MDA-MB-157 cells were obtained from the American Type Culture Collection (ATCC) and regularly cultured in high-glucose Dulbecco’s Modified Eagle’s Medium (DMEM) (5 g/L glucose, Sigma-Aldrich, D6429). The media was supplemented with 10% fetal bovine serum (FBS, Gibco, 26140079) and 1% penicillin/streptomycin (Gibco, 15140-122), and cells were maintained in a humidified incubator at 37 °C and 5% CO₂. For induction of MIR503HG or GFP with doxycycline (Dox, Sigma Aldrich, D9891), the cells were treated with 250 ng/mL of Dox for the indicated times before collection. For glucose treatments, cells were incubated in low-glucose DMEM media (1 g/L glucose, Sigma-Aldrich, D6046) for 48 h, before being treated with the experimental glucose concentration (1, 5, or 15 g/L glucose) and co-treated with Dox for 4 h.

Lentiviruses were generated by transfecting HEK293T cells (ATCC, CRL-3216) with pInducer20 [32] containing MIR503HG or GFP, along with pMD2.G, psPAX2 plasmids using the Lipofectamine 3000 kit (Thermo Fisher Scientific, L3000015) according to the manufacturer’s protocol. Medium from transfected 293T cells was collected and filtered through a 0.45 µm filter to transduce MDA-MB-157 cells, and polybrene (0.5 μg/mL) was added. Stably transduced cells were selected under drug selection by Geneticin™ (G418, Gibco, 11-811-031) at 1 μg/mL and were used for Dox-induced ectopic expression of MIR503HG and GFP to perform a variety of experiments described herein.

Cells expressing inducible MIR503HG or GFP were seeded in six-well plates and treated with 250 ng/mL Dox and media with the indicated glucose concentrations for 4 h. The total RNA was isolated using EZ-10 DNAway RNA Mini-Preps Kit (Bio Basic, Canada). Agilent Technologies 4200 TapeStation was used to analyze RNA integrity of the preparation, and Nanodrop (Thermo Scientific, USA) was used to conduct RNA quantification. Only isolated RNA samples with >9 RIN (RNA Integrity Number) were utilized for RNA-seq library preparation. RNA-seq libraries for total RNA-sequencing were prepared and sequenced at Novogene Corporation (Sacramento, CA). Briefly, ribosomal RNA was removed from the total RNA and then precipitated using ethanol. After fragmentation, the cDNA was synthesized using random hexamer primers, followed by the addition of a second-strand synthesis buffer (Illumina), dNTPs, RNase H, and DNA polymerase I to initiate the second-strand synthesis. Next, terminal repair, A-tailing, and sequencing adaptor ligation steps were carried out. Finally, the double-stranded cDNA library was completed through size selection and PCR enrichment. The libraries then went through QC checks, which include Qubit 2.0, Agilent 2100, and real-time PCR. Finally, the quantified libraries were pooled according to the concentration and data amount required and fed into Illumina sequencers (NovaSeq X Plus).

The transcriptomic data for this study were submitted to NCBI’s Gene Expression Omnibus and are accessible through NCBI-GEO-GSE297214. This dataset represents RNA-seq-based gene expression in the breast cancer cell line MDA-MB-157, which includes samples of G1, G5, and G15 glucose treatments for cells overexpressing MIR503HG or GFP.

In this study, the raw reads (fastq files) from total RNA-sequencing were uploaded to Genialis™ Expressions 3.0 [33] and processed using the General RNA-seq pipeline (featureCounts). Reads were trimmed using BBDuk (BBMap 37.90), followed by alignment to the human genome (hg38) with annotation from Ensembl release version 109 (February 2023) using STAR aligner (v2.7.10b) [34]. BAM files were then sorted and indexed using Samtools (v1.14) [35]. Read counts were quantified using featureCounts (1.6.3) [36], with reference annotation from Ensembl version 109. Normalization of expression values was calculated using rnanorm (1.3.1). The DESeq2 tool was used for count normalization and differential gene expression analyses. Filtering criteria for differentially expressed genes were set by applying a p-value (≤0.05).

Utilizing the Molecular Signatures Database (MSigDB) [37], Gene Set Enrichment Analysis (GSEA) version 4.4.0 was conducted to examine the relationship among top DEGs to specific processes and mechanisms. Top 20 gene sets (gene ontologies) with FDR q-value less than 0.05 for gene ontology (GO), KEGG, Hallmark, and CGP were generated for up- and downregulated genes. Genes were ranked by log2 Fold Change value, and the top 500 were used as the input gene set for the GSEA.

To evaluate the prognostic value of identified differentially expressed genes specific to glucose treatment, we explored the expression of merged gene sets across breast tumor tissues. The Kaplan-Meier plots estimating overall survival (OS), distant metastasis-free survival (DMFS), and relapse-free survival (RFS) in three quantiles over 10 years were generated utilizing the Gene Expression-Based Outcome for Breast Cancer Online (GOBO) tool [38]. Genes were ranked by log2 Fold Change value, and the top 500 up- and downregulated genes were used as the input gene sets for the GOBO. Boxplots of merged gene sets across molecular breast tumor subgroups was also attained using the GOBO tool, using the same gene sets. Each box depicts the combined expression score within a subgroup, with the dotted line indicating the overall median expression across all samples.

To investigate the role of glucose in MIR503HG-dependent gene expression, we conducted RNA-sequencing to identify condition-specific gene sets. Dox-inducible MDA-MB-157 cells were generated to overexpress MIR503HG or GFP. We resorted to ectopically expressing MIR503HG (mature and processed RNA), enabling us to identify MIR503HG-regulated gene sets independently of miRNAs. These cells were treated with different glucose concentrations 1 g/L, 5 g/L, and 15 g/L (G1, G5, G15, respectively) for 4 h, while co-treated with doxycycline to induce overexpression of MIR503HG or GFP (Figure 1A). At 4 h, the cells were collected, and total RNA was isolated (Figure 1A). Differential expression analyses of RNA-sequencing data identified distinct up- and downregulated gene sets for MIR503HG-regulated genes compared to GFP (Figures 1A,B). The condition-specific differentially expressed genes (DEGs) were utilized to conduct downstream analyses, as shown in Figure 1A, highlighting the biological and clinical significance of glucose on these genes. Figures 1B,C show Venn diagrams for up- and downregulated gene sets, the circled ones of which were used for all subsequent analyses. It was noticed that there were fewer up- and downregulated genes specific to G1, 55 and 47 respectively, compared to the gene sets for G5 and G15 (Figures 1B,C; Supplementary Tables S1–S3).

To assess the clinical significance of the identified MIR503HG-regulated gene sets affected by specific glucose doses, an 1881-sample breast cancer dataset from Gene Expression-Based Outcome for Breast Cancer Online (GOBO) was used. Boxplots were generated to show expression levels of condition-specific gene sets across PAM50 tumors, indicating significant differences among molecular subtypes of breast cancer (Figure 2). Most markedly, the expression levels of G1 and G5 upregulated genes showed attenuation in basal and HER2-enriched tumors, when compared to G15 (Figures 2A,C,E). It was observed that basal tumors displayed similar expression levels to those observed in our condition-specific downregulated in vitro analysis (Figures 2B,D,F). Additionally, relapse-free survival (RFS) among molecular breast cancer subtypes was examined using the same condition-specific gene sets used previously. Kaplan-Meier survival analysis indicated varying RFS with increased expression of these genes, however, high expression of these gene sets was significantly associated with better RFS rates in All Tumors for G1, G5 and G15 (Figure 3). Similarly, Kaplan-Meier survival analysis was conducted on condition-specific downregulated gene sets, showing a significant correlation between low gene expression and low RFS rates for G1 and G15, specifically, for All Tumors (Figure 4). Overall survival (OS) and distance metastasis-free survival (DMFS) were also analyzed, however, there was a wider variation of correlation among tumor types and gene set expression levels (Supplementary Figures S1–S4). Together, these findings demonstrate that glucose concentrations exert distinct regulatory effects on MIR503HG-regulated genes across breast tumor subtypes, with significant implications for predicting clinical outcomes in breast cancer patients.

Gene Set Enrichment Analysis (GSEA) was conducted to assess the biological impact of each glucose condition on MIR503HG-regulated gene sets. To do this, the previously mentioned condition-specific gene sets were used to derive the top 20 gene ontology (GO) terms, KEGG pathways, Hallmark annotations, and chemical and genetic perturbations (CGP) for each condition, giving insight as to what processes these top up- and downregulated genes are associated with (Figures 5–9; Supplementary Tables S4–S14). Biological processes (BP) for genes upregulated in G5 were mainly involved in metabolic processes relating to small molecules, organophosphates, lipids, cellular catabolic processes, and organic acids (Figure 5A). On the other hand, top BP categories for G5 downregulated genes pertained to regulation of RNA metabolism, signaling, cell proliferation, gene expression, and transcription (Figure 5B). BP for G15 downregulated genes found similar categories to G5 downregulated categories, such as regulation of cell proliferation and signaling, with differences such as proteolysis and phosphorylation observed (Supplementary Figure S5).

Molecular functions (MF) were then obtained to identify molecular roles associated with condition-specific gene sets. Main MF GO terms associated with G5 upregulated genes include lyase activity, purine nucleotide binding, hydrolase activity acting on ester bonds, carbon carbon lyase activity, and guanyl ribonucleotide binding (Figure 6A). Downregulated genes exposed to the G5 condition revealed MF related to purine nucleotide binding, cadherin binding, signaling receptor binding, kinase binding, and transcription factor binding (Figure 6B). Our study found that G15 upregulated genes were primarily associated with MF protein containing complex binding, actin filament binding, and cytoskeletal protein binding (Figure 6C). G15-specific downregulated genes were found to be related to protein-containing complex, protein kinase activity, histone modifying activity, ATP-dependent activity, and PDZ domain binding (Figure 6D).

Similarly, we utilized GSEA to identify KEGG pathways associated with condition-specific gene sets. G5 upregulated genes were observed to be related to Fc gamma R mediated phagocytosis, oxidative phosphorylation, regulation of actin cytoskeleton, insulin signaling pathway, and focal adhesion (Figure 7A). It is also interesting to note that genes related to neurological diseases such as Parkinson’s, Huntington’s, and Alzheimer’s diseases were enriched in the G5 condition. Our study revealed that several KEGG pathways associated with cancer were downregulated in G5, including pathways related to cancer, chronic myeloid leukemia, small cell lung cancer, bladder cancer, and pancreatic cancer (Figure 7B).

Hallmark annotations were also generated for this study. Upregulated genes in G5 were found to be associated mainly with oxidative phosphorylation, androgen response, heme metabolism, bole acid metabolism, and glycolysis (Figure 8A). On the other hand, G5 downregulated Hallmark annotations were related to TNFa signaling via NFKb, estrogen response, inflammatory response, hypoxia, and P53 pathway (Figure 8B). Interestingly, G1 upregulated gene sets were associated with IL2 STAT5 signaling and adipogenesis (Supplementary Figure S6A), while G15 upregulated genes were related to IL6 JAK STAT3 signaling, glycolysis, and myogenesis (Supplementary Figure S6B).

Additionally, our GSEA analysis included investigating chemical and genetic perturbations (CGP). Downregulated genes in G1 condition were mainly associated with ovarian cancer tumors and xenografts down (dn), ELAVL1 targets up, liver cancer ciprofibrate up, liver cancer E2F1 up, liver cancer ACOX1 up, and SATB1 targets dn (Supplementary Figure S7). Top CGP related to G5 upregulated genes include Alzheimer’s disease dn, liver cancer up, chronic myelogenous leukemia up, ESR1 targets up, and apoptosis by doxorubicin dn (Figure 9A). Downregulated G5 genes were primarily associated with bronchial epithelial cells influenza A NS1 up, Alzheimer’s disease up, apoptosis by doxorubicin up, ovarian cancer survival suboptimal debulking, and metabolic syndrome (Figure 9B). CGP related to G15 upregulated genes were mainly associated with brain HCP with H3K4me3 and H3K27me3, nasopharyngeal carcinoma dn, adrenocortical tumor dn, endocrine therapy resistance 4, and neuroblastoma copy number up (Figure 9C). Finally, our study showed that genes related to photodynamic therapy stress up, serum response dn, thyroid carcinoma anaplastic up, EZH2 targets up, and Nanog targets CGP were downregulated in the G15 condition (Figure 9D).

Triple-negative breast cancer (TNBC) is associated with aggressive tumor phenotypes and higher recurrence rates than other subtypes.

Glucose regulates MIR503HG-dependent genes in a dose-dependent manner.

This work represents an advance in biomedical science because it reveals that glucose affects genes regulated by noncoding RNAs, specifically MIR503HG, suggesting their potential use as therapeutic targets in treating TNBC.