Cancer Stem Cell Division: When the Rules of Asymmetry Are Broken

Normal conditions

During normal interphase, aPKC localizes apically upon forming a complex with PAR6 and L(2)GL (Fig. 1). Aurora A, a serine-threonine protein kinase initiates apical signaling by phosphorylating PAR6, which in turn activates aPKC [1]. Activated aPKC phosphorylates L(2)GL, reducing its affinity with the complex and leading to its replacement by PAR3 [1]. Under normal conditions, activation of L(2)GL and the entry of PAR3 leads to the critical event of Numb phosphorylation, inactivating and releasing it from the apical plasma membrane (Fig. 1). Numb is a well-established Notch signaling suppressor, and its inactivation therefore upregulates Notch signaling, providing self-renewal properties to the apical daughter [1].

OPEN IN VIEWER

FIG. 1.

During asymmetric cell division, two distinct molecular programs take place on the apical and basal pole. Apical pole: At the apical side, aPKC/PAR6/PAR3 complex formation initiates during interphase, giving apical pole the identity of self-renewal. Aurora A protein kinase phosphorylates aPKC leading to its activation. Active aPKC phosphorylates L(2)GL which releases L(2)GL from the complex and PAR3 enters. This complex phosphorylates Numb releasing it from the apical membrane and increasing Numb concentration at the basal side. This keeps the Notch signal active on the apical side causing stemness. Wnt signal also takes part in the self-renewal, though the detailed mechanism is not known. In metaphase and telophase, apical microtubule arrangement is maintained by Inscuteable/Pins/Gα_i complex. Cytoskeletal adapter protein binds with Pins and Gα_i whereas Mud forms a complex with Dlg and Knc73. These two complexes come together that arranges the microtubules attached to Knc73. Basal pole: Adapter protein Miranda on the basal pole binds with Prospero and Brat. Degradation of Miranda releases transcription factor Prospero that turns on the genes driving differentiation. Brat on the other hand, acts as a translational repressor, possibly suppressing protein production needed for proliferation. brat ortholog TRIM32 also transports cMYC to endosomes for its degradation. Higher accumulation of NUMB on the basal membrane leads to the degradation of NOTCH1 inside endosomes. This suppresses Notch signal and its proliferative effect on the basal side. aPKC, atypical protein kinase C. Color images available online at www.liebertpub.com/scd

Results concordant with Drosophila have been established using mammalian systems: during asymmetric cell division of radial glia within the mouse brain ventricular zone, the aPKC/PAR6/PAR3 complex accumulates at the apical side [1] with the help of a small GTP-binding protein, CDC42 [1]. This complex ensures apical adherens junction integrity and establishes apico-basal polarity [8].

In neoplasia

In neoplastic disease, disruption of signaling networks involved in asymmetric division typically gives rise to a proliferative state and accumulation of stem-like cells with limited capacity to differentiate. For example, a mutant form of Drosophila aPKC that is constitutively active initiates Notch signaling through reduction of active Numb on apical and basal sides, thereby promoting neuroblast self-renewal and tumor formation [1,9]. In contrast, suppression of aPKC results in reduced numbers of neuroblasts, establishing aPKC as a protumorigenic protein. Similarly, Aurora A mutants generate tumors by enhanced Notch signaling [7]. L(2)GL mutants also display a neoplastic proliferation of stem-like cells [10], likely through the formation of a nonfunctional aPKC/PAR6/PAR3 complex and activation of Notch on both apical and basal sides.

In mammalian systems, disruption of apical determinants has similar effects. Overexpression of PAR3 drives radial glial cells toward symmetric division and retention of stem-like properties of both daughter cells [5] by keeping Numb inactive, thereby enhancing Notch signaling [11]. PAR6 has been established as a causal factor for breast cancer epithelial–messenchymal transition (EMT) through transforming growth factor-beta (TGF-β) signaling. Mutated PAR6 attenuates TGF-β signaling in mice and suppresses lung metastasis of mammary tumors [12]. Loss of LGL1, a mammalian ortholog of fly L(2)GL, disrupts neural progenitor cell differentiation in the ventricular region of the mouse cerebellum, and promotes proliferation [13].

There are numerous examples of human cancers that harbor alterations in orthologs of Drosophila genes related to asymmetric cell division. For example, deletions in PARD6 (PAR6), PARD3 (PAR3), and DLG2 are prevalent in human epithelial neoplasms. PARD6 deletion disrupts the cell polarity causing neoplastic growths in lung and adrenal cancer (Table 1). PARD3 deletion in squamous cell carcinomas and glioblastomas leads to the loss of cell polarity, symmetric cell division, and enhanced proliferation. DLG2 is most frequently deleted in cervical and lung cancer, and generates a phenotype similar to PARD3 deletion [14,15]. AURKB, the human ortholog of Aurora kinase is frequently deleted in metastatic prostate adenocarcinoma, strongly supporting a connection between neoplastic disease and deregulation of symmetric cell division (Table 1). In some human cancer types, genes implicated in asymmetric cell division have been shown to alter properties of cancer stem cells [1,16], while in other instances this has not been demonstrated.

Normal conditions

On the basal side during asymmetric division, the combination of apical Numb phosphorylation and the absence of aPKC/L(2)GL/PAR3 complex leads to accumulation of active Numb, which inhibits Notch signaling and directs this daughter toward differentiation. In addition to active Numb, a highly coil-coiled adapter protein, Miranda, localizes basally during interphase and binds Prospero, a transcription factor, and brain tumor (Brat), a translational repressor [7]. Eventually, Miranda is degraded, releasing Prospero and initiating a transcriptional program that promotes differentiation [7]. Besides being a translational repressor, Brat also attenuates self-renewing Wnt signaling in neuroblasts through β-catenin/Armadillo inhibition [17].

Uncovering precise functions of Brat has been challenging due to its complex structure and reports of numerous, seemingly unrelated functions [18]. Nevertheless, Brat has been firmly established as a driver of differentiation in neuroblasts (Fig. 1). One of the first roles described for Brat was as a translational repressor of Hunchback in the embryo [19], yet deeper mechanistic detail have not emerged. Based on sequence homology and similar protein domains, the mammalian ortholog of Brat belongs within the tripartite motif containing protein family called TRIM. TRIM32, a ubiquitin ligase, has been reported to drive neurogenesis in a mouse model [1] and has several functions that include activation of microRNAs (mi-RNAs), such as Let-7, and binding of c-MYC to promote its degradation [1].

Similar to Drosophila, studies of the mammalian ortholog of Prospero, PROX1, has established its role in supporting neurogenesis in mouse models [20].

In neoplasia

Numb localization on the basal side is critical, since it attenuates Notch signaling (Fig. 1) driving differentiation. As apical fate determinant proteins, such as aPKC and L(2)GL, primarily regulate Numb activation, Numb localization at the basal side depends on them. Under conditions where aPKC is inhibited, phosphorylation of L(2)GL does not occur [1] on the apical side, resulting in attenuated Numb phosphorylation and its failure to localize to the basal membrane [1]. Thus, Numb's aberrant maintenance causes inappropriate Notch signaling after division [21] resulting in excessive proliferation of neuroblasts and tumor growth.

Mutations in miranda, prospero, or brat lead to a brain tumor phenotype in Drosophila, characterized by massively enlarged brains containing neuroblasts that fail to differentiate. These mutants have remarkable similarity, pointing to the common regulatory mechanism shared by these proteins. Transplantation of mutant neuroblasts into the abdomen of wild-type Drosophila results in tumor growth that is ultimately fatal, suggesting they are more that benign or self-limited proliferations [22 –24]. Transplanted cells also show metastatic behavior, with primary abdominal neuroblasts forming secondary tumors disseminating far from the primary tumor site [7].

Mutations of mammalian PROX1, the ortholog of Prospero have been noted in pancreatic, esophageal, and colon cancer. Mice xenograft experiments with pancreatic cancer cells show reduced proliferation and self-renewal with the introduction of PROX1 protein [25].

PROX1 alterations and reduced expression are seen in several forms of human neoplasia (Table 1), including hematologic malignancies, where DNA methylation of PROX1 is predominant (56.3%) and contributes to the neoplastic behavior [26]. In hepatocellular malignancies, PROX1 transcription is altered in a manner that promotes progression, while in breast cancer, epigenetic silencing has been described [27,28]. NUMB gene mutation and deletion are hallmark of many cancer types, such as uterine endometrial carcinoma and adenoid cystic carcinoma, where mutation induces self-renewal through Notch activation (Table 1). NUMB protein is reduced in selective subsets of breast cancer [8].

Orthologs of Brat in human cancer have also received increasing attention, especially as related to brain tumors. Among TRIM genes, TRIM3 has greatest sequence homology to Drosophila brat. TRIM3 is also exclusively expressed in the brain, similar to brat in Drosophila. TRIM3 is deleted in about 25% of human glioblastomas and has a similar deletion frequency in lower grade gliomas, suggesting it is an early event in gliomagenesis [29]. Recent studies suggest that TRIM3 also has E3 ubiquitin ligase activity and is a tumor suppressor at least partially through ubiquitination and degradation of p21. However, TRIM3 overexpression suppresses proliferation in oligodendrogliomas in both presence and absence of p21, suggesting other signaling pathways are also impacted [30]. Recent work has corroborated that ∼25% of GBMs have at least one TRIM3 allele deleted and also suggested that nondeleted gliomas have methylated TRIM3 that could account for its lower gene and protein expression in nearly all GBMs [16]. Low TRIM3 protein expression in GBM neurospheres was associated with upregulated Notch signaling, possibly through reduced expression of the negative regulator NUMB. Reconstitution of TRIM3 in TRIM3 null neurospheres led to increased NUMB levels and suppressed Notch signaling. In both glioma cell lines and neuropsheres, the lack of TRIM3 was also tightly coupled with elevated expression and activity of c-MYC. Functional studies showed that TRIM3 expression within a glioma stem cell population (as defined by PKH-26 high) led to reduced symmetric cell division and rescued the asymmetric cell division phenotype. Altogether, these data suggest that the Brat mammalian ortholog TRIM3 regulates cell proliferation and self-renewal signaling within a glioma stem cell population [16].

Another Brat ortholog is TRIM32, which has been shown to ubiquitinate MYC protein, thereby attenuating self-renewal during asymmetric cell division in mice and inhibiting over proliferation and tumor formation [31].

Normal conditions

During mitotic spindle formation in metaphase and telophase, in preparation for asymmetric cell division, microtubules are arranged in a manner to support division into unequal daughters. The process starts with apical localization of the cytoskeletal adapter protein Inscuteable [32], which then binds with partner of inscuteable (Pins), and the G protein receptor substrate Gα_i, in a receptor-independent manner [1]. Also on the apical side, microtubule-bound kinesin Khc73 binds with disc large (Dlg) and forms a complex with Gα_i, Pins, and Mud [1]. Altogether, this large protein complex is linked to the apical cytoskeleton through Inscuteable and establishes the correct spindle alignment and the orthogonal cleavage plane along the apical/basal axis (Fig. 1).

In mammals, the ortholog of PINS is LGN. LGN binds with a microtubule-associated protein NuMA and Gα_i to establish asymmetric microtubule orientation [33]. In the absence of LGN, PAR3 localizes exclusively to one cell and drives excessive production of outer radial glia [1].

In neoplasia

Dlg is an established tumor suppressor, with Drosophila mutants demonstrating loss of apico-basal polarity in imaginal discs during development, resulting in cellular enlargement. Absence of Dlg disrupts microfilament and microtubule network and disrupts the localization of basal determinants [1,34]. Inscuteable protein anchors mitotic spindles that assist with localization of basal determinants, such as Miranda and Prospero. In the absence of Inscuteable, symmetric cell division occurs rather than asymmetric, mostly due to Miranda and Prospero localizing randomly [7]. The Pins and Mud also interact with one another to establish proper spindle orientation. Loss of either (or both) results in a lack of spindle alignment with the asymmetric fate determinants, leading to their haphazard localization and resulting in the formation of proliferative neuroblasts-like cells [35].

Mouse LGN is closely related to Drosophila Pins and can rescue Pins mutants to maintain asymmetric division. Genetic mutation and siRNA experiments have shown that functional loss of LGN leads to disruption of asymmetry in radial glial cell division and leads to proliferation through oblique division as opposed to vertical division [8]. Overexpression of mammalian NuMA, an ortholog of Mud, results in aneuploidy in mouse myeloid cell [36] and may cause neoplastic growth in the setting of additional alterations, such as TP53 mutation [37].

A large number of human breast cancer cells show high levels of LGN protein expression [38], suggesting that its enhanced function may play a role in tumor development and progression. LGN is a target of PBK/TOPK for phosphorylation and subsequent activation of LGN drives symmetric spindle orientation, resulting in proliferation and tumor formation. Another well-known tumor suppressor in colorectal carcinoma, adenomatous polyposis coli (APC), binds with human DLG1 protein and negatively regulates cell cycle progression [39]. APC also binds with β-catenin through GSK-3β and axin complex, eventually leading to β-catenin ubiquitination and degradation. This process normally attenuates WNT-signaling and suppresses cell proliferation, but the downstream effects are reversed in cancer following the genomic alteration of signaling network members [40,41].