To identify genes that could serve as targets for novel cancer therapeutics, we used a bioinformatic analysis of microarray data comparing gene expression between normal and tumor-derived primary human tissues. From this approach, we have found that maternal embryonic leucine zipper kinase (Melk), a member of the AMP serine/threonine kinase family, exhibits multiple features consistent with the potential utility of this gene as an anticancer target. An oligonucleotide microarray analysis of multiple human tumor samples and cell lines suggests that Melk expression is frequently elevated in cancer relative to normal tissues, a pattern confirmed by quantitative reverse transcription-PCR and Western blotting of selected primary tumor samples. In situ hybridization localized Melk expression to malignant epithelial cells in 96%, 23%, and 13% of colorectal, lung, and ovarian tissue tumor samples, respectively. Expression of this gene is also elevated in spontaneous tumors derived from the ApcMin and Apc1638N murine models of intestinal tumorigenesis. To begin addressing whether Melk is relevant for tumorigenesis, RNA interference–mediated silencing within human and murine tumor cell lines was done. We show that Melk knockdown decreases proliferation and anchorage-independent growth in vitro as well as tumor growth in a xenograft model. Together, these results suggest that Melk may provide a growth advantage for neoplastic cells and, therefore, inactivation may be therapeutically beneficial.
Progress toward effective disease management in the clinic has been achieved by an understanding of the relevant signaling pathways. The clinical benefit derived from rationally designed, targeted therapies, such as trastuzumab (1, 2), STI571 (3), gefitinib (4, 5), cetuximab (6), and erlotinib (7, 8), has shown the value of defining the molecular phenotypes of a given tumor type. Therefore, we have sought to identify additional genes that serve a vital role in the growth and maintenance of a given tumor, thereby hopefully revealing novel therapeutic targets. In addressing this subject, we propose that maternal embryonic leucine zipper kinase (Melk), an AMP-activated protein kinase (AMPK)–related serine/threonine kinase, may serve a key functional role within multiple cancers.
The AMPK family comprises two isoforms of AMPK (α1 and α2) and 12 AMPK-related kinases (9). Although the AMPK isoforms have been labeled the cellular “fuel gauge” because of their ability to respond to slight changes in the ATP-to-AMP ratio (10), the functional significance of the AMPK-related members is not as well characterized. Melk is conserved among multiple species [pEg3, Xenopus (11); KIAA0175, human (12); murine protein serine-threonine kinase 38/Melk, murine (13, 14)] and was initially identified as a maternal message in both mouse and Xenopus oocytes (11, 14). Although little is known concerning the mechanisms of Melk regulation or activity in the cell, recent studies have suggested a role in mitosis. Melk expression and kinase activity have been shown to be maximal during mitosis in Xenopus embryos and mammalian cells (15, 16). The identification of potential Melk substrates has provided additional links to the cell cycle and some hints with respect to its signaling pathway. First, Melk has been shown to phosphorylate CDC25B on Ser323 in vitro (16), a critical 14-3-3 binding site (17, 18). Binding of 14-3-3 proteins to CDC25 is thought to negatively regulate CDC25 phosphatase activity during the cell cycle (19). Additional reported Melk substrates include ZPR9, a novel zinc finger–like protein (20), and NIPP1, a splicing factor involved in spliceosome assembly (21). Both proteins were initially discovered to associate with Melk through yeast two-hybrid screens and build on the likely involvement of Melk in the cell cycle. For example, Seong et al. (20) showed that Melk phosphorylates ZPR9 in vitro and promotes its nuclear localization in vivo. In agreement with its predicted role as a transcription factor, ZPR9 was subsequently shown to interact with and enhance the transcriptional activity of B-MYB (22), a protein whose activity is tightly regulated by the cell cycle (23). Finally, Vulsteke et al. (21) found that the interaction between Melk and NIPP1 is largely dependent on the presence of a phosphorylated Thr478 within the COOH terminus of Melk. This association between Melk and NIPP1 was maximal during mitosis and prevented spliceosome assembly in cell extracts. Interestingly, the binding of NIPP1 to Melk, as well as the resulting inhibition of spliceosome assembly, did not seem to require an active Melk kinase (21). Although it is unclear whether NIPP1 is a substrate of the kinase activity of Melk, the interaction between these two proteins again supports a close link between Melk and the cell cycle.
We have identified Melk as a potential anticancer target from a bioinformatics analysis of oligonucleotide microarray data comparing gene expression between human tumor and normal tissues. The correlation between Melk expression and cancer is supported by quantitative reverse transcription-PCR (qRT-PCR), in situ hybridization, and Western analyses of tumor samples and cell lines. Whereas the most consistent increase in expression was observed in colon tumors, elevated Melk levels were found in a broad range of cancer types. Using RNA interference (RNAi), we show that Melk knockdown results in an accumulation of cells at the G2-M transition. Furthermore, Melk knockdown decreased the transformed phenotype of multiple tumor cells lines as measured by both in vitro (proliferation, anchorage-independent growth) and in vivo (xenograft) assays. Together, these results suggest that Melk is required for tumorigenesis and, therefore, targeting this serine/threonine kinase with a small molecule inhibitor may serve as an effective cancer therapy.
Materials and Methods
Cell lines and tissue samples. All cell lines used in this study were obtained from American Type Culture Collection (Manassas, VA) and were maintained under recommended conditions. Anonymized fresh-frozen and formalin-fixed, paraffin-embedded human tissues were obtained from the Genentech pathology archives. Tissue microarrays were constructed as described previously (24, 25). ApcMin and Apc1638N mice (26) were euthanized at ∼3 months of age. At autopsy, the intestines were removed, washed with PBS, and either snap frozen or formalin fixed and paraffin embedded.
Microarray data. Melk was identified from a bioinformatics screen of the Gene Logic (Gaithersburg, MD) expression database of Affymetrix HG-U133 data, representing 3,600 normal and 1,701 neoplastic human tissue samples. Probe set 204825_at was chosen to represent Melk expression.
Microarray analysis was done on fresh-frozen normal tissues from five regions of normal mucosa from each ApcMin mouse (n = 3) and two from each Apc1638N mouse (n = 2), representing the full length of the small and large intestines. In addition, seven fresh-frozen intestinal tumors were selected from each ApcMin mouse (two duodenum, nine jejunum, and ten ileum) and three from each Apc1638N mouse (two duodenum and four jejunum). RNA was extracted using the RNeasy Micro kit (Qiagen, Valencia, CA), amplified, labeled, and hybridized to Affymetrix MOE430A v.2 GeneChip as described previously (27). Probeset 1416558_at was chosen to represent Melk expression.
Small interfering RNA synthesis and preparation. The target sequences within Melk to which the RNA oligonucleotides were designed are the following: AACCCAAGGGTAACAAGGA (si1); CAGGCAAACAATGGAGGAT (si2); and TACTCACTACGCCAAATCG (si3). The control small interfering RNAs (siRNA) were designed against green fluorescent protein (GFP) with target sequences of: GCAAGCTGACCCTGAAGTTCAT (siC) and AAGATCCGCCACAACATCGA (siC_2). From the target sequences, 21-nucleotide RNAs were chemically synthesized in-house using TOM-RNA phosphoramidites and 2′-deoxynucleotides (Glen Research, Sterling, VA). Oligonucleotides were deprotected, gel-purified, and annealed as 10 μmol/L double-stranded RNA by incubating in annealing buffer [100 mmol/L potassium acetate, 30 mmol/L HEPES-KOH (pH 7.4), and 2 mmol/L magnesium acetate] for 1 minute at 90°C followed by 1 hour at 37°C. siRNAs were designed according to published guidelines (28, 29) and analyzed by Basic Local Alignment Search Tool (National Center for Biotechnology Information, Bethesda, MD) to confirm sequence specificity.
Small interfering RNA transfection and phenotypic analysis. Cells were plated at a density to achieve 80% confluence and then transfected 8 hours later with the appropriate siRNA (30 nmol/L) using LipofectAMINE 2000 (Invitrogen, Carlsbad, CA) according to the recommendations of the manufacturer. To analyze the effect of knockdown on anchorage-independent growth, cells were collected 12 to 15 hours posttransfection, suspended in medium with 0.3% agar (Difco, Franklin Lakes, NJ), and plated in replicates of six using six-well tissue culture plates. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma-Aldrich Corp., St. Louis, MO) was added 15 to 30 days postplating at a final concentration of 300 μg/mL and the number of colonies was visualized. Colony number was documented with a Fluorchem 8900 digital camera (Alpha Innotech, San Leandro, CA) and quantified using the autocount feature of the accompanying AlphaEase software. The effect of gene knockdown on cell proliferation was determined by plating cells 12 to 15 hours post–siRNA transfection at 2,000 per well of 96-well tissue culture plate in replicates of six for each time point. Cell growth was quantified by addition of MTT (590 μg/mL, final) at the appropriate time point. After at least 4 hours at 37°C, cells were then lysed with 7.5% SDS/3.7 N HCl and A570/690 nm was determined. Following assay setup, the remaining cells were replated and subsequently collected 72 hours posttransfection for real-time qRT-PCR and Western immunoblot confirmation of Melk knockdown. To analyze the effect of knockdown on cell cycle, cells were transfected in the middle of a double thymidine block as described by Hirota et al. (30). Briefly, after treating cells with 2 mmol/L thymidine for 18 hours, fresh medium was added and cells were allowed to recover for 3 hours before transfecting with the appropriate siRNA as described above. Four hours post–siRNA transfection, fresh medium containing 2 mmol/L thymidine was added and cells were cultured for an additional 14 hours at 37°C. Cells were then harvested and washed 24 hours after the final release, fixed with 70% ethanol, and resuspended in PBS with 3% fetal bovine serum and 40 μg/mL propidium iodide (Sigma-Aldrich). The effect of Melk overexpression on cell cycle was analyzed by cotransfection of HeLa cells with a Melk expression construct or empty vector mixed at a 5:1 molar ratio with pEGFP-C1 (Clontech, Palo Alto, CA). Cells were dispersed with trypsin 24 hours posttransfection, replated, and cultured for an additional 24 hours and then harvested as above for cell cycle analysis except for the addition of a 5-minute incubation with 0.5% paraformaldehyde before fixation with 70% ethanol. Cell cycle status was collected using a Coulter Epics XL-MCL (Beckman Coulter, Hialeah, FL) and data were analyzed with ModFit LT (Verity Software House, Topsham, ME).
Real-time quantitative reverse transcription-PCR (Taqman). RNA was isolated from human tumor cell lines and fresh-frozen tissues using the RNeasy Mini kit (Qiagen) according to the recommendations of the manufacturer. Samples were subjected to on-column DNase treatment during purification to remove genomic DNA contamination. For real-time qRT-PCR assays, 100 ng/well of total RNA was used as a template using the One-Step qRT-PCR kit (Qiagen). Probes consisted of a 5′-FAM reporter and 3′-Black Hole quencher. Duplicate wells were assayed for Melk expression (forward primer: 5′-AGAAGTGTGCCAGCTTCAAA-3′; reverse primer: 5′-CTAGATAGGATGTCTTCCACTAATCTTT-3′; probe: 5′-CCAGGCATCGCCCTTAAGCC-3′) using SDS7700 (Applied Biosystems, Foster City, CA) and the relative abundance of Melk transcript was normalized to RPL19 levels (forward primer: 5′-GCGGATTCTCATGGAACACA-3′; reverse primer: 5′-GGTCAGCCAGGAGCTTCTTG-3′; probe: 5′-AAGCTGAAGGCAGACAAGGCCC-3′) using the 2−ΔΔCt method (31). Tumor RNA from xenograft samples was isolated and amplified similarly but after tissue homogenization and by using primers for the murine orthologue of Melk (forward primer: 5′-CAGAGACCTGACGTGGTAGGC-3′; reverse primer: 5′-CACTAATCTCTTGTAAACCCAGGCAT-3′; probe: 5′-ACCCTTCAGCCGCTGTCTCCGG-3′) and murine RPL19 (forward primer: 5′-TTCTTGGTCTCTTCCTCCTTG-3′; reverse primer: 5′-ATGTATCACAGCCTGTACCTG-3′; probe: 5′-GGTCTAAGACCAAGGAAGCACGCAA-3′). Analysis was done using SDS software (version 1.7; Applied Biosystems).
Polyclonal antibody generation and Western blotting. A rabbit polyclonal antibody generated and affinity purified against the keyhole limpet hemocyanin (KLH)–conjugated peptide CSQGYAHRDLKPENLLFD (corresponding to amino acids 125-137 of Melk) was used for Western blot analysis. Cell lysates were prepared by addition of 50 mmol/L HEPES (pH 7.4), 100 mmol/L NaCl, 50 mmol/L NaF, 5 mmol/L β-glycerophosphate, 2 mmol/L EGTA, 1 mmol/L Na vanadate, 1% Triton X-100 to a cell pellet, resolved on a 4% to 12% acrylamide gel, and transferred to a polyvinylidene difluoride (PVDF) membrane (VWR). Fresh-frozen primary human tissue samples were prepared for Western analysis by the following method. Frozen samples were homogenized in liquid nitrogen with a mortar and pestle until tissue was a fine powder. Tissue powder was resuspended in 0.1 mol/L β-mercaptoethanol, 4 mol/L guanidium thiocyanate, 25 mmol/L Na citrate, and 0.5% N-laurylsarcosine. Samples were then layered over 5.7 mol/L CsCl/50 mmol/L EDTA and centrifuged at 280,000 × g in a SW55 Ti rotor (Beckman Coulter, Fullerton, CA) for 24 hours. Protein was recovered from the resulting supernatant by first incubating for 3 minutes at 90°C in the presence of 0.1 mol/L DTT and 2% SDS followed by precipitation with acetone (85% v/v) for 15 minutes at room temperature. The resulting protein pellet was resuspended in SDS sample buffer before being resolved on a 4% to 12% acrylamide gel and transferred to a PVDF membrane. Membranes were blocked in 10% powdered milk [made up in TBS with 0.1% Tween 20 (pH 7.4)], incubated with the anti-Melk antibody followed by probing with a horseradish peroxidase–conjugated goat anti-rabbit IgG (DakoCytomation, Carpinteria, CA), and visualized with a Fluorchem 8900 digital camera (Alpha Innotech). Blots were subsequently probed with an antitubulin antibody (Santa Cruz Biotechnology, Santa Cruz, CA) as a loading control.
Expression constructs. A cDNA encoding full-length human Melk was generated by PCR amplification from a pool of 107 cDNA libraries from various human tissues and cell lines. Based on sequences obtained from Genbank (accession number NM_014791), sense (5′-CCAAATAAACTTGCAAGAGGACTATGAAAGATTATGAT-3′) and antisense (5′-CCATCAATTATACCTTGCAGCTAGATAGGATGTCTT-3′) primers were designed to produce a PCR product of 1,986 bp using 200 ng of the pooled cDNA library as template. The amplification reaction was catalyzed by Pfu turbo DNA polymerase (Stratagene, La Jolla, CA) for 35 cycles under the following conditions: denaturing at 94°C for 1 minute, annealing at 65°C for 4 minutes, and extension at 72°C for 1 minute. The amplified product was then ligated into the pcDNA3.1 V5-HIS-TOPO Vector using TA-Cloning strategy (Invitrogen). This construct was then PCR amplified using sense primer (5′-CCACCATGAAAGATTATGATGAACTTCTCAAATAT-3′) and antisense primer (5′-TTACTTGTCATCGTCATCCTTGTAGTCTACCTTGCAGCTAGATAGGATGTCTTCCACTA-3′) to allow the addition of a Kozak consensus sequence and COOH-terminal FLAG tag, and TOPO-cloned into the pcDNA3.1D.v5-His-TOPO vector (Invitrogen). A kinase-dead Melk was created by mutation of the critical DFG motif, replacing the aspartic acid at amino acid 150 to an alanine. This mutation was created by a nested PCR strategy using the overlapping 5′-GCTGATTGCCTTTG-GTCTC-3′ sense and 5′-GAGACCAAAGGCAATCAGC-3′ antisense primers to introduce this mutation into the above wild-type Melk construct. Inactivation of Melk kinase activity by this D150A mutation has been shown by Vulsteke et al. (21).4
Our unpublished observations.
In situ hybridization and microdissection. cDNA templates were generated by PCR amplification with the Advantage polymerase kit (Clontech) using the sense (5′-CGCCAAATCGTTACACTACACCC-3′) and antisense (5′-CCGCTGCCTCCTGATACCC-3′) primers for human Melk (NM_014791) and the sense (5′-CGCCAGCACCACTCCAAAG-3′) and antisense (5′-AGCGCAGTGAGAACCTTATCCAG-3′) primers for murine Melk (NM_010790). Sense and antisense 33P-labeled riboprobes were transcribed in vitro from cDNA templates for human and murine Melk. Detailed methods for in situ hybridization have been previously described (24, 25). Formalin-fixed, paraffin-embedded whole sections and tissue microarray cores hybridized to antisense riboprobes were visualized by bright/dark field microscopy and scored positive or negative for Melk expression. Sense riboprobes were used to assess the specificity of hybridization.
To analyze RNA levels within isolated human colonic crypts, a fresh-frozen normal colon tissue sample was sectioned, fixed in ethanol, and stained with H&E. Microdissection of colonic crypts (luminal, middle, and basal thirds) was done with the SLμ-cut instrument (Molecular Machines and Industries AG, Glattbrugg, Switzerland). RNA was extracted using the Picopure RNA kit (Arcturus, Mountain View, CA) and amplified with the MessageAmp II kit (Ambion, Austin, TX) according to the instructions of the manufacturer. qRT-PCR was done as described above with 10 ng of amplified RNA.
In vivo tumor growth assays. To study the effect of Melk knockdown on the in vivo growth of transformed cells, we created a doxycycline-regulated short hairpin RNA (shRNA) expression construct.5
D. Gray et al., in preparation.
K. Hoeflich et al., submitted for publication.
Melk expression in cancer. Using the recently collated gene collection representing a complete listing of human kinases (the kinome; ref. 9), we first selected kinases with at least a 2-fold higher expression in one of the major tumor types compared with corresponding normal tissues. Data were analyzed by gene expression profiling in silico, a previously described expressed sequence tag–based, expression analysis tool (36). This gene list was then surveyed against a microarray data set comparing RNA levels between normal and cancer samples by hybridization to the Affymetrix HG-U133 GeneChip. As shown in Fig. 1A, Melk RNA levels were consistently increased in neoplastic tissues compared with nondiseased samples of the same organ. To confirm this finding, a qRT-PCR analysis of selected tissues was done. Figure 1B illustrates a 5- to 50-fold increase in Melk message from colon, breast, ovary, and lung tumor samples. In agreement with data derived from the microarray and RT-PCR analyses, an in situ hybridization analysis of human tumors scored the following as positive for Melk expression: 96% of colorectal adenocarcinomas (65 of 68 samples), 5% ductal breast carcinomas (3 of 57 samples), 23% lung cancers (4 of 40 adenocarcinomas, 6 of 26 squamous cell carcinomas, and 2 of 25 neuroendocrine tumors), 7% pancreatic adenocarcinomas (4 of 60 samples), and 13% of ovarian cancers (3 of 25 samples). Where present, the Melk in situ hybridization signal localized to areas of malignant epithelium as opposed to the surrounding stroma (Fig. 2). Analysis of normal tissues found low levels of Melk expression detectable only within epithelial cells of the lower portions of colonic crypts. However, expression at the base of normal crypts was barely detectable compared with the easily detected Melk signal within areas of malignant cells (see Supplementary Fig. S1).
Next, we investigated Melk expression in normal and neoplastic samples from a murine model of intestinal cancer. Samples were collected from histologically normal and spontaneous intestinal tumors that had developed within the ApcMin or Apc1638N heterozygotic mouse models. In both cases, one allele of the Apc gene has been truncated because of a nonsense mutation at codon 850 (ApcMin) or by insertion of the neomycin gene in the opposite orientation of the Apc transcription after codon 1638 (Apc1638N; ref. 26). Tumors that arise in either strain exhibit a loss or truncation of the wild-type Apc allele. RNA was harvested from normal and neoplastic intestine and hybridized to the Affymetrix Mouse Genome 430A GeneChip. As shown in Fig. 3A, relative expression for Melk is 2- to 6-fold higher in tumors from both ApcMin–and Apc1638N–derived mice compared with normal intestine. This finding was confirmed by qRT-PCR of selected samples (data not shown). Finally, an in situ hybridization inspection of formalin-fixed, paraffin-embedded tissues found the Melk signal to be restricted to the proliferating regions of the crypts in normal jejunum with increased expression in corresponding cancer samples (Fig. 3B-E). Interestingly, analysis of microdissected normal human colon tissue found a graded increase in Melk expression with descent into the crypt (Fig. 3F-G). Together, these findings suggest that Melk may play some role in the proliferation and/or self-renewal of progenitor cells that reside in the basal region of the crypt. This role may not be unique to the colon, as Melk expression has been shown to coincide with other stem cell–specific genes within neuronal tumors (37).
To confirm that Melk expression can also be detected at the protein level within tumor samples, a rabbit polyclonal antibody was generated and affinity purified against the KLH-conjugated peptide CSQGYAHRDLKPENLLFD (corresponding to amino acids 125-137 of Melk). Primary human tumor and normal tissue samples were resolved on 4% to 12% gradient gels and Western blotted with the polyclonal anti-Melk antibody. Figure 4 shows the presence of a band in the majority of lung, breast, and ovarian tumors that has the same mobility as a recombinant Flag-tagged Melk (Fig. 4B and C, lanes IP). The specificity of this antibody was further shown by the ability of the immunizing peptide to effectively compete against the anti-Melk but not an antitubulin antibody (Supplementary Fig. S2).
A comparison of Melk expression between primary tumor and normal samples derived from lung, breast, and ovary tissues (Fig. 4A-C) was then done. A representative Western blot for each tissue type with the densitometric ratio of Melk to tubulin within each sample listed at the top of every lane is shown. These data show that, on average, there is a 2-fold increase in Melk protein within primary lung- and breast-derived tumors as opposed to normal matched tissue (Fig. 4A and B), whereas expression levels within the ovary-derived samples are more closely matched between tumor and normal (Fig. 4C). Although the increased protein expression may not be as large as what would be expected from the elevated Melk mRNA levels (Figs. 1–3), we believe that this apparent discordance is a characteristic phenomenon surrounding the regulation of many other genes. In these cases, posttranscriptional mechanisms may also be simultaneously regulating expression depending on the cellular context. Examples include the effect of the 5′ untranslated region on the translational efficiencies of c-myc and fibroblast growth factor-2 (38), cell cycle–dependent changes in translational efficiency (p27Kip; refs. 39, 40) or protein degradation (Aurora; ref. 41), and differential polyribosomal recruitment of specific mRNA subsets following activation of Ras and Akt signaling (42). Similar mechanisms may be at play here, limiting the steady-state level of Melk protein visible within the primary tumor samples (Fig. 4).
Further supporting the notion of posttranscriptional mechanisms regulating Melk expression, we observed the presence of a slower migrating version of Melk isolated from multiple primary tumor samples (mainly lung; Fig. 4A). This is consistent with hyperphosphorylation of Melk, a posttranslational modification that has been shown to correlate with increased Melk kinase activity (16).4 Finally, it is interesting to note that an additional faster migrating band is present at the same intensity as Melk within the representative normal ovary samples (Fig. 4C). This band is consistent with a previously described, ovary-specific, and alternative spliced form of Melk (43). Although the authors found that the splicing event removed a portion of the kinase catalytic domain, the effect on activity has not been determined. However, it is intriguing that the ratio of full-length to truncated Melk seems to be skewed from an approximately equal molar ratio within normal ovary to a more predominant expression of the full-length version of Melk within ovarian tumor samples (Fig. 4C). Although the significance of these posttranslation modifications on Melk activity within cancer cells will need to be determined, it seems that Melk protein expression may be deregulated at multiple levels within cancer cells.
Due to the strong association between elevated Melk expression and primary cancer samples described above, we next inspected Melk expression in tumor cell lines. As it has been reported that Melk is regulated during the cell cycle, we analyzed RNA and protein levels of asynchronous and nocadazole-induced, mitotically arrested cells (Fig. 5; Supplementary Fig. S3). Upon mitotic arrest, we observed an increase in Melk at both the protein (Fig. 5A) and RNA levels (Fig. 5B). Interestingly, the level of Melk expression also seems to correlate with the transformed phenotype of the cell. Tumor-derived cell lines, such as HCT116 and Panc-1, consistently display increased Melk expression compared with IMR90 and Wi-38, both nontransformed diploid fibroblasts (Fig. 5A; Supplementary Fig. S3). Similarly, an increase in Melk expression is observed with V12N H-Ras–transformed NIH3T3 cells (3T3-Ras; Fig. 5A) compared with vector control cells (3T3-EV; Fig. 5A). These results provide additional evidence that elevated Melk expression correlates with the transformed phenotype of both murine and human cells.
Effect of Melk knockdown. The above association between Melk expression and cancer suggests that this kinase may be relevant for the growth and/or maintenance of a cancer cell. Conversely, the observed expression profile may simply be symptomatic of tumorigenesis with no ascribable functional consequence. If the latter scenario were true, inhibition of Melk expression would not be expected to hinder tumor growth. To begin distinguishing between these possibilities, we used RNAi to silence Melk expression in human tumor cell lines. Four 21-mer siRNA duplexes were designed and prepared according to previous reports (28, 29). The duplexes were transfected as described in Materials and Methods and the degree of silencing was assessed. Melk targeting duplexes si1 and si2 reduced RNA levels >80%, with si2 exhibiting the most prolonged reduction in Melk message (Fig. 6A). A minor, transient reduction in message was seen with si3, whereas no detectable effect was observed with a control GFP-targeting duplex (siC; Fig. 6A). Melk depletion was also confirmed by Western blot analysis. Therefore, significant knockdown of Melk protein was observed with si1 and si2, whereas si3 did not alter protein levels relative to the control siRNA. The lack of appreciable knockdown with si3, however, enables its use as an additional negative control.
Using these siRNA reagents, we sought to determine the consequence of Melk silencing within the cell. Because Melk expression (Fig. 5) and activity (15, 16) are regulated by the cell cycle, the effect of RNAi-mediated knockdown on cell proliferation was quantified. As shown in Fig. 7A-D, the growth rates of HEK293, PANC-1, BT549, and HCT116 cells were all inhibited when transfected with si1 and si2 duplexes, whereas the si3 duplex had little or no effect on growth relative to siC or mock-transfected cells. This result agrees with the level of Melk knockdown in Fig. 6B; the greatest level of silencing was consistently observed with si1 and si2, whereas a more modest and transient level of silencing was observed with si3. To determine if the decreased proliferation with Melk knockdown corresponds to an alteration of the cell cycle, the effect of silencing Melk on cell cycle progression was assessed. HeLa cells were transfected with siRNA duplexes in the middle of a double thymidine block as described in Materials and Methods. The cells were harvested 24 hours after release from the final thymidine block and the cell cycle status was determined by a propidium iodide–based fluorescence-activated cell sorting (FACS) analysis. An increase in the G2-M population was observed with si1 and si2 Melk targeting duplexes, whereas little or no effect was seen with si3 relative to two control duplexes (Fig. 7E; Supplementary Fig. S5). This result agrees with the degree of silencing observed for these duplexes, indicating that the phenotype is the result of Melk knockdown.
We next explored the effect of Melk overexpression on the cell cycle. To this end, HeLa cells were cotransfected with a wild-type or a kinase-dead (D150A) Melk mixed with a GFP expression construct at a 5:1 molar ratio. Forty-eight hours posttransfection, the cell cycle status of GFP-expressing cells was determined by a propidium iodide–based FACS analysis. As shown in Fig. 7F, we found an increase in the G2-M peak of cells overexpressing wild-type or D150A Melk compared with cells transfected with empty vector. This apparent G2-M delay upon overexpression of wild-type Melk (and to a lesser extent overexpression of a K40R inactive Melk) has been previously reported (16). It is intriguing to note that both Melk knockdown as well as overexpression of wild-type or kinase-dead Melk result in a G2-M delay in cell cycle progression. Because of the relative low resolution of FACS-based cell cycle analysis, a possible reconciliation of these data is that Melk may be required for progression through a G2-M checkpoint but once the cell has overcome this checkpoint, Melk expression and/or activity becomes inhibitory. In this way, a similar G2-M delay phenotype could result from both RNAi knockdown and forced overexpression of wild-type or a kinase-dead Melk mutant. Further work will be required to test this hypothesis and determine how Melk regulates this phase of the cell cycle.
Finally, we asked whether Melk expression is required for the transformed phenotype of tumor cell lines. We first used anchorage-independent growth, an in vitro assay that has been shown to serve as a good indicator of tumor growth in vivo (44, 45). Tumor-derived human cell lines were transfected with Melk-targeting siRNA duplexes and then plated on agar. Two to four weeks postplating, the number of viable colonies was counted. Figure 8A shows that colony formation is decreased by at least 50% in cells transfected with the si2 duplex relative to the si3 duplex or mock-transfected cells. Interestingly, except for a 30% reduction in colony number with BT549 cells, the si1 duplex did not inhibit PANC-1 or HEK293 colony formation. This result suggests that whereas the level of knockdown mediated by the si1 duplex is adequate to inhibit proliferation and cell cycle progression, it is not sufficient to prevent anchorage-independent growth.
To determine whether Melk expression is relevant for tumor growth in a more in vivo setting, we have created pHUSH, a H1 or U6 shRNA expression vector.5 As described in Materials and Methods, pHUSH is comprised of a modified H1 promoter (containing the TetO2 operon; ref. 34) with a RNA polymerase II promoter-tetracycline repressor expression cassette (35) on a single retroviral plasmid. With this vector, we can easily obtain stable clones with at least 90% knockdown of a target gene upon culturing cells with doxycycline. To study the effect of Melk knockdown, we selected the SVT2 (SV40-transformed 3T3) murine cell line due to its robust growth as a xenograft model and the fact that Melk RNA levels in these cells are ∼6-fold higher than nontransformed 3T3 cells (data not shown). Individual clones containing pHUSH with a Melk-targeting hairpin were generated. Those with at least 90% knockdown in Melk RNA after 72-hour incubation with 1 μg/mL doxycycline were selected for phenotypic analysis. As shown in Fig. 8B, tumor growth is reduced upon addition of doxycycline to the drinking water relative to mock treated (sucrose only) or tumors derived from cells containing the pHUSH vector with no hairpin oligonucleotide insert (EV). As with transient Melk knockdown in the human tumor lines described above, we have also observed decreased proliferation and anchorage-independent growth in vitro after doxycycline-induced Melk knockdown in the SVT2 clones (data not shown). Therefore, Melk expression seems to be relevant for the transformed phenotype of both human and murine tumor cell lines.
As part of an effort to identify novel drugable targets, we have found that Melk, an AMPK-related serine/threonine kinase, is overexpressed in multiple tumor types. Originally discovered through an analysis of microarray data, we have confirmed this pattern of expression using qRT-PCR, in situ hybridization, and Western analyses of human and mouse tumor samples. We provide evidence that Melk plays an important role in cell division as RNAi knockdown inhibited proliferation and slowed the transition of cells through the G2-M phase of the cell cycle. This result, together with the finding that RNAi-mediated Melk knockdown inhibited colony formation as well as tumor growth in vivo, suggests that Melk expression may be relevant with respect to establishing and/or maintaining certain types of cancer.
Other members of the AMPK family have also recently been shown to provide a vital function within cancer cells. First, numerous reports have suggested an important link between AMPK and the ability of a cell to survive a hypoxic challenge. Activated AMPK signaling within endothelial cells was shown to be required for initiating an angiogenic response to hypoxia, including phosphorylation of epithelial nitric oxide synthase and nitric oxide production (46, 47). In addition, hypoxic activation of AMPK has been shown to positively regulate hypoxia-inducible factor 1 transcriptional activity (48). AMPK activation has also been correlated with the ability of tumor cells to survive nutrient withdrawal (49). Finally, signaling via ARK5/NUAK1, a more recently discovered AMPK-related kinase, has been suggested to play a role in tumor survival and invasion. ARK5 was originally identified as a component of the Akt signaling pathway (50). Its expression seems to be elevated in colon tumors where it may play a positive role in Akt-stimulated tumor cell invasion (51, 52). Whether Melk impinges on any of the signaling pathways mentioned above remains to be determined. Interestingly, a recent study provided evidence that the tumor suppressor LKB1 may serve as a master activating kinase for all AMPK family members except Melk (53). This result suggests that Melk activity is regulated by a distinct mechanism relative to the other AMPK members.
In addition to our current results showing Melk expression in multiple tumor types, previous studies have reported elevated expression in pediatric brain tumors (37, 54, 55). Hemmati et al. (37) went on to show that Melk expression correlated not only with the presence of other stem cell–specific genes but also depended on the multipotency of the explanted primary tumor. From their findings, the authors suggested that pediatric brain tumors may arise from progenitor cells with multipotent stem cell–like attributes. The possibility that Melk may be a stem cell marker for other cell types is consistent with our analysis of Melk expression between normal and tumor-derived colon tissue from both humans and mice. As shown in Fig. 3, we found that Melk is expressed within spontaneous tumors as well as the basal regions of crypts within normal gastrointestinal epithelium. The basal crypt region is the stem cell compartment for the normal colon, providing the necessary multipotency and regenerative capacity to replace the differentiated cells found along the crypt walls and lumen. Although it is unclear whether colon tumors arise primarily from the basal crypts or may also derive from differentiated cells found higher up the crypt wall or within the luminal surface, intestinal tumors are frequently found to contain a high proportion of cells with an undifferentiated, crypt-like morphology and gene expression profile (26). Therefore, the presence of Melk within normal stem cells of the basal crypt region as well as intestinal tumor cells supports the hypothesis that this kinase plays a stem cell–specific role. If this model is correct, the ability to target and inhibit Melk activity may be an effective method for eliminating tumor cell progenitors.
The function of Melk within normal or tumorigenic stem cells remains to be elucidated. The work described herein, as well as other studies, suggests a role in maintaining proper progression during meiosis and mitosis. First, Heyer et al. (43) proposed that Melk has a sex-specific role in meiotic maturation of the oocyte based on expression profiling. Second, the finding that Melk expression and kinase activity is maximal in Xenopus embryos and mammalian cells during mitosis (15, 16), as well as its interaction with reported putative substrates (16, 20, 21), further implicates Melk as an important player within the cell cycle. Finally, we show that Melk is expressed in a number of primary tumors and tumor-derived cell lines, consistent with an established elevated mitotic index for these cells. This correlation agrees with the recent report that expression of Melk and other mitotic-specific genes is a hallmark of retinoblastoma loss (56). We have provided additional evidence for a functional connection to the cell cycle by showing that RNAi-mediated Melk knockdown decreases the proliferative capacity and delays G2-M progression of tumor cell lines. Interestingly, we show that forced overexpression of either wild-type or a kinase-dead mutant of Melk also results in a G2-M delay. Although these results seem to be in apparent disagreement, a possible reconciliation of the data is that Melk expression and/or kinase activity regulates both G2-M progression and exit. Melk may be required for progression through a G2-M checkpoint but once the cell has overcome this checkpoint, Melk expression or activity becomes inhibitory. In this way, a similar G2-M delay phenotype could result from both RNAi knockdown and forced overexpression of wild-type or a kinase-inactive mutant Melk. Further work will be required to determine whether these checkpoints exist and how Melk regulates this phase of the cell cycle. Regardless of the mechanism of action, it is becoming clear that Melk has a functional role in the cell cycle. Our expression analysis, combined with in vitro and in vivo siRNA knockdown studies, suggests that this activity is functionally maintained in many cancer cells. Therefore, blocking Melk-related signaling may provide a therapeutic advantage in cancer treatment.
In summary, we have provided evidence that links Melk expression to cancer, the first step in evaluating the potential of this kinase as a therapeutic target. The apparent role of Melk within highly proliferative tumor cells (and possibly tumor stem cells) may provide a therapeutic benefit by targeting not only the proliferating tumor cells but also the progenitors from which they are derived. Further studies will be required to explore these possibilities in detail and to determine whether a small molecule should be developed that targets the kinase-active site or inhibits via an allosteric mechanism.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank James Cupp for expert help with cell cycle analysis of FACS data, Ron Smits for providing RNA samples fom APCMin and APC1638N mice, and Mark Vasser for the synthesis of DNA and siRNA oligonucleotides.