Molecular Biomarkers of Cancer Stem/Progenitor Cells Associated with Progression, Metastases, and Treatment Resistance of Aggressive Cancers

Significant advancement in basic and clinical oncology during the last few years has led to earlier diagnosis and more effective therapeutic management of patients with leukemias and organ-confined tumors in the clinics (1–3). Although the surgical tumor resection may result in some cases to a complete remission, the rapid cancer progression of aggressive cancers to locally invasive and metastatic stages is generally associated with the development of resistance mechanisms by cancer cells to current antihormonal, radiation, and/or chemotherapeutic treatments and disease relapse (1–3). At the present time, the metastatic cancers remain the leading cause of the death of patients with cancer. Therefore, many research efforts have been made to identify and validate novel molecular biomarkers and therapeutic targets in cancer cells at primary and secondary tumors to prevent cancer progression and metastases and optimize the genetic- and proteomic-based individualized treatments of patients with cancer (Fig. 1; refs. 4–28).

Importantly, accumulating lines of evidence have revealed that the shedding of cancer cells from the primary tumors into the lymphatic vessels and peripheral circulation can occur very early during the cancer development and be dependent of cellular origin, genetic alterations, and aggressiveness of cancer subtypes (16, 29–41). Hence, some patients who undergo a complete surgical tumor resection with negative margins may show the presence of circulating tumor cells (CTC) in the peripheral blood and disseminated tumor cells at the regional lymph nodes and distant tissues and organs (Fig. 1; refs. 16, 29–41). Consequently, CTCs that are able to survive in the bloodstream and spread at distant sites can persist and contribute to metastases and disease relapse even after an effective and apparently curative surgical resection of the primary tumor. In this regard, a growing body of experimental evidence has also revealed that cancer stem/progenitor cells endowed with stem cell–like properties, also designated as cancer-, tumor-, and metastasis-initiating cells, can provide critical functions for tumor growth, metastases at near and distant tissues and organs, treatment resistance, and disease relapse. In fact, it has been shown that the most cancers may originate from the malignant transformation of immature tissue-resident stem/progenitor cells or their early differentiated progenies endowed with a high self-renewal ability and aberrant differentiation potential (2, 42–44). The cancer stem/progenitor cells expressing specific stem cell–like markers such as CD133, CD44^high, nestin, aldehyde dehydrogenase (ALDH^high), and high levels of ATP-binding cassette (ABC) multidrug transporters have also been identified and isolated from primary and secondary neoplasms, including leukemias, melanomas, brain tumors, and the most epithelial cancers and cancer cell lines (9,17, 24, 44–76). It has been shown that cancer stem/progenitor cells were able to give rise to the total tumor cell mass, including differentiated cancer cells that reconstituted the histological architecture and molecular characteristics of primary and secondary tumors closely resembling to original patient's tumors in vivo (9, 17, 45–57, 59–66, 68, 69, 71, 77). Moreover, the data from recent studies have indicated that cancer stem/progenitor cells may be more resistant than their differentiated progenies to current antihormonal, radiation and chemotherapeutic treatments, and targeted therapies (17, 22–25, 44, 52, 53, 56–64, 68, 70, 72, 77–94).

We review here recent advances on the characterization of gene products often altered in cancer stem/progenitor cells and their differentiated progenies during primary cancer progression and dissemination through the peripheral circulation and metastases. The emphasis is on molecular gene signatures, epithelial–mesenchymal transition (EMT)-like and stem cell–like biomarkers detected in cancer cells and exosomes associated with cancer progression, treatment resistance, disease relapse, and poor outcome of patients with cancer.

The most cancers are heterogeneous diseases and include some biological subtypes characterized by specific genetic and/or epigenetic alterations occurring in cancer stem/progenitor cells and their differentiated progenies during cancer initiation and progression to locally invasive and metastatic stages (4, 6, 7, 44, 60, 78, 80, 95–109). Although the molecular events that govern the spread of cancer stem/progenitor cells and their differentiated progenies at preferential metastatic sites remain not precisely defined, some lines of evidence suggest that particular gene signatures of cancer cells at primary cancers may be associated with their propensity at metastasizing at specific tissues and organs (4, 6, 7, 95–102). More particularly, the alterations in the gene products involved in the acquisition of aggressive and migratory properties by highly tumorigenic cancer stem/progenitor cells during the EMT program at the primary tumor may result in their invasion to near lymph nodes and tissues, dissemination through the peripheral circulation, and metastases at distant sites (Fig. 1; refs. 30, 42, 90, 91, 96, 97, 110–126). Moreover, different gene products have also been identified in exosomes released by tumor cells and CTCs with stem cell–like features that may be used as diagnostic, prognostic, and/or predictive biomarkers. In this regard, we review the gene products frequently altered in cancer cells, during cancer progression to locally advanced and metastatic states, exosomes and CTCs that have been associated with treatment resistance, disease relapse, and poor overall survival of patients with cancer.

Some tissue-specific and whole genome approaches for molecular profiling of complex gene signatures of tumors and cancer cell lines have been developed in the last few years and led to the discovery of specific gene sets related to particular cancer types and which may predict the risk of metastatic dissemination at specific tissues and organs and overall outcome of patients with cancer (Fig. 1; refs. 4–15). Moreover, the in vitro and in vivo characterization of stem cell–like markers and functional features of cancer stem/progenitor cells in tumor tissue specimens from patients with cancer and cancer cell lines has also revealed their major implications in tumor growth, metastases, treatment resistance, and disease relapse (Fig. 1; refs. 15–27). We are reporting here the gene signatures and stem cell–like biomarkers of cancer cells associated with particular subtypes of aggressive cancers, including hormone-independent prostate and breast cancers.

Prostate cancer is among the most commonly diagnosed malignancies affecting men (127–129). Although an earlier diagnosis of patients with prostate cancer has led to an improvement of their effective treatment by tumor resection surgery and/or radiation therapy, the locally advanced and metastatic castration-resistant prostate cancers (CRPC) still represent the second leading cause of cancer-related death (127–129). In fact, antihormonal treatments and first-line docetaxel-based chemotherapies of metastatic CRPCs are only palliative and result in the death of most patients after about 12 to 19 months (127–129). Therefore, this inefficacy of current treatments against CRPCs underlines the importance to identify novel biomarkers to predict the risk of propagation to CRPCs. Of clinical relevance, a 11-gene signature related to the activation of polycomb-group protein BMI-1 pathway, which provides critical functions for the self-renewal of diverse normal and cancer stem cells, has notably been identified through a comparative genomic approach using primary and metastatic tumor specimens from transgenic adenocarcinoma of the mouse prostate model and patients with prostate cancer (4). The 11-gene signature includes 8 upregulated genes (Ki67, cyclin B1, gastrulation brain homeobox 2, budding uninhibited by benzimidazoles 1 “BUB1” that acting as serine/threonine-protein kinase, kinetochore associated 2, ubiquitin-specific protease 22, host cell factor 1, and ring finger protein 2) and 3 downregulated genes (ankyrin 3, fibroblast growth factor receptor 2, and carboxylesterase 1; ref. 4). This 11-gene signature, which is associated with an upregulation of stem cell–resembling expression profile of the BMI-1–regulated pathway, has been shown to be a powerful predictor of patients with prostate cancer progression to locally invasive and metastatic stages (4). The analyses of this 11-gene signature of BMI-1–driven pathway in 1,153 clinical tumor specimens from patients with cancer diagnosed with 11 different types of cancer have also indicated that the expression of these gene products was associated with a short interval to disease recurrence, distant metastases, and death after therapy of patients with cancer (4). These cancer types comprise acute myeloid leukemia, lymphoma, mesothelioma, medulloblastoma, glioma, and prostate, breast, lung, ovarian, and bladder epithelial cancers (4). Importantly, high expression levels and/or activities of stem cell–like markers such as CD133, CD44, integrin-α₂, nestin, ALDH1, and CD49f in primary prostatic adenocarcinomas and bone metastases from patients and prostate cancer cell lines have also been associated with their aggressive and invasive phenotypes and treatment resistance (17–26, 28). Particularly, the data from immunohistochemical analyses of integrin-α₂, integrin-α₆, and hepatocyte growth factor receptor (MET) in prostate cancer cells have revealed that an increase of expression levels of these markers in prostatic adenocarcinoma specimens from patients with prostate cancer was associated with an enhanced risk of local invasion and bone metastasis (130). It has also been reported that several ALDH isoforms were expressed in specimens of primary prostate tumors with matched bone metastases and prostate cancer cell lines, and ALDH^high PC3M-Pro4 cells rapidly formed primary prostatic tumor and metastasized at distant sites in vivo as compared with ALDH^low PC3M-Pro4 cells (18, 19). Of therapeutic interest, the knockdown of ALDH7A1 expression in PC3M-Pro4 cells has also been observed to result in a decrease of the integrin-α₂^high/integrin-α_v^high/CD44⁺ PC3M-Pro4 stem/progenitor cell subpopulation and gene products involved in the EMT process and invasion, including snail, snail2, twist, and extracellular molecule, osteopontin (19). Moreover, the ALDH7A1 silencing in PC3M-Pro4 cells significantly inhibited their clonogenic and migratory abilities in vitro as well as their capacity to form bone metastases in vivo (19). In the same way, it has also been observed that prostate-specific antigen (PSA^−/lo) prostate cancer cells from LNCaP, LAPC4, or LAPC9 cell lines or tumor specimens from patients with prostate cancer expressing stem cell–like markers such as ALDH, CD44, integrin-α₂β₁, and with a long-term tumor-propagating ability were able to give rise to the total mass of PSA⁺ prostate cancer cells in vivo (17). It has also been noted that PSA^−/lo LAPC9 cells were resistant to androgen deprivation in castrated hosts, and high grade and recurrent prostatic adenocacinoma specimens of patients were enriched in PSA^−/lo prostate cancer cells (17). These observations suggest that PSA^−/lo prostate cancer cells may play critical functions for the tumor regrowth after antihormonal treatment initiation.

In addition, we have observed that sonic hedgehog (SHH/GLI-1), epidermal growth factor receptor (EGFR), pAkt, nuclear factor-κB (NF-κB), and macrophage inhibitory cytokine-1 (MIC-1) are frequently overexpressed in prostate cancer cells, including CD133⁺ prostate cancer stem/progenitor cells, detected in prostatic adenocarcinoma and bone metastasis specimens from patients with prostate cancer (23, 24). The high expression levels and/or activities of these oncogenic products were also detected in the side population from tumorigenic and invasive WPE1-NB26 cells and metastatic PC3M-LN4 endowed with a high prostasphere-forming ability (23–25). Moreover, the targeting of hedgehog, EGFR, pAkt, NF-κB, or MIC-1 by using cyclopamine, gefitinib, Akt inhibitor III, partenolide, or anti-MIC-1 antibody was effective at inducing the apoptotic effects on side population and non-side population cells from WPE1-NB26 and PC3M-LN4 cell lines and improving cytotoxic effects induced by docetaxel on these side population and non-side population cells (23, 24). Recent studies have also revealed that the changes in the expression levels and activities of hypoxia-inducible factors (HIF) and microRNAs (miRNA), which are short noncoding RNA molecules of about 22 nucleotides that are involved in the stringent control of normal stem cell behavior often occur in cancer cells including prostate cancer cells (27, 28, 131, 132). HIFs and miRNAs may be involved in the epigenetic reprogramming and acquisition of stem cell–like features by cancer cells (27, 28, 131, 132). More specifically, the upregulation of HIF-1α and HIF-2α has been associated with the increased expression of pluripotency-associated transcription factors (Nanog, Oct4, and Sox2) and EMT program-associated molecules in prostate cancer cells under normoxic and hypoxic conditions during prostate cancer progression to locally invasive and metastatic stages (27, 28). The downregulation of some tumor-suppressor miRNAs such as mi-R34a, let-7b, miR-106a, and miR-141 combined with an upregulation of miR-301 and miR-452 has also been observed to regulate the proliferation and tumorigenicity of CD133/CD44/integrin-α₂β₁ prostate cancer stem/progenitor cell population and side population cells in vitro and in vivo (131, 132).

Breast cancer is among the most common cancers in women (127, 133, 134). Although a high survival rate is observed for early-stage patients with breast cancer treated by tumor resection surgery and adjuvant treatments, the propagation to locally invasive and metastatic disease state is generally associated with the development of resistance mechanisms by cancer cells to current antihormonal treatments, targeted therapies, radiation, and chemotherapies (127, 133, 134). Typically, breast cancer is a complex and highly heterogeneous disease that encompasses distinct cancer subtypes characterized by the malignant transformation of the basal and/or luminal breast epithelial cells in the mammary gland (135, 136). In this regard, the data from analyses of the expression levels of gene products in breast cancer cells, including breast cancer stem cells (BCSCs) with stem cell–like features, have indicated that breast cancer subtypes may exhibit cellular and functional heterogeneities (5–14, 137, 138). It has been observed that CD44^high/CD24^−/low BCSCs are frequently detected in basal-like subtypes of breast cancer whereas these immature cancer cells are very low in erbB2/HER2⁺ breast tumors (8). Moreover, a transcriptional analysis of basal-like breast cancer cell lines has led to the identification of 2 major cancer subclasses in which basal B seemed mesenchymal-like, whereas basal A may have either a luminal-like or basal-like morphology. It has been observed that a CD44^high/CD24^low epithelial basal A cell subpopulation isolated from triple-negative HMT3909S13 breast cancer cell line exhibited cancer stem cell features such as mammosphere formation and resistance to the standard chemotherapy as well as tumor-initiating capacity in vivo (6). In contrast, CD44^high/CD24⁻ mesenchymal-like basal B cell fraction does not exhibit these functional features (6). The data from comparative quantitative proteomic and gene array analyses of CD44^high/CD24^low BCSCs with an epithelial-like morphology have also indicated that these immature cancer cells expressed lipolysis-stimulated lipoprotein receptor, RAB25, S100A14, and mucin 1 (MUC1; ref. 6). These CD44^high/CD24^low BCSCs also expressed a novel 31-gene signature that was related at the presence of distant metastases in estrogen receptor negative (ER⁻)-human breast cancer specimens (6). Conversely, the results from another investigation have revealed that CD44^high/CD24^low BCSCs that were detected in patient's primary tumors exhibited a gene signature as the claudin-low molecular subtype, which is characterized by the expression of several EMT program-associated molecules (10). It has been noted that the CD44^high/CD24^low BCSCs expressing mesenchymal markers were enriched in patient's primary tumors after either endocrine letrozole-based therapy or docetaxel-based chemotherapy (10). These findings suggest that breast cancer cells endowed with an epithelial or mesenchymal phenotype can exhibit cancer stem cell properties dependent of breast cancer subtypes and be involved in the metastasis formation and treatment resistance.

On the other hand, the data from analyses of global gene expression profiles have also indicated that a common subset of 81 genes was differentially expressed between ER⁻ and ER⁺ breast tumor specimens and breast cancer cell lines as well as in androgen-independent versus androgen-sensitive prostate cancer cell lines (11). It has been noted that some signaling elements of EGFR pathway were enriched in ER⁻ breast and androgen-independent prostate cancer cell lines, suggesting that the sustained activation of this tumorigenic cascade may contribute to the growth and survival of these hormone-independent cancer cells (11). Interestingly, it has been reported that an 8-gene signature (EGFR, integrin-α₃, myosin light chain kinase, retinoic acid–induced protein 14, AHNAK nucleoprotein, glutaminase, interleukin-32, and nicotinamide N-methyltransferase) was associated with the invasion and sensitivity of a panel of 78 tumor cell lines to chemotherapeutic drugs, including paclitaxel, docetaxel, erlotinib, everolimus, and dasatinib (14). This set of 8 genes also predicted the relapse-free survival for patients with breast cancer (14).

Recent investigations have also led to the discovery of specific gene signatures and potential biomarkers in cancer tissue specimens for improving the personnel management of patients with other aggressive cancers such as brain, melanoma, esophageal squamous cell carcinoma (ESCC) and lung, pancreatic, colorectal, esophageal, and ovarian cancers (Table 1; refs. 3, 14, 70, 139–149). For instance, the real-time PCR data from gene expression profiles of brain tumor stem-like cells (BTSCs) from 5 human glioblastoma multiforme (GBM) and matched GBM tissue specimens have indicated that SHH and GLI1 hedgehog signaling elements were coexpressed only at high levels in BTSCs 1 to 3 and not in BTSCs 4 and 5 (141). It has also been noted that the SHH/GLI1 cascade-induced BTSC proliferation via phosphatidylinositol-3′ kinase (PI3K)/Akt/mTOR, was inhibited by small hairpin RNA (shRNA) against human GLI1 or rapamycin in vitro and in an intracranial tumor model in athymic mice (141). Of clinical interest, higher expression levels of SHH, patched receptor 1 (PTCH1), and GLI1 have also been detected in phosphatase and tensin homolog deleted on chromosome 10 (PTEN)-expressing brain tumors than in PTEN-deficient brain tumors and associated with a poorer survival time of GBM patients (141). Furthermore, it has been shown that melanoma cells from patient with melanoma specimens expressing high levels of ALDH1A1 and ALDH1A3 isoforms displayed a high self-renewal ability, and were more resistant to chemotherapeutic agents and more tumorigenic than ALDH⁻ melanoma cells in nonobese diabetic (NOD)/severe combined immonodeficient (SCID) mice (139). It has also been noticed that the silencing of the ALDH1A gene by shRNA led to an arrest in cell cycle and apoptosis of ALDH1A⁺ melanoma cells, inhibition of tumor growth in vivo, and sensitized the ALDH1A⁺ melanoma cells to cytotoxic effects of chemotherapeutic drugs (139). These findings implicate that ALDH1A and ALDH3A isozymes are putative biomarkers for melanoma stem cells and constitute attractive therapeutic targets for treating melanoma patients. In the same way, human pancreatic ductal adenorcinoma (PDAC) specimens have also been observed to contain a side population exhibiting the gene expression signatures that are associated with pancreatic cancer stem cell–like characteristics and resistance to chemotherapeutic drugs (144). Among these genes, expression levels of ABCB1 multidrug transporter and CXC chemokine receptor 4 (CXCR4) were correlated with a worse survival of patients with pancreatic cancer (144).

Altogether, these studies have led to the identification and validation of gene expression signatures and potential diagnostic, prognostic, and predictive biomarkers related to stem cell–like cancer cells, EMT program, treatment resistance, and relapse of different aggressive cancers. Future investigations are however necessary to validate the expression of these gene signatures and their implications in the treatment resistance on larger cohorts of tumor tissue specimens from patients with cancer. In this regard, recent studies have also revealed the possibility to analyze the phenotypic features of exosomes secreted by cancer cells and CTCs, including circulating cancer stem/progenitor cells, to assess the tumor stage and risk of cancer progression, metastases, and disease relapse of patients with cancer.

Accumulating lines of evidence have indicated that exosomes that are secreted by normal and cancer cells can act as key mediators in signal transduction mechanisms that are involved in short- and long-distance cell-to-cell communication (150–157). In fact, the exosomes are small microvesicles (sized about 50–100 nm) derived from cellular endocytosis that may contain specific lipids, proteins, mRNAs, and miRNAs encapsulated in a cholesterol-rich phospholipid membrane (150–157). More particularly, the exosomes that are released by cancer cells in their surrounding microenvironment and peripheral circulation may adhere via their antigenic surface molecules or transfer their cytoplasmic contents through membrane fusion to their targeted host cells, including other cancer cells, stromal cells, endothelial cells, and immune cells (Fig. 1; refs. 150–157). Hence, the shedding of exosomes by cancer cells under normoxic and hypoxic conditions may modulate immune system and promote the malignant behavior of recipient cells in their tumor microenvironment and angiogenesis as well as the formation of premetastatic niches, metastases at distant tissues and organs, and treatment resistance (Fig. 1 and Table 2; refs. 150–160). Importantly, recent advances in the development of isolation techniques of exosomes from some types of body fluids or cancer cell culture supernatants using centrifugation, ultrafiltration, immunoprecipitation, magnetic activated cell sorting (MACS), and immunoaffinity capture methods have led to an improvement of the purity of exosomal preparations (150–159, 161–164). Hence, several studies carried with these isolation methods have indicated the presence of exosomes in blood, plasma, serum, lymph, urine, cerebrospinal fluid, or ascites from patients with melanoma and brain, ovarian, breast, pancreatic, colon, bladder, prostate, and lung cancers as well as conditioned medium of different cancer cell types (Fig. 1 and Table 2; refs. 150–159, 161–164). It has also been observed by proteomic profiling that different exosomal preparations of different cancer types frequently expressed a set of biomarkers. The common biomarkers of most exosomes include tubulin, actin, tetraspanins (CD9, CD63, CD81, and CD82), Alix, TSG101, clathrin, small GTPases (Rab proteins), annexins, heat-shock proteins (Hsp70 and Hsp90), 14-3-3 proteins, and MHC class I molecules that are involved in the biosynthesis, cytoskeletal structure, and trafficking of exosomes (159, 162–164). Furthermore, tumor cell–derived exosomes also harbored specific RNAs, miRNAs, and proteins, including stem cell–like markers (CD133, CD44, and/or CD105) and oncogenic products (EGFR, EGFRvIII mutant, erbB2/HER2/Neu, MET, claudins, and/or FASN) that are related to cancer progression, metastases, and treatment resistance (150–159, 161–166). For instance, the data from proteomic analyses of highly purified exosomes derived from 2 ovarian cancer cell lines, OVCAR-3 and IGROV1, have indicated that several proteins of tumorigenic pathways that are overexpressed in ovarian cancer tissues and associated with migration, invasion, immune modulation, angiogenesis, and metastasis were also detected in exosomes (162). These oncogenic products comprise EGFR, epithelial cell surface antigen (EpCAM), proliferation cell nuclear antigen (PCNA), tubulin β-3 chain (TUBB3), apolipoprotein E (APOE), and claudin 3 (CLDN3; ref. 162). Moreover, the exosomes expressing T-cell receptor (TCR), CD20, HLA-DR, B7-2, HER2, CA125, and histone H2A have been detected in 85.4% (35/41) of ascites samples from patients with ovarian cancer (167). It has also been shown that the exosomes induced the apoptosis of precursor and mature dendritic cells and peripheral blood mononuclear cells (PBMC) and impaired the cytotoxic activity of PBMCs in the presence of dendritic cells suggesting their implication in the modulation of antitumor immunity (167).

In addition, the tumor-specific and constitutively active form of EGFR designated as EGFRvIII mutant has also been detected in serum exosomes from 7 of 25 patients with glioblastoma (152). It has been noted that EGFRvIII protein expressed by glioma cells may be transferred via exosomes to other glioma cells lacking EGFRvIII, and thereby promoted their oncogenic transformation via the activation of PI3K/Akt and mitogen-activated protein kinase (MAPK) pathways (Figs. 1 and 2A and Table 2; ref. 153). These molecular events induced the changes in the expression of EGFRvIII-regulated genes such as vascular growth factor “VEGF,” Bcl-xL, and p27 in host glioma cells and increased their anchorage-independent growth capacity (153). It has also been observed that exosomes released from glioblastoma tumor cells also contained mRNA, miRNA, and angiogenic proteins that could be transferred to host cells, including normal brain microvascular endothelial cells, and thereby stimulate the tubule formation (153). In the same way, cancer stem cells expressing the mesenchymal stem cell marker CD105 from human renal cell carcinoma also released exosomes that contained proangiogenic mRNAs and miRNA (168). It has also been noted that the exosomal mRNAs and miRNA in turn activated the growth of normal endothelial cells and induced the vessel formation and angiogenesis after implantation in SCID mice (168). Moreover, the treatment of SCID mice with exosomes from CD105⁺ cells was effective at promoting the formation of lung metastases induced by intravenous injection of renal carcinoma cells (168).

Importantly, several lines of experimental evidence have also indicated that the exosomes may be involved in the formation of premetastatic niches and transfer of more aggressive and drug-resistance phenotypes to cancer cells and modulation of treatment responses of patients with cancer (Table 2; refs. 150, 169–175). For instance, CD133-expressing exosomes purified from FEMX-I cell line, which was derived from a lymph node metastasis of a patient with malignant melanoma, have been observed to express several prometastatic proteins, including CD44, MAPK4K, GTP-binding proteins, ADAM10, and annexin A2 and high levels of miRNAs with cancer-related functions (173). It has been shown that primary melanoma-derived exosomes can induce the recruitment and growth of melanoma cells to sentinel lymph nodes in vivo by inducing extracellular matrix deposition and vascular proliferation in the lymph nodes (174). Moreover, the exosomes from highly metastatic melanomas also contributed to the premetastatic niche formation through the upregulation of the MET oncoprotein in CD45⁻/KIT^low/+/tyrosine kinase with immunoglobulin-like and EGF-like domains-2 (TIE2⁺) bone marrow-derived progenitor cells (Fig. 1; ref. 169). The enhanced expression of MET by bone marrow progenitor cells resulted in their reprogramming toward provasculogenic and prometastatic phenotypes (169). The gene signature of exosomes released by melanoma, including MET, tyrosinase-related protein-2, very late antigen-4, HSP70, and HSP90, has also been associated with a poor prognosis of patients with melanoma (Table 2; ref. 169). Moreover, it has been reported that the exosomes isolated from serum samples of patients with prostate cancer promoted the proliferation and invasion of prostate cancer cell lines whereas the exosomes from age-matched healthy controls have not significant effects (150). The exposure of DU145 and 22Rv1 prostate cancer cells to exosomes from serum samples of patients with prostate cancer treated with docetaxel was also effective at enhancing their docetaxel resistance as compared with exosomes from untreated prostate cancer patient's sera (150). It has also been noticed that docetaxel-resistant DU145RD and 22Rv1RD cell lines secreted the exosomes expressing P-glycoprotein, which could be transferred and conferred docetaxel resistance to parental DU145 and 22Rv1 cell lines (Fig. 1; ref. 150). The PCR-based analyses of urine samples from patients with prostate cancer have also indicated that the transcripts of prostate cancer gene-3 (PCA-3) and TMPRSS2;ERG gene fusion were detected in urinary exosomes suggesting their potential use as a noninvasive method for the diagnosis and follow-up of patients with prostate cancer (161). Moreover, it has also been shown that the exosomes released by SKBR3 breast cancer cells overexpressing HER2 oncoprotein interfered with the antitumoral activity of the anti-HER2 antibody, trastuzumab (Herceptin; ref. 172).

The molecular targeting of tumor cell–derived exosomes by interfering with their biosynthesis, secretion, and/or interactions with cell antigens or receptors also represents a promising approach to prevent the cancer progression, premetastatic niche formation, and metastases at distant sites (Fig. 2; refs. 160, 169, 176–179). For example, the targeting of neutral sphingomyelinase 2, which is involved in the ceramide biosynthesis and upregulation of exosome secretion, with chemical drug GW4869 has been shown to inhibit the lung metastases in lung cancer cell-bearing mice in vivo (176). Moreover, the interference with molecular pathways of exosome secretion by shRNA silencing of small GTPase Rab27a or Rab27b or 2 Rab27 effectors, known as Slp4 and Slac2b, also inhibited the exosome secretion by cervical adenocarcinoma HeLa cells, 4T1 and TS/A mammary carcinoma cells, and melanoma cells (169, 177, 178). It has also been noted that Rab27a blockade resulted in decreased primary tumor growth and lung dissemination of a metastatic carcinoma (4T1), but not of a nonmetastatic carcinoma (TS/A; ref. 177). In A431 human tumor xenografts in mice, angiogenic endothelial cells stained positively for human EGFR and phospho-EGFR, whereas treatment with annexin homodimer, diannexin which cloak phosphatidylserine residues on the surfaces of exosomes led to a reduction of tumor growth rate and microvascular density (Fig. 2A; ref. 160). Of particular clinical interest, a novel medical device using an affinity plasmapheresis platform termed as Aethlon ADAPT system was also effective to perform extracorporeal hemofiltration and elimination of tumor cell–derived exosomes from the entire circulatory system (Fig. 2B; ref. 179).

Recent investigations have also revealed that exosomes could be used as natural cell–derived membrane vehicles for systemic delivery of therapeutic agents interfering with the functions of tumor-specific mRNAs, miRNAs, and oncogenic products in target cells as well as immune regulatory tools in cancer immunotherapy (Fig. 2C; refs. 180–183). As a matter of fact, exosomes engineered to deliver spherical nucleic acid constructs into PC3 prostate cancer cells downregulated miR-21 target by approximately 50% (182). Furthermore, it has been shown exosomes that are released by tumor cells and dendritic cells may be used to elicit-specific immune responses against established tumors (180, 181). For instance, the data from a phase I clinical trial have indicated that a combination of autologous ascites-derived exosomes with granulocyte macrophage colony-stimulating factor (GM-CSF) was safe and well tolerated and induced tumor antigen-specific cytotoxic T-cell responses without side effects in patients with advanced colorectal cancer (180).

Together these data suggest important roles of exosomes secreted by cancer cells in their surrounding microenvironment in the acquisition of a more malignant behavior by host cancer cells and promoting immune escape and tumor neovascularization during cancer progression. Moreover, the exosomes can play critical functions in the establishment of premetastatic niches, recruitment, and homing of cancer cells at distant tissues and organs, metastases, and treatment resistance. Hence, these results support the interest to use the exosome-specific biomarkers as new diagnostic, prognostic, and predictive indicators and to develop new exosome-based-targeted therapies for treating aggressive and metastatic cancers.

The enumeration of CTCs isolated from whole blood samples of patients with metastatic cancers before or posttreatment initiation and characterization of their molecular biomarkers may constitute a noninvasive and real-time method for improving accuracy of diagnosis and prognosis and choice of their therapeutic management, including use of appropriate targeted therapy (16, 31–35, 184). Several strategies, including CellSearch platform, microfluidic devices, and cell size–based separation methods, have been developed to detect, enrich, and isolate rare CTCs from peripheral blood samples (Fig. 1 and Table 3; refs. 16, 31–34). These detection techniques involve the CTC immunocapture from peripheral blood samples using different specific antibodies or size-based filters (Fig. 1 and Table 3; refs. 16, 31–34). More particularly, the CellSearch technology, which consists to use immunomagnetic beads coated with an antibody directed against EpCAM molecules, can allow researchers to isolate epithelial cancer cells expressing EpCAM surface marker (16, 31, 33–35, 184, 185). In a second step, the analyses of different biomarkers of the interest expressed by captured CTCs can be performed by quantitative real-time reverse transcription PCR (qRT-PCR) using DNA extracted from CTCs or immunofluorescence staining of CTCs (Fig. 1; refs. 16, 31, 34, 184). The biomarkers of CTCs include stem cell–like markers (CD133, CD44, ALDH, and ABC multidrug transporters), cytokeratins, EGFR, EGFRvIII, HER2, and K-Ras as well as the absence of the leukocyte-specific marker CD45 (Table 3; refs. 16, 31, 34, 184). Hence, it has been observed that the number and phenotypic features of CTCs captured using CellSearch assay, microfluidic CTC-ships, and other detection techniques given significant information on the treatment benefit and helped to predict the overall outcome of patients with locally advanced and metastatic breast, prostate, ovarian, pancreatic, lung, colorectal, and pancreatic cancers (Table 3; refs. 16, 31–35, 37, 184, 186–189). For instance, the characterization of CTCs from patients with metastatic CRPC has indicated that the majority of CTCs coexpressed epithelial proteins (EpCAM, cytokeratins, and E-cadherin), mesenchymal markers (vimentin, N-cadherin, and O-cadherin), and the stem cell–like marker CD133 (34).

In addition, the results of qRT-PCR analyses of CTCs from patients with breast cancer have also indicated a direct relationship between the expression of stem cell–like markers such as Nanog, Oct3/4, Sox-2, nestin, and CD34 on CTCs and the stage of the disease progression (36). Moreover, the analyses of blood samples from 502 patients with breast cancer with the AdnaTest BreastCancer Select for the detection of EpCAM, MUC1, HER2, and β-actin transcripts have revealed that CTCs were detected in 19% of samples, and 29% samples expressed at least one of EMT markers (twist1, Akt2, and PI3Kα) whereas ALDH1 stem cell–like marker was detected in 14% of the samples (38). In another study, CTCs were also detected in 31% of 226 blood samples from 39 patients with metastatic breast cancer and 62% were positive for at least one of EMT markers and 69% for ALDH1 (39). It has also been noticed that the expression of EMT markers and ALDH1 was detected in 62% and 44% of patients with breast cancer that were nonresponders to the therapy whereas only 10% and 5% of the responders expressed these markers, respectively suggesting their potential use as predictive biomarkers (39).

More recently, single CTCs and multicellular clusters in a viable state have also been isolated from bloodstream of patients with metastatic lung cancers by dean flow fractionation (40). This size-based separation method may overcome the problems associated with the immunocapture with antibodies that is dependent of antigens expressed by CTCs (40). The number of CTCs detected by dean flow fractionation in whole blood samples from 20 patients with metastatic lung cancer was 39.1 ± 24.8 CTCs/mL whereas CTCs in blood samples from 20 healthy individuals was negligible (0.79 ± 0.42 CTCs/mL; ref. 40). It has also been observed that the CTC population detected in blood samples from patients with advanced lung cancer contained cells coexpressing cytokeratin and CD133 stem cell–like marker (40). Moreover, the qRT-PCR analyses of peripheral blood samples from patients with primary melanoma (early stages 0–II, n = 154) and metastatic melanoma (late stages III–IV, n = 76) have indicated that all samples contained CTCs (41). It has also been noted that melan-A known as melanoma antigen and stem cell–like marker, ABCB5 multidrug transporter were expressed in 45% and 49% of patients showing melanoma recurrence after treatment initiation, respectively (41).

The use of specific biomarkers of CTCs, including stem cell–like markers, represents a promising strategy for improving the diagnosis, prognosis, and therapeutic management of patients with aggressive and metastatic cancers.

The characterization of gene products expressed in tumor tissue specimens and CTCs from patients with cancer and exosomes released by cancer cells has led to the identification and validation of some gene signatures and potential stem cell–like markers. These altered gene products could be used as noninvasive diagnostic, prognostic, and predictive tools in conjunction for monitoring the risk of tumor progression and metastases and determining the choice of therapeutic treatments of patients with cancer. Furthermore, the molecular targeting of specific markers of cancer stem/progenitor cells, CTCs, and exosomes also represents a potential strategy could prevent cancer progression and metastases and overcome the treatment resistance and disease relapse of patients with cancer.

Future investigations are however required to optimize the sensitivity and reproducibility of immunocapture methods used for the enrichment and enumeration of CTCs and circulating exosomes from whole blood samples to consider their low number, potential heterogeneity, and specific biomarkers. Indeed, the detection and purification of rare CTCs, including circulating cancer stem cells, and circulating exosomes may be obscured by the nonspecific background of other hematopoietic cell types and overwhelming material in whole blood samples. Therefore, the optimization of in vivo detection platforms suitable for a continuous monitoring of whole blood volume could markedly improve the probability to detect rare CTCs, including circulating cancer cells with stem cell–like features. Moreover, the use of additional antibodies directed against epithelial, mesenchymal, and/or stem cell–like surface markers of CTCs could be included in immunoaffinity capture methods, such as CellSearch platform and microfluidic ships, to detect the EpCAM⁻ CTCs or exosomes. It will also be of interest to further establish the relationship between the expression levels of EMT process- and stem cell–like markers detected on captured CTCs and exosomes and the stage of tumor progression, treatment response, and survival rate using a large cohort of patients with cancer. Hence, the combined use of these new biomarkers detected in primary and metastatic tumor tissues, CTCs, and exosomes should lead to the development of multiplex biomarker approaches for improving the accuracy of current diagnostic and prognostic tests and efficacy of individualized treatments for patients with cancer. Of clinical relevance, the development of new nanotheranostic platforms for the combined in vivo detection by imaging and nanobubble-based targeted therapies of CTCs and circulating exosomes also may constitute a promising approach for real-time diagnosis and treatment of patients with aggressive and metastatic cancers.

No potential conflicts of interest were disclosed.

The authors were supported by the grants from the NIH (R01CA138791, EDRN U01, and SPORE P50 CA127297).