Cancer stem cells (CSCs) alone can initiate and maintain tumors, but the function of non-cancer stem cells (non-CSCs) that form the tumor bulk remains poorly understood. Proteomic analysis showed a higher abundance of the extracellular matrix small leucine-rich proteoglycan fibromodulin (FMOD) in the conditioned medium of differentiated glioma cells (DGCs), the equivalent of glioma non-CSCs, compared to that of glioma stem-like cells (GSCs). DGCs silenced for FMOD fail to cooperate with co-implanted GSCs to promote tumor growth. FMOD downregulation neither affects GSC growth and differentiation nor DGC growth and reprogramming in vitro. DGC-secreted FMOD promotes angiogenesis by activating integrin-dependent Notch signaling in endothelial cells. Furthermore, conditional silencing of FMOD in newly generated DGCs in vivo inhibits the growth of GSC-initiated tumors due to poorly developed vasculature and increases mouse survival. Collectively, these findings demonstrate that DGC-secreted FMOD promotes glioma tumor angiogenesis and growth through paracrine signaling in endothelial cells and identifies a DGC-produced protein as a potential therapeutic target in glioma.
The authors shed light on the role that differentiated glioma cells exerts in promoting cancer progression, revealing that the secreted fibromodulin by differentiated glioma cells is crucial in mediating angiogenesis in glioma via integrin-dependent Notch signaling. The results are important for gaining insight into the less concerned differentiated glioma cells in promoting cancer and would potentially enrich the treatment strategy for glioma.https://doi.org/10.7554/eLife.78972.sa0
Tumors and their microenvironment form an ecosystem with many cell types that support tumor growth. The key constituents of this ecosystem include cancer stem cells (CSCs), non-cancer stem cells (non-CSCs) representing the differentiated cancer cells, and various other cell types present in the tumor stroma (Prager et al., 2019). It is well established that the tumor-initiating capacity lies solely with CSCs, thereby making them the crucial architects of tumor–stroma interactions that favor tumor growth and progression (Rheinbay et al., 2013). CSCs have a dichotomous division pattern as they are capable of self-renewal and give rise to differentiated cells that form the bulk of the tumor (Olmeda and Ben Amar, 2019). The indispensable role of CSCs, which usually constitute only a minority population within tumors, is well documented in many solid tumors (Galli et al., 2004; Ignatova et al., 2002; Singh et al., 2004; Yang et al., 2020a).
The tumor microenvironment is a vital driver of plasticity and heterogeneity in cancer (Carnero and Lleonart, 2016; Heddleston et al., 2010). The presence of hypoxic and necrotic regions is the hallmark of very aggressive tumors like glioblastoma (GBM), which have a highly vascular niche that supplies nutrients to cancer cells and makes a conducive environment for the tumor cells to thrive (Hambardzumyan and Bergers, 2015; Huang et al., 2016). Paracrine signaling mediated by proteins secreted from tumor cells, particularly glioma stem-like cells (GSCs), helps acquire this highly vascular phenotype by attracting blood vessels and inducing pro-angiogenic signaling in endothelial cells through extracellular matrix (ECM) remodeling (Dittmer and Leyh, 2014; Rupp et al., 2016). A reciprocal relationship exists between GSCs and endothelial cells by which endothelial cells induce stemness phenotype in cancer cells through activation of Notch, sonic hedgehog (SHH), and nitric oxide synthase signaling pathways (Jeon et al., 2014; Yan et al., 2014; Zhu et al., 2011), while GSCs drive vascularization of the tumor via endogenous endothelial cell stimulation, vascular mimicry, and GBM-endothelial cell transdifferentiation (Hardee and Zagzag, 2012; Soda et al., 2011). Recent reports have shown that CSCs induce high vascularization of tumors like GBM by migrating along blood vessel scaffolds to invade novel vascular niches, thereby ensuring surplus and continuous blood supply at their disposal (Prager et al., 2020). In GBM, CD133+ and Nestin+ cells (representing GSCs) are located in close proximity to the tumor microvascular density (MVD), whereas a lower number of CD133- and Nestin- cells (representing differentiated glioma cells [DGCs]) are located in the vicinity of the blood vessels. It has also been reported that the depletion of brain tumor blood vessels causes a decrease in the number of tumor-initiating GSCs (Calabrese et al., 2007).
While CD133 marker expression was reported to be associated with GSCs initially (Singh et al., 2004; Galli et al., 2004), later reports documented CD133- cells exhibiting GSC-like properties (Beier et al., 2007; Chen et al., 2010; Joo et al., 2008; Ogden et al., 2008; Wang et al., 2008). CXCR4-dependent SHH-GLI-NANOG signaling promotes stemness in GSCs. This study also showed that the miR302-367 cluster could suppress stemness and promote differentiation by targeting CXCR4/SDF1 (Fareh et al., 2012). The above group subsequently showed that miR18A* promotes GSC stemness by activating Notch-dependent SHH-GL1-NANOG signaling, targeting DLL3, an autocrine inhibitor of Notch 1 signaling (Turchi et al., 2013). In contrast to these observations, Dirkse et al. showed the existence of stem cell-associated heterogeneity in GBM, which results in tumor plasticity and is orchestrated by the microenvironment (Dirkse et al., 2019).
Besides CSC self-renewal, their differentiation to form the bulk cancer cells also plays a crucial role in tumor growth and maintenance (Jin et al., 2017). Epigenome unique to CSCs compared to differentiated cancer cells has been documented (Suvà et al., 2014; Zhou et al., 2018). Reciprocally, a set of four reprogramming transcription factors, POU3F2, SOX2, SALL2, and OLIG2, is identified in GBM that is sufficient to reprogram DGCs and create the epigenetic landscape of native GSCs, thus creating ‘induced’ CSCs (Suvà et al., 2014). The epigenetic regulation forms the basis of cellular plasticity, which creates a dynamic equilibrium between CSCs and differentiated cancer cells (Safa et al., 2015). Oncogene-induced dedifferentiation of mature cells in the brain was also reported using a mouse model of glioma, and the reprogrammed CSCs were proposed to contribute to the heterogeneous cell state populations observed in malignant gliomas (Friedmann-Morvinski et al., 2012; Friedmann-Morvinski and Verma, 2014). Lineage-tracing analyses revealed the reprogramming of DGCs to GSCs that act as a reservoir for initiating relapse of the tumors upon temozolomide chemotherapy (Auffinger et al., 2014; Chen et al., 2012). Hypoxia has also been reported to reprogram differentiated cells to form CSCs in glioma, hepatoma, and lung cancer (Prasad et al., 2017; Wang et al., 2017). Spontaneous conversion of differentiated cancer cells to CSCs has also been reported in breast cancer (Klevebring et al., 2014; Zhou et al., 2019).
Collectively, these studies highlight the crucial role of CSCs in cellular crosstalk in the tumor niche and establish CSCs as critical drivers of tumorigenesis. However, the massive imbalance in the proportions of CSCs and non-CSCs or differentiated cancer cells in tumors raises several important questions. Considering that differentiated cancer cells constitute the bulk of tumors, do they have specific functions, or do they only constitute the tumor mass? Do they contribute to the complex paracrine signaling occurring within the tumor microenvironment? Do they support tumor growth by promoting CSC growth and maintenance? It has been recently shown in GBM that DGCs cooperate with GSCs through a paracrine feedback loop involving neurotrophin signaling to promote tumor growth (Wang et al., 2018). While this study suggests a supporting role for differentiated cancer cells in tumor growth, the large proportion of them in tumors suggests a role in paracrine interactions with other stromal cells in the tumor niche.
We used quantitative proteomics to identify DGC-secreted proteins that might support their paracrine interactions within the tumor microenvironment. We show an essential role of fibromodulin (FMOD) secreted by DGCs in promoting tumor angiogenesis via a crosstalk with endothelial cells. FMOD promotes integrin-dependent Notch signaling in endothelial cells to enhance their migratory and blood vessel-forming capacity. These findings indicate that DGCs are crucial for supporting tumor growth in the complex tumor microenvironment by promoting multifaceted interactions between tumor cells and the stroma.
While GSCs alone can initiate a tumor, the overall tumor growth requires functional interactions between GSCs and DGCs (Singh et al., 2004; Wang et al., 2018). To further understand the respective roles of GSCs vs. DGCs in tumor growth, we compared the conditioned medium (CM) derived from three patient-derived human GSC cell lines (MGG4, MGG6, and MGG8) (Wakimoto et al., 2009) and their corresponding DGCs, using a quantitative proteomic strategy. Proteins in CMs were systematically analyzed by nano-flow liquid chromatography coupled to Fourier transform tandem mass spectrometry (nano-LC-FT-MS/MS), and their relative abundance in DGC vs. GSC CM was determined by label-free quantification. We found that 119 proteins are more abundant in GSC CM, while 185 proteins are more abundant in the DGC CM (p<0.05, Figure 1A; Supplementary file 1). Analysis of overrepresented functional categories among proteins exhibiting differential abundances in GSC vs. DGC CMs using Perseus with a p-value <0.05 revealed that the DGC CM is enriched in proteins known to exhibit extracellular or cell surface localization, such as proteins annotated as ECM organization while terms related to DNA replication and many signaling pathways are enriched in GSC CM (Figure 1; Figure 1—figure supplement 1).
Figure 1—source data 1
Figure 1—source data 2
Figure 1—source data 3
Figure 1—source data 4
Figure 1—source data 5
Figure 1—source data 6
Figure 1—source data 7
Figure 1—source data 8
Figure 1—source data 9
Figure 1—source data 10
Figure 1—source data 11
The enrichment of the ECM annotation among proteins exhibiting higher abundance in DGC secretome prompted us to focus on ECM proteoglycans in line with their critical role in facilitating cancer cell signaling through their interaction with growth factor receptors, extracellular ligands and matrix components, and in promoting tumor–microenvironment interactions (Winkler et al., 2020). Six ECM proteoglycans were found to be more abundant in DGC CM compared with GSC CM (Figure 1B). The role of five of them (LAMB2, SERPINEE1, ITGB1, TNC, and LAMA5) in tumor growth has been well established (Angel et al., 2020; Bartolini et al., 2016; Long et al., 2016; Wang et al., 2021; Yang et al., 2020b). We thus focused on FMOD, which exhibited the highest DGC CM/GSC CM protein ratio. FMOD is a small leucine-rich repeat proteoglycan upregulated in GBM due to the loss of promoter methylation orchestrated by TGF-β1-dependent epigenetic regulation (Mondal et al., 2017). FMOD promotes glioma cell migration through actin cytoskeleton remodeling mediated by an integrin-FAK-Src-Rho-ROCK signaling pathway but does not affect colony-forming ability, growth on soft agar, chemosensitivity, and glioma cell proliferation (Mondal et al., 2017). We first confirmed the higher abundance of FMOD seen in DGC CM compared to GSC CM (Figure 1C) both at the transcript level (Figure 1D) and at the protein level (Figure 1E and F) in three GSC cell lines (MGG4, MGG6, and MGG8).
In line with our previous findings indicating that TGF-β signaling controls FMOD expression in glioma (Mondal et al., 2017), we next explored the possible role of this pathway in FMOD overexpression in DGCs. Gene set enrichment analysis (GSEA) of differentially regulated transcripts in GSC vs. DGC showed significant depletion of several TGF-β signaling pathway genes (Figure 1—figure supplement 2; Supplementary file 2), suggesting an enhanced TGF-β signaling in DGCs. Likewise, GSEA revealed an enrichment of several TGF-β signaling pathway genes in most GBM transcriptome datasets (Figure 1—source data 8), further supporting the activation of TGF-β signaling in DGCs that represent the bulk of GBMs. In addition, the DGCs used in this study that express a high level of FMOD showed enrichment in mesenchymal signature compared to GSCs (Figure 1—source data 10), consistent with the elevated TGF TGF-β signaling and FMOD levels we observed in the mesenchymal GBM subtype (Figure 1—figure supplement 3A B). Moreover, treating MGG8-DGCs with the TGF-β inhibitor (SB431542) significantly decreased luciferase activity of SBE–Luc (a TGF-β-responsive reporter and contains Smad-binding elements) and FMOD Promoter-Luc reporters (Figure 1—figure supplement 3C D). We also found higher levels of FMOD and TGM2 (a bonafide TGF-β target gene) transcripts and FMOD and pSMAD2 (an indicator of activated TGF-β signaling) proteins in MGG8-DGCs than MGG8-GSCs (Figure 1G–I). The addition of a TGF-β inhibitor (SB431542) significantly decreased transcript levels of FMOD and TGM2 and protein levels of FMOD and pSMAD2 in MGG8-DGCs (Figure 1G–I). Further, chromatin immunoprecipitation experiments revealed pSMAD2 occupancy on FMOD promoter in MGG8 DGCs that was significantly reduced by pretreating cells with SB431542 (Figure 1J). These results demonstrate a predominant expression and secretion of FMOD by DGCs that are promoted by TGF-β signaling.
Toward exploring the possible role of DGC-secreted FMOD in glioma tumor growth, we first investigated the role of FMOD in GSC and DGC growth and interconversion between both cell populations in vitro using two human (MGG8 and U251) and two murine (AGR53 and DBT-Luc) glioma cell lines. We found that the absence of FMOD neither affected GSC growth and differentiation to DGC (Figure 2—figure supplement 1, Figure 2—figure supplement 2, Figure 2—figure supplement 3) nor DGC growth and reprogramming to form GSCs (Figure 2—figure supplement 4, Figure 2—figure supplement 5, Figure 2—figure supplement 6; more details in Appendix 1), consistent with our previous findings showing that FMOD does not affect glioma cell proliferation in vitro (Mondal et al., 2017). In line with previous findings that DGCs cooperate with GSCs to promote tumor growth (Wang et al., 2018), we then evaluated the ability of DGCs silenced for FMOD to support the growth of tumors initiated by GSCs in co-implantation experiments in a syngeneic mouse model using GSCs and DGCs derived from DBT-Luc glioma cells. Reminiscent of our observations in MGG8 cell line, DBT-Luc-DGCs express higher levels of FMOD than DBT-Luc-GSCs (Figure 2—figure supplement 5C D). To silence the expression of FMOD in DBT-Luc-DGCs, we used a doxycycline-inducible construct that contains an inducible mCherry-shRNA downstream of the Tet-responsive element (Angel et al., 2020; Figure 2A). The scheme of the co-implantation experiment is described on Figure 2B. DBT-Luc-GSC cells were coinjected with either DBT-Luc-DGC/miRNT (nontargeting shRNA) or DBT-Luc-DGC/miRFMOD (FMOD shRNA). In both groups, 50% of the mice received doxycycline on alternated days from day 9 post-injection until the end of the experiment. Tumors in mice coinjected with DBT-Luc-GSCs and DBT-Luc-DGCs/miRNT grew much faster and reached a significantly larger size (measured by bioluminescence) than tumors in mice injected with DBT-Luc-GSCs alone regardless of doxycycline treatment (Figure 2B–D, compare black and purple lines with blue line; Supplementary file 3). Notably, mice treated with doxycycline did show mCherry expression in tumors (Figure 2C). In contrast, injected DBT-Luc-DGC/miRFMOD cells failed to support the growth of tumors initiated by DBT-Luc-GSCs in doxycycline-treated mice compared to doxycycline-untreated mice (Figure 2B–D, compare red line with orange line; Supplementary file 3). While mice injected with DBT-Luc-GSCs+DBT-Luc-DGCs/miRFMOD (Dox+) showed an increase in tumor growth until the onset of doxycycline treatment (as seen in the rise in bioluminescence), subsequent tumor growth was drastically reduced. As expected, mice injected with DBT-Luc-DGCs alone developed substantially small tumors (Figure 2C and D). The small tumors formed in animals injected with either DBT-Luc-GSC+DBT-Luc-DGC/miRFMOD (Dox+) or DBT-Luc-GSC alone expressed significantly less FMOD protein than other tumors (Figure 2—figure supplement 7). These results indicate that FMOD secreted by DGCs is essential for the growth of tumors initiated by GSCs.
Tumor cell interactions with stromal cells are critical for glioma tumor growth (Pine et al., 2020). Small leucine-rich proteoglycans such as FMOD promote angiogenesis in the context of cutaneous wound healing (Pang et al., 2019; Zheng et al., 2014). In addition, we previously found a significant enrichment of the term ‘angiogenesis’ among differentially regulated genes in FMOD-silenced U251 glioma cells (Mondal et al., 2017). In light of these observations, we next examined the impact of FMOD on tumor angiogenesis. First, we tested the ability of glioma cell-derived FMOD to induce angiogenic network formation by immortalized human pulmonary microvascular endothelial cells (ST1). We used LN229 and U251 glioma cells, which express low and high levels of FMOD, respectively, for overexpression and silencing studies (Mondal et al., 2017). We found that the CM derived from LN229 cells stably expressing FMOD (LN229/FMOD) induced more angiogenesis than LN229/Vector stable cells (Figure 3A and B). Further, the CM of FMOD-silenced U251 cells was less efficient in promoting angiogenesis than the CM of cells expressing nontargeting siRNA (Figure 3C, Figure 3—figure supplement 1A) or shRNA (Figure 3—figure supplement 1B). The addition of recombinant human FMOD (rhFMOD) to the CM of U251/siFMOD cells rescued its ability to induce angiogenesis (Figure 3C, Figure 3—figure supplement 1A). More importantly, the addition of rhFMOD directly to endothelial cells induced angiogenesis in the presence of a control antibody (IgG) but not in the presence of an FMOD neutralizing antibody ( Figure 3—figure supplement 1C).
Figure 3—source data 1
Figure 3—source data 2
Figure 3—source data 3
Figure 3—source data 4
Figure 3—source data 5
Figure 3—source data 6
Figure 3—source data 7
Figure 3—source data 8
Figure 3—source data 9
Figure 3—source data 10
Both CMs derived from three DGCs and their corresponding GSCs also induced angiogenesis efficiently (Figure 3—figure supplement 1D). Further, pretreating cells with an FMOD antibody significantly reduced the ability of MGG8-DGC CM, but not that of MGG8-GSC CM, to induce angiogenesis (Figure 3D). The reduced ability of the FMOD antibody-pretreated DGC CM to promote angiogenesis was rescued by the exogenous addition of an excess of rhFMOD (Figure 3D). Moreover, CM derived from FMOD-silenced MGG8-DGCs was less efficient in promoting angiogenesis than the CM of shNT (Figure 3D) or siNT (Figure 3—figure supplement 1E) transfected MGG-DGCs. The effect of FMOD silencing was rescued by adding exogenous rhFMOD (Figure 3D, Figure 3—figure supplement 1E). Both rhFMOD and FMOD present in the CM collected from MGG8-DGC induced the migration and invasion but not the proliferation of ST1 cells (Figure 3—figure supplement 2A–E). Further, CM from MGG8-DGC/shNT cells was more efficient than CM from MGG8-DGC/shFMOD cells in promoting angiogenic network formation by human brain-derived primary endothelial cells (HBMECs) and mouse brain-derived immortalized endothelial cells (B.End3) (Figure 3E and F, Figure 3—figure supplement 2F). Again, the effect of FMOD silencing was rescued by adding rhFMOD (Figure 3E and F, Figure 3—figure supplement 2F).
Vascular mimicry (VM) is one of the alternative mechanisms of angiogenesis wherein tumor-derived endothelial cells (TDECs) originate from GBM cells (Angara et al., 2017; Ricci-Vitiani et al., 2010; Soda et al., 2011). To assess the ability of FMOD to induce TDECs derived from DGCs to form angiogenic networks, we used MGG8-DGC and U87 cells. MGG8-DGC/shNT cells grown in endothelial media (M199) under hypoxia (1% O2) differentiated to TDECS as evidenced by an increase in CD31 (Figure 3G and H). MGG8-DGC/shFMOD also differentiated to form TDECs, albeit with less efficiency (Figure 3G and H). Further, the addition of rhFMOD induced both MGG8-DGC/shNT-TDEC and MGG8-DGC/shFMOD-TDEC cells to form angiogenic networks efficiently (Figure 3I and J). Similarly, U87 cells differentiated to TDECs (Figure 3—figure supplement 3A B), which readily formed angiogenic networks in the presence of rhFMOD (Figure 3—figure supplement 3C D). Collectively, these results demonstrate that DGC CM can induce angiogenesis and identify FMOD as a critical mediator of DGC-induced angiogenesis.
To dissect the signaling mechanisms underlying FMOD-induced angiogenesis, we subjected protein extracts derived from ST1 endothelial cells treated or not with rhFMOD to reverse phase protein array (RPPA). A total of 12 proteins exhibited differential abundance in a time-dependent manner in rhFMOD-treated ST1 endothelial cells (Figure 4A). These include HES1, a downstream target of the Notch signaling pathway that has been shown to promote angiogenesis (Zhao et al., 2017). We thus investigated the possible involvement of Notch signaling in FMOD-induced angiogenesis. The addition of rhFMOD induced luciferase activity of Notch-dependent CSL-Luc and HES-Luc reporters in ST1 cells but not in gamma-secretase inhibitor (GSI; a Notch pathway inhibitor) pretreated cells (Figure 4—figure supplement 1A B). The addition of rhFMOD also increased HES1 mRNA and protein levels in ST1 cells, an effect abolished by GSI pretreatment of cells (Figure 4B and C). rhFMOD treatment also resulted in the translocation of NICD (Notch intracellular domain) from the cytosol to the nucleus, as shown by subcellular fractionation and confocal microscopy (Figure 4—figure supplement 1C D). Furthermore, rhFMOD failed to induce angiogenic network formation by GSI-pretreated ST1 cells (Figure 4D). In addition, ST1 cells having a stable expression of NICD (ST1/NICD) showed enhanced angiogenic network formation than ST1 vector stable (ST1/Vector) cells (Figure 4—figure supplement 2A–D). FMOD present in CM from MGG8-DGCs induced ST1/Vector cells, but not ST1/NICD cells, to form more angiogenic networks, suggesting that Notch activation in endothelial cells is an essential step in FMOD-induced angiogenesis.
Figure 4—source data 1
Figure 4—source data 2
Figure 4—source data 3
Figure 4—source data 4
Figure 4—source data 5
Figure 4—source data 6
Figure 4—source data 7
Figure 4—source data 8
Figure 4—source data 9
The increase in phosphorylated FAK (pFAK, FAK_Py397-R-V; the molecule downstream of integrin signaling) levels in rhFMOD-treated endothelial cells, as shown by RPPA (Figure 4A), also suggested a possible role of integrin signaling in FMOD-induced angiogenesis. This is consistent with our previous findings indicating that FMOD activates integrin signaling via type I collagen to engage the FAK-Src-Rho-ROCK pathway and promote the migration of glioma cells (Mondal et al., 2017). We first confirmed the activation of integrin signaling by FMOD, as assessed by increased pFAK in rhFMOD-treated ST1 cells, but not in cells pretreated with RGD peptide, an integrin inhibitor (Figure 4E). The addition of RGD peptide inhibited angiogenesis induced by LN229/FMOD CM (Figure 4F). Likewise, angiogenesis induced by LN229/FMOD CM was completely inhibited when ST1 cells were pretreated with inhibitors of FAK (FAKi) or Src (PP2), two signaling molecules downstream of integrin (Figure 4F). These treatments also strongly reduced the basal level of angiogenesis elicited by the CM of LN229/Vector cells. In contrast, an inactive analog of Src inhibitor (PP3), as well as inhibitors of RAC1 and ROCK, failed to inhibit the ability of CM derived from LN229/FMOD cells to induce angiogenesis (Figure 4F). Our previous report also demonstrated that the interaction of FMOD with type I collagen is essential for integrin activation (Mondal et al., 2017). The C-terminal region of FMOD comprises 11 leucine-rich-repeats (LRRs), of which the 11th repeat binds to type I collagen (Oldberg et al., 2007). A synthetic interfering peptide (RLDGNEIKR) corresponding to the 11thh LRR of type 1 collagen, but not a modified peptide (RLDGNQIMR), competes with rhFMOD for binding to type I collagen to activate integrin signaling in glioma cells (Mondal et al., 2017; Oldberg et al., 2007). Consistent with these findings, rhFMOD-induced luciferase activity of CSL-Luc and HES-Luc (Figure 4—figure supplement 3A B) and angiogenesis by ST1 cells (Figure 4—figure supplement 3C) were significantly inhibited by the interfering peptide, but not the modified peptide, suggesting a crucial role of type I collagen-dependent activation of integrin signaling in FMOD-induced angiogenesis. To identify the α and β subunits of integrin involved in FMOD-mediated activation of integrin signaling in endothelial cells, we chose ITGA6, ITGB1, and ITGAV for investigation based on the analysis of transcriptome data derived from laser capture-dissected microvessels from the human brain (more details in Appendix 1). Silencing either of the selected three integrin subunits in ST1 cells reduced significantly the ability of rhFMOD to activate integrin as assessed by reduced pFAK levels (Figure 4—figure supplement 3D–I), thus demonstrating the involvement of αv/β1 and α6/β1 heterodimeric integrin receptors in FMOD activation of integrin signaling in endothelial cells.
Next, to examine a possible crosstalk between integrin and Notch signaling in FMOD-treated endothelial cells, we tested the effect of the RGD peptide on the ability of rhFMOD to induce Notch signaling. Pretreatment of ST1 cells with RGD peptide significantly reduced rhFMOD-elicited CSL-Luc and HES-Luc activation (Figure 4—figure supplement 1E F). Likewise, pretreatment of cells with FAKi or PP2, but not PP3, significantly reduced rhFMOD-induced CSL-Luc and HES-Luc activity in ST1 cells (Figure 4—figure supplement 4A–D). Further, rhFMOD failed to increase HES1 transcript and protein levels in ST1 cells treated with either RGD peptide, FAKi, or PP2, but not in cells treated with PP3 (Figure 4G and H, Figure 4—figure supplement 4E–H).
We next investigated the mechanistic link between integrin and Notch signaling in FMOD-treated ST1 cells. Since Notch activation by FMOD is sensitive to GSI treatment, we explored the activation of Notch ligands by integrin-FAK signaling in FMOD-treated cells. While the addition of rhFMOD induced DLL3 and JAG1 transcripts in ST1 cells, RGD peptide pretreatment abolished JAG1 induction (Figure 4—figure supplement 5A), suggesting that JAG1 might be a potential linking molecule. Consistently, pretreatment of cells with either FAKi or PP2, but not PP3, also abolished the ability of rhFMOD to induce JAG1 transcript in ST1 cells (Figure 4—figure supplement 5B C). Further, JAG1 silencing significantly decreased rhFMOD-induced activity of CSL-Luc and HES-Luc reporters and HES1 transcript/protein levels in ST1 cells (Figure 4—figure supplement 5D E, Figure 4I–K). JAG1 silencing also abolished the ability of rhFMOD to induce angiogenic networks by ST1 cells (Figure 4L). In addition, rhFMOD induced the transcript levels of KLF8, a FAK-inducible transcription factor in ST1 cells, but not in RGD peptide pretreated cells (Figure 4—figure supplement 5F). Further, KLF8 silencing in ST1 cells significantly reduced the ability of rhFMOD to increase JAG1 level (Figure 4—figure supplement 5G H), suggesting that KLF8 activates JAG1 through an integrin-dependent pathway in FMOD-treated endothelial cells. Collectively, these results identify JAG1 as a molecular link between integrin and Notch signaling pathways in FMOD-treated endothelial cells and demonstrate a key role of integrin-FAK-JAG1-Notch-HES1 signaling in FMOD-induced angiogenesis.
To further explore the clinical relevance of these findings, we interrogated transcriptome datasets from multiple sources (more details in Appendix 1). We found a significant upregulation of transcript levels of FMOD, JAG1, and HES1 in GBM from various datasets (Figure 4—figure supplement 6A–C). We also found a positive correlation between FMOD and HES1 transcripts and between FMOD and JAG1 transcripts in the majority of GBM datasets analyzed (Figure 4—figure supplement 6D E), which further substantiates the functional link between FMOD, JAG1, and HES1. We also show that high FMOD transcript levels and hypomethylation of FMOD promoter are associated with poor prognosis in most datasets (Figure 4—figure supplement 7A B). These observations provide additional support for the activation of integrin-Notch signaling in FMOD-treated endothelial cells.
While GSCs alone can initiate a tumor, tumor growth requires continuous differentiation to form DGCs, which form the bulk of the tumor mass. In line with our co-implantation experiments (Figure 2), we sought to define the importance of FMOD secreted by DGCs generated through a differentiation program initiated by GSCs during tumor growth in vivo using a syngenic intracranial glioma mouse model. We injected AGR53-GSC-miRNT and AGR53-GSC-miRFMOD cells intracranially into C57/black mice and allowed them to form tumors. 13 days after intracranial injections, both groups received doxycycline as indicated (Figure 5A). The understanding is that as the tumors start growing, GSCs, in addition to their self-renewal, will start differentiating de novo to form DGCs, which would express high levels of FMOD. However, doxycycline treatment would inhibit FMOD expression, and thus one could investigate the importance of DGC-secreted FMOD in tumor growth. AGR53-GSC/miRNT and AGR53-GSC/miRFMOD cell-initiated tumors showed a similar size 7 days after doxycycline treatment as shown by mCherry fluorescence (Figure 5B and C; day 21). However, upon subsequent follow-up, doxycycline administration significantly inhibited the growth of AGR53-GSC/miRFMOD-initiated tumors over time but not that of AGR53-GSC/miRNT tumors (Figure 5B, C and E). Further, doxycycline administration increased the survival of mice injected with AGR53-GSC/miRFMOD cells compared to AGR53-GSC/miRNT cells (Figure 5D). AGR53-GSC/miRFMOD-initiated tumors showed decreased FMOD, mCherry, and GFP expression compared to AGR53-GSC/miRNT-initiated tumors (Figure 5F and G). DBT-Luc GSC/miRFMOD, another murine glioma GSC cell line, produced similar results: FMOD silencing after doxycycline administration resulted in reduced tumor growth (Figure 5—figure supplement 1A B), increased mice survival (Figure 5—figure supplement 1C), and decreased FMOD expression (Figure 5—figure supplement 1D).
Figure 5—source data 1
Figure 5—source data 2
Figure 5—source data 3
Figure 5—source data 4
Figure 5—source data 5
Figure 5—source data 6
To determine the relevance of our findings to the human pathology, we investigated the importance of DGC-secreted FMOD in the growth of tumors initiated by MGG8 and U251 cells using a xenograft mouse glioma model. The higher expression of CD133 confirmed the enrichment of CSCs in MGG8-GSC neurospheres compared to MGG8-DGCs (Figure 6A). We established orthotopic xenografts using MGG8-GSC/shNT and MGG8-GSC/shFMOD cells. Reminiscent of results obtained using transplantation of murine glioma cells, MGG8-GSC/shNT-transplanted mice readily developed intracranial tumors, whereas MGG8-GSC/shFMOD mice showed impaired tumor formation and increased mice survival (Figure 6B and C). Immunostaining and confocal microscopy analysis showed high expression of FMOD in MGG8-GSC/shNT tumors, while FMOD was barely detectable in MGG8-GSC/shFMOD tumors (Figure 6D). Likewise, U251/shFMOD cells that showed reduced FMOD protein levels compared to U251/shNT (Figure 2—figure supplement 6A) developed smaller tumors, and mice bearing U251/shFMOD tumors had longer survival than those carrying U251/shNT tumors (Figure 6—figure supplement 1A–D). As expected, the expression of FMOD was more elevated in U251/shNT tumors than in U251/shFMOD tumors (Figure 6—figure supplement 1E). Collectively, these findings indicate that DGC-secreted FMOD is essential for the growth of both human and mouse glioma.
Figure 6—source data 1
Figure 6—source data 2
Figure 6—source data 3
Figure 6—source data 4
Next, we investigated the cellular differentiation and angiogenesis in the tumors toward understanding the mechanism underlying reduced tumor growth in FMOD-silenced conditions. First, we measured the expression of CD133 and GFAP markers as the representation of GSCs and DGCs in the tumors formed in the animal models. The CD133-positive cells are much less in proportion compared to GFAP-positive cells in tumors formed by AGR53-GSC, DBT-Luc-GSCs, and MGG8-GSCs under FMOD nontargeting conditions (Figure 7—figure supplement 1A B, Figure 7—figure supplement 2A B, Figure 7—figure supplement 3A B), in good correlation to the low proportion of GSCs seen in brain tumors (Singh et al., 2004; Galli et al., 2004; Calabrese et al., 2007). Further, most GFAP-positive cells are also positive for FMOD expression compared to CD133-positive cells (Figure 7—figure supplement 1A C, Figure 7—figure supplement 2A C, Figure 7—figure supplement 3A C), recapitulating the results we obtained in vitro, where FMOD is expressed specifically by DGCs (Figure 1). We also found a similar expression pattern of CD133 and GFAP markers in small tumors formed under FMOD-silenced conditions in all three tumor models (Figure 7—figure supplement 1D E, Figure 7—figure supplement 2D E, Figure 7—figure supplement 3D E). The GFAP staining in these tumors confirms the occurrence of an efficient differentiation program even in FMOD-silenced conditions, which confirms our results obtained in vitro, where the absence of FMOD failed to affect the GSC differentiation to form DGCs (Figure 2—figure supplement 1, Figure 2—figure supplement 2, Figure 2—figure supplement 3). As expected, the small tumors formed in FMOD-silenced conditions showed substantially reduced FMOD staining (Figure 7—figure supplement 1D F, Figure 7—figure supplement 2D F, Figure 7—figure supplement 3D F).
We next evaluated the extent of blood vessel formation by measuring the endothelial cell marker immunostaining in the tumors. The small size tumors formed by AGR53-GSC/miRFMOD cells after doxycycline treatment showed reduced staining for CD31 and von Willebrand factor (vWF) compared to AGR53-GSC/miRNT tumors (Figure 7A-C, Figure 7—figure supplement 4A–C). Tumors formed by DBT-Luc-GSC/miRFMOD cells in doxycycline-treated mice also showed significantly reduced CD31 staining compared to that measured in the absence of doxycycline treatment (Figure 7—figure supplement 4D–F). Reminiscent of murine glioma tumors, tumors induced by MGG8-GSC/shFMOD cells also showed reduced CD31 staining compared to MGG8-GSC/shNT cells (Figure 7D–F). We then tested the extent of blood vessel formation by TDECs. In all three tumor models (AGR53-GSCs, DBT-Luc-GSCs, and MGG8-GSCs), blood vessels formed by TDECs were significantly reduced in tumors formed in FMOD-silenced conditions (Figure 7G–L, Figure 7—figure supplement 4G–I). These findings confirm our previous results obtained in vitro, where the absence of FMOD decreased the ability of host-derived endothelial cells and TDECs to form angiogenic networks (Figure 3—figure supplement 1, Figure 3—figure supplement 2, Figure 3—figure supplement 3). Next, we investigated the involvement of integrin-FAK-JAG1-Notch-HES1 signaling in FMOD-induced angiogenesis in the context of glioma tumors. Confocal microscopy analysis in tumors formed by AGR53-GSCs, DBT-Luc-GSCs, and MGG8-GSCs revealed a significant colocalization of the endothelial cell marker CD31 with FMOD, pFAK, JAG1, and HES1 markers in blood vessels (Figure 7—figure supplement 5A–C). From these results, we conclude that angiogenesis induced by DGC-secreted FMOD is essential for glioma tumor growth.
Figure 7—source data 1
Figure 7—source data 2
Figure 7—source data 3
Figure 7—source data 4
Figure 7—source data 5
Figure 7—source data 6
Figure 7—source data 7
The importance of GSCs in tumor initiation, growth, immune escape, angiogenesis, invasion into the normal brain, and resistance to therapy is also well established (Bao et al., 2006; Wakimoto et al., 2009). GSCs are known for their ability to self-renew and differentiate to form DGCs, the bulk cells of tumors (Galli et al., 2004; Ignatova et al., 2002; Singh et al., 2004; Suvà et al., 2014; Yang et al., 2020a). However, the role of DGCs in tumor growth remains poorly understood. The key requirement of tumor cells for self-maintenance in a novel tumor niche is the supply of nutrients. GSCs are known to promote the establishment of a highly vascularized microenvironment by being in close physical contact with endothelial cells (Calabrese et al., 2007). In addition, GSC-secreted factors such as VEGF-A, GDF15, IL8, and miR21 also have been shown to induce tumor angiogenesis (Conroy et al., 2018; Sun et al., 2017; Thirant et al., 2013; Treps et al., 2017; Zhu et al., 2021).
The massive proportion of DGCs in the tumor suggests that GSC-initiated angiogenesis might not be sufficient to meet the large nutrient requirement of the entire tumor. Wang et al. demonstrated that DGC-secreted BDNF is essential for GSC growth and maintenance through DGC-GSC paracrine signaling, which highlights the crucial role of DGC-secreted proteins in tumor formation (Wang et al., 2018). Further, a possible interaction between DGCs and stromal cells, such as endothelial cells, cannot be ruled out. We hypothesized that in addition to GSCs, DGCs might play an essential role in autocrine and paracrine signaling involving different types of cells to augment tumor growth. This study demonstrates the existence of paracrine signaling between DGCs and endothelial cells, which promotes angiogenesis and glioma tumor growth.
Previously, we have demonstrated that FMOD is highly expressed in GBMs compared to normal brain tissues. The loss of FMOD expression hampers the migratory function of glioma cells but has no impact on glioma cell proliferation (Mondal et al., 2017). This study identifies that FMOD is expressed exclusively by DGCs and further showed that FMOD is not needed for GSC/DGC growth and their plasticity to form one from the other. Given these facts, the inability of DGCs silenced for FMOD to support the growth of tumors initiated by GSCs was unexpected. However, it enlightened us that DGC-secreted FMOD has some essential yet unidentified functions supporting GSC-initiated tumor growth.
Here, we explored the mechanisms underlying the role of DGCs acting as critical support for GSC-initiated tumor growth. We demonstrate that DGC-secreted FMOD promotes tumor growth by inducing angiogenesis through integrin-dependent Notch signaling in endothelial cells, thus highlighting the importance of DGCs in tumor–stroma interactions that contribute to a sustainable niche for tumor growth. We further investigated the mechanism by which FMOD activates integrin-dependent Notch signaling in endothelial cells. Based on RPPA data that showed upregulation of HES1 in endothelial cells upon FMOD treatment, we demonstrated that activation of Notch signaling is essential for FMOD-induced angiogenesis. The importance of Notch signaling in glioma, especially in GSC growth and angiogenesis, is well documented (Bazzoni and Bentivegna, 2019; Stockhausen et al., 2010; Teodorczyk and Schmidt, 2014). Similarly, activation of Notch signaling in endothelial cells has been involved in tumor development and angiogenesis (Gridley, 2007; Kofler et al., 2011; Phng and Gerhardt, 2009). RPPA data also showed an increase in pFAK in FMOD-treated glioma cells. Our previous study showed that FMOD acts on glioma cells via the integrin-FAK-Src-Rho axis to promote migration (Mondal et al., 2017). This study demonstrates that FMOD-activated integrin signaling is essential for Notch pathway activation in ST1 cells. Integrin signaling in endothelial cells has been shown to play a crucial role in angiogenesis (Short et al., 1998; Silva et al., 2008). Based on our results indicating that FMOD-activated Notch signaling in endothelial cells is inhibited by GSI, we predicted that FMOD-elicited activation of the Notch pathway could involve Notch ligand-dependent process. Our experiments identified that JAG1 is the linker molecule that integrates integrin signaling to the Notch pathway. We also found a significant colocalization of endothelial cell marker CD31 with FMOD, pFAK, JAG1, and HES1 in blood vessels of tumors formed in mouse models. Finally, endothelial cells stably expressing NICD showed enhanced angiogenic network formation in an FMOD-independent manner, suggesting that Notch activation is an essential step in FMOD-induced angiogenesis. Thus, our results show that integrin-FAK-Src-KLF8-JAG1-dependent Notch signaling activation in endothelial cells mediates FMOD-induced angiogenesis.
Finally, in an orthotopic intracranial GBM mouse model, we show that conditional silencing of FMOD in newly generated DGCs during tumor growth leads to a significant reduction of tumor growth. Supporting our in vitro data indicating that FMOD is not required for GSC growth and differentiation, we found that the small tumors formed in FMOD-silenced conditions show differentiation of GSCs. However, these tumors exhibited poorly developed blood vasculatures of host-derived endothelial cells and TDECs. The lower tumor burden in the absence of FMOD might be attributed to insufficient nutrient supply to sustain tumor growth due to the reduced blood vessel density. Thus, our study establishes an essential role of paracrine signaling between the DGCs and the stroma in the context of tumor growth in the natural tumor niche complexity. It also demonstrates the importance of FMOD secreted by DGCs in promoting human glioma tumor growth in a mouse model. We propose a tumor evolution model (Figure 7M), whereby GSCs, in addition to their self-renewal, continuously differentiate to form DGCs, which secrete protein factors like FMOD that mediate paracrine signaling in the different cell types of the tumor, thus creating a niche favorable to tumor growth.
In conclusion, this study demonstrates that in addition to GSCs, DGCs have an essential role in tumor growth and maintenance. While the therapy-resistant and self-renewing GSCs trigger the early events of transformation and growth, DGCs, the proportion of which continues to increase during tumor growth, progressively become essential. Thus, targeting both CSCs and differentiated cancer bulk cells is vital to achieving a durable therapeutic response. The study also highlights the potential of GSC and DGC CM analysis to uncover novel targets in cancer therapy and the critical influence of DGC-secreted FMOD in glioma tumor growth.
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Dr. Kumaravel Somasundaram (firstname.lastname@example.org).
This study did not generate any new unique reagents.
Experiments were performed in C57&BL/6J female mice and Athymic Nude female mice (6–8 weeks old) following the approval by the Institute Ethical Committee for Animal Experimentation (Institute Animal Ethics Committee [IAEC] Project Number: CAF/Ethics/752/2020). The mice were kept in a 12 hr light and dark cycle, fed ad libitum with a normal diet, and the experiments were done in the light phase of the cycle.
Primary human tumor-derived GSCs MGG4, MGG6, and MGG8 were extensively characterized and kindly gifted to us by Dr. Wakimoto, Massachusetts General Hospital, Boston, USA (Wakimoto et al., 2009) . Characterization of AGR53 mouse-derived cell line has been described before (Angel et al., 2020; Friedmann-Morvinski et al., 2016). DBT-Luc cells were extensively characterized (Jiang et al., 2014) and gifted to us by Dr. Dinesh Thotala, Washington University in St. Louis, St Louis, MO. ST1 endothelial cells were extensively characterized (Krump-Konvalinkova et al., 2001) and gifted to us Dr. Ron Unger, Johannes Gutenberg University, Germany. B.End3 cell line was purchased from The American Type Culture Collection (ATCC, #CRL-2299, Certification of Analysis can be downloaded from https://www.atcc.org/products/crl-2299 with the lot# 70030469). The HBMECs were purchased from Cell Biologics, USA (#H-6023). The Certificate of Analysis can be found at https://cellbiologics.com/index.php?route=product/product&product_id=2959. The U251 cell line was bought from the European Collection of Authenticated Cell Cultures (ECACC, # 09063001). The Certificate of Analysis can be found at https://www.culturecollections.org.uk/products/celllines/generalcell/detail.jsp?refId=09063001&collection=ecacc_gc. The U87 cell line was bought from the ECACC (#89081402). The Certificate of Analysis can be found at https://www.culturecollections.org.uk/products/celllines/generalcell/detail.jsp?refId=89081402&collection=ecacc_gc. The LN229 cell line was bought from ATCC (# CRL-2611, https://www.atcc.org/products/crl-2611). All cell lines were obtained from the sources mentioned above within a year of experimentation, and additional yearly authentications were carried out wherever necessary. All cell lines are verified to be mycoplasma free by EZdetect PCR Kit for Mycoplasma Detection (HiMedia).
The RAR3G vector (in which miRNT and miRFMOD are cloned for the inducible shRNA experiments) was previously described (Friedmann-Morvinski et al., 2012). shRNA for FMOD (miRFMOD) was cloned following the mir30 (miRNT)-inducible backbone under the TRE (tetracycline response element) promoter, which is placed downstream of the mCherry reporter gene. The m2RtTA transactivator is expressed from the same vector in the opposite direction under the EF1α promoter, and following a 2A peptide, the puromycin gene is placed for selection in vitro. The FMOD overexpression plasmid was bought from Origene, USA (#LY419579). NICD pCMV Neo/intracellular domain of human Notch1 (NIC-1) and CSL Luc were kind gifts from Prof. Thomas Kandesch, Department of Genetics, University of Pennsylvania School of Medicine. Hes-Luc plasmid was bought from Addgene, USA (#43806).
FMOD (Abgent AP9243b, 1:2000), FAK (Cell Signaling Technology 3285S, 1:1000), pFAK (Cell Signaling Technology #3283S, 1:500), GAPDH (Sigma #G8795, 1:20,000), Actin (Sigma A3854, 1:20,000), HES1 (Cell Signaling Technology #D6P2U, 11988S, 1:1000), vWF (Abcam #6994, 1:2000), CD31 (Cell Signaling Technology #89C2, mouse mAb 1:200 for IHC), JAG1 (Cell Signaling Technology, Jagged1 [D4Y1R] XP rabbit mAb #70109, 1:200 for IHC, 1:1000 for WB), FMOD (fibromodulin polyclonal antibody PA5-26250, Invitrogen, IHC 1:100, WB 1:1000), CD133 (recombinant anti-CD133 antibody-EPR20980-104 #ab216323 Abcam, Flow Cyt. 1:100, Prom1 monoclonal antibody-2F8C5, #MA1-219, IHC 1:100), SOX2 (Cell Signaling Technology #3579 rabbit mAb, IHC 1:100), GFAP (anti-GFAP antibody ab7260, Abcam ICC 1:200, recombinant anti-GFAP antibody [EPR1034Y] – mouse IgG2a [ab279290], IHC 1:200), NICD (NOTCH1 (cleaved Val1744) polyclonal antibody PA5-99448, WB 1:200), SMAD2 (Smad2 [D43B4] XP rabbit mAb #5339, WB 1:1000), pSMAD2 (phospho-SMAD2 [Ser465/Ser467] [E8F3R] rabbit mAb #18338, WB 1:1000), integrin beta-1 (Cell Signaling Technology #9699 rabbit mAb, WB 1:1000), integrin alphaV (Cell Signaling Technology #4711 rabbit Ab, WB 1:1000), integrin alpha-6 (recombinant anti-integrin alpha 6 antibody [EPR18124] [ab181551], WB 1:500), and KLF8 (anti-KLF8 antibody [ab168527], WB 1:500).
Goat anti-mouse HRP conjugate (Bio-Rad #170-5047, WB 1:5000), goat anti-rabbit (H+L) secondary HRP conjugate (Invitrogen, #31460, WB 1:5000), goat anti-mouse IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 488 (Invitrogen, #A-11029), goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 488 (Invitrogen, Cat# A-11034), goat anti-mouse IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 594 (Invitrogen, Cat# A-11032), goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 594 (Invitrogen, Cat# A-11037), goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 405 Plus (Invitrogen, #A48254), and goat anti-mouse IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 405 Plus (Invitrogen, Cat# A48255). All Alexa Fluor conjugated antibodies were used at a dilution of 1:500 for IHC and ICC.
Fibromodulin (FMOD) (NM_002023) Human, Origene, Cat# TP306534.
The GSCs were obtained by dissecting GBM tumor tissue and then treating with trypsin, followed by a trypsin inhibitor. The chopped tissue was then passed through a cell strainer to remove the debris. The obtained filtrate was plated in ultra-low attachment plates using the stem cell for neurosphere formation media described next. Then, neurospheres were grown in Neurobasal medium (#21103049, Gibco) supplemented with 1× l-glutamine (#25030081, stock 200 mM, i.e., 100×, Gibco), heparin 2 μg/ml (#H3393 Sigma), 1× B27 supplement (#17504044, Gibco, stock concentration 50×), 0.5× N2 supplement (#17502048, Gibco, stock concentration 100×), 20 ng/ml rhEGF (#g5021, Promega), 2 ng/ml rhFGF-basic (#100-18B-100 μg, PeproTech), and penicillin and streptomycin. To make single-cell suspensions for re-plating, the spheres were chemically dissociated after 7 days of plating, using NeuroCult Chemical Dissociation Kit (mouse, #05707) from Stem Cell Technologies, according to the manufacturer’s instructions.
For differentiation of GSCs to DGCs, fully grown neurospheres were collected from the ultra-low attachment plates (Corning, USA) and plated on normal cell culture dishes in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and antibiotics, as mentioned earlier, for 15–20 days (Suvà et al., 2014).
The differentiated counterparts of GSCs and differentiated GBM cell lines were removed from 10% FBS containing DMEM, spun down twice in phosphate-buffered saline (PBS) to remove any trace of FBS, and plated on ultra-low attachment plates in Neurobasal medium containing all the supplements mentioned earlier and antibiotics for 7–10 days.
Viral infection was done in the GSCs using lentivirus transduction for nontargeting shRNA (shNT) or shRNA for the gene of interest. The small pellets of cells were collected 48 hr after viral infection, dissociated to form single cells, counted, and replated in six-well plates (3 × 104 cells/well) in complete Neurobasal medium. At the same time, cells were harvested and checked for specific gene manipulation (like knockdown verification). Media were replenished every 2–3 days, and neurosphere formation was monitored till the sixth or seventh day after replating. The number of neurospheres, sphere diameter, and size was analyzed using ImageJ software. Spheres having an area less than 50 µm2 were excluded from the analysis.
For each condition, 1, 10, 50, 100, and 200 GSCs (single cells) were plated in 10 wells each, respectively, of a 96-well plate, and sphere formation was assessed over the next 5–7 days. The number of wells not forming spheres was counted and plotted against the number of cells per well. Extreme limiting dilution assay was done using the ELDA software available online (https://www.elda.at/cdscontent/?contentid=10007.854970&portal=eldaportal).
GSC cell lines were grown as neurospheres in a complete Neural Stem Cell medium containing glutamine, heparin, N2, B27, EGF, and FGF for 6 days. They were thoroughly washed using PBS and replated without disruption in Neural Stem Cell medium devoid of supplements and growth factors. 36 hr after plating in incomplete medium, the CM was collected, spun at 1500 rpm for 15 min, filtered using 0.45 µm syringe filters, and stored at –80°C. The DGCs were grown in complete DMEM supplemented with 10% FBS, then washed with PBS, and CM was collected after 36 hr in an incomplete medium and similarly stored.
The CM were concentrated using Centricons (3 kDa cutoff; Merck) from 8 ml to 100 μl, followed by precipitation of proteins with trichloroacetic acid (TCA 10% for 30 min at 4°C). Equal amounts of proteins from each condition were run on gradient (4–20%) SDS gels and stained with Colloidal Coomassie blue. The gels were then destained, and each lane was cut into five equal pieces. Proteins were digested in-gel using trypsin (Gold Promega, 1 μg per band, overnight at 30°C) as previously described (Thouvenot et al., 2008).
Tryptic peptides were analyzed online by nano-flow liquid chromatography coupled to tandem-mass spectrometry using a Q-Exactive+ mass spectrometer (Thermo Fisher Scientific, Waltham, USA) coupled to an RSLC-U3000 HPLC (Thermo Fisher Scientific). Desalting and pre-concentration of samples were performed online on a PepMap pre-column (0.3 mm × 10 mm, Dionex). A gradient consisting of 2–25% B for 80 min, 25–40% B for 20 min, and finally 40–90% B for 5 min (A = 0.1% formic acid; B = 0.1% formic acid in 80% acetonitrile) at 300 nl/min was used to elute peptides from the capillary reverse-phase column (0.075 mm × 150 mm, Acclaim PepMap 100 C18, Thermo Fisher Scientific). Eluted peptides were electro-sprayed online at a voltage of 1.5 kV into the Q-Exactive+ mass spectrometer. A cycle of one full-scan mass spectrum (MS1, 375–1500 m/z) at a resolution of 70,000, followed by 12 data-dependent tandem-mass (MS2) spectra, was repeated continuously throughout the nano-LC separation. Parameters used for MS2 spectra were the resolution of 17,500, AGC target of 1e5, normalized collision energy of 28, and an isolation window of 1.2 m/z. Mass spectra were processed using the MaxQuant software package (v 22.214.171.124, Cox and Mann, 2008) against the UniProtKB Reference proteome UP000005640 database for Homo sapiens (release 2018_11) and contaminant database. The following parameters were used: enzyme specificity set as trypsin/P with a maximum of two missed cleavages, oxidation (M) and acetyl (protein N-term) set as variable modifications, and carbamidomethyl (C) as fixed modification and a mass tolerance of 0.5 Da for fragment ions. The maximum false peptide and protein discovery rate was specified as 0.01. Match between runs was used to transfer identification from run to run. Relative protein quantification was performed using LFQ intensity in MaxQuant. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD032958. For statistical analysis, missing values were defined using the imputation tool of the Perseus software (v. 126.96.36.199, Tyanova et al., 2016).
HEK-293T cells were seeded in 60 mm or 90 mm poly-l-llysine-coated cell culture dishes. The cells were transfected with shRNA plasmid and helper plasmids psPAX2 and pMD2.G using Opti-MEM (Invitrogen #22600-050) medium and Lipofectamine (https://www.thermofisher.com/order/catalog/product/12566014; Invitrogen) when the cells were 60–70% confluent. 6 hr after transfection, the Opti-MEM medium was replaced by a fresh DMEM supplemented with 10% FBS. 60 hr post-transfection, the supernatant from the transfected cells was collected in 15 ml Falcon tubes, centrifuged at 5000 rpm for 10 min, filtered through 0.45 μm, and stored at –80°C for future use.
For the virus used in endothelial cells, DMEM was removed after 24 hr of transfection and changed to complete M199, and the virus was collected after 60 hr, as mentioned previously. The same method was used for the virus for GSCs (DMEM was changed to NBM). The shRNA construct number TRCN0000152163 from Sigma human TRC shRNA library was used for knockdown studies of human FMOD. The shRNA construct number TRCN000094248 from Sigma mouse TRC shRNA library was used for knockdown studies of mouse FMOD. A pooled lentivirus using the construct numbers TRCN000024441, TRCN000024441, TRCN000024441, and TRCN000024420 was used for knockdown studies of JAG1 in the ST1 cells.
ST1 cells were grown in Medium 199 (Sigma #M4530) supplemented with 20% FBS and endothelial cell growth factor (ECGS) (Sigma #E2759), heparin, glutamine (1×) and 1× antibiotic-antimycotic solution (Sigma, # A5955, stock concentration 100×).
The MGG8-DGCs were grown in Medium 199, having the same composition as mentioned above, for 3 days and subjected to hypoxia (1% O2) for 8 hr. The cells then formed transdifferentiated endothelial cells (TDECs), showed endothelial morphology (data not shown), were harvested for checking markers levels, and were also plated for the in vitro angiogenesis assay.
For NICD stable cell line generation, the ST1 cells were transfected with NICD pCMV Neo/intracellular domain of human Notch1 (NIC-1) plasmid using Lipofectamine 200 (Invitrogen). They then were harvested 48 hr after transfection for Western blot and in vitro angiogenesis assay.
Total RNA from the cells was isolated from cells using the TRI reagent (#T9424 Sigma) according to the manufacturer’s instructions. Next, the integrity of the RNA was checked by running it on 2% MOPS-formaldehyde gel. RNAs were quantified by NanoDrop. 2 µg of total RNA were used for cDNA conversion using the high-capacity cDNA reverse transcription kit (Applied Biosystems, USA) according to the manufacturer’s protocol. The cDNAs were diluted in a ratio of 1:10 with nuclease-free water to make their final concentration 10 ng/µl. Subsequently, real-time quantitative PCR was done using the ABI Quant Studio 5 and 6 (Life Technologies, USA). The cDNA was used as a template, and DyNAmo Flash SYBR Green qPCR kit (#F-416L) was used. Gene-specific primer sets were used for the reaction (see table at the end) under the following conditions: 95°C for 15 min, 40 cycles of 95°C for 20 s, 60°C for 25 s, and 72°C for 30 s followed by the dissociation cycle for melt curve generation. Each sample was run either in duplicate or triplicate. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), ACTB (beta-actin), 18S rRNA, RPL35a (ribosomal protein L35a), and ATP5G1 (ATP synthase, H+transporting, mitochondrial F0 complex, subunit C1 [subunit 9]) were used as reference genes for human gene expression analysis. For mouse gene expression analysis, cyclophilin was used as a housekeeping gene. ΔΔCT method was used to calculate gene expression, which was transformed to log2 ratio and then to absolute scale for plotting.
For Western blotting analysis, cell pellets were lysed in RIPA lysis buffer (containing 1 mM sodium orthovanadate, 5 mM sodium fluoride, 1 mM phenylmethanesulfonyl fluoride, and 1× protease inhibitor cocktail, Sigma, USA), and proteins were isolated from the cells by spinning at 13,000 rpm for 30 min. The supernatant containing the proteins was collected. Protein concentrations were measured using Bradford’s reagent, and a standard BSA curve was used to determine the protein concentrations. Equal amounts of proteins from all conditions were mixed with protein loading dye (1×), denatured at 95°C for 15 min, loaded in each well of an SDS polyacrylamide gel, and the gel was run for around 8 hr. To prepare SDS- polyacrylamide gel, resolving and stacking gels were prepared at a concentration of 10–12 concentrations. The gel was run at 70–100 V, and then the proteins were transferred onto a polyvinyl fluoride (PVDF) membrane using the semi-dry transfer method. After the transfer, the membrane was blocked using 5% skimmed milk in 1× Tris-buffered saline-Tween (TBST) for 1 hr. Subsequently, the membrane was washed in TBST for 30 min and probed initially with primary antibodies in 5% BSA-TBST for 14–16 hr at 4°C. Then, the membrane was washed in TBST for 30 min, and the secondary antibody, diluted in 5% skimmed milk in TBST, was added and incubated at room temperature for 2–3 hr. Finally, the blot was washed and developed using PerkinElmer ECL Plus lightning and Bio-Rad Clarity and Clarity Plus ECL chemiluminescent reagent using GE Image Quant machine.
ChIP assay was conducted with chromatin isolated from ST1 cells treated with 10 μM TGF-β RI inhibitor (SB431542) and DMSO for 6 hr. Briefly, after cross-linking, the nuclei were prepared and sonicated to generate chromatin fragments between 100 and 10,000 bp following the manufacturer’s protocol using the Simple Chromatin Immunoprecipitation kit (CST; Cat# 9003). The sheared chromatin was collected by centrifugation (10,000 × g for 10 min at 4°C), and a 10 μl aliquot was removed to serve as a positive input sample. Aliquots of 100-μl-sheared chromatin were incubated with 2 μg of the required antibody/antibodies followed by Protein G magnetic beads for the stipulated time. An equal amount of IgG and H3 antibodies was used as control. The eluted DNA was analyzed by quantitative PCR using the FMOD promoter-specific primer set ChIP-F/ChIP-R (in the list of primers) to amplify the desired region in the FMOD promoter. Conditions of linear amplification were determined empirically for this primer. The PCR conditions were as follows: 95°C for 5 min; 95°C for 30 s, 56°C for 30 s, and 72°C for 30 s for 35 cycles. PCR products were resolved by electrophoresis on a 2% agarose gel and visualized after ethidium bromide staining. Real-time qPCR was performed with the same eluted DNA. The conditions were as follows: 95°C for 3 min; 95°C for 10 s, 56°C for 30 s, 72°C for 30 s for 40 cycles and 72°C for 5 min. The Ct values of different conditions were normalized to Ct values in IgG control.
Trans-well assay was done in 24-well Boyden chambers with 8 μm pore size polycarbonate membranes (BD Biosciences, San Diego, USA). ST1 cells (5 × 104) were resuspended in 500 μl serum-free Medium 199 and placed in the upper chamber. The lower chamber was filled with serum-free Medium 199 with 750 μl CM or 400 nM final concentration of recombinant protein dissolved in an incomplete medium (serving as a chemoattractant). After 24 hr of incubation, cells remaining on the upper surface of the membrane were removed with a wet cotton bud. The cells that migrated to the lower surface of the membrane were fixed in ice-cold methanol, stained with crystal violet, and imaged in a light microscope.
Trans-well assay was done in 24-well Boyden chambers with 8 μm pore size polycarbonate membranes coated with Matrigel (BD Biosciences). ST1 cells (5 × 104) were resuspended in 500 μl serum-free Medium 199 and placed in the upper chamber. The lower chamber was filled with serum-free Medium 199 with 750 μl CM or 400 nM final concentration of recombinant protein dissolved in an incomplete medium (serving as a chemoattractant). After 24 hr of incubation, cells remaining on the upper surface of the membrane were removed with a wet cotton bud. The cells that have invaded the lower surface of the membrane were fixed in ice-cold methanol, stained with crystal violet, and imaged in a light microscope.
1.5 × 103 ST1 cells were plated with 2% FBS-containing M199 Medium in each well of a 96-well plate. MTT assay was performed as per the established protocol. MTT was added to each well, and Formazan crystals formed after 3 hr of incubation were dissolved in DMSO, and the absorbance was measured at 420 nm. The first reading served as the untreated condition (0th time point). After this reading, the cells were treated with 400 nm rhFMOD or an equivalent concentration of bovine serum albumin (BSA), and readings were taken every 24 hr, till 96 hr. The cell viability was then plotted as a line graph.
ST1 endothelial cells were seeded in 96-well plates (10 × 103–15 × 103 cells per well), coated with Geltrex (Invitrogen), and grown in Medium 199 (Sigma) without growth factors. Equal protein amounts (50–100 µg) from serum-free CM from different conditions were added on top of the cells. After 10–12 hr of incubation, endothelial cells form tube-like structures. Each complete circular structure was considered one complete network (well-connected and not broken at any place). The total number of networks for each condition was counted double-blindly. For positive control, cells were plated in complete endothelial cell media (Medium 199) supplemented with ECGS and 20% FBS, and in the negative control, cells were plated in incomplete Medium 199 (without serum and ECGS).
Cells were plated on coverslips in 12-well plates and allowed to attach. The cells were then fixed with 4% paraformaldehyde (PFA) and permeabilized using PBS supplemented with 0.25% Triton-X100. Cells were then washed with PBS and blocked using PBS supplemented with 1% BSA, 0.3% Triton X-100, and 5% goat serum for 2 hr at room temperature. After blocking, the primary antibody, diluted in the blocking buffer, was added to the coverslips overnight at 4°C. For dual staining, the primary antibodies (one anti-mouse and the other anti-rabbit) were simultaneously added to the samples in the required dilutions. After removing the primary antibody, cells were thoroughly washed with PBS three times for 5 min each. Fluorescence-conjugated secondary antibody was dissolved in the blocking buffer and added to the cells for 3 hr, after which the cells were washed three times with PBS for 5 min. The cells were stained with DAPI (1 µl/ml) for 5 min at room temperature. Coverslips were mounted on glass slides using glycerol as a mounting agent and imaged using a Zeiss LSM 880 confocal microscope.
Cells were seeded in six-well plates and co-transfected with the reporter luciferase construct and pCMV-beta gal (as control) using Lipofectamine according to the manufacturer’s instructions. 24–48 hr after transfection, cell extracts were made in reporter lysis buffer (Promega). Protein concentrations of the cell lysates were measured by Bradford assay reagent (Bio-Rad). 10 µg of protein were mixed with 30 µl of luciferase assay reagent (LAR) to determine the luciferase activity, and values were normalized to β-galactosidase activity units.
The GSCs and DGCs are pelleted down and stained with live dead fluorescent stain (L34955 LIVE/DEAD Fixable Violet Dead Cell Stain Kit, for 405 nm excitation) for 15 min at 37°C. Then, the cells are blocked using 10% FCS and 1% sodium azide in PBS. After blocking, the cells are washed with PBS thrice, incubated with the primary antibody for 2 hr at room temperature, washed again thrice, and incubated with the secondary antibody for 2 hr in the dark. The final pellet was washed thrice with PBS, dissolved in 200 μl 3% BSA in PBS, and analyzed using the BD FACS Verse flow cytometer. The data were analyzed using the FlowJo software.
Untreated ST1 cells and cells treated for 10, 30, and 60 min with 400 nM rhFMOD were lysed using RPPA lysis buffer. Lysates were serially diluted in 5 twofold dilutions using lysis buffer and printed on nitrocellulose-coated slides using an Aushon Biosystem 2470 arrayer. Slides were probed with 304 validated primary antibodies followed by detection with appropriate biotinylated secondary antibodies. Slides were scanned, analyzed, and quantified using Array-pro Analyzer software (Media Cybernetics) to generate spot intensity (level 1 data). Signals were visualized by a secondary streptavidin-conjugated HRP antibody and DAB colorimetric reaction. The list of 304 antibodies can be found at http://www.mdanderson.org/research/research-resources/core-facilities/functional-proteomics-rppa-core/antibody-information-and-protocol.html.
Mice were perfused intracardiacally using 4% PFA solution, and the brains were harvested and stored in PFA for 12 hr and subsequently in 30% sucrose solution. The brains were then embedded in Poly-freeze solution (Sigma, #35059990) and sectioned into 20-μm-thick sections using a Leica Cryostat. The sections were stored at –80°C in Tissue Cutting Solution (TCS).
The brain sections were removed from TCS, put in 96-well plates, and washed thoroughly with PBS for immunofluorescence. Then, the same protocol was followed for immunofluorescence staining of tissue sections as that followed for monolayer cells grown on coverslips. At the final step, after adding the secondary antibody, DAPI or H33342, and PBS washes, the sections were individually mounted on glass slides using the ProLong Glass Antifade Mountant (Invitrogen, #P36980) and covered by coverslips. Images were taken using the Zeiss LSM 880 confocal microscope using ×10, ×20, and ×40 objectives for various conditions.
For immunofluorescence of FFPE sections, an extra step of antigen retrieval was performed by de-paraffinizing the sections in xylene, then boiling in distilled water twice for 5 min each. After this, the rest of the steps from permeabilization to mounting the same steps were followed as that of immunofluorescence of free-floating sections.
The areas of the blood vessels and the fluorescence intensities for all the fluorophores used in the tissue sections were measured using the Zeiss Black software (https://www.zeiss.com/microscopy/int/products/microscope-software.html), converted to percentages, and plotted as bar diagrams. For determining the extent of vascular mimicry in the blood vessels of the tissues, the overlap of green (coming from GFP or GFAP staining of the tumor cells) and red (a measure of CD31 expression of the endothelial cells) forming yellow color was calculated as a measure of colocalization coefficients in the Zeiss Black software. The absolute values of the colocalization coefficients were plotted for each condition.
All animal procedures followed were approved by the Institute Ethical Committee for Animal Experimentation (Institute Animal Ethics Committee [IAEC] Project Number: CAF/Ethics/752/2020). Mouse cell line DBT-Luc-DGC, which grows as a monolayer cell line (stable for luciferase expression), was reprogrammed to form DBT-Luc-GSCs. A combination of 105 DBT-Luc-GSCs (without any shRNA) and 106 DBT-Luc-DGCs that carried either the miRNT or miRFMOD constructs (referred to as DBT-Luc-DGC/miRNT or DBT-Luc-DGC/miRFMOD, respectively) was injected subcutaneously in the mice. Two groups of mice (n = 5 each) received a combination of 105 DBT-Luc-GSCs+106 DBT-Luc-DGC/miRNT and another two groups (n = 5 each) received a combination of 105 DBT-Luc/GSCs +106 DBT-Luc-DGC/miRFMOD (only one of the two groups for both miRNT and miRFMOD received doxycycline). Doxycycline injection (intraperitoneal, 100 μg per animal) began on 9th day post the injection and was given on every day up to the 16th day, after which it was given on every alternate day till the end of the experiment. Since the doxycycline administration induced the mCherry expression along with the shRNA expression, the animals were imaged for both bioluminescence (marked by red on the timeline) and fluorescence (marked by orange on the timeline) on the days indicated in the timeline. Two other groups (n = 5 each) received either only 105 DBT-Luc-GSCs or 106 DBT-Luc-DGCs (not stables for any shRNA) as controls.
Cells were harvested in incomplete DMEM or NBM depending on the cell type to be injected. 0.25 × 106 DGCs or 1 × 105 GSCs were injected intracranially (in the hippocampus, 2.5 mm deep) in each animal using a stereotaxic apparatus. The animals were imaged on the third day after injection and subsequently, every 5–6 days until the end of the experiment. In experiments using inducible shRNAs, doxycycline was injected every day for 10 days (from the 13th day after injection) and on every alternate day until the end of the experiment. MGG8-GSC (MGG8-GSC/shNT vs. MGG8-GSC/shFMOD), AGR53-GSCs (AGR53-GSC/miRNT vs. AGR53-GSC/miRFMOD), and DBT-Luc-GSCs (DBT-Luc-GSC/miRFMOD) cells were intracranially injected in this study.
In vivo imaging was done for bioluminescence or fluorescence with the PerkinElmer IVIS Spectrum by using mild gas anesthesia (using isoflurane) for the animals.
Brains from perfused mice were paraffin-embedded and sectioned using a microtome (Leica Biosystems, 5 μm sections). Sections were mounted on glass slides and removed from paraffin, rehydrated, and stained with Harris Hematoxylin (Merck, 6159380051046) for nuclear staining and Eosin Y (SDFCL, 44027G25) solution for cytoplasmic staining. The sections were then mounted using DPX mounting medium and imaged at 0.8× using a Lawrence and Mayo digital microscope.
The differentially expressed genes between the GSC and DGC (as identified in GSE54792) were pre-ranked based on fold change and used as an input to perform GSEA. All the gene sets available in the Molecular Signature Database (MSigDB, roughly 18,000 gene sets) were used to run the GSEA. We filtered out the TGF-β pathway-related gene sets to identify that most of them were significantly enriched in the DGCs over the GSCs. Similarly, the same analysis was carried out in multiple publicly available GBM vs. normal samples datasets to show the significant enrichment of TGF-β gene sets in GBM over normal. We acknowledge our use of the GSEA software and MSigDB (Subramanian et al., 2005; http://www.broad.mit.edu/gsea/).
Gene set variation analysis (GSVA) was performed using ssGSEA to determine the enrichment of the different molecular subtypes of GBM in GSE54792 and also the enrichment of the TGF-β Hallmark gene set from MSigDb. The higher GSVA score indicates the highest enrichment, which gradually decreases.
Kaplan–Meier survival analysis was done using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA).
The heatmaps were generated using the Multiple Experiment Viewer (MEV) software (http://www.tm4.org/mev.html) version 4.8.1. LFQ values for protein expression from mass-spec data and mean pixel density for the RPPA were used as inputs for heatmap generation. A nonparametric t-test was performed with a false discovery rate (FDR) and a p-value cutoff of 0.05.
Bar diagrams were generated using Microsoft Excel. The box plots were generated using GraphPad Prism 5.0 (GraphPad Software). The p-value was calculated by unpaired t-test with Welch’s correction or Student’s t-test done using Microsoft Excel indicated. ANOVA p-value was calculated using GraphPad Prism 5 software. Significance for Kaplan–Meier survival analysis as calculated by the log-rank test. Correlation coefficients were calculated using Pearson’s correlation coefficient using GraphPad Software. All experiments (except the ones involving the animal models) were done at least thrice with duplicates for each condition within the experiments. The data are represented as mean ± standard deviation. p-Value <0.05 is considered significant with *, **, ***, and **** representing p-values <0.05, 0.01, 0.001, and 0.00001, respectively. ns indicates nonsignificance.
|27||cPPT (for sequencing)||GAAGGAATAGAAGAAGAAGGT GGAGAG|
|S. no||Gene name||Clone 1||Clone 2||Clone 3||Clone 4||Clone 5|
|S. no.||Inhibitor name||Target molecule||Catalog number|
|1||Gamma-secretase inhibitor (GSI)||Gamma secretase||565750 (Merck)|
|2||RGD Peptide||Integrins||A5082 (Sigma-Aldrich)|
|4||PP3||Structural analog to PP2||529574 (Calbiochem)|
|6||RAC1 inhibitor||RAC1||553502 (Sigma-Aldrich)|
|8||SB431542||TGFB RI||616464 (Sigma-Aldrich)|
Before studying the importance of DGC secreted FMOD on tumor growth, we decided to investigate its requirement for the GSC, DGC growth, and their plasticity to form the other type. We have earlier shown that FMOD is not required for the proliferation of established glioma cell lines (Mondal et al., 2017). We used two human glioma cell lines, MGG8 and U251, two murine glioma cell lines, AGR53 (Angel et al., 2020) and DBT-Luc (Yuan et al., 2007).
MGG8-GSCs transduced with a small hairpin RNA targeting FMOD (MGG8-GSC/shFMOD) grew as neurospheres with equal efficiency as measured by neurosphere formation and limiting dilution assays and also differentiated to form DGCs as efficiently as control MGG8-GSCs transduced with non-targeting shRNA (MGG8-GSC/shNT) (Figure 2—figure supplement 1A–D). As expected, the differentiation was accompanied by the downregulation of glioma reprogramming factors in both MGG8-GSC/shNT and MGG8-GSC/shFMOD cells (Figure 2—figure supplement 1). In concordance with human GSCs, we found a higher expression of FMOD in AGR53-DGCs than AGR53-GSCs (Figure 2—figure supplement 2A–C). To silence the expression of FMOD in AGR53-DGC, we used a doxycycline-inducible FMOD shRNA (miRFMOD) construct that contains an inducible mCherry-shRNA cassette downstream of the Tet-responsive element (Angel et al., 2020; Figure 2A). Efficient silencing of FMOD in doxycycline-treated AGR53-DGC/miRFMOD cells was observed compared to AGR53-DGC/miRNT cells (Figure 2—figure supplement 2D E). Next, we investigated the impact of FMOD silencing on AGR53-GSC growth and differentiation to DGCs in vitro. Like human GSCs, AGR53-GSC/miRNT and AGR53-GSC/miRFMOD grew as neurospheres with equal efficiency both in the absence and presence of doxycycline (Figure 2—figure supplement 3A D). Further, both AGR53-GSC/miRNT and AGR53-GSC/miRFMOD differentiated efficiently to grow as a monolayer (Figure 2—figure supplement 3C). The differentiation resulted in the significant upregulation of astrocytic markers (Figure 2—figure supplement 3). These results confirm that FMOD is overexpressed in both human and murine DGCs but is not required for GSC growth and their differentiation to DGCs.
To study the role of FMOD on reprogramming of DGCs to form GSCs, AGR53-DGC and DBT-Luc DGC were tested for their ability to reprogram. DBT-Luc is a luciferase-expressing glioblastoma-derived cell line, which grows as a differentiated monolayer (referred to here onward as DBT-Luc-DGC) in an FBS-containing medium (Pang et al., 2019). AGR53-DGC/miRNT and AGR53-DGC/miRFMOD (both in the absence and presence of doxycycline) cells grow as a monolayer efficiently and reprogrammed to form neurospheres (Figure 2—figure supplement 4A). The reprogramming resulted in the significant upregulation of stem cell markers (Figure 2—figure supplement 4B). Similarly, DBT-Luc-DGC could readily reprogram to form DBT-Luc neurospheres (DBT-Luc-GSC; data not shown) with concomitant upregulation of stem cell markers and downregulation of astrocyte markers (Figure 2—figure supplement 5A B). DBT-Luc-DGCs expressed higher levels of FMOD transcript and protein compared to DBT-Luc-GSCs (Figure 2—figure supplement 5C D, respectively). To silence the expression of FMOD in DBT-Luc-DGC, we used the doxycycline-inducible shFMOD (miRFMOD) construct as explained above (Angel et al., 2020; Figure 2A). The addition of doxycycline resulted in downregulation of FMOD transcript and protein in DBT-Luc-DGC/miRFMOD but not in DBT-Luc-DGC/miRNT cells (Figure 2—figure supplement 5E F). As expected, both DBT-Luc-DGC/miRFMOD and DBT-Luc-DGC/miRNT cells showed mCherry expression after doxycycline treatment (Figure 2—figure supplement 5G, ). Both DBT-Luc-DGC/miRNT and DBT-Luc-DGC/miRFMOD (both in the absence and presence of doxycycline) cells reprogrammed to form neurospheres with equal efficiency (data not shown). We next explored the impact of FMOD silencing on the ability of U251 cells, an established human glioma cell line, to form neurospheres through reprogramming. U251 cells, which grow as differentiated monolayer cells (referred to here as U251-DGC) in an FBS-containing medium, can reprogram and form neurospheres enriched in CD133 expression (Tao et al., 2018). Both U251-DGC/shNT and U251-DGC/shFMOD cells could reprogram with equal efficiency, as measured by neurosphere formation and limiting dilution assays (Figure 2—figure supplement 6A–C). The neurospheres formed through reprogramming showed an upregulation of glioma reprogramming factors in U251-GSC/shNT and U251-GSC/shFMOD cells (Figure 2—figure supplement 6D). Collectively, these observations indicate that FMOD is not required for GSC or DGC growth and differentiation or reprogramming processes.
Integrins are a family of α/β heterodimeric transmembrane adhesion receptors. To identify the integrin α and β subunits that are involved in rhFMOD activation of integrin signaling in endothelial cells, we resorted to an unbiased approach where we analyzed the transcriptome data of microvessels isolated through laser capture microdissection (LCM) from human brain samples (Song et al., 2020). The transcript abundance of all integrins subunits α subunits (n = 19) and β subunits (n = 13) is shown in the following table. Three integrin subunits with maximum transcript abundance – ITGA6, ITGB1, and ITGAV – were chosen for testing. The ability of rhFMOD to induce pFAK levels in ST1 cells silenced for any of the above three integrins was investigated. The results show that all three integrins are essential for rhFMOD activation of integrin signaling in endothelial cells (Figure 4—figure supplement 3).
|High-abundant integrins in endothelial cells*|
|S. no.||Microvessel 1 FPKM (log2)||Microvessel 2 FPKM (log2)||Microvessel 3 FPKM (log2)||Average FPKM (log2)|
|*Song et al., 2020.|
The transcript data of FMOD, HES1, and JAG1 in different data sets were used for deriving (1) transcriptional upregulation in GBM over the control brain, (2) survival prediction by Kaplan–Meier analysis, and (3) correlation between transcripts (see following table for more details). We used a total of 1885 samples that included GBMs (n = 1833) and control brain samples (n = 52). Wherever the data is not available (NA), the specific analysis is not carried out. A p-value <0.05 is considered significant, with ****<0.0001, ***<0.001, **<0.01, and *<0.05. The nonsignificant data is denoted as ‘ns.’ The data is shown in Figure 4—figure supplement 6 and Figure 4—figure supplement 7.
|Expression, correlation, and survival of fibromodulin (FMOD), HES1, JAG1 in glioblastoma (GBM)|
|S. no||Datas et||Source||Control||Survival information||Control brain (n)||GBM samples (n)||FMOD: GBM vs. control||HES1: GBM vs. control||JAG1: GBM vs. control||FMOD survival Kaplan–Meier||FMOD to HES1 correlation||FMOD to JAG1 correlation|
|log2 fold change||p-Value||log2 fold change||p-Value||log2 fold change||p-Value||Hazard ratio||Low vs. high p-value||R-value||p-Value||R-value||p-Value|
|4||TCGA RNA Seq||TCGA||Yes||Yes||4||156||2.91||0.00||1.01||0.01||2.32||<0.000||1.31||0.13||0.22||0.01||0.05||0.58|
|5||CGGA RNA Seq||CGGA||Yes||Yes||20||105||12.94||0.01||9.62||0.01||NA||2.25||0.00||0.06||0.58||NA|
For survival prediction based on the methylation status of FMOD promoter, the samples were divided into methylation high (above median β value) and low (below median β value) for two CpG IDs, cg03764585 and cg04704856, derived from TCGA and GSE48461 datasets. The data is presented in Figure 4—figure supplement 7.
Label-free mass spectrometry data between the GSC and DGC showing protein ratios in the GSC and DGC secretome and p values are shown in Supplementary File 1 for proteins exhibiting significant differences in abundance in both conditions. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD032958.
EBI PRIDEID PXD032958. Functional genomics of glioblastoma: from epigenetics to proteomic investigation of tumor-initiating cell secretome.
Conversion of differentiated cancer cells into cancer stem-like cells in a glioblastoma model after primary chemotherapyCell Death and Differentiation 21:1119–1131.https://doi.org/10.1038/cdd.2014.31
The hypoxic microenvironment: A determinant of cancer stem cell evolutionBioEssays: News and Reviews in Molecular, Cellular and Developmental Biology 38:S65–S74.https://doi.org/10.1002/bies.201670911
Paracrine effects of stem cells in wound healing and cancer progression (ReviewInternational Journal of Oncology 44:1789–1798.https://doi.org/10.3892/ijo.2014.2385
Dedifferentiation of neurons and astrocytes by oncogenes can induce gliomas in miceScience (New York, N.Y.) 338:1080–1084.https://doi.org/10.1126/science.1226929
Notch signaling in vascular development and physiologyDevelopment (Cambridge, England) 134:2709–2718.https://doi.org/10.1242/dev.004184
Mechanisms of glioma-associated neovascularizationThe American Journal of Pathology 181:1126–1141.https://doi.org/10.1016/j.ajpath.2012.06.030
A GSK-3β inhibitor protects against radiation necrosis in mouse brainInternational Journal of Radiation Oncology, Biology, Physics 89:714–721.https://doi.org/10.1016/j.ijrobp.2014.04.018
Clinical and biological implications of CD133-positive and CD133-negative cells in glioblastomasLaboratory Investigation; a Journal of Technical Methods and Pathology 88:808–815.https://doi.org/10.1038/labinvest.2008.57
Notch signaling in developmental and tumor angiogenesisGenes & Cancer 2:1106–1116.https://doi.org/10.1177/1947601911423030
Generation of human pulmonary microvascular endothelial cell linesLaboratory Investigation; a Journal of Technical Methods and Pathology 81:1717–1727.https://doi.org/10.1038/labinvest.3780385
Small Leucine-Rich Proteoglycans in Skin Wound HealingFrontiers in Pharmacology 10:1649.https://doi.org/10.3389/fphar.2019.01649
The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidencesNucleic Acids Research 50:D543–D552.https://doi.org/10.1093/nar/gkab1038
Glioblastoma Stem Cells: Driving Resilience through ChaosTrends in Cancer 6:223–235.https://doi.org/10.1016/j.trecan.2020.01.009
Hypoxia-Mediated Epigenetic Regulation of Stemness in Brain Tumor CellsStem Cells (Dayton, Ohio) 35:1468–1478.https://doi.org/10.1002/stem.2621
Integrin-mediated signaling events in human endothelial cellsMolecular Biology of the Cell 9:1969–1980.https://doi.org/10.1091/mbc.9.8.1969
Transcriptomic comparison of human and mouse brain microvesselsScientific Reports 10:12358.https://doi.org/10.1038/s41598-020-69096-7
The functional role of Notch signaling in human gliomasNeuro-Oncology 12:199–211.https://doi.org/10.1093/neuonc/nop022
Notching on Cancer’s Door: Notch Signaling in Brain TumorsFrontiers in Oncology 4:341.https://doi.org/10.3389/fonc.2014.00341
Enhanced detection of CNS cell secretome in plasma protein-depleted cerebrospinal fluidJournal of Proteome Research 7:4409–4421.https://doi.org/10.1021/pr8003858
Glioblastoma stem-like cells secrete the pro-angiogenic VEGF-A factor in extracellular vesiclesJournal of Extracellular Vesicles 6:e1359479.https://doi.org/10.1080/20013078.2017.1359479
Tumorigenic potential of miR-18A* in glioma initiating cells requires NOTCH-1 signalingStem Cells (Dayton, Ohio) 31:1252–1265.https://doi.org/10.1002/stem.1373
CD133 negative glioma cells form tumors in nude rats and give rise to CD133 positive cellsInternational Journal of Cancer 122:761–768.https://doi.org/10.1002/ijc.23130
Endothelial cells promote stem-like phenotype of glioma cells through activating the Hedgehog pathwayThe Journal of Pathology 234:11–22.https://doi.org/10.1002/path.4349
Targeting cancer stem cell pathways for cancer therapySignal Transduction and Targeted Therapy 5:e8.https://doi.org/10.1038/s41392-020-0110-5
Notch Signaling Pathway Regulates Angiogenesis via Endothelial Cell in 3D Co-Culture ModelJournal of Cellular Physiology 232:1548–1558.https://doi.org/10.1002/jcp.25681
Fibromodulin Enhances Angiogenesis during Cutaneous Wound HealingPlastic and Reconstructive Surgery. Global Open 2:e275.https://doi.org/10.1097/GOX.0000000000000243
Caigang LiuReviewing Editor; Shengjing Hospital of China Medical University, China
Mone ZaidiSenior Editor; Icahn School of Medicine at Mount Sinai, United States
Caigang LiuReviewer; Shengjing Hospital of China Medical University, China
Herve ChneiweissReviewer; Sorbonne University INSERM CNRS, France
Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.
Decision letter after peer review:
Thank you for submitting your article "Non-cancer stem cell-derived Fibromodulin activates Integrin-dependent Notch signaling in endothelial cells to promote tumor angiogenesis and growth" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Caigang Liu as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Mone Zaidi as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Herve Chneiweiss (Reviewer #2).
The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.
1) Non-CSC as the key formulation needs to be specified and clearly clarified;
2) Statement regarding the biomarker of GSC in the Introduction part may require correction, and some important findings should be included as Reference;
3) Outline the reasons for the selection of the employed cell lines;
Reviewer #2 (Recommendations for the authors):
1) Be more precise on why you select a cell type and a given method. May be some results are not absolutely needed and removing them could ease reading the manuscript
2) Justify why the patient-derived cell lines were not mainly used along the work and why you used "circus" cell lines such as U87 or U251.https://doi.org/10.7554/eLife.78972.sa1
1) Non-CSC as the key formulation needs to be specified and clearly clarified;
We understand the possibility that the use of "non-cancer-stem cells (non-CSCs)" may create confusion as it may sound like referring to stromal cells of cancer that are non-cancer cells of the tumor. This was rightly pointed out by reviewer # 2. After obtaining advice from Dr. Caigang Liu, reviewing Editor, we decided to use "glioma stem-like cells" (GSCs) and "differentiated glioma cells" (DGCs) in place of CSCs and non-CSCs respectively.
We also considered the Reference (Qiao X, et al. Association of human breast cancer CD44-/CD24- cells with delayed distant metastasis. eLife. 2021 Jul 28;10:e65418) referred by Dr. Caigang Liu, reviewing Editor for formulating the non-CSCs. This is a good suggestion.
However, we thought it would be difficult to use this terminology as many GSC markers for glioma have been identified, resulting in heterogeneity within the GSC population.
2) Statement regarding the biomarker of GSC in the Introduction part may require correction, and some important findings should be included as Reference;
The statement regarding biomarker of GSC is addressed appropriately, taking the recommendations of Reviewer #2. Accordingly, references that reported the stemness properties of CD133- glioma cells (Beier et al., 2007; Chen et al., 2010; Joo et al., 2008; Ogden et al., 2008; Wang et al., 2008), involvement of miR302-367 cluster and miR18A* in the regulation of stemness (Fareh et al., 2012, Turchi et al., 2013), and the existence of stem cell-associated heterogeneity in GBM (Dirkse et al., 2019) are referred in the revised manuscript.
3) Outline the reasons for the selection of the employed cell lines;
1) We have used three patient-derived GSC lines (MGG4, MGG6, and MGG8) for the initial discovery and in vitro validation of the DGC-specific expression of FMOD.
Please refer to Main Figure 1.
2) We have used one human patient-derived GSC line (MGG8) and two murine GSC lines (AGR53 and DBT-Luc) in an intracranial orthotopic mouse model experiment to prove the importance of FMOD induced angiogenesis in tumor growth. While MGG8 line silenced for FMOD using short hairpin RNA (shRNA) was used to prove the importance of FMOD in tumor angiogenesis and growth (Please refer to Main Figure 6), AGR53 and DBT-Luc lines carrying doxycycline-inducible shRNAs proved the importance of FMOD secreted by de novo generated DGCs from a GSC initiated tumor in angiogenesis and growth.
Please refer to Main Figure 5 and Main Figure 5—figure supplement 1, respectively.
3) We have also used the murine GSC line (DBT-Luc-GSC) for co-implantation experiment Please refer to Main Figure 2.
4) The only mouse model experiment carried out using the U251 cell line is now moved to the supplementary section as these cell lines are highly unrepresentative since they are grown for a long time in serum conditions, as recommended by reviewer # 2
Please refer to Main Figure 6—figure supplement 1.
5) The other places where the established glioma cell lines, LN229, U251, and U87 used, are in the main figure 3 and its figure supplements. In these experiments, we used these cells only as a source of secreted FMOD to study the angiogenesis-inducing property. However, we would like to point out that conditioned media from MGG4, MGG6, and MGG8 also showed the angiogenesis-inducing property of FMOD.
Please refer to Main Figure 3 and its figure supplements.
[Editors' note: we include below the reviews that the authors received from another journal, along with the authors’ responses.]
We thank the reviewers for their constructive comments, based on which we have revised the manuscript. All questions have been appropriately addressed and additional data have been included when requested. Please find below point-to-point responses to the reviewers’ comments describing how they were addressed.
Answers to comments from Reviewer 1
1) Large GBM or therapy-resistant GBM is characterized by at least three forms of neovascularization, angiogenesis, vasculogenesis, and vascular mimicry. This manuscript only dealt with angiogenesis. GSC is known to be involved in pericyte development to support angiogenesis and vasculogenesis. Recent reports also indicate the involvement of GSC in making vascular mimicry. The authors should at least show the effect of FMOD on vascular mimicry in vitro or in vivo (I believe the authors already have tissues from different conditions and staining for VM markers will prove the relationship).
In the first version, we showed the important role played by FMOD on tumor angiogenesis as measured by the ability of endothelial cells to form angiogenic networks. In the revised manuscript, we extended our studies to vascular mimicry, another form of neovascularization wherein the tumor cells transdifferentiate to form endothelial cells (TDECs) and obtain more nutrients.
– in vitro, we demonstrate that MGG8-DGCs and U87 readily differentiated to form TDECS when grown in M199 media, which correlated with increased CD31 expression and decrease in FMOD expression. Both TDECs formed angiogenic networks more efficiently in the presence of rhFMOD. These data are illustrated on Figures 3G-J and Supplementary Figure 15. See page no. 9, lines 259 to 271 for more detail in the revised manuscript.
– in vivo, we demonstrate that large tumors formed by MGG8-GSC, AGR53-GSC, and DBTLuc-GSC in FMOD+ condition show the presence of TDECs, as evidenced by significant colocalization of CD31 and either GFAP (for MGG8) or GFP (for AGR53 and DBT-Luc) on blood vessels. This process was significantly reduced in small tumors formed by the above cells in FMOD-silenced conditions. These data are illustrated on Figures 7 G-L and Supplementary Figures 27G-I; see page no. 14, lines 422 to 428 for more detail in the revised manuscript.
2) Larger GBM and recurrent GBM release many cytokines and attractants that mobilize and accumulate BMDPC in the tumor and cause vasculogenesis. The author should investigate or discuss the effect of FMOD on vasculogenesis.
We appreciate the suggestion of the reviewer to investigate the role of BMDPC and vasculogenesis. We tried to examine this phenomenon by staining the tumor sections using antiCD34 (a BMDPC-specific endothelial cell marker). While we found colocalization of CD34 and CD31 in a few locations on the blood vessels (the data are shown in Author response image 1), we are not comfortable with the quality of the CD34 antibody available. Hence, we would that these results are not included in the manuscript and to consider this line of investigation in the future.
3) ST1 is an endothelial cell line that originated from the lungs. These are not a replica of GBM or brain associated microvascular derived EC. The author should prove that there are no differences between EC from the brain and lungs in respect of in vitro angiogenesis with or without FMOD.
To address this issue, we have used two endothelial cell lines originating from human and murine brain -primary human brain microvascular endothelial cells (HBMECs) and murine immortalized brain endothelial cells (B.End3), respectively. Our results show that CM derived from MGG8 DGC/shNT is more efficient to promote formation of angiogenic networks by both B.End3 and HBMECs than CM derived from MGG8 DGC/shFMOD cells (Supplementary Figure 14F; Figures 3E and F). The effect of FMOD silencing was rescued by adding rhFMOD exogenously to both cell types (Supplementary Figure 14F; Figures 3E and F). This indicates that the pro-angiogenic effect of FMOD is conserved across endothelial cells, irrespective of their tissue of origin, and validates our initial findings using the human pulmonary microvascular endothelial cells (ST1). See page no. 9, lines no. 253 to 258 for more detail in the revised manuscript.
4) If the development of GBM from GSC is related to DGC, then targeting GSC will not be the answer for therapeutic purposes. The authors should give a guideline on how to deal with the scenario in the clinical context. Antiangiogenic therapy is not fruitful, therefore, the guideline should emphasize breaking the link between DGC and GSC.
The present study demonstrates that in addition to GSCs, DGCs have an essential role in tumor growth and maintenance. While the therapy-resistant and self-renewing GSCs trigger the early events of transformation and growth, DGCs, the proportion of which continues to increase during tumor growth, progressively become essential. Thus, targeting both cancer stem cells and differentiated cancer bulk cells is crucial to achieve a durable response. The study also highlights the potential of GSC and DGC CM analysis to uncover novel targets in cancer therapy and the critical influence of DGC-secreted FMOD in glioma tumor growth. This aspect is discussed in the Discussion section. See page no. 17, lines no. 504 to 510 for more detail in the revised manuscript.
Answer to comments from Reviewer 2
1) The manuscript is very hard to follow in places and takes leaps of logic. It would be far more helpful if the authors would focus on generating data to support a coherent story that is logically presented.
To address this issue, we have modified the manuscript to make it more coherent. The experiments showing that FMOD is not required for GSC growth and differentiation as well as DGC growth and reprogramming have been moved to supplementary information. These data are now illustrated on supplementary figures 6-11.
2) One issue is that the proteomics study seems a bit too easy that the authors happened to identify a protein that they have already studied. It would be better to do a secondary screen or some other prioritization.
The choice of FMOD from the proteomic data was motivated by the objective to identify proteins preferentially secreted by DGCs involved in ECM remodelling. This is the reason why we considered proteins enriched in the DGC secretome, compared to the GSC secretome, and annotated to be ECM proteoglycans (see the Venn diagram on Figure 1B). Among the six proteins exhibiting these characteristics, FMOD was the most enriched in the DGC secretome, compared to the GSC secretome. These results and our earlier work on FMOD led us to choose FMOD to further investigate its role in tumor formation by inducing ECM remodelling and associated phenotypes such as angiogenesis.
See page no. 5 and 6, lines no 148 to 165 for more detail in the revised manuscript.
3) One major concern is that GSC and DGC are cultured differently. It is not clear to me that FMOD is strictly connected to differentiated cells, but rather adherent cells grown in serum, which can induce ECM expression. The in vivo studies in Figure 2 are the best evidence but should be better developed.
Several lines of evidence support that “specific” FMOD expression in DGCs (compared with GSCs) is not due to specific culture conditions (grown as adherent cells in serum-containing medium):
1) Our previously published study (Mondal et al., 2017, Oncogene, 36, 71–83) showed a highly variable pattern of FMOD expression by several established adherent GBM cultures even though all cell lines used in the study were grown as adherent cultures in serum-containing medium. U251, U87, and U343 cells show very high FMOD expression while others like LN229 and LN18 cells show minimal or no detectable FMOD expression. This observation suggests that FMOD expression is not related to specific culture conditions such as the use of adherent cells and serum-containing media.
2) Both GSCs and DGCs (MGG8) are grown as a monolayer on Geltrex-coated cell culture dishes. In this condition, we show that (see Author response image 2 for detail):
a) FMOD RNA and protein (intracellular and secreted) are still more abundant in DGCs than in GSCs (see panels A and B of Author response image 2).
b) RNA level of four reprogramming factors are upregulated in GSCs, compared with DGCs (see panel C of Author response image 2).
c)Confocal staining of SSEA1, SOX2, and CD133 show that they are upregulated in GSCs, while GFAP and FMOD are upregulated in DGCs (see panel D of Author response image 2).
d) FACS analysis of CD133 expression by GSCs and DGCs shows an enrichment of CD133 positive-cells in the GSC monolayer but not in the DGCs (see panel E of Author response image 2).
Our results show that even when GSCs and DGCs are grown in adherent conditions, DGCs express higher levels of FMOD than GSCs. Thus, the DGC-specific expression of FMOD compared to GSC is not related to adherent growth conditions.
3) We also demonstrate that in FMOD+ tumors (in all three orthotopic tumors using AGR53, DBT-Luc, and MGG8 GSCs), FMOD colocalizes with GFAP, further supporting that preferential FMOD expression and secretion by DGCs takes place in vivo.
These data are illustrated on Supplementary Figures 24, 25, and 26;. See page no. 14, lines no. 405 to 414 for more detail in the revised manuscript.
4) FMOD has been strongly connected to TGF-β signaling (and invasion and angiogenesis). It is not clear from the studies how the proposed signaling integrates into the prior modeling of protein function.
As suggested by the reviewer, we firmly believe that FMOD expression in DGCs is under the control of TGF-β signaling. In the revised manuscript, the following aspects have been addressed:
1) GSEA analysis of DEGs (differentially regulated genes) in GSCs over DGCs (Suva et al., 2014, 157, 580-594, Cell) showed a significant depletion of many gene sets associated to TGF-β signaling (Supplementary Figure 2).
2) GSEA analysis of GBM DEGs (GBM/Control) showed a significant enrichment of many gene sets associated to TGF-β signaling (Supplementary Figure 3)
3) FMOD and TGM2 RNA levels are more elevated in DGCs than in GSCs, and decreased upon treatment of cells with the TGF-β inhibitor (SB431542) (Figure 1G)
4) FMOD (intracellular and secreted, Figure 1H) and pSMAD (Figure 1I) levels are more elevated in DGCs than in GSCs, and decreased upon treatment of cells with the TGF-β inhibitor SB431542
5) TGF-β1-induced luciferase activities of SBE-Luc and FMOD Promoter-Luc reporters are higher in MGG8-DGCs than in MGG8-GSCs, and are inhibited upon SB431542 treatment (Supplementary Figures 5C, D).
6) pSMAD2 occupancy in FMOD promoter is higher in MGG8-DGCs than in MGG8GSCs and is reduced by SB431542treatment (Figure 1J)
All the above data are presented in pages no. 6 and 7, lines 166-188 in the revised manuscript.
5) The authors bounce between cell lines. I am not clear why the authors use MGC lines for the discovery then use other cell types for all the experiments. The rationale for using LN229 and U251 is particularly weak, as these cells have been extensively cultured in serum. I would drop the studies with these lines or relegate them solely to supplemental figures.
In the revised manuscript, we have included experiments in an intracranial orthotopic tumor mouse model using one human GSC cell-line (MGG8; Figures 6 A-D), and a new murine glioma cell line (DBT-Luc; Supplementary Figure 23) in addition to the previously used murine (AGR53; Figures 5A-G) and human glioma cell lines (U251; Figure 6E –I). This data is described in page no. 12-14, lines no 363 to 402 in the revised manuscript.
6) To address the concerns about culture conditions complicating interpretation of comparing GSC and DGC, I would suggest performing more studies with cell sorting and comparison with the same culture conditions. MGC have been cultured for prolonged periods and may resemble cell lines.
We agree in principle with the reviewer. However, we cultured GSCs as neurospheres in neural stem cell media supplemented with growth factors so that the enrichment in CD133-positive cells of the neurospheres is maintained as reported by Suva et al., Cell 157, 580-594, 2014. The enrichment of CD133-positive cells was verified at regular intervals of culturing.
Further, we now show that the MGG8 GSC neurospheres used in our experiments (Figure 6) are indeed enriched in CD133-positive cells. This data is described in page no. 13, lines no 389 to 391 in the revised manuscript.
Moreover, the “DGC-specific” expression of FMOD was demonstrated when both GSCs and DGCs are grown as monolayers in Geltrex-coated cell culture dishes. More detail is given in answer to comment no # 2 of the reviewer
7) The studies only use single knockdown approaches (sometimes siRNA, sometimes shRNA, etc.). It would be helpful to demonstrate better controls and rescue.
In the previous version of the manuscript, siRNA was used to silence FMOD expression in U251 cells while shRNA was used in MGG8-DGCs for the purpose of collecting conditioned media to set up in vitro angiogenesis experiments. In the revised manuscript, silencing of FMOD was achieved by using shRNA in U251 cells and siRNA in MGG8-DGCs. (Supplementary Figures 13 B and E). This data is described in page no. 8, lines no 235 to 238, and page no. 9, lines no. 248-251 in the revised manuscript.
We had already illustrated the “rescue” experiments in U251/siFMOD and MGG8DGC/shFMOD cells (see Figures 3C, D) in the first version of the manuscript. In the revised manuscript, additional rescue experiments were carried out in human brain-derived primary endothelial cells (HBMECs) and mouse brain-derived immortalized endothelial cells (bEnd.3) (see Figure 3 E and F; Supplementary Figure 14F). In all these cellular modles, the reduced angiogenic network formation in presence of CM from FMOD-silenced cells is rescued by the exogenous addition of rhFMOD. This data is described in pages no. 8 and 9, lines no 224 to 257 in the revised manuscript.
8) I would strongly suggest consideration of FMOD analysis in the absence of culture. They should consider double staining of tumor specimens for stem/differentiation markers and FMOD. Also, FMOD expression should be near blood vessels and pFAK and Hes1.
To address this question, we used tissue samples from various intracranial tumor models (human and murine) to show that FMOD is indeed colocalized with GFAP (used as differentiation marker), to validate our initial finding that FMOD is “solely” secreted by the DGCs. These new data are illustrated on Supplementary Figures 24-26, described in page no. 14, lines no. 403 to 414 in the revised manuscript.
We also show a co-localization of CD31 and FMOD, CD31 and pFAK, CD31 and JAG1, and CD31 and HES1 in all three tumor models (Supplementary Figures 28A-C). This data is described in pages no. 14 and 15, lines no. 428 to 434 in the revised manuscript.
9) It would be helpful to consider the choice of lines with the studies of FMOD. FMOD appears to be lower in proneural tumors, which is not surprising due to the association with TGF-β. It would be important to consider the subtype of all the lines used.
We agree with the reviewer that FMOD expression is not associated with proneural tumors. We show in the revised manuscript that the expression of FMOD and the enrichment of TGFβ signaling gene set are more pronounced in mesenchymal subtype GBM (Supplementary Figures 5A and B). We also show that DGCs used in this study, which exhibit high FMOD expression, are also of the mesenchymal subtype, with an enrichment of genes of the TGF-β signaling pathway (Supplementary Figure 4). This data is described in pages no. 6 and 7, lines no. 166 to 188 in the revised manuscript.
10) ST1 cells are not ideal as endothelial cells, as they have been immortalized, which may change the signaling.
To address this issue, we have used two endothelial cells lines of brain origin, primary human brain microvascular endothelial cells (HBMECs) and murine immortalized brain endothelial cells (B.End3). Our results show that CM of MGG8-DGC/shNT cells is more efficient to promote angiogenic network formation by bEnd.3 and HBMECs than CM of MGG8DGC/shFMOD cells (Figures 3E and F; Supplementary Figure 14F), reminiscent of our initial observations using ST1 cells. Again, the effect of FMOD silencing was rescued by adding rhFMOD to both cell types (Figures 3 E and F; Supplementary Figure 14F). This indicates that the pro-angiogenic effect of FMOD is conserved across endothelial cells, irrespective of their tissue of origin, and confirms the validity of our initial findings using the human pulmonary microvascular endothelial cells (ST1). See page no. 9, lines no. 253 to 257 for more detail in the revised manuscript.
11) It would seem that the upstream regulation of FMOD should be better investigated to understand why DGC, not GSC, express FMOD
As suggested by the reviewer, we carried out several experiments to confirm that FMOD expression in DGCs is under the control of TGFβ signalling. The detail is provided in the answer to comment #3 above.
12) The key experiments would seem to be determining if over expression of FMOD in GSC could replace DGC with in vivo growth. This is the more important finding. I was particularly concerned with the design of Figure 6. The authors assume that the effects of FMOD are due to the DGC, but there is no direct evidence to this effect. I was not even sure why there was not greater study of the xenografts for expression of FMOD and the other signaling molecules.
While we show that FMOD is an important mediator of angiogenesis induced by DGCs, thereby contributing to tumor growth, DGCs may also influence tumor growth through FMODindependent mechanisms. For example, Wang et al. showed a reciprocal signaling between GSCs and DGCs in promoting tumor growth whereby DGC-secreted BDNF acts on NTRK2 receptors on the GSC membrane to activate self-renewal and activation of GSCs. (Wang et al., Cell Stem Cell, 22, 514-528, 2018). Hence, it is unlikely that overexpressing FMOD in GSCs would replace DGCs in tumor growth in vivo.
To address the reviewer’s other concerns, we utilized tissue samples from our various intracranial tumor models (human and murine) to show that FMOD is colocalized with GFAP (used as a differentiation marker) and thus validating our initial finding that FMOD is “solely” secreted by the DGCs in vitro. These data are illustrated on Supplementary Figures 24-26. We further demonstrate that the small tumors formed in FMOD-silenced conditions show GFAP staining, suggesting that differentiation has indeed occurred in the absence of FMOD, which supports our in vitro data indicating that FMOD is not needed for GSC differentiation to DGC.
Further analysis showed reduced angiogenesis in the small tumors formed in FMOD-silenced conditions (Figures 7A-L; Supplementary Figures 27 A-I). From these results, we conclude that compromised angiogenesis is indeed the main reason why tumor growth is reduced in FMOD-silenced condition. We also show the co-localization of CD31 and FMOD, CD31 and pFAK, CD31 and JAG1, and CD31 and HES1 in all three tumor models used in our study, which confirms (in addition to in vitro data) the activation of integrin-Notch signaling in FMOD-treated endothelial cells (Supplementary Figures 28 A-C). See pages no. 14 and 15, lines no. 403 to 434 for more detail in the revised manuscript
13) FMOD appears to be prognostic. This has been reported by the authors but could be better developed.
We have obtained the following data highlighting the prognostic significance of FMOD:
1) Significant upregulation of FMOD, JAG1, and HES1 in GBM in 5 data sets (Supplementary Figure 21 A-C)
2) Correlation between FMOD and HES1 transcripts and between FMOD and JAG1 transcripts in 8 data sets (Supplementary Figures 21 D and E)
3) Association of high FMOD transcript levels with poor prognosis in 5 data sets (Supplementary Figure 22 A)
4) Association of FMOD promoter hypomethylation with poor prognosis in 4 data sets (Supplementary Figure 22 B).
See page no. 12, lines no. 353 to 362 for more detail in the revised manuscript.
14) With FMOD already being reported to be connected to invasion, I was not sure why this was not examined. Also, the effects of FMOD on the immune system should be considered.
To address this issue, we performed invasion assays of ST1 cells treated with BSA vs. rhFMOD or CM of either MGG8-DGC/shNT or MGG8-DGC/shFMOD cells. We show that the presence of FMOD favors invasion of the ST1 cells (Supplementary Figures 14 C and D). See page no. 9, lines no. 251 to 253 for more detail in the revised manuscript.
We agree with the reviewer that investigating the effect of FMOD on the immune system would be of great interest and certainly warrants further exploration. However, we have been successful in addressing almost all comments by producing an important amount of new data both in vitro and in vivo. We would like to address this point in the future. We hope that the reviewer will share our opinion that it is beyond the main scoop of the current article.
15) The summary figure appears premature, as there are a number of molecules that have not been tested and the design of the Notch and KAK studies were not ideal. The authors should have engineered endothelial cells to have activated Notch and FAK to see if this rescues the effects of loss of FMOD in DGC.
To address the reviewer’s concerns, we carried out the following experiments in the revised manuscript.
1) We showed that αVβ1 and α6β1integrins play an important role in rhFMOD-mediated induction of pFAK levels (Supplementary Figures 18 D-I).
2) Pre-treatment of ST1 cells with FAK inhibitor or a Src inhibitor (PP2) reduced significantly the rhFMOD-mediated activation of CSL-Luc and HES-Luc (Supplementary Figures 19 AD), and induction of HES1 transcript and protein in ST1 cells (Supplementary Figure 19 E-
3) Pre-treatment of ST1 cells with FAK inhibitor or PP2 (a Src inhibitor) significantly reduced the activation of JAG1 by rhFMOD (Supplementary Figures 20 B and C).
4) Silencing JAG1 in endothelial cells inhibited the ability of rhFMOD to activate CSL-Luc and HES-Luc (Supplementary Figures 20 D and E), increase HES1 transcript and protein levels, and induce angiogenic network formation by the ST1 cells (Figures 4 I-L).
5) We have identified KLF8 as the FAK-activated transcription factor responsible for JAG1 activation in rhFMOD-treated endothelial cells (Supplementary Figures 20 G and H).
6) We have also engineered ST1 cells exhibiting constitutively active Notch signaling. We stably expressed the Notch-intracellular domain (NICD) or the Vector Control (VC) in ST1 (ST1/NICD, ST1/VC) cells. (Supplementary Figure 17A and B). ST1/NICD cells showed higher capacity to form angiogenic networks by themselves, compared with ST1/VC cells (Supplementary Figures 17C and D). ST1 cells were then be subjected to angiogenesis network formation assay in the presence or absence MGG8-DGC/shNT vs. MGG8DGC/shFMOD CM. We observed that the lack of FMOD in MGG8-DGC/shFMOD CM caused a significant decrease in network formation by ST1/VC cells but not by ST1/NICD cells (Supplementary Figures 17 C and D). These results confirm that Notch activation is an essential step in FMOD-induced angiogenesis. See pages no. 9-12, lines no. 272 to 362 for more detail in the revised manuscript.
7) We also show co-localization of CD31 with FMOD, pFAK, JAG1 and HES1 in the three tumor models used in our study, which confirms (in addition to in vitro data) the activation for integrin-Notch signalling in FMOD-treated endothelial cells (Supplementary Figures 28 AC). See page no. 14, lines no. 428 to 434 for more detail in the revised manuscript.
1) The data presentation is not professional. The figures do not comply to Cell journal standards.
Efforts were made to make appropriate changes in the Figures and to improve their readability.
2) Better statistical testing with correction for multiple comparisons should be performed.
We have now systematically performed ANOVA in experiments where multiple samples were compared.
Answer to comments from Reviewer 3
1) The authors use 10% FBS to induce GSC differentiation into DGC, and the corresponding two types of conditioned medium were harvested and analyzed by mass spectrometry. It is unclear if these two cell types were cultured in same medium (stem cell medium or normal cancer cell medium?). Furthermore, FMOD was identified to be enriched in DGC-conditioned medium, which could be simply due to the effects of floating (cancer stem cells) vs adherent cells (differentiated cells). As this is the foundation of the whole study, the authors need to verify it in a well-controlled system, e.g., glioma stem cells (CD133+) vs non-stem glioma cells (CD133-), both cultured as neuro spheres, from the same patients. Since many datasets of global gene expression in GSCs and DGCs have been published, the authors could easily verify the specific mRNA expression of FMOD in DGCs.
Overall, we agree with the reviewer. However, we cultured GSCs as neurospheres in neural stem cell media supplemented with growth factors so that the enrichment of CD133-positive cells in neurospheres in was maintained (see Suva et al., Cell 157, 580-594, 2014). We also verify this point at regular intervals of culturing and show that the MGG8-GSC neurospheres used in our experiments (Figure 6A) are indeed enriched in CD133-positive cells.
Moreover, the DGC-specific expression of FMOD is also shown when both GSCs and DGCs are grown as monolayers in Geltrex-coated cell culture dishes. More detail is given in the answer to comment #2 of Reviewer #2.
2) Figure 3. The in vitro angiogenesis analysis is weak, solely depending on tube formation assay. The results need to be further verified by EC proliferation and migration assays.
To address the reviewer’s comment, we have included a number of new data:
1) We show that FMOD (rhFMOD or FMOD coming from MGG8-DGC/shNT CM) induces migration, invasion, but not proliferation of ST1 cells (Supplementary Figures 14A-E). See page no. 9, lines no. 249 to 251 for more detail in the revised manuscript.
2) Using two additional endothelial cell lines of brain origin (namely primary human brain microvascular endothelial cells (HBMECs) and murine immortalized brain endothelial cells (b.End3), we show that CM from MGG8-DGC/shNT cells promotes formation of angiogenic networks by HBMECs (Figures 3E and F) and B.End.3 (Supplementary Figure 14F) more efficiently than CM from MGG8-DGC/shFMOD cells. The effect of FMOD silencing was rescued by adding rhFMOD to both cell types (Figures 3 E and F; Supplementary Figure 14F). This indicates that the pro-angiogenic effect of FMOD is conserved across endothelial cells, irrespective of their tissue of origin, and confirms the validity of our initial findings using human pulmonary microvascular endothelial cells (ST1). See page no. 9, lines no. 253 to 258 for more detail in the revised manuscript.
3) The small size tumors formed by AGR53-GSC/miRFMOD cells after doxycycline treatment showed reduced staining for CD31 and vWF (von Willebrand factor) compared to tumors formed by AGR53-GSC/miRNT cells (Figures 7A-C; Supplementary Figures 27 A-C). Likewise, tumors formed by DBT-Luc GSC/miRFMOD cells in doxycycline-treated mice showed significantly reduced CD31 staining, compared with those developed by DBT-Luc GSC/miRFMOD cells in the absence of doxycycline treatment (Supplementary Figures 27 D-F). Like murine glioma tumors, tumors developed by MGG8-GSC/shFMOD cells showed reduced CD31 staining compared to tumors formed by MGG8-GSC/shNT cells (Figures 7 DF). We also tested the extent of blood vessel formation by TDECs (Tumor-derived endothelial cells). In all the three tumor models used, blood vessels formed by TDECs were significantly reduced in tumors formed in FMOD-silenced condition (Figures 7G-L; Supplementary Figures 27 G- I). See page no. 14, lines no. 414 to 428 for more detail in the revised manuscript.
These findings confirm our results in vitro, where the absence of FMOD decreased the ability of host-derived endothelial cells and TDECs to form angiogenic networks (Figure 3; Supplementary Figures 13-15). From these results, we conclude that angiogenesis induced by DGCsecreted FMOD is essential for glioma tumor growth.
4) We also show the co-localization of CD31 and FMOD, pFAK, JAG1, and HES1 in all the three tumor models used, which confirms (in addition to in vitro data) the activation of integrin-Notch signaling in FMOD-treated endothelial cells (Supplementary Figures 28 A-C). See page no. 14, lines no. 428 to 434 for more detail in the revised manuscript.
3) The mechanism study is relatively weak. The authors should identify the integrin that interacts with FMOD and test its role in FMOD-induced angiogenesis. To support the proposed model (Figure 7), the authors may define the precise mechanism by which FMOD induces FAK/Src activation and Jag1 expression, such as assays for dependency of Src/FAK on Jag1 expression, transcription factor/promoter binding, and transcriptional activation.
To address the reviewer’s comment, the following data have been included in the revised manuscript.
1) With respect to integrin signaling, we show that ITGA6, ITGAV and ITGB1 are highly expressed in endothelial cells (See supplementary information); Silencing experiments proved that all three integrins are important for the increase in pFAK levels induced by rhFMOD. This suggests the involvement of αVβ1 and α6β1 integrins in rhFMOD effects in endothelial cells (Supplementary Figures 18 D-I).
2) Pre-treatment of ST1 cells with FAK inhibitor or PP2 (Src inhibitor) significantly reduced the rhFMOD-mediated activation of CSL-Luc and HES-Luc (Supplementary Figures 19 AD), as well as induction of HES1 transcript and protein in ST1 cells (Supplementary Figures 19 E-H).
3) Pre-treatment of ST1 cells with FAK inhibitor or PP2 reduced the activation of JAG1 by rhFMOD (Supplementary Figures 20 B and C).
4) Silencing JAG1 in endothelial cells inhibited the ability of rhFMOD to activate CSL-Luc and HES-Luc (Supplementary Figures 20 D and E), induce HES1 transcript and protein, and induce angiogenic network formation by the ST1 cells (Figures 4 I-L).
5) We identified KLF8 as the FAK-activated transcription factor responsible for JAG1 activation in rhFMOD-treated endothelial cells (Supplementary Figures 20 G and H).
6) We engineered ST1 cells exhibiting constitutively active Notch signaling. We stably expressed the Notch-intracellular domain (NICD) or the Vector Control (VC) in ST1 (ST1/NICD, ST1/VC) cells (Supplementary Figures 17A and B). ST1/NICD cells showed higher capacity to form angiogenic networks by themselves, compared with ST1/VC cells (Supplementary Figures 17C and D). ST1 cells were then be subjected to in vitro angiogenesis network formation assay in the presence or absence of MGG8-DGC/shNT or MGG8-DGC/shFMOD CM. We observed that the lack of FMOD in MGG8-DGC/shFMOD CM caused a significant decrease in network formation by ST1/VC cells but not by ST1/NICD cells (Supplementary Figures 17 C and D). These results confirm that Notch activation is an essential step in FMODinduced angiogenesis. See pages no. 9-12, lines no 272 to 362 for more detail in the revised manuscript.
4) The authors should validate their proposed molecular mechanism in the in vivo study, such as FMOD-induced and endothelial Jak1 expression in tumor ECs.
We show a co-localization of CD31 and FMOD, pFAK, JAG1, and HES1 in all three tumor models used in the studies, which confirms (in addition to in vitro data) the activation for integrin-Notch signaling in FMOD-treated endothelial cells (Supplementary Figures 28 A-C).
See page no. 14, lines no. 428 to 434 for more detail in the revised manuscript.
1. Numerous mistakes/typos in references, e.g., Wang et al., 2918 (from future?) and Mondal et al., 2017 (could be seen in the ref list) in Page 6 and etc.
We carefully proofread the entire manuscript (including references) and corrected the mistakes.
2. Figure S1. The shRNA knockdown efficiency needs to be shown.
The knockdown of FMOD in MGG8 GSCs and DGCs is shown in the revised manuscript (Supplementary Figure 6A). In addition, FMOD-silencing in tumor models is shown on Figures 5 and 6, and Supplementary Figures 23-26.
3) Figure S2.D.E. The results of FMOD knockdown were not consistent in S2D and S2E, or the lanes were wrongly labeled.
This mistake has been rectified and shown in Supplementary Figure 7 of the revised manuscript.
4) Figure S5. Considering several patient-derived GSCs are available in the authors' laboratory, why authors use U251 and mouse GSCs to test the role of FMOD in stemness and differentiation?
In the revised manuscript, we have included experiments in an intracranial orthotopic tumor model in the mouse using one human GSC cell-line (MGG8; Figures 6 A-D), and an additional murine glioma cell line (DBT-Luc; Supplementary Figure 23) in addition to existing experiments using murine (AGR53; Figures 5A-G) and human (U251; Figures 6E–I) glioma cell lines. This data is described in pages no. 12-14, lines no. 363 to 402.
5) Figure S10. The in vivo angiogenesis data is important for the whole study, and, therefore, is suggested to move to main figure. Furthermore, the main findings need to validate by additional assays, such as vessel perfusion or tumor hypoxia. In addition, the changes in representative images, but not in the quantified data, are not impressive. Additional staining with CD31 immunofluorescence would strengthen the results.
To address this concern, we have carried out the following experiments in the revised manuscript.
1) We have done additional angiogenesis-related experiments in vivo and the results are illustrated on Figure 7 and Supplementary Figure 27.
2) In the revised manuscript, we also demonstrate that FMOD induces angiogenic network formation by tumor-derived endothelial cells (TDECs) both in vitro and in vivo.
in vitro, we demonstrate that MGG8-DGCs and U87 cells readily differentiate to form endothelial cells (tumor-derived endothelial cells; TDECs) when grown in M199 media, which correlated with increased CD31 expression and decreased FMOD expression. In both cases, TDECs were induced to form angiogenic networks by the presence of rhFMOD. These new data are presented on Figures 3G-J and Supplementary Figure 15.
in vivo, we demonstrate that large tumors formed by MGG8-GSC, AGR53-GSC, and DBTLuc-GSC in FMOD+ condition show the presence of TDECs, as evidenced by significant colocalization of CD31 and GFAP (for MGG8), GFP (for AGR53 and DBT-Luc) on the blood vessels. This was significantly reduced in small tumors formed by the above cells in FMODsilenced conditions (Figures 7 G-L; Supplementary Figures 27 G-I).https://doi.org/10.7554/eLife.78972.sa2
- Kumaravel Somasundaram
- Kumaravel Somasundaram
- Kumaravel Somasundaram
- Philippe Marin
- Dinorah Friedmann-Morvinski
- Philippe Marin
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
The results published here are in whole or part, based upon data generated by The Cancer Genome Atlas pilot project established by the NCI and NHGRI. Information about TCGA and the investigators and institutions that constitute the TCGA research network can be found at http://cancergenome.nih.gov/. We acknowledge the shRNA consortium (Dr. Subba Rao), IISc, India, for shRNA constructs. The Central Animal Facility (CAF), Indian Institute of Science is acknowledged for all animal experiments. SS acknowledges IISc for fellowship. KS acknowledges CEFIPRA DBT, DST, ICMR, and CSIR (Govt. of India) for research grants. Infrastructure supported by DST FIST, DBT-IISc partnership program, and UGC is acknowledged. KS is awarded J. C. Bose Fellowship from DST. This research was supported by grants to DFM from the Israel Science Foundation (1315/15 and 1429/20). PM was supported by grants from Fondation pour la Recherche Médicale and CEFIPRA (IFC/5603-C/2016/503). Mass spectrometry experiments were carried out using the facilities of the Montpellier Proteomics Platform (PPM, BioCampus Montpellier).
The Institute Ethical Committee for Animal Experimentation (Institute Animal Ethics Committee [IAEC] Project Number: CAF/Ethics/752/2020).
- Mone Zaidi, Icahn School of Medicine at Mount Sinai, United States
- Caigang Liu, Shengjing Hospital of China Medical University, China
- Caigang Liu, Shengjing Hospital of China Medical University, China
- Herve Chneiweiss, Sorbonne University INSERM CNRS, France
© 2022, Sengupta et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Sequencing of cell-free DNA (cfDNA) is currently being used to detect cancer by searching both for mutational and non-mutational alterations. Recent work has shown that the length distribution of cfDNA fragments from a cancer patient can inform tumor load and type. Here, we propose non-negative matrix factorization (NMF) of fragment length distributions as a novel and completely unsupervised method for studying fragment length patterns in cfDNA. Using shallow whole-genome sequencing (sWGS) of cfDNA from a cohort of patients with metastatic castration-resistant prostate cancer (mCRPC), we demonstrate how NMF accurately infers the true tumor fragment length distribution as an NMF component - and that the sample weights of this component correlate with ctDNA levels (r=0.75). We further demonstrate how using several NMF components enables accurate cancer detection on data from various early stage cancers (AUC = 0.96). Finally, we show that NMF, when applied across genomic regions, can be used to discover fragment length signatures associated with open chromatin.
The transformation of normal to malignant cells is accompanied by substantial changes in gene expression programs through diverse mechanisms. Here, we examined the changes in the landscape of transcription start sites and alternative promoter (AP) usage and their impact on the translatome in TCL1-driven chronic lymphocytic leukemia (CLL). Our findings revealed a marked elevation of APs in CLL B cells from Eµ-Tcl1 transgenic mice, which are particularly enriched with intra-genic promoters that generate N-terminally truncated or modified proteins. Intra-genic promoter activation is mediated by (1) loss of function of ‘closed chromatin’ epigenetic regulators due to the generation of inactive N-terminally modified isoforms or reduced expression; (2) upregulation of transcription factors, including c-Myc, targeting the intra-genic promoters and their associated enhancers. Exogenous expression of Tcl1 in MEFs is sufficient to induce intra-genic promoters of epigenetic regulators and promote c-Myc expression. We further found a dramatic translation downregulation of transcripts bearing CNY cap-proximal trinucleotides, reminiscent of cells undergoing metabolic stress. These findings uncovered the role of Tcl1 oncogenic function in altering promoter usage and mRNA translation in leukemogenesis.