Characterization of stem cell and cancer stem cell populations in ovary and ovarian tumors

Tissue samples were collected from ovarian cancer patients admitted to the James Graham Brown Cancer Center, University of Louisville. Procedures followed were as per the ethical protocols approved by the Institutional Review Board (IRB). Informed patient consent was obtained prior to surgery. Of total 50 samples, 34 ovarian tissue samples were obtained from the Department of Pathology, University of Louisville after thorough examination by an experienced pathologist (n = 4 [NO], n = 10 [BN], n = 10 [BL] and n = 10 [HG]) [NO: Normal Ovary, BN: Benign tumor, BL: Borderline tumor, HG: High grade metastatic tumor]. Sixteen fresh/frozen samples (n = 4 for each category - NO, BN, BL and HG) were obtained from institutional bio-repository and were used for reverse transcription polymerase chain reaction (RT-PCR) analysis. All the experiments were performed in triplicates. Chemicals, reagents and plastic-ware were procured from Sigma-Aldrich or stated otherwise.

Paraffin embedded ovarian sections from NO, BN, BL and HG were prepared by standard histology procedures. Sections were de-paraffinized in xylene. Endogenous peroxidase activity (abundant in the red blood cells) was quenched by incubating the sections in 3% hydrogen peroxide in 100% methanol for 60 min followed by rehydration in alcohol grades. Antigen retrieval and staining of each tissue section for specific stem cell/CSC marker was performed as described previously [10]. Immuno-staining pattern was assessed by scanning the slides using an Aperio Scanscope digital pathology scanner (Leica Biosystems) and Olympus IX71 inverted microscope (Olympus, Tokyo, Japan). Representative images were photographed and specific staining signals were documented.

Protocol for immunofluorescence staining and dual staining (co-localization studies) was similar to immuno-histochemistry. Endogenous peroxidase (3% H₂O₂) quenching step was omitted during immunofluorescence staining and intermediate washing steps were performed using PBS containing 0.5% BSA as described previously [10]. Antibody dilution for each was optimized to provide specific binding with no background as detailed in Table 1. Tissue sections were viewed using Nikon (Eclipse TI) confocal microscope and documented using NIS Elements AR software (version 4.5.1). The slides were thoroughly scanned under the confocal microscope and representative 10–12 images for each section were captured at 40X with 2.5X zoom to focus upon positive signals emitted by small 5–10 μm (cancer) stem cells of interest in each sample. Dual immuno-fluorescence protocol was identical to immuno-fluorescence protocol, except each of the two primary and secondary antibodies were applied simultaneously to each tissue section on day 1 and 2 respectively. Alexa Fluor 488 and 568-labeled anti-mouse and anti-rabbit IgG were used at 1:500–1:1000 dilution followed by counter staining with nuclear dye DAPI. The sections were examined under confocal microscope. Uniform instrument parameters were employed for confocal microscopy imaging for each antibody and on an average 7–10 images (per sample/antibody) were captured, further analysed and compiled to prepare the composite panels comprising of individual cells captured from various regions of a single field or from multiple fields that were scanned and documented.

RNA from NO, BN, BL and HG tissues was prepared, purified and subjected to RT-PCR amplification using specific primers (detailed in Table 2) and conditions (described in [10]). Amplified products were analyzed on 2% agarose gels.

A characteristic but unique distribution pattern for each marker namely C-KIT/CD117, CD133, CD44, CD24 and ALDH1/2 (Figs. 1A, 2A and 3A, Additional file 1: Figure S1, Additional file 3: Figure S3) within OSE layer and cortex regions both were noticed (detailed in Additional file 4: Table S1).

To confirm our results from immuno-histochemical analysis and to define the sub-cellular (membrane, cytoplasmic and nuclear) localization, we performed immuno-fluorescence analysis using specific antibody for each of the marker followed by high resolution confocal microscopy. Our analysis revealed the existence of small, spherical (~ 5 μm size) cells as well as large elliptical (~ 10 μm size) cells and multiple cell clusters in both OSE layer and cortex in NO and BN, BL and HG tumors (Figs. 1B, 2B, 3B, Additional file 2: Figure S2, Figs. 4, 5, 6, 7 and 8). Distribution pattern of various markers is described below and detailed in Additional file 5: Table S2.

It is vital to understand cellular identity of proliferating cell fraction in HG tumors. Since CSCs have been reported to play critical role in metastasis, therefore, to comprehend the phenomenon of metastasis and ascites formation, we performed co-expression of KI67 with stem cell and CSC specific markers. In our study, we observed that pluripotency related stem cell marker OCT4 was predominantly expressed in both OSE layer (Fig. 4a-d) and cortex (Fig. 4e-h). In both locations, OCT4⁺ cells also expressed proliferation marker KI67 with mostly nuclear signals for both markers while few cells with cytoplasmic KI67 signals (blue asterisk) were detected. Expression of both OCT4 and KI67 was found to be variable. A few OCT4⁺ cells expressed low/undetectable levels of KI67 (white asterisk), possibly signifying different stages of differentiation of CSCs. Other surface marker typical of pluripotent stem cells that is SSEA4 (Fig. 5) was found to be expressed in both OSE layer (Fig. 5a-d) and cortex (Fig. 5e-h) of HG. Some of the SSEA4⁺ cells in cortex showed co-expression of KI67. On the other hand some cells showed only expression of KI67 (were SSEA4⁻ indicated by white asterisk). Cortex region of normal ovary revealed more of these dynamically mixed (SSEA4⁺/KI67⁺, SSEA4⁻/KI67⁺, SSEA4⁺/KI67⁻) cell populations and revealed cytoplasmic signals (blue asterisk) similar to previous report in NO [36, 39].

CD44 (CSC-surface marker) positive cells showed co-expression of KI67. These cells revealed nuclear as well as cytoplasmic and membrane bound (blue asterisk) KI67 signals in cortex. A large number of bright dual positive cells were detected in the cortex region (Fig. 6). Some of the KI67⁺ cells revealed undetectable CD44 staining (white asterisk), suggesting variable expression of CD44 marker at various stages of proliferation and differentiation of cancer cells as well as CSCs. Similarly LGR5 (a surface G-protein coupled receptor) revealed its presence in OSE layer as well as cortex region of HG tumor. LGR5⁺ cells in OSE layer exhibited membrane localization, whereas, KI67⁺ signals revealed both nuclear as well as cytoplasmic localization (blue asterisk). Cell clusters positive for expression of both markers (LGR5⁺/KI67⁺) were found in OSE layer as well as in cortex region suggesting proliferating CSCs.

ALDH1/2⁺/KI67⁺ cells were mainly localized in OSE layer. Many spindle shaped KI67⁻ cells (white asterisk) expressing ALDH1/2 were observed in the cortex (Fig. 8), probably suggesting non-proliferating CSC population within the cortex. However further in-depth study to confirm this hypothesis is required. Negative controls for mouse and rabbit antibodies were employed by omitting primary antibody for both immuno-staining methods (Additional file 6: Figure S4).

To confirm for the existence of various CSC markers in ovarian tissues at various stages of ovarian tumorigenesis, we performed RT-PCR amplification of each marker gene using specific primers. As shown in Fig. 9, amplified gene products of expected size for CSC and proliferation specific markers were detected in most samples analysed. We used four normal ovarian (NO) tissues and four sets of patient tissues belonging to BN, BL, and HG tumors. RT-PCR analysis without cDNA template served as a negative control. RT-PCR analysis presents a semi-quantitative measure of mRNA transcript copy number expressed across various tissues. RT-PCR analysis of normal and tumor tissues confirmed the expression observed by both immuno-histochemical and immuno-fluorescence methods for each marker. As expected, some variation in the expression of each marker gene was observed among patients, but tumor stage-specific differences in expression of marker genes was not obvious due to immense cell heterogeneity within tumor tissues and lack of RT-PCR analysis on enriched cell populations per se (Fig. 9).

In present study, CSC sub-populations were detected by immuno-staining for CSC specific markers and co-expression analysis of same with proliferation marker KI67 to analyse the distribution, abundance and heterogeneity of persisting cell types within OSE and cortex regions of both normal ovary and ovarian cancer tissues at various stages of tumorigenesis. We did not investigate tumor progression from a particular stage to the next and associated changes in terms of CSC profile, but employed archival and fresh patient samples surgically excised at the Cancer Center, meeting the criteria of representing various stages of tumor development (BN, BL, HG) besides normal sample (NO) for immuno-staining and RT-PCR studies. We investigated by RT- PCR the presence of transcripts for each marker using specific primers. Immuno-staining of each marker (Figs. 1, 2 and 3, Additional file 1: Figure S1, Additional file 2: Figure S2 and Additional file 3: Figure S3) with respect to differential marker expression, distribution of signals and co-localization with proliferation marker KI67, detected heterogeneous sub-populations of cells (Figs. 4, 5, 6, 7 and 8). This scenario correlates to certain extent (if not completely) and possibly justifies the variation in expression of various markers observed by RT- PCR (Fig. 9). A variable level of expression for each marker gene was observed in patients’ tumor tissues, which could be the result of presence of heterogeneous mixture of tumor cells in various stages of differentiation along with a tiny fraction of CSCs. Such variability in expression of each CSC marker, distinct phenotype, heterogeneity in stemness status etc. have been recently reviewed in ovarian CSCs [52]. On similar lines the findings by Inoue et al. [53] where they demonstrated a variable expression pattern of epithelial mesenchymal transition (EMT) marker E-cadherin within female genital tissues from benign and malignant tumors compared to normal ovaries substantiates the variability factor in present study results. Precise gene expression profiles of individual cells comprising of tumor tissue may be quite variable and hence representing differential levels of expression and hence collectively all tumor cells contribute to heterogeneous cell types and thus variable representation of gene expression.

The CSC hypothesis proposes bi-directional inter-conversion between stem and non-stem compartments in order to sustain tumorigenesis and establish multi-lineage differentiation into different tumor cell types [54]. However, exact mechanism to explain such lineage differentiation remains unknown. OCT4^med/low, CD44^hi, CD24⁺ hierarchy of breast cancer cells have been reported recently [55]. Schwede et al. [56] characterized serous ovarian tumor cells, and co-related populations of stem cell like subtype in HG serous tumors to poor patient survival than the other differentiated subtype identified with mixed histology. Expression of stem cell markers (OCT4, SOX2, NANOG, NESTIN, ABCG2, CD133 and CD117) in ovarian cancer (sphere cells) represent their tumorigenic potential, and resistance to cisplatin, paclitaxel, adriamycin and methotrexate, which highlights the significance of “stemness” of cancer cells [17]. Similar to our current observations of two distinct populations of (spherical 5 μm and elliptical 10 μm) cells, recently two cell subsets with epithelial and stromal gene signatures respectively were detected in HG serous ovarian cancer exhibiting variable gene expression [57]. More recently, Ahmed et al. [58] based on proteomic signatures, reported activation/enrichment of specific pathways related to energy, metabolism and DNA repair in ascites derived tumor cells. Collectively these studies indicate persistence of heterogeneous CSC populations following a dynamic hierarchy possibly ordained to perform specific functions or cell rescue operations upon their exposure to onco-therapy. However, these reports fail to present a clear scenario of tumor progression associated phenotypic and genotypic changes in CSCs (specifically) in a stage-specific fashion. Recently, we reported presence of small spherical (~ 5 μm) and elliptical/spindle-shaped (≥ 10 μm) cells variably expressing CSC, and germline stem cell specific markers in normal ovarian and tumor tissues [10]. However, proliferation status of these (cancer) stem cell subsets especially in HG tumors (with high metastatic potential) remains unknown to date. To address this question, in our present study, we co-localized KI67 with stem and CSC specific markers and observed co-expression of KI67 and OCT4⁺, SSEA4⁺, CD44⁺, LGR5⁺, ALDH1/2⁺, CSCs, suggesting their highly proliferative nature.

Various CSC sub-populations viz. ALDH⁺, CD133⁺, CD44⁺/CD117⁺ presenting differential response to chemotherapeutic drug were recently isolated from ovarian cancer cell line SKOV3 [59]. Stemberger-Papic et al. [60] examined the expression of CD133⁺/CD117⁺ CSC markers in 64 serous ovarian cancer cases in HG serous carcinoma and peritoneal metastasis using IHC and found them relevant in disease progression and prognosis. Consistent with these investigators, presence of chemo-resistant CD44⁺/CD117⁺ CSCs from SKOV3 epithelial ovarian cancer cell line was reported by Chen et al. [61]. Similarly, Qin et al. [62] reported over-expression of Nestin unlike CD133, implicated in cisplatin resistance and shorter overall patient survival. In other cancer types dynamic and interconverting sub-populations of intestinal CSCs with LGR5⁺/KI67⁺, LGR5⁺/KI67⁻, LGR5⁻/KI67⁺, LGR5⁻/KI67⁻ expression profiles were reported in patient derived organoid xenografts generated in NOD/SCID mice [63], suggesting existence of various CSC sub-populations in patient tumors. Co-expressing populations of CD24⁺/CD44⁺/EpCAM⁺/CD133⁺ with pro-tumorigenic gene expression profile were reported in pancreatic ductal adenocarcinoma [64, 65]. In an independent study, Paula et al. [66] reported the presence of CD24⁺/CD44⁺/CD133⁺ CSCs in pancreatic ductal adenocarcinoma with higher expression of Notch1 and Sonic hedgehog. Expression of KI67 in these CSCs indicated tumor aggressiveness. In similar context, Paula et al. [66] reported ALDH1⁺/CD44⁺ co-expressing cells in non-malignant and neoplastic breast tissues. Similarly, Kalantari et al. [67] reported CD44⁺/CD133⁺ co-expressing putative CSC markers in prostate carcinoma. Consistent with these studies, expression of various stem and CSC-specific markers such as OCT4, SSEA4, ALDH, CD117, CD133, CD44, CD24, and LGR5 in our study, suggests presence of various CSC populations with characteristics which may explain heterogeneity, aggressiveness and metastatic potential of ovarian cancer. However, it remains unknown from our studies, if CSCs present in OSE layer migrate to ovarian cortex and progress to cancer or the stem cells locally present in the cortex region (or any migrating from extra-ovarian sites) undergo transformation and result into tumor development. The possible involvement of stem cells/CSCs from organs of extra-ovarian origin requires thorough and systematic investigation which may add novel dimension in understanding the complicated disease of ovarian cancer and explore the putative cells of origin. Tracking specific (cancer) stem cell sub populations across different tumor stages in comparison with normal ovary and further in-depth studies may provide greater clarity of this phenomenon. However, the choice of stem/CSC population (s) to chase their trafficking is interesting and will be part of our future study.