Despite marked advances in breast cancer therapy, basal-like breast cancer (BBC), an aggressive subtype of breast cancer usually lacking estrogen and progesterone receptors, remains difficult to treat. In this study, we report the identification of MELK as a novel oncogenic kinase from an in vivo tumorigenesis screen using a kinome-wide open reading frames (ORFs) library. Analysis of clinical data reveals a high level of MELK overexpression in BBC, a feature that is largely dependent on FoxM1, a master mitotic transcription factor that is also found to be highly overexpressed in BBC. Ablation of MELK selectively impairs proliferation of basal-like, but not luminal breast cancer cells both in vitro and in vivo. Mechanistically, depletion of MELK in BBC cells induces caspase-dependent cell death, preceded by defective mitosis. Finally, we find that Melk is not required for mouse development and physiology. Together, these data indicate that MELK is a normally non-essential kinase, but is critical for BBC and thus represents a promising selective therapeutic target for the most aggressive subtype of breast cancer.https://doi.org/10.7554/eLife.01763.001
Not all cancers are the same. There are, for example, at least five types of breast cancer. Different types of cancer can have different mutations and express different genes that determine how aggressively the tumors grow and how well they respond to different therapies. By exploiting these differences, scientists have developed therapies that target specific tumor types, and these targeted therapies have proven useful against most breast cancers.
One type of breast cancer, however, has proven hard to treat. Basal-like breast cancer grows rapidly and there are few treatment options for women with this type of cancer. One reason for this is that, unlike other forms of breast cancer, these cancers do not have the hormone receptors that are the targets of existing therapies.
Enzymes called kinases are promising alternate targets, and many kinase-inhibiting drugs can kill tumor cells in mice. Nevertheless, it has proven difficult to develop kinase inhibitors that are safe for use in humans because these drugs can also kill normal cells. To avoid this side effect, cancer researchers have been searching for a kinase that is active in cancer cells but not in normal cells.
Wang et al. tested a large collection of kinases and found that one called MELK caused tumors to grow in the mammary glands of mice. Further examination of tumor samples collected from hundreds of women in previous clinical studies revealed that MELK expression was increased in basal-like breast cancers and other breast cancer tumors that lack the usual hormone receptor targets.
When Wang et al. treated tumor cells and mice with tumors with a chemical that stops MELK working, basal-like breast cancer cells stopped multiplying and died. On the other hand, tumor cells that had the usual hormone receptors continued to multiply. To see if MELK is important in healthy mice, Wang et al. genetically engineered mice to delete the MELK gene and found that these mutant mice appear normal. The next challenge will be to test if drugs that inhibit MELK can kill basal-like breast cancer cells without having the side effect of harming normal cells.https://doi.org/10.7554/eLife.01763.002
Breast cancer is a heterogeneous disease with a high degree of diversity in histology, therapeutic response, and treatment outcomes. Transcriptional profiling analyses have reproducibly identified at least five major ‘intrinsic’ subtypes of breast cancer: normal breast-like, luminal A, luminal B, HER2/Neu-enriched, and basal-like breast cancer (BBC) (Perou et al., 2000; Sorlie et al., 2001). These molecular subtypes have recently been confirmed in a comprehensive characterization of human breast tumors at the genomic, epigenetic, transcriptomic, and proteomic levels (Cancer Genome Atlas Network, 2012). Among these subtypes, basal-like breast cancer (BBC) is strongly associated with an aggressive phenotype and poor prognosis (Rakha et al., 2008). Unlike their luminal counterparts, BBC cells lack expression of estrogen receptor (ER) and progesterone receptor (PR). Most BBC tumors also lack expression of HER2 and thus this subtype largely overlaps with the clinically defined ‘triple-negative’ breast cancer (TNBC), which is also characterized by the lack of ER, PR, and HER2 expression (Rakha et al., 2008; Foulkes et al., 2010). The lack of these molecular targets renders BBC or TNBC cells relatively unresponsive to the targeted therapies that are highly effective in the treatment of luminal or HER2 positive breast cancer. Thus, establishing the molecular pathogenesis of this subtype and identifying potential targets for treatment remains a key challenge for BBC/TNBC.
Kinases comprise a large family of proteins that is frequently involved in tumor pathogenesis. Indeed, a large number of mutations, alterations in copy number, and/or expression level have been observed in genes encoding kinases across multiple types of human cancers. In addition, kinases have proven to be pharmacologically tractable, making inhibition of kinase activity with small molecules a highly effective strategy for cancer treatment (Zhang et al., 2009). Therefore, identifying kinases critical for the growth and survival of BBC cells could not only provide valuable insights into the pathogenesis of BBC, but also define potential druggable targets for therapeutic interventions.
Kinases that regulate progression through mitosis, including Aurora A, Aurora B and PLK1, are essential for cell proliferation. Inhibiting them in cancer cells causes mitotic arrest and/or abnormalities in chromosome segregation and cytokinesis, which in turn trigger apoptosis (Taylor and Peters, 2008; Lens et al., 2010). Inhibitors of these kinases are effective at eradicating human cancer cells in culture and in mouse xenograft models, but their efficacy in the clinic has been limited by killing of normal proliferating cells especially the bone marrow (Dar et al., 2010). If a kinase exists that is required for mitosis in a specific type of cancer cell, but not other tumor cells or in normal cells, inhibitors of that kinase might make highly effective and safe drugs. To date, this type of cancer-specific mitotic kinase has not been identified for any cancer.
In this study, we report the identification of MELK as a novel oncogenic kinase that emerged from an in vivo tumorigenesis screen. Analyses of breast cancer patient data according to subtypes revealed a remarkable overexpression of MELK in BBC. We further demonstrate that MELK is directly regulated by the FoxM1 transcription factor, a master mitotic regulator also found to be overexpressed in BBC. We discover that MELK is essential in basal-like, but not in luminal breast cancer cells. Notably, mice in which MELK has been genetically ablated display normal development and hematopoiesis. Together, our data establish MELK as a mitosis-regulating kinase involved in the pathogenesis of BBC and a promising molecular target for patients with basal-like breast malignancy.
Transformation of primary human cells with defined genetic elements is a powerful method for identifying specific genes or pathways that are involved in oncogenic transformation (Hahn et al., 1999; Zhao et al., 2004). To this end, we first developed an in vivo tumorigenesis system that models the pathogenesis of human breast cancer, using a previously established human mammary epithelial cell (HMEC)-based transformation system (Zhao et al., 2003). To further optimize this system, we engineered telomerase-immortalized HMECs to express a dominant negative form of p53 (p53DD), NeuT and PI3KCA H1047R. The resulting cells, termed HMEC-DD-NeuT-PI3KCA, were fully transformed as evaluated by their ability to form orthotopic tumors in the mammary fat pads of mice (Figure 1—figure supplement 1). Our model recapitulates the concurrent activation of HER2/Neu and PI3KCA that is prevalent in breast cancer (Stephens et al., 2012).
The HMEC transformation system described above provided us with a platform to identify novel oncogenic events capable of replacing the mutant PIK3CA in cooperating with NeuT to drive HMECs to form tumors in mice. To this end, we infected HMEC-DD-NeuT cells (lacking the mutant PIK3CA) with subpools of a kinome-wide retroviral library consisting of 354 human kinases and kinase-related open reading frames (Boehm et al., 2007). The library was screened as a series of subpools of 10–12 kinase ORFs in HMEC-DD-NeuT cells. The infected cells were injected into the inguinal mammary fat pads of mice, and recipient mice were followed for tumor formation. Kinases in 12 pools induced tumor formation with latencies of 2–4 months. Genomic DNA was extracted from harvested tumor specimen as well as HMECs infected with matched pools of kinases prior to injection. We then used quantitative PCR to determine the relative abundance of each kinase in these paired samples. In total, 26 kinases were found specifically enriched in the tumors in vivo (Figure 1, Figure 1—figure supplement 2). Several candidate kinases that scored in the screen have previously been implicated as proto-oncogenes or cancer-associated genes, such as the inhibitor of nuclear factor kappa-B kinase subunit epsilon (IKBKE) (Boehm et al., 2007), rearranged during transfection (RET) (Takahashi et al., 1985), casein kinase 1 epsilon (CSNK1E) (Kim et al., 2010), NIMA-related serine/threonine kinase 6 (NEK6) (Nassirpour et al., 2010), and polo-like kinase 1 (PLK1) (Liu et al., 2006). At least three of them, PLK1 (Golsteyn et al., 1994), NEK6 (Yin et al., 2003), and MELK (maternal embryonic leucine zipper kinase) (Le Page et al., 2011), have been previously implicated in regulating mitotic progression.
One of the top-scoring hits from our genetic screen was MELK (Figure 1), an atypical member of AMPK serine/threonine kinase family (Lizcano et al., 2004). While little is known about the exact biological functions of MELK, this kinase has been reported to be overexpressed in a variety of tumors (Gray et al., 2005). When we analyzed MELK expression in the breast cancer data set of The Cancer Genome Atlas (TCGA) (Cancer Genome Atlas Network, 2012), a large cohort consisting of 392 invasive ductal breast carcinomas and 61 samples of normal breast tissues, the level of MELK transcript was approximately eightfold higher in breast tumors compared to their normal counterparts (Figure 2A). The p value for this differential expression (4.6 × 10−54) places MELK in the top 1% overexpressed genes in breast cancer (Figure 2A). The overexpression of MELK in breast tumors relative to normal breast tissues was further confirmed by analyzing two other independent data sets (Figure 2—figure supplement 1A; Ma et al., 2009; Richardson et al., 2006).
To gain insights into the potential relevance of MELK overexpression in breast cancer, we asked whether MELK expression correlates with the status of disease. By analyzing gene expression data across five independent cohorts totaling more than 1500 patients (Desmedt et al., 2007; Hatzis et al., 2011; Schmidt et al., 2008; Wang et al., 2005b; Supplementary file 1), we found that higher expression of MELK was strongly associated with higher histologic grade in breast cancer (Figure 2B, Figure 2—figure supplement 1B); the p values for this correlation rank in the top 1% of a total 12,624 or more genes measured in all these cohorts.
We also examined whether MELK expression is correlated with metastatic recurrence. We analyzed three independent cohorts in which patients with early-stage breast cancer were followed for metastasis-free survival and had not received adjuvant systemic treatment after surgery (van 't Veer et al., 2002; Wang et al., 2005b; Schmidt et al., 2008; Supplementary file 1). In all three cohorts, higher MELK expression levels were strongly associated with earlier metastasis in women initially diagnosed with lymph-node-negative tumors (all p values<0.001, hazard ratios >2; Figure 2C, Figure 2—figure supplement 1C). We further analyzed two cohorts, where a majority of patients had high grade and lymph-node-positive breast cancer and nearly all patients received neoadjuvant chemotherapy and/or hormone therapy (Hatzis et al., 2011; Loi et al., 2007; Supplementary file 1). Again, high expression level of MELK robustly correlates with metastasis in breast cancer patients (both p values<0.001, hazard ratios >2; Figure 2C). Thus MELK overexpression appears to have a strong predictive value for breast cancer metastasis irrespective of tumor grade or treatment regimen.
We next asked if MELK expression also correlates with the survival of breast cancer patients. In five independent large cohorts in which more than 1100 total patients were followed for overall survival (Desmedt et al., 2007; Esserman et al., 2012; Kao et al., 2011; Pawitan et al., 2005; van de Vijver et al., 2002; Supplementary file 1), high expression level of MELK strongly correlated with increased rates of mortality (all p values<0.05, hazard ratios >2) (Figure 2D, Figure 2—figure supplement 1D). Together, these data show that MELK may serve as a prognostic indicator in predicting breast cancer patients' likelihood of metastasis and overall survival rate.
Given the heterogeneity of breast cancer, we analyzed MELK expression in different subtypes of breast cancer as defined by gene expression profiling (Perou et al., 2000; Sorlie et al., 2001). We categorized samples in multiple breast cancer data sets by PAM50 gene signature (Parker et al., 2009). In five independent cohorts with more than 1500 patients in total, we observed a strikingly similar pattern of MELK expression among these different subtypes of breast tumors (Figure 2E, Figure 2—figure supplement 1E). While luminal A and normal-like subtypes displayed the lowest expression of MELK, basal-like breast cancers (BBC) showed the highest expression level of MELK among all subtypes (p<0.0001). Given that there are more high-grade tumors in the BBC than the other subtypes, we sought to determine the correlation of MELK with subtypes of breast tumors within the same grade. We performed statistical analysis of a large cohort of breast cancer for grade 1, 2 and 3 across all subtypes, respectively, and found that MELK is most highly expressed in BBC (Figure 2—figure supplement 2A), suggesting that MELK expression is most pronounced in basal-like breast tumors with the same pathological grade. Moreover, a significant association of MELK expression with disease status also exists within the subtype of BBC (Figure 2—figure supplement 2B), suggesting that MELK expression is associated with tumor aggressiveness and poor prognosis in this disease.
Consistent with this observation of MELK overexpression in BBC, we found that MELK expression in breast tumors has a significant inverse correlation with the expression of luminal markers, including estrogen and progesterone receptors (ER, PR) (Figure 2F, Figure 2—figure supplement 1F). To confirm this observation at the protein level, we analyzed primary tumors samples for MELK expression. Strikingly, all the four ER/PR+ tumor samples lacked detectable signal of MELK expression. In contrast, ER/PR-negative tumors had abundant MELK protein (Figure 2G). Given that ER/PR expression varies within the molecular HER2+ subtype, we analyzed MELK expression within this subtype. We found that MELK expression was significantly higher in ER/PR− tumors than in those with ER/PR+ status (Figure 2H).
An alternate categorization of breast cancers uses the expression of ER/PR and HER2. Triple-negative breast cancer (TNBC), a subtype lacking ER/PR and HER2 expression, largely overlaps with basal-like breast cancer (Rakha et al., 2008; Foulkes et al., 2010). Because this subtype-categorization has been routinely used in the clinic for diagnosis and selection of treatment strategies, we also examined whether MELK expression correlates with this alternate subtype categorization. In two independent cohorts, the expression level of MELK is the highest in TNBC (Figure 2I, Figure 2—figure supplement 1G). Again, within the HER2+ sub-group, ER/PR− tumors have much higher MELK expression than ER/PR+ ones (Figure 2I). Together, these data indicate that MELK expression is highly elevated in breast tumors lacking the expression of ER and PR luminal markers.
To investigate the mechanism underlying MELK overexpression in BBC, we first analyzed the copy number of MELK in breast cancer. Gene amplification of MELK occurs in both primary tumors and human breast cancer cell lines, especially in ER-negative samples (Figure 3—figure supplement 1). These tumors or cells with increased copy number of MELK also exhibit high level of MELK expression, suggesting that gene amplification contributes to the overexpression of MELK.
However, gene amplification of MELK occurs at low frequency, and does not explain the widespread overexpression of MELK in BBC. Recent comprehensive profiling of breast cancer suggests a role for FoxM1 activation in the transcriptional maintenance of BBC (Cancer Genome Atlas Network, 2012). We found that like MELK, FoxM1 is most highly expressed in the BBC or TNBC subtypes (Figure 3A, Figure 3—figure supplement 2A,B). Moreover, an extremely tight correlation between FoxM1 and MELK expression was observed in multiple large-sized cohorts (Figure 3B, Figure 3—figure supplement 2C). FoxM1 downregulation via gene silencing or a chemical inhibitor, thiostreptoin (Hegde et al., 2011), reduced MELK expression (Figure 3C,D, Figure 3—figure supplement 2D). Furthermore, we found that the promoter of MELK contains a putative FoxM1 binding motif (Wierstra and Alves, 2007), and chromatin immunoprecipitation assays using a FoxM1-specific antibody recovered a MELK promoter region that included the putative binding site (Figure 3E). Together these data suggest that FoxM1 is a transcription factor that is enriched in BBC and regulates MELK expression, providing a molecular mechanism underlying the overexpression of MELK in BBC.
Given that FoxM1 is a master transcription factor for many genes that are essential for mitosis (Laoukili et al., 2005; Wang et al., 2005a), our finding also suggests that MELK is a mitotic factor in BBC cells. Proteins required for mitotic progression typically accumulate during G2 and M phase, and are destroyed by ubiquitin-dependent proteolysis at the end of cytokinesis. A previous study reported that MELK is stabilized in mitosis and partially degraded upon mitotic exit in HeLa cells and Xenopus embryos (Badouel et al., 2010). We found that, in BBC cells, MELK was highly expressed during mitosis, and its protein abundance decreased dramatically when mitotic cells progressed into G1 phase (Figure 3F, Figure 3—figure supplement 3). This expression pattern of MELK, which is similar to that of Cyclin B1 and Aurora kinases, indicates that MELK is a mitotic kinase in BBC cells. Interestingly, while luminal breast cancer cells have a similar pattern of MELK expression during cell cycle, their MELK protein levels in the M phase are much lower than those of BBC cells (Figure 3G). The expression levels of other mitotic factors including Cyclin B1 and Aurora A are comparable between basal-like and luminal cancer cells (Figure 3G), suggesting that MELK may play a unique role during mitosis in BBC cells.
MELK was scored in our kinase library screen and is overexpressed in breast cancer, particularly in basal-like breast tumors. Therefore, we sought to further determine the potential oncogenic role of MELK. To this end, we re-engineered HMEC-DD-NeuT cells to express wild type (WT-) or myristoylated (myr-) MELK (the kinases in our initial screen were myristoylated, Boehm et al., 2007). While HMEC-DD-NeuT cells expressing the empty vector failed to form tumors in mice, overexpression of either WT- or myr-MELK in these cells drove tumor formation with 100% penetrance within 2 months (Figure 4—figure supplement 1), demonstrating that overexpression of MELK was able to confer the tumorigenicity of HMEC-DD-NeuT cells.
MELK expression strongly correlates with cell proliferation (Venet et al., 2011), indicating a functional role of MELK for cell growth. Indeed, we found that MELK overexpression in non-transformed HMEC-DD cells resulted in increased cell proliferation in suspension culture (Figure 4—figue supplement 2A). While oncogenic PIK3CA, Ras, or NeuT alone can induce colony formation of HMEC-DD cells in soft agar, two oncogenic events (e.g., PIK3CA plus NeuT), are usually required to fully transform HMEC-DD cells to form tumors in mice (Zhao et al., 2003, 2004). Similar to these oncogenes, over-expression of MELK alone can also promote anchorage-independent growth of HMEC-DD and MCF10A cells (Figure 4—figure supplement 2B–E). Likewise, wild-type MELK cooperates with a second oncogene, for example NeuT, to induce tumor formation in vivo.
In contrast to the transformation of HMEC-DD cells, one oncogenic event is sufficient to transform Rat1 rodent fibroblasts expressing p53DD (Rat1-DD) cells, as indicated by both anchorage-independent growth in vitro and tumor formation in vivo (Ni et al., 2012). To determine whether MELK has a transforming activity as a single event in this system, we engineered Rat1-DD cells expressing MELK (Rat1-DD-MELK), or PI3KCA H1047R (Rat1-DD-PI3KCA H1047R) as a positive control (Figure 4A). As expected, Rat1-DD cells transduced with an empty vector failed to grow as colonies in soft agar or to form tumors in mice. Strikingly, Rat1-DD-MELK cells displayed a robust transformed phenotype comparable to Rat1-DD-PI3KCA H1047R cells, as evidenced by both colony growth in vitro and tumor formation in vivo (Figure 4B,C).
To determine whether the transforming ability of MELK requires its kinase activity, we introduced catalytically inactive alleles of MELK, D150A or T167A (Lizcano et al., 2004; Vulsteke et al., 2004), into Rat-DD cells. Unlike Rat1-DD-MELK cells, both Rat1-DD-MELK-D150A and Rat1-DD-MELK-T167A cells exhibited only limited growth in soft agar or in mice (Figure 4D–F). Together, these studies indicate that MELK can be a potent oncogenic driver, when it is aberrantly overexpressed and that this oncogenic potential relies on its kinase activity.
Since MELK is predominantly overexpressed in basal-like breast tumors, we sought to determine whether MELK plays a role in the proliferation of BBC cells. We first analyzed a set of breast cancer cell lines that mirror the molecular subtypes of clinical tumors (Neve et al., 2006), and found that the expression level of MELK is much higher in the cohort of 23 BBC cell lines than in the cohort of 24 luminal breast cancer cell lines (Figure 5A). This is consistent with the expression pattern of MELK in primary human breast tumors. We also confirmed that the protein abundance of MELK is much higher in BBC cells compared to luminal cells (MCF7 and T47D) (Figure 5B). These cell lines thus, provide an excellent platform to assess potential roles of MELK in BBC.
To examine the role of MELK in the proliferation of these cells, we utilized a tetracycline-inducible gene knockdown technique in which shRNA transcription (and consequently target gene silencing) is induced upon exposure of the targeted cells to doxycycline (Wiederschain et al., 2009). Among the multiple inducible shRNAs that we generated to target MELK, two shRNAs were found to efficiently reduce MELK expression in cells treated with doxycycline (Figure 5C). We then stably introduced both shMELKs into basal or luminal breast cancer cell lines. Cell proliferation was measured upon induction of the shRNA in the presence of doxycycline. As hypothesized, MELK knockdown strongly impaired the growth of all six BBC cell lines tested, including BT549, MDA-MB-468, MDA-MB-231, MDA-MB-436, HCC70, and HCC1954 (Figure 5C,D, Figure 5—figure supplement 1A,B). In contrast, in five luminal breast cell lines, MELK knockdown with equivalent efficiency did not result in obvious inhibition on cell growth (Figure 5E, Figure 5—figure supplement 1A,B). Similarly, MELK knockdown had little effect on non-transformed HMECs, which have low level of MELK expression (Figure 4—figure supplement 2).
To further validate the essential role of MELK in BBC cells, we carried out rescue experiments with both WT and kinase inactive MELK. We found that the proliferation of MDA-MB-468 cells expressing shMELK was restored, when the MELK expression level in these cells was rescued by expression of a shMELK-resistant allele of MELK (MELK-R) (Figure 5F, Figure 5—figure supplement 3), confirming that the effects of shRNA are due to the specific knockdown of MELK. Notably, expression of a kinase inactive version of MELK-R, MELK-R (T167A), failed to restore cell proliferation in these cells (Figure 5F), indicating that kinase activity of MELK is critical for the proliferation of these BBC cells.
To understand the mechanism(s) underlying the MELK function in BBC cells, we examined how inducible shRNA-mediated MELK depletion affects various cellular processes. In the presence of doxycycline, BBC cells underwent cell death indicated by increased apoptotic markers, including cleaved caspase 3, cleaved PARP, and DNA fragmentation (Figure 6A,B, Figure 6—figure supplement 1A). zVad, a pan-caspase inhibitor, was able to rescue cell death, indicating an active role of caspases in executing cell death upon MELK depletion (Figure 6C, Figure 6—figure supplement 1B). In contrast, MELK knockdown has little effect on luminal tumor cells, such as MCF7 (Figure 6D). To complement our RNAi-mediated MELK knockdown studies, we used a recently developed chemical inhibitor of MELK, OTSSP167 (Chung et al., 2012), to evaluate the functional dependency on MELK by basal and luminal breast cancer cells. Consistently, OTSSP167 induced apoptotic cell death selectively in basal breast cancer cells (Figure 6—figure supplement 2A–C).
The finding that MELK is a mitotic kinase in BBC cells prompted us to hypothesize that the cell death observed upon MELK inhibition might be due to an altered cell cycle progression. MELK knockdown by doxycycline induced an accumulation of cells with 4n DNA content (Figure 6E, Figure 6—figure supplement 3A), indicating an induction of G2/M arrest or failure of cytokinesis. By immunoblotting, we found that cells exposed to doxycycline exhibited an elevation of Cyclin B1 and Aurora A kinase, two markers of G2/mitosis (Figure 6E, Figure 6—figure supplement 3A). We next used microscopy to define the cell division defect in more detail. Doxycycline induced a nearly twofold increase in the percentage of cells with two or more nuclei (Figure 6F,G, Figure 6—figure supplement 3B), indicating a failure of cytokinesis. Indeed, our time-lapse microscopic analysis revealed binucleated cells forming after impaired cytokinesis (Video 1). Furthermore, cells in which MELK had been depleted displayed asymmetric division (Figure 6G), characterized by an unequal allocation of cell mass into daughter cells. Interestingly, Caenorhabditis elegans with mutations in the MELK homologue, PIG-1, demonstrate impaired asymmetric cell division (Cordes et al., 2006), supporting a critical role of MELK in the late stage of cell division.
We next used time-lapse microscopy of GFP-Histone 2B expressing cells (Kanda et al., 1998) to determine whether apoptosis and defective mitosis due to MELK knockdown are functionally associated. Cell death events were dramatically increased upon MELK knockdown (5 out of 235 cells in control, 151 out of 317 in doxycycline-treated cells during 10 hr of imaging). Moreover, cell death events were often preceded by division abnormalities in doxycycline-treated populations. Cells with double nuclei, which had presumably failed cytokinesis, often underwent cell death (Figure 6H, middle panel; Video 2). Some cells with an apparently normal metaphase plate were unable to progress towards anaphase, instead entering into the process of cell death directly from mitosis (Figure 6H, bottom panel; Video 3). Overall, following MELK knockdown, out of 27 examples we noted 16 failed mitoses among which 10 proceeded to cell death after the formation of metaphase plate and six gave rise to binucleated cells. By contrast, mitosis in 36 out of a total of 37 control cells appeared normal (Figure 6H, top panel; Video 4). The morphological events associated with failed cell division and ensuing cell death resembled the previous reports of the effects of inhibiting essential mitotic kinase such as Aurora B (Keen and Taylor, 2009). Together, these data suggest a model in which BBC cells rely on MELK for proper mitosis; inhibiting MELK in these cells causes impaired mitosis and consequent cell death.
Since MELK is selectively required for the survival of BBC cells, we sought to determine whether MELK also supports the oncogenic growth of BBC cells using both in vitro colony formation and in vivo xenograft tumor growth assays. While MDA-MB-468 and MDA-MB-231 cells readily grew into macroscopic colonies in soft agar, MELK knockdown in these cells upon doxycycline treatment caused a nearly complete inhibition in colony formation (Figure 7A). To determine whether MELK is also important for BBC cells to grow as tumors in vivo, we transplanted BBC cells expressing inducible shMELK into the mammary fat pads of athymic mice to allow orthotopic tumor formation. While all recipient mice in the control group without doxycycline treatment developed tumors within 2 months, mice treated with doxycycline immediately following transplantation failed to develop tumors (Figure 7B), suggesting that MELK is required for the proliferation of these BBC cells in vivo. To further examine whether MELK is required for the maintenance of established tumors, we administered doxycycline to mice bearing xenograft tumors derived from basal-like or luminal breast cancer cells. Remarkably, down-regulation of MELK led to a substantial regression of tumors arising from BBC cells but had little effect on tumors derived from luminal cancer cells (Figure 7C,D, Figure 7—figure supplement 1).
To determine if the pharmacological inhibition of MELK would recapitulate the effect of MELK knockdown in xenografts, we administered OTSSP167 or vehicle control to mice that have tumors derived from basal or luminal breast cancer cell lines. While the growth of luminal breast tumors were largely unaffected by the treatment of OTSSP167, the chemical caused significant inhibition on the growth of basal breast tumors (Figure 7E,F). Together, these data indicate the MELK is selectively required for the oncogenic growth of BBC cells, and suggest that MELK inhibition could be an effective approach in treating basal-like breast cancer.
While MELK has a critical role in basal-like breast cancer, it is not clear whether this kinase is important for the proliferation of normal cells or tissue growth in vivo. This question is critical to the toxicity of any potential MELK-targeted therapy. To address this question, we generated mice with germ-line knockout (KO) of Melk (Figure 7G). Notably, Melk-deficient mice are viable and appear normal without any noticeable phenotypes in the development of the embryos or adult mice. Both male and female mice are fertile and produce litters of normal size. A study by Hebbard et al. found a high activity of Melk promoter in mouse mammary progenitor cells (Hebbard et al., 2010). Therefore, we anticipated an impairment of mammary gland development in Melk KO mice. However, mammary glands from mice with Melk KO appear normal in both morphology and function (e.g., lactation).
Since bone marrow toxicity is a major side effect of most anti-cancer therapies and, in particular, of those drugs targeting mitotic kinases/machinery, we characterized the immune system in Melk-deficient mice. We isolated cells from bone marrow, spleen and thymus, and analyzed immune cell populations including monocytes, neutrophils, B and T cells. We also analyzed the hematopoietic stem cells and various progenitor cells in the bone marrow. In both cases, virtually no differences were found between wild-type and KO mice (Figure 7H,I). Together, these data suggest that Melk is not essential for normal development and physiological functions in mice, providing compelling evidence for MELK as a highly selective target for therapeutic intervention of basal-like breast cancer.
Patients with basal-like breast cancer remain faced with limited treatment options due to the aggressive nature of the disease and the current lack of suitable molecular targets for therapeutic intervention. In this study, we report that MELK, a novel oncogenic kinase that emerged from an unbiased, in vivo tumorigenesis screen, may indeed be a therapeutic target in this tumor type. In a comprehensive analysis of databases with multiple cohorts of breast cancer, we find MELK to be highly overexpressed in breast cancer lacking the expression of ER/PR, including basal-like breast cancer. Remarkably, overexpression of wild-type MELK induces robust oncogenic transformation both in vitro and in vivo with a transforming potency comparable to that of the highly oncogenic mutant allele of PIK3CA. Even more striking is the finding that only basal-like, but not luminal breast cancer cells, depend on MELK for proliferation. In addition, the dispensable nature of Melk in normal development and hematopoiesis in mice underlines its selective role in BBC. Notably, the kinase activity of MELK is required for its transforming activity as well as for the survival and proliferation of BBC cells. Thus, MELK is potentially a novel oncogenic driver of basal-like breast carcinoma and a promising target for small molecule-based therapeutic intervention.
Our data point to a potential role for MELK as a marker in predicting disease outcome. In multiple independent breast cancer cohorts analyzed, we found a strong association of high expression levels of MELK with a higher grade of malignancy and an unfavorable prognosis regardless of the treatment modality. While high MELK expression seems to be a unique phenomenon for BBC in breast cancer, MELK overexpression has been associated with tumor aggressiveness and poor outcome in a number of other cancer types, including glioblastoma (Nakano et al., 2008), astrocytoma (Marie et al., 2008), and prostate cancer (Kuner et al., 2013). The prognostic feature of MELK expression is likely due to its correlation with cell proliferation. In fact, MELK and other proliferation-related genes are major components of multi-gene signature for predicting disease outcome. For example, a recent study developed a cell proliferation signature that consists of the top 1% genes whose expression is most positively correlated with that of proliferating cell nuclear antigen (PCNA). The authors found that adjusting breast cancer expression data for this cell proliferation signature causes a dramatic reduction in outcome association of most published breast cancer signatures (Venet et al., 2011). Notably, MELK expression strongly correlates with cell proliferation, and in fact is one of the top-ranking signature genes of cell proliferation that correlate with PCNA expression (Venet et al., 2011).
Previous studies demonstrated that, while MELK is a member of the AMPK family, it is not activated via phosphorylation by the tumor suppressor kinase LKB1 (Lizcano et al., 2004). Recombinant MELK expressed in bacteria is catalytically active (Davezac et al., 2002; Lizcano et al., 2004; Beullens et al., 2005). Consistent with these findings, overexpression of wild-type MELK readily drives transformation in vitro and in vivo. This behavior is similar to that of other established proto-oncoproteins, such ERBB2 (Di Fiore et al., 1987; Hudziak et al., 1987), and Aurora A kinase (Bischoff et al., 1998; Zhou et al., 1998), the transforming activity of which is driven by overexpression of the wild-type protein. While a number of substrates have been proposed for MELK, such as Bcl-G (Lin et al., 2007), CDC25B (Davezac et al., 2002), p53 (Seong and Ha, 2012), and PDK1 (Seong et al., 2012), the substrates that mediate the oncogenic activity of MELK in breast cancer remain to be identified.
An intriguing question is how the selective overexpression of MELK is achieved in BBC. Our finding that the mitotic transcription factor FoxM1 (Laoukili et al., 2005; Wang et al., 2005a) plays a major role in regulating MELK expression has shed some light on this enigma. Notably, the expression levels of FoxM1 and MELK demonstrate a striking correlation across all breast cancer samples and subtypes examined. Like MELK, FoxM1 is significantly overexpressed in BBC. Consistent with our results, FoxM1 was recently proposed as a transcriptional driver of proliferation-associated genes in BBC (Cancer Genome Atlas Network, 2012).
However, why MELK is selectively required for cell division in BBC cells, but not in other types of breast cancer or normal cells, remains an open question. MELK was not observed as a hit in systematic screens for essential cell division proteins in HeLa cells (Kittler et al., 2004). Both C. elegans and mice are tolerant of mutation or deletion of MELK ortholog (Cordes et al., 2006; Figure 7G–I). However, MELK can be essential in some circumstances. It is expressed in early frog embryos, where it seems to play some role in cell division (Le Page et al., 2011), and we observed it accumulating in dividing cells (Figure 3F), and playing an important role during cell division in BBC cells (Figure 6). To reconcile these apparently disparate findings, we propose that one or more MELK-related kinase is required for cell division in many, if not all vertebrate cells. In BBC cells, MELK must play this role uniquely and is selectively overexpressed, perhaps because redundant kinases are down-regulated. In other cells, MELK may function during division, but it is not essential due to redundancy with related kinases. Consistent with this hypothesis, the MELK-related kinase AMPK was recently shown to play a role in mitosis (Vazquez-Martin et al., 2009). Perhaps AMPK, or other kinases in the same family, can substitute for MELK in some cells, but not in BBC cells, which seem to have become addicted to MELK for proper execution of cell division. Determining the precise function of MELK in cell division, and the reason this function is selectively required in BBC cells, will require further analysis. Nevertheless, our studies firmly establish MELK as a molecular target for the treatment of BBC. Unlike other mitotic factors like Aurora A, Aurora B, and PLK1, that are normally essential, MELK presents a unique mitotic kinase that is only required by a subset of cancer cells, and is therefore an excellent therapeutic target.
In summary, recent comprehensive characterization of basal-like breast cancer demonstrates that this subtype of disease has high genetic heterogeneity, but lacks commonly occurring genetic alterations, with the exception of the frequent inactivation of p53 (Cancer Genome Atlas Network, 2012). In contrast, the relative uniform overexpression of MELK in basal-like breast cancer makes it a potential common target in an otherwise heterogeneous disease. Thus our data on MELK provide important information for guiding the development of targeted therapies in basal-like breast cancer.
The human MELK was amplified using the template DNA deposited in the described kinase library, and cloned into pWZL retroviral vector (Zhao et al., 2003), in which target gene expression is driven by the long terminal repeat of Moloney murine leukemia virus. The MELK mutants (D150A, T167A, or shMELK-resistant MELK with silent mutations) were generated via Quickchange XL Site-directed Mutagenesis (Stratagene, La Jolla, CA). Primers were listed in Supplementary file 2.
To construct a tetracycline-inducible gene expression system, GFP or mutated MELK was amplified using the primers listed in Supplementary file 2. The PCR products were digested with AgeI and PacI, and ligated with digested pLKO-TREX (Wee et al., 2008).
To construct pWzl-H2B-GFP, human Histone 2B was amplified using the genomic DNA of HEK293T cells as templates. Primers for cloning were listed in Supplementary file 2. PCR products following digestion with BamHI and XhoI were ligated with digested pWzl-GFP.
To generate pLKO-tet-on-shRNAs targeting human MELK, oligonucleotides were designed and synthesized (IDT, Coralville, Iowa). Following annealination, double-stranded oligonucleotides were directly ligated with pLKO vector that was digested with AgeI and EcoRI. The sequences for scramble, shMELK1, shMELK2 are listed in Supplementary file 2.
Retroviruses were generated by transfecting HEK293T cells with pWzl plasmids and packaging DNA. Typically 1.6 μg pWzl DNA, 1.2 μg pCG-VSVG and 1.2 μg pCG-gap/pol, 12 μl lipid of Metafectene Pro (Biontex, Martinsried, Germany) were used; DNA and lipid were diluted in 300 μl PBS respectively and mixed; and following 15 min of incubation, they were added to one 6-cm dish that was seeded with 3 million HEK293T cells 1 day earlier. Viral supernatant was collected 48 hr and 72 hr after transfection. After the supernatant was filtered through 0.45-μm membrane, it was added to target cells in the presence of 8 μg/ml polybrene (Millipore, Billerica, MA). Lentiviruses were generated with a similar approach with the exception of HEK293T cells that were transfected with 2 μg pLKO DNA, 1.5 μg pCMV-dR8.91, and 0.5 μg pMD2-VSVG. Cells were selected with antibiotics starting 72 hr after initial infection. Puromycin and blasticidin were used at the final concentrations of 1.5 μg/ml and 4 μg/ml respectively.
Human mammary epithelial cells (HMECs) were maintained in DMEM/F-12 supplemented with EGF (10 ng/ml), insulin (10 μg/ml), and hydrocortisone (0.5 μg/ml) under 5% CO2 and 37°C. Rat1 and HEK293T cells were maintained in DMEM supplemented with 10% FBS (Invitrogen, Carlsbad, CA). All breast cancer cell lines (MCF7, T47D, MDA-MB-468, MDA-MB-231, MDA-MB-436, HCC1197, BT549) were cultured in RPMI 1640 medium supplemented with 10% FBS. For cells stably introduced with tetracyclin-inducible genes/shRNAs, Tet-approved FBS (Clontech, Mountain View, CA) was used.
Typically, breast cancer cells were seeded in 12-well plates (1–2 × 104) in 1 ml medium. On the next day, wells were added with 110 μl medium without or with 1 μg/ml doxycycline (to reach a final concentration of 100 ng/ml), which was repeated every 2 days. 6 days after the initial treatment, cells were fixed with formaldehyde, and stained with crystal violet (0.05%, wt/vol), a chromatin-binding cytochemical stain. The plates were washed extensively, and imaged with a flatbed scanner. For quantification of the staining, 1 ml 10% acetic acid was added to each well to extract the dye. The absorbance was measured at 590 nm with 750 nm as a reference.
The assays were typically performed in a 12-well plate unless otherwise mentioned. Cells were suspended in medium containing 0.3% agar and plated onto a layer of 0.6% agar (for each well, 4000 cell in 800 μl medium, 1 ml bottom agar). The wells were added with medium (without or with 100 ng/ml doxcycycline) on the next day. 3 weeks after seeding, the colonies were fixed with formaldehyde and imaged. The number of colonies in each well was quantified using ImageJ (National Institutes of Health).
All xenograft studies were conducted in accordance with the animal use guidelines from the National Institutes of Health and with protocols approved by the Dana-Farber Cancer Institute Animal Care and Use Committee. The recipient mice used were NCR-nude (CrTac:NCr-Foxn1nu, Taconic, Hudson, NY). Cells were resuspended in 40% of Matrigel-Basement Membrane Matrix, LDEV-free (BD Biosciences, San Jose, CA) and sit on ice until injection. For transplanting human cell lines, mice were γ-irradiated with a single dose of 400 rads on the same day of injection. Mice were anesthesized by inhalation of isoflurane, and were injected with 150 μl cells (5 × 106) per site. Tumors were measured in two dimensions by a caliper. Tumor volume was calculated using the formula: V = 0.5 × length × width × width. All xenograft data are presented as mean ± SEM. Comparison between groups of treatment were conducted using two-tailed Student's t test. Calculations were performed using either Openoffice or GraphPad Prism version 5.0b.
For tumorigenesis study, 5 × 106 HMEC cells were injected into the mammary fat pad, and 5 × 106 Rat1 cells subcutaneously. Tumor growth was monitored twice a week. Rat1 xenografts were harvested 3 weeks after injection.
To study the impact of MELK knockdown on tumor growth, mice were randomly sorted into groups on the second day of injection, and were untreated or treated with doxycycline (2 mg/ml in 5% dextrose in drinking water, refreshed twice a week) for the duration of the study. Tumor was measured twice a week.
To study the roles of MELK in tumor maintenance, mice with established tumors (≥200 mm3) derived from orthotopic injections of MDA-MB-231, or MDA-MB-468, or MCF-7, or T47D cells were randomly sorted into two groups, with one group receiving doxcycline in drinking water. Tumors were calipered twice per week to monitor the effect of MELK knockdown on tumor growth.
Time-lapse imaging was performed on a Nikon Ti motorized inverted microscope, which was equipped with a perfect focus system and a humidified incubation chamber (37°C, 5% CO2) (Nikon Imaging Center, Harvard Medical School). Cells stably expressing H2B-GFP were pre-seeded in 24-well glass-bottom plate, and either untreated or treated with doxycyline (100 ng/ml final). Images were captured every 5 min with a 20× objective lens, and a Hamamatsu ORCA-AG cooled CCD camera. Images were analyzed using ImageJ (National Institutes of Health).
Cells were seeded on No. 1.5 coverslips (12 mm round) that were pre-placed into 24-well plates. Upon harvest, cells were fixed with 4% formaldehyde for 10 min. After washing, cells were permeablized with 0.1% Trition X-100 for 10 min. Cells were then washed and blocked with 1% bovine serum for 30 min before incubated with primary antibody (anti-β-tubulin, #2128; Cell Signaling Technology, Beverly, MA) prepared in PBS containing 1% bovine serum albumin. After overnight incubated at 4°C, the samples were washed and incubated with Alexa 488-conjugated secondary antibody (Invitrogen) for 1 hr at room temperature. After extensive washing, the samples were dried and mounted with ProLong Antifade reagent (Invitrogen). The images were acquired with a Nikon 80i upright microscope at the Nikon Imaging Center (Harvard Medical School), which is equipped with a Hamamatsu C8484-03 monochrome camera. ImageJ was used for analysis of the images, which includes merging channels with different colors and cropping.
Cells were lysed with RIPA buffer (25 mM Tris, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate) supplemented with protease inhibitors cocktail (Roche) and phosphatase inhibitors cocktail (Thermo Scientific, Waltham, MA). Cleared lysates were analyzed for protein concentration using BCA kit (Thermo Scientific). Equal amount of protein (10–20 μg) was resolved on SDS-PAGE, and was subsequently transferred onto a nitrocellulose or polyvinylidene difluoride membrane. The membrane was blocked with 5% non-fat milk and was then incubated with primary antibodies overnight at 4°C. After washing, the membrane was incubated with fluorophore-conjugated secondary antibodies for 1 hr at room temperature. The membrane was then washed and scanned with an Odyssey Infrared scanner (Li-Cor Biosciences, Lincoln, NE). Primary antibodies used in this study include anti-MELK, anti-α-tubulin (Abcam, Cambridge, MA), anti-cyclin B1 (Millipore), anti-Vinculin (Sigma, St. Louis, MO), anti-FoxM1 (Santa Cruz, Dalla, TX), anti-β-tubulin, anti-phopho-Akt (S473), anti-phospho-Akt (T308), anti-total Akt, anti-Flag, anti-cleaved PARP (Asp214), anti-cleaved Caspase-3, anti-AURKA, anti-AURKB, anti-p27, anti-Estrogen Receptor α (all from Cell Signaling Technology). Secondary antibodies used were IRDye700-conjugated anti-rabbit IgG and IRDye800-conjugated anti-mouse IgG (Rockland, Gilbertsville, PA).
Primary human breast cancer samples were obtained from the Dana-Farber Cancer Institute with patients' consent and institutional review board approval. These samples were deidentified and are not considered human subject research. Samples were homogenized in RIPA buffer supplemented with protease/phosphatease inhibitors using Bullet blender (Next advance, Averill Park, NY). After clearing, tissue lysates were subjected for protein concentration determination. 20 micrograms of lysates were used for immunoblotting.
Mouse embryonic stem cells with one allele of Melk inserted with lacZ and neomycin-resistance genes between exon 2 and 3 were obtained from the Knockout Mouse Project (KOMP; ID: CSD33136). Cells with normal karyotype were injected into blastocysts isolated from C57BL/6 mice. The procedure of injection was performed at the Transgenic Core Facility, Brigham and Women's Hospital (Boston, MA). Germline transmission was subsequently observed, and further cross was made to generate Melk homozygous knockout mice. Melk knockout was confirmed by long-rang PCR, qPCR, and immunoblotting.
Cells were isolated from bone marrow, spleen, and thymus of mice, and stained with the following antibodies: B220 (APC; BD Pharmingen), cKit (PE-Cy7; BioLegend, San Diego, CA), CD3 (PE-Cy7; BD Bioscience), CD4 (APC-H7; BD Pharmingen), CD8 (ECD; Beckman Coulter), CD11b (PE; BD Bioscience), CD16/32 (PE; eBioscience, San Diego, CA), CD34 (FITC; BD Pharmingen), CD45.2 (PerCP-Cy5.5; BD Pharmingen), CD48 (APC-Cy7; BD Pharmingen), CD127 (ECD; BD Pharmingen), CD150 (PerCP-Cy5.5; BioLegend), Gr1 (APC-Alexa700; BD Bioscience), Lineage Cocktail (APC; BD Pharmingen), Sca1 (Brilliant Violet 421; BioLegend). Dead cells were excluded using either DAPI or Vivid-Aqua (Invitrogen) staining. All data acquisition was performed on a LSRII (BD) flow cytometer, and results were analyzed using FlowJo v.8.8.7 (TreeStar).
Cells were harvested by trypsinization, and repeatedly pipetted into single-cell suspension. After centrifugation, cells were fixed by adding 70% ethanol (−20°C) dropwise while vortexing. Cells were then stained with propidium iodide (50 μg/ml, Sigma) solution containing 50 μg/ml DNase-free RNase A (Sigma) and 0.5% bovine serum albumin (BSA). After 30 min of incubation, the samples were washed and resuspended in 0.5% BSA. The analysis was performed on a LSRFortessa (BD Biosciences) at the DFCI Flow Cytometry Core Facility. Single cells were gated via plotting FL3-A to FL3-H to exclude cell debris and doublets. At least 10,000 single cells were collected for each sample.
Chromatin immunoprecipitation was performed as previously described (Lee et al., 2006). Upon harvest, medium in cell culture dishes was added with 16% formaldehyde (Electron Microscopy Sciences, Hatfield, PA) to reach a final concentration of 1%, and quenched with glycine (125 mM final, 5 min incubation) after incubation at room temperature for 10 min. Cells were harvested by scrapping into cold PBS, and centrifuged. Cell pellets were lysed with LB1 (50 mM HEPES, pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton-X-100), then after centrifugation with LB2 (10 mM Tris–HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA), and again after centrifugation resuspended in LB3 (10 mM Tris–HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate, 0.5% N-lauroylsacosine). Samples were sonicated using a Q800R DNA Shearing Sonicator (Qsonica, Newtown, CT) at 50% amplitude for 10 min with a pulse of 30 s on and 30 s off. Samples were then supplemented with 10% Triton-X 100 to a final concentration of 1%, and centrifuged at 20,000×g for 10 min at 4°C. The cleared lysates were used for the following immunoprecipitation, with 50 µl of lysate saved as input.
Protein G-conjugated Dynabeads (Invitrogen) were washed with block solution (0.5% bovine serum albumin in PBS) and incubated overnight with 5 µg anti-FoxM1 (SC-502, Santa Cruz Biotechnology), or 5 µg rabbit IgG in block solution, and on the next day washed three times with block solution. Cell lysates were incubated with the antibody/magnetic bead, rotating at 4°C overnight. On the next day, the beads were collected with magnetic stand, and washed six times with RIPA buffer (50 mM HEPES pH 7.6, 500 mM LiCl, 1 mM EDTA, 1% NP-40, 0.7% Na-deoxycholate). After a single wash with Tris-EDTA buffer containing 50 mM NaCl, samples were resupsended with elution buffer (50 mM Tris–HCl pH 8.0, 10 mM EDTA, 1% SDS) for incubation at 65°C overnight. Also, the 50 µl input was mixed with 150 µl elution buffer and incubated at 65°C overnight for reverse crosslinking. On the next day, RNase A was added to the samples (0.2 µg/ml final), followed by incubation for 1 hr at 37°C. Samples were then treated with Proteinase K (0.2 µg/ml final) and incubated at 56°C for 1 hr. DNA were purified with a QIAquick PCR purification kit (Qiagen), and eluted with 30 µl water. PCR was performed using Quick-Load Taq 2X Master Mix (New England BioLabs, Beverly, MA), using primers listed in Supplementary file 2.
Total RNA was extracted from cultured cells with RNeasy Mini kit (Qiagen), with the use of QIAshredder spin column for homogenization and an on-column DNase digestion. 2 μg of the total RNA was reversely transcribed using a High Capacity RNA-to-cDNA Kit (Applied Biosystems, Foster City, CA). cDNA were analyzed quantitatively using Power SYBR Green PCR Master Mix (Applied Biosystems) on an ABI7300 Real-time PCR system. Primers used were listed in Supplementary file 2. Cycling conditions were 95°C for 15 min, 40 cycles of 15 s at 94°C, 30 s at 55°C and 30 s at 72°C. Ct values were generated using the default analysis settings. ΔCT was defined as Ct gene of interest − Ctβ-actin. ΔΔCT was defined as ΔCt treated sample − Ct control sample. Relative quantification (RQ) was calculated as 2−ΔΔCT. Statistical analysis was performed by Student's t test.
Gene expression data were downloaded from Oncomine (Rhodes et al., 2004). Information of the clinical data sets is listed in Supplementary file 1. Analyses and figures were made in GraphPad Prism. In dot plot graphs, each dot indicates an individual sample, with results expressed as median with interquartile range.
Independent cohorts of breast cancer patients with overall survival or metastasis-free survival data available were examined. Information of the cohorts is listed in Supplementary file 1. Data of MELK expression and associated survival were downloaded from Oncomine (Rhodes et al., 2004). For each cohort, patients were divided into top 60% ‘MELK high’ and bottom 40% ‘MELK low’ groups based on the expression of MELK. Kaplan–Meier curves, as well as the log-rank (Mantel–Cox) test and the hazard ratio were analyzed by GraphPad Prism.
Two-tailed Student's t test and ANOVA (Analysis of Variance) were used for differential comparison between two groups and among three groups, respectively. Survival and correlation analysis were performed in GraphPad Prism.
Substrate specificity and activity regulation of protein kinase MELKThe Journal of Biological Chemistry 280:40003–40011.https://doi.org/10.1074/jbc.M507274200
Comprehensive molecular portraits of human breast tumours61–70, Nature, 490, no. 7418.
Development of an orally-administrative MELK-targeting inhibitor that suppresses the growth of various types of human cancerOncotarget 3:1629–1640.
Aurora kinase inhibitors–rising stars in cancer therapeutics?Molecular Cancer Therapeutics 9:268–278.https://doi.org/10.1158/1535-7163.MCT-09-0765
erbB-2 is a potent oncogene when overexpressed in NIH/3T3 cellsScience 237:178–182.
Cell cycle analysis and chromosomal localization of human Plk1, a putative homologue of the mitotic kinases Drosophila polo and Saccharomyces cerevisiae Cdc5Journal of Cell Science 107:1509–1517.
A genomic predictor of response and survival following taxane-anthracycline chemotherapy for invasive breast cancerThe Journal of the American Medical Association 305:1873–1881.https://doi.org/10.1001/jama.2011.593
Increased expression of the putative growth factor receptor p185HER2 causes transformation and tumorigenesis of NIH 3T3 cellsProceedings of the National Academy of Sciences of the United States of America 84:7159–7163.https://doi.org/10.1073/pnas.84.20.7159
The maternal embryonic leucine zipper kinase (MELK) is upregulated in high-grade prostate cancer237–248, Journal of Molecular Medicine, 91, no. 2, 10.1007/s00109-012-0949-1.
FoxM1 is required for execution of the mitotic programme and chromosome stabilityNature Cell Biology 7:126–136.https://doi.org/10.1038/ncb1217
Shared and separate functions of polo-like kinases and aurora kinases in cancerNature Reviews Cancer 10:825–841.https://doi.org/10.1038/nrc2964
Normal cells, but not cancer cells, survive severe Plk1 depletionMolecular and Cellular Biology 26:2093–2108.https://doi.org/10.1128/MCB.26.6.2093-2108.2006
Definition of clinically distinct molecular subtypes in estrogen receptor-positive breast carcinomas through genomic gradeJournal of Clinical Oncology 25:1239–1246.https://doi.org/10.1200/JCO.2006.07.1522
Maternal embryonic leucine zipper kinase transcript abundance correlates with malignancy grade in human astrocytomas807–815, International journal of cancer, 122, no. 4, 10.1002/ijc.23189.
Maternal embryonic leucine zipper kinase is a key regulator of the proliferation of malignant brain tumors, including brain tumor stem cells48–60, Journal of neuroscience research, 86, no. 1, 10.1002/jnr.21471.
Nek6 mediates human cancer cell transformation and is a potential cancer therapeutic targetMolecular Cancer Research 8:717–728.https://doi.org/10.1158/1541-7786.MCR-09-0291
Supervised risk predictor of breast cancer based on intrinsic subtypesJournal of Clinical Oncology 27:1160–1167.https://doi.org/10.1200/JCO.2008.18.1370
Murine protein serine-threonine kinase 38 activates p53 function through Ser15 phosphorylationThe Journal of Biological Chemistry 287:20797–20810.https://doi.org/10.1074/jbc.M112.347757
PDK1 protein phosphorylation at Thr354 by murine protein serine-threonine kinase 38 contributes to negative regulation of PDK1 protein activityThe Journal of Biological Chemistry 287:20811–20822.https://doi.org/10.1074/jbc.M111.331827
Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implicationsProceedings of the National Academy of Sciences of the United States of America 98:10869–10874.https://doi.org/10.1073/pnas.191367098
Polo and Aurora kinases: lessons derived from chemical biologyCurrent Opinion in Cell Biology 20:77–84.https://doi.org/10.1016/j.ceb.2007.11.008
A gene-expression signature as a predictor of survival in breast cancerThe New England Journal of Medicine 347:1999–2009.https://doi.org/10.1056/NEJMoa021967
Most random gene expression signatures are significantly associated with breast cancer outcomePLOS Computational Biology 7:e1002240.https://doi.org/10.1371/journal.pcbi.1002240
Inhibition of spliceosome assembly by the cell cycle-regulated protein kinase MELK and involvement of splicing factor NIPP1The Journal of Biological Chemistry 279:8642–8647.https://doi.org/10.1074/jbc.M311466200
FOXM1, a typical proliferation-associated transcription factorBiological Chemistry 388:1257–1274.https://doi.org/10.1515/BC.2007.159
The serine/threonine kinase Nek6 is required for cell cycle progression through mitosisThe Journal of Biological Chemistry 278:52454–52460.https://doi.org/10.1074/jbc.M308080200
Targeting cancer with small molecule kinase inhibitorsNature Reviews Cancer 9:28–39.https://doi.org/10.1038/nrc2559
Functional genetics and experimental models of human cancerTrends in Molecular Medicine 10:344–350.https://doi.org/10.1016/j.molmed.2004.05.005
Louis StaudtReviewing Editor; National Cancer Institute, United States
eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.
[Editors’ note: this article was originally rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision.]
Thank you for choosing to send your work entitled “The oncogenic protein kinase MELK is essential for mitotic progression in basal-like breast cancer cells” for consideration at eLife. Your full submission has been evaluated by a Senior editor and 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the decision was reached after discussions between the reviewers. We regret to inform you that your work will not be considered further for publication at this time.
In particular, the reviewers were not convinced that MELK is a bona fide oncogene in basal breast cancer and were further not persuaded that the evidence warrants development of therapies targeting MELK. The functional data provided from a limited series of breast cancer cell lines does not allow generalizable conclusions to be drawn. Moreover, the role of MELK in a wide variety of proliferating cells in many organisms argues against its selective role in basal breast cancer.
The manuscript entitled “The oncogenic protein kinase MELK is essential for mitotic progression in basal-like breast cancer cells” begins with a functional overexpression screen utilizing a library of open reading frames encoding human kinases. The assay was the ability of a partially transformed breast epithelial cell line to form tumors when injected into the mammary fat pad. MELK was one of several kinases that scored in this assay. In several breast cancer gene expression profiling datasets, MELK mRNA expression was higher in the basal subtype, and was associated with poor prognosis and early metastasis – these observations have been made in part by others previously.
MELK protein levels were higher in basal cell lines that in luminal cell lines, although only 2 luminal cell lines were tested. Knockdown of MELK impaired the proliferation of basal cell lines, but had less effect on luminal cell lines, with some caveats (see below). Moreover, MELK knockdown caused tumor regression in xenografts of basal cell lines, but had a more modest effect on xenografts of luminal cell lines. In rescue experiments, kinase-dead MELK was incapable of preventing the toxicity of MELK knockdown. MELK knockdown prevented mitotic progression, leading frequently to apoptosis in basal cell lines. High expression of MELK was correlated with expression of the transcription factor FOXM1. The authors conclude that MELK is a therapeutic target for basal breast cancer, although there is no discussion of the role of MELK in normal cell division and what side effects a MELK inhibitor might be expected to have.
1) The level of interest in these findings relies on whether MELK is selectively utilized for proliferation in basal breast cancer cells versus other proliferating cells. This selectivity is necessary in order to propose that MELK might deserve attention as a target for therapeutic development. The proposed selectivity of MELK for basal breast cancer cells is difficult to reconcile with an essential role for MELK in diverse proliferation systems, including Xenopus oocytes, C. elegans and various proliferating progenitor populations in the mouse (Hebbard et al. Cancer Research 2010 70:8863 and references therein). Indeed, Hebbard et al previously showed that MELK is required for breast cancer tumor initiating cells in a mouse breast cancer model, a fact that is not cited by the authors. Functional studies were restricted to 2 basal and 2 luminal cell lines, so it is difficult to generalize from this small sample set. Moreover, it is unclear whether the knockdown of MELK was equivalent in these experiments between the basal and luminal cell lines since the Western blots were not quantified.
By eye, the degree of knockdown of MELK by shMELK 2 in the basal cell lines (Figure 4C) appears stronger than in the luminal cell lines (Figure 4D). Getting functional effects by knocking down kinases can be difficult given their enzymatic nature. It could be that the degree of knockdown of MELK in the luminal cell lines did not go below a necessary threshold to cause a proliferation effect. Figure 4C demonstrates that two different MELK shRNAs that have modest differences in their degree of knockdown have very large differences in their effect on proliferation. For these studies to be strengthened, several more cell lines of each type need to be examined and equivalent knockdown needs to be demonstrated by both quantitative protein and mRNA analysis. The xenograft study suffers from the same caveats. In Figure 5–figure supplement 1, the degree of knockdown in the luminal cell lines was not as great as in the one basal cell line presented, which could explain the difference in effect on tumor growth. These xenograft experiments should be repeated under conditions in which MELK knockdown is equivalent, ideally with additional luminal and basal cell lines.
2) The association of MELK with survival in breast cancer is likely due to its correlation with proliferation, which has previously been shown to be a component of gene expression profiles that predicts adverse survival (Dai et al. Cancer Res. 2005 65:4059-66). The authors should determine whether MELK correlates with other published signatures of proliferation and whether its expression adds any statistical power to a proliferation signature in predicting adverse prognosis in breast cancer.
The manuscript describes the dependence on the cell cycle kinase MELK in basal-like breast cancer (BBC) for survival. The authors used a cDNA kinome transformation screen method previously developed in their lab to identify MELK as a potential oncogene. They provided evidence that MELK expression is elevated in BCC and depletion of MELK in BCC cells leads to mitotic defects. Untransformed breast epithelial cells are insensitive to MELK depletion due to low expression levels. These findings indicate that MELK could be a potential target for BCC and other cancer types that show high levels of MELK expression. The data presented in the manuscript is generally of high quality. Previous studies characterizing MELK expression in triple negative breast cancer and other aggressive tumor types somewhat diminish the novelty of this work. The authors showed that the transcription factor FoxM1, which positively regulates MELK expression, is also upregulated in BCC, although the mechanism by which this happens is unclear.
1) Informatics for MELK's association with BCC. The authors showed that MELK expression is higher in high-grade breast cancers (Figure 2B), thus it is not clear whether MELK expression is higher in in BCC because there are more high-grade tumors in the BCC sub-types than the other 4 sub-types in the dataset used for the analysis. The authors can perform an ANOVA analysis of the TCGA dataset to address this issue. Although the authors suggest that high MELK expression might be a unique phenomenon for BCC, MELK over-expression has been associated with tumor aggressiveness and poor prognosis in several cancer types, including melanoma (Ryu et al. PMID 17611626), glioblastoma and astrocytoma (Nakano et al. PMID 17722061; Marie et al. PMID 17960622), cervical cancer (Rajkumar et al. PMID 21338529), and prostate cancer (Kuner et al. PMID 22945237).
In these prior studies higher MELK expression in more aggressive tumor sample/cell lines is often correlated with other cell cycle genes. Thus it is likely that MELK expression is generally associated with more poorly differentiated tumors that show higher proliferation rate. The authors should discuss their finding in light of these works in other cancer types. Recently MELK has been reported to be over-expressed in TNBC tissues compared to normal ductal cells (Komatsu et al. PMID 23254957), this somewhat diminish the novelty of the authors' finding.
2) The role of MELK in HMEC transformation. The authors identified MELK's transforming activity in HMEC-DD-NeuT cells and demonstrated that MELK can transform Rat1-DD fibroblasts in a Neu-independent fashion. The authors did not present data to show whether MELK can transform HMEC-DD cells without the NeuT oncogene. Since MELK has not been significantly mutated or amplified in breast cancers, the evidence for MELK as a bona fide oncogene is relatively weak.
3) The dependency on MELK in BCC cell lines. Given MELK is a cell cycle regulated genes, it would be important for the authors to test whether MELK expression levels in luminal and basal cell lines (Figure 4B) is correlated with their doubling times. Given there are a large number of well-characterized breast cancer cell lines available, testing additional luminal or ER/PR positive cell lines for their dependency on MELK should further strengthen the authors conclusions. MELK was not identified as a BCC lethal gene in the Neel lab's screen using the TRC shRNA library (Marcotte et al., PMID 22585861).
4) The role of MELK in cell cycle. Prior studies indicate that MELK is a cell cycle gene involved in mitosis. It is expressed in tissue progenitor cells and plays a role in embryogenesis. The authors' data indicate that not all proliferating cells require MELK, as knockdown of MELK in MCF7, T47D and HMECs does not impair proliferation. Have the authors investigated whether MELK knockdown in these insensitive cells affect their cell cycle distribution? Have the authors also looked at whether MELK over-expression in HMEC-DD or HMEC-DD-NeuT cells affect their cell cycle and proliferation rate? In xenopus embryo both the loss and over-expression of MELK causes mitotic defects. This would help address the question whether MELK's transforming activity in HMECs is due to a cell cycle effect or a cell-cycle independent mechanism.https://doi.org/10.7554/eLife.01763.032
We appreciate the insightful and constructive comments from reviewers and editors. In response to these critiques, we have extensively revised the manuscript, in most cases with new experimental data, including data with a MELK inhibitor showing selective activity against basal-like breast cancer (BBC), the finding of gene amplification of MELK in addition to extensive overexpression of MELK in BBC from a recent TCGA dataset, and a MELK knockout mouse model showing that MELK is dispensable for general proliferation and development. In so doing, we feel that the fundamental conclusions of our manuscript have been significantly strengthened.
1) Reviewer #1: “The level of interest in these findings relies on whether MELK is selectively utilized for proliferation in basal breast cancer cells versus other proliferating cells. This selectivity is necessary in order to propose that MELK might deserve attention as a target for therapeutic development...”
Reviewer #2 expresses a similar concern.
We thank the reviewers for their insightful comments. Indeed, the selective requirement of MELK in basal breast cancer cells versus other proliferating cells is one line of evidence supporting MELK as a therapeutic target. We now have new data from newly developed mouse models demonstrating that MELK is not essential for normal physiological functions.
We have recently generated mice with germ-line knockout (KO) of Melk. Surprisingly, mice with Melk germ-line KO are viable and appear normal without any noticeable phenotypes in the development of the embryos or adult mice. Both male and female mice are fertile and produce litters of normal size. While we were disappointed by these “negative” results, these data suggest that Melk is not an essential gene in mice. This stands in contrast to other mitotic kinases, such as Aurora A (worm ortholog AIR1), Aurora B (worm ortholog AIR2), and PLK1, all of which are essential for the embryonic development of both mice and C. elegans (Chase et al., 2000; Schumacher et al., 1998; Severson et al., 2000).
Since bone marrow toxicity is a major side effect of most anti-cancer therapies and, in particular, of those drugs targeting mitotic kinases (e.g., aurora kinases and PLK1), we characterized the immune system in Melk KO mice. We isolated cells from bone marrow, spleen and thymus, and analyzed immune cell populations including monocytes, neutrophils, B and T cells. There were no differences found between wild-type and KO mice (Figure 7G, 7H). We further analyzed the hematopoietic stem cells and various progenitor cells in the bone marrow of both wild-type and KO mice, and again found no differences between the two groups of mice (Figure 7I).
We understand the reviewers’ concern that it is difficult to reconcile the previously observed roles for MELK in diverse proliferating systems, including Xenopus oocytes, C. elegans and various proliferating progenitor populations in the mouse, with our findings. However, previous studies did not point uniformly to a MELK requirement. For example, while Xenopus MELK (xMELK) is required for the development of Xenopus oocytes (Le Page et al., 2011), the C. elegans MELK ortholog, PIG-1, is not an essential gene in the worm. Similar to mice, worms with mutated or deleted PIG-1 are viable and relatively healthy albeit with an altered number of certain neurons (Cordes et al., 2006).
A study by Hebbard et al. (2010), where GFP expression was driven by the Melk promoter in a mouse model, found a high expression of Melk in mouse mammary progenitor cells, but the functional role of Melk in these cells was not determined. Given these data, we anticipated an impairment of mammary gland development in Melk KO mice. But mammary glands from mice with Melk KO appear completely normal in both morphology and function (e.g., lactation).
Furthermore, the growth of normal human mammary epithelial cells was largely unaffected by shRNA knockdown of MELK with ∼90% efficiency (Figure 5–figure supplement 2).
In summary, while MELK is expressed in a variety of proliferating cells in many organisms, the roles of MELK in these cells and organisms are divergent. Our mouse genetic data strongly suggest that MELK is not essential in mice and deserves attention as a target for therapeutic development. We now have included these data in the revised manuscript.
2) The functional data provided from a limited series of breast cancer cell lines does not allow broad conclusions to be drawn.
Both reviewers consider that since the functional studies were only tested in 2 basal and 2 luminal breast cancer cell lines, the sample set was too small, and testing additional cell lines for their dependency on MELK should further strengthen the authors conclusions. Reviewer #1 also had concerns that the degree of knockdown of MELK by shMELK in the basal cell lines appears stronger than in the luminal cell lines, equivalent knockdown needs to be demonstrated by both quantitative protein and mRNA analysis.
We agree with the reviewers that quantification of MELK knockdown is important for the evaluation of functional dependence of MELK in these cells. We have re-constructed these luminal cell lines expressing shMELK and were able to achieve an equivalent efficiency of MELK knockdown in these cells (Figure 5–figure supplement 1A, 1B). Similar to what we observed previously, proliferation of luminal cells were largely unaffected by MELK knockdown (Figure 5E).
We also agree that the sample set used in the study was small, and we have expanded it to include 6 basal and 5 luminal breast cancer cell lines. In addition to cancer cell lines, we have also examined MELK expression in primary tumor specimens resected from breast cancer patients. Notably, while robust MELK protein expression was observed in all ER/PR-negative breast tumors, MELK protein was not detectable in these ER/PR-positive breast tumors (Figure 2G), suggesting that MELK is largely dispensable for the growth of these luminal tumors.
To complement the RNAi-mediated knockdown approach, we are also using pharmacological inhibition of MELK to evaluate the functional dependency of MELK in basal and luminal breast cancer cells. We tested OTSSP167, a newly developed potent MELK inhibitor (Chung et al., 2012), on a panel of 4 basal and 4 luminal breast cancer cell lines, and found that all these basal cancer cells are significantly more sensitive to the drug than are the luminal cell lines. OTSSP167 at 100 nM induced dramatic cell death of basal cells, but had little effect on luminal cells (Figure 6–figure supplement 2A-C). Notably, in mice bearing breast cancer cell line-derived tumors, administration of OTSSP167 significantly impaired the growth of basal breast tumors, or even caused tumor regression; by contrast, the growth of luminal tumors were largely unaffected by the treatment (Figure 7E, 7F). Together, these data indicate that chemical inhibition of MELK preferentially targets basal breast cancer cells. We now have included these new data in the revised manuscript as indicated above.
3) Since MELK has not been significantly mutated or amplified in breast cancers, the evidence for MELK as a bona fide oncogene is relatively weak.
Reviewer #2: “The authors identified MELK's transforming activity in HMEC-DD-NeuT cells and demonstrated that MELK can transform Rat1- DD fibroblasts in a Neu-independent fashion. The authors did not present data to show whether MELK can transform HMEC-DD cells without the NeuT oncogene. Since MELK has not been significantly mutated or amplified in breast cancers, the evidence for MELK as a bona fide oncogene is relatively weak.”
We thank the reviewer for these comments. Regarding the transformation of HMECs, we and others have shown that mutant PIK3CA, Ras or NeuT alone can induce colony formation of HMEC-DD cells in soft agar, but is not sufficient to promote tumor formation of these cells in mice (Hahn et al., 1999; Zhao et al., 2003; Zhao et al., 2004). It usually takes two oncogenic events, e.g. PIK3CA plus NeuT or Ras plus NeuT, to fully transform HMEC-DD cells to form tumors in mice. Similar to these potent oncogenes, over-expression of MELK alone can also robustly promote anchorage-independent growth of HMEC-DD cells (Figure 4–figure supplement 2C), and again similar to other known oncogenes wild-type MELK cooperates with a second oncogene, e.g., NeuT, to induce these cells to form tumors in vivo.
In contrast to the transformation of HMEC-DD cells, one oncogenic event is sufficient to transform Rat1-DD cells, indicated by both anchorage-independent growth in vitro and tumor formation in vivo (Ni et al., 2012) . We have shown in our manuscript that MELK overexpression alone in Rat1-DD cells was able to effectively promote colony formation in soft agar or tumor growth in nude mice (Figure 2).
In addition to HMEC-DD and Rat1-DD cells, we also show that overexpression of MELK is able to induced anchorage-independent growth of MCF10A cells (Figure 4–figure supplement 2E).
In addition to HMEC-DD and Rat1-DD cells, we also show that overexpression of MELK is able to induced anchorage-independent growth of MCF10A cells (Figure 4–figure supplement 2E).
We agree with the reviewer that, in general, genes with hot spot mutations or amplification in human tumors are more readily accepted as oncogenes. Thus, a gene that is over-expressed in tumors needs rigorous testing in relevant systems to demonstrate a causal role in oncogenic transformation. Therefore, we have carried out extensive functional validation of MELK in our study as described above and in the manuscript. In fact, it is striking that the transforming activity of overexpressed MELK, is comparable or superior to that of the highly potent oncogenic PIK3CA-H1047R, in all the transformation assays that we have performed.
With data recently available for the analysis of copy number in large cohort of breast cancer (TCGA, 2012), we find that gene amplification of MELK occurs in breast cancer, especially in ER-negative breast cancer (Figure 3–figure supplement 1A). The single ER-positive tumor with MELK amplified is also positive for HER2 expression. Similarly, out of a total 56 established breast cancer cell lines, three are found harboring MELK amplification and are all ER-negative (Figure 3–figure supplement 1B). Notably, tumors or cells with MELK amplified tend to express high level of MELK, suggesting that gene amplification contributes to the overexpression of MELK in breast cancer.
Additional points from Reviewer #1:
The association of MELK with survival in breast cancer is likely due to its correlation with proliferation, which has previously been shown to be a component of gene expression profiles that predicts adverse survival (Dai et al. Cancer Res. 2005 65:4059-66). The authors should determine whether MELK correlates with other published signatures of proliferation and whether its expression adds any statistical power to a proliferation signature in predicting adverse prognosis in breast cancer.
We agreed with the reviewer that the prognostic feature of MELK expression is likely due to its correlation with cell proliferation. In fact, MELK and other proliferation-related genes are major components of multi-gene signature in predicting disease outcome. For example, a recent study developed a cell proliferation signature composing of top 1% genes whose expression is most positively correlated with proliferating cell nuclear antigen (PCNA) expression, and found that adjusting breast cancer expression data for this cell proliferation signature causes a dramatic reduction in outcome association of most published breast cancer signatures (Venet et al., 2011). Notably, MELK expression is strongly correlated with cell proliferation, e.g., it’s correlation with PCNA expression in the above-mentioned cell proliferation signature is 10th of the total 129 genes (Venet et al., 2011).
Additional points from Reviewer #2:
Informatics for MELK's association with BCC. The authors showed that MELK expression is higher in high-grade breast cancers (Figure 2B), thus it is not clear whether MELK expression is higher in in BCC because there are more high-grade tumors in the BCC sub-types than the other 4 sub-types in the dataset used for the analysis. The authors can perform an ANOVA analysis of the TCGA dataset to address this issue. Although the authors suggest that high MELK expression might be a unique phenomenon for BCC, MELK over-expression has been associated with tumor agressiveness and poor prognosis in several cancer types, including melanoma (Ryu et al. PMID 17611626), glioblastoma and astrocytoma (Nakano et al. PMID 17722061; Marie et al. PMID 17960622), cervical cancer (Rajkumar et al. PMID 21338529), and prostate cancer (Kuner et al. PMID 22945237). In these prior studies higher MELK expression in more aggressive tumor sample/cell lines is often correlated with other cell cycle genes. Thus it is likely that MELK expression is generally associated with more poorly differentiated tumors that show higher proliferation rate. The authors should discuss their finding in light of these works in other cancer types.
Following the reviewer’s suggestion, we performed analyses for the correlation of MELK expression with subtypes of breast cancer with the same pathological grade. Through analysis of a large dataset of breast cancer for grades 1, 2 and 3 across all subtypes, respectively, we found that MELK is most highly expressed in BBC (Figure 2–figure supplement 2A), suggesting that MELK expression in most pronounced in basal-like breast tumors concerning the same grade of the disease. Moreover, a significant association of MELK expression with disease status also exists within the subtype of BBC (Figure 2–figure supplement 2B), suggesting that MELK expression is associated with tumor aggressiveness and poor prognosis in this disease as the reviewer mentioned.
Recently MELK has been reported to be over-expressed in TNBC tissues compared to normal ductal cells (Komatsu et al. PMID 23254957), this somewhat diminish the novelty of the authors' finding.
This study, which is based on gene expression profiling of 30 samples of triple negative breast cancer samples and 13 samples of normal breast ductal cells, found that MELK is one of the 301 genes whose expression is upregulated in TNBC. But MELK was not further functionally analyzed, as MELK was not a focus of this study. Our bioinformatics analysis, based on multiple independent large-size cohorts shows that MELK is strongly overexpessed in BBC/TNBC, not only compared to normal breast tissues, but also compared to luminal or ER/PR+ breast cancer. Our data suggest that MELK represent a bona-fide TNBC gene, rather than a general breast cancer gene. More importantly, we have carried out extensive functional analysis of MELK in breast cancer both in vitro and in vivo.
The dependency on MELK in BCC cell lines. Given MELK is a cell cycle regulated genes, it would be important for the authors to test whether MELK expression levels in luminal and basal cell lines (Figure 4B) is correlated with their doubling time.
We thank the reviewer for the insightful suggestion of testing any correlation between MELK expression and the doubling time of cell lines. A recent study from Dr. Joe Gray’s group has systemically analyzed doubling time among 49 breast cancer cell lines. Indeed, basal breast cancer cells have a statistically shorter doubling time (Figure S3A, Heiser et al., 2012). However, we would like to point out that untransformed cells such as HMEC and MCF-10A have much lower expression of levels of MELK than basal breast cancer cells, but have a very similar doubling times. Therefore, it is difficult to conclude on a universal correlation between MELK expression and the rate of cell proliferation.
MELK was not identified as a BCC lethal gene in the Neel lab's screen using the TRC shRNA library (Marcotte et al., PMID 22585861).
As pointed by the reviewer, a recent study (Marcotte et al., 2012) carried out a genome-wide shRNA screen in a total 72 cancer cell lines, including 28 breast cancer cell lines. The criteria applied might explain why MELK was not identified as a breast cancer essential gene. The top-ranked genes from the screens were selected and overlapped against two sets of genes that are likely enriched in essential genes (housekeeping genes (n=1,722), and genes with highly conserved orthologs in eight species (n=1,617)). Interestingly, MELK is not such a housekeeping genes (defined here as expressing in more than 73 of 79 tissues in a human expression compendium), or highly conserved (e.g., MELK ortholog has not been identified in Saccharomyces cerevisiae).
The role of MELK in cell cycle. Prior studies indicate that MELK is a cell cycle gene involved in mitosis. It is expressed in tissue progenitor cells and plays a role in embryogenesis. The authors' data indicate that not all proliferating cells require MELK, as knockdown of MELK in MCF7, T47D and HMECs does not impair proliferation. Have the authors investigated whether MELK knockdown in these insensitive cells affect their cell cycle distribution? Have the authors also looked at whether MELK over-expression in HMEC-DD or HMEC-DD-NeuT cells affect their cell cycle and proliferation rate? In Xenopus embryo both the loss and over-expression of MELK causes mitotic defects. This would help address the question whether MELK's transforming activity in HMECs is due to a cell cycle effect or a cell-cycle independent mechanism.
We thank the reviewer for the insightful comments. We now have performed experiments and found that neither MELK knockdown, nor MELK overexpression in untransformed HMECs alters cell cycle or proliferation as assayed in regular tissue culture. However, the information acquired from regular two-dimensional culturing might not be directly applied to the process of cell transformation, which is typically assayed in three-dimensional conditions. Indeed, when we assessed these cells in suspension culture, overexpression of MELK significantly increased the proliferation of HMEC cells (Figure 4–figure supplement 1A). Likewise, MELK overexpression induces colony formation in soft agar (Figure 4–figure supplement 1C). Further study is needed to address the correlation of the transforming activity of MELK overexpression with cell cycle.
In summary, recent comprehensive characterization of basal-like breast cancer demonstrates that this subtype of disease has high genetic heterogeneity, but lacks commonly occurring genetic alterations, with the exception of the frequent inactivation of p53 (Cancer Genome Atlas Network, 2012). In contrast, the relative uniform overexpression of MELK in basal-like breast cancer makes this gene a unique target if its functional importance can be established. Thus our studies on MELK are critical for guiding the development of targeted therapies in basal-like breast cancer.
Our new study finds that MELK is not an essential gene for many types of normal murine and human cells, but is important for the growth of basal-like breast cancer cells. We plan to add to the current manuscript our data from knockout mice, to strengthen the point that MELK is not a universally essential gene for cell proliferation and only selectively required in basal-like breast cancer. We will also add new data from small molecule studies that use a chemical inhibitor of MELK, to demonstrate the selectivity of targeting MELK in basal versus luminal breast tumors.
The finding of a non-essential mitotic kinase that can be targeted in basal-like breast cancer is novel and unexpected. We are very enthusiastic about this finding, and believe that our work will have enormous impact on guiding the development of therapy for basal-like breast cancer, the most aggressive subtype of breast cancer lacking any proven targeted therapy. In addition, as mentioned by the reviewers, MELK over-expression has been associated with tumor aggressiveness and poor prognosis in several other cancer types, including melanoma, glioblastoma and prostate cancer, though functional studies are lacking. Thus MELK represents a highly promising therapeutic target for BBC and potentially other aggressive tumors.https://doi.org/10.7554/eLife.01763.033
- Jean J Zhao
- Jean J Zhao
- Jean J Zhao
- Nathanael S Gray
- Jean J Zhao
- Yubao Wang
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Drs TM Roberts and DM Livingston for scientific discussions. We thank the Nikon Imaging Center at Harvard Medical School, DFCI Flow Cytometry Core Facility, and the Transgenic Core Facility at the Brigham and Women's Hospital for technical assistance and the use of instruments. This work was supported by Friends of Dana-Farber Cancer Institute (YW), DFCI/Accelerator Fund (NSG, JJZ), and NIH grants (JJZ).
Animal experimentation: This study was conducted in accordance with the animal use guidelines from the National Institutes of Health and with protocols (#02-127, #06-034) approved by the Dana-Farber Cancer Institute Animal Care and Use Committee.
- Louis Staudt, National Cancer Institute, United States
© 2014, Wang et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.