Background

Ovarian cancer (OC) is one of the most dangerous gynecologic malignancies. In 2020, the morbidity of OC was estimated to be 3.4% worldwide for women, and ranked eighth among female malignant tumors and third among gynecological malignancies. Mortality from OC accounts for 4.7% of all female malignant tumors around the whole world and is ranked second among female genital tumors [1]. Serous ovarian cancer (SOC) is the most common type of gynecological malignancy [2]. The National Comprehensive Cancer Network (NCCN) has recommended poly ADP-ribose polymerase inhibitors (PARPi) and angiogenesis inhibitors as treatment for patients with SOC. Although PARPi and antiangiogenics have been shown to prolong progression-free survival (PFS), they do not reflect an obvious improvement in overall survival (OS) of patients with SOC [3, 4].

The tumor microenvironment (TME) denotes the niche where tumor cells interact with the surrounding stroma, including various immune cells, stroma cells, lymph-vascular space, and the extracellular matrix (ECM) [5]. Mesenchymal stem cells (MSCs) are a key component of stromal cells and can mediate the immune response by inhibiting the activity of T lymphocytes, interfering with the proliferation and differentiation of B lymphocytes, and inducing macrophage phenotypic switching [6]. In addition to immune regulation, MSCs can promote angiogenesis by releasing soluble factors and can be a source of carcinoma-associated fibroblasts (CAF). CAFs, in turn, can directly release angiogenesis-related factors and indirectly modulate pathophysiological processes, including ECM stiffness, elasticity, and interstitial fluid pressure [7, 8]. In OC, MSCs regulate cancer cell proliferation, metastasis, phenotype, and response to chemotherapy by binding directly to target cells, secreting soluble factors, or discharging exosomes from MSCs [9,10,11]. Cancer-associated MSCs (CA-MSCs) can be isolated and identified in tumor tissue, and exhibit a unique gene expression profile from MSCs compared to that of healthy individuals [12]. Patients with the CA-MSC phenotype have a significantly worse PFS than those with the normal MSC phenotype [10]. It is reported that CA-MSCs of an immune 'hot' mouse OC drived CD8 + T cell tumor immune evasion of CD8 + T cells from tumors and these mouse exhibited a poor response to anti-programmed death ligand 1 (PD-L1) immune checkpoint blockade therapy (ICB) through the secretion of multiple chemokines, such as CCL2, CX3CL1, and TGF-β1 [13].

Cobalt chloride (CoCl2)-induced polyploid giant cancer cells (PGCC) exhibit the same characteristics as cancer stem cells (CSC) and express markers related to CSC (CD133 and CD44). Daughter cells produced by PGCC undergo an epithelial–mesenchymal transition (EMT) and gain a mesenchymal phenotype and are thus endowed with strong abilities for migration and invasion [43], which may be related to a poor prognosis of the mesenchymal phenotype of SOC. For different MSC phenotypes, high-risk scores indicated the presence of abundant infiltration of immunocytes with inhibitory activity, such as immature dendritic cells and regulatory T cells (Fig. 4C). Except for tumor-associated immunosuppressive cells, immunity inhibition factors were also differentially expressed (Fig. 4D; Additional file 16: Fig. S4A). MSCs have been reported to secrete immunomodulatory factors that influence other immune or stromal cells, such as transforming growth factor-beta (TGF-β1) on macrophages, vascular endothelial growth factor receptor 2 (KDR) in endothelial cells, and galectin-9 (LGALS9) in T cells [51]. TAMs are a heterogeneous cell population and broadly classified into pro-inflammatory M1 and anti-inflammatory M2 macrophages. In OC, M2 macrophages comprise 39% of the immune cells and are associated with adverse clinical outcomes [52]. Single cell spatial analysis discloses intimate interactions of exhausted CD8+ T cells and PD-L1+ macrophages that are considered mechanistic determinants of response to niraparib and pembrolizumab treatment, which are PARP and immune checkpoint inhibitors, respectively [53]. Among stromal cells, CAFs originate from MSCs or by transdifferentiation of other cells. With the expression of specific molecules and receptors, CAFs promote angiogenesis, metastasis, and infiltration of immunosuppressive cells, thus fueling tumor growth and progression [54]. SOC stromal fibroblasts exhibit intrinsic resistance to PARPi and increased further after PARPi administration [55]. Except for the PARPi response, patients with high CAF infiltration exhibit chemoresistance and contribute to the insensitivity to immunotherapy [54, 56, 57]. Similarly, the increased dispersion of MSCs in SOCs tends to shorten survival and attenuates the response to immunotherapy of patients in our study.

In terms of the clinical relevance, the expression of MSC markers and the existence of soluble factors derived from MSCs are negatively correlated with the prognosis of patients. CD105 + MSCs were associated with a reduced OS of patients with brain neoplasm, lung cancer, and gastric cancer [58,59,60]. MMP9 and IL-6 are secreted proteins of MSCs. High expression of MMP9 has been associated with low survival rates in lung adenocarcinoma [61]. Patients with high IL-6 levels have significantly a poorer survival rate than those with low IL-6 levels [62]. The exosomal microRNAs released by MSCs were positively related to survival time in colorectal, myeloid leukemia, nasopharyngeal carcinoma, and glioma [63,64,65,66].

Essentially, MSCs exert immunomodulatory effects on both innate and adaptive cells through cell-to-cell contact and paracrine activity, including T cells, natural killer (NK) cells, and DCs. Induction of regulatory T cells (Tregs) is a main mechanism of immunosuppression by MSCs. MSCs can convert conventional T cells (T convs) to Forkhead box P3 (Foxp3) expressing Tregs [67]. Foxp3 is a transcription factor that inimitably defines Tregs and is a requirement for Tregs differentiation [68]. The immature dendritic cells are a subset of dendritic cells that selectively promote the proliferation of Tregs, and both take part in immunosuppressive activity [69]. Moreover, mature DCs co-cultured with MSCs skew to immature status and show a reduced stimulatory activity on T cells [70]. Therefore, in our study, patients with the mesenchymal phenotype tended to have an immunosuppressive state characterized by a richness of Tregs and immature DCs, and responded poorly to anti-PD-1 therapy. However, there are other immunoeffector cells that assembled in the high-risk score group. NK cells are innate cytotoxic lymphocytes and MSCs modulate their inhibitory effects on cell proliferation, altered cytotoxicity and cytokine production, and induction of apoptosis by MSC secreted cytokines such as prostaglandin E2 (PGE2), indoleamine 2,3-dioxygenase (IDO), TGF-β1, IL-6, and nitric oxide (NO) [71]. In our study, NK cells were enriched in the high MSC risk score group. This change largely resulted from the fact that there was heterogeneity among NK cells, and CD56bright NK cells and CD56dim NK cells are two main subsets of circulating human NK cells. CD56bright NK cells are more immature and are more enriched in the tumor, and exhibit more limited cytotoxicity responses compared to CD56dim NK cells [72], which is likely to occupy the majority of the mesenchymal phenotype of OC. In addition, Wan et al. revealed that the unique bispecific anti-programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) antibody induced NK cells to transition from inert to more active and cytotoxic phenotypes, implicating NK cells as the key missing component of the current ICB-induced immune response in SOC [73].

With the exception of the prognostic response to ICB, our MSC score system is able to provide guidance for anti-angiogenesis therapy. As for patients with high-risk scores, standard chemotherapy plus angiogenic inhibitor is superior to chemotherapy alone. Stefani et al. found that low-dose irradiated MSCs showed antiangiogenic properties and infiltrated predominantly the perivascular niche, leading to rejection of established tumors [74]. MiR-16, a microRNA targeting VEGF, was enriched in MSC-derived exosomes and partially resulted in an antiangiogenic effect in breast cancer cells [75]. The MSC score system was a prospective marker for the administration of PARPi. The higher the risk scores, the lower the number of defects in homologous recombination, for which repair deficiency is closely associated with sensitivity to PARPi therapy in epithelial OC [76, 77]. Therefore, patients with a mesenchymal phenotype may not be suitable for treatment with PARPi.

In our validation experiment, three genes related to poor outcomes were prone to accumulate in the healthy ovary and in epithelial cells. However, the PER1 content is distinct between HEY and SKOv3 cells, which may be explained by the role of TP53. TP53 in HEY cells is wild-type and in SKOv3 cells is deleted. PER1 knockdown influences pancreatic cancer cell lines with mutated TP53, but does not alter cells containing wild-type TP53 [78], the reason for this finding is that p53 represses PER1 transcription [79]. MSCs can act directly on metastasis of tumor through production of pro-metastatic cytokines or regulation of epithelial-mesenchymal transition [80, 81]. Not only in SOC, the elevated proteins of PER1, AKAP12 and MMP17 are also associated with the migration and invasion in triple-negative breast cancer, melanoma and colon cancer [82,83,84].

Because of its multipotency, low immunogenicity, easy accessibility and ethical advantage compared to pluripotent stem cells or embryonic stem cells, MSCs are desirable candidates for in degenerative and inflammatory diseases, auto-immune diseases, such as joint injury, atopic dermatitis, and multiple sclerosis [85]. Infrapatellar fat pad-derived mesenchymal stem cells, proximal to the knee joint and similar to adipose cells, own proliferation and differentiation potential independent of age and promote hyaline-like cartilage formation without integration into the surrounding cartilage [86]. Currently, there are several phase I/II and III clinical trials involving immunomodulatory MSCs aimed at treating graft-versus-host disease and tumors. In combination with ganciclovir, genetically-modified autologous MSC were found to be safe and tolerable in patients with advanced gastrointestinal adenocarcinoma [87]. A similar trial confirmed that allogeneic MSC infusions showed safety and feasibility in patients with prostate cancer [88]. Another trial in which endovascular superselective intraarterial (ESIA) MSC infusions loaded with an oncolytic adenovirus Delta-24 (MSC-D24) were used to treat glioblastoma is currently underway [89]. The direction of treatment for MSCs mainly includes the delivery of various anticancer biological agents or suicide genes using an extracellular vesicle derived from MSCs [90, 91]. However, we have to face the possibility about the latent pro-metastasis functions and the promotion of immune evasion if anticancer agents or suicide genes in MSCs cannot function. In addition, MSCs combined with drug nanoparticles were used to induce the death of cancer cells. The conjugation forms between MSCs and nanoparticles include MSCs loading nanoparticles, nanoparticles attached to MSCs surface, nanoparticles coated with MSCs membrane, and vectors of anti-tumor genes in MSCs [92]. Because of tumor tropism of MSCs, the conjugation between MSCs and nanoparticles solved the problem of low target specificity and minimized side effects of conventional medicine. However, the toxicity of nanoparticles and the uncertainty of pharmacokinetics are still existent, including accumulation in organs followed by inflammation or binding with blood constituent followed by coagulation [93]. If these disadvantages about the safety of MSCs, nanoparticles or drugs can be solved, MSC may gradually be applied in clinical practice.

Conclusions

Using a comprehensive transcriptomic analysis of genes characterizing MSCs, this study constructed an MSC-related prognostic model that could separate patients into two groups. The high-risk group was associated with a worse prognosis, a different immunosuppressive phenotype, and a weak response to anti-PD-1 treatment. There was also instructive significance of this sorting system for anti-angiogenesis therapy.