A new way to treat brain tumors: Targeting proteins coded by microcephaly genes?

Summary

New targets for brain tumor therapies may be identified by mutations that cause hereditary microcephaly. Brain growth depends on the repeated proliferation of stem and progenitor cells. Microcephaly syndromes result from mutations that specifically impair the ability of brain progenitor or stem cells to proliferate, by inducing either premature differentiation or apoptosis. Brain tumors that derive from brain progenitor or stem cells may share many of the specific requirements of their cells of origin. These tumors may therefore be susceptible to disruptions of the protein products of genes that are mutated in microcephaly. We highlight the potential for the products of microcephaly genes to be therapeutic targets in brain tumors by reviewing research on EG5, KIF14, ASPM, CDK6, and ATR. Treatments that disrupt these proteins may open new avenues for brain tumor therapy that have increased efficacy and decreased toxicity.

Graphical Abstract

Microcephaly and brain cancer exist on opposite ends of a spectrum defined by regulated proliferation of neural stem/progenitor cells. Several genes that are under-expressed in microcephaly are either over-expressed or critically important in brain cancers. We propose studying the products of these genes to develop new brain cancer treatments.

1. Introduction

During brain development, a small number of pluripotent stem cells generates a much larger number of brain cells through many rounds of cell division. This process involves the generation of neural progenitors, which are transit amplifying cells that are highly proliferative but less pluripotent ^[1]. Strict control of neural stem and progenitor cell proliferation prevents both developmental malformations resulting from hypo-proliferation and tumor formation resulting from hyper-proliferation.

Microcephaly is a neurodevelopmental disorder in which the growth of the brain is disproportionately reduced, relative to the growth of the body. Diverse pathologic processes can cause microcephaly, ranging from genetic conditions to fetal trauma to maternal exposures. The common link between these processes is insufficient proliferation, specifically in neural stem or progenitor cells. Reduced proliferation may be caused by some combination of inappropriate cell death and loss of proliferative potential through premature differentiation or senescence ^[2]. Heritable microcephaly syndromes result from mutations in genes that are specifically required by neural stem or progenitor cells to maintain their survival and proliferation. These hereditary conditions identify gene products that critically regulate processes necessary for the normal expansion neural stem and progenitor cell populations.

The cells of primary brain tumors descend from neural stem and progenitor cells and their proliferation may depend on the same genes that are required by the normal proliferative cells within their lineage. The most common primary, malignant brain tumors are glioblastoma in adults and medulloblastoma in children. The cell of origin for glioblastoma is still under investigation, but candidates include Nestin⁺ neural stem cells, multipotent progenitors of the subventricular zone, or lineage-restricted progenitors that give rise to astrocytes or oligodendrocytes ^[3–5]. Medulloblastomas of specific subgroups arise from either neural stem cells of the cerebellar ventricular zone or progenitor cells of the external granule layer of the cerebellum or rhombic lip ^[6]. In both glioblastoma and medulloblastoma, either a majority or a distinct population of tumor cells retain genetic and molecular characteristics of the stem/progenitor cell of origin ^[7]. Therefore, proteins that support growth and survival in neural stem/progenitor cells may retain their function in brain tumors. Familial microcephaly and congenital microcephalic disorders identify these proteins, which may be attractive targets for brain tumor treatment.

In this review, we focus on five genes that are each mutated in a different form of hereditary microcephaly and either over-expressed or uniquely important in a primary brain tumor: KIF11, KIF14, ASPM, CDK6, and ATR. The products of these genes all exert some form of control over mitotic entry or progression, underscoring the particular importance of mitosis for normal brain development and tumor growth. For each of the five candidates, preclinical studies have demonstrated efficacy in targeting the gene products for attenuating tumorigenesis in glioblastoma and/or medulloblastoma. These examples ultimately highlight the importance of studying normal brain development for understanding the pathogenesis and treatment of brain tumors.

2. KIF11 is mutated in primary microcephaly and upregulated in glioma

KIF11 mutations were initially identified in patients with primary, autosomal dominant microcephaly ^[8]. Examination of ten families found diverse KIF11 mutations that were associated with microcephaly, chorioretinopathy, and/or lymphedema. Heterozygous nonsense mutations and single- and di-nucleotide deletions were observed, with no clear association between genetic alteration and phenotype, including in later analyses encompassing more subjects ^[9]. All of the mutations were predicted to alter the KIF11 protein product, EG5, which is a bipolar molecular motor protein of the kinesin family. KIF11 expression begins early in mammalian development, at the blastula stage and EG5 is required for initial cleavage events in embryogenesis ^[10]. The autosomal dominant pattern of inheritance shows that heterozygous mutations are sufficient to disrupt EG5 function, and suggest that homozygous deletion may not be viable. Consistent with this conclusion, Kif11-null mouse embryos perish prior to gastrulation ^[11].

EG5 has been shown to interact with microtubules of the mitotic spindle during mitosis ^[12]. Chromatid separation in mitosis requires formation of a bipolar spindle apparatus with two centrosomes located on opposite sides of the nucleus (Fig. 1A). The hetero-tetrameric shape and plus-end-directed motor activity of EG5 allow it to crosslink antiparallel microtubules and push apart the duplicated centrosomes in prophase ^{[13, 14]}. Continued EG5 activity into prometaphase completes establishment of a bipolar mitotic spindle that is capable of pulling apart sister chromatids ^{[15, 16]}. Indeed, early studies of EG5 disruption using the small molecule inhibitor monastrol resulted in monoasters with radial spokes of microtubules surrounded by a band of duplicated, unsegregated chromosomes (Fig. 1A′) ^{[12, 17]}.

Figure 1.

The roles of specific microcephaly proteins in normal proliferation and the effects of their disruption. (A) EG5 (blue and tan coils) is a plus-end-directed microtubule motor protein that pushes apart the duplicated centrosomes to create a bipolar mitotic spindle in prophase and prometaphase, allowing for separation of sister chromatids. (A′) In the absence of EG5 activity, the centrosomes fail to separate, resulting in the formation of a monoaster, which triggers mitotic arrest that can lead to apoptosis. (B) ASPM (green octagon) is a microtubule minus-end-associated protein found at mitotic spindle poles that has been implicated in both mediating symmetric cell divisions (indicated by same colored cells) and orienting the cell division plane (dotted line) in neural stem and progenitor cells. (B′) The absence of ASPM creates more asymmetric divisions (indicated by different colored cells) and altered division planes (top vs. bottom), both of which can promote differentiation over continued proliferation. (C) CDK6 (red octagon) associates with mitotic centrosomes in the cell’s (purple) nucleus (blue), where it may mediate centrosome duplication, spindle assembly, or cellular motility (indicated by dotted lines and arrows). (C′) The CDK6 mutation associated with mirocephalic patients did not reduce CDK6 (grey polygon) protein levels but rather altered its function and behavior such that it no longer associated with centrosomes, which led to supernumerary centrosomes, inappropriate mitotic spindle alignment, abnormal cellular polarity, and reduced cell motility. (D) Endogenous, replication-associated DNA damage (lightning bolt) activates the DNA damage response protein ATR, which recognizes single-stranded DNA coated by RPA (orange circles) at stalled replication forks. Activated ATR can promote G₂/M arrest by barring mitotic entry through preventing CDK1/CDC2 activation and by targeting CEP63 to block assembly of the mitotic spindle. (D′) Absence of ATR function limits DNA repair while also permitting CGNPs with DNA damage to enter mitosis. Progression through mitosis exacerbates DNA damage leading to severe chromosomal aberrations. Overwhelming DNA damage causes ATM to activate p53, resulting in widespread apoptosis of CGNPs.

Formation of this aberrant mitotic spindles on EG5 disruption leads to activation of the spindle assembly checkpoint via MAD2 and subsequent mitotic arrest ^{[12, 18]}. Prolonged mitotic arrest triggers the cell death program through the intrinsic apoptotic pathway and activation of Caspase-3 ^[19]. The exact mechanism, however, for how mitotic arrest actually leads to apoptosis is an area of active investigation. One theory proposes that prolonged mitotic arrest produces cellular stress, which induces activation of Caspases-9 and -7 and the caspase-activated DNase ^[20]. This nuclease damages the DNA of arrested cells, triggering full engagement of the p53-dependent apoptotic pathway.

Although it is still unclear how exactly KIF11 mutation leads to microcephaly, it follows that based on the known mechanism of EG5 and the consequences of its inhibition in other cells, that reduced EG5 activity may increase apoptosis in mitotically active neural stem and progenitor cells. In the developing zebra fish nervous system, eg5 inhibition was found to attenuate the rate of proliferation and increase the rate of cell death in neural stem and progenitor cells ^[21]. Most likely through a similar mechanism, KIF11 mutations that reduce rather than completely abrogate EG5 activity specifically impair brain growth in humans.

Importantly, while EG5 acts as an essential microtubule motor during mitosis, this function is limited to mitosis ^{[12, 17]}. That is, EG5 does not seem important in for the non-mitotic functions of microtubules, which include diverse cellular processes, among them axonal transport in the central nervous system ^[22]. Thus, EG5 inhibitors may disrupt the mitotic spindle without producing the neurotoxicity of microtubule-directed agents such as vinca alkaloids (vinblastine and vincristine) and taxanes (paclitaxel and docetaxel) ^{[23, 24]}. Overexpression of KIF11 has been identified in bladder and pancreatic cancers and its expression was correlated with higher clinical grades and stages ^{[25, 26]}. A recent study showed that KIF11 knockdown reduced proliferation and increased apoptosis in triple-negative breast cancer cells ^[27]. Increased KIF11 expression in triple-negative breast cancer was associated with shortened disease free survival. A 2015 study found KIF11 upregulation in human glioblastoma and showed that EG5 inhibition by the small molecule inhibitor ispinesib (SB-715992) reduced the proliferation of neural stem-like glioblastoma cells in vitro and in mouse xenograft models by preventing mitotic progression ^[28]. In this study, Ispenisib treatment increased the survival of mice with glioblastoma from 24 to 36 days.

KIF11/EG5 inhibitors have not yet been tested for primary brain cancers in clinical trials in the United States. However, several small molecule EG5 inhibitors have been tested in patients with various other cancers, including ispinesib, litronesib (LY-2523355), filanesib (ARRY-520), SB-743921, 4SC-205, AZD-4877, MK-0731, and ARQ-621 (Table 1). For all of these agents, patients experienced predictably minimal neurotoxicity – neutropenia and leukopenia were more common side-effects since mitotic inhibition targets all rapidly dividing cells ^[48]. Unfortunately, EG5 inhibition as monotherapy had limited benefit in the cancers in which it has been tested, due to the short half-life of many of the inhibitors, which prevented targeting relatively slower cycling neoplastic cells in those cancers ^[49]. ARRY-520 and 4SC-205 have shown longer half-life, but in general, EG5 inhibitors are being tested in combination with other chemotherapeutics ^[50]. There is currently only one active clinical trial for an EG5 inhibitor: ARRY-520 for multiple myeloma. Regardless of these findings in other cancers, EG5 inhibition could potentially still hold great promise even as monotherapy for glioblastoma since the cancer cells of glioblastoma are very highly proliferative, which may allow even short half-life inhibitors to exert a potent effect. Moreover, the recent reports on EG5 inhibition in preclinical models of glioblastoma ^{[19, 28]} are encouraging and recommend further exploration of inhibiting EG5 for the treatment of high grade gliomas.

Table 1.

Several KIF11/EG5 inhibitors were tested in clinical trials for a diverse range of non-primary brain cancers, with only one currently active trial for multiple myeloma.

Compound	Cancer	Combination	Phase*	Study start	Outcomes; Dose-limiting toxicities
Ispinesib (SB-715992)	Recurrent head & neck cancer	None	II (C)	2004	No benefit; neutropenia and infection toxicities
	AML, ALL, CML, or MDS	None	I (C)	2004	Unavailable
	Metastatic prostate cancer	None	II (C)	2004	Drug failure [29]
	Metastatic malignant melanoma	None	II (C)	2004	35% subjects with stable disease; well-tolerated [30]
	Metastatic liver cancer	None	II (C)	2004	46% subjects with stable disease; granulocytopenia toxicity [31]
	Breast cancer	None	II (C)	2004	Unavailable
	Relapsed ovarian cancer	None	II (C)	2004	45% subjects with stable disease; neutropenia toxicity [32]
	Metastatic NSCLC	None	II (C)	2004	Unavailable
	Solid tumors	Capecitabine, Carboplatin, or Docetaxel	I (C)	2005	29% subjects with stable disease; neutropenia toxicity [33]
	Metastatic colorectal cancer	None	II (C)	2005	Neutropenia toxicity
	Metastatic kidney cancer	None	II (C)	2006	30% subjects with stable disease; neutropenia, hyperuricemia toxicities [34]
	Refractory solid tumors or lymphomas	None	I (C)	2006	Unavailable
	Advanced/metastatic breast cancer	None	I/II (C)	2008	7% subjects with partial response; neutropenia toxicity [35]
SB-743921	Recurrent solid tumors or lymphomas	None	I (C)	2005	Neutropenia toxicity [36]
	NHL	None	I/II (C)	2006	Hematological toxicities [37]
Litronesib (LY-2523355)	SCLC	None	II (C)	2009	5% subjects with progression-free survival; neutropenia toxicity
	All advanced/metastatic solid cancers	Pegfilgrastim	I (C)	2010	2% subjects with partial response; neutropenia toxicity [38]
	Ovarian, NSCLC, prostate, colorectal, gastroesophageal, or head & neck cancer	Pegfilgrastim	II (C)	2010	1% subjects with progression-free survival; neutropenia toxicity
	Acute leukemia	None	I (T)	2010	Unavailable
	Advanced/metastatic solid cancer	None	I (C)	2011	No benefit; neutropenia toxicity [39]
	Metastatic/recurrent breast cancer	(Peg)filgrastim	II (C)	2011	2% subjects with progression-free survival; neutropenia toxicity
Filanesib (ARRY-520)	Advanced solid tumors	Filgrastim	I (C)	2007	18% subjects with stable disease; neutropenia toxicity [40]
	AML or MDS	None	I/II (C)	2008	Unavailable
	MM or PCL	Dexamethasone & Filgrastim	I/II (C)	2009	Unavailable
	MM or PCL	Bortezomib, Dexamethasone, & Filgrastim	I (C)	2010	Unavailable
	MM or PCL	Carfilzomib, Dexamethasone, & Filgrastim	I (C)	2011	Unavailable
	Advanced MM	Carfilzomib, Dexamethasone, & Filgrastim	II (C)	2013	Unavailable
	Advanced MM	Filgrastim	II (C)	2014	Unavailable
	Relapsed/refractory MM	Pomalidomide & Dexamethasone	I/II	2015	Neutropenia toxicity [41]
4SC-205	Advanced solid tumors or lymphomas	None	I (C)	2010	28% subjects with 6-week stable disease; neutropenia toxicity [42]
AZD-4877	Advanced solid tumors or lymphomas	None	I (T)	2007	Well-tolerated [43]
	AML	None	I (T)	2007	No benefit; stomatitis and hyperbilirubinemia toxicities [44]
	Recurrent AML, ALL, NHL, or MM	None	I (T)	2007	Unavailable
	Advanced bladder cancer	None	II (C)	2008	3% subjects with partial response, 18% with stable disease; neutropenia toxicity [45]
MK-0731	Advanced solid tumors	None	I (C)	2005	Extended stable disease; neutropenia toxicity [46]
ARQ-621	Advanced solid or hematologic cancers	None	I (C)	2009	Anemia, sepsis, pneumonia, pleural effusion, thrombosis toxicities [47]

Information from clinicaltrials.gov: accessed 12/17/17.

^*Indicates current active phase or most recently reported completed phase. C = completed, T = terminated.

3. KIF14 is mutated in primary microcephaly and upregulated in glioma and medulloblastoma

KIF14 is another member of the kinesin family and, like KIF11, acts as a mitotic microtubule motor protein. Microarray analysis in synchronized cells shows maximal KIF14 expression during mitosis ^[51]. In addition, the minimal amount of KIF14 in interphase is localized to the cytoplasm, whereas mitotic progression both increases KIF14 expression and induces protein entry into the nucleus. siRNA silencing of KIF14 prevents cytokinesis, leading to either immediate apoptosis or the formation of unstable, multinucleated cells that later undergo apoptosis.

The first suggestion that KIF14 mutation could induce microcephaly came about incidentally during the breeding of genetically engineered mice to study neurite outgrowth ^[52]. Homozygous KIF14 mutation arose spontaneously and mutant mice were found to have microcephaly and severely disrupted CNS myelination. The authors found that cerebral and cerebellar hypoplasia were a result of increased apoptosis and reduced proliferation during embryogenesis. Recently, KIF14 mutation was identified as the driver of a form of primary, autosomal recessive microcephaly in four families ^[53]. These independent families were from Pakistan, Saudi Arabia, and Germany, and encompassed a total of ten individuals with microcephaly and homozygous and compound heterozygous KIF14 mutations that resulted in aberrant splicing or truncated protein. Molecular analyses showed that in primary patient-derived fibroblasts, mutant KIF14 failed to appear at the cytokinesis midbody. Absence of KIF14 was associated with concurrent absence of the critical cytokinesis regulator CRIK ^[54]. As a result, KIF14-mutated cells from these microcephalic patients tended to be either binucleated or apoptotic. Finally, two disparate inactivating biallelic mutations in KIF14 were found in two non-viable human fetuses that had severe microcephaly, intrauterine growth restriction, renal dysplasia/agenesis, and other genitourinary abnormalities ^[55]. In contrast to mutations detected in living patients with microcephaly, these mutations were detected in post-mortem analysis of non-viable fetuses ^[55], and were thus selected for greater severity, rather than for viability. All of the KIF14 phenotypes require homozygous or compound heterozygous KIF14 mutations, demonstrating that unlike KIF11, a single unaffected copy of KIF14 is sufficient to sustain neural development.

In cancer, KIF14 has been found to be up-regulated in diverse types of tumors. Retinoblastoma was among the first tumors where KIF14 overexpression was noted ^[56], and KIF14 overexpression was found to accelerate retinoblastoma tumorigenesis in a murine model ^[57]. KIF14 overexpression has also been found in both ovarian and chemo-resistant breast cancers, where KIF14 inhibition leads to chemo-sensitization ^{[58, 59]}. Interestingly, one study found reduced levels of KIF14 in lung adenocarcinomas, and levels of KIF14 mRNA and protein were inversely correlated with overall survival and metastasis ^[60]. In vitro and in vivo experiments demonstrated that increased expression of KIF14 inhibited lung adenocarcinoma growth.

In gliomas, KIF14 was first found to be upregulated in 99/128 patient-derived samples, and increased levels of both KIF14 mRNA and protein corresponded to worse pathological grade ^[61]. KIF14 expression correlated with higher proliferative rate and mitotic index in gliomas. These characteristics were associated with poorer outcomes in glioma patients with KIF14 overexpression. Results from a separate study also found increased KIF14 expression in glioblastoma cell lines and 53 human astrocytoma samples, where again, higher levels of KIF14 corresponded to worse pathological grade ^[62]. siRNA knockdown of KIF14 in glioblastoma cell lines increased G₂/M arrest, cytokinesis failure leading to binucleated cells, and apoptosis. Furthermore, tumors in mice xenografted with cells from glioblastoma cell lines were smaller if treated with KIF14 siRNA.

In a miRNA expression analysis by small RNA-Seq on 416 pediatric gliomas, miR-137 and miR-6500-3p were found to be significantly downregulated ^[63]. Reduced expression of these miRNA was associated with increased expression of KIF14. As in the prior studies, higher grade pediatric gliomas had greater levels of KIF14, and these patients had worse outcomes. KIF14 knockdown in pediatric high-grade glioma cell lines led to reduced proliferation due to G₁ arrest, and the anti-proliferative effect was enhanced by concomitant temozolomide treatment.

Finally, KIF14 expression was found to be significantly elevated in 7 medulloblastoma cell lines and 32/32 human medulloblastoma tumor samples ^[64]. Analysis on a separate group of 93 medulloblastoma samples demonstrated that, as in glioma, greater KIF14 expression corresponded to shorter progression-free survival and overall survival. Interestingly, lower KIF14 levels were seen in the WNT subgroup of medulloblastoma, which tends to have better outcomes, and higher KIF14 levels were seen in the other medulloblastoma subgroups. The authors also found increased KIF14 expression in tumor samples with gain of chromosome 1q, which is associated with non-WNT medulloblastoma. In vitro, KIF14 siRNA knockdown in two medulloblastoma cell lines attenuated proliferation, impaired cell migration and invasion, and promoted apoptosis. Taken together, these findings should inspire further research on KIF14 inhibition in animals with primary, spontaneous brain cancer. As of now, there are currently no KIF14 inhibitors in clinical trials in the United States.

4. ASPM is mutated in primary microcephaly and upregulated in glioma and medulloblastoma

Like EG5, ASPM has a role in orienting and maintaining the mitotic spindle, and it is especially important in neural stem/progenitor cells ^{[65, 66]}. ASPM is a microtubule minus-end-associated protein that is found at the mitotic spindle poles from prophase through telophase ^{[65, 67]}. In vitro studies in non-neural human cells demonstrated that ASPM is required for cell division and that cytokinesis fails in its absence ^[67]. In Drosophila neuroblasts, the ASPM homolog asp is required for proper completion of mitosis and its disruption leads to mitotic arrest at metaphase ^[68]. In contrast, loss of ASPM in mammalian neural stem and progenitor cells does not prevent mitotic progression ^[69–71]. Rather, ASPM is thought to control symmetric divisions in epithelial stem cells of the neuroectoderm and its acute inhibition by RNAi led to increased asymmetric divisions that depleted the stem cell population ^[69]. A conflicting report by the same group found no such changes to the proportion of symmetric versus asymmetric neuroepithelial divisions when Aspm was mutated. However, the authors suggested that this discrepancy could be due to acute, complete blockage of ASPM function in the ASPM-RNAi experiments versus persistent, partially preserved ASPM function in the experiments with Aspm mutation ^[70]. In the more lineage-committed cerebellar granule neuron progenitor (CGNP), ASPM mediates cell cleavage plane orientation ^[71]. Conditional Aspm deletion in CGNPs led to a relative increase in the proportion of divisions oriented transverse to the pial surface in the external granule layer of the early postnatal mouse cerebellum, consistent with inappropriately increased progenitor cell cycle exit and differentiation (Fig. 1B, B′).

Homozygous mutations in ASPM are the most common cause of familial microcephaly ^{[72, 73]}. Nearly all of the identified 150+ identified mutations lead to ASPM truncation, primarily as a result of nonsense or frameshift mutations ^[74]. Although the mutations affect all tissues, for unclear reasons, ASPM is particularly important to neural stem/progenitor cells compared to other proliferating cells ^{[65, 75]}. Indeed, Aspm-null mice develop microcephaly and hypogonadism, but other tissues seem relatively unaffected ^{[70, 76]}. Similarly, individuals with homozygous ASPM mutations uniformly exhibit microcephaly and neurocognitive disorders, with varying degrees of other pathology such as behavioral problems, seizures, and rarely various organ malformations ^[74]. Research into the molecular function of ASPM indicates that microcephaly associated with ASPM mutation is due a combination of increased apoptosis and impaired self-renewal of neural stem and progenitor cells ^[69–71]. In brain progenitors, Aspm deletion provoked DNA damage and p53-dependent cell death ^[71], while increased cell death did not appear to occur in ASPM-inhibited neuroepithelial cells ^[70]. Altogether, these studies emphasize the disproportionate importance of ASPM in the developing brain compared to other tissues and demonstrate how ASPM disruption decreases proliferation and increases cell death in neural stem/progenitor populations.

In contrast to microcephaly, ASPM expression is increased in epithelial ovarian cancer and malignant gliomas. ASPM protein levels were correlated with epithelial ovarian cancer grade, as seen in cells isolated from human ovarian ascites samples ^[77]. The same authors validated these findings in a later study on primary patient tissue samples ^[78]. In an early study using 120 patient samples of glioblastoma, an unbiased screen identified ASPM as being significantly overexpressed ^[79]. A separate study using RT-PCR (reverse transcription polymerase chain reaction) and immunohistochemistry on tumor samples from 15 patients with low-grade (II) astrocytoma and 15 patients with glioblastoma showed that ASPM mRNA and protein were low in the astrocytomas but much higher in glioblastomas ^[80]. Similar results were obtained from 175 patient glioma samples, representing 8 grade II, 75 grade III, and 92 grade IV gliomas: ASPM mRNA and protein levels increased with tumor grade and were also higher in recurrent tumors ^[81]. siRNA knockdown of ASPM in glioblastoma cell lines reduced proliferation, increased cell death, and sensitized cells to DNA damaging agents like x-rays and cisplatin ^{[79, 81, 82]}. ASPM is also upregulated in medulloblastoma ^[83]. In a primary mouse model of medulloblastoma, conditional Aspm deletion was found to increase DNA damage in tumors and attenuate overall tumorigenesis ^[71]. Further research in vivo is needed to illuminate the exact mechanism by which ASPM disruption reduces tumor formation or destroys tumors. Nonetheless, these studies highlight the shared dependence of neural stem/progenitor and brain tumor cells on ASPM. Despite these promising findings, there are currently no ASPM inhibitors in clinical trials in the United States.

5. CDK6 is mutated in primary microcephaly and upregulated in glioma and medulloblastoma

Data collected on eight generations of a large, consanguineous family in which 10 individuals had apparent primary, autosomal recessive microcephaly found recurrent mutations in CDK6 ^[84]. CDK6 protein levels were not reduced in the microcephalic patients, however its function and behavior were altered. In control human fibroblasts, CDK6 was associated with centrosomes in mitosis (Fig. 1C) – this association was absent in CDK6 mutant fibroblasts, which displayed abnormal mitotic spindle alignment and supernumerary centrosomes that persisted into interphase (Fig. 1C′). Mutant patient fibroblasts had abnormal cell polarity and reduced cell motility in vitro. While these observations have yet to be confirmed directly in neural stem or progenitor cells, the authors detected CDK6 by immunofluorescence in neuroepithelial and neural progenitor cells of the embryonic mouse brain. They hypothesized that microcephaly was a consequence of increased neural stem/progenitor apoptosis and decreased cell proliferation due to reduced RB phosphorylation–phenomena that were seen in the mutant fibroblasts.

The most well-studied role for CDK6 is in control of the cell cycle. CDK6 interacts with cyclin D in G₁-phase to phosphorylate RB ^[85]. This phosphorylation allows cell cycle progression from G₁- to S-phase ^{[86, 87]}. Cyclin D binding is required for CDK6 activity, and the levels of this cyclin are controlled by growth factors and mitogens ^[88]. In turn, CDK6 levels are thought to be regulated by p21 (CIP1), p27 (KIP1), and p57 (KIP2), which may control CDK6 assembly and nuclear translocation ^[89–91]. Besides cyclin D, CDK6 activity is also regulated by its inhibitors p15 (INK4B), p16 (INK4A), p18 (INK4C), and p19 (INK4D) ^[92].

Tight control of CDK6 cell cycle activity is imperative as it is expressed at some point in almost all tissues. Nevertheless, CDK6 has a homolog in mammals, CDK4, that is able to carry out many of the same functions ^{[93, 94]}. Indeed, Cdk6-null mice are viable and mostly develop normally, but suffer from anemia and defective T-cell maturation ^{[95, 96]}. Cdk4-null mice are also viable and instead experience pancreatic and pituitary hypoplasia ^{[97, 98]}. In contrast, loss of both Cdk4 and Cdk6 results in embryonic lethality, underscoring redundant roles for CDK4 and CDK6 ^[96]. The distinct phenotypes of the two null mice, however, suggest some non-overlapping functions for the kinases. Apart from a role in hematopoiesis, CDK6 may regulate the apoptotic program ^[99]. Interestingly, cultured mouse embryonic fibroblasts overexpressing Cdk6 upregulated Tp53 and subsequent exposure to UV irradiation caused rapid, widespread, BAX-mediated apoptosis.

In the CNS, Cdk6 expression is high in cortical progenitor and striatal stem cells and decreases once the cells cease proliferating and differentiate ^[100]. Cdk6 disruption prolonged G₁-phase in these cells and inhibited entry into S-phase. G₁-phase lengthening has been shown in mouse neuroepithelial cells to induce premature differentiation, explaining the decreased proliferation associated with Cdk6 disruption ^[101]. Juvenile mice lacking Cdk6 similarly showed precursor cells in the subventricular zone and subgranular zone of the dentate gyrus with prolonged G₁-phase, premature cell cycle exit, and attenuated proliferation ^[102]. Another role for CDK6 in the CNS is suggested by an experiment where astrocytes explanted from neonatal mice that were induced to overexpress Cdk6 appeared more bi-polar and fibroblast-like than normal, multi-polar astrocytes ^[103]. Unlike in fibroblasts, increased expression of Cdk6 in glia and neurons was not associated with apoptotic priming.

CDK6 upregulation is found in many cancers, including lymphoma, leukemia, bladder cancer, pancreatic cancer, and prostate cancer, as has been reviewed elsewhere ^[104]. In the brain, increased CDK6 expression can be found in malignant glioma and medulloblastoma ^[105–107]. CDK6 expression in gliomas increases with higher tumor grade and is correlated with a worse prognosis ^{[108, 109]}. siRNA knockdown of CDK6 in cultured glioblastoma cells reduced proliferation and increased apoptosis following treatment with temozolomide ^[109]. Two miRNAs, miRNA-495 and miRNA-340, have been identified that suppress CDK6 activity ^{[108, 110]}. They are both downregulated in human glioblastoma and their expression was found to arrest tumor cells in G₁, leading to less proliferation. SUMO1 has also recently been identified to interact with CDK6 in glioblastoma, preventing CDK6 degradation, which permits tumor cell hyper-proliferation ^[111]. As in glioblastoma, CDK6 inhibition in medulloblastoma cell lines was found to induce G₁ arrest due to reduced RB phosphorylation, which attenuates cell proliferation ^[112]. Moreover, CDK6 inhibition significantly limited the survival of medulloblastoma cells exposed to radiation. The role of CDK6 in mitotic progression, as suggested by microcephaly, has yet to be investigated in these cancers.

A multitude of CDK6 inhibitors, though not all specific for CDK6, have been tested in clinical trials over the past twenty years for a variety of cancers ^[113]. Few, however, have been tried in primary brain cancers (Table 2). The CDK4/6 inhibitor palbociclib (PD-0332991) has the most ongoing clinical trials for mostly breast cancers, but also several for primary or metastatic brain tumors. Most promisingly, palbociclib reduced tumor growth in a patient-derived xenograft model of medulloblastoma, extending animal survival ^[118].Ribociclib (LEE-011), another CDK4/6 inhibitor, has active Phase I/II clinical trials for gliomas. Finally, abemaciclib (LY-2835219), also a CDK4/6 inhibitor, has one active Phase I trial for children with diffuse intrinsic pontine glioma. Early results with these agents as monotherapy in non-brain cancers have shown good clinical outcomes, neutropenia and some gastrointestinal toxicity being the most common limiting side-effects. However, anti-tumor effects have generally been modest and not all cancers are responsive. Further research will reveal the feasibility of targeting CDK6 in humans to treat primary brain tumors.

Table 2.

Three CDK6 inhibitors are in active clinical trials for CNS cancers.

Compound	Cancer	Combination	Phase*	Study start	Outcomes; Dose-limiting toxicities
Palbociclib (PD-0332991)	RB+ CNS germ cell tumor	None	II	2009	28% with 24-week progression-free survival; hematological toxicity [114]
	Recurrent RB+ glioblastoma	None	II (T)	2010	No benefit [115]
	Recurrent, refractory, or progressive RB+ CNS tumors in young patients	None	I	2014	Neutropenia/lymphopenia toxicities [116]
	Recurrent oligodendroglioma or oligoastrocytoma	None	II	2015	Unavailable
	HER2+ breast cancer w/brain metastasis	Trastuzumab	II	2016	Unavailable
	Metastatic malignant neoplasm to brain	None	II	2016	Unavailable
Ribociclib (LEE-011)	High grade gliomas after radiation	XRT	I/II	2016	Unavailable
	Recurrent glioblastoma or anaplastic glioma	None	I	2016	Unavailable
	Preoperative glioma and meningioma	None	I	2016	Unavailable
Abemaciclib (LY-2835219)	Solid tumors, including glioblastoma	None	I	2009	18% glioblastoma subjects achieved stable disease; fatigue toxicity [117]
	Diffuse intrinsic pontine glioma	None	I	2015	Unavailable
	Recurrent brain tumors	None	II	2017	Unavailable
	Recurrent glioblastoma	None	II	2017	Unavailable

Information from clinicaltrials.gov: accessed 12/17/17.

^*Indicates current active phase or most recently reported completed phase. S: suspended, T: terminated.

6. ATR is mutated in Seckel syndrome and plays an important role in medulloblastoma tumorigenesis

ATR is a serine/threonine kinase that activates substrates involved in DNA repair, cell cycle arrest, and apoptosis (Fig. 1D) ^[119]. Specifically, ATR responds to the presence of single-stranded DNA, which forms as a normal part of DNA replication and pathologically as a result of DNA damage ^[120]. In vitro studies demonstrate that in undamaged dividing cells, ATR checks genomic integrity at S-phase and slows cell cycle progression to prevent replication stress from causing DNA damage ^[121]. Conversely, single-stranded DNA resultant from DNA damage can cause ATR activation in all phases of the cell cycle ^[122–124]. The presence of DNA damage prompts ATR to act in two possible ways: cell cycle arrest and DNA repair if the damage is minimal, or apoptosis if the damage is extensive ^[125]. Although ATR can promote apoptosis, loss of ATR is actually associated with increased cell death due to the greater role of ATR in mitigating DNA damage and preventing cell death ^[126]. There is some debate as to how cell death proceeds from ATR deficiency. Cleaved Caspase-3 activation in Atr-deleted cells indicates that cell death occurs via apoptosis ^{[127, 128]}. In contrast to typical apoptosis after genotoxic stress, however, ATR-deficient cell death has been suggested to be p53-independent ^{[127, 129]}, which is unexpected, given the pivotal role of p53 in bridging DNA damage with apoptosis ^[130].

The role of ATR in proliferating cells is especially important due to its ability to relieve physiologic replication stress, which would otherwise cause DNA damage ^[131]. Complete loss of ATR is embryonic lethal in mice, as widespread chromosome fragmentation precipitates caspase-mediated apoptosis ^[132]. In the brain, ATR is essential to cerebellar growth, where proliferation of CGNPs continues longer than almost all other neural cell types ^[133]. Conditional deletion of Atr in the embryonic mouse CNS was found to almost exclusively affect two areas of the developing brain – the ganglionic eminence, which produces cortical, striatal, and olfactory cells, and the nascent cerebellum ^[129]. Neural stem cells of the embryonic brain required ATR for normal proliferation and expansion, and its loss resulted in microcephaly, with especially prominent cerebellar hypoplasia, and early postnatal lethality. In the absence of ATR, neural stem cell populations of the embryonic ganglionic eminence and nascent cerebellum experienced extensive DNA damage that led to either apoptosis or impaired proliferation, respectively. Interestingly, inactivation of p53 neither significantly rescued nor exacerbated the ATR phenotype, despite partial attenuation of apoptosis in the forebrain, leading the authors to conclude that neuropathology from ATR loss is not p53-dependent.

The function of ATR appears to be slightly different in more lineage-committed neural progenitors at later stages of development. Our group has found that Atr deletion in CGNPs causes chromosomal damage and widespread apoptosis but no defect in proliferation at the early postnatal stage, when CGNP proliferation is maximal ^[134]. Cerebellar hypoplasia was partially rescued by co-deletion of either Bax and Bak or p53. However, in the Atr/Bax/Bak-deleted mice, pathology persisted due to continued activation of p53 by ATM in CGNPs, leading to p21-mediated cell cycle arrest and premature cell cycle exit. CGNPs in Atr/p53-deleted mice experienced caspase-independent cell death that likely induced cerebellar hypoplasia. Furthermore, ATR loss was found to increase the fraction of CGNPs in M-phase, suggesting an important role for ATR in mediating mitotic entry in CGNPs. This checkpoint role is supported by the finding that Atr deletion significantly increased the fraction of CGNPs with a spectrum of chromosomal abnormalities (Fig. 1D′). Finally, sequencing and pathway analyses demonstrated the critical importance of p53 in mediating the response to ATR loss in CGNPs.

While ATR may have slightly different functions in neural stem cells compared to neural progenitors, the impact of its loss in these cell types clearly demonstrates its importance in brain development. This importance is highlighted by the congenital disorder Seckel syndrome, which is caused by hypomorphic mutations in ATR and characterized by intrauterine growth restriction, cryptorchidism, distinctive facies, pancytopenia, and microcephaly ^{[135, 136]}. Research into the pathogenesis of Seckel syndrome, and especially the effect of ATR deletion on neural stem/progenitor cells, as discussed above, has implications for developing novel brain cancer therapeutics. Apart from Seckel syndrome, ATR mutation has only been associated with one other condition in man, per a single report ^[137]. The authors found in 24 individuals a heterozygous mutation in ATR that attenuated ATR activation of p53, which led to a phenotype consisting of oropharyngeal cancer, skin telangiectasias, and abnormal hair, teeth, and nails.

The unique requirement of cerebellar progenitors for ATR ^[129] suggests that ATR may be an effective therapeutic target for the cerebellar cancer medulloblastoma ^[138]. Activating mutations in the SHH signaling pathway are responsible for 25% of human medulloblastoma ^[139], and CGNPs are thought to be the cells of origin for this molecular subset ^[6]. Indeed, genetically engineered mice with neuronal hyper-activation of the SHH pathway reproducibly develop cerebellar tumors that resemble human medulloblastoma ^[140]. As with other types of cancer cells, medulloblastoma cells suffer from genetic instability owing to rapid rounds of proliferation ^[141]. adiation and chemotherapy, the current standard of care for medulloblastoma, take advantage of this genomic instability by further damaging the DNA of neoplastic cells – enough to push them towards cell death ^{[142, 143]}. However, radiation can be extremely toxic to the normal brain, and chemotherapy damages healthy, endogenous dividing cells such as found in the GI tract and bone marrow ^[144]. These effects can be particularly devastating in young patients.

We have found that Atr deletion in two different mouse models of medulloblastoma either significantly reduced tumorigenesis or completely abrogated tumor formation ^[145]. Our group is currently testing pharmacological ATR inhibition in mice with established medulloblastoma using a novel formulation in which the small molecule ATR inhibitor VE-822 is packaged in polymeric micelle nanoparticles (pVE-822). Preliminary results show that pVE-822 delivered by intraperitoneal injection effectively crosses the blood-brain barrier and induces DNA damage and apoptosis in brain progenitors of neonatal wild-type mice. In older, medulloblastoma-bearing mice, pVE-822 treatment increased tumor DNA damage and reduced tumor size. Ongoing research aims to determine whether pVE-822 produces a survival benefit in mice with medulloblastoma and how pVE-822 may be combined with other therapies to improve treatment outcomes. At this time, VE-822, also known as VX-970 is in Phase I clinical trials with radiation therapy for secondary brain cancer (metastases from non-small cell lung cancer). VX-970 is also being actively tested in Phase I and II trials for other non-brain malignancies (Table 3). In addition, the ATR inhibitor AZD-6738 is currently in Phase I trials for a variety liquid and solid tumors. Similarly, BAY1895344 is in Phase I trials for advanced solid tumors and lymphomas. Thus far, increased ATR activity has only been described in male breast cancer ^[147].

Table 3.

Three ATR inhibitors are in active clinical trials, but not for primary brain malignancies.

Compound	Cancer	Combination	Phase*	Study start	Outcomes; Dose-limiting toxicities
VX-970 (VE-822, M6620)	Advanced solid tumors	Gemcitabine	I	2012	Unavailable
	Advanced solid tumors	None	I	2014	Unavailable
	Uterine/cervical cancer, neuroendocrine tumor, NSCLC, SCLC, ovarian cancer	Topotecan	I/II	2015	Unavailable
	Head & neck squamous cell carcinoma	Radiation & Cisplatin	I	2016	Unavailable
	Metastatic urothelial cancer	Gemcitabine & Cisplatin	II	2016	Unavailable
	Advanced/metastatic solid tumors	Topotecan	I	2016	10% subjects with partial response, 33% with stable disease; hematological toxicities [146]
	Advanced/metastatic solid tumors	Irinotecan	I	2016	Unavailable
	Recurrent ovarian, primary peritoneal, or fallopian tube cancer	Gemcitabine &/or Carboplatin	I/II	2016	Unavailable
	Brain metastases from NSCLC, SCLC, or neuroendocrine tumors	XRT	I	2016	Unavailable
	Refractory solid tumors	Veliparib & Cisplatin	I	2016	Unavailable
AZD-6738	Relapsed/refractory CLL, PLL, or B-cell lymphoma	None	I	2013	Unavailable
	Advanced solid tumors	Carboplatin, Olaparib, or MEDI-4736	I/II	2014	Unavailable
	Metastatic NSCLC	Durvalumab	II	2017	Unavailable
BAY1895344	Advanced solid tumors/lymphomas	None	I	2017	Unavailable

Information from clinicaltrials.gov: accessed 12/17/17.

^*Indicates current active phase or most recently reported completed phase.

A particular benefit of ATR inhibition for the treatment of brain cancers is that in the adult brain, ATR seems to be dispensable for the survival and function of post-mitotic neurons ^{[145, 148, 149]}. It has been shown in mice, however, that ATR is still required for normal proliferation of remaining neural stem cell populations in the adult mouse brain, such as in the hippocampus ^[150]. However, our findings in juvenile medulloblastoma-bearing mice suggest fairly minimal CNS toxicity. In gliomas, although ATR upregulation has yet to be reported, the deleterious effect of ATR disruption in the neural stem cells from which gliomas may derive supports investigating ATR inhibition in these tumors. Here especially, the apparent CNS tolerability for ATR disruption suggests the possibility of an improved therapeutic window for ATR inhibition over traditional chemotherapy.

7. Targeted disruption of microcephaly gene products, combined with standard therapies

A concern remains that inhibitors of mitotic gate-keepers may not synergize well with current treatments involving radiation and cytotoxic chemotherapy. The basis for this concern is that blocking mitosis would be expected to halt cell cycle progression and current therapies for brain tumors target cycling cells ^{[151, 152]}. However, the anti-mitotic agent vincristine has long been paired with alkylating agents and cisplatin in medulloblastoma therapy ^[153]. Moreover, disruption of ASPM and ATR does not block the cell cycle progression of neural stem/progenitor cells ^{[69–71, 145]}. The potential for combining the targeted disruption of microcephaly proteins with conventional radiation and chemotherapy is unknown, and must be investigated before conclusions can be drawn. An obvious advantage of targeted therapy would be reduced toxicity, such as seen in preclinical studies with EG5 inhibition in glioblastoma and ATR inhibition in medulloblastoma.

8. Conclusions

We propose that the proteins encoded by genes that are mutated in microcephaly are critical for growth within the nervous system and may thus be important targets for brain tumor treatment. Of the five microcephaly associated proteins we discuss, three – EG5, ASPM, and CDK6 – play a role in orienting the mitotic spindle in neural stem and progenitor cells ^{[12, 65, 66, 84, 154]}. Our group has shown that ATR similarly has an important function related to neural stem/progenitor cell mitosis, in that ATR controls mitotic entry in CGNPs ^[145]. KIF14, is also upregulated in mitosis and controls cytokinesis ^[51]. That multiple genes mutated in microcephalic disorders code for proteins involved with mitosis is not necessarily surprising since primary microcephaly has been described as essentially a mitotic disorder ^[155]. Rather, these examples emphasize the similar molecular dependencies of neural stem/progenitor cells and brain cancer cells and suggest that the disproportionate importance of mitosis in neural stem/progenitor cells should provide incentive to study mitotic inhibition more intensively as a potential targeted treatment for brain cancers. While prior clinical trials with anti-mitotic agents like Aurora kinase inhibitors and polo-like kinase inhibitors have shown limited success, these studies were almost entirely in non-brain tumors. The limited efficacy of these inhibitors in humans was thought to be due in part to the longer doubling time of tumors in humans (average 114–391 days) compared to in animals (average 3.4–5.7 days) ^[156]. However, primary brain tumors in humans have much shorter doubling times (average 25–50 days) ^[157,158]. This difference in doubling time is likely just one of many reasons that disrupting microcephaly-associated proteins may be more effective in brain tumors than in other tumor types. We propose that inhibiting targets revealed by microcephalic disorders to be particularly required for growth in the brain will bring new efficacy and reduced toxicity to brain tumor treatment.

List of abbreviations

ALL

acute lymphoblastic leukemia

AML

acute myeloid leukemia

CGNP

cerebellar granule neuron progenitor

CLL

chronic lymphocytic leukemia

CML

chronic myeloid leukemia

CNS

central nervous system

ECM

extracellular matrix

MDS

myelodysplastic syndrome

miRNA

micro-RNA

MM

multiple myeloma

NHL

non-Hodgkin’s lymphoma

NSCLC

non-small cell lung cancer

PCL

plasma cell leukemia

PLL

prolymphocytic leukemia

RNAi

RNA interference

SCLC

small cell lung cancer

siRNA

small interfering RNA

XRT

x-ray therapy