Isocitrate dehydrogenase 1–snail axis dysfunction significantly correlates with breast cancer prognosis and regulates cell invasion ability

This study was approved by the institutional review board (IRB) of Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan (IRB number: VGHKS13-CT10-10). The requirement for written informed consent from patients was waived by the hospital IRB because all the data and specimens were previously collected and anonymously analyzed. Tissue microarrays containing 309 paraffin-embedded samples of breast tissue from invasive ductal carcinoma (IDC) were established, and the pathological information on the patients was summarized in our previous study [13].

All level-3 expression data for breast cancer were downloaded from The Cancer Genome Atlas (TCGA) database. The transcriptome data on 1102 and 113 breast cancer and adjacent normal tissues, respectively, were downloaded from the TCGA database. In addition, the small RNA profiles of 778 and 87 breast cancer and adjacent normal tissues, respectively, were obtained from TCGA database. The expression levels of protein-coding genes were shown as reads per kilo base million and those of miRNA were shown as transcripts per million.

A Novolink max polymer detection system (Leica, Newcastle upon Tyne, UK) was used for immunohistochemical (IHC) analysis in this study. The slides were deparaffinized in xylene and rehydrated in graded alcohol. Antigen retrieval was performed by immersing the slides in Tris-ethylenediaminetetraacetic acid (10 mM and pH 9.0) at 125 °C for 10 min in a pressure boiler. Endogenous peroxidase activity was blocked by incubating the slides with 3% hydrogen peroxide in methanol for 30 min. After blocking at room temperature, primary antibodies were immediately applied, and the slides were incubated overnight at 4 °C in a wet chamber. The following primary antibodies were used in this study: rabbit polyclonal anti-IDH1 (1:400; GeneTex, San Antonio, Texas, USA) in Tris-buffered saline solution with 5% bovine serum albumin. After being washed with phosphate-buffered saline, the slides were incubated with horseradish peroxidase-labeled secondary antibody for 10 min at room temperature, and the sections were counterstained with hematoxylin.

First, a senior pathologist accompanied a technician to evaluate the slides until all the discrepancies were resolved. Subsequently, the technician independently reviewed all the slides. Finally, 5–20% of core samples at each intensity were randomly selected for re-evaluation by the pathologist. During the evaluation, the pathologist and technician were blinded to the clinical outcomes of the patients. We graded the immunoreactivity using a semiquantitative approach. Marker scores for staining were calculated based on the staining intensity (0, no signal; 1, mild; 2, moderate; and 3, strong) and the proportion of positively stained tumor cells in the 5 high-power field (0, < 5%; 1, 5–25%; 2, 26–50%; 3, 51–75%; and 4, > 75%). The marker score is the sum of the staining intensity score and percentage of the positive tumor cell score. The score was graded as follows: –, 0–1; +, 2–3; ++, 4–5; and +++, 6–7.

Two micrograms of total RNA was reverse-transcribed with oligo (dT)₁₅ primers and SuperScript III Reverse Transcriptase according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). The reaction was performed by incubating cells at 42 °C for 1 h; the Reverse Transcriptase was subsequently inactivated by incubation at 85 °C for 5 min. The cDNA was used for real-time polymerase chain reaction (PCR) analysis with gene-specific primers, and gene expression was detected using a SYBR Green I assay (Applied Biosystems, Foster City, CA, USA). To determine their expression levels, the genes were subjected to the following conditions: 94 °C for 10 min, followed by 40 cycles of 94 °C/1 min, 60 °C/1 min, and 72 °C/30 s, with a final extension at 72 °C for 10 min. GAPDH expression was used as an internal control, and the expression levels of epithelial–mesenchymal transition (EMT)-relative genes were normalized to those of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (difference in cycle threshold (ΔCt) = Ct_candidates − Ct_GAPDH). The real-time PCR primers used in this study are listed in Additional file 1: Table S1.

We reverse-transcribed 1 μg of total RNA with a stem-loop reverse transcription (RT) reaction using microRNA RT primers and SuperScript III Reverse Transcriptase according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). The reaction was performed under the following incubation conditions: 30 min at 16 °C, followed by 50 cycles of 20 °C/30 s, 42 °C/30 s, and 50 °C/1 s. The enzyme was subsequently inactivated by incubation at 85 °C for 5 min. Gene expression was detected using the SYBR Green I assay (Applied Biosystems, Foster City, CA, USA), and the expression levels of microRNA were normalized to those of U6 small RNA (ΔCt = target miRNA Ct-U6 Ct).

Breast cancer cells were transfected with 10-nM miRNA-32-5p mimics, miR-92b-3p mimics, or the appropriate miRNA mimic control (GenDiscovery Biotechnology Inc., Taiwan) using the Lipofectamine RNAiMAX reagent (Invitrogen, Carlsbad, CA, USA). Twenty-four hours after transfection, cells were harvested, and their expression levels were examined using stem-loop RT quantitative PCR.

Breast cancer cells were transfected with 10-nM si-IDH1#1 (sense: 5′- CAUUAAAGGUUUACCCAAUtt-3′ and antisense: 5′- AUUGGGUAAACCUUUAAUGca-3′, si-IDH1#2 (sense: 5′- CCAACGACCAAGUCACCAAtt-3′ and antisense: 5′- UUGGUGACUUGGUCGUUGGtg-3′) or scramble control (Invitrogen, Carlsbad, CA, USA) using the Lipofectamine RNAiMAX reagent (Invitrogen, Carlsbad, CA, USA). Twenty-four hours after transfection, cells were harvested, and their expression levels were examined through western blotting.

Stable MDA-MB-231 cells with IDH1 knockdown were generated by infecting MDA-MB-231 cells with lentiviruses expressing sh-IDH1 in the presence of 8 μg/mL of polybrene for 24 h, followed by puromycin (4 μg/mL) selection for 3–5 days. The shLuc vector targeting the luciferase gene provided puromycin resistance and was used as the control. IDH1 expression was confirmed through a western blotting assay. In this study, we designed two shRNA sequences targeting IDH1, and the individual sequences of shRNA used for constructs in this study were as follows: for sh-IDH1 #1, sense: 5′-CCGGCCTATCATCATAGGTCGTCATCTCGAGATGACGACCTATGATGATAGGTTTTT-3′ and antisense: 5′-AAAAACCTATCATCATAGGTCGTCATGAGCTCATGACGACCTATGATGATAGGCCGG-3′; for sh-IDH2#2, sense: 5′-CCGGGCTTTGGAAGAAGTCTCTATTCTCGAGAATAGAGACTTCTTCCAAAGCTTTTT-3′ and antisense: 5′-AAAAAGCTTTGGAAGAAGTCTCTATTGAGCTCAATAGAGACTTCTTCCAAAGCCCGG-3′.

Breast cancer cells (2.5 × 10³ cells/mL) were seeded in a 96-well plate and transfected with siIDH1, shIDH1 or scramble control. After transfection, cell growth was determined at 0, 1, 2, 3, and 4 days using the CellTiter-Glo® One Solution Assay (Promega Corporation, Madison, WI, USA). All the experiments were performed in triplicate.

For the wound healing assay, after transfection with si-IDH1 or scramble control for 24 h, cells (1.5 × 10⁶) were seeded on six-well plates. A straight line was scratched on the monolayer in the middle of the well using the tip of a 200-mL pipette. A culture medium of 10% fetal bovine serum (FBS) was replaced with a serum deprivation culture medium, and cells were then incubated at 37 °C. Wound closure was monitored and photographed at different time points under a microscope. Subsequently, the open area was assessed for quantifying cell migration ability.

Cells were assessed for their invasion ability in vitro using a Transwell assay, according to our previous study [14]. Briefly, breast cancer cells (4.5 × 10⁵) transfected with si-IDH1, miR-32-5p mimics, miR-92b-3p mimics, or scramble control were suspended in 2% FBS and seeded on the upper chamber of the Transwells (Falcon, Corning Incorporated, USA) with a coating of Matrigel (BD Biosciences, MA, USA) for the invasion assay. After incubating in a CO₂ incubator at 37 °C for 12 or 24 h, the remaining cells in the upper chamber were removed with cotton swabs, and cells on the undersurface of the Transwells were fixed with 10% formaldehyde solution. Cells were stained with crystal violet solution, and the numbers of breast cancer cells were calculated by counting three fields under a phase-contrast microscope. All experiments were performed in triplicate.

The putative miRNAs targeting the 3′-UTR of IDH1 were determined using TargetScan (http://​www.​targetscan.​org/​vert_​71/​) and microRNA.org (http://​34.​236.​212.​39/​microrna/​home.​do). In this study, we identified that 15 miRNAs were predicted for binding at the 3′-UTR of IDH1. The full-length 3′-UTR of IDH1 was cloned into the PGL3 vector. Subsequently, pGL3–IDH1–3′-UTR vectors were co-transfected with miR-32-5p mimics, miR-92b-3p mimics, or scramble controls in breast cancer cells using the Lipofectamine RNAiMAX reagent (Invitrogen, Carlsbad, CA, USA). After transfection for 24 h, cell lysates were used for measuring the luciferase activity by using the Dual-Glo Luciferase Reporter Assay System (Promega Corporation, Madison, WI, USA).

Total cell lysates were extracted with radioimmunoprecipation assay (RIPA) buffer (50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% deoxycholic acid, and 0.1% sodium dodecyl sulfate). Total proteins were separated through 6–10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred onto nitrocellulose filter membranes (Millipore, Billerica, USA). After the transfer, the membranes were blocked with a blocking buffer for 1 h at room temperature and probed with a primary antibody for the desired molecule overnight at 4 °C, followed by treatment with horseradish peroxidase-conjugated secondary antibody for 1 h at room temperature. Finally, the proteins were visualized using WesternBright™ ECL (Advansta Inc., Menlo Park, CA, USA) and detected using the BioSpectrum^TW 500 Imaging System (UVP, USA). The antibodies are shown in Additional file 2: Table S2.

α-KG was analyzed through ultra-high performance liquid chromatography (UPLC; Acquity UPLC system, Waters Corporation, MA, US) coupled with a quadrupole time-of-flight mass spectrometer (QTof MS) (Xevo, Waters). Details of the procedures were described in another study [15]. Raw data on all samples were processed using MassLynx software.

Cells subjected to different treatments were cultured in the presence of 20 μg/mL of cyclohexamide for the indicated time periods and then harvested in RIPA buffer containing protease inhibitors (Roche, CA, USA). Protein lysates were quantified using the bicinchoninic acid (BCA) method for the subsequent western blotting analysis.

Many statistical methods were used for data analysis. The chi-squared test, Student t test, analysis of variance (ANOVA), Mann–Whitney U test, and Kruskal–Wallis one-way ANOVA were used to test correlation between the expression levels of each protein and different types of breast tissues or clinicopathological parameters. Studies on breast cancer have typically defined the outcomes as the time from diagnosis or surgery until a particular event of interest (endpoint). Disease-specific survival (DSS) is measured from the time of the initial resection of the primary tumor to the date of cancer-related death or last follow up. Disease-free survival (DFS) is defined as the time from surgery to an event (local recurrence, regional recurrence, and distant metastasis, but not disease-related death). The cumulative survival curves were estimated using the Kaplan–Meier method, and the survival curves were compared using the log-rank test. A Cox proportional hazards model was used to determine the independent predictors of survival using factors significant in univariate analysis as covariates. We considered p < 0.05 (two-sided) as significant.

First, we examined the expression levels of IDH1 and IDH2 in breast cancer tissues and adjacent normal tissues from 10 patients using western blotting. The results revealed that IDH1 expression levels were significantly lower in breast cancer tissues than in adjacent normal tissues (p = 0.004; Fig. 1a and b). However, IDH2 expression levels did not differ significantly. We further assessed the mRNA expression levels of IDH1 using a real-time PCR approach, which revealed that IDH1 expression levels and transcript levels were significantly lower in breast cancer tissues than in adjacent normal tissues (p = 0.0042; Fig. 1c). By analyzing TCGA database, we observed the same result that IDH1 expression levels were significantly lower in breast cancer compared to adjacent normal tissues (p = 0.00003; Fig. 1d). To assess whether the lower IDH1 expression levels contributed to breast cancer progression, we performed IHC analysis of tissue microarrays containing adjacent normal tissues (n = 309), ductal carcinoma in situ (DCIS) tissues (n = 120), and IDC tissues (n = 309) obtained from 309 patients with breast cancer. The results revealed that IDH1was expressed in the cytoplasm; representative photomicrographs show negative (−), weak (+), moderate (++), and strong (+++) staining in IDC tissues (Fig. 1e). The IHC data revealed that IDH1 expression levels were significantly lower in DCIS tissues (p < 0.001) and IDC tissues (p < 0.001) than in adjacent normal tissues. Furthermore, IDH1 expression levels were significantly lower in IDC tissues than in DCIS tissues (p < 0.001; Fig. 1f and g). Together, these results indicated that the protein expression levels of IDH1 were gradually decreased during breast cancer progression.

We investigated whether IDH1 protein dysfunction contributes to the clinicopathological features of breast cancer. As shown in Table 1, low IDH1 expression levels were significantly correlated with the early pathology stage (p = 0.012) and early perineural invasion (pN) stage (p = 0.018). Furthermore, high IDH1 expression levels were significantly correlated with human epidermal growth factor receptor (HER)-positive human breast cancer (p = 0.046), whereas no difference was observed in estrogen receptor (ER) and progesterone receptor (PR) status (Additional file 3: Table S3). Univariate logistic analysis revealed that low IDH1 expression levels resulted in poor DSS (crude hazard ratio (CHR), 1.64; 95% confidence interval (CI), 1.14–2.38; p = 0.008) and DFS (CHR, 1.58; 95% CI 1.07–2.34; p = 0.023) in patients with breast cancer (Table 2 and Fig. 2). Multivariate logistic analysis and plotting of survival curves showed that low IDH1 expression levels were significantly associated with poor DSS (adjusted HR (AHR), 1.57; 95% CI, 1.08–2.30; p = 0.02) but not DFS (AHR, 1.40; 95% CI, 0.93–2.10; p = 0.106) in patients with breast cancer (Table 2). An analysis of TCGA database revealed that low IDH1 mRNA expression levels were significantly associated with poor overall survival in patients with breast cancer (CHR, 1.80; 95% CI% 1.10-2.94; p = 0.02) (Additional file 4: Table S4). These data suggest that IDH1 might be involved in the growth, migration, and invasion of breast cancer cells and that low IDH1 expression correlates with poor prognosis in patients with breast cancer. In addition, several studies have showed that the IHC antibody raised against mutant IDH1 protein with R132H mutation can specifically recognize this mutant protein in the tumor tissues using immunochemistry staining (IHC). Therefore, we examined the presence of R132H mutant IDH1 in breast cancer tissues using the IHC approach, and this mutation was absent in breast cancer tissue from the 309 women (data not shown). This finding was in accordance with a previous study [12]. Together, our data revealed that aberrantly low expression levels of IDH1, rather than IDH1 mutation, had crucial effects on the prognosis of breast cancer.

A number of studies have reported that dysfunction of miRNA is often the cause of the aberrant expression of cancer-associated genes [14, 16, 17]. Therefore, we further identified putative miRNAs that suppress IDH1 expression. Both microRNA.​Org and TargetScan predicted 15 miRNA candidates to potentially target the 3′-UTR of IDH1 (Fig. 3a). Because the expression levels of IDH1 were lower in breast cancer, those miRNA candidates with elevated expression in breast cancer would have the potential to regulate IDH1. Therefore, the expression levels of the predicted candidates were examined using TCGA database, showing that miR-32-5p and miR-92b-3p expression was significantly higher (fold change > 2) in breast cancer tissues (n = 778) compared with adjacent normal tissues (n = 87; Fig. 3b). According to the aforementioned criteria, miR-32-5p and miR-92b-3p might suppress IDH1 expression by targeting its 3′-UTR. To examine this notion, miR-32-5p and miR-92b-3p were ectopically expressed in MDA-MB-231 cells by transfecting them with miR-32-5p and miR-92b-3p mimics. As shown in Fig. 3c and d, the expression levels of miR-32-5p and miR-92b-3p were clearly increased in MDA-MB-231 cells transfected with the respective mimics compared with those transfected with scramble control (N.C). Furthermore, the ectopic expression of miR-32-5p and miR-92b-3p suppressed the mRNA and protein expression of IDH1 in breast cancer cells (Fig. 3e and f). We further constructed the wild-type IDH1 full-length 3’UTR, the mutant IDH1 3’UTR with mutated miR-32-5p binding site and with mutated miR-92-3p binding site respectively into the pMIR reporter vector and performed luciferase reporter assays in MDA-MB-231 cells (Fig. 3g and h). As shown in Fig. 3i and j, miR-32-5p and miR-92b-3p significantly suppressed the luciferase activity of wild-type pMIR–IDH1-3′-UTR (wild-type), but not of mutated pMIR–IDH1-3′-UTR.

We examined the expression levels of miR-32-5p and miR-92b-3p in breast cancer using a real-time PCR approach. The expression levels of miR-32-5p and miR-92b-3p were significantly higher in breast cancer tissues than in the adjacent normal tissues (miR-32-5p, p < 0.001 and miR-92b-3p, p < 0.001, respectively; Additional file 5: Figure S1a and b). The ectopic expression of miR-32-5p or miR-92b-3p significantly accelerated the invasion ability of breast cancer cells (Additional file 5: Figure S1c and d). In summary, the aberrant overexpression of oncogenic miR-32-5p and miR-92b-3p resulted in the reduction of IDH1 by directly binding at the 3′-UTR of IDH1 to promote invasion in breast cancer cells.

Our data revealed that IDH1 expression levels were closely associated with early pathology and pN stages, suggesting that IDH1 contributes to the growth and migration/invasion of breast cancer cells. We further examined the expression levels of IDH1 in eight breast cancer cell lines, including cells with low invasive ability (MCF7, T47D, SK-BR3, MDA-MB-468, and MDA-MB-453) and cells with high invasion ability (BT549, Hs578T, and MDA-MB-231; Fig. 4a). Our data indicated that IDH1 expression levels were higher in cells with low invasive ability than in cells with high invasive ability. Endogenous IDH1 expression was knocked down in MDA-MB-231 cells through transfecting with si-IDH1 and scramble control. As shown in Fig. 4b and Additional file 6: Figure S2a, IDH1 expression levels were significantly lower in MDA-MB-231 cells transfected with si-IDH1 than in those transfected with scramble control. We further investigated the effects of IDH1 on the proliferation, migration, and invasion of MD-MB-231 cells. The growth ability of MDA-MB-231 cells was not significantly affected after IDH1 knockdown (Fig. 4c and Additional file 6: Figure S2b). However, the wound healing assay indicated that IDH1 knockdown significantly increased the migration ability of MDA-B-231 cells (Fig. 4d). Furthermore, the invasion ability of MDA-MB-231 cells significantly increased after IDH1 knockdown (p < 0.001; Fig. 4e and f and Additional file 6: Figure S2c and d). The expression levels of slug, snail, and twist proteins and mRNA were significantly increased, whereas those of vimentin were not significantly changed in MDA-MB-231 cells transfected with si-IDH1 (Fig. 4g and h and Additional file 6: Figure S2e). We also established an IDH1 stable knockdown with shIDH1 MDA-MB-231 cells. Our data revealed that a stable knockdown could accelerate the invasion ability of MDA-MB-231 through elevating snail and slug protein expression levels (Additional file 7: Figure S3). Similar results also were observed in HS578T and BT549 cells with IDH1 knockdown (Additional file 8: Figure S4). We also examined the role of IDH1 in ER-expressing MCF7 cells. As depicted in Additional file 9: Figure S5, the knockdown of IDH1 can significantly accelerate the migration ability of MCF7 cells through increased snail and twist expression (Additional file 9: Figure S5b and c). In contrast to MDA-MB-231 cells, the proliferation of MCF7 was promoted after IDH1 knockdown (Additional file 9: Figure S5d), which suggested that the role of IDH1 in MCF7 cell proliferation might be slightly different in breast cancer cells with different subtypes. However, these findings require further investigation. Our present study revealed that IDH1 knockdown significantly increased the migration and invasion ability of breast cancer cells by triggering their EMT.

We performed the transcriptome profiling of si-IDH1 and scramble control cells using a microarray approach to evaluate the putative signaling pathways regulated by IDH1. Pathway enrichment analysis revealed that IDH1 knockdown significantly altered cancer-related pathways such as the mitogen-activated protein kinase (MAPK), peroxisome proliferator-activated receptor signaling pathways, regulation of cell adhesion, and cell migration (Additional file 10: Figure S6). Mutant IDH1 can reportedly influence the activity of MAPK signaling in melanoma cells [18]. In addition, HIF-1a and nuclear factor kappa B (NFkB) are involved in EMT by elevating snail expression in cancer cells [19‐23]. Based on the previous evidence, we examined whether IDH1 knockdown could potentially accelerate the invasion ability of breast cancer cells through MAPK–HIF-1a/NFkB signaling (Additional file 10: Figure S6b). As shown in Fig. 5a-c, MAPK signaling was activated in IDH1 knockdown MDA-MB-231 cells and thus promoted cell invasion. Treatment of U0126, the inhibitor of MAPK signaling, abolished IDH1 knockdown-induced cell invasion (Fig. 5b and c), implying that the effect of IDH1 knockdown on the promotion of cell invasion may be attributed to MAPK signaling. Further, the knockdown of IDH1 expression revealed that the expression levels of HIF-1α and phosphorylated NFkB were clearly elevated in MDA-MB-231 cells (Fig. 6a and b and Additional file 6: Figure S2e). The reporter assay showed that both HIF-1α and NFkB signaling were activated in IDH1 knockdown cells. Meanwhile, the activation of HIF-1α and NFkB signaling induced by IDH1 knockdown was abolished in the presence of U0126 (Fig. 6c and d). These results indicated that IDH1 possibly modulated the motility of breast cancer cells through MAPK–HIF-1a/NFkB signaling.

We subsequently investigated the molecular mechanism underlying the IDH1 knockdown-mediated upregulation of HIF1α. First, we examined whether the elevated HIF1α expression upon IDH1 depletion was regulated at the mRNA level. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) indicated that IDH1 depletion caused a slight decrease in HIF1α mRNA (Fig. 7a), which implied that the IDH1 depletion-induced upregulation of HIF1α may not be explained by the increased mRNA expression in HIF1α. Next, the effect of IDH1 depletion on the protein stability of HIF1α was examined. A protein degradation assay was performed to compare the protein stability of HIF1α between the IDH1-depleted cells and control cells. Western blotting indicated that the protein level of HIF1α was relatively stable during 60 min of cycloheximide treatment in the IDH1-depleted cells compared with the control cells, in which the protein level of HIF1α began to rapidly decrease at 30 min after cycloheximide treatment (Fig. 7b and c). This suggested that the IDH1 depletion-mediated upregulation of HIF1α was attributed to the increased protein stability of HIF1α. Several studies have demonstrated that the protein degradation of HIF1α is mainly regulated by an enzyme called prolyl hydroxylase domain-containing protein 2 (PHD2), the activity of which is highly dependent on α-KG under normoxic conditions [24, 25]. IDH1 is the major cellular enzyme responsible for catalyzing isocitrate into α-KG in the tricarboxylic acid (TCA) cycle [26], therefore, we hypothesized that IDH1 depletion caused a reduction in the cellular level of α-KG to impair the activity of PHD2, leading to the stabilization of HIF1α proteins under normoxic conditions. To address this assumption, the intracellular level of α-KG in IDH1-depleted cells was measured through UPLC coupled with a quadrupole time-of-flight mass spectrometer (QT of MS). The MS revealed that the intracellular level of α-KG decreased twofold in IDH1-depleted cells compared with the control cells (Fig. 7d). Furthermore, 1 mM α-KG reversed the increase in the HIF1α protein level observed in the IDH1-depleted cells (Fig. 7e and f). Taken together, these results indicated that IDH1 depletion resulted in a low level of cellular α-KG, which in turn impaired the activity of the α-KG-dependent enzyme PHD2, ultimately leading to the accumulation of HIF1α protein.

Using the IHC data, we examined the correlation among IDH1, HIF-1a, twist, and snail. As shown in Table 3, the expression levels of HIF-1a and twist were positively associated with those of IDH1 (r = 0.221, p = 0.001 and r = 0.172, p = 0.011, respectively), whereas snail expression levels were inversely associated with IDH1 expression levels (r = − 0.141, p = 0.037). Notably, the IDH1–snail axis supported our finding in breast cancer on the negative correlation between IDH1 and snail expression in clinical samples. We next combined the expression levels of IDH1 and snail to assess the correlation with breast cancer survival. We classified the combined status of IDH1 and snail into four groups: IDH1 high–snail low, low levels of both IDH1 and snail, high levels of both IDH1 and snail, and IDH1 low–snail high. As shown in Fig. 8 and Table 4, low IDH1 expression levels and high snail expression levels were significantly associated with poor survival in patients with breast cancer (CHR for DSS, 2.10; 95% CI, 1.22–3.63; p = 0.008 and CHR for DFS, 2.49; 95% CI, 1.42–4.37, p = 0.001), even after adjustment for clinicopathological factors (AHR for DSS, 2.34; 95% CI, 1.32–4.16; p = 0.004 and AHR for DFS, 2.50; 95% CI, 1.39–4.50; p = 0.002). The combination of IDH1 and snail expression levels had a linear trend for their effects on DSS (p for linear trend = 0.006) and DFS (p for linear trend = 0.011).

In summary, we observed that miR-32-5p and miR-92b-3p overexpression resulted in tumor-suppressive IDH1 silencing in breast cancer tissues compared with that in adjacent normal tissues. Furthermore, depletion of IDH1 in breast cancer cells results in accelerating breast cancer migration and invasion activities by activating snail expression (Fig. 9). Low IDH1 expression levels, particularly in combination with snail expression, served as a potential prognosis biomarker for breast cancer.