The neuroepithelial origin of ovarian carcinomas explained through an epithelial-mesenchymal-ectodermal transition enhanced by cisplatin

Cisplatin drives morphological remodeling of SKOV3 ovarian carcinoma cells into neuron-like cells

Morphological remodeling of cancer cells upon cytostatic treatments is observed in various tumors of different histology. Usually, phenotype dedifferentiation in malignant cells is attributed to processes, such as EMT, which greatly contribute to clonal evolution and tumor cell heterogeneity.

Mounting evidence suggests that malignant cells influence the tumor microenvironment (TME) by releasing a plethora of neurotrophic factors, e.g. NGF, BDNF, GDNF, Neuturin and Ephrin B1 to promote axonogenesis, neurogenesis and neural reprogramming³⁷. However, little is known about the ability of some cytostatics to convert certain cancer cells into neuron-like cells which may explain the increase in neurotrophic factors in the TME.

Rat brain cell cultures grown in 2D Neuropan induction medium allow the growth and differentiation of giant cells that develop into multipolar neuron-like cells. Figure 1, Panel A illustrates the formation of dendritic trees via cytoplasm shrinkage (pictures A-D, white arrows). In a similar way, cisplatin induces a morphological conversion of ovarian carcinoma cells into neuron-like cells as observed in normal rat cells isolated from the forebrain as well as cells isolated from colon (Fig. 1, Panel B). This effect is independent of the type of medium employed, indicating that plays a key role in this process.

Morphological remodeling of cells with neuron-forming ability. Panel A: progressive cytoplasm shrinkage in rat brain cells in vitro with formation of dendritic trees, typical of multipolar neurons. The cells were cultivated in Neuropan medium after the enzymatic isolation from rat brains. A population of giant, mono- or binucleated cells (A) undergoing progressive cytoplasmic/cytoskeletal structural changes (A–H) and forming “branches” (white arrows, B–D) which ultimately give the cells a multipolar neuron-like appearance. Panel B: cellular remodeling towards neuronal morphologies of SKOV3 cells treated with sub-cytotoxic concentrations of platinum- derived drugs. The formation of neuronal structures reminiscent of multipolar neurons is clearly analogous to normal neuron development observed in brain cells isolated from rats (Panel A). Pictures are representative of several independent experiments.

Full size image

Ovarian carcinoma cells with induced neuronal morphology form cell–cell communication networks

Communication networks between neurons and other cell types in 2D-cultured normal rat enteric cells are illustrated in Fig. 2, Panel A showcasing interactions between neuronal and somatic cells. SKOV3 cells after exposure to cisplatin undergo a neuronal de-differentiation and initiate a prominent network with adjacent cells or with other neuronal like cells (Fig. 2, Panel B) similar to the cell–cell interactions observed in cultures of differentiated neurons isolated from normal colon. Neuron-like cells attach to somatic cells and establish communication through cellular protrusions resembling those found in neuromuscular junctions (Fig. 2, Panel B close-up pictures C & D and Videography A).

Communication networks and differentiation of cells in 2D and 3D cell cultures. Panel A: normal enteric neurons isolated from Wistar rats interacting with adjacent cells via neuron-somatic cell (A, B, C, E, F & H) or neuron-neuron (D & G) communication, with the contact being mediated by dendrite-like structures. Panel B: cell–cell communication among SKOV3^CP cells similar to normal tissue (neuron-like somatic cells: close-up from pictures C & D; neuron-neuron, H) via dendrite-like protoplasmic extensions. Panel C: culture of primary OC cells in 3D Geltrex gel supplemented with organoid medium. The clusters of OC cells maintained in organoid conditions revealed that in the early stages, these cells develop a neuronal appearance similar to 2D models (A–E). In this context, we observed cells with a morphology resembling colonic neuroendocrine cells of the crypts characterized by the typical basal granule accumulation (C). The neuronal nature of these cells is confirmed by the expression of βIII-tubulin (F). In 3D cultures, we observed the recapitulation of ChgA expressing cells (G, white arrows). Vimentin was widely expressed in these cultures (H), while EpCAM was differentially expressed among cell clusters (I, white arrows). The transcription factors PAX6, PAX8 and PAX9 were differentially expressed by cells within the organoids (J–L, white arrows). Panel D: Primary ovarian carcinoma 3D cell cultures with neuronal morphology and phenotype. OC236 subpopulations grow in nest-like forms with EpCAM^negative cells sorted from the EpCAM^positive fraction using immunomagnetic MicroBeads. In organoid cultures derived from EpCAM^negative fraction, cells with a neuronal phenotype, similar to those observed in 2D cell cultures (Panel C, A–C) appeared. Contrarily, in EpCAM^positive cells this phenotype is rare to find (Panel D, E–H). Of note is the tendency of EpCAM^positive cells to generate clusters similar to organoids. Pictures are representative of several experiments. Magnification is reflected in each picture.

Full size image

Generation of OC organoids initiated by cells with neuronal-like morphology

In 2D cultures of OC cell lines, a small population of cells with a neuronal-like morphology is recurrently observed, as previously demonstrated. Contrarily, during the organoid differentiation, particularly within the first two days of culture, a prominent neuronal-like population is observed (Fig. 2, Panel C, A-E & SF4). It is known that the formation of organoids in 3D models typically originates from stem cells³⁸. Consequently, those cells with an initial neuronal-like appearance from which an organoid arise, are the progenitor cells. This suggests that organoid differentiation, including the development of cell heterogeneity, is driven by these neuronal-like cells, which initially resemble neuroendocrine cells of the colonic crypts. Regarding the morphology of OC organoids, it reflects the phases of cell cluster formation observed directly in freshly collected ascites, where a cystic formation differentiates to solid cell clusters (see Videography B).

In the Fig. 2, Panel C, we observed the cell heterogeneity recapitulation in organoids generated from primary ovarian carcinoma cells, isolated from ascites. Cells, expressing high amounts of Chromogranin A (ChgA) were detected (G, white arrows). Notably, many cells within the organoids express basal levels of ChgA. The expression of Vimentin, a type III intermediate filament expressed in mesenchymal cells, was observed in all organoids scrutinized (H), indicating an EMT process in those entities. Contrarily, EpCAM was differentially expressed among organoids (I, white arrows) revealing the different fates of cells in distinct clusters. The expression of PAX transcription factors was observed in a subset of cells within the cellular clusters (J-L, white arrows).

Acquisition of neuronal morphology by ovarian carcinoma cells begins with the propagation of a distinctive cell type

Microscopic inspection of SKOV3^WT cells reveals a naturally occurring, distinctive cell type with a morphology resembling that of bipolar and multipolar neurons. These cells are clearly different in shape and color from the rest of the population and attach to adjacent cells during cytokinesis (Fig. 3, Panel A, white arrows in pictures A-D or Panel B, pictures A-C). We called these cells “twin” because they become more prominent during cell division, and analyzed their transcriptional signature for stemness and neuronal markers (see Fig. 6 and Videography B & C). Interestingly, we found similar cell types in primary cultures of ovarian carcinomas obtained from ascites (Figure SF5 & SF6).

Cells with the capacity to convert into neuron-like cells among SKOV3 cell populations and time-dependent changes in gene expression. Panel A: naturally occurring, dark-colored cells generating cells with different morphology than the whole cell population (arrows in pictures A–D). Another cell type are cells with two or multiple protrusions (arrows in pictures E–H). Panel B: changes in SKOV3 cell morphology upon platinum-derived drug exposure. Naturally occurring cells with distinctive morphology in SKOV3^WT cells (A–C). Multikaryocyte formation (D–F) was a frequent event in SKOV3 cells treated with cisplatin. SKOV3^CP cell populations cultured in 2D Neuropan neuronal induction medium form typical iPSC-like colonies which generate cells which are morphologically clearly different from the parental SKOV3 cell population (G–I). The expression of Notch1, E-cadherin and Vimentin was studied by Western blot, revealing that EMT takes place in SKOV3^CP cells. Lane 1, SKOV3^WT cells cultured in 2D DMEM; lane 2, SKOV3^CP cells in 2D DMEM; lane 3, SKOV3^CP cells in 2D Neuropan neuronal induction medium. Interestingly, Notch1 was downregulated in Neuropan-cultured cells. We named the cells having a different morphology than the rest in wildtype populations as “twin” cells (Panel A, A–D; Panel B, A–C). Panel C: Inducible expression of stem cell related markers upon cisplatin exposure. The time-dependent expression of the transcription factors PAX6, PAX9 and Slug, as well as the mesenchymal marker Vimentin were monitored by immunoblotting. Despite of the oscillatory expression of those markers upon cisplatin exposure, the trend observed in the expression kinetics clearly revealed an accumulation of those markers in treated cells. This effect was accompanied with morphological changes towards a “neuronal” shape. Values represent the expression of such proteins relative to GAPDH after densitometric analysis. AU: arbitrary units Representative experiments for n = 3.

Full size image

SKOV3^CP cell subpopulations undergo karyokinesis without cytokinesis, resulting in the formation of giant multikaryocytes (Fig. 3, Panel B pictures D-F) which, via cellularization, lead to an increase of colonies once platinum-based drugs are eliminated from the medium. These events were reported by our group and others³⁹. Incubating SKOV3^CP cells in neuronal induction medium drives the formation of colonies very similar to the ones obtained by induction of pluripotent cells (Fig. 3, Panel B pictures G-I).

Western blot analysis of the stemness features in cells treated with cisplatin and maintained in normal DMEM medium revealed that, SKOV3^CP cells upregulate Notch1, Vimentin and N-Cadherin while downregulating the epithelial protein E-Cadherin, compared to SKOV3^WT cells (Western blots from Fig. 3, Panel B), indicating an EMT process. This effect is also partially observed in cells treated with cisplatin maintained in Neuropan induction medium, though notably Notch1 expression is significantly reduced. However, detecting E-Cadherin with commercial antibodies often reveals bands at different molecular weights, depending on the cell type and its physiological state^40,41.

Stem cell related-markers are accumulated during cisplatin exposure

Transcription factors (TFs) often function synergistically to regulate gene sets essential for specific biological processes. In the OVCAR-3 ovarian carcinoma cell line, we observed a time-dependent increase in certain proteins that are key regulators or indicators of stem cell processes. OVCAR-3 cells exposed to cisplatin for 14 days revealed profound changes in the cell morphology towards “neuronal” phenotype. These changes were primarily accompanied by the time-dependent accumulation of some TFs such as PAX6, PAX9 and Slug, however, are probably supported by EMT as indicated by the accumulation of Vimentin (Fig. 3, Panel C). Nevertheless, in other ovarian carcinoma cell types, the kinetics of such markers oscillated. Notably, PAX6 accumulation was consistent across different cell lines.

Epithelial-neuronal gene expression and phenotypic switch in SKOV3 cells induced by cisplatin

To determine the molecular basis of the morphological conversion of epithelial cancer cells into neuron-like cells under the genotoxic stress of cisplatin, we first conducted a gene expression analysis focusing on markers indicative of neuronal conversion.

Figure 4, Panel A reflects the data gathered by qPCR with a TaqMan system. We inferred from the genetic expression analysis that the cells, at the population level, had lost their epithelial state, as judged by the downregulation of key epithelial markers like E-Cadherin, EpCAM, Occludin, CD24, and ALDH1L1. Concurrently, transcription factors associated with stemness and EMT like Slug, SOX2, Klf-4, Notch1 and Nanog were clearly upregulated, as was Vimentin. Of particular note is the upregulation of Tribbles-1 and 2 (TRB 1 & 2), which act as tumor suppressors. Occludin, a NADH oxidase typically localized at tight junctions in epithelial cells, plays crucial roles in tight junction assembly and barrier maintenance. However, in cancer cells, high levels of Occludin are associated with reduced invasiveness and motility. Therefore, its downregulation by cisplatin in ovarian carcinoma cells may promote migration and invasion. This downregulation aligns with the loss of epithelial traits and the acquisition of mesenchymal features (Fig. 4, A). It appears that the exposure to cisplatin enhances the transcription factor Neurogenin 2 (Ngn2), which is involved in the development of motor neurons (Fig. 4, A).

Expression of EMT-related proteins in SKOV3^CP. Panel A: ΔΔCt values for transcription factors and neuronal markers (RT-qPCR TaqMan system). Transcriptional signatures defining a neuronal phenotype are clearly demonstrated in SKOV3^CP cells. Of note, epithelial markers like Occludin, ALDH1L1, EpCAM, CD24 and E-cadherin were significantly downregulated. Green arrows denote an upregulation, red arrows a downregulation at the transcriptional level. Data represents only those genes which were differentially expressed below the ΔCt threshold of 29 cycles. A1: Cluster analysis of the interaction networks of differentially expressed genes using STRING (http://www.string-db.org) with default parameters, including a minimum interaction score of 0.4 (medium confidence) and various interaction sources including Text mining, Experiments, Databases, Co-expression, Neighborhood, Gene Fusion, and Co-occurrence. The colored “halo” around the bubbles indicates differential expression values, with red indicating higher values in SKOV3^WT and blue indicating higher values in SKOV3^CP. Panel B: 1, Transcriptome analysis of neuronal-related differentially expressed genes in SKOV3^WT vs. SKOV3^CP cells using NGS. SKOV3 cells were exposed to cisplatin for 72 h to study the first biological response to this agent. We extracted from the transcriptome results the neuronal-related genes found enhanced in SKOV3 cells treated with cisplatin for 72 h and depicted the existence or until unknown interconnection of more than 32 different genes related to neuronal processes found upregulated in cisplatin exposured cells. One relevant gene, which has not escaped our attention, was PNMA1 that codes the paraneoplastic Ma1 antigen a truly specific neuronal protein. 2, Gene expression clustering analysis using the SynGO 2022 database (https://maayanlab.cloud/Enrichr/) revealing the upregulation of several genes related to neurodegenerative diseases and neuronal physiology. Panel C: Gene expression analysis using the SynGO 2022 database from the Enrichr online platform revealed that there are significant enrichments of gene sets associated with synaptic function.

Full size image

The transcription of Snail (SNAI1) was upregulated upon cisplatin exposure. Snail, similar to Slug (SNAI2), is a transcriptional repressor of ectodermal genes within the mesoderm and also represses E-Cadherin. Its upregulation contributes to EMT, inducing a downregulation of some ectodermal genes, which were upregulated after cisplatin exposure (Fig. 4, A). In summary, the cells responded to changes in ectodermal gene expression, and the activation of the Snail pathway served as a mechanism to restore homeostasis.

Another relevant finding is the upregulation of typical neuronal markers involved in neurogenesis like PAX6, along with structural proteins like βIII-tubulin and MAP2, which are almost exclusively found in neurons. Tenascin C, a protein mostly restricted to the extracellular matrix (ECM) in neurogenic domains of the nervous system was also upregulated. Moreover, certain key players in the physiology of neuronal transmission such as synaptophysin and ACHE (acetylcholinesterase), or the neuronal receptor NMDAR2B and the nuclear receptor 4A2 (NR4A2) or Nurr1 showed increased expression (Fig. 4, A).

The protein association network derived from the transcriptional data revealed an interconnected system, where proteins such as Tenascin, SNURF, and SCG5 appeared to be not functionally related to the majority of other proteins (Fig. 4, Panel A1). It is worth to mention that this network was created considering pre-existing data. Consequently, we identified proteins involved in the global transcriptional regulation induced by cisplatin in the SKOV3 ovarian carcinoma cell line. The extensive expression of genes associated with neuronal systems, along with their physiological characteristics, strongly suggests a neuronal background for these cells. Notably, the presence of the paraneoplastic Ma antigen 1 (PNMA1), which is specific for neurons, provides clear evidence of the neuronal origin of these cells. Figure 4, Panel B1 depicts the KEGG interconnection of neuron-related genes, extracted from transcriptome analysis performed in SKOV3^WT and SKOV3^CP. It is worth mentioning is that this analysis of the whole genome was performed after 72 h of cisplatin exposure. Of note, PNMA1 is present in both SKOV3^WT and SKOV3^CP ovarian carcinoma cells and was enhanced in cisplatin-exposed cells, as corroborated using different methods (Fig. 4, Panel A; Fig. 7, Panel A, K & O). This is clearly indicative of the neural ontogeny of this cell line. Moreover, other genes related to the neuromuscular junction like the MuSK activator Agrin, Membralin and SLC25A1 are expressed in SKOV3^WT ovarian carcinoma cells and overexpressed in SKOV3 cells exposed to cisplatin. We have to keep in mind that in the transcriptional study conducted by qPCR, the transcription factor Neurogenin 2, which is implied in the development of motor neurons, is also upregulated upon cisplatin exposure. In this context, it is noteworthy that the transcriptome analysis of both wildtype and cisplatin-exposed cells revealed the expression of numerous genes associated with the muscle system and morphogenesis. (see Figure SF7).

Figure 4, Panel B2, illustrates a significant multitude of genes associated with various diseases, notably showcasing a prevalence of genes linked to degenerative or neurological conditions characterized by impaired movements, including Parkinson’s, Huntington’s, Amyotrophic lateral sclerosis (ALS), Alzheimer’s, and Bovine spongiform encephalopathy (BSE).

We observed in 2D & 3D cultures of OC cells, the tendency of cells to build communication networks similar to those formed between neurons (Fig. 2) or neurons and other cell types. In fact, some genes related to synaptic physiology e.g. Synaptophysin (Fig. 4, Panel A) were detected in cells of this cancer entity. The analysis of the transcriptome revealed the expression of genes related to neuronal synapsis (Fig. 4, Panel C). However, electrophysiological studies revealed no electrical activity in such cells and therefore the dendritic-like structures (data not shown) remain dysfunctional. This could be based on the imbalance of up- and down-regulated genes, which form a proper functional synapsis.

In all of the transcriptome data resulting from NGS, we decided to search for the fate of the HERV expression, given the fact that in resistant cancer cells the expression of envelope proteins mediate cell–cell fusion, a phenomenon often observed in ovarian carcinoma cells (see Videography E). This process is thought to be involved in metastization events. Ovarian carcinoma cells express a plethora of endogenous retrovirus elements. Using qPCR, we validated the differential expression of HERV elements in SKOV3 wildtype and cisplatin resistant cells (Figure SF8, I, II & III). It is worth mentioning that we globally observed a downregulation of several HERV elements in SKOV3^CP compared to its wildtype. However, we detected some upregulated HERV genes, which code for envelope proteins like, HERV-W_E1 (Syncytin 1), HERV-FRD₁ (Syncytin 2) and HERV-W_env, which have retained their fusogenic properties. These proteins have been widely reported to be highly fusogenic in addition to their elucidated involvement in immune escape, which is mediated by the immunosuppressive domain (ISD). HERV3.1 was also upregulated in cisplatin resistant cells. Although this HERV has lost its fusogenic properties, the immunosuppressive activity has been retained because the ISD is highly similar to the CKS-17 immunosuppressive peptide.

The neuronal morphology of SKOV3 cells exposed to cisplatin correlates with the expression of neuronal markers

Gene expression analysis using qPCR and NGS of SKOV3^WT and SKOV3 cells exposed to cisplatin for 72 h revealed a clear switch from an epithelial to a mesenchymal and from this to an ectodermal state. Nevertheless, these findings suggest that the conversion of mesoderm-derived cell type like ovarian carcinoma cells into a neuronal type had occurred at the transcriptional level.

To determine if effector proteins are synthesized by cells challenged with cisplatin, we investigated the transcript-protein correlation and observed that the protein expression of relevant neuronal markers was, in fact, congruent with the gene expression data (Fig. 5, Panels A, B and C).

Figure 5, Panel A reflects the expression of key transcription factors crucial for neuronal development, e.g. SOX2 and HES1 TFs, both involved in the generation and maintenance of progenitor and stem cells. Notably, HES1 is vital for nervous system development, influencing neural stem cells and neuroepithelial cells by repressing basic helix-loop-helix (bHLH) activators^42,43. The transcription factor PAX6, which regulates neurogenetic and neural stem cell self-renewal processes, was overexpressed⁴⁴. The bHLH transcription factor NeuroD, which is a critical controller of the neuronal vs. glial fate decision and responsible for dendrite formation and maintenance, was overexpressed in SKOV3^CP cells^45,46. The occurrence of EMT in the SKOV3^CP cells can also be inferred by the loss of the epithelial marker EpCAM and the gain of Vimentin, a type III intermediate filament (IF) protein exclusively expressed in mesenchymal cells.

Ovarian carcinoma cells exposed to cisplatin also overexpressed certain structural proteins which are characteristic of neuronal tissues. Figure 5, Panel B reflects the increase in several cytoskeletal structural proteins like βIII-tubulin and MAP2 in SKOV3^CP cells. Both are almost exclusively expressed by neurons and play important physiological roles in neuron development^47,48. In addition, the low molecular weight neurofilament protein (NF-L) is clearly upregulated in SKOV3^CP cells. This neurofilament is mainly localized in the axonal regions of neurons where it helps to stabilize the axons and facilitates axonal transport^49,50. N-Cadherin, which plays a decisive role during neural development⁵¹, was also overexpressed in ovarian cancer cells treated with cisplatin. However, the expression of N-Cadherin mediates the migration of mesenchymal cells from the bone marrow towards tumors, which is associated with poor prognosis^52,53. Stathmin, also known as oncoprotein 18, plays a crucial role in cytoskeleton regulation specifically by modifying microtubule dynamics. This protein was overexpressed in SKOV3^CP cells. The expression of Stathmin is upregulated during neuronal development^54,55 and has been associated with carcinogenesis and metastatic processes^56,57. Furthermore, Tenascin C, which is expressed in the extracellular matrix in neurogenic domains of the central nervous system and also temporarily during organogenesis, was overexpressed in SKOV3^CP cells^58,59. Tenascin C expression is clinically associated with the metastatic process^60,61. Some neuronal transmitters use Synaptophysin as a mediator. This synapse-associated protein was overexpressed in SKOV3^CP cells. Importantly, Synaptophysin is expressed in neuroendocrine cells or neurons^62,63.

Using flow cytometry (Fig. 5, Panel C, and Figure SF9), we investigated the expression or loss of certain EMT markers in SKOV3^CP cells vs SKOV3^WT. The expression of proteins indicative of EMT was correlated to a significant reduction of E-Cadherin and increase of Vimentin (to approx. 75% and 96%, respectively, Fig. 4, Panel C). Likewise, the EMT-initiating factors Snail and Slug were upregulated in about 29% and 23% respectively. A notable finding was the upregulation of Doublecortin, a key protein expressed by neuronal precursor cells. Moreover, together with the upregulation of Goosecoid, a protein involved in driving EMT, chemotherapy refraction and metastasis formation was observed⁶⁴.

Cells isolated from OC ascites revealed potential origins of the tumor

It is necessary to consider that the majority of commercial OC cell lines were derived from cisplatin refractory patients and therefore not truly virgin to this therapy. Even though cell lines provide valuable research models, they do not recapitulate the tumor biology as primary cell cultures do. For this reason, we sought to investigate the existence of cells with a neuronal signature in fresh cells isolated from ascites from ovarian carcinoma patients.

Figure 5, Panel C (merged channels) and Figure SF10 (RBG decomposition) shows the growth features and the expression of different markers in non-treated OC cells isolated from ascites. The first characteristic observed in this primary cell culture is the existence of islands of nested cells, which depict a distinct morphology from the surrounding cells.

Nested cells express a vast number of pivotal transcription factors typically observed in embryonic development. Notch 3 is mainly expressed in the nested cells (Fig. 5, Panel D, A). PAX6 and PAX8 were expressed in both nested and their adjacent cells (B & C). PAX9 was highly expressed by nested cells and to a lesser extent in surrounding cells. Only few cells express the TFs Klf-4 and SOX2 (E & F), mainly detectable in the nested cells. Contrarily, Slug and Oct-4 are strongly expressed in nested cells (G & H) with Oct-4 being restricted to these cells, while Slug is also detectable in adjacent cells.

It is evident that nested cells show an EMT signature, indicated by the exclusive localization of Vimentin in these cells (I). Their neuronal nature is further revealed by the expression of βIII-tubulin (J), Doublecortin (K), Stathmin (also found in the mixed populations) (L), the neuroendocrine marker ChgA (M) and PNMA1 (N). Intriguingly, nested cells also overexpress keratin 80 (KRT80) (O), an epithelial related keratin. However, KRT80, despite being an epithelial keratin, is clearly expressed in the nested cells, indicating that this keratin may also serve as a marker for cells in an EMT transit. However, this was previously observed in the context of colorectal carcinomas and associated with high metastatic potential⁶⁵. Consequently, it raises the question of whether this keratin is also shared by cells with a neuronal signature.

In summary, these findings suggest two key points i: the nested cells in OCs exhibit a strong neuronal signature, indicating that the tumors surely arose from this cell type and ii: these cells are in fact the cancer stem cells based on the expression of plethora of TFs which are cardinal for both stem cell maintenance and cell differentiation.

The nuclei of cells undergoing an EMT are impermeable to pivotal transcription factors

In OC236 cell model, PAX8 TF is observed without restriction in both nucleus and cytoplasm, contrary to PAX6, PAX9 and Notch3 (Fig. 5, Panel C, merged channels and Figure SF10, RBG decomposition). This cytoplasmic restriction was also observed for PAX6 in cisplatin resistant SKOV3^CP cells (Fig. 5, Panel A). Consequently, this TF is the only one that can properly regulate the transcription of its downstream target genes. PAX8 is involved in the embryonic development of the thyroid gland, specifically its rudimentary lateral portion, which develops from neural crest cells.

Concerning the subcellular localization of PAX TFs in particular, it is true that in “neuronal” or stem cells, such TFs are ectopically found in the cytoplasm instead in the nucleoplasm as expected. We verified that both gene expression and protein synthesis of these TFs occurred and that at the protein level they remain functional. In this regard, the more epithelial the cells appear, the more likely it is to find PAX TFs in the nucleoplasm. Conversely, in cells depicting a “neuronal” or adult stem cell phenotype, PAX TFs are less likely to be located in the nucleoplasm (Fig. 5, Panel C, merged channels and Figure SF10, RBG decomposition). This clearly indicates that the nuclear membranes of both populations are functionally different, with those nuclei from cells that are phenotypically more stem or “neuronal” being less permissive to these TFs, fact that was unknown until now. This is an important observation that merits further biochemical studies in order to elucidate the underlying mechanism. If these cells do not utilize these TFs themselves, it is possible that they export them to other target cells.

In summary, we have established that several proteins, which are biologically characteristic of neuronal development, were overexpressed in ovarian carcinoma cells treated with platinum-derived drugs. Importantly, the cell shape was consistent with a neuron-like morphology, indicating that the upregulation of the neuronal proteins was partly responsible for the observed changes in cell shape. Moreover, as already shown in Fig. 4, the gene expression of neuronal markers strongly correlated with the protein synthesis. We have thus demonstrated a relationship between protein expression and the acquisition of a neuronal-like morphology in ovarian carcinoma cells treated with platinum compounds.

Are “twin” cells the source of the conversion of epithelial to neuronal-like cells upon platinum- derived drugs exposure?

We hypothesized that a naturally recurrent cell type, described above and called “twin”, present among SKOV3^WT, OVCAR-3 cells and primary ovarian carcinomas obtained from ascites (Figure SF5 & SF6), might be the source of the expansion of cells with neuron-forming capacity (Fig. 3, Panel A). These cells maintain the dual formation for long periods; especially in cisplatin-exposed cultures (see Videography C & D).

To explore this hypothesis, we isolated “twin” cells using a micromanipulator and analyzed their stemness and neuronal gene expression at the single-cell level (Fig. 6, Panel A). We collected about 20 cells during cell division, extracted their RNA, and performed qPCR transcriptional analysis. A limit of 29 cycles was established, with ΔCt values above this limit excluded from consideration. However, we reported genes that showed significant expression in at least one cell type, either SKOV3^WT or “twin” cells. This was the case with NeuroN and SOX2, which were not detectable in wildtype cells but were detected in “twin” cells. In line with this, both CD24 and EpCAM were not detected in “twin” cells, but were highly expressed in SKOV3^WT cells (Fig. 6, Panel B). The (CD44⁺/CD24^–) SKOV3-derived “twin” cells lack expression of CD24, often associated with cancer stemness. It is worth mentioning that CD24, though specific to B-lymphocytes, remains ambiguous as a definitive stemness marker due to tumor heterogeneity and its inconsistent expression across tumor types⁶⁶.

Isolation and analysis of “twin” cells present in SKOV3 cultures. Panel A: “twin” cells as shown in Fig. 3 (Panel A, A–D and Panel B, A–C) were subjected to mechanical isolation using a one-arm micromanipulator. At least 20 “twin” cells and a proportional amount of surrounding normal cells were collected for single-cell RNA isolation. Panel B: qPCR transcriptional analysis of “twin” cells vs. normal epithelia-shaped cells. A dual neuronal-stem signature is detected at the transcriptional level. Some markers were not detected (red triangles): NeuroN, SOX2 in normal cells and CD24 and EpCAM in “twin” cells. Panel C: I: Cartoon representing OC cells acquiring a neuron-like morphology under the influence of platinum-derived compounds as seen in Fig. 1. A large cell (A) gradually condenses its cytoplasm and starts forming dendritic tree-like structures (B & C). The cytoplasm is finally dismantled until fully formed dendrite-like structures are left (C & D). These morphological changes correlate with the expression of a plethora of neuronal-defining markers. II: Diagrammatic representation of the modulation of pathways involved in EMT and from this point to a “neuro-mesenchymal” inter-condition or EMET based on the transcriptional and protein expression profiles. SKOV3 cells have a basal expression of several neuronal markers, which is enhanced by exposure to sub-toxic concentrations of platinum-derived drugs. The expression of cardinal stemness markers like Nanog, Notch1, SOX2, and Klf-4 suggests the acquisition of cancer stem cells status. The cells that have moved to a neuroepithelial condition are in a transitory status of dormancy. Results are representative of n = 3 experiments.

Full size image

Single-cell RNA analysis further confirmed that “twin” cells possess stemness characteristics alongside the expression of neuronal genes (Fig. 6, Panel B). In addition, essential transcription factors involved in self-renewal, pluripotency and/or neurogenesis were significantly upregulated, including Klf-4, Oct-4, SOX2, Nanog and Notch1 being the most prominent ones. Moreover, the transition from an epithelial to a neuronal-like steady state was confirmed by the loss of EpCAM (epithelial), the gain of βIII-tubulin (neuronal) and Vimentin (see Figures SF10-13). The expression of such genes was confirmed in ovarian carcinoma primary cultures (Figure SF14).

“Twin” cells likely represent a cell type with dual characteristics: both stem and neuronal. Nevertheless, while the entire population of SKOV3 and other ovarian carcinoma cell lines exhibit a neuronal background at the protein and transcriptional levels (summarized in Fig. 6, Panel C), exposure to platinum compounds unmask this signature. However, this “neuronal” background and the more defined neuron-like signature caused by cisplatin exposure probably do not render in functional neurons at this stage, since preliminary experiments for the excitatory physiology of mature neurons were unsuccessful (data not shown). Thus, the shift to a neuronal condition is incomplete.

Investigating the neuroepithelial origin of ovarian cancers in ovarian tissues.

The precise anatomical origin of serous ovarian carcinoma remains a topic of considerable debate. While some evidence suggests that OCs may arise from the terminal portion of the fallopian tube or the ovarian surface epithelium, a definitive explanation of the specific cell type undergoing malignant transformation has been elusive^67,68.

Given the extensive data showing that ovarian cells can undergo morphological and gene expression changes towards an ectodermal state under cisplatin exposure, we aimed to identify cells with distinct neuronal-related features in ovarian tissues. We suspected that the darkness of some cells among 3D cell cultures (hanging drops, data not shown) could be a result of melanin pigments, therefore we employed the Fontana-Masson silver staining commonly used in histopathology to detect argentaffin pigments, including melanin. This staining revealed that nascent “twin” cells were particularly positive for argentaffin (Fig. 7, Panel A, pictures A-H). However, since this method does not distinguish between types of melanin, we looked for the expression of Tyrosine hydroxylase (TH), a key enzyme in the production of neuromelanin (NM) and catecholamines in the central nervous system (CNS), peripheral sympathetic neurons and the adrenal medulla Our findings indicated that TH expression was limited to a few cells in SKOV3^WT cultures but was significantly increased in cells exposed to cisplatin (Fig. 7, Panel A, pictures J vs. N). Likewise, the PNMA1 neuronal marker was also overexpressed in cisplatin treated cells (Fig. 7, Panel A, pictures L vs. P). In addition, in the primary OC cells OC236, TH was primarily detected mainly in nested cells (Q) which also exhibited stronger Fontana-Masson staining (R).

We then scrutinized ovarian tissues, both regular (non-tumoral) and tumoral for argentaffin cells or regions using the above mentioned histochemical method. We detected argentaffin cells in the ovarian surface epithelium of regular tissues (Fig. 7, Panel B, upper block of histochemical pictures, OR -ovarian regular-, A-D). These cells appeared morphologically larger compared to the surrounding cells and presented stronger pigments. In contrast, in tumoral tissues, argentaffin cells were not as concurrent and the few ones that were found appeared smaller and more dispersed (Fig. 7, Panel B, upper block of histochemical pictures, OC –ovarian carcinoma-, E–H). Notably, the distribution of TH expression mirrored that of Fontana-Masson staining in normal tissues (Fig. 7, Panel B, inferior block of IHC pictures, OR, I-L). However, TH was also specifically expressed in tumor areas, forming distinct islands (Fig. 7, Panel B, inferior block of IHC pictures, OC, M-P) matching the expression of the neuroendocrine marker ChgA. This discrepancy between the staining patterns led us to hypothesize that melanin pigments might be subject to tissue clearance via degradation, despite the noticeable expression of TH.

These observations reinforced the neuroepithelial nature of OCs, but the precise origin of “neuronal-like” cells contributing to tumor proliferation remained unclear. The experiments with Fontana-Masson staining and TH expression alone were insufficient to conclusively classify these cells as neuronal. Further investigation revealed that some rare cells in normal ovarian tissues expressed KRT80, a keratin previously associated with epithelial tissues, alongside ChgA, suggesting a potential neuroendocrine function (Fig. 7, Panel C, upper block of IHC pictures, OR A-D). It is worthy to mention, that ChgA is an activatable protein, functioning as an on/off switch in response to specific stimuli, rather than constitutively produced. We found a significant amplification of KRT80-ChgA dual-positive cells, indicating that these rare cells in non-tumoral ovarian tissues could be the progenitors that expand to form the tumor mass (Fig. 7, Panel C, upper block of IHC pictures, OR E–H). Moreover, we found that PNMA1-ChgA double positive cells were significantly amplified in the ovarian carcinoma tissues (Fig. 7, Panel D, I-P). However, in regular ovarian tissues, these cells are rarely found (Fig. 7, Panel D, A-H).

Notch inhibitors are insufficient to prevent “axonal or dendritic arborization” in SKOV3^CP cells

Notch receptors are key regulators of signaling pathways that influence various biological processes, including embryogenesis, maintenance of neuronal progenitor cells and neurite development. Notch signaling also plays a role in the cellular plasticity necessary for the upkeep of the nervous system and ongoing neurogenesis in adult tissues. To investigate the possible interference of Notch signaling with the observed “axonal or dendritic arborization” in induced neuronal-like OC cells, we treated SKOV3^CP cells with two GSIs, both indirect inhibitors of this cascade [RG-4733 (1 µM) or DAPT (5 µM)] together with 2 µg/mL of cisplatin for 5 days. We previously corroborated that these compounds inhibit Notch pathway, since HES1 was downregulated under the influence of both small molecules (Figure SF15A). Despite the inhibition of Notch signaling, there were no significant changes in the morphology of the treated cells; they retained the same “neuronal” morphology observed in untreated cells (Figure SF15B). This indicates that Notch signaling does not play a role in driving the dendritic “arborization” process in this cell model. These findings are consistent with previous observations that Notch does not translocate to the nucleus in the nested cells of the OC236 primary cells.

Curiously, in Neuropan medium (Fig. 3, Panel B), which is commonly used to develop neuronal tissues, the Notch1 appears downregulated and the cells gained the “neuronal” morphology. Consequently, it appears that this “arborization” in OC is independent from Notch signaling.

Stemness potential of nested cell population in OC236

Nested cells represent a unique cellular subpopulation capable of binary differentiation: on one hand, they maintain the condition of cancer stem cell in a small fraction (probably via multinucleation) as seen by the expression of stem cell markers like Oct-4, SOX2 and Klf-4 and, on the other hand, they differentiate towards an epithelial condition as seen by gradual loss of mesenchymal characteristics in parallel with the acquisition of the epithelial phenotype, thus shaping the cellular heterogeneity present in tumors (Fig. 5, Panel D, SF10). It should be noted that over time, the population of nested cells gradually decreases, with these cells becoming dispersed among the predominant epithelial cells. Though this could be the results of different grow kinetics.

Interfering Notch and MAP4K4 signaling pathways render in more neuronal background in OC cells

To confirm the activity of both RG-4733 and DAPT in wildtype SKOV3 and OVCAR-3 cells, we monitored the expression of HES1, whose expression is governed by the Notch pathway. Indeed, both inhibitors repressed HES1 expression (see SF15A).

In SKOV3^CP cells, the simultaneous incubation of cisplatin and GSI do not revert their neuronal morphology (Fig. 8, Panel A, A-D and SF15B). The expression of Notch3 remains unaltered regardless the type of treatment (E). Both small molecules RG-4733 and DAPT effectively stimulate the protein levels of both PAX6 and ChgA (F & G). Consequently, both GSI enhance the neuronal background in the resistant subline SKOV3^CP.

Interfering Notch and MAP4K4 signaling pathways in OC cells. Panel A: combination treatment of cisplatin and GSIs. SKOV3^CP cells were treated under the following conditions: (A) medium control, (B) 2 µg/ml cisplatin, (C) 2 µg/ml cisplatin + 5 µM DAPT, and (D) 2 µg/ml cisplatin + 1 µM RG-4733 for 120 h. The simultaneous administration of cisplatin and GSIs did not reverse the neuronal morphology of SKOV3^CP cells (A–D). Additionally, this combination did not lead to significant changes in Notch3 expression (E). However, the incubation with both GSIs induced an upregulation of PAX6 and ChgA, with RG-4733 exerting the most pronounced effect. Panel B: combination treatment of MAP4K4i and GSIs in OC Cells. The exposure of OC cells to MAP4K4i did not induce general changes as expected in HES1 protein levels across various OC cell lines. However, both GSIs significantly downregulated HES1 protein expression (A–D). Notch3 expression was stimulated by all inhibitors in both SKOV3^WT and SKOV3^CP cells (E). In OVCAR-3 and primary OC cells OC236, only the combination of MAP4K4i with RG-4733 synergistically enhanced Notch3 expression (G–H). Furthermore, βIII-tubulin expression was strongly stimulated by the MAP4K4i in SKOV3^WT cells, an effect not observed in the resistant SKOV3^CP subline. Nevertheless, the combination of MAP4K4i with GSIs led to an enhanced expression of this neuronal-related structural protein (I–J). ChgA expression was slightly repressed in SKOV3^WT cells but enhanced in SKOV3^CP cells after 24 h of MAP4K4 inhibitor treatment. Both GSIs increased ChgA expression, and this effect was slightly potentiated in the presence of the MAP4K4 inhibitor (K–L). Abbreviations: M: MAP4K4i; D: DAPT; RG: RG-4733; AU: arbitrary units. Bar graphs represent densitometric analyses of western blots performed to measure protein expression following drug exposure. Results are representative of n = 3 independent experiments.

Full size image

We next sought to study the effect of MAP4K4i and GSI on OC cells in the absence of cisplatin. The MAP4K4i revealed ambiguous expression of HES1 in different OC cells. Nevertheless, both GSI alone or in combination with the MAP4K4i effectively repressed the expression of HES1 (Fig. 8, Panel B, A-D). In SKOV3 cells, wildtype and cisplatin resistant all drugs alone or in combination stimulated the expression of Notch3 at protein level in absence of cisplatin (Fig. 8, Panel B, E–F). Contrarily, a synergistic or additive effect of the MAP4K4i and both RG-4733 and DAPT, GSIs on Notch3 expression was detected in OVCAR-3 and the primary ovarian carcinoma cells OC236 (Fig. 8, Panel B, G-H).

Interestingly, the inhibition of MAP4K4 and Notch pathways induces the upregulation of both βIII-tubulin and ChgA, two classical neuronal markers. This was observed in SKOV3^WT and more accentuated in the cisplatin resistant SKOV3^CP subline (Fig. 8, Panel B, I-L).

• SEO Powered Content & PR Distribution. Get Amplified Today.
• PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
• PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
• PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
• PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
• Source: https://www.nature.com/articles/s41598-024-76984-9