CTC: physiologic characteristics and analysis
CTCs are cancer cells found in peripheral blood that intravasate or are passively shed from a primary or metastatic solid tumour site. Several analytic methods for CTC isolation have been developed and validated for ovarian cancer that are based on various biological (i.e., positive epithelial markers, negative hematopoietic markers) or physical properties (i.e., size, density, deformability, electric charges, and invasive capacity) [
14‐
29]. The ability to detect CTCs in the bloodstream has important prognostic implications in ovarian cancer for identifying potential micrometastasis, pre-neoplastic lesions, tumour heterogeneity, and tumour evolution over time [
30‐
33].
Isolation of CTCs from peripheral blood samples is technically challenging given the low concentration with approximately 1 CTC in 1,000,000 circulating cells [
34‐
36]. Following release from the primary tumour, CTCs overcome several obstacles to survive in the systemic vasculature and spread to distant organs [
37]. First, tumour cells shed from solid tumours often traverse the endothelium to enter the circulation by undergoing the epithelial-to-mesenchymal transition (EMT) process. EMT is a phenotypic transformation of epithelial cells with loss of polarity, morphology, and cell markers such as the epithelial cell adhesion molecule (EpCAM), to gain the migratory and invasive properties of mesenchymal cells [
38]. After dissociation from a primary model V
2 carcinoma (established from skin carcinoma of cottontail rabbits) site of 1 cm, approximately one million CTCs intravasate via dermal invasion into the peripheral circulation each day, of which < 1% typically remain viable for metastasis [
39]. Most CTCs undergo apoptosis or necrosis due to the profound environmental challenges in the bloodstream such as starvation, shear stress, and immunological detection [
40]. Only a small proportion of CTCs can survive through upregulation of several signalling pathways, including increased secretion of growth factors, downregulation of death receptors, and over-expression of anti-apoptotic ligands [
39]. CTCs must also evade natural immune system defences and natural killer (NK) cell recognition. According to the adaptive immune resistance theory, tumour cells avoid activation of the antitumor response from NK cells and T cells by upregulating the naturally occurring programmed death-ligand 1 (PD-L1). They also avoid phagocytosis by macrophages through the upregulation of CD47 [
41‐
43].
Currently, the CellSearch detection system is the most widely used isolation strategy with US Food and Drug Administration (FDA) approval. CellSearch uses an immunoaffinity-based isolation strategy to identify CTCs based on positive EpCAM expression [
44]. However, the application of CellSearch in ovarian cancer may be limited in the setting of low EpCAM expression. In a study with newly diagnosed or recurrent ovarian cancer patients, Liu et al. found no correlation between serial CTC enumeration by the CellSearch system and clinical outcomes [
15]. One possible explanation is the downregulation of EpCAM during EMT or the heterogeneous expression of cell surface markers in ovarian cancer [
15]. Obermayr et al. also reported EpCAM expression in a small proportion of EOC patients [
21]. The researchers used RT-qPCR to analyze EpCAM expression of CTCs isolated using density gradient centrifugation from 216 EOC patients before and after primary treatment compared to 39 healthy controls. The researchers reported EpCAM expression in only 8% of patients at baseline before treatment and 4% for patients after six months of adjuvant chemotherapy.
To overcome these limitations, alternative approaches targeting various biophysical properties of CTCs such as cell size and invasive potential have been developed [
14,
29]. For example, Yang et al. recently used a technique called CanPatrol enrichment, in which CTCs were filtered through 8-μm porous membranes and detected using RNA in situ hybridization (RNA-ISH). CTC subpopulations were identified using epithelial (EpCAM and CK8/18/19), mesenchymal (vimentin and Twist), and epithelial-mesenchymal hybrid markers [
14]. The researchers used hybrid markers and found a mean CTC count of 8.70 ± 5.69 detected in 5 mL of blood among stage 1-IV EOC patients that was significantly higher compared to controls with benign gynecologic diseases (1.04 ± 0.73). Multivariate analyses demonstrated both higher CTC counts and higher percentage of mesenchymal CTC were independent prognostic factors for significantly lower OS (
p = 0.012 and
p = 0.009 respectively). Fan et al. proposed another novel enrichment method utilizing the unique property that blood containing CTCs will invade and ingest Cell Adhesion Matrix (CAM) while non-tumour and dead tumour cells do not [
29]. The researchers used a cell invasion assay that enriches and identifies tumour cells based on CAM invasion (CAM +) and expression of standard epithelial markers (Epi +) to analyze peripheral blood samples of 71 suspected ovarian cancer patients. The study found a significantly higher mean CTC count in stage III/IV patients at 41.3 CTCs/ml compared to 6.0 CTCs/ml in stage I/II patients and 0 CTCs/ml in benign patients (
p-value = 0.001). Kaplan Meier analysis showed a significantly lower disease-free survival in patients with detectable CTCs with a median survival of 15.0 months compared to 35.0 months in those without detected CTCs (
p= 0.042). Other novel techniques include modified immunoaffinity-based strategies targeting several ligands at once (e.g. EpCAM, folate receptor alpha, Human epidermal growth factor receptor 2) and nanoparticles conjugated with the antibody against Mucin 1 (MUC1) [
45,
46]. Given the rarity of CTCs in peripheral circulation despite their prognostication potential, further studies are required to optimize the detection and isolation of CTCs in ovarian cancer.
cfDNA/ctDNA: physiological characteristics and analysis
Normally, plasma contains cfDNA that is passively released from necrotic or apoptotic cells, while ctDNA is the cfDNA secreted from cancer cells. In healthy individuals, cfDNA concentrations are elevated following tissue damage such as intense exercise, inflammation, sepsis, surgery, radiotherapy, trauma, or during pregnancy [
47‐
49].
Compared to CTCs, cfDNA concentrations are higher in blood, making them suitable targets for liquid biopsy [
50]. Tumours harbour unique somatic genetic alterations that help in distinguishing ctDNA from noncancerous cfDNA [
47,
48]. The majority of cfDNA is expected to originate from healthy cells, while a variable amount of cfDNA (0.01–93%, depending on the tumour size) can originate from cancer cells (ctDNA) [
51,
52]. However, a popular hypothesis posits that a large fraction of cfDNA is released from cells in the tumour microenvironment that were destroyed due to hypoxia or the antitumour response [
53]. Recent studies confirmed that cfDNA levels in the blood are higher among ovarian cancer patients with an average of 180 ng/mL compared to 30 ng/mL in healthy controls or individuals with benign ovarian pathologies [
54‐
56]. Therefore, increased amounts of cfDNA may serve as a diagnostic tool for ovarian cancer, while genomic analysis of ctDNA may provide valuable prognostic and predictive information [
57,
58].
The mechanism of cfDNA released from cells into the circulation remains unclear, although apoptosis and necrosis are the most widely accepted hypotheses based on cfDNA properties. Previous studies have estimated the size of cfDNA to vary from ~ 40–200 base pairs (bp), with a peak at around 166 bp [
53,
59‐
61]. Agarose gel electrophoresis to separate extracted cfDNA has found fragment ladders ranging from 160 bp up to 21 kbp [
61,
62]. The size of these fragments corresponds primarily to mono- and oligonucleosomes that are characteristic of caspase-dependent cleavage during apoptosis [
59,
61‐
63]. In contrast, DNA fragments larger than 10 kb are thought to be a result of necrotic cell death in tumours with different kinetics and amount of cfDNA released from different necrosis-inducing agents [
64‐
66]. However, this theory has been called into question following studies reporting that radiation therapy, which typically induces tissue necrosis, results in a reduction of cfDNA levels by up to 90% in the plasma of cancer patients [
67,
68]. Other proposed cfDNA release mechanisms include active secretion in living cells with the expulsion of nuclei, phagocytosis, neutrophil extracellular trap release (NETosis), and excision repair [
69‐
74].
Once released into the bloodstream, the size, integrity, and half-life of cfDNA have important clinical implications in diagnosis and tumour detection. One challenge currently is the small amount of ctDNA in the blood compared to cfDNA released from normal cells, particularly when the tumour size is small. Since a significant proportion of ctDNA is released from necrotic cancer cells, the cfDNA size in cancer patients is generally longer than those of healthy individuals. However, the length of ctDNA released from apoptotic cancer cells is shorter than cfDNA released from the normal cells due to apoptosis, with a mean of 133–144 bp [
75]. ctDNA enrichment may therefore be possible based on a size selection approach. Selecting shorter DNA fragments between 90–150 bp improved the detection of ctDNA with up to 11-fold enrichment of mutation allele fraction [
61,
75,
76]. In addition, the distribution of differently sized DNA fragments has implications for disease staging as an indicator of cfDNA Integrity (cfDI). The cfDI is defined as the ratio of long (released from necrotic cells) to short (released from apoptotic cells) cfDNA fragments. cfDI is calculated by measuring long and small ALU sequence fragments (ALU
247 and ALU
115respectively) using qPCR [
77]. Studies have shown that cancer patients have a higher cfDI compared to healthy controls or individuals with benign disease [
78,
79]. Higher integrity is associated with increased levels of necrotic cell death in advanced disease with larger and more aggressive tumours [
78,
80].
The short half-life of cfDNA present in the bloodstream allows for real-time analyses of the tumour mutational profile. The level of cfDNA in circulation at any given time is determined by the net amount of DNA released minus DNA clearance. cfDNA clearance may occur in organs including the liver, spleen, kidney, or lymph nodes [
81]. In the bloodstream, circulating enzymes such as DNase I, plasma factor VII–activating protease (FSAP), and factor H are responsible for cfDNA breakdown [
82,
83]. Rapid clearance of apoptotic cells and cfDNA normally allow for healthy individuals to have low levels of cfDNA. In cancer patients, cfDNA accumulates due to impaired clearance that is currently poorly understood. Using fetal DNA in postpartum maternal circulation, Lo et al. estimated the half-life of cfDNA to be approximately 4 to 30 min, which has been consistent across other studies [
84‐
87]. However, the half-life of cfDNA may vary depending on several factors, including interactions with molecular complexes that interfere with cfDNA degradation, tumour stage and subtype, and treatment [
69,
81]. Interestingly, one study used next-generation sequencing (NGS) technology to examine the kinetics of cfDNA and found that the clearance of cfDNA may occur in a bi-phasic manner. The first rapid phase has a mean half-life of an hour, followed by a second slow phase with a mean half-life of 13 h [
88].
Several technologies have been developed for ctDNA detection in blood, including quantitative PCR, digital droplet PCR (ddPCR), and NGS for targeted sequencing or whole-genome sequencing (WGS). In addition to quantitative changes, these technologies detect qualitative changes in ctDNA, which include tumour-specific variants (TSVs), gene fusion, copy number variations, aberrant DNA methylation, and chromosomal instability. The development of NGS and digital polymerase chain reaction (dPCR) has improved the sensitivity and specificity of ctDNA detection. To date, most ctDNA detection methods have focused on high-grade serous ovarian cancer (HGSOC) patients with targeting TP53 mutations [
45,
89‐
91]. One study used targeted error correction sequencing (TEC-Seq) to examine 58 cancer-related genes encompassing 81 kb and reported the highest sensitivity and specificity at 75–100% and > 80%, respectively [
89]. In stage I-II disease, the highest detection rate was 68% with a specificity of 100% that was achieved using TEC-Seq and ddPCR combined. The high specificity achieved in this study may be attributable to TEC-Seq advantages for using deep sequencing for more direct evaluation of sequence changes. In fact, deep sequencing using random unique molecular barcodes annealed to each DNA template fragment has been the preferable method for detecting low-level signatures of TSVs in liquid biopsies [
92]. Duplex sequencing using molecular barcodes on both DNA strands for removing sequencing errors that are in one strand only has improved variant detection accuracy by > 10,000 times compared to conventional NGS [
93,
94].
Although ctDNA analysis with plasma samples is currently the preferred method, alternative approaches have utilized different sources. In 2013, Kinde et al. examined the ability of the liquid Pap test with uterine cervix sampling to detect ovarian and uterine cancers. The researchers used massively parallel sequencing for TSVs using a 12-gene panel and found ctDNA in 41% (9 of 22) of ovarian cancer patients [
95]. Similarly in 2018, Wang et al. analyzed Pap brush samples from 245 ovarian cancer patients using PapSEEK with an assay for mutation in 18 genes and reported a limited detection sensitivity of 33%, including 34% for patients with stage I–II disease [
96]. Likewise, Maritschnegg et al. conducted a study with uterine cavity lavage samples from EOC patients and benign gynecologic patients [
97]. The researchers used NGS for sequencing
AKT1,
APC,
BRAF,
CDKN2A,
CTNNB1,
EGFR,
FBXW7,
FGFR2,
KRAS,
NRAS,
PIK3CA,
PIK3R1,
POLE,
PPP2R1A,
PTEN, and
TP53 genes to analyze lavage samples. Using NGS, the researchers reported detectable mutations, mainly in
TP53, in 60% of ovarian cancer patients and an improved ctDNA detection rate of up to 80% with more sensitive methods of digital droplet polymerase chain reaction (ddPCR) and the Safe-sequencing system (SafeSeqS). Interestingly, the study also found
TP53 mutations in lavage samples of all 5 patients with stage IA disease. Building on these results, the same team conducted a study in 2018 demonstrating the feasibility of this technique that found a median absolute amount of 2.23 μg cfDNA in uterine and tubal lavage samples and TSVs using deep-sequencing in 80% (24 of 30) of ovarian cancer patients [
27]. Given these findings, the molecular analysis of uterine lavage samples may be a potential technique for the early diagnosis of ovarian cancer. Other novel techniques including peritoneal washing, urine sampling, and vaginal sampling have been utilized for ctDNA profiling. However, such methods require more research to elucidate their diagnostic utility in ovarian cancer [
98‐
101].
cfRNA: cell-free mRNA, miRNA, circRNA and lncRNA
The rapid turnover of tumours results in high gene transcription and shedding of high amounts of cfRNA consisting of mRNA and microRNA (miRNA) into the circulation [
102]. Normal and tumour cells secrete miRNAs into various body fluids, including plasma, urine and vaginal discharge, and breast milk [
103]. In the blood, mRNA and miRNA are bound to specific ribonucleoprotein complexes, high-density lipoproteins, platelets, or packaged in extracellular vesicles (EV) such as exosomes to avoid degradation and acquire more stability [
103,
104]. Several studies have suggested the role of miRNAs in tumorigenesis, cell differentiation, proliferation, inhibition of angiogenesis, metastasis, and apoptosis. Importantly, the biogenesis and activation of miRNAs are faster with longer half-lives compared to mRNA and proteins, which may make miRNAs more suitable for earlier diagnosis of ovarian cancer [
105‐
108].
The diagnostic, prognostic, and therapeutic potential of circulating miRNAs in ovarian cancer have been explored in many studies. In 2008, Taylor et al. first reported that higher levels of 8 exosomal miRNAs (miR-21/141/200a/200b/200c/203/205/214) were found in the serum of ovarian patients compared to healthy controls, although there was no significant difference in early versus late-stage ovarian cancer [
109]. These findings were subsequently supported by several other studies reporting that serum miRNAs (miRNA-141/200a/200b/200c) were upregulated in ovarian cancer patients compared to normal or benign tumour controls [
110,
111]. Gao et al. also found that different miRNA-200c expression levels may correlate with ovarian cancer staging, with more advanced tumours having lower miRNA-200c levels and higher miRNA-141 [
110]. However, Kim S. et al. analyzed seven serum exosomal miRNAs and concluded that the expression of miRNA-141, 200a, and 200b were too low to be an appropriate serologic biomarker [
112].
Although miRNA-145 was identified as the best-performing single marker with a sensitivity of 91.7% and accuracy of 86.8%, similar changes in miRNA-145 levels were observed in other malignancies besides ovarian cancer [
112,
113]. The lack of discrimination between cancer types suggests that single miRNAs are unlikely to be a reliable biomarker. To overcome these challenges, a recent study by Elias et al. was the first to combine NGS analysis of serum circulating miRNA with a machine learning technique called a neural network model and developed a diagnostic algorithm for EOC. The study authors reported an AUC value of 0.90 for this model, which was significantly higher compared to CA-125. This study suggests the potential for a new era of machine-learning application in biomarker discovery [
114].
In 2017, Yokoi et al. performed miRNA sequencing to identify the optimal combination of candidate circulating miRNAs for the early detection of ovarian cancer [
115]. This study identified eight miRNAs with RT-qPCR validation and statistical cross-validation with a large research cohort. The predictive model using a combination of 8 circulating serum miRNAs was able to differentiate early-stage ovarian cancer from benign tumours with 86% sensitivity and 83% specificity, and from healthy controls with 92% sensitivity and 91% specificity [
115]. In a later study, the same research team analyzed 4,046 serum samples from 333 ovarian patients, 95 benign or borderline ovarian tumours, 2,759 healthy controls and 859 other solid cancers using miRNA microarray [
116]. The study found that combined miRNAs can successfully discriminate ovarian from lung, gastric, breast, hepatic, colorectal, and pancreatic cancers, but not sarcoma or esophageal cancer. In this study, utilization of circulating miRNA yielded a sensitivity of 99% and a specificity of 100% for discriminating between ovarian cancer and healthy controls. This was the first large-scale comprehensive study examining circulating miRNAs in ovarian cancer and reported promising miRNA combinations for the detection of early-stage disease.
In addition to miRNA, circular RNAs (circRNAs) and long non-coding RNAs (lncRNAs) also demonstrated potential utility as biomarkers for liquid biopsy in ovarian cancer. circRNAs have a covalently closed loop structure and lncRNAs have transcript sizes of > 200 nucleotides, which allow increased stability and resistance against RNase degradation in the peripheral circulation. circRNAs are abundant and diverse, with a half-life > 48 h, that facilitate easier detection [
117‐
119]. circRNA expression differs between primary and metastatic sites and is thought to play a role in regulation of ovarian cancer. A recent study found that the expression levels of circular RNAs are inversely associated with activating many signalling pathways involved in tumour metastasis (i.e., NF-κB, PI3k, AKT, and TGF-β) [
120]. Using RT-qPCR in a sample of 83 EOC patients compared to 166 benign or healthy controls, Hu and colleagues found that CircBNC2 was associated with histological grade, serous subtype, and distant metastasis [
121]. Similarly, lncRNAs were found to contribute to the early pathogenesis, progression, metastasis and chemoresistance of recurrent ovarian cancer [
122‐
124]. Although there is emerging evidence suggesting an association between differing expression levels of lncRNAs (H19, LSINCT5, XIST, CCAT2, HOTAIR, AB073614, and ANRIL) and clinical progression or treatment response of ovarian cancer, the diagnostic sensitivity and specificity of lncRNAs remain to be fully elucidated [
125‐
130]. To date, no lncRNA has been approved for clinical utility and further research is required to identify the most clinically relevant candidates with cancer-enriched or specific signatures in ovarian cancer.
TEPs: RNA content
Tumour-educated platelets (TEPs) play an important role in local and systemic responses to tumour growth. Platelets are normally anucleate, although they may contain residual mRNA and miRNA derived from their megakaryocyte precursors or captured from intercellular interactions in the circulation. Platelet education denotes the transfer and sequestration of biomolecules from tumour cells into platelets [
131,
132]. External factors in the tumour microenvironment such as stromal and immune cell signals may activate platelet surface receptors to induce specific splice events of pre-messenger RNAs (pre-mRNAs) in circulating platelets [
133,
134]. Key advantages of TEPs include their high abundance, easy isolation, and high-quality RNA that may be processed according to external signals. Therefore, TEPs have a dynamic mRNA repertoire with both specific splice events in response to external signals and direct ingestion of spliced circulating mRNA that may provide useful diagnostic information in ovarian cancer. Best et al. first studied the diagnostic potential of TEPs by mRNA sequencing in patients with various cancers [
135]. This study found that TEPs were able to distinguish cancer patients from healthy controls with a high accuracy of up to 96% and detect the primary tumour location with 71% accuracy. Later, Piek et al. concluded that TEPs can differentiate early stage ovarian cancer from benign pathologies with 80% accuracy [
136]. An ongoing clinical trial (NCT04022863) may further build upon these results by examining the accuracy of TEPs and ctDNA in determining the nature of ovarian tumours and provide information on its diagnostic potential [
137]. Interestingly, a recent retrospective cohort study by Giannakeas et al. examined the association between thrombocytosis (platelet count greater than 450 × 109/L) and cancer. By studying 53,339 adults aged 40–75 years who developed thrombocytosis with normal platelet count in the previous 2 years and no malignancy history, they estimated the risk of cancer in a 10-year follow-up period [
138]. The authors reported that the 2-year relative risk (RR) was highest for ovarian cancer (RR = 7.11; 95% CI, 5.59–9.03), while the 6-month RR for developing ovarian cancer was even higher (RR = 23.33; 95% CI, 15.73–34.61). In the future, TEPs profiling with complementary ctDNA/CTC analysis and platelets quantification may potentially be a blood-based method for cancer diagnostics.
Exosomes: content analysis
Interest in the diagnostic and prognostic potential of exosomes has increased in recent years. Exosomes are extracellular vesicles (EVs) typically 30–100 nm in diameter. Such vesicles are extremely stable and can survive under extreme conditions. Exosomes are released from both normal and tumour cells. Likewise, they are found in various body fluids, such as saliva, plasma, urine, ascites, and cerebrospinal fluid [
139]. Exosomes can participate in local and distant signalling by fusing with the membrane of the recipient cell or attaching to receptors on the cell’s surface. In cancer, exosomes have the ability to enhance tumorigenesis [
140], help tumour cells escape the immune system [
141], and cause treatment resistance [
142]. Likewise, exosomes can enter the circulation and increase the likelihood of metastasis by preparing distant tumour microenvironments [
143,
144]. Exosomes have been also used as therapeutics to successfully eliminate tumour cells [
145]. Furthermore, exosomes contain tumour-specific proteins, lipids, DNA, and RNA making them potential diagnostic biomarkers in cancer. For example, exosomes with heat shock protein (HSP70) expressed on their membrane, are observed more in ovarian, breast, and lung cancer samples compared to healthy controls [
146]. Additionally, studies have shown increased total exosome concentrations in serum samples of EOC patients [
109,
111].
Exosomes can carry significant quantities of miRNAs. Multiple studies have observed differences between the miRNA profiles of exosomes in EOC patients and healthy controls. Meng et al. showed that the concentrations of miR-200b and miR-200c are higher in exosomes obtained from patients with stage III–IV EOC and are associated with significantly shorter OS [
111]. Another study showed that the miRNA profiling of circulating exosomes using a modified magnetic-activated cell sorting (MACS) technique can differentiate between benign and malignant ovarian tumours [
109]. Additionally, exosomes derived from EOC patients have higher concentrations of TGFB1 and melanoma-associated antigen 3 (MAGE3) and MAGE6 [
147]. EOC exosomes also have a higher concentration of Claudin 4 that is associated with tumour stage and CA125 levels [
148]. CD24 and EpCAM were also shown to be elevated in exosomes isolated from EOC plasma samples [
149]. Furthermore, Liang et al. identified 2,230 proteins in exosomes secreted from OVCAR-3 and IGROV1 ovarian cancer cell lines. Many of these identified proteins were involved in tumorigenesis and metastasis, indicating the prognostic potential of exosomal profiling [
150]. Overall, exosomal profiling can act as a cancer-specific diagnostic and prognostic biomarker and replace invasive cell biopsies. However, more comprehensive clinical studies are required to determine the clinical value of this approach.