Prediction of Chemotherapeutic Resistance in Serous Ovarian Cancer with Low-Density Custom Microarray

Ovarian cancer remains the deadliest gynecologic malignancy, largely due to the development of chemotherapeutic resistance. This resistance involves complex molecular mechanisms driven by thousands of genes. Addressing this challenge, a recent study utilized a single-channel low-density custom microarray to identify gene expression profiles differentiating chemotherapy-resistant and chemotherapy-sensitive serous ovarian cancer. The research identified 87 differentially expressed genes that accurately predict chemotherapeutic resistance, offering a foundation for personalized treatment strategies.

Study Design and Microarray Development

The study focused on advanced serous ovarian cancer patients admitted between 2001 and 2008. Patients were categorized into two groups based on disease-free interval (DFI): Group R (chemotherapy-sensitive, DFI >12 months, median DFI = 27.36 months) and Group N (chemotherapy-resistant, DFI <6 months, median DFI = 3.25 months). Out of 23 patients, 12 were resistant (Group N) and 11 were sensitive (Group R). Six patients from each group were randomly selected for training datasets (R1, R3, R5, R9, R11, R12 and N2, N4, N5, N8, N11, N12), while the remaining 11 (5 sensitive, 6 resistant) formed testing datasets to validate prediction models.

A custom low-density microarray was designed using Agilent Technologies’ eArray 5.0 platform, containing 1,366 genes selected through two sources: 95 genes previously associated with chemoresistance from the Gynecologic Oncology Center’s genomic studies and additional genes identified via literature mining from PubMed, HighWire, Elsevier, ProQuest, BlackWell, and CNKI. This targeted approach minimized noise from unrelated genes, enhancing specificity compared to whole-genome arrays.

RNA Extraction and Data Acquisition

Total RNA was isolated from tumor tissues using TRIZOL Reagent (Cat#15596-018, Life Technologies) and assessed for quality via RNA Integrity Number (RIN) scores using an Agilent Bioanalyzer 2100. Cy3-labeled cRNA (1.65 μg per sample) was hybridized onto microarrays with Agilent’s Gene Expression Hybridization Kit (Cat#5188-5242). Data were processed using GeneSpring Software 11.0 and SAS, with statistical analyses conducted in R-software.

Identification of Differentially Expressed Genes

Comparative analysis revealed 87 genes differentially expressed between resistant and sensitive groups (Supplementary Table 1 and Table 2). Among these, 71 genes were upregulated (N/R >2, P <0.05) and 16 downregulated (N/R <0.5) in resistant tumors. Chromosomal mapping showed these genes distributed across chromosomes 1–22, with none on sex chromosomes. Notably, chromosome 16q harbored seven metallothionein genes (MT2A, MT1L, MT1E, MT1B, MT1G, MT1H, MT1X), all upregulated in resistant samples. Metallothioneins, known for metal ion binding and oxidative stress response, may contribute to chemoresistance mechanisms.

Gene Ontology (GO) enrichment analysis categorized the 87 genes into molecular functions (87 genes), cellular components (80 genes), and biological processes (81 genes), highlighting pathways linked to drug metabolism, cell adhesion, and apoptosis regulation.

Validation of Predictive Models

Hierarchical clustering of the 11 testing patients (5 sensitive, 6 resistant) using the 87-gene panel achieved 90.9% accuracy (10/11 correct classifications; Figure 1). Only one resistant patient was misclassified. Partial Least Squares (PLS) analysis correctly identified 5/6 resistant patients (83.3% sensitivity) and 4/5 sensitive patients (80% specificity). Support Vector Machine (SVM) models showed slightly lower specificity, correctly classifying 5/6 resistant patients (83.3% sensitivity) but only 2/5 sensitive patients (40% specificity). These results underscore the panel’s robustness in predicting resistance, particularly with PLS-based models.

Comparative Analysis with Prior Studies

Previous efforts to predict chemoresistance in ovarian cancer employed broader genomic approaches. Komatsu et al. (2006) used a 19,981-probe oligomicroarray to identify eight genes predictive of progression-free survival (r = 0.683, P = 0.042). Helleman et al. (2006) utilized 18K cDNA microarrays to define a nine-gene panel (FN1, TOP2A, LBR, ASS, COL3A1, STK6, SGPP1, ITGAE, PCNA) with 89% sensitivity and 59% specificity for platinum resistance. In contrast, the current study’s focused microarray design improved specificity by excluding extraneous genes, achieving comparable or superior performance with a smaller gene set.

Januchowski et al. (2017) highlighted drug-specific resistance mechanisms by profiling cisplatin-, doxorubicin-, topotecan-, and paclitaxel-resistant cell lines. While cell-based studies provide mechanistic insights, the current study’s use of patient-derived tissues preserves in vivo gene expression patterns, avoiding artifacts from in vitro culture.

Advantages and Limitations

The study’s strengths include homogeneity in histologic subtype (serous adenocarcinoma only), tissue-based profiling, and a tailored microarray reducing data complexity. However, limitations include a small sample size (23 patients) and lack of prospective validation. Future directions involve expanding cohorts and leveraging databases like The Cancer Genome Atlas (TCGA) for external validation.

Clinical Implications and Future Directions

The 87-gene panel lays groundwork for a clinical assay predicting chemoresistance. A low-density microarray based on these genes could guide first-line therapy selection, avoiding ineffective treatments and reducing side effects. For instance, patients predicted as resistant may benefit from alternative regimens or targeted therapies. Additionally, mechanistic studies on highlighted genes, particularly metallothioneins, could unveil novel therapeutic targets.

Conclusion

This study demonstrates the efficacy of a low-density custom microarray in identifying chemoresistance-associated genes in serous ovarian cancer. The 87-gene signature achieved high predictive accuracy, offering a translatable tool for personalized oncology. Larger prospective studies are needed to confirm clinical utility, but this approach represents a significant stride toward overcoming chemoresistance in ovarian cancer.

doi.org/10.1097/CM9.0000000000000717

Was this helpful?

YesNo

0 / 0