Abstract
Background/Aim: Ovarian cancer (OC) is one of the leading gynecological causes of death among women. The current standard treatment for OC is debulking surgery followed by platinum-based chemotherapy treatments; however, despite initial success to treatment many patients experience relapses. Currently, there are no available tests to predict sensitivity or resistance to chemotherapy. The aim of this review is to investigate the literature regarding prediction of chemotherapy resistance in patients with OC using gene expression patterns. Materials and Methods: The literature research on PubMed resulted in a total of 490 articles published from November 20th, 2018, to November 20th, 2024. We selected only the original studies that described the comparison of mRNA profiles between platinum-sensitive and -resistant OC patients. Studies were included if mRNA expression was measured in human tissue by gene expression microarray, RNA sequencing and quantitative real-time PCR. Results: Forty-four articles were included covering data from discovery cohorts obtained from hospitals and universities, as well as additional data obtained from online datasets from Gene Expression Omnibus and The Cancer Genome Atlas Program that provided either single- or multiple mRNA signatures that could discriminate between chemotherapy-sensitive and chemotherapy-resistant OC patients. OCs at all stages and histological subtypes were used but most articles included exclusively high-grade serous OC patients. Conclusion: mRNA-based biomarkers to predict chemotherapy resistance in patients have not yet been clinically implemented, but many differentially expressed genes between chemotherapy-resistant and -sensitive patients have been reported, such as ABCG2, DOCK4, DUSP1, DUSP4, DUSP5, GADD45B, HELQ, HOXA9, KLF4, and NR4A1, which could be compelling biomarker candidates for further studies.
Ovarian cancer (OC) has been among the deadliest gynecological diseases worldwide for years. In 2020, it accounted for approximately 314,000 new cases and 207,000 deaths, ranking OC as the seventh most common cancer type among women (1). OC is a heterogeneous disease on both molecular and histological levels. It can originate from epithelial cells, sex-cord stromal cells, germ cells, or mixed-cell type (2), with epithelial ovarian cancer (EOC) being most prevalent accounting for about 90% cases. OC can histologically be divided into clear cell, endometrioid, mucinous and serous adenocarcinomas (2, 3).
According to clinicopathology and molecular differences EOC can be divided into type I and type II (2). Type I includes low-grade serous OC (LGSOC), endometrioid carcinoma, clear cell carcinoma (CCC), mucinous carcinoma, malignant Brenner tumor, and seromucous carcinoma. Type II includes high-grade serous ovarian cancer (HGSOC), undifferentiated carcinoma, and carcinosarcoma. Type I is associated with histological serous low tumor grade, possible early diagnosis, good overall clinical outcome, slow progression, while type II is associated with histological serous high tumor grade, diagnosis at late stage, poor overall clinical outcome, and rapid and aggressive progression (4, 5). HGSOC is the most aggressive and most common subtype of EOC, making up 75% of all EOC subtypes (3). Of all OC diagnosed patients about 75% are diagnosed at advanced stage due to lack of specific symptoms in early stages (3) giving the high mortality seen in OC.
Currently, the standard treatment for EOC includes primary debulking surgery of removal of all tumor mass followed by adjuvant chemotherapy treatment. If it is not expected to remove all visible tumor, then neoadjuvant chemotherapy is given. After the neoadjuvant chemotherapy, the patient is evaluated for possible surgery (interval surgery) followed by additional series of chemotherapy (3). Initially 80% of OC patients do respond to platinum-based chemotherapy treatment, but unfortunately up to 80% of patients relapse (6, 7).
The chemotherapy resistance of patients can be refractory/resistant (R), or patients may be partially sensitive (PS)/sensitive (S) to treatment. Chemotherapy drug resistance might potentially be caused by changes that occur within cancer cells, such as development of survival pathways, occurrence of mutations, as well as changes in gene expression.
For the last years, many studies focused on the discovery of new biomarkers for predicting sensitivity to chemotherapy in OC subtypes based on their molecular, histological, imaging, or physiological characteristics (8, 9). It has been shown that the gene expression in human tissue might vary between chemotherapy-R and chemotherapy-S patients (10, 11). Therefore, one of the promising strategies is based on gene expression measurements, due to the development and improvement of high-throughput technologies, such as quantitative real-time PCR (qRT-PCR), microarrays, and later RNA sequencing (RNA-seq).
The aim of this review is to investigate the literature regarding prediction of chemotherapy resistance in patients with OC using gene mRNA expression pattern. This review was restricted to studies using human tissues to ensure the gene signatures are based on clinical relevance and biological variability by capturing the diversity, such as histological type, age, and ethnicity thus giving a more precise representation of the complexity of chemotherapy resistance in human populations. Moreover, this review was restricted to studies that used tissue from patients that have received chemotherapy treatment and compared the gene expression levels between chemotherapy-R and chemotherapy-S patients to capture their possible implication in chemotherapy resistance development. All studies included in this review included patients treated with both adjuvant and neoadjuvant platinum-based chemotherapy.
Materials and Methods
A literature review was conducted in the PubMed database by using the following keywords: (((((mRNA[Title/Abstract]) OR (gene expression[Title/Abstract])) AND (ovarian[Title/Abstract])) AND (cancer[Title/Abstract])) AND (resistance[Title/Abstract])) NOT (Review[Publication Type]) AND (“2018/11/20”[Date - Publication]: “2024/11/20”[Date - Publication]), which identified 490 studies.
The aim of this review is to present gene expression patterns in patients with OC focusing on their potential relevance to chemotherapy resistance. The including criteria for studies were as follows: (i) Only original OC studies regarding mRNA expression measurements were included; reviews were excluded; (ii) Gene expression was measured in patient tumor tissue by one of these three methods: gene expression microarray, RNA sequencing or qRT-PCR. The studies that analyzed gene expression by directly measuring protein levels were not included; (iii) Gene expression pattern was compared in chemotherapy resistant and chemotherapy sensitive patients that received platinum-based treatment; (iv) The study was published during the last 6 years.
The PubMed results (490 articles) were filtered by two-step procedure: (i) For the abstract-based selection, the papers were excluded if they failed to fulfill the above-mentioned four requirements or were written in other languages than English. The number of articles after the examination was 161; (ii) For the full-text-based selection, the papers were excluded if they did not meet all the requirements – or if needed information was missing. After examination this resulted in 44 papers that were used in this review. The study selection was performed and archived on each of these two steps by MT and other reviewers.
The data was extracted by pre-creating tables specifically for this purpose, and the tables were subsequentially filled with the relevant data/information.
Relevant data included in the tables were: mRNA source, histological subtype, FIGO stage, tissue preservation, tumor percentage, number of patients in discovery cohort, age (years), chemotherapy drug, chemotherapy treatment, strategy for differentially expressed genes (DEGs) selection, study reproducibility, availability of gene lists, final mRNA candidates, definitions of resistance, publicly available datasets, and number of samples in publicly available datasets.
DEGs from the studies were not directly compared due to variations in gene annotations. Instead, the rentrez R package was used to standardize and convert all gene identifiers to official gene symbols. After this standardization, the gene names were compared across studies to identify overlapping genes.
Results
Overview of included studies. Figure 1 and Table I, Table II, Table III and Table IV give an overview of the studies included in this review. Figure 1 shows the chart flow of study selection process.
Flow chart showing the process of study selection.
Overview of clinicopathological characteristics of patients of studies included in this scoping review.
Description of strategy for DEGs selection of each study and the list of genes identified in the selected articles.
The definition of platinum-based chemotherapy resistance/sensitivity of included studies.
List of publicly available datasets used in the selected studies.
An overview of the clinical information is shown in Table I. The studies are organized according to the OC histological subtype and to the use of own cohort, publicly available cohorts, or both own and publicly available cohorts to achieve the results. Based on the information in Table I, the 44 articles included in this review cover a total of 1,724 unique patient-samples from own cohorts collected from either hospitals or universities. The tissue samples used in the selected articles were collected between 1988-2021.
Of the 44 studies included, 21 studies used tissue-samples from own cohorts as the only source of mRNA extraction for mRNA expression analyses, 6 studies used tissue-samples from own cohorts and repositories, such as TCGA and GEO for mRNA expression analyses, and 17 studies used exclusively publicly available datasets, such as TCGA and GEO for mRNA expression analyses.
All patients received platinum-based chemotherapy treatments. Adjuvant chemotherapy treatment was given in 90.9% (n=40) of the studies, neoadjuvant chemotherapy was given in 6.8% (n=3) of the studies, and 2.3% (n=1) of the studies used a mixed cohort, with some patients receiving neoadjuvant and others adjuvant chemotherapy. Patients have received platinum (n=28), cisplatin (n=5), paclitaxel in combination with carboplatin (n=3), carboplatin (n=2), platinum in combination with cisplatin (n=2), paclitaxel in combination with cisplatin (n=2), or taxane in combination with platinum derivates (n=2).
Gene expression measurement methods used in the selected articles were microarray (n=18), qRT-PCR (n=14), RNA-seq (n=9), and combination of RNA-seq and microarray (n=3). Patient age was stated in 16 articles. Although most of the studies had a focus on a specific histological subtype of OC, there were some studies that either looked at several histological subtypes or did not define the histological subtype of OC. The majority of the studies (59%, n=26) focused exclusively on HGSOC. Eight (18%) studies focused on EOC without describing the histological subtype, only that samples were retrieved from patients with EOC. Four (9%) had no description of OC type, two (4.5%) studies focused on CCC, two (4.5%) used EOC cohorts containing both CCC, serous OC, endometrioid OC, mucinous OC, and mixed type OC, one (2.27%) study focused on SOC, and one (2.27%) study focused on mixed cohort containing SOC and HGSOC.
Table II provides information about the strategy for DEGs selection as described in the original articles by their author, study reproducibility, availability of gene list, and final mRNA candidates. The mRNA candidates were defined as the ones that were differentially expressed between S and R patients, but also associated with sensitivity/resistance to platinum-based chemotherapy after additional filtering steps described in the articles. The studies were organized according to histological subtype. Based on Table II, there were only 6 reproducible studies, while the rest of the studies were not reproducible due to lack of detailed description of the selection of DEGs, lack of inclusion of barcodes for TCGA patients, lack of complete list of DEGs, lack of raw data for studies using own cohort, and other. We investigated whether the final mRNA candidates from Table II were highlighted in various studies, and we found 10 genes (ABCG2, DOCK4, DUSP1, DUSP4, DUSP5, GADD45B, HELQ, HOXA9, KLF4, and NR4A1) that were mentioned in more than one study.
Table III provides information about platinum-based chemotherapy resistance/sensitivity definition among the selected studies. Based on the included studies in this review, there were 12 different definitions of platinum-based chemotherapy resistance, and seven studies did not provide any definition. The studies were organized according to the definition of resistance used in their study.
Information about publicly available datasets used in the selected articles, the number of samples in each dataset and dataset title is displayed in Table IV. The listed publicly available datasets contain OC patients that have received platinum-based chemotherapy treatment or are relevant to OC research. The studies were organized according to the online dataset they used in their study.
Discussion
This review revealed that a total of 267 unique candidate DEGs were identified between R and S OC patients, based on the criteria applied by the authors of the included 44 original studies. However, only 10 candidate DEGs were identified in more than a single study, and these are: ABCG2, DOCK4, DUSP1, DUSP4, DUSP5, GADD45B, HELQ, HOXA9, KLF4, and NR4A1.
Notably, nine of these genes were not identified in more than two independent studies, while KLF4 was identified in three independent studies (12-14). The level of overlap may be lower than expected, particularly given that certain articles utilized the same publicly available datasets. Therefore, we investigated the reasons behind the insufficient consensus regarding mRNA biomarkers for chemotherapy resistance based on the information extracted from 44 selected articles.
First, it is important to notice that, among the studies included in this review, there are 12 different platinum-based chemotherapy resistance definitions that are defined based on either progression free survival (PFS), platinum free interval (PFI), disease free survival (DFS), or time to recurrence (TTR), as shown in Table III. It is also important to notice that 25% (n=11) of the studies lack a definition in their studies. Furthermore, even among studies using PFI as a measure for platinum-based chemotherapy resistance, there are differences. The inconsistence and lack of definition across studies leads to significant challenges in reproducibility of findings.
One would expect that studies using the same publicly available datasets would find the same or a high degree of the same candidate genes being differentially expressed. According to the findings in this review there is no overlapping of results between the genes in different studies that used the same publicly available dataset, as for example GSE15622 shown in Table III (15, 16). One of the reasons might be that even if the studies use the same datasets, the different ways in which data is processed (e.g., normalization methods or various differential expression analysis tools) lead to variation in the outcome.
Selected studies in this review used different RNA extraction and purification kits that might give different levels of expression and purity. Furthermore, the studies also used tissues in different conditions to perform the mRNA expression analyses. Only one of the studies have used fresh tissues without any freezing for the analysis (17), while all the other studies have used tissues preserved in different ways, such as fresh frozen tissues (16, 18), or formalin-fixed paraffin-embedded (8).
When measuring mRNA expression level in a tissue sample, it is important to be aware about the tumor percentage because the mRNA expression levels are measured as an average across the sample. Among selected articles, some studies use samples containing ≥70% tumor cells for mRNA measurement (19, 20), some use samples containing ≥50% tumor cells (21), and some use samples containing ≥20% tumor cells (22). If a sample contains a lower percentage of tumor cells, it might dilute the significance of mRNA expression level in tumor cells. OC is known for its heterogeneity, and mRNA expression levels vary across tumor cells from different histological subtypes of OC.
One should also notice that the studies using the TCGA OC dataset, which originally provided mRNA profiles for 489 patients based on microarray technology (23), vary in the number of patients analyzed. However, details on how these patients were selected or how their clinical data was accessed are often not provided. Therefore, including TCGA patient barcodes for each TCGA-based study together with platinum sensitivity or resistance definition applied should be required.
It has been shown that OC incidence and mortality is affected by demographic factors, such as age, race, and ethnicity (24). From the included studies in this review only one study had included ethnicity (25), no studies have included race, and only 16 have included age.
Besides the missing information about demographic factors, four of the studies in this review do not clarify the histological subtype of OC and only 22 studies have included OC FIGO stage. Therefore, it would be advisable to specify the histological type and stage in future studies to get a better and more precise understanding of chemotherapy resistance development that is subtype specific.
This review also revealed that some findings were consistent between different research groups. For example, the two studies addressing ABCG2 gene expression (26, 27) showed that the upregulated mRNA level of ABCG2 is associated with platinum-based chemotherapy resistance, besides the fact that the studies might have different patient characteristics, since one was done on EOC patient samples and the other does not specify the OC subtype. Unfortunately, in contrast to the consistent results addressing ABCG2, the findings of the studies regarding HELQ and DOCK4 showed opposing results (18, 28-30). One of the studies presents that upregulated mRNA levels of HELQ are associated with platinum sensitivity, while the other study suggests a contradictory association. Similarly, the two studies investigating DOCK4 conclude the opposite results about whether upregulated mRNA levels are associated with platinum-resistance or platinum-sensitivity (18, 30).
When comparing the two studies regarding DOCK4, it is important to notice that one of the studies has used RNA-seq technology, while the other has used qRT-PCR to estimate gene expression. Of notice is that one of the studies uses an online dataset consisting of only 28 patients, while the other study is based on 149 patient tissue samples. Furthermore, these two studies have different definitions of chemotherapy resistance and might have different patient and tumor characteristics as one of the studies is done on HGSOC patient samples and the other study does not specify the EOC subtype. Neither of the studies mention the tumor percentage used for analyses.
When comparing the two studies regarding HELQ, even though both studies have used RNA-seq and same definition of chemotherapy resistance, it is important to notice that tissue samples used in one of the studies have been collected from a hospital in Czech Republic, Europe, while the tissue samples used in one of the studies have been collected from a hospital in China, Asia. This might affect the findings taking the great importance of ethnicity and race into consideration. Moreover, these two studies might also have different patient and tissue characteristics, since one of the studies is done on HGSOC patient samples and the other does not specify the EOC subtype, and neither of the studies mention the tumor percentage used for analyses.
These opposing findings highlight the need for further investigation of these genes in larger well characterized studies to clarify the potentials of DEGs as biomarkers for platinum-based chemotherapy resistance in HGSOC.
In this review, while we conducted a broad literature search, it is possible that certain studies - particularly those examining specific OC subtypes - were missing if their titles, abstracts, or keywords did not explicitly mention “ovarian cancer.” This limitation could result in incomplete coverage of relevant literature. Another constraint involves the language filter applied, as we included only English-language articles, which may have excluded valuable research published in other languages. Additionally, we aimed to map the overlapping DEGs identified across included studies; however, many studies have not provided raw data or detailed analysis description. Therefore, to identify overlapping DEGs, we focused on those selected by authors of the original studies, based on their criteria (Table II), rather than the complete list of DEGs discovered through various technologies. Our scoping methodology also excluded formal quality assessments of the included studies, as the goal of this review was to map existing research rather than evaluate study rigor. Furthermore, our search was conducted within a limited time frame, which may have unintentionally excluded recent publications. Finally, we did not impose a minimum patient/sample size for studies, as this review was designed to catalog the breadth of current research, regardless of sample size.
Conclusion
Many studies have investigated the gene expression patterns between OC patients sensitive or resistant to chemotherapy to achieve a better understanding of the development of platinum-based chemotherapy resistance, yet an optimal mRNA biomarker for the prediction of chemotherapy resistance is not found. This review included 44 original studies, in which, only 10 mRNAs were found in more than one study. The 10 potential mRNA candidates to be further validated as biomarkers of resistance in platinum-based chemotherapy in OC patients are ABCG2, DOCK4, DUSP1, DUSP4, DUSP5, GADD45B, HELQ, HOXA9, KLF4, and NR4A1. Moreover, this review highlights the importance of consensus across studies regarding chemotherapy resistance definition, tissue handling, patient characteristics, and treatment protocols to find a clinically approved mRNA as biomarker for prediction of platinum-based chemotherapy resistance.
Footnotes
Author’s Contributions
M.T. contributed to the study planning, data collection, data analysis, results discussion; and writing of the final manuscript. J.L.J. contributed to the study planning, data collection, data analysis, results discussion, and final manuscript review. E.H. contributed to the study planning, data collection, data analysis, results discussion, and final manuscript review.
Conflicts of Interest
All Authors declare no conflicts of interest.
- Received November 8, 2024.
- Revision received December 3, 2024.
- Accepted December 6, 2024.
- Copyright © 2025 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
