2.1 CAF-mediated ECM remodeling

Within the tumor microenvironment, fibroblasts are the predominant regulator of the ECM, and are responsible for secreting, crosslinking, and remodeling structural proteins such as collagens, fibronectin, and laminins (Figure 2A) [23]. Apart from providing structure and stability within the TME, these structural molecules can bind many types of extracellular-sensing receptors on cancer cells to regulate their stemness [24]. For example, CAFs are known to secrete laminins such as laminin-332, which induces a stem-like phenotype in hepatocellular carcinoma (HCC), promoting quiescence and chemoresistance in an mTORC2-dependent manner [25,26]. Other laminins interact with integrins, such as integrin α6β4, on cancer cells to activate downstream Notch signaling, and promote cell proliferation and migration [27]. Similarly, activation of integrin ITGA2 on cancer cells upon interaction with collagen I promotes tumorigenesis and sphere formation in oral squamous cell carcinoma (OSCC) [28].

Other ECM components, such as fibronectin, interact with integrin αvβ3 on cancer cells to activate PI3K/Akt/SOX2 and CDC42/YAP-1/NUPR1/Nestin axes, promoting a stem-like phenotype in glioblastoma (GBM) [29]. Similarly, Wu et al. demonstrate that CAF-secreted fibronectin activates PI3K/Akt/SOX2 signaling in non-small cell lung cancer (NSCLC), and that inhibition of αvβ3 can sensitize cancer cells to chemotherapy [30]. Fibronectin also activates Wnt/β-catenin and MAPK/ERK signaling within NSCLC cells to promote angiogenesis and invasion in a WISP3-dependent manner [31]. Additionally, fibronectin-integrin interaction can upregulate the epithelial-mesenchymal transition (EMT) markers snail, N-cadherin, and vimentin [32], while fibronectin can promote cell migration, invasion, and matrix metalloproteinase 2 (MMP-2) expression in breast cancer [33]. CAF-derived MMPs are responsible for remodeling tumoral extracellular matrix, promoting angiogenesis, and degrading basement membrane to help promote EMT and invasion [34,35]. Given that MMP-7 has been shown to promote chemoresistance in numerous types of CSCs by mediating MUC-1 shedding [36], CAF-derived MMPs may also play a role in regulating stemness independent of canonical MMP-EMT pathways.

CAFs have been shown to secrete other structural molecules such as the polysaccharide hyaluronic acid (HA) in response to TGF-β [37]. In breast cancer, HA activates the stemness-related pathways Ras-MAPK and PI3K/Akt by binding to CD44, a widely acknowledged marker for CSCs [38,39]. Similarly, in head and neck squamous cell carcinoma (HNSCC), HA/CD44 mediated signaling enhances SOX2 expression through the PI3K/4EBP1 axis [40]. Apart from secreting structural proteins, CAFs can also regulate ECM stiffness by lysyl oxidase-mediated crosslinking of the existing ECM [41]. In HCC, matrix stiffening in 3D hydrogel culture is associated with increased expression of stemness markers CD44, EpCAM, and Nanog [42], while matrix stiffening promotes EMT in lung cancer in a DDR2-dependent manner [43]. Similarly, in colorectal cancer (CRC) cells, matrix stiffness is associated with increased CSC markers like CD133, ALDH1, and LGR-5, and knockdown of the Hippo transcription factor, Yes-associated protein (YAP), could reverse this effect and reduce sphere formation in vitro [44]. Together, these findings suggest that the ECM serves as more than a passive scaffold within the TME; the ECM is dynamically deposited and remodeled by CAFs to regulate cancer cell stemness via structural proteins and ECM-sensing receptors.

2.2 CAF-secreted growth factors

Apart from regulating the ECM, fibroblasts also secrete a variety of soluble factors to influence cancer cell migration, proliferation, and stemness (Figure 2B) [45]. Breast cancer-associated fibroblasts secrete fibroblast growth factor 5 (FGF5) to upregulate stemness markers ID3 and SOX10, and enhance sphere formation in vitro [46]. Similarly, breast cancer CAFs express high FGF2, and CAF-derived conditioned media promotes cancer cell proliferation in vivo, while knockdown of FGF2 in CAFs abrogated this effect [47]. Mechanistically, CAF FGF2 secretion is activated by estrogen via GPER/EGFR/ERK signaling [48]. Given that FGF2 helps maintain an undifferentiated state in mesenchymal stem cells marked by high expression of Oct4, SOX2, and Nanog [49], is also reported to convert iPSCs into CSCs, and is used in spheroid culture to enrich CSC populations [50,51], CAF-derived FGF2 may also play a larger role in regulating other cells within the TME.

Unsurprisingly, cancer cell and CAF communication is bidirectional. Cancer cells can secrete hedgehog ligands to promote FGF5 secretion from CAFs [46], while cancer cell-derived TGF-β can activate normal fibroblasts to adopt a more CAF-like phenotype (Figure 2C) [52]. In breast cancer, cancer cells secrete tumor necrosis factor α (TNFα) to induce CAFs secretion of hepatocyte growth factor (HGF), which promotes cancer cell proliferation, EMT, and radioresistance in vitro. Interestingly, radiation-exposed breast cancer cells secrete more TNFα to further enhance radioresistance in a CAF-dependent manner [53]. In gastric cancer, CAF-derived HGF promotes cell proliferation and migration in only Met-unamplified cells, while CAF-derived HGF enriched the proportion of CD44⁺ CRC stem cells in vitro and promoted metastasis through increased adhesion to endothelial cells in vivo [54,55]. In HCC, both fibroblast and cancer cell-derived TGF-β enhance expression of FSTL1 within CAFs. CAF-derived FSTL1 promotes tumorigenesis, metastasis, and sorafenib resistance in vivo [56]. In breast cancer, FSTL1 promotes stemness markers, sphere formation, and chemoresistance in vitro, while also promoting tumorigenesis and metastasis in gastric cancer [57,58]. In HCC, spatial multiomics identifies a subset of CAFs marked by high expression of FSTL1 that are localized within regions of cancer stemness. Unsurprisingly, these regions demonstrate enrichment of TGF-β and Notch signaling [59].

2.3 CAF-derived cytokines and chemokines

In gastric cancer, CAFs secrete IL-6 to promote EMT and metastasis by activating the JAK/STAT3 axis, while IL-6 promotes sphere formation and the expression of stemness markers Oct4, SOX2, CD44, and Nanog in HCC [60,61]. Similarly, breast cancer CAFs secrete IL-6 to promote radioresistance via STAT3 signaling, while CAF-derived IL-6 enhances p53 ubiquitination and degradation, allowing cells to escape doxorubicin-induced apoptosis [62,63]. High levels of other CAF-derived cytokines such as IL-8 are associated with poor response to chemotherapy and cisplatin resistance in gastric cancer [64]. In lung cancer, IL-8-secreting CAFs are marked by high levels of the filament Desmin, and are activated by cancer cell-derived lactate. Normal fibroblasts co-injected with lung cancer cells enhanced tumor growth in vivo, while knockout of the Desmin transcription factor JUNB-B in normal fibroblasts attenuated this effect [65]. Similarly, in ovarian cancer, CAF-secreted IL-8 promotes sphere formation and stemness in vitro, while CAF co-injection with ovarian cancer cells promotes tumor growth in vivo [66]. In OSCC, CAFs also produce the chemokine CXCL12 to increase sphere formation and expression of stemness markers in cancer cells [67]. CXCL12 has also been shown to stimulate normal fibroblasts to adopt CAF-like properties in vitro [68], suggesting that CAF secretion of CXCL12 may have a positive feedback effect. Given that CXCL12/CXCR4 signaling activates oncogenic ERK and PI3K/Akt signaling, and promotes chemotherapy resistance [69], it is unsurprising that high expression of CXCL2 is associated with metastasis to the brain, bone, lung, and liver in breast cancer [70]. Additionally, there are a number of other CAF-derived cytokines that regulate CSCs such as CAF-derived CCL2, which promotes CSC self-renewal via NOTCH1 signaling in breast cancer [71].

3.1 Tumor associated macrophages

Tumor associated macrophages (TAMs) are macrophages found within the TME, and originate either from tissue resident macrophages or circulating monocytes, the latter of which is the main source of TAM replenishment [76]. Monocyte-to-macrophage differentiation is driven by complex environmental cues and immune/cancer cell secreted cytokines such as colony stimulating factor (CSF) 1, CSF-2, and IL-34 [77]. Once differentiated, these monocytes can be further polarized by signals within the TME into two general categories, M1-polarized and M2-polarized macrophages. M1-polarized macrophages are recognized as tumor-suppressive, while M2-polarized macrophages are known for their tumor-promoting abilities [78]. However, these categories are likely an oversimplification, and macrophages in vivo exist on a spectrum of activation states influenced by a complex network of immune signals [79], of which CSCs play a critical role (Figure 3A). For example, in GBM, CSCs preferentially express periostin (POSTN) to recruit monocytes to the tumor site and promote polarization to an M2, tumor-promoting state via integrin αvβ3 signaling, a finding recapitulated in ICC [80,81]. Similarly, glioma stem cells (GSCs) induced STAT3 phosphorylation and STAT1 dephosphorylation in macrophages to promote M2-polarization and secretion of immunosuppressive IL-10 [82]. Inflammatory breast CSC-like cells also recruit monocytes and promote M2-polarization via IL-8 and GRO-mediated STAT3 phosphorylation in monocytes [83]. Additionally, numerous other studies have demonstrated the ability for CSCs to polarize monocytes to M2-like macrophages [84,85]. CD133⁺ GSC-derived IL-6 and IL-10 activate STAT3 signaling within macrophages to activate downstream expression of VTCN1. VTCN1 inhibits macrophage phagocytosis, T cell IL-2/IFN-γ production, and promotes T cell apoptosis, and silencing of VTCN1 in macrophages reduces tumor growth in vivo [86]. GSCs also preferentially express WISP1, which promotes stemness in an autocrine manner while promoting M2 TAM survival via integrin α6β1/Akt signaling [87]. Likewise, GSC-derived FSTL1 promotes stemness by activating PI3K-Akt signaling in an autocrine manner, while also promoting macrophage recruitment and M2 polarization in vivo [88].

CD14⁺ M2-like macrophages enhance sphere formation, stemness marker expression in GBM cells, and enrich the population of CD133⁺/CD44⁺ CSCs via CCL2 secretion. Co-injection of these M2-like macrophages with cancer cells enhances tumor growth in vivo, an effect reversed by inhibition of CCR2 [84]. Similarly, TAM-derived CCL8 promotes stemness of GBM cells cultured in vitro, while promoting tumor growth in vivo. Interestingly, cells cultured under sphere conditions overexpressed the CCL8 receptors CCR1 and CCR5 [89]. TAM-derived cytokines are also implicated in CSC regulation in numerous other cancer types. In breast cancer, high infiltration of CD163⁺ M2 macrophages in the TME is associated with lymph node metastasis and decreased survival, and TAM-derived CCL2 is responsible for enhancing the proportion of CD24^-/CD44⁺ or ALDH⁺ CSCs via CCL2/Akt/β-catenin signaling in vitro [90]. Additionally, TAM-derived IL-6 promotes stemness via STAT3 signaling in HCC cells, and TAM-derived CCL5 similarly promotes stemness via β-catenin/STAT3 signaling in prostate cancer. Knockdown of CCL5 in TAMs inhibited tumor growth, metastasis, and CSC populations in vivo [91,92]. In HCC, spatial transcriptomic analysis suggests that CSCs are colocalized with SPP1⁺ macrophages within the TME, which are typically regarded as tumor-promoting M2-like macrophages [93,94]. Consistent with their role in tumor progression, SPP1⁺ macrophage and CSC colocalization is associated with poorer survival in patients with HCC. Interestingly, Fan et al. also show enhanced HIF-1α expression within colocalized areas [93]. Accordingly, HCC cells have been shown to secrete miR-1290 to induce M2-macrophage polarization [95]. Moreover, in HCC, SPP1⁺ macrophages are frequently found in close proximity to FAP⁺ CAFs [96,97], suggesting that macrophages, CAFs, and CSCs may form a complex network to regulate stemness within the TME. Given that M2 macrophages also suppress immune response in HCC by promoting the expression of exhaustion markers PD1/TIM3 on CD8⁺ T cells [98], CSCs may also modulate immune response via M2 macrophages.

While M1-macrophages, as the counterpart to M2-macrophages, were originally classified as tumor-suppressive, high levels of iNOS⁺ M1-polarized macrophages have been suggested to be associated with improved survival in HER2⁺ breast cancer [99]. However, Oshi et al. demonstrate that upon higher dimensional, transcriptomic classification of M1/M2 macrophages, M1 or the ratio of M1/M2 macrophages are not associated with improved survival in breast cancer [100], suggesting that the role of M1 macrophages may be more complex. In vitro, M1-polarized macrophages enrich the populations of CD24^-/CD44⁺ and ALDH1⁺ breast CSCs during co-culture, while increasing the expression of stemness markers Lin-28B and Nanog in breast cancer cells [101]. Similarly, conditioned media from LPS/IFNγ-induced M1 macrophages enhances the expression of stemness markers in breast cancer cells via NFκB signaling [102]. In CRC, ID1-expressing M1-like macrophages have pro-tumorigenic and stemness promoting properties. Tumors grown in myeloid-specific ID1 knockout mice contain fewer ALDH⁺ and CD44⁺/Lgfr5⁺ CSCs, and inhibition of ID1 deubiquitination in M1 macrophages with ML323 reduces tumor growth, metastasis, and sensitizes tumors to chemo/immunotherapy [103]. These findings suggest that M1 macrophages, traditionally classified as tumor-suppressing, can also influence CSC maintenance and promote tumor growth.

3.2 CSCs and checkpoint inhibition

In addition to macrophages, CSCs also interact with other immune cells in the TME such as CD8⁺ and CD4⁺ T cells (Figure 3B). While activated CD8⁺ T cells are highly effective in clearing cancer cells, exhausted CD8⁺ T cells exhibit diminished effector function marked by high expression of checkpoint inhibitor receptors such as PD-1, CTLA-4, and TIM-3 [104]. While immunotherapies targeting these pathways can restore immune cell activity and enhance tumor clearance [105], CSCs have been shown to undermine these responses by mediating the expression of checkpoint inhibitors and activating additional mechanisms of immune evasion. For example, immunohistochemistry (IHC) analysis of CD44 expression, a common stem cell marker in lung adenocarcinoma (LUAD), reveals a positive association with PD-L1 expression [106], while CD133⁺/CD44⁺ CRC stem cells likewise exhibit higher PD-L1 expression via enhanced STAT3 signaling in patient-derived xenograft (PDX) models [107]. In breast cancer, CD24^-/CD44⁺/EpCAM⁺ CSCs express higher levels of PD-L1 via Notch3/mTOR signaling in vitro [108], while ALDH1A1 expression by IHC in breast cancer is also associated with increased PD-L1 levels [109]. However, these associations may be context/cancer dependent. IHC analysis in NSCLC demonstrates a negative correlation between ALDH1 and PD-L1, and no significant relationship between CD44 and PD-L1 expression [110], suggesting that the relationship between stemness and PD-L1 expression may be more complex. Moreover, TGF-β-induced EMT increased PD-L1 levels in breast CSCs via β-catenin/STT3 glycosylation and stabilization of PD-L1 in vitro [111]. In HNSCC, IFN-I/IFNAR1 signaling promotes stemness and production of exosomal PD-L1 and Galectin-9, while IFNAR knockdown reduces the proportion of CD44⁺/ALDH⁺ CSCs, inhibits tumor growth, and promotes T-cell exhaustion in vivo [112]. However, CSC-derived PD-L1 not only acts on immune cells, but also on other cancer cells. In CRC, CD133⁺/CD44⁺ CSCs express higher levels of PD-L1, which interacts with HMGA1 to enhance sphere formation and Oct4 expression via Akt and MEK signaling [113], suggesting that CSCs can induce stemness within other cells in addition to suppressing immune response via PD-L1 expression.

3.3 Other immune cells

Apart from regulating CD8⁺ cells, CSCs also influence the activation and function of various other numerous immune cell populations. In breast cancer, CSCs promote NK cell migration but inhibit NK cell-mediated cytotoxicity. These CSCs express higher levels of NK-inhibiting ligand HLA-E, while downregulating the NK-activating ligands MICA and MICB [114]. However, other studies suggest that NK cells preferentially target CSCs in various cancers including breast cancer, glioma, sarcoma, and pancreatic cancer [115]. Additionally, CD4⁺ T cells cultured in the presence of breast CSC-conditioned media contain a larger proportion of CD4⁺CD25⁺CD127⁻ Tregs, possibly via CSC-derived TGF-β (Figure 3B) [116]. Breast CSCs also secrete exosomal FOXP3 to transform CD4⁺ T cells into CD4⁺/CD25⁺/FOXP3⁺ Tregs in vitro [117]. In GBM, CSCs not only promote the generation of Tregs, but also inhibit CD4⁺ and CD8⁺ cell activation and proliferation via activation of STAT3 signaling in T cells [118]. In HNSCC, CD44⁺ CSCs produce higher levels of numerous cytokines including TGF- β, and also promote Treg generation, MDSC generation, and inhibit T cell activation from PBMCs [119]. Ovarian CSCs not only induced Tregs to produce higher levels of the immunosuppressive cytokine IL-10, but also promoted the recruitment and chemotaxis of Tregs in vitro via CCL5 secretion [120]. In vivo, ovarian CD133⁺ CSCs express higher levels of the IL-17 receptor compared to non-CSCs, and colocalize with IL-17-producing cells such as CD4⁺ TH17 cells and macrophages. This IL-17 promotes sphere formation in vitro and tumorigenesis in vivo via NF-κB and p38 MAPK signaling in cancer cells [121]. Similarly, IL-17 promotes stemness in pancreatic cancer cells via NF-κB signaling [122].

3.4 Myeloid-derived suppressor cells

Other important immunosuppressive cells in the TME include MDSCs. MDSCs are a heterogeneous population of cells broadly classified for their immunosuppressive ability. MDSCs arise from myeloid progenitors in the setting of abnormal maturation and development, often due to cancer, chronic inflammation, or infection [123]. MDSCs are typically classified into two categories, polymorphonuclear and monocytic MDSCs (PMN-MDSC and M-MDSC). Although some typical surface marker sets are used to distinguish the two, higher dimension analysis has revealed many subpopulations within each category [124]. In the setting of cancer, MDSCs are recruited to the TME by chemokines such as CCL2 and CXCL12, where they can activate and proliferate via stimulation of STAT3, STAT6, and NF-κB pathways, among many others [125]. In breast cancer, CSCs preferentially express G-CSF via enhanced mTOR signaling, which results in MDSC accumulation within the TME [126], while melanoma CSC-derived TGF-β similarly recruits PMN-MDSCs to the tumor site [127]. In prostate cancer, Yap1-high CSC-like cells express higher levels of CXCL5, which promotes MDSC recruitment via CXCR2 signaling [128] (Figure 3C).

MDSCs are well-established mediators of T-cell immunosuppression, employing mechanisms such as L-arginine depletion, nitric oxide (NO) production, and reactive oxygen species (ROS) generation [129], but also regulate cancer cell stemness within the TME. In breast cancer, PMN-MDSCs enrich the proportion of ALDH⁺ CSCs, increase sphere formation, and upregulate expression of stemness markers via CXCL2/CXCR2 signaling [130]. Similarly, MDSCs cocultured with ovarian cancer cells enhance sphere formation ability and stemness markers via CSF-2/STAT3 signaling [131]. MDSC-derived IL-6 and NO activate STAT3 and NOTCH signaling pathways in vitro to enrich the population of ALDH⁺ CSCs in breast cancer [132]. Given that inducible nitric oxide synthase (iNOS) expression and subsequent NO production within cancer cells are associated with CSC maintenance via Notch1 signaling in numerous types of cancers [133,134], it is possible that MDSC-derived NO may exert a similar effect in cancers other than breast cancer. Similarly, while few studies have examined the role of MDSC-derived ROS in relation to CSCs, cancer cell-derived ROS are known to play multiple important roles in CSC maintenance, including ROS-mediated HIF stabilization [135]. In ovarian cancer, tumor-derived G-CSF promotes the generation and survival of MDSCs in vivo [136]. These MDSCs, in turn, secrete prostaglandin E2 (PGE2) to increase stemness and PD-L1 expression via the PI3K/Akt axis. Accordingly, G-CSF overexpressing tumors contain elevated levels of ALDH⁺ CSCs in vivo [136,137]. Additionally, in a mouse breast cancer model, PNM-MDSCs maintain the CD49f⁺ mouse CSC population via netrin-1 signaling. Notably, while anti-CTLA-4 immunotherapy enriched the population of CD49f⁺ CSCs, combination with netrin-1 blockade abolished this effect [138], suggesting that MDSC-mediated regulation of stemness may also contribute to therapy-induced CSC expansion.

4.1 MSCs-secreted cytokines and chemokines

Numerous cytokines secreted by MSCs have been shown to promote stemness (Figure 4A). MSCs express higher levels of BMP2 upon coculture with ovarian cancer cells, which acts in a paracrine manner to increase the number of ALDH⁺/CD133⁺ CSCs and enhance sphere formation ability in vitro, an effect abolished upon treatment with the BMP signaling inhibitor, Noggin [145]. Furthermore, primary MSCs co-cultured with GSCs promote sphere formation in vitro, and tumorigenesis in vivo via MSC-derived IL-6/STAT3 signaling [146]. Similarly, MSCs isolated from human bone marrow (BM-MSCs) secrete IL-6 to promote cell invasion in HCC [147]. IL-6 has also been shown to enrich CSC populations and promote treatment resistance in several cancers including breast cancer, HNSCC, and colorectal cancer [148–150], suggesting that MSC-derived IL-6 may play a central role in promoting stemness across cancer types. In breast cancer, ALDH⁺ MSCs secrete CXCL7 to induce cancer cell IL-6 secretion, which can then promote MSC chemotaxis in a positive feedback loop. Accordingly, ALDH⁺ MSCs injected into the mouse tibia were targeted to the breast tumor site, promoting tumor growth and enriching the proportion of ALDH⁺ or CD24^-/CD44⁺ CSCs in vivo [151]. Additionally, while MSCs are recognized to secrete PGE2 to promote recovery after liver injury [152], MSC-derived PGE2 also mediates CSC maintenance. When co-cultured with MSCs, CRC cells secrete IL-1, which stimulates MSCs to produce PGE2, IL-6, and IL-8. Co-injection of MSCs with CRC cells promotes tumorigenesis and enriches the ALDH⁺ CSC population via PGE2/Akt/β-catenin signaling [153]. PGE2 has also been demonstrated to promote cancer cell stemness in lung cancer via PI3K/Akt signaling, suggesting the effects of MSC-derived PGE2 may not be restricted to only CRC [154]. Similar to the findings reported by Liu et al. in breast cancer, BM-MSCs injected into the bloodstream are recruited into mouse prostate tumors, which then increase the number of CD44⁺/CD133⁺ CSCs. In vitro, co-culture of BM-MSCs with prostate cancer cells promotes sphere formation and the expression of stemness markers via downregulation of AR signaling [155]. In GBM mouse models, co-injection of MSCs with GBM cells likewise enhances tumor growth and enriches the population of tumoral CSCs [156]. In HCC, tumor-specific CD34⁺/CLDN5⁺ endothelial cell-derived IGF can recruit MSCs into the TME, which induces a stem-like phenotype in HCC cells [157], further supporting the notion that cells within the TME can actively recruit MSCs to enhance cancer cell stemness.

4.2 Other roles of MSCS

Apart from directly mediating cancer stemness, MSCs have also been suggested to convert to a CAF-like phenotype (Figure 4B). Long-term stimulation with TNF-α and IL-1b imparts MSCs with CAF-like morphology, increased vimentin and FAP expression, and an increased ability to contract collagen gels [158]. MSCs cultured in stiff, ECM-mimicking conditions acquire a CAF-like phenotype upon treatment with breast cancer cell-derived conditioned media. These CAF-like cells downregulate the MSC marker Stro-1 while upregulating CAF markers α-SMA and LMNA [159]. Similarly, transcriptomic analysis of MSCs treated with exosomes derived from colorectal cancer cells reveals an increase in CAF-related genes such as α-SMA and FAP [160]. Considering the substantial evidence supporting the role of CAFs in promoting stemness, context-dependent transformation of MSCs to CAFs may represent an additional mechanism of MSC-mediated enhancement of cancer stemness. Additionally, MSCs may also fuse with non-CSCs to generate CSC-like cells (Figure 4C). Dornin et al. provide evidence to suggest that breast cancer cells fused with MSCs contain a higher proportion of ALDH⁺ CSCs and also express higher EMT markers [161]. Similarly, lung cancer-MSC and normal epithelial cell GES1-MSC fusion hybrids express higher levels of EMT markers and possess increased tumorigenicity in vivo [162,163]. Fusion hybrids of GBM cells and MSCs demonstrate higher stemness markers SOX2, Oct4, and Nanog in vitro, while displaying increased tumorigenicity in vivo [164].

5.1 Cancer cell-derived exosomes

Additionally, exosomes have also been shown to promote stemness in recipient cells. For example, pancreatic ductal adenocarcinoma (PDAC) CSCs secrete exosomes enriched in tetraspanin 8 (TSPAN8) compared to their non-CSC counterparts. Exosomal TSPAN8 activates Hedgehog (Hh) signaling in recipient cells to promote colony formation, stemness marker expression, and tumorigenesis [170]. Similarly, in HCC, non-CSC-derived exosomes containing the Hh ligand Sonic hedgehog (Shh) could enhance recipient cell stemness [171]. In esophageal squamous cell carcinoma, CD44⁺ CSCs secrete exosomes containing the lncRNA FMR-AS1, which acts as a TLR7 ligand to promote TLR7/NF-κB mediated stemness in cancer cells [172]. In GBM, CSC-derived exosomes containing the Notch1 receptor enhance sphere formation, stemness marker expression, and tumorigenesis in recipient cells [173], suggesting that exosome-mediated stemness may not only be restricted to the transfer of ligands, but functional receptors as well.

In addition to exosomal proteins, exosomal RNA has also been widely studied in the context of CSCs. In breast cancer, exosomal miR-105-5p produced by CSCs promotes cancer cell migration and tumor growth by inhibiting the translation of GPR12 [174], while exosomal miR-454 derived from breast cancer non-CSCs activates Wnt signaling to expand the population of CD44⁺/CD133⁺ CSCs in vitro while enhancing tumor growth in vivo [175]. In melanoma, CSC-derived exosomes containing miR-1268a and miRNA-4535 have been shown to promote metastasis in non-CSCs by inhibiting autophagy [176,177], while lung CSCs secrete miR-210-3p via exosomes to promote EMT by targeting fibroblast growth factor receptor 1 (FGFRL1) translation in recipient cancer cells [178]. Similarly, in renal cell carcinoma (RCC), CSCs produce exosomal miR-19b-3p to promote EMT in recipient cells by inhibiting PTEN translation. Interestingly, these RCC CSC-derived exosomes have differential effects on tumor growth based on size and CD103⁺ incorporation, with high CD103⁺ exosomes being able to more efficiently target tumor cells [179]. miRNAs are known to be regulated by lncRNAs, which can sponge miRNAs to inhibit their activity [180], representing another layer of post-transcriptional regulation that CSCs can exploit to promote tumor progression. LUAD CSC-like cells produce exosomes enriched in the lncRNA Mir100hg, which can promote metastasis by enhancing ALDOA activity and N3K14 lactylation [181].

5.2 CSC exosome-mediated immune modulation

Recipients of CSC and non-CSC-derived exosomes are not only limited to cancer cells. CSC-derived exosomes have been demonstrated to mediate immune cell activity and promote immunosuppressive TME. While Wang et al. demonstrate that exosomal miR-210-3p can inhibit FGFRL1 in cancer cells [178], CSC-derived exosomes containing miR-210 also target macrophages in prostate cancer. Inhibition of FGFRL1 in macrophages promotes M2 polarization, while co-injection of these M2-polarized macrophages with prostate cancer cells promotes tumor growth and gemcitabine resistance in vivo [182]. In oral squamous cell carcinoma, CSCs produce exosomes containing the lncRNA UCA1, which inhibits miRNA-134-mediated silencing of laminin γ2 (LAMC2) in macrophages. The resulting upregulation of LAMC2 in macrophages and CD4⁺ T cells upon exosomal delivery of UCA1 promotes M2-polarization and reduces the proportion of activated IFNγ⁺/CD4⁺ T cells [183]. Similarly, tenascin-C, an ECM glycoprotein, is packaged into exosomes by glioblastoma stem cell-like cells to reduce T cell proliferation and activation via tenascin-C/α5β1 and αvβ6-mediated signaling [184].

Furthermore, Gabrusiewicz et al. suggest that only monocytes and activated T cells are able to uptake glioblastoma stem cell-derived exosomes. Recipient monocytes exhibit M2-polarization and increased PD-L1 expression [185], suggesting that CSC-derived exosomes can selectively target certain immune cells within the TME. In CRC, CSC-derived exosomes localize to mouse bone marrow and induce bone marrow-derived neutrophils to secrete IL-1β. These IL-1β-secreting neutrophils exhibit increased lifespan and ability to enhance tumorigenicity of CRC cells in vivo [186]. Similarly, CRC CSCs produce exosomes enriched with β-catenin, IL-6, TGF-β1, and pSTAT3, which induce M2 macrophage polarization and normal fibroblast to CAF conversion, and reduction of this exosomal cargo upon ovatodiolide treatment reduces tumor growth in vivo [187]. In vitro, monocytes that differentiate into dendritic cells in the presence of RCC CSC-derived exosomes containing the human leukocyte antigen G (HLA-G) exhibit lower levels of dendritic cell activation markers CD83 and CD40, and the MHC-II molecule, HLA-DR [188].

5.3 CSC exosome-mediated chemoresistance

CSC-derived exosomes have also been suggested to mediate chemoresistance. Oral squamous cell carcinoma exosomes promote resistance to cisplatin in vitro, while reduction of exosomal cargo via ovatodiolide treatment sensitizes tumors to cisplatin in vivo [189]. Exosomes derived from gemcitabine-resistant pancreatic CSCs promote gemcitabine resistance in gemcitabine-sensitive cells by transferring miR-210 [190]. Similarly, in hepatocellular carcinoma, CSC-derived exosomes enhance regorafenib resistance in vitro in a Nanog-dependent manner [191], suggesting that CSC-derived exosomes may be another mechanism through which tumors can develop chemoresistance. Additionally, more substantial evidence demonstrates dysregulation of exosome production upon chemotherapy treatment. Breast CSC-like cells treated with cisplatin differentially express numerous miRNAs whose miRNA transcripts include numerous cancer-relevant proteins such as MYC and JAG1 [192]. Likewise, docetaxel and doxorubicin treatment of non-CSC breast cancer cells induces production of exosomal miRNAs targeting the transcription factor ONECUT2. While in vivo data suggests that ONECUT2 may promote tumor growth, miRNA-mediated downregulation of ONECUT2 upon docetaxel/doxorubicin treatment induces a stem-like phenotype by upregulating Notch1, Oct4, and SOX9 in breast cancer cells [193]. In gastric cancer cells, cisplatin and paclitaxel treatment induces exosomal lncFERO secretion, which promotes SCD1 mRNA translation to protect recipient CSCs against ferroptosis [194]. In bladder cancer, treatment of non-CSCs with gemcitabine/cisplatin promotes the production of exosomes that can protect recipient CSCs from chemotherapy [195].

5.4 Non-cancer cell-derived exosomes

Within the TME, the production of exosomes is not limited to cancer cells alone. In colorectal cancer, CAFs secrete exosomal lncRNA H19, which acts as a miR-141 sponge to promote tumor growth and resistance to oxaliplatin in recipient cells [196]. These non-cancer cell-derived exosomes have also been demonstrated to mediate stemness. In CRC, CAF-derived exosomes containing Wnt3a enrich stemness markers in cancer cells, promote their sphere formation ability, and drive in-vivo chemoresistance and tumor proliferation [197], while MDSC-derived exosomes containing S100A9 promote stemness of CRC cells in vitro and tumor growth in vivo [198]. Wang et al. demonstrate that S100A9 binds NADPH oxidase to increase ROS production and subsequently activate STAT3 and NF-κB signaling. Subsequent work by Wang et al. demonstrates that MDSC-derived exosomes also promote M2-macrophage polarization by transferring miR-93-5p to downregulate STAT3 signaling in recipient monocytes [199]. In addition to CAFs and MDSCs, BM-MSCs also regulate CSCs via exosomal signaling. BM-MSCs exosomes containing miR-142-3p promote sphere formation, stemness marker expression, and the proportion of CD133⁺/LGR5⁺ CSCs in vitro, while CRC cells overexpressing miR-142-3p exhibit increased tumorigenesis in vivo. Mechanistically, miR-142-3p targets mRNA transcripts of DUMB, a negative regulator of Notch signaling, to promote Notch pathway activity [200]. Similarly, in acute myeloid leukemia, BM-MSCs-derived exosomes upregulate S100A4 expression in recipient cancer cells, enhancing their sphere formation ability and expression of stemness markers [201]. However, BM-MSC exosomes containing the circular RNA circ-0030167 inhibit stemness in vitro, while also reducing tumor growth in vivo by inhibiting miR-338-5p/WIF1/Wnt8/β-catenin signaling in pancreatic cancer cells [202]. Altogether, these findings suggest that apart from CSCs and non-CSCs, various stromal and immune cells also contribute to the exosomal secretome within the TME.

Ethics Statement

Not applicable

Consent for Publication

Not applicable

Availability of Data and material

Not applicable

Funding

This work was supported by grants from the NIH (R01CA248027, R01CA211175, and R01CA234124) and CDMRP (W81XWH211028, and HT9425-24-1-1034)

Competing Interests

The authors have declared that no competing interests exist.