The vulnerable primed cancer stem cells in disguise: demystifying the role of Maspin

Epithelial-specific Maspin is widely known as a tumor suppressor. However, while the level of maspin expression is inversely correlated with tumor grade and stage, emerging clinical evidence shows a correlation between seemingly better differentiated tumor cells that express Maspin in both the nucleus and the cytoplasm, (n + c)Maspin, with a poor prognosis of many types of cancer. Biological studies demonstrate that Maspin plays an essential role in stem cell differentiation. In light of the recently established characterization of primed stem cells (P-SCs) in development, we propose, for the first time, that cancer stem cells (CSCs) also need to undergo priming (P-CSCs) before their transition to various progeny phenotypes. We envisage major differences in the steady state kinetics between P-SCs and P-CSCs. We further propose that P-CSCs of carcinoma are both marked and regulated by (n + c)Maspin. The concept of P-CSCs helps explain the apparent dichotomous relationships of (n + c)Maspin expression with cancer diagnosis and prognosis, and is supported by the evidence from mechanistic studies. We believe that the potential utility of (n + c)Maspin as a molecular marker of P-CSCs may significantly accelerate the advancement in our understanding of the genesis of tumor phenotypic plasticity in response to changes of tumor microenvironments (TME) or drug treatments. The vulnerabilities of the cellular state of (n + c)Maspin-expressing P-CSCs are also discussed as the rationale for future development of P-CSC-targeted chemotherapeutic and immunotherapeutic strategies.


The role of Maspin as a tumor suppressor challenged
Maspin is a clay B member in the serine proteinase inhibitor (serpin) superfamily, also known as Serpin B5 [1]. The maspin expression is epithelial-specific in normal somatic tissues, and is differentially regulated in carcinomas. Since its discovery, Maspin is widely known as a class II tumor suppressor, i.e., the loss of its expression, but not its genetic mutation or deletion, is a gain of function in tumor progression [2]. The evidence in support to this notion includes the inverse correlation between the total levels of maspin protein and mRNA with tumor grade and stage [3]. For example, the level of maspin mRNA in primary prostate carcinoma (PCa) is comparable or slightly higher than that in adjacent benign prostate epithelial tissues, but is significantly down-regulated in invasive and metastatic diseases ( [4], Fig. 1A). Accumulated biological evidence further supports a causative relationship between Maspin and tumor suppression, demonstrating that Maspin blocks tumor invasion and metastasis without affecting cell proliferation [3]. However, clinical investigations into the relationship between Maspin and clinical outcome revealed that better or moderately differentiated primary tumors can be distinguished based on the subcellular localization patterns of Maspin. Specifically, as compared to the tumor cells that express maspin in the nucleus (nMaspin), the subpopulation of tumor cells that expresses Maspin in the nucleus plus cytoplasm ((n + c)Maspin) is specifically correlated with malignant progression and/or earlier tumor relapse [5] of ovarian cancer [6], non-small cell lung cancer [7], and PCa [8], among other types of carcinoma. It is worth noting that the majority of the experimental approaches employed established tumor cell lines or immortalized noncancerous epithelial cell lines, wherein the expression or silencing of Maspin is ectopically controlled constitutively. In in vitro culture, these cell lines express Maspin in the (n + c) pattern, irrespective of its subcellular localization in the original tissue specimens. The apparent dichotomous relationships of (n + c)Maspin expression with tumor cell differentiation and cancer prognosis bring into focus the following important questions: is Maspin exclusively a tumor suppressor? If the (n + c)Maspin-expressing tumor cells represent a bulk tumor cell population that is inhibited in invasion and metastasis, how can they lead to poor clinical outcomes? Is it possible that the (n + c)Maspin-expressing tumor cells in natural tumor progression represent an intermediary cellular state that is not aggressive it itself but has the potential to become aggressive?

The (n + c)Maspin-expressing cellular state in stem cells differentiation
Cancer mortality is predominantly caused by metastasis and drug resistance, both of which are facilitated by tumor cell genetic instability and phenotypic plasticity. To date, one of the most supported hypotheses for tumor heterogeneity and phenotypic plasticity comes from the concept that a small number of naïve CSCs carry driver oncogenic mutations and, when activated, are capable of self-renewal (SR) and multilineage differentiation. To this end, CSCs share with normal stem cells (SCs) many common cellular characteristics and the underling molecular mechanisms. Maspin is not expressed in naïve embryonic stem cells (ESCs) or somatic SCs, but is involved in their differentiation. During human pregnancy, detection of fetus-derived hypermethylated maspin DNA in the plasma predicts a higher risk of preeclampsia [9]. In mouse embryogenesis, the level of maspin mRNA and protein gradually increases from embryonic day (ED) 1 to ED5, and then decreases on ED6 and ED7. Treatment of pregnant mice with anti-Maspin polyclonal antibodies significantly reduced the number of implanted embryos [10]. Mouse ESCs with homozygous deletion of the Maspin gene were viable in vitro. However, after the implantation, they died at the peri-implantation stage, with disorganized endodermal cell mass and no basement membrane layer in the embryo body [11]. The embryonic lethality of Maspin knockout was also shown by a Cre-recombinase-based breeding scheme at the step of oogenesis [12]. Forced tissue-specific transgenic expression of Maspin curtailed mouse mammary gland development, which was accompanied by increased epithelial apoptosis [13]. On the other hand, through an alternative breeding scheme wherein cytoplasmic Maspin could be carried over from oocyte to the zygote, viable Maspin knockout (Maspin −/− ) mice were successfully generated. However, these Maspin −/− mice developed a spectrum of epithelial abnormalities including adenoma and tumors of primitive basal epithelial features in the mammary and prostate glands, and poorly differentiated lung adenocarcinoma [12].
According to the theory of CSC in tumor progression, the activation of CSC differentiation should be expected whenever new colonies of a surviving tumor progeny phenotype emerges. The (n + c)Maspin-expressing tumor cells appear in the same time window. In normal somatic tissues, nMaspinexpression is strictly restricted to terminally differentiated epithelial cells (e.g., luminal cells in the mammary gland and secretory epithelial cells in the prostate gland), whereas the (n + c)Maspin-expressing cells reside in the basal layer of these glands [1,4]. It is important to note that the basal epithelial layer contains both basal or myoepithelial cells of primitive differentiation and multipotent epithelial stem (precursor) cells [14]. Carcinogenesis is commonly marked by the loss of the normal basal epithelial cells. However, the (n + c)Maspinexpressing cells did not disappear with the basal epithelial layer. Instead, they are detected abundantly at the earliest in situ tumors [1,15] (Fig. 1Ba-c), exhibiting a gene expression profile similar to that of epithelial precursor cells [16]. In breast cancer lymph node metastasis, the re-appearance of the primary-liked tumor cells was associated with the expression of (n + c)Maspin [17], whereas in PCa bone metastasis, (n + c)Maspin-expressing tumor cells of moderate differentiation are detected sporadically among the majority of tumor cells that do not express Maspin (Fig. 1Bd). In radical prostatectomy specimens of PCa patients who received neoadjuvant anti-androgen treatment, (n + c)Maspin-expressing cells were detected among the debris of the bulk androgen-sensitive tumor cells [18].

The (n + c)Maspin-expressing tumor cells as primed CSCs (P-CSCs)
In the last couple of decades, significant advancements have been made to demonstrate the capability of CSCs to initiate tumor, and led to the identification of their cell surface molecular markers [19]. However, there is yet an overarching conceptual framework for how to connect the central control of the CSCs in producing heterogeneous bulk tumor cell populations, shifting the differentiation lineages in different tumor microenvironments (TME), and supporting cancer drug resistance and recurrence. To this end, a recent breakthrough in the field of development demonstrates that naïve SCs that are stimulated to exit the symmetrical self-renewal (SR) are primed prior to making the commitment to their destined differentiation [20]. The primed SC (P-SC) step seems preserved across species, and may be triggered by stress signals, including oxidative stress or nutrition deprivation. The gene expression profiles of P-SCs seem distinct from those of naïve SCs, but have no clear lineage directionality [21]. Cells in the P-SC state are considered epigenetically "paused," yet with a landscape of chromatin modification poised to respond swiftly to differentiation-specific transcription factors [22]. The P-SCs of adult epithelial cells reside among the basal epithelial cells [23] and exhibit a hybrid basal and luminal cell differentiation profile [14]. Other P-SC features include a decrease in membranous integrity of intracellular organelles and an increase in nuclear-cytoplasmic shuttling of proteins. These cells also have a higher level of tolerance of increased endoplasmic reticulum (ER) stress without elevating unfolded protein response (UPR) which may otherwise prematurely prevent the establishment of the upcoming proteome in the destined phenotype [24]. While P-SCs may survive such contemporaneous stress at least in part by the mechanism of autophagy, they may exhibit an elevated level of immunogenicity which is marked by the increased expression of major histocompatibility complex class I (MHC-I) molecules [25]. These characteristics of P-SCs are also the key features of the (n + c)Maspin-expressing tumor cells that have an expression profile overlapping with that of inflammatory [26] basal epithelial cells [16]. In some cancer patients, anti-Maspin antibodies were also detected [27]. We propose that the (n + c)Maspin-expressing tumor cells are the P-CSCs ( Fig. 2A). The epigenetic pausing in primed stem cells involves transient asynchronous histone modification [22]. Starting from the step of oogenesis and morphogenesis in development, a dominant histone-modifying enzyme is histone deacetylase 1 (HDAC1) that controls the expression of key regulators of cellular stemness, including Nanog which is a driver of stem cell priming [28,29]. It came as no surprise that HDAC1 is essential and cannot be complemented by other HDACs in early embryogenesis [30]. Interestingly, HDAC1 may be translocated to the cytoplasm in stressed stem cells [31]. Cytoplasmic HDAC1 may reside in the cytosol and ER [32] to deacetylate proteins such as heat shock protein 90 (Hsp90) [33] and glucose regulated protein 78 (GRP78). In turn, deacetylated Hsp90 and GRP78 are biochemically active to protect new proteomic and transcriptomic profiles to capacitate the phenotypic transition. Conversely, acetylation inactivates Hsp90 and GRP78. A hallmark of Hsp90 inactivation is the decrease of its translocation from the cytoplasm to the nucleus, whereas acetylation-mediated inactivation of GRP78 may alter ER integrity and trigger its dissociation from the ER without initiating UPR [34]. In tumor progression, HDAC1, Hsp90 and GRP78 are commonly upregulated.
Maspin is one of the most ancient and highly conserved clay B serpins [38]. Consistent with the X-ray crystallographic analysis [39], Maspin does not inhibit activated serine proteases [40]. Instead, it inhibits the activation of serine protease zymogens such as single-chain tissue-type plasminogen activator and pro-urokinase-type plasminogen activator [41][42][43], and inhibits serine protease-like enzyme HDAC1 [44]. Maspin is the only endogenous polypeptide enzymatic inhibitor of HDAC1 identified thus far. As shown in Fig. 2B, Maspin depends on amino acid (aa) residue R 340 in its reactive center loop (RCL) to interact with the catalytic domain of the target molecule. As a 42 kDa protein, Maspin can pass the nuclear envelope pores through diffusion. It also has an intramolecular KDEL motif (aa 345-348) for contingent ER retention. Among all serpin homologues, maspin is the only one with this KEDL sequence. Conservative substitution of D 346 with E in this motif rendered exclusive nuclear localization of both Maspin and HDAC1 [45]. Thus, the synchronized nuclear-cytoplasmic shuttling of HDAC1 and Maspin may be driven by Maspin.
The (n + c)Maspin-associated gene expression profile was highly sensitive to changes of TME. When cells were cultured as three-dimensional (3D) organoids, (n + c)Maspin expression resulted in increased expression of antigen processing and histocompatibility genes, as well as inflammatory cytokines. Under several experimental conditions, (n + c)Maspin commonly upregulated the expression of 31 genes and downregulated the expression of 29 genes [46]. These common changes are involved in stem cell differentiation, adhesion and tumor progression. Bioinformatics analysis identified transcription factors (TFs) that directly control Maspin-regulated gene expression [47]. A general upstream analysis revealed that Maspin downregulates key nodes of the TGF-β signaling network, and upregulates key nodes in the networks of inflammation and selective cell death.
Glutathione s-transferase (GST), Hsp90 and GRP78 have also been identified as predominant Maspin-associated proteins [45,48]. The inhibitory effects of Maspin on HDAC1 is enhanced by its molecular interaction with GST, leading to reduction of cellular oxidative stress [48]. Maspin also abrogates the HDAC1-mediated epigenetic silencing of GST-Pi [49]. Interestingly, a mutually exclusive subcellular localization pattern of Maspin and GRP78 was revealed by dual immunofluorescent staining [45]. This result would be expected if Maspin acts to inhibit HDAC1-mediated deacetylation of GRP78, which leads to the exit of acetylated GRP78 from the ER [34]. These data suggest that the Maspin/HDAC1 axis may exert a contemporaneous and concerted control of cell differentiation, stress response, tumor inflammation and selective cell death pathways at the step of stem cell priming.

The "Disguise" of (n + c)Maspin-expressing P-CSCs
If priming is a common steady state required for stem cell differentiation, its kinetics may vary under different pathophysiological conditions. In somatic tissues, the elusive quiescent multipotent SCs can go through the steps of selfrenewal and differentiation with a high level of fidelity to replenish the function of terminally differentiated cells that are injured or lost [50]. This highly hierarchical process may quickly pass through the step of P-SC to reach the predestined differentiated products. A major disadvantage of CSC study is the lack of clearly defined timing and loci of CSC activation and priming since the onset of cancer is driven by genetic mutations which disrupt the normal predestined temporal and spatial program of differentiation. To maximize the survival of the progeny cells, CSCs may be perpetually primed by their mutator genotypes and the evolving TME, rather than by the finite need for normal tissue morphogenesis or damage repair. Consequently, P-CSCs may be substantially accumulated in the tumor tissues, and easily mistaken as one of the final products of CSC differentiation. In this new thought framework, the evidence that (n + c)Maspinexpressing tumor cells are less invasive and metastatic as compared to more aggressive bulk tumor cells that do not express Maspin makes biological sense, since these cells represent an intermediary steady state between naïve CSCs and their established progeny cells. To this end, the past interpretation of the same data as evidence for a tumor suppressive role of Maspin was inaccurate due to the assumption that the (n + c)Maspin-expressing cells represented an established progeny phenotype of CSC differentiation. In any experimental model wherein ectopic (n + c)Maspin expression is driven by a strong constitutive promoter, the surviving clonal cell lines may resemble a P-CSC state. However, since these experimental P-CSCs are prevented from downregulating the level of maspin expression, their commitment to non-epithelial phenotypes, such as mesenchymal-like or neuroendocrine phenotypes may not ensue or reach fruition. On the other hand, while these results underscore the limitation of using ectopic Maspin expression to study the differentiation of CSCs wherein the temporal and spatial dynamics of Maspin expression is expected to play a key role, they serendipitously helped to exaggerate the manifestation of certain aspects of the P-CSC state.
For example, as expected of P-CSCs, (n + c)Maspinexpressing tumor cells, but not the bulk tumor cells that do not express Maspin, have the capacity to undergo induced differentiation. In an earlier study, we developed a 2D/ Susp/3D (or xenografts in vivo) scheme [51]. In brief, 2D culture for exponential growth can provide attachment anchors to trigger CSCs to exit from the SR state. Naïve CSCs in 2D culture can be enriched by consecutive passaging of mammospheres in suspension culture (Susp). Cells harvested from either 2D culture or Susp culture can be grown as 3D organoids or as xenografts in immunecompromised mice. In our experiments, a PCa cell line that expresses an undetectable level of Maspin was stably transfected with the maspin cDNA or a mock control. We showed that the "CSC's" of maspin-transfected cells lost the ability to form symmetrical mammospheres in Susp culture. Instead, they grew as asymmetric cell aggregates with an elevated level of autophagy for up to 7 consecutive passages before dying of senescence. The same cell population failed to produce tumors in the limiting-dilution tumor initiation (LDTI) assay. In contrast, the CSCs of the mock-transfected cells were immortal in Susp culture, were tumorigenic in the LDTI assay, and produced invasive progeny tumor cells. When cells harvested directly from the 2D culture were grown as 3D organoids or as xenografts in a model for PCa bone metastasis, maspin-transfected cells gave rise to better differentiated in situ like tumors whereas the mocktransfected control formed poorly differentiated tumor sheets [52]. In addition, we noted that the maspin-transfected cells in 3D culture and xenografts featured Maspin protein predominantly in the nucleus. These experimental data demonstrate that as the (n + c)Maspin-expressing tumor cells are averted from immortality, they acquire a greater aptitude for differentiation.
Further, the elevated levels of proteostatic and ER stress in (n + c)Maspin-expressing P-CSCs also help explain the evidence that maspin-transfected cells elicited a dramatic anti-tumor inflammatory response when they were subcutaneously injected into nude mice. Approximately 40% of injections were quickly "absorbed" and never developed tumors [53]. The "gross tumor volumes" of the maspintransfected tumors resulting from the remainder injections were significantly greater than those of the mock controls. However, these maspin-transfected tumors sustained dramatic tissue lysis and were infiltrated extensively by neutrophils [53,54], which targeted maspin-transfected tumor cells specifically in vitro. In addition, anti-Maspin immunoglobulin G (IgG) was detected only in mice bearing maspintransfected tumor nodules.

The vulnerabilities of the P-CSCs and future directions
To our knowledge, this is the first time that the concept of P-CSC is raised. It does not contradict the current framework for CSCs, but rather highlights a previously unrecognized step between the naïve CSCs and the bulk cancer cells. It is likely that future studies will identify additional molecular features that are specifically associated with P-CSCs in different organ and tissue microenvironments. For example, a large body of evidence demonstrates that carcinoma is a disease resulting from the combination of genetic mutations and aging. The cellular characteristics and gene expression profiles of (n + c)Maspin-expressing P-CSCs, overlap significantly with those of senescing epithelial cells [51,55]. In fact, replicatively senescent normal human mammary epithelial strains [56] expressed (n + c)Maspin at a significantly higher level than immortalized normal breast epithelial cell lines [1]. A possibility to be tested is that P-SCs derived from aging SCs with insufficient reprogramming capacity may not complete the full differentiation process, thus may remain in the P-SC state which is otherwise called atrophy [57]. We propose that the (n + c)Maspin-expressing P-CSCs in aging tissues may be similarly accumulated like P-SCs and serve as the immediate precursors of bulk tumor cells of divergent phenotypes. It was reported that the expression of (n + c)Maspin, vs. nMaspin, predicts a worse prognosis in laryngeal cancer patients who were 65 years of age or older [58]. We believe that P-CSCs are at the center in connecting CSCs to the genesis of phenotypic plasticity, tumor metastasis, tumor dormancy and tumor recurrence. It has long been recognized that metastasis is an inefficient process. The chances for circulating disseminated tumor cells to establish metastasis at secondary organ sites is estimated to be less than 0.01%. Such a low efficiency demonstrates that the bulk tumor mass may not be responsible for the colony formation at metastatic sites. The probability of metastasis may depend not only on how many CSCs are disseminated but also on how sensitive these CSCs are to the priming signals from different tissue microenvironments. These additional biological variables may explain why some disseminated "tumor cells" remain dormant for a long time and when they exit the dormancy their phenotypes resemble more closely their matching primary tumors [59]. We propose that the combination of cancer genetics and the quantification of (n + c)Maspin-expressing P-CSCs may accurately predict the propensity of tumor metastasis and cancer lethality.
Ideally, naïve CSCs, the seeds of all the bulk tumor cells in the continuum of tumor progression and metastasis, should be eliminated in order to cure the disease. This notion may explain why androgen deprivation therapy (ADT) of PCa always leads to cancer recurrence, since ADT may not be cytotoxic to PCa CSCs that do not express androgen receptor [60]. Yet, naïve CSCs may be only a nominal fraction of tumor cell population and can remain quiescent without distinct phenotypic features to aid specific and effective drug targeting. Our new concept of P-CSC suggests that P-CSCs, not naïve CSCs, may be accumulated, thus presenting a widow of opportunity to nip the devil in the bud. To this end, we have shown that (n + c)Maspinexpressing P-CSCs are resistant to a number of chemotherapeutic agents. However, two characteristics of P-CSCs may be particularly promising drug targets. First, the "intracellular stress" that the P-CSCs have to tolerate may be pushing the cells to the limit of viability, which may prove to be an Achilles' heel. In particular, the decreased membranous integrity of ER and other organelles may perturb the ion gradients, and render the cells more sensitive to drugs like salinomycin (SAL). SAL is a broad-spectrum ionophoretype antibiotic that is generally nontoxic to healthy people and has been approved by the US Food and Drug Administration (FDA) for treating patients infected by the coronavirus disease 2019 (COVID-19). Of note, SAL is cytotoxic to CSCs both in preclinical studies and in clinical trials [61]. In 2020, HSB-1216, a novel formulation of SAL, was approved by FDA as an orphan drug for treating small cell lung cancer. In our study, we found that the (n + c)Maspin-expressing P-CSCs of PCa were resistant to docetaxel and HDAC inhibitor, but were hyper-sensitive to SAL in the Susp culture, 2D culture and 3D culture [51]. It is intriguing to speculate that the effectiveness of SAL may be due to its dual targeting to both CSCs and (n + c)Maspin-expressing P-CSCs.
Second, the (n + c)Maspin-expressing P-CSCs may have an elevated level of immunogenicity of MHC class I binding peptides, which may induce cytotoxic T lymphocyte response. Major breakthroughs in cancer immunotherapies have benefited patients with many types of cancer. However, it remains a challenge to make personalized prediction of the therapeutic response. To this end, in non-small cell lung cancer patients treated with immune checkpoint blockade inhibitors (ICIs), the development of a CD8 + /IFN-γ + T lymphocyte population that is also positive with T cell receptor (TCR) for Maspin peptides was specifically associated with partial or complete remission [62]. While ICIs are not effective in treating PCa, the autologous dendritic cell therapy Sipuleucel-T (Sip-T), the FDA-approved immunotherapy for asymptomatic metastatic castration resistant PCa (mCRPC) [63], produced remission in some patients in clinical trials. Of note, although the overall mortality rate of African American men (AAM) with PCa is more than double any other racial groups, the median overall survival of the AAM treated with Sip-T was 20.6 months longer than that of the European American men (EAM) [64]. It has been shown that the effectiveness of Sip-T depends on tumor epitope spreading (i.e., the broadening of the immune recognition to additional tumor antigens) [65], thus may be dictated by the baseline tumor immunogenicity and inflammation. Indeed, the highly heterogenous PCa is commonly associated with an elevated level of inflammation [66]. In particular, low-grade inflammation is more prevalent in the primary tumors of AAM PCa patients, and AAM PCa patients who responded favorably to Sip-T showed a higher baseline level of IFN-γ than EAM patients [64]. It is intriguing to postulate that the tumor cell population that contributed to antigen spreading and Sip-T sensitivity may be the chemo-resistant (n + c)Maspin-expressing P-CSCs.
To explore the possibility of blocking P-CSCs by maspintargeted therapeutic strategies, it is essential to identify the molecular features of maspin involved in the specific inhibition of HDAC1 in the nucleus and the cytoplasm, respectively. Based on the X-ray crystallographic analysis, future structural-functional relationship studies of maspin may be focused on its the RCL. In addition, the G-helix of Maspin is another highly flexible secondary structure and has been implicated in protein phosphorylation and molecular interaction with co-factors [39]. Our revelation about the role of (n + c)Maspin in the steady state of P-CSCs may serve as a reminder that tumor progression involves changes of transcriptomes, proteomes, as well as protein subcellular distribution. Earlier, we have shown that Maspin is also secreted to extracellular space both as a free molecule and a cargo of exosomes [67]. It is important to understand how Maspin secreted in these two different forms may differentially regulate the biological activities of stem cells and their progenies, respectively. To this end, the effect of (n + c)Maspin on HDAC1 may also shift the trafficking routes of HDAC1 substrates such as Hsp90 and GRP78, both of which may also be secreted to extracellular space. The specific contributions of compartmentalized Hsp90 and GRP78 to the steady state of P-CSCs need to be delineated.