Microtubule Affinity-Regulating Kinase 2 is associated with DNA damage response and cisplatin resistance in non-small cell lung cancer

Abstract

Microtubule affinity-regulating kinases (MARKs) are involved in several cellular functions but few studies have correlated MARK kinase expression with cancer, and none have explored their role in lung cancer. In this study, we identified MARK2 as frequently disrupted by DNA hypomethylation and copy gain, resulting in concordant overexpression in independent lung tumor cohorts and we demonstrate a role for MARK2 in lung tumor biology. Manipulation of MARK2 in lung cell lines revealed its involvement in cell viability and anchorage-independent growth. Analyses of both manipulated cell lines and clinical tumor specimens identified a potential role for MARK2 in cell cycle activation and DNA repair. Associations between MARK2 and the E2F, Myc/Max, and NF-κB pathways were identified by luciferase assays and in-depth assessment of the NF-κB pathway suggests a negative association between MARK2 expression and NF-κB due to activation of non-canonical NF-κB signaling. Finally, we show that high MARK2 expression levels correlate with resistance to cisplatin, a standard first line chemotherapy for lung cancer. Collectively, our work supports a role for MARK2 in promoting malignant phenotypes of lung cancer and potentially modulating response to the DNA damaging chemotherapeutic, cisplatin.

INTRODUCTION

Lung cancer is the number one cause of cancer death worldwide¹. The five year survival rate for lung cancer patients is only 18%, and the dismal prognosis is attributable to the lack of early detection methods and curative treatments². Although progress has been made over the last decade in identifying actionable mutations driving a small percentage of lung tumors, much remains to be understood about lung cancer biology^3–5. Thus, a better understanding of disease biology is needed to improve patient outcome and to develop new treatment strategies designed to inhibit the molecular mechanisms contributing to lung tumorigenesis.

Microtubule affinity-regulating kinases (MARKs) are serine/threonine protein kinases involved in the phosphorylation of microtubule-associated proteins such as Tau, cell cycle-regulating phosphatases such as CDC25, and class IIa histone deacetylases such as HDAC7^{6, 7}. Murine knockout studies revealed MARKs to have multiple functions, including roles in neuronal differentiation, neurodegeneration, cell polarity, intracellular transport, and cell migration (reviewed by Matenia et al.⁷); some of these functions are commonly deregulated in cancer cells. There are four MARK genes (MARK1-4) in humans. MARK3 is overexpressed in hepatocellular carcinoma cell lines and associated with nuclear accumulation of beta-catenin⁸, and MARK2 overexpression has been observed in cisplatin resistant cancer cell lines⁹. Since cisplatin is part of standard first line therapy for lung cancer patients^10–12 the relevance of MARK2 in lung cancer warrants exploration. In this study we sought to investigate the potential oncogenic role of MARK2 in lung cancer.

MATERIALS AND METHODS

Molecular Profiling

Genetic alterations were analysed in lung tumors from the BCCRC (77 lung adenocarcinoma (LUAD), http://edrn.nci.nih.gov/science-data) and The Cancer Genome Atlas (TCGA) cohorts (230 LUAD, 482 squamous cell carcinoma (LUSC)). Sample acquisition, processing, clinical information, genomic profiling and data analyses details for the BCCRC cohort are described elsewhere^{13, 14}. Level 3 processed data from the TCGA was downloaded from the TCGA data portal. In the BCCRC cohort, genetic alterations were defined relative to patient matched non-malignant tissue, whereas alterations in TCGA tumors were defined with reference to the beta value median (DNA methylation) and RSEM distribution (mRNA expression: RNA-Seq relative abundance estimation by Expectation-Maximization) of available normal tissues (125 with methylation and 108 cases with RNAseq). We used standard thresholds for defining alterations in expression (fold change >2 for BCCRC, z-score >2 for TCGA), methylation (delta-beta-value (∆β) >0.2 for BCCRC and TCGA), and copy number (log-ratio >0.3 for TCGA, and as previously described for the BCCRC¹³). Copy number data was generated using Affymetrix SNP 6.0 arrays (BCCRC, TCGA), methylation by Illumina HM27K (BCCRC, TCGA) and HM450K arrays (TCGA) (MARK2 probe: cg06204948), and expression by Illumina Human WG6v3.0 arrays (BCCRC, MARK2 probe: ILMN_1736747) and RNA sequencing (TCGA). Gene expression profiles for lung cancer cell lines were generated using Affymetrix Human PrimeView expression microarrays and processed using Affymetrix Expression Console software. Cancer cell line cisplatin IC50 data were retrieved from the Sanger drug sensitivity project¹⁵.

In vitro assays

LUAD cell lines (H1437, H1395, H1650, H1993, H2228, H1693) were cultured in RPMI-1640 media supplemented with 10% FBS and 0.1% penicillin-streptomycin. Non-malignant lung bronchial epithelial cells immortalized by the introduction of hTERT and CDK4 (HBEC-KT)¹⁶ were cultured in KSFM media supplemented with 50μg/μl bovine pituitary extract and 5ng/μl EGF. MARK2 levels were modulated in cell lines using stable lentiviral shRNA constructs and a tetracycline inducible expression system using a non-tagged wild-type MARK2 expression vector or a kinase-dead MARK2 (T208A/S212A) mutant expression vector¹⁷. Both expression vectors were verified by sequencing. Five different shRNA constructs were assessed, and the shRNAs producing the greatest magnitude knockdown were used for subsequent experiments (Supplementary Figure 1, Supplementary Table 1). RT-qPCR, Western blots, cell viability assays, colony formation, and dose response assays were performed as previously described^18–21. To investigate cancer pathways associated with MARK2, transcription factor reporter assays were performed using the Cignal Finder Cancer 10-Pathway Reporter assay which assesses: WNT, NOTCH, p53, TGFβ, E2F, NFκB, Myc/Max, HIF1A, ERK, and JNK. Cell cycle and phosphorylated-γH2AX intensity after cisplatin treatment were assessed using a flow cytometry protocol modified from Huang et al²². Briefly, cultured cells were treated for 3 hours with cisplatin at 5 μM, cisplatin media was replaced with normal growth media, and then cells were incubated for 6, 14 or 24 hours before assessing cell cycle and phosphorylated-γH2AX. At each time point, cells were collected and fixed with 70% ethanol at −20°C for at least 12 hours. The staining protocol was modified from²². The working solution was 1% BSA-PBS, and anti-γ-H2AX (Cell Signaling #9718S) and FITC-conjugated anti-rabbit immunoglobulin (BD Pharmingen #554020, San Diego, CA, USA) antibodies were both diluted at 1/500. Cells resuspended in PBS containing propidium iodide (0.02 mg/ml) and RNase A (0.05 mg/ml) were left 30 min at room temperature in the dark and cellular fluorescence was measured using a FACSCanto (Becton-Dickinson). Doublet exclusion was performed using PE-A versus PE-H gating. Cells in the sub-G0/1 phase (characterized by low DNA content due to loss of fragmented DNA caused by ethanol permeabilization) that were also within the gate for viable cells (forward versus side scatter dot plot) were considered apoptotic. All reagents, vendors, and catalogue numbers are listed in Supplementary Table 1. Statistical analyses of experimental results were conducted using GraphPad Prism v6. All assays were performed at least in duplicate.

Transcriptomic analyses

Pre-ranked gene set enrichment analyses (GSEA) were performed on gene expression data generated for PLKO and shMARK2 cell lines (H1650, H1993, H1693). Fold change data (shMARK2 relative to PLKO) for all genes were input and tested for enrichment in the C3 transcription factor target and C2 canonical pathway gene sets for each cell line²³. The same pre-ranked analyses were performed on expression data for tumor and non-malignant BCCRC tissue samples separately, using fold change data generated by comparing the 10 highest and 10 lowest MARK2 expressing cases. Only the gene sets with concordant enrichment (i.e. positive or negative) in all cell lines, an FDR q-value <0.05 in at least 2 of the 3 cell lines, and a concordant enrichment trend in tumor samples were considered further. A list of genes significantly associated with MARK2 expression was generated using Significance Analysis of Microarrays (SAM) in R statistical software²⁴. Genes from the SAM analysis with an estimated FDR q-value less than 0.05 were used as input for Ingenuity Pathway Analysis (IPA).

RESULTS

MARK2 is frequently and concurrently altered at the DNA and mRNA levels in NSCLC

To assess gene expression changes of the four MARK kinases, we analyzed 77 LUAD tumors with patient matched non-malignant tissue with expression profiles (BCCRC cohort, Supplementary Table 2). MARK1 displayed similar frequencies of over- and underexpression and aberrant expression of MARK3 and MARK4 was rare, suggesting these MARKs are not critical to lung tumorigenesis (Supplementary Figure 2). Conversely, MARK2 was overexpressed in 46.8% of LUAD, and expression was significantly higher in tumors relative to non-malignant tissue (Fig.1A&D, p<0.0001). To determine whether MARK2 expression is associated with DNA level alterations, we assessed copy number and DNA methylation status of MARK2 in the same 77 tumors. We observed a high frequency of hypomethylation (33.8%) and copy number gain (20.8%) with 30% of tumors displaying concurrent and significantly correlated copy number gain or hypomethylation and overexpression (Fig.1A–C, Supplementary Table 2). In the fraction of tumors not harboring copy number gains or hypomethylation alterations, we suspect MARK2 overexpression may be driven by alternative epigenetic mechanisms, such as miRNA or other non-coding RNAs, which we were unable to assess.

Figure 1. MARK2 is frequently altered in clinical NSCLC specimens.

(A) Frequencies of DNA methylation, gene dosage and mRNA expression alterations for tumors in the BCCRC (77 LUAD) and TCGA cohorts (230 LUAD, 482 LUSC). The OverExpr/UnderExpr+DNA columns indicate tumors with concurrent mRNA and DNA (either copy number or methylation) alterations. (B) Spearman’s correlation between MARK2 copy number and expression levels in the BCCRC tumors. (C) Spearman’s correlation between MARK2 delta-β-value and expression levels in the BCCRC tumors. Box plots with 97.5 and 2.5 percentiles showing the distributions of MARK2 expression in BCCRC samples (D), TCGA samples (E), and methylation β-values for the MARK2 promoter CpG site assessed (F). ****= Mann Whitney U test p-value was < 0.0001.

To assess the reproducibility of MARK2 overexpression in our cohort, we validated our findings in the LUAD (n=230) and LUSC (n=482) cohorts from TCGA. Similar to the BCCRC cohort, MARK2 expression was significantly higher in tumor samples relative to non-malignant samples (Fig.1E, p<0.0001). Comparison of MARK2 expression in LUAD and LUSC revealed no difference, suggesting MARK2 expression is not subtype specific. At the DNA level, TCGA LUAD tumors displayed similar frequencies of MARK2 hypomethylation (27.8%), copy number gain (26.1%) and overexpression (50.9%) as the BCCRC cohort, validating our findings (Fig.1A&F). While the 482 LUSC samples showed overexpression (60.8%) and hypomethylation (33.4%) frequencies similar to the two LUAD cohorts, copy number alterations of MARK2 were rare (4.6% showed gain) in LUSC (Fig.1A&F). Examination of DNA sequence data revealed mutations to be a rare mechanism of MARK2 disruption in lung cancer (2/230 LUAD and 3/482 LUSC, TCGA data, Fig.1A). Taken together, these results demonstrate MARK2 is frequently overexpressed in NSCLC, irrespective of histological subtype, and that overexpression is likely mediated by DNA hypomethylation and to a lesser extent DNA copy gain.

MARK2 displays oncogenic properties in vitro

Given the high frequency of MARK2 overexpression observed in tumors, we assessed the role of MARK2 in lung cancer cell lines (LCCLs) and in an immortalized human bronchial epithelial cell line (HBEC) using lentiviral shRNAs and expression vectors. To assess the phenotypic effects of MARK2 expression, we chose LUAD cell lines with high (H1693, H1650 and H1993) and low (H2228, H1395 and H1437) MARK2 protein expression (Fig.2A). MARK2 knockdown (shMARK2, efficiency shown in Fig.2B and Supplementary Fig.1&3) significantly decreased cell viability in all cell lines (Fig.2C); however, the cell lines expressing higher endogenous levels of MARK2 were more sensitive to knockdown than those with low expression. HBECs were extremely sensitive to MARK2 knockdown, indicating basal levels of MARK2 may be required for viability in non-malignant bronchial epithelial cells, and that the effect of MARK2 knockdown is not cancer-specific. Conversely, overexpression of MARK2 in HBECs (to levels comparable to those of high expressing LUAD cell lines) resulted in a small but significant increase in viability at the last time point, suggesting MARK2 may impair contact inhibition rather than enhance cell proliferation when overexpressed in non-malignant lung epithelial cells (Fig.2C). Similar results were observed for colony formation assays in soft agar (CFSA), with knockdown significantly impairing anchorage-independent growth of all cell lines to similar extents, suggesting MARK2 may play a role in establishing anchorage-independent growth (Fig.2D).

Figure 2. MARK2 expression is associated with a malignant phenotype in NSCLC cells.

(A) MARK2 expression in the 7 cell lines used for in vitro experiments. (B) shRNA knockdown (shMARK2) and ectopic expression (MARK2) efficiency relative to control vectors (PLKO and LACZ). GAPDH was used as a loading control. (C) Cell viability as assessed by the MTT assay following shRNA knockdown in lung cancer cell lines or ectopic expression in HBECs. (D) Soft agar colony formation ability of shMARK2 cell lines relative to controls. H1693 and HBEC lines do not form colonies in soft agar and therefore are not shown. (E) Ectopic expression of MARK2 using a tetracycline inducible expression system. The table shows a color code for the different conditions assessed and the blot results below demonstrate the levels of MARK2 protein expression. (F) Cell viability as assessed by MTT assays following induction of MARK2 expression. (G) Colony formation ability of H1437 MARK2 and control lines before and after expression induced by tetracycline. Experiments were performed in triplicates and two-tailed t tests between manipulated lines and controls were performed on day 5 for viability assays and after 4 weeks for colony formation (*:p<0.05, **:p<0.01, ***:p<0.001). shRNA clone 4 targeting the MARK2 coding sequence was used for knockdowns (Supplemental Table 1).

To further elucidate the involvement of MARK2 in cell viability and anchorage independence, a tetracycline inducible expression system in combination with shRNA-mediated knockdown was used. For these experiments, we chose H1437 (low MARK2 expression) as it enabled us to obtain biologically relevant MARK2 expression levels (Fig.2E and Supplementary Fig.3C) without drastically reducing cell viability. As expected, in the absence of tetracycline, knockdown of MARK2 significantly decreased viability (Fig.2F). Tetracycline-induced MARK2 expression abrogated the effect of knockdown and led to increased viability relative to LACZ controls, which showed no difference in cell viability upon exposure to tetracycline (Fig.2F). Similar results were obtained for CFSA, with overexpression of MARK2 enhancing anchorage-independent growth relative to knockdown and LACZ controls (Fig.2G). These results further confirm a role for MARK2 in cell viability and anchorage-independent growth, supporting an oncogenic role for MARK2 in LUAD.

Lastly, we aimed to determine whether the detrimental effect of MARK2 knockdown on lung cancer cell viability was dependent on its kinase activity. First we treated lung cancer cells with high endogenous MARK2 expression (i.e. those that are most sensitive to MARK2 knockdown) with an ATP competitive inhibitor which has been demonstrated to elicit a potent ability to specifically inhibit the kinase activity of MARK2²⁵. In all three lines, the MARK2 kinase inhibitor failed to reduce cell viability below 50% of control cells, suggesting the dramatic reduction in cell viability upon MARK2 knockdown is not kinase-dependent (Fig. 3A). To confirm this observation, we assessed whether a kinase-dead form of MARK2 was capable of rescuing the shRNA-mediated MARK2 knockdown phenotype. We generated the well-characterized T208A/S212A kinase-dead MARK2 mutant^{7, 17, 25}. Non-synonymous mutations at the threonine 208 and serine 212 residues abrogate MARK2 kinase activity by preventing phosphorylation of MARK2, rendering its substrate cleft in an open conformation which prevents MARK2 catalytic activity^{7, 17, 25}. MTT assays were then conducted to assess cell viability in H1437 cells co-transduced with LACZ (control)/wild-type-MARK2/kinase-dead-MARK2 (T208A/S212A) and shMARK2. As previously shown (Fig.2C), MARK2 knockdown (shMARK2-LACZ) resulted in significantly reduced cell viability compared to control cells (PLKO-LACZ), however, both wild-type (shMARK2-WT) and kinase-dead mutant (shMARK2-MUT) cells showed significantly improved viability compared to cells with knockdown alone (shMARK2-LACZ), demonstrating the ability of both wild-type and kinase-dead MARK2 to rescue the knockdown phenotype (Fig.3B). These findings are consistent with the MARK2 kinase inhibitor results, suggesting MARK2’s effect on contributing to the malignant phenotype of lung cancer cells is independent of its kinase function.

Figure 3. MARK2’s contribution to the lung cancer phenotype is independent of its kinase activity.

(A) The effect of MARK2 inhibition on viability of high MARK2-expressing cells (H1693, H1650, H1993) was analyzed using a small molecule MARK2 kinase inhibitor (MKi). Cells were incubated with serial dilutions of MKi for 72 hours and viability was assessed using the MTT assay. Viability is presented as a proportion of untreated control cells. Standard deviations are indicated. (B) Cell viability was assessed using MTT assays and compared between H1437 control cells (PLKO-LACZ) and cells stably co-transduced with shMARK2 and LACZ, wild-type MARK2 (WT) or a kinase-dead (T208A/S212A) MARK2 mutant (MUT). P-values indicating significant differences in viability on Day 5 are indicated, as calculated using a Student’s t-test. shRNA clone 1 targeting the 3′UTR of MARK2 was used for knockdown (Supplemental Table 1).

MARK2 expression is associated with NF-κB activity, DNA damage repair, and cisplatin resistance

In an attempt to elucidate the cellular pathways through which MARK2 functions in lung cancer, we performed luciferase-based reporter assays on 10 transcription factors of known cancer-related signaling pathways (Fig.4A). Across all cell lines, WNT, HIF1A, E2F, and Myc/Max activity were noticeably reduced while NF-κB activity was significantly increased in knockdown relative to control lines. HBECs overexpressing MARK2 displayed the opposite pattern of transcription factor activity, substantiating the luciferase results in LCCLs.

Figure 4. MARK2 expression is associated with oncogenic pathways such as NF-κB, and DNA damage repair in NSCLC cells.

(A) Assessment of MARK2 knockdown (left, H1693) and MARK2 overexpression (right, HBEC) on the activity of 10 cancer signaling pathways, as assessed by luciferase reporter assays. (B) Selected gene sets enriched for genes significantly associated with MARK2 expression, as determined by Gene Set Enrichment Analysis (GSEA). Genes with differential expression associated with MARK2 in manipulated cell lines, high and low MARK2-expressing tumors, and high and low MARK2-expressing normal lung tissues were used as input for GSEA. (C) Measurement of NF-κB pathway components and phosphorylated-γH2AX in MARK2-manipulated cell lines by Western blot. GAPDH was used as a loading control.

To gain further insight into the signaling pathways associated with MARK2, we performed genome wide expression profiling on control (PLKO) and knockdown (shMARK2) cells from the three cell lines with high endogenous expression of MARK2. Expression profiles were also assessed in the 10 highest and 10 lowest MARK2 expressing tumor and non-malignant cases from the BCCRC dataset to identify genes significantly associated with MARK2 expression (Supplementary Table 3, SAM analysis). Pre-ranked GSEAs were performed on all genes with a 2-fold or greater expression change in 1) cell lines, 2) tumors, and 3) non-malignant tissues. 56 gene sets passed our stringent criteria of concordance in results across cell line and tumor analyses (outlined in the methods). These gene sets included those related to DNA repair, E2F, Myc/Max and NF-κB transcription factors (Fig.4B, Supplementary Fig.4, Supplementary Table 4), further supporting the luciferase assay findings and implicating MARK2 in these pathways. To further investigate the negative association between MARK2 expression and NF-κB observed in both the luciferase and GSEA analyses, we assessed the NF-κB pathway by Western blotting. A significant increase in phospho-p65 in shMARK2 cells relative to controls was observed, confirming the inverse relationship between MARK2 and NF-κB pathway activity (Fig.4C). Interestingly however, none of the canonical NF-κB pathway components upstream of p65 were affected upon manipulation of MARK2 expression (Fig.4C), which led us to hypothesize that MARK2 may be involved in non-canonical NF-κB activity, which can be activated by DNA damage²⁶. Since DNA repair was among the most significant gene sets and pathways associated with MARK2 in our analyses (Fig.4B, Supplementary Fig.4 and Supplementary Tables 4–6), we investigated whether MARK2 knockdown influenced DNA damage by assessing γH2AX levels, an indicator of impaired DNA damage response (DDR)^{27, 28}. Indeed, we observed an increase of γH2AX in MARK2 knockdown cells compared to controls, providing direct evidence of a potential role for MARK2 in DDR (Fig.4C).

Our observations of impaired DDR upon MARK2 knockdown and in MARK2 lowly expressing tumors combined with a previously described association of MARK2 expression with cisplatin sensitivity⁹ prompted us to investigate the potential clinical relevance of MARK2 in the context of lung cancer treatment with cisplatin – a DNA damaging agent and standard first line chemotherapy for the treatment of NSCLC. Dose-response assays revealed 2 out of 3 lines with low MARK2 expression (H2228 and H1437) had significantly lower IC50s than the 3 cell lines with high MARK2 expression (p=0.012, Supplementary Figure 5 and Supplementary Table 7). The drastic change in cell viability after MARK2 knockdown in cell lines with high MARK2 expression precluded us from comparing the sensitivity of these cells to cisplatin before and after MARK2 knockdown. However, using the tetracycline inducible expression system, we observed a significant decrease in the cisplatin sensitivity of H1437 cells expressing MARK2 compared to LACZ control cells (Fig.5A, p=7.0×10⁻⁴). Integration of LCCL expression and IC50 data from the Sanger drug sensitivity projects¹⁵ comparing cell lines with the highest (n=15) and lowest (n=15) MARK2 expression further validated our observations; LCCLs with high MARK2 expression had significantly higher cisplatin IC50 values than those with low expression (Fig.5B, p=1.45×10⁻²).

Figure 5. MARK2 expression is associated with cisplatin sensitivity of NSCLC cells.

(A) Cisplatin dose response curves for H1437 control (LACZ, green) and MARK2-overexpressing (MARK2, blue) cells. (B) Comparison of cisplatin IC50s in NSCLC lines with high (n=15) and low (n=15) MARK2 expression (Wellcome Trust Sanger Institute, Genomics of Drug Sensitivity project).

To determine whether the observed difference in cisplatin sensitivity might be attributable to the amount of DNA damage sustained, cells were treated with 5μM cisplatin for 3 hours then incubated in regular growth media for 6, 14 or 24 hours and phosphorylated-γH2AX levels and cell cycle phase analyzed by flow cytometry (Fig.6 and Supplementary Fig.6). These results clearly showed enhanced blockage of cells in S phase in lines with lower MARK2 expression (Fig.6C&E), which was associated with a higher intensity of γH2AX and therefore increased DNA damage (Fig.6D&F). Importantly, we noted fewer than 5% of cells in the Sub-G1 apoptotic phase, suggesting the reduced γH2AX levels we observed in cell lines with low MARK2 expression is not due to apoptosis-induced DNA damage. Furthermore, we noted that there was an increased population of cells in the G2/M phase in high compared to low MARK2 expressing cells. Taken together, these results suggest that the reduced DNA damage sustained in high MARK2 expressing cells likely contributes to the increased resistance of these cell lines to cisplatin, highlighting the potential clinical relevance of MARK2 expression.

Figure 6. MARK2 is potentially involved with DNA damage repair and cell cycle regulation.

(A, B) Difference in cell cycle distribution (top) and γH2AX intensity (bottom) at 0, 6 and 24 hours post cisplatin treatment in cells with low (H2228) and high (H1693) endogenous MARK2 expression levels. (C) Graphical representation of cell distribution among the Sub-G1, G1, S, and G2/M phases of the cell cycle. (D) Statistical comparison of phosphorylated-γH2AX intensity between high and low MARK2-expressing cells. The p-value indicated is based on triplicate experiments. To ensure comparison of cells with the same amount of DNA, intensity of γH2AX is shown for the G1 subpopulation 24 hours after cisplatin treatment. (E) Comparison of the percentage of cells in S or G2/M phase of the cell cycle 24 hours post cisplatin treatment in high and low MARK2-expressing cells. (F) Comparison of phosphorylated-γH2AX intensity levels in cells in S or G2/M phase of the cell cycle 24 hours post cisplatin treatment in high and low MARK2-expressing cells. Results for each cell line, time point and cell phase are presented in Supplementary Figure 6.

DISCUSSION

MARKs have been implicated in multiple cellular processes, however despite their involvement in critical functions, few studies have assessed their role in cancer. In this study we performed an integrated genetic analysis on multiple independent datasets and found that MARK2 was frequently overexpressed in NSCLC, irrespective of histological subtype. Moreover, its overexpression appears to be mediated by DNA alterations, specifically DNA hypomethylation and to a lesser extent DNA copy gains. Manipulation of MARK2 expression in multiple LCCLs and HBECs revealed its involvement in cell viability and anchorage-independent growth, DNA damage and cisplatin sensitivity.

Knockdown of MARK2 via shRNA caused a marked reduction in cell viability in LUAD cell lines with high endogenous levels of MARK2. Treatment of cells exhibiting the highest MARK2 levels (H1993, H1650, and H1693) with an ATP competitive inhibitor specific for MARK2’s kinase activity²⁵ revealed only a small reduction in cell viability, as opposed to the large effect we observed upon shRNA-mediated knockdown. This could suggest that MARK2’s involvement in the malignant phenotypes we observed could be independent of its kinase function and could possibly be due to non-catalytic functions such as scaffold protein interactions, which represent known and putative functions of MARK family members including MARK2^{7, 25, 29, 30}. To address this hypothesis, we assessed whether a kinase-dead mutant form of MARK2 (T208A/S212A) was capable of rescuing the drastically reduced viability we observed upon shRNA-mediated MARK2 knockdown. Indeed, both wild-type and kinase-dead forms partially rescued the MARK2 knockdown phenotype. Thus, our findings suggest that MARK2’s contribution to the malignancy of lung cancer cells likely occurs, at least in part, through a kinase-independent function.

Transcription factor reporter assays further suggested an oncogenic role for MARK2 in lung cancer by promoting WNT, HIF1A, E2F and Myc/Max activity. Of note, MARK2 has been previously implicated in canonical and non-canonical WNT signaling^31–33. Conversely, NF-κB activity was negatively correlated with MARK2 expression. Transcriptomic analyses of both manipulated cell lines and clinical tumor specimens provided further evidence supporting a role for MARK2 in the NF-κB pathway, cell cycle, and DDR. There are several possible ways MARK2 could be involved in DDR. MARK2 could phosphorylate and thereby regulate the activity of DDR or DNA repair proteins, or could interact with DDR proteins via a scaffold function. Alternatively, MARK2 could contribute indirectly to DDR by regulating the cell cycle to allow time for damaged DNA to be repaired. Supporting this hypothesis, our transcriptomic analyses implicated MARK2 in the cell cycle, and MARK proteins have previously been implicated in cell cycle progression via phosphorylation of cell cycle regulators such as CDC25⁷. A recent study found that activation of the MARK2 homologue par-1 in C. elegans negatively regulates DNA replication in S phase of two-cell embryos, although the effector of MARK2’s involvement in replication was not identified³⁴. Based on this established evidence, MARK2’s role in DDR could be attributable to its regulation of cell cycle kinetics.

As DNA repair and the NF-κB pathways were among the most significant, positively enriched gene sets correlated with MARK2 expression, and since we observed a significant negative correlation between MARK2 expression and DNA damage levels, we hypothesized that MARK2 may be linked to non-canonical NF-κB activation via DNA damage. Although we validated the negative correlation between MARK2 expression and NF-κB activation by assessing p65 levels, none of the NF-κB pathway components upstream of p65 were altered upon manipulating MARK2 expression, suggesting MARK2 does not function through the canonical NF-κB pathway. Intriguingly, non-canonical NF-κB signaling can be activated through DNA damage leading to p65 phosphorylation, thereby weakening its affinity for the inhibitor IkBα, resulting in gradual accumulation of active p65/p50 heterodimers and NF-kB transcriptional activity²⁶. While NF-κB activation is generally associated with tumor promotion, in certain contexts it may suppress tumor growth^{35, 36}. Our experiments revealed an increase in NF-κB activation upon MARK2 knockdown, and the latter significantly reduced lung cancer cell viability. NF-κB can exert pro- or anti-apoptotic effects through upregulation or repression of its target genes^{35, 36}. Thus, it is possible that in the context of MARK2 knockdown/DNA damage, NF-κB behaves as a tumor suppressor and consequently, could potentially impair cell growth in MARK2 knockdown cells. However, this speculation requires experimental validation.

Taken together, these observations led us to assess the association between cisplatin sensitivity, DNA damage and MARK2 expression, as cisplatin is commonly used in lung cancer treatment. A significant association between MARK2 expression and cisplatin resistance was observed in all 3 independent analyses (Sanger’s drug sensitivity project, comparison of low versus high expressing MARK2 cell lines, and ectopic expression of MARK2), highlighting the potential clinical relevance of MARK2 expression as a marker of cisplatin resistance. Of interest, the low expressing MARK2 cell line that was least sensitive to cisplatin (H1395) was the only cell line in the panel with wild-type TP53. Given the importance of TP53 in DNA repair, it is not surprising that this cell line was more resistant. Based on the strong correlation between MARK2 expression and DDR components and our observations of higher DNA damage levels in cell lines with low MARK2 expression, we suspect MARK2 may contribute to cisplatin resistance through modulation of the DDR pathway. However, further investigation to understand how MARK2 specifically interacts with DDR components is needed to unequivocally demonstrate this relationship.

In summary, we describe for the first time highly frequent DNA and RNA level disruption of MARK2 in clinical NSCLC cases, and provide multi-faceted evidence supporting a novel potential role for MARK2 in the DNA damage response pathway as well as cisplatin sensitivity.

Supplementary Material

Supp FigureS1-S6

Supplementary Figure 1 – Efficacy of 5 shRNA constructs in knocking down MARK2 expression.

shRNA Clone 4 which targets the coding sequence of MARK2 (mature sequence: TAAGGCTTTAGTTCATCATCT) and shRNA Clone 1 which targets the 3’UTR of MARK2 (mature sequence: TTAGGCGAAATACTCTGTGCA) produced the greatest knockdowns of MARK2 at the protein level, and were used to generate stable lung cancer cell lines with MARK2 knockdown as specified in the respective figure legends.

Supplementary Figure 2 – Expression status of MARK genes in the BCCRC cohort.

Frequencies of over- and underexpression are indicated for MARKs 1–4 in 77 lung adenocarcinomas. MARK expression was considered in individual tumors and a fold-change was calculated by dividing expression in the tumor by expression in matched non-malignant tissue. Fold-changes exceeding 2-fold were considered aberrantly expressed.

Supplementary Figure 3 – RT-qPCR of MARK2 mRNA expression.

Assessment of MARK2 mRNA expression was performed by RT-qPCR using the ΔΔCt method for (A) cell lines transduced with PLKO and shMARK2 vectors and (B) HBEC transduced with ectopic LACZ and MARK2 expression vectors. C) Expression levels for the various conditions assayed using the tetracycline-inducible system established in H1437 cells. Results for each line are normalized to expression levels in control cells (PLKO LACZ without tetracycline). PLKO: empty vector pLKO.1, shM: pLKO.1 – shRNA construct targeting MARK2, LACZ: tetracycline-regulated β-galactosidase control vector, MARK2: tetracycline-regulated MARK2 vector (wild-type).

Supplementary Figure 4 – Enrichment plots from gene set enrichment analyses.

GSEA enrichment plots depicting positive enrichment for the DNA repair reactome gene set, and transcription factor targets of E2F, NF-κB and MYC/MAX. Enrichment scores (green line) represent the probability that the gene set is positively (left side) or negatively (right side) enriched in a ranked gene list. Genes are ranked based on differences between MARK2 high and low expressing groups. Ranked genes that appear in the defined gene set are indicated as “hits” below the enrichment profile (black hash marks). Genes ranked near the top of the gene list are underscored with a red bar, whereas genes near the bottom of the list are underscored with a blue bar.

Supplementary Figure 5 – Comparison of cisplatin sensitivities in lung cancer cell lines with low and high MARK2 expression.

Cipslatin dose response curves for cell lines with high endogenous MARK2 levels (blue – H1650, H1693 and H1993) and with low endogenous MARK2 levels (green – H1395, H1437, H2228). H1395 was the only cell line tested with wild-type TP53.

Supplementary Figure 6 – Cell cycle and γH2AX intensity results for individual cell lines tested.

Analyses of cell cycle and γH2AX intensity in low and high MARK2 expressing cells after treatment with cisplatin. The left side of the figure shows the distribution of cells in various phases of the cell cycle for each cell line and at different time points as indicated. The right side of the figure shows the mean intensity levels of γH2AX in the different subpopulations of the cell cycle for each cell line at the different time points. Experiments were performed in triplicate for each cell line.

Supp TableS1-S7

Supplementary Table 1 – Reagents information.

Supplementary Table 2 – Genomics data indicating MARK2 disruption in the BCCRC cohort.

This table lists the frequency of tumor samples with expression fold changes (tumor/matched normal) greater than 2-fold for each of the MARK genes, and the copy number, methylation delta-β-values, and expression fold-changes for MARK2 in the BCCRC lung adenocarcinoma cohort.

Supplementary Table 3 – Genes identified by significance analysis of microarrays (SAM).

This table lists over- and underexpressed genes identified by the SAM analysis (http://statweb.stanford.edu/~tibs/SAM/sam.pdf) performed on shMARK2 versus control cell lines, and high versus low MARK2 expressing tumors and normal tissues.

Supplementary Table 4 – Gene set enrichment analysis (GSEA) results.

Gene sets listed in the table have passed the threshold criteria of FDR q value < 5% in at least 2 of the 3 MARK2 knockdown cell lines assessed and are observed in all 3 lines as either “negatively correlated with shMARK2 phenotype and positively correlated with PLKO phenotype (positive NES, pink)” or “positively correlated with shMARK2 phenotype and negatively correlated with PLKO phenotype (negative NES, green)”, with the same trend observed in tumor samples.

Supplementary Table 5 – Genes comprising the REACTOME_DNA_REPAIR geneset.

This table lists the genes that comprise the GSEA geneset, “Reactome_DNA_Repair.”

Supplementary Table 6 – Ingenuity pathway analysis.

Genes from the SAM analysis (Supplementary Table 3) with estimated FDR q values less than 0.05 were considered for target gene analysis using Ingenuity Pathway Analysis software (http://www.ingenuity.com/). This table lists the pathways associated with MARK2 expression. Table A indicates pathways significantly enriched by genes associated with MARK2 (identified by comparing expression profiles of PLKO-control and MARK2-knockdown cancer cell lines). Table B indicates pathways significantly enriched by genes associated with MARK2 (genes common to analyses comparing expression profiles of PLKO-control and MARK2-knockdown cell lines AND high and low MARK2-expressing non-malignant tissues).

Supplementary Table 7 – Cisplatin IC50s for the 6 NSCLC cell lines assessed.

IC50s for cisplatin sensitivity were calculated using GraphPad Prism v6 (least squares fit linear regression model). Two way ANOVAs were performed to compare the two lines with low MARK2 expression (H1437 and H2228) to the three lines with high expression (H1650, H1693, H1993).

Novelty and impact.

We describe for the first time highly frequent DNA and RNA level disruption of MARK2 in clinical NSCLC cases, and provide evidence supporting a novel role for MARK2 in affecting the oncogenic properties of several lung tumor phenotypes including response to the DNA damaging chemotherapeutic agent, cisplatin.

Abbreviations

BCCRC

British Columbia Cancer Research Centre

CDC25

cell division cycle 25A

E2F

E2 transcription factor

HDAC

Histone deacetylase

H2AX

histone 2A family member X

LCCL

lung cancer cell line

LUAD

lung adenocarcinoma

LUSC

squamous cell carcinoma

MARK

Microtubule affinity-regulating kinase

Max

Myc-Associated Factor X

Myc

Myelocytomatosis oncogene cellular homolog

NF-κB

Nuclear factor κ B

NSCLC

Non-small cell lung cancer

RT-qPCR

Real time quantitative polymerase chain reaction

sh

short hairpin RNA

DDR

DNA damage response

TCGA

The Cancer Genome Atlas

TP53

tumor protein p53