in Additional file 1: Table S1. Normal tissues were obtained from a quadrant other than the one harboring cancer. Tissue samples were snap frozen in liquid nitrogen and kept at −80 °C until use.

Study II

Stage II or III primary breast cancer patients (n = 107), who had been treated with neoadjuvant chemotherapy (NAC) consisting of paclitaxel (80 mg/m2) weekly for 12 cycles followed by 5-FU (500 mg/m2), epirubicin (75 mg/m2) and cyclophosphamide (500 mg/m2) every 3 weeks for four cycles at Osaka University Hospital between 2004 and 2009, were retrospectively included in this study. Each patient underwent vacuum assisted biopsy (VAB) of the tumors, and the tumor samples were snap frozen in liquid nitrogen and kept at −80 °C until use. Histological grade, ER, PR, and HER2 status were determined as described in a previous report of ours (Miyake et al. 2012). Ki67 was classified as “high” when ≥20 % of tumor cells were immunohistochemically positive (clone; MIB-1). Pathological complete response (pCR) was defined as no evidence of invasive cancer components in breast irrespective of any axilla lymph nodes metastases. Intrinsic subtypes were determined by means of DNA microarray using the PAM50 method as previously described (Naoi et al. 2011; Parker et al. 2009). The clinicopathological characteristics of these patients are summarized in Table 1. These studies were approved by the Ethical Review Board of Osaka University Hospital and the Research Ethics Committee of Osaka University, and informed consent was obtained from each patient before sampling.
Table 1

Comparison of TRIM9 methylation ratio with various clinicopathological parameters of breast tumors

Characteristics

Total

TRIM9

Methylation ratio mean ± SE

P value*

All cases

107

  

 Age (years)

  <50

49

11.2 ± 1.54

0.052

  ≥50

58

7.44 ± 1.10

 

 Menopausal status

  Pre

51

10.7 ± 1.51

0.117

  Post

56

7.73 ± 1.14

 

 Tumor size

  T1+2

84

8.97 ± 1.04

0.709

  T3+4

23

9.83 ± 2.19

 

 Lymph node metastasis

  Negative

30

10.9 ± 2.27

0.338

  Positive

77

8.48 ± 0.96

 

 Stage

  II

88

9.02 ± 1.02

0.768

  III

19

9.75 ± 2.50

 

 Histological type

  IDC

97

8.66 ± 0.99

0.104

  ILC

10

13.9 ± 2.68

 

 Estrogen receptor

  Negative

42

3.47 ± 0.56

<0.0001

  Positive

65

12.8 ± 1.32

 

 Progesterone receptor

  Negative

65

6.93 ± 1.01

0.005

  Positive

42

12.6 ± 1.70

 

 HER2 receptor

  Negative

76

9.07 ± 1.10

0.893

  Positive

31

9.35 ± 1.85

 

 TNBC

  No

82

11.2 ± 1.13

<0.0001

  Yes

25

2.39 ± 0.32

 

 Subtype (IHC)

  LumA

51

12.3 ± 1.42

<0.0001

  LumB

14

14.6 ± 3.40

 

  HER2

17

5.06 ± 1.21

 

  TN

25

2.39 ± 0.32

 

 Subtype (PAM50)

  LumA

29

13.7 ± 1.89

<0.0001

  LumB

21

10.1 ± 2.24

 

  HER2

16

9.90 ± 2.65

 

  Basal-like

23

3.22 ± 0.84

 

  Normal-like

18

7.70 ± 2.24

 

 Histological grade

  1+2

86

9.89 ± 1.09

0.112

  3

21

6.12 ± 1.64

 

 Ki67

  Low (<20 %)

44

10.2 ± 1.31

0.366

  High (≥20 %)

62

8.43 ± 1.33

 

  Unknown

1

  

 Clinical response

  No CR

70

8.62 ± 1.13

0.434

  CR

37

10.2 ± 1.70

 

 Histological response

  Grade 1, 2a

58

11.4 ± 1.37

0.006

  Grade 2b, 3

49

6.45 ± 1.17

 

 Pathological response

  No pCR

74

11.0 ± 1.20

0.001

  Pcr

33

5.02 ± 1.15

 

 Recurrence

  No

90

9.88 ± 1.08

0.007

  Yes

17

5.30 ± 1.12

 

* t test

DNA extraction and sodium bisulfite treatment

Total DNA from cell lines was isolated using TRIzol® reagent (Invitrogen, Carlsbad, CA, USA) and total DNA from the breast tissues was extracted using the DNeasy® Blood and Tissue Kit (QIAGEN, Valencia, CA, USA). 1 μg of genomic DNA was then subjected to sodium bisulfite treatment with the EpiTect® Bisulfite Kit (QIAGEN), and the QIAamp® Circulating Nucleic Acid Kit (QIAGEN) was used to extract plasma DNA from a 2 ml plasma sample, which was then subjected to sodium bisulfite treatment as previously described (Fujita et al. 2012).

Quantitative TRIM9 promoter methylation analysis using NGS

The NGS methylation assay was performed with the GS Junior system (Roche Diagnostics, Basel, Switzerland) according to the manufacturer’s instructions, and data was analyzed with GS Amplicon Variant Analyzer (AVA) software (version 2.7; Roche Diagnostics). The methylation index (MI) was calculated by dividing the number of cytosines by that of the total reads at each CpG site. NGS primers used for TRIM9 methylation of frozen tissues or cell lines were designed as follows: forward 5′-TGTTTGGAGTGAAATATTGAGATTT-3′, reverse 5′-ACAATAAAACTTTTCTCCTTCTCC-3′ (long primer; Fig. 1). The average methylation ratio of 12 of the 26 CpG sites (6th–17th CpG), which showed the most significant difference between cancer and normal tissues, was used for methylation analysis. NGS primers used for DNA from formalin-fixed paraffin embedded (FFPE) specimens were designed as follows: forward 5′-AGTTTAGTTAGGTGTTTTGGGAGGT-3′, reverse 5′-ACATTAATCAAAATCTATAACCCCTTC-3′ (short primer; Fig. 1). The NGS short primer included 7 CpG sites, corresponding to 6th–12th CpG.
Fig. 1

Primer designs for DNA methylation analysis of TRIM9 using NGS and for detection of TRIM9 methylated ctDNA using real-time PCR. Long primer sets were designed for DNA methylation analysis of TRIM9 by means of NGS (solid arrow). TRIM9 methylated ctDNA in plasma was detected by means of real-time PCR using the short primers (dashed arrow) and probes (dotted line)

In situ hybridization (ISH) for TRIM9 mRNA and immunohistochemical staining (IHC) for TRIM9 protein

The QuantiGene®ViewRNA ISH Tissue Assay kit (Affymetrix, Santa Clara, CA, USA) was used according to the manufacturer’s protocol. FFPE Sections (4 μm) of tumor tissues were incubated at 98 °C with a pretreatment solution for 20 min, followed by protease digestion for 10 min. The TRIM9-specific View RNA™ Probe set (Affymetrix) was hybridized for 2 h. A TRIM9 specific probe set was then designed to hybridize the common sequence of TRIM9_v1 and TRIM9_v2 (1319 bp). ISH images were obtained under fluorescent microscopy (BZ9000; Keyence, Osaka, Japan). Signal intensity was semi-quantitatively determined based on the number of cytoplasmic fluorescent dots in five non-overlapping fields at high-power magnification (×400).

Formalin-fixed paraffin Sections (3 μm) of the tumor tissues were obtained for immunohistochemical staining with rabbit anti-TRIM9 polyclonal antibody (ProteinTech Group, Inc., Chicago, IL, USA) at a dilution of 1:400 according to a previously described method for ER, PR and Ki-67, with a slight modification in that antigen retrieval was accomplished by incubating at 98 °C in citrate buffer (pH 9.0) for 40 min (Shimomura et al. 2009; Tanei et al. 2009). Immunohistochemical staining for TRIM9 was classified as 3+ (strongly positive), 2+ (intermediately positive), 1+ (weakly positive) or 0 (negative). The sections were counterstained with hematoxylin.

Isolation of breast tumor cells by magnetic-activated cell sorting (MACS)

Breast tumor cells were separated from the FFPE tumor tissues with the magnetic-activated cell sorting (MACS) method using the EasySep Human EpCAM Positive Selection Cocktail, the EasySep Human MUC1 Positive Selection Cocktail and EasySep Magnetic Particles (Stem Cell Technologies, Vancouver, BC, Canada) as previously described (Otani et al. 2014). Total DNA was extracted from these isolated tumor cells using the QIAamp® DNA FFPE Tissue Kit (QIAGEN).

Demethylation study of cell lines using 5-aza-2′-deoxycytidine

Twelve breast cancer cell lines (BCCs) and one normal breast cell line were cultured under the conditions shown in Additional file 2: Table S2. For demethylation studies, the cultured cells were treated with 10 μmol/L 5-aza-2′-deoxycytidine (5-aza; Sigma-Aldrich, St Louis, MO, USA) or with dimethylsulfoxide (DMSO) as control for 72 h, with the medium changed every 24 h.

RNA extraction and real-time qRT-PCR

Total RNA was isolated from cell lines using TRIzol® reagent (Invitrogen), and 1 μg of total RNA was reverse-transcribed for single strand cDNA, using random primers and the ReverTra Ace® qPCR RT kit (Toyobo, Osaka, Japan). Reverse-transcription reaction was performed first at 65 °C for 5 min and then at 37 °C for 15 min and at 98 °C for 5 min. Quantitative mRNA expression was measured using the Light Cycler 480 Real-time PCR System (Roche Applied Science, Mannheim, Germany) at 95 °C (10 min), followed by 50 cycles at 95 °C (15 s) and at 60 °C (60 s), and 1 cycle at 50 °C (10 s). TRIM9 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) TaqMan® Gene Expression Assays (assay identification numbers: Hs00364838_m1 and Hs02758991_g1. Applied Biosystems, Foster City, CA, USA) were used for the real time qPCR assay. The expression of TRIM9 was normalized to that of GAPDH, and each assay was performed in duplicate. For the 5-aza treated BCCs, each treated cell line was normalized to the value of its control, which was set at 1.

Measurement of TRIM9 methylated ctDNA in plasma

Two ml of plasma samples was obtained from healthy controls (n = 60) and from metastatic breast cancer (MBC) patients (n = 56), 41 cases of recurrent and 15 of primary advanced breast cancer, before they had been treated at Osaka Police Hospital or Osaka University Hospital between 2012 and 2014. TRIM9 methylated ctDNA in plasma was measured by using quantitative methylation-specific PCR (MSP) with the TRIM9 short primer (Fig. 1). The double-dye probe, including TRIM9 8th–10th CpG sites (5′-TCGTGGAAGGAAGTACGTTTATATTAC-3′; Fig. 1) for detection of TRIM9 methylated ctDNA in plasma, is shown in Fig. 1. 9 µl aliquot each of the bisulfite DNA, eluted for a total PCR reaction volume of 20 µl, was placed in 96-well plates for the TRIM9 PCR reactions. TRIM9 methylated ctDNA in plasma was classified as positive when quantification cycles were less than 50 cycles for TRIM9.

Statistical analysis

The JMP statistical software package (version 11.2.1; SAS Institute, Cary, NC, USA) was used for statistical analyses. Association between the various parameters and TRIM9 methylation ratio was evaluated using the t test for two groups or the Kruskal–Wallis test for more than two groups. The paired t test was used for comparison of frozen cancer and normal tissue MI in matched-pair samples. The Tukey test was used for comparison of the TRIM9 methylation ratio for each subtype. The univariate and multivariate analysis of various parameters for the association with pCR were performed with the logistic regression model. All statistical analyses were two-sided and P values <0.05 were considered to be statistically significant.

Results

Promoter methylation of TRIM9 and its impact on gene expression in BCCs

To study the methylation status of TRIM9, we performed an NGS methylation assay of the TRIM9 promoter in 12 BCCs and a normal breast cell line (HMEC). The methylation ratio of the TRIM9 gene promoter varied greatly from 10.3 to 92.6 % in 11 of the BCCs and was relatively hypomethylated in the HMEC cells (Fig. 2a; Additional file 3: Table S3).
Fig. 2

TRIM9 methylation status and mRNA expression of 13 breast cancer cell lines. a Correlation between TRIM9 methylation index and mRNA expression. b Fold changes in TRIM9 mRNA expression of 8 cell lines after 5-aza treatment

We next used TRIM9-specific primers and probes to investigate the TRIM9 mRNA expression by qRT-PCR. An inverse correlation between the TRIM9 mRNA expression and methylation ratio was clearly observed (Pearson’s correlation coefficient: −0.753) (Fig. 2a). We then treated eight of these cell lines with a demethylating reagent (10 μM 5-aza) and compared the mRNA expression of the treated and untreated cells. 5-aza treatment induced a 16- to 110-fold up-regulation of mRNA expression in all four hypermethylated BCCs (MDA-MB-453, BT474, SKBR3 and MDA-MB-361), while no up-regulation was detected in any of the four hypomethylated BCCs (MDA-MB-468, ZR75-30, T47D and MDA-MB-231), demonstrating that the TRIM9 gene was re-expressed by the demethylation of its promoter region (Fig. 2b).

Methylation and expression of TRIM9 in human breast cancer tissues

To study the methylation status of TRIM9 in human breast cancer and normal breast tissues, we performed an NGS methylation assay using the 19 paired tumor and normal tissues (study I). The methylation ratio was significantly higher for the tumor tissues than the normal tissues (median values, 19 and 1.8 %, respectively, P = 0.00067; Fig. 3a). The ratio of TRIM9 hypermethylated tumors (methylation ratio ≥8.2 %) was 68 %.
Fig. 3

Methylation status of TRIM9 in breast cancer and normal breast tissues. a Comparison of TRIM9 methylation index for 19 paired normal breast and cancer tissues. b Comparison of TRIM9 methylation index for whole breast cancer tissues and tumor cells isolated with the MACS method

For a more accurate assessment of the cancer cell-specific methylation status, we isolated the tumor cells from the FFPE tumor tissues with the MACS method, and the isolated tumor cells were then subjected to an NGS methylation assay. Five tumor tissues with a low methylation ratio (<25 %) were further analyzed since it was thought the low methylation ratio of some of them was due to contamination by the normal stromal and inflammatory cells. Methylation ratios increased in the tumor cells isolated from whole tumor tissues (Fig. 3b).

Relationship between TRIM9 methylation and clinicopathological characteristics

An NGS methylation assay of TRIM9 was performed using the biopsy specimens obtained before NAC (study II) to examine the relationship between TRIM9 methylation and the various clinicopathological parameters including response to NAC (Table 1). TRIM9 hypermethylation (methylation ratio ≥8.2 %) was observed in 40 % (43/107) of the specimens. TRIM9 hypermethylation was significantly associated with ER positivity, PR positivity, low histological grade and no pCR (Table 1). Furthermore, the methylation ratio was significantly lower for basal type (9 %) than for luminal A type (P = 0.0007; Fig. 4).
Fig. 4

Methylation status of TRIM9 in 107 breast cancer tissues. Breast tumors were classified into five intrinsic subtypes (luminal A, luminal B, HER2, basal-like, normal breast-like) by PAM50 for comparison of their methylation index of TRIM9. *Tukey’s test

Next, to determine whether methylation is related to gene expression, we subjected the TRIM9 hypermethylated (n = 10) and hypomethylated tumors (n = 10) to ISH and IHC and found that neither ISH signals nor IHC scores in tumor cells were significantly associated with methylation ratios (Fig. 5).
Fig. 5

Association between TRIM9 mRNA expressions obtained with ISH and IHC analysis and methylation ratios in breast cancer tissues. TRIM9 hypermethylated or hypomethylated breast cancer tissues were subjected to ISH (a) and IHC (b) for TRIM9 mRNA. Each immunoreactivity was one of four scores (0, 1+, 2+ and 3+). ISH, in situ hybridization; IHC, immunohistochemistry

Relationship between TRIM9 methylation and response to NAC

The clinicopathological parameters were assessed by means of univariate analysis for their association with pCR (Table 2). Age, Ki67, ER, PR, HER2, and TRIM9 methylation were found to be significantly associated with pCR. The multivariate analysis showed that only ER, but not TRIM9 methylation, was a significant and independent predictor for pCR.
Table 2

Univariate and multivariate analysis of clinicopathological factors for pCR

Characteristics

Univariate analysis

Multivariate analysis

Odds ratio

95 % CI

P value

Odds ratio

95 % CI

P value

Age (≥50 vs <50)

3.13

1.29–7.66

0.0091

2.24

0.80-6.56

0.1236

T stage (T1+2 vs T3+4)

1.80

0.61–5.35

0.2743

   

Lymph node status (positive vs negative)

1.68

0.64–4.41

0.2858

   

ER (negative vs positive)

10.5

4.00–27.4

<0.0001

5.14

1.23–27.2

0.0240

PgR (negative vs positive)

5.60

1.95–16.1

0.0004

0.98

0.17–4.86

0.9823

HER2 (positive vs negative)

2.47

1.03–5.94

0.0440

1.98

0.69–5.76

0.2002

TRIM9 methylation (<8.2 vs ≥8.2 %)

10.6

2.96–37.6

<0.0001

3.96

0.97–20.1

0.0545

CI confidence interval

Detection of TRIM9-methylated ctDNA in MBC patients

TRIM9-methylated ctDNA in plasma of 56 MBC patients and 60 healthy controls was assayed by using MSP. An amplification curve of the eight standards was obtained by diluting the methylated human control DNA (diluted to 10, 3, 1, 0.3, 0.1, 0.03, 0.01 and 0 ng/ml plasma). The limit of detection for methylated TRIM9 DNA was 0.1 ng/ml in plasma. TRIM9 methylated ctDNA was detected in 18 % (10/56) of MBC patients but not in any of the healthy controls. Primary breast tumor tissues for determination of TRIM9 methylation status were available for 27 of the 56 cancer patients, and TRIM9 methylated ctDNA was detected in 44 % (4/9) of the MBC patients with TRIM9 hypermethylated tumors but in only 6 % (1/18) of the MBC patients with TRIM9 hypomethylated tumors (Table 3).
Table 3

Sensitivity for detection of methylated TRIM9 in plasma of Stage IV and metastatic breast cancer patients

 

Total

Methylated TRIM9 in plasma

 

Positive

Negative

No.

(%)

No.

(%)

Healthy control

60

0

(0)

60

(100)

MBC patients (total)

56

10

(18)

46

(82)

MBC patients with TRIM9 hypermethylated tumors

9

4

(44)

5

(56)

MBC patients with TRIM9 hypomethylated tumors

18

1

(6)

17

(94)

MBC metastatic breast cancer

Discussion

For this study, we selected the TRIM9 gene as a breast cancer specific methylation marker by referring to the methylation array database and observed TRIM9 hypermethylation in 92 % (11/12) of the BCCs and 68 % (13/19) of breast tumor tissues but not in any of the normal breast epithelial cell line (HMEC) cells or normal breast tissues. Several methylation markers for breast cancers have been investigated, such as GSTP1, RASSF1A and RARβ2, and hypermethylation of these genes has been reported as, respectively, 17–48 %, 43–90 % and 26–78 % in breast cancer tissues and as 2–3 %, 3–8 % and 0 % in normal breast tissues (Yamamoto et al. 2012; Jung et al. 2013; Hagrass et al. 2014; Pirouzpanah et al. 2015), indicating the equally high sensitivity and specificity of TRIM9 as a methylation marker for breast cancer. Although methylation of these other genes has reportedly been detected in other types of cancers than breast cancer (Zhang et al. 2015; Li et al. 2015a, b; Grote et al. 2005), TRIM9 is methylated specifically in breast cancer according to the Illumina Human Methylation 450 database (http://cancergenome.nih.gov/), implying that TRIM9 may function as a methylation marker that is specific to breast cancer. The fact that the methylation ratio was lower in tumor tissues than BCCs seems to be explained by the contamination of tumor tissues by the normal stromal and inflammatory cells, since the tumor cells isolated by MACS showed an evidently higher methylation ratio than the tumor tissues from which they derived. These results indicate that breast tumor cells, but not normal breast epithelia, actually harbor TRIM9 methylation.

We found that TRIM9 mRNA expression correlated inversely with TRIM9 methylation ratio in BCCs, and that treatment of TRIM9 hypermethylated BCCs with a demethylating reagent resulted in the reactivation of TRIM9 mRNA expression. Although these findings suggest that TRIM9 expression is epigenetically regulated by promoter methylation in BCCs, we could not confirm the occurrence of such an epigenetic regulation in breast tumor tissues. No reports have been published so far on possible associations between promoter methylation and gene expression in the TRIM family, including TRIM9. TRIM9 is known to be up-regulated by interferons, suggesting that another mechanism than promoter methylation may be more important in the regulation of gene expression in breast cancer tissues (Carthagena et al. 2009) although promoter methylation seems to play a significant role in vitro as we have shown in the present study.

The TRIM9 methylation ratio was significantly lower in basal type tumor than in the other intrinsic subtypes, which is consistent with the report that basal type tumors are more globally hypomethylated than the other subtypes (Cancer Genome Atlas Network 2012). Although TRIM9 hypermethylation was found to be significantly associated with no pCR, this does not necessarily mean that TRIM9 hypermethylation plays a significant role in resistance to chemotherapy. Multivariate analysis failed to demonstrate any statistical significance for TRIM9 hypermethylation as an independent predictor for no pCR. It is thus speculated that TRIM9 hypermethylation may be indirectly associated with no pCR via its strong association with ER, which is a well-established predictor for no pCR (Carey et al. 2007; Rouzier et al. 2005; Ignatiadis and Sotiriou 2013). Putting these considerations together suggests that TRIM9 is unlikely to play a significant role in chemotherapy resistance or is at least, not a clinically useful predictor for no pCR.

Our study detected TRIM9 methylated ctDNA in only 18 % (10/56) of MBC patients. However, this sensitivity was as high as 44 % (4/9) when only the MBC patients with TRIM9 methylated tumors were taken into consideration, but it was only 5.6 % (1/18) for those without TRIM9 methylated tumors. Previous studies have reported that aberrant promoter methylation in serum DNA of MBC patients was 18–25 % for GSTP1 (Yamamoto et al. 2012; Müller et al. 2003; Sharma et al. 2010), 33–39 % for RASSF1A (Yamamoto et al. 2012; Müller et al. 2003; Kim et al. 2010) and 20–87 % for RARβ2 (Yamamoto et al. 2012; Sharma et al. 2010; Kim et al. 2010). Although methylated TRIM9 in blood was less sensitive than the existing methylation markers, the specificity was 100 % which was superior to that of any other genes (2–10 % for GSTP1, 0–10 % for RASSF1A and 5–6 % for RARβ2) (Yamamoto et al. 2012; Müller et al. 2003; Kim et al. 2010). TRIM9-methylated ctDNA may thus be a potential tumor marker and might work better in combination with other blood biomarkers for breast cancers to compensate for its lower sensitivity. However, the number of patients in our study was limited and further prospective studies are needed to verify our findings.

Conclusions

We found that TRIM9 promoter hypermethylation occurred in 68 % of breast tumors but not in normal breast tissues. Methylated TRIM9 was detected in the plasma from 44 % of metastatic breast cancer patients with TRIM9 methylated tumors. Although the regulatory mechanism of TRIM9 gene expression and its biological functions remain unclear, our preliminary results suggest that methylated TRIM9 may serve as a novel blood biomarker specific to breast cancer patients.

Abbreviations

ER: 

Estrogen receptor

PR: 

Progesterone receptor

HER2: 

Human epidermal growth factor receptor 2

CEA: 

Carcinoembryonic antigen

CA15-3: 

Carbohydrate antigen15-3

Declarations

Authors’ contributions

NK and SN participated in the design of the study. SM acquired the data and performed the statistical analysis. TT and KS contributed the immunohistochemical staining. YN advised the statistical analysis. MS, AS, SK conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.

Competing interests

The authors declare that they have no conflict of interest.

Ethical standards The study complied with the current laws of Japan.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Department of Breast and Endocrine Surgery, Osaka University Graduate School of Medicine

References

  1. Arai T, Miyoshi Y, Kim SJ, Taguchi T, Tamaki Y, Noguchi S (2006) Association of GSTP1 CpG islands hypermethylation with poor prognosis in human breast cancers. Breast Cancer Res Treat 100:169–176View ArticleGoogle Scholar
  2. Berti C, Messali S, Ballabio A, Reymond A, Meroni G (2002) TRIM9 is specifically expressed in the embryonic and adult nervous system. Mech Dev 113:159–162View ArticleGoogle Scholar
  3. Cancer Genome Atlas Network (2012) Comprehensive molecular portraits of human breast tumours. Nature 490:61–70View ArticleGoogle Scholar
  4. Cao T, Shannon M, Handel MA, Etkin LD (1996) Mouse ret finger protein (rfp) proto-oncogene is expressed at specific stages of mouse spermatogenesis. Dev Genet 19:309–320View ArticleGoogle Scholar
  5. Carey LA, Dees EC, Sawyer L, Gatti L, Moore DT, Collichio F, Ollila DW, Sartor CI, Graham ML, Perou CM (2007) The triple negative paradox: primary tumor chemosensitivity of breast cancer subtypes. Clin Cancer Res 13:2329–2334View ArticleGoogle Scholar
  6. Carthagena L, Bergamaschi A, Luna JM, David A, Uchil PD, Margottin-Goguet F, Mothes W, Hazan U, Transy C, Pancino G, Nisole S (2009) Human TRIM gene expression in response to interferons. PLoS One 4:e4894. doi:https://doi.org/10.1371/journal.pone.0004894 View ArticleGoogle Scholar
  7. Chambon M, Orsetti B, Berthe ML, Bascoul-Mollevi C, Rodriguez C, Duong V, Gleizes M, Thénot S, Bibeau F, Theillet C, Cavaillès V, Tsai WW, Wang Z, Yiu TT, Akdemir KC, Xia W, Winter S, Tsai CY, Shi X, Schwarzer D, Plunkett W, Aronow B, Gozani O, Fischle W, Hung MC, Patel DJ, Barton MC (2011) Prognostic significance of TRIM24/TIF-1α gene expression in breast cancer. Am J Pathol 178:1461–1469View ArticleGoogle Scholar
  8. Fujita N, Nakayama T, Yamamoto N, Kim SJ, Shimazu K, Shimomura A, Maruyama N, Morimoto K, Tamaki Y, Noguchi S (2012) Methylated DNA and total DNA in serum detected by one-step methylation-specific PCR is predictive of poor prognosis for breast cancer patients. Oncology 83:273–282View ArticleGoogle Scholar
  9. Grote HJ, Schmiemann V, Geddert H, Rohr UP, Kappes R, Gabbert HE, Böcking A (2005) Aberrant promoter methylation of p16(INK4a), RARB2 and SEMA3B in bronchial aspirates from patients with suspected lung cancer. Int J Cancer 116:720–725View ArticleGoogle Scholar
  10. Hagrass HA, Pasha HF, Shaheen MA, Abdel Bary EH, Kassem R (2014) Methylation status and protein expression of RASSF1A in breast cancer patients. Mol Biol Rep 41:57–65View ArticleGoogle Scholar
  11. Hatakeyama S (2011) TRIM proteins and cancer. Nat Rev Cancer 11:792–804View ArticleGoogle Scholar
  12. Ignatiadis M, Sotiriou C (2013) Luminal breast cancer: from biology to treatment. Nat Rev Clin Oncol 10:494–506View ArticleGoogle Scholar
  13. Jung EJ, Kim IS, Lee EY, Kang JE, Lee SM, Kim DC, Kim JY, Park ST (2013) Comparison of methylation profiling in cancerous and their corresponding normal tissues from korean patients with breast cancer. Ann Lab Med 33:431–440View ArticleGoogle Scholar
  14. Kim JH, Shin MH, Kweon SS, Park MH, Yoon JH, Lee JS, Choi C, Fackler MJ, Sukumar S (2010) Evaluation of promoter hypermethylation detection in serum as a diagnostic tool for breast carcinoma in Korean women. Gynecol Oncol 118:176–181View ArticleGoogle Scholar
  15. Li JY, Huang T, Zhang C, Jiang DJ, Hong QX, Ji HH, Ye M, Duan SW (2015a) Association between RASSF1A promoter hypermethylation and oncogenic HPV infection status in invasive cervical cancer: a meta-analysis. Asian Pac J Cancer Prev 16:5749–5754View ArticleGoogle Scholar
  16. Li QF, Li QY, Gao AR, Shi QF (2015b) Correlation between promoter methylation in the GSTP1 gene and hepatocellular carcinoma development: a meta-analysis. Genet Mol Res 14:6762–6772View ArticleGoogle Scholar
  17. Miyake T, Nakayama T, Naoi Y, Yamamoto N, Otani Y, Kim SJ, Shimazu K, Shimomura A, Maruyama N, Tamaki Y, Noguchi S (2012) GSTP1 expression predicts poor pathological complete response to neoadjuvant chemotherapy in ER-negative breast cancer. Cancer Sci 103:913–920View ArticleGoogle Scholar
  18. Müller HM, Widschwendter A, Fiegl H, Ivarsson L, Goebel G, Perkmann E, Marth C, Widschwendter M (2003) DNA methylation in serum of breast cancer patients: an independent prognostic marker. Cancer Res 63:7641–7645Google Scholar
  19. Naoi Y, Kishi K, Tanei T, Tsunashima R, Tominaga N, Baba Y, Kim SJ, Taguchi T, Tamaki Y, Noguchi S (2011) Development of 95-gene classifier as a powerful predictor of recurrences in node-negative and ER-positive breast cancer patients. Breast Cancer Res Treat 128:633–641View ArticleGoogle Scholar
  20. Otani Y, Miyake T, Kagara N, Shimoda M, Naoi Y, Maruyama N, Shimomura A, Shimazu K, Kim SJ, Noguchi S (2014) BRCA1 promoter methylation of normal breast epithelial cells as a possible precursor for BRCA1-methylated breast cancer. Cancer Sci 105:1369–1376View ArticleGoogle Scholar
  21. Parker JS, Mullins M, Cheang MC, Leung S, Voduc D, Vickery T, Davies S, Fauron C, He X, Hu Z, Quackenbush JF, Stijleman IJ, Palazzo J, Marron JS, Nobel AB, Mardis E, Nielsen TO, Ellis MJ, Perou CM, Bernard PS (2009) Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 27:1160–1167View ArticleGoogle Scholar
  22. Pirouzpanah S, Taleban FA, Mehdipour P, Atri M (2015) Association of folate and other one-carbon related nutrients with hypermethylation status and expression of RARB, BRCA1, and RASSF1A genes in breast cancer patients. J Mol Med (Berl) 93:917–934View ArticleGoogle Scholar
  23. Rouzier R, Perou CM, Symmans WF, Ibrahim N, Cristofanilli M, Anderson K, Hess KR, Stec J, Ayers M, Wagner P, Morandi P, Fan C, Rabiul I, Ross JS, Hortobagyi GN, Pusztai L (2005) Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res 11:5678–5685View ArticleGoogle Scholar
  24. Sharma G, Mirza S, Parshad R, Srivastava A, Gupta SD, Pandya P, Ralhan R (2010) Clinical significance of promoter hypermethylation of DNA repair genes in tumor and serum DNA in invasive ductal breast carcinoma patients. Life Sci 87:83–91View ArticleGoogle Scholar
  25. Shi M, Cho H, Inn KS, Yang A, Zhao Z, Liang Q, Versteeg GA, Amini-Bavil-Olyaee S, Wong LY, Zlokovic BV, Park HS, García-Sastre A, Jung JU (2014) Negative regulation of NF-κB activity by brain-specific TRIpartite Motif protein 9. Nat Commun. 5:4820. doi:https://doi.org/10.1038/ncomms5820 View ArticleGoogle Scholar
  26. Shimomura A, Miyoshi Y, Taguchi T, Tamaki Y, Noguchi S (2009) Association of loss of BRCA1 expression with centrosome aberration in human breast cancer. J Cancer Res Clin Oncol 135:421–430View ArticleGoogle Scholar
  27. Suzuki T, Urano T, Tsukui T, Horie-Inoue K, Moriya T, Ishda T, Muramatsu M, Ouchi Y, Sasano H, Inoue S (2005) Estrogen-responsive finger protein as a new potential biomarker for breast cancer. Clin Cancer Res 11:6148–6154View ArticleGoogle Scholar
  28. Tanei T, Morimoto K, Shimazu K, Kim SJ, Tanji Y, Taguchi T, Tamaki Y, Noguchi S (2009) Association of breast cancer stem cells identified by aldehyde dehydrogenase 1 expression with resistance to sequential Paclitaxel and epirubicin-based chemotherapy for breast cancers. Clin Cancer Res 15:4234–4241View ArticleGoogle Scholar
  29. Tanji K, Kamitani T, Mori F, Kakita A, Takahashi H, Wakabayashi K (2010) TRIM9, a novel brain-specific E3 ubiquitin ligase, is repressed in the brain of Parkinson’s disease and dementia with Lewy bodies. Neurobiol Dis 38:210–218View ArticleGoogle Scholar
  30. Tsai WW, Wang Z, Yiu TT, Akdemir KC, Xia W, Winter S, Tsai CY, Shi X, Schwarzer D, Plunkett W, Aronow B, Gozani O, Fischle W, Hung MC, Patel DJ, Barton MC (2010) TRIM24 links a non-canonical histone signature to breast cancer. Nature 468:927–932View ArticleGoogle Scholar
  31. Winkle CC, McClain LM, Valtschanoff JG, Park CS, Maglione C, Gupton SL (2014) A novel Netrin-1-sensitive mechanism promotes local SNARE-mediated exocytosis during axon branching. J Cell Biol 205:217–232View ArticleGoogle Scholar
  32. Yamamoto N, Nakayama T, Kajita M, Miyake T, Iwamoto T, Kim SJ, Sakai A, Ishihara H, Tamaki Y, Noguchi S (2012) Detection of aberrant promoter methylation of GSTP1, RASSF1A, and RARβ2 in serum DNA of patients with breast cancer by a newly established one-step methylation-specific PCR assay. Breast Cancer Res Treat 132:165–173View ArticleGoogle Scholar
  33. Zhang W, Jiao H, Zhang X, Zhao R, Wang F, He W, Zong H, Fan Q, Wang L (2015) Correlation between the expression of DNMT1, and GSTP1 and APC, and the methylation status of GSTP1 and APC in association with their clinical significance in prostate cancer. Mol Med Rep 12:141–146Google Scholar

Copyright

© Mishima et al. 2015