CKAP2L Promotes Non–Small Cell Lung Cancer Progression through Regulation of Transcription Elongation

Lung cancer remains the leading cause of cancer-related death worldwide (1). Although the recent outstanding improvements in the survival rates due to tyrosine kinase inhibitors and immunotherapy (2), only 16% of patients affected by non–small cell lung cancer (NSCLC) survive the disease after 5 years from diagnosis, with most patients being diagnosed at advanced stages characterized by metastatic dissemination (3).

Part of the challenge in treating lung cancer, as many other cancer types, resides in the high complexity and heterogeneity of the cancer genetic landscape. Indeed, the bulk tumor consists of a diverse collection of cells characterized by distinct molecular signatures, giving rise to a dissimilar sensitivity and response to treatments. Chromosomal instability (CIN), the prevalent form of genome instability, has been recognized as an important contributor to tumor heterogeneity (4). CIN is defined as the ongoing acquisition of structural (mutations, translocations, deletions) and numerical (aneuploidy) chromosomal aberrations as a result of errors in chromosome segregation during mitosis, and it has been shown to drive metastatic dissemination (5, 6) and correlate with poor prognosis (7). Therefore, the mechanisms underlying CIN have attracted more interest in the past years as a potential source for the identification of targets to tackle tumor progression and offer new therapeutic opportunities in cancers such as lung adenocarcinoma, which harbors high rates somatic mutation and genomic rearrangements (8).

Thus, with the aim to discover new players associated with CIN in the progression of lung adenocarcinoma, we performed a bioinformatics analysis of The Cancer Genome Atlas (TCGA) dataset LUAD comparing primary lesions versus normal lung and identified 58 genes that are overexpressed in lung adenocarcinoma and correlate with high chromosomal instability. Among those genes, CKAP2L (Cytoskeleton-associated protein 2-like) emerged as an attractive target for further investigation given its strong correlation with poor patient outcome and in consideration of the insufficient understanding of its functions and mechanisms of regulation, especially in a cancer setting.

Experimental evidence on its biological role has been initially provided by the study of Radmis, the mouse orthologous of CKAP2L. Radmis was found to function as a regulator of microtubule dynamics in neural stem/progenitor cells. Overexpression of Radmis in vitro was reported to induce defects in mitotic spindle formation and mitotic arrest, while Radmis downregulation resulted in multipolar mitotic spindle structure and chromosome missegregation (9).

Here, we provide evidence of a previously unrecognized role of CKAP2L in transcription elongation. Through direct binding to RNA Pol II and regulation of phosphorylation at serine 2, CKAP2L promotes the expression of genes involved in the cell cycle, E2F signaling, and microtubule–cytoskeleton dynamics, thus enabling tumor progression and dissemination. Furthermore, we suggest that combining CKAP2L depletion with inhibition of the transcriptional cyclin-dependent kinase 9 (CDK9) using alvocidib should be further investigated as a therapeutic approach in lung cancer. Overall, we propose that targeting CKAP2L might expose NSCLC, and likely other cancer types, to transcriptional vulnerabilities.

TCGA data version 2016_01_28 for LUAD (mRNASeq and SNP6 copy number datasets for hg19 assembly) were downloaded from Broad GDAC Firehose. Differentially expressed genes (DEG) were determined from mRNASeq datasets by comparing normal and tumor samples using R/Bioconductor package DESeq2 v.1.26.0 (10). Correlation analysis was then performed to identify a set of DEGs whose expression patterns are significantly correlated with their copy number ratio in tumor samples by using Pearson correlation coefficient ≥0.17 as the cutoff criteria (11). CIN (expr) was calculated by summing the normalized expression across these genes. The proportion of altered genome was calculated by summing the length of regions with copy number events.

All human samples subject to The Human Tissue Act 2004 consent provisions have been obtained with written informed consent of the donor and collected under the MCRC biobank Research Tissue Bank ethics (ref: 18/NW/0092) in accordance with the Declaration of Helsinki.

IHC was performed using tissue microarrays (TMA) containing 1-mm cores of 103 human samples to include normal lung tissue, lung adenocarcinomas, and metastatic tissues. Cores were provided in duplicates for all samples except for two, where sufficient material was not available. TMA sections were stained using anti-CKAP2L antibody (Thermo Fisher Scientific, PA5–58778) and the Olympus VS120-L100-W-12 (Olympus Corporation) was utilized to image under transmitted light via built-in Koehler illumination for a BX61VS frame. Digitization was achieved via an Olympus VC50 camera (2/3″ CCD camera, 3.45 μmol/L × 3.45 μmol/L pixel size).

Chromogenic staining was quantified using HALO 3.0.331.299. Regions of interest (ROI) for each core were identified from hematoxylin and eosin staining by user-specified tissue annotations to teach a machine learning classifier. CKAP2L expression levels were then quantified on a serial section as percentage of DAB-positive cells within the ROI using the Multiplex IHC v2.3.4 module. For duplicated cores, the average value was calculated.

All animal experiments were approved by Cancer Research UK Manchester Institute's Animal Welfare and Ethical Review body in accordance with the Animals Scientific Procedures Act 1986 and according to the ARRIVE guidelines and the Committee of the National Cancer Research Institute guidelines. Experimental procedures were conducted under the project license P72E31537 (M. Garofalo). For the generation of the lung orthotopic model, 6- to 8-week-old female NOD/SCID mice (Charles River Laboratories) were anesthetized with isoflurane and injected percutaneously into the left thorax with H1299 cells (2 × 10⁶) stably transduced with nontargeting shRNA or shCKAP2L (7 mice/group), using a 0.5-mL syringe. Animals were then sacrificed at clinical endpoint when showing significant signs of physical (bodyweight loss greater than 10%, severe ascites), behavioral (e.g., grooming activity), or respiratory distress.

H1299, A549, Calu-1, H460, HBEC3-KT, and 293T cell lines were purchased from ATCC. H1299, A549, and H460 were cultured in RPMI1640 media, Calu-1 and 283T cells were cultured in DMEM, and HBEC-3KT were maintained in Airway Epithelial Cell Basal Medium (ATCC PCS-300–030). All media were supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin except for HBEC-3KT cells, where media were supplemented with Bronchial Epithelial Cell Growth Kit (ATCC PCS-300–040). Cells were expanded initially upon receipt and aliquoted into frozen stocks. All cell lines were validated by STR profiling using PowerPlex 21 (Promega) and routinely tested for Mycoplasma and maintained in culture for <3 months, at which point, a new early-passage vial was thawed.

Custom FlexiPlate siRNAs (Qiagen) were dissolved in RNase-free water according to the manufacturers' instructions. H1299 cells, stably transduced with a GFP lentiviral vector (Santa Cruz Biotechnology, sc-108084), were reverse transfected with 50 nmol/L individual siRNAs and 48 hours after transfection cells were reseeded in the Oris Pro Cell Migration plates (PROCMACC1, Platypus Technologies). Cells were incubated in a humidified chamber with 5% CO₂ at 37°C for 2 hours to permit cell attachment, after which, premigration images were acquired. Cells were further incubated for 24 hours at 37°C with 5% CO₂ to allow migration and postmigration images captured. All images were acquired using the PerkinElmer Operetta high-content imaging system. Images were analyzed using ImageJ 2.0 software and cell migration was calculated as percent gap closure applying the equation (Premigration_area–Migration_area)/Premigration_area × 100. The siRNA screen was performed in duplicate and the mean of gap closure was calculated and represented in the heatmap.

RNA extraction for subsequent RNA sequencing (RNA-seq) analysis was performed with the RNeasy Kit (Qiagen) with on-column DNase digestion. Indexed PolyA libraries were prepared using 200 ng of total RNA and 14 cycles of amplification with the SureSelect Strand Specific RNA Library Prep Kit for Illumina Sequencing (Agilent, cat no: G9691B). Libraries were quantified by qPCR using the Kapa Library Quantification Kit for Illumina sequencing platforms (Roche, cat no: 07960336001). Single-read 75-bp sequencing was carried out by clustering 2.0 pmol/L of the pooled libraries on a NextSeq 500 sequencer (Illumina Inc.).

RNA-seq reads were quality checked and aligned to the human genome assembly (GRCh37) using the RSubread v1.32.4 package (12) with the default settings. Mapped data were converted to gene level integer read counts using featureCounts (13) available within RSubread package and the Ensemble GTF annotation (Homo_sapiens.GRCh37.74).

DEGs were identified by comparing the gene level integer read count data of the knockdown to the control samples using R/Bioconductor package DESeq2 v.1.26.0 (10). DESeq2 DEGs determination involved multiple steps. In brief, it uses normalization factors to model read counts to account for sequencing depth. Then, it estimates gene-wise dispersions and performs shrinkage analysis to generate more accurate estimates of dispersion to model the counts. Finally, it fits the negative binomial model and performs hypothesis testing using the Wald test. The resulting P values were adjusted using the Benjamini and Hochberg approach for controlling the FDR. Genes with adjusted P values ≤ 0.05 (FDR ≤ 0.05) were considered as differentially expressed. All calculations were performed in R v.3.5.3.

Gene set enrichment analysis (GSEA) was performed using hallmark gene sets from the Molecular Signatures Database (MSigDB) version 7.0. Functional analysis of downregulated genes was performed using the Bioconductor package clusterProfiler (14) and Volcano plot generated with the Bioconductor package EnhancedVolcano (https://github.com/kevinblighe/EnhancedVolcano).

Chromatin immunoprecipitation (ChIP) was performed as previously described (15). Briefly, H1299 cells were crosslinked using 1% formaldehyde for 10 minutes at room temperature and then quenched with glycine for 5 minutes at room temperature. Chromatin was sonicated using Diagenode Bioruptor and immunoprecipitated overnight with antibodies that were preincubated with Dynabeads Protein G. Immunoprecipitated samples were washed once in low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris-HCl pH 8.0, 150 mmol/L NaCl), once in high salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris-HCl pH 8.0, 500 mmol/L NaCl), once in LiCl wash buffer (0.25 mol/L LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mmol/L EDTA, 10 mmol/L Tris-HCl pH 8.0), and then DNA was eluted in elution buffer (1% SDS, 100 mmol/L NaHCO₃). Samples were reverse cross-linked by overnight incubation at 65°C and then treated with RNase A and proteinase K. DNA was finally purified using a PCR Purification Kit (Qiagen). The sequences of primers used for ChIP quantitative PCR (ChIP-qPCR) are listed in Supplementary Table S1.

Sequencing reads from ChIP and Input samples were quality checked using FASTQC. All bases with a Phred quality score ≤ 20 and any adapter sequences present in the data were removed using FASTX-Toolkit. The cleaned and trimmed FASTQ files were then mapped to the hg19 reference assembly using Bowtie2 (v2.2.1; ref. 16). The aligned data in SAM format were processed using samtools. v1.3.1 (17) to remove unmapped reads, and to retain reads that mapped in proper pair with a mapping quality ≥ 20. Finally, the SAM files were converted into BAM files and indexed.

For peak calling and annotation, we used MACS2 v2.1.2 (18) to identify genomic regions enriched for ChIP signal. Homer v4.10 (19) was used to annotate the significant peaks and TSS/gene body profiles were generated by ngsplot v2.61 (20).

The differential binding analysis was performed using diffReps v1.55.4 (21). Because of the lack of biological replicates, we performed differential analysis test using G-test. Functional enrichment analysis of differential binding peaks was performed using R and Bioconductor package ChIP-Enrich, v.2.10.0 (22).

Pausing Index/Travel Ratio was calculated using “analyzeRNA.pl” program available in Homer v4.10 (19). Promoter–proximal pause region was set to 500 bp and “-normMatrix 1e6” parameter was used to get RPKM (reads per kilobase of transcript per million mapped reads) counts.

Cells were seeded at 1 × 10⁴ cells/well in CellCarrier Spheroid ULA 96-well microplates (PerkinElmer) and allowed to form spheroids over 48 to 72 hours. Spheroids were then embedded by gently removing 100 μL of growth medium, which was replaced with 100 μL of Matrigel (Corning, 354230) and preinvasion images were captured. Spheroids were subsequently incubated for 48 hours at 37°C with 5% CO₂ to allow invasion. Ninety-six–well microplates were imaged using high content screening via the PerkinElmer Opera Phenix (PerkinElmer), a confocal spinning disk 4 laser (405 nm 50 mW, 488 nm 50 mW, 591 nm 50 mW, 640 nm 50 mW) fixed light path system, with a range of emission filters (435–550 nm, 435–480 nm, 500–550 nm, 570–630 nm, 650–760 nm). Four Zyla sCMOS cameras, 2,160 × 2,160 pixels, 6.5 μm pixel size (Andor) are set up for each dedicated lightpath. Using the Zeiss EC Plan Neofluar x5 air objective NA 0.16 WD 12.1 mm images were acquired for the entire area. Data were captured using the Harmony software (PerkinElmer) and then transferred to the Columbus (PerkinElmer) server for data analysis and archival.

Global RNA synthesis was analyzed using the Click-iT RNA Alexa Fluor 488 Imaging Kit (Thermo Fisher Scientific). Briefly, cells were seeded in a black CellCarrier-96 Ultra microplate (PerkinElmer) and on the day of the analysis they were incubated with 1 mmol/L 5-ethynyl uridine (EU) for 90 minutes, then fixed with 3.7% formaldehyde in PBS for 15 minutes at room temperature and permeabilized with a solution of 0.5% Triton X-100. Cells were then stained using the Click-iT reaction cocktail, according to the manufacturer's instructions and images acquired with a 40× water objective using Opera Phenix (PerkinElmer). Image analysis was performed using Columbus software by measuring the mean fluorescence intensity in the 488 channel for each nucleus.

Statistical analyses were performed using GraphPad Prism 7. For comparison of two groups, a two-tailed unpaired t test was used, whereas multiple groups were compared using ANOVA with a post hoc Tukey test. Log-rank P value was used for the survival curve (Kaplan–Meier) analyses. *, P < 0.05; **, P < 0.01.

ChIP sequencing (ChIP-seq) and RNA-seq data reported in this study have been deposited with links to BioProject accession number PRJNA689472 in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/). Mass spectrometry data have been deposited in the PRIDE database under the accession number PXD023425. A list of proteins interacting with CKAP2L identified by mass spectrometry is provided in Supplementary Table S2.

To identify new vulnerabilities associated with CIN, we analyzed gene expression data from TCGA for primary tumors of 517 patients with lung adenocarcinoma and compared their profile with 59 normal lung samples. Bioinformatics analyses identified a number of genes that were significantly overexpressed in the tumor samples and correlated with high copy number variation and CIN (Fig. 1A). A further correlation analysis was carried out to confirm a possible involvement of these genes in CIN. The results showed a significant positive correlation between the expression of the majority of these genes and CIN calculated from genome-wide copy number alteration in tumor samples (Supplementary Fig. S1A and S1B; ref. 23). To select genes implicated in both CIN and migratory capability and identify novel mechanisms interconnecting those two hallmarks of cancer, we performed a small-scale RNAi screen in H1299 cells, a human non–small cell lung carcinoma cell line, using an siRNA‐based library comprising three separate siRNAs targeting 48 CIN-associated genes. By using a gap closure assay analyzed by high-content microscopy, we assessed the effect of each siRNA individually for the ability to impair cell migration. The results of the analysis are summarized in Fig. 1B, where the cells migratory activity is expressed as a percentage of gap closure. Genes were chosen as potential “candidates” if at least two siRNAs reduced cell migration by an arbitrary threshold of 40% and, after applying the selection criteria, a list of 10 candidates was generated, which included multiple genes already implicated in genome stability, such as WD repeat and HMG-box DNA-binding protein 1 (WDHD1), kinetochore protein NDC80 homolog (NDC80), targeting protein for Xklp2 (TPX2), and a microtubule-associated protein (CKAP2L).

Analysis of the RNAi screen indicated that all three siRNAs against CKAP2L strikingly reduced cell migration (Supplementary Fig. S2A). Although CKAP2L has been previously reported to be indispensable for mitotic spindle formation, thus playing a role in chromosome segregation (9), and to promote invasion of human lung adenocarcinoma cell lines (24), its function still remains largely unexplored. Furthermore, CKAP2L expression is significantly enriched in lung adenocarcinoma samples (Supplementary Fig. S2B) and high expression has been shown to be a strong predictor of poor patient outcome (24). Therefore, integration of the RNAi screen results with clinical data prompted us to further validate and characterize the role of CKAP2L in lung cancer.

Examination of a 103-sample tissue microarray of human lung adenocarcinoma revealed significantly higher CKAP2L expression levels in metastatic tissues and in primary lung adenocarcinoma samples in comparison with normal lung tissue (Fig. 1C). Analysis of the TCGA lung adenocarcinoma cohort also indicated elevated CKAP2L levels in advanced (T stage 4) versus early (T stage 1 and 2) disease and in patients with lymph node metastasis versus patients without distant metastases (M1 vs. M0; Supplementary Fig. S2C). Finally, the analysis of 33 different datasets from the TCGA revealed higher levels of CKAP2L across multiple tumors in comparison with normal samples obtained from the TCGA and GTEX datasets (Supplementary Fig. S2D). Taken together, these data support the clinical relevance of CKAP2L in lung adenocarcinoma and, possibly, in other tumor types.

Alterations such as amplifications or mutations in the CKAP2L gene are infrequent events in human lung adenocarcinoma (Fig. 2A), implicating that the elevated mRNA levels found in tumor samples occur as a result of deregulation at the transcriptional level. To identify transcription factors that might regulate expression of CKAP2L, we interrogated a large collection of ChIP-seq experiments performed by the ENCODE project and found MYC occupancy at the promoter region of CKAP2L. We confirmed MYC binding to two sites on CKAP2L promoter by ChIP-qPCR (Fig. 2B) and, in addition, we observed a significant decrease in CKAP2L mRNA levels upon MYC silencing in three different NSCLC lines (H1299, Calu-1, and H460; Fig. 2C; Supplementary Fig. S3A), while overexpression of exogenous MYC resulted in increased CKAP2L expression (Fig. 2D). Notably, downregulation of CKAP2L did not affect MYC at mRNA or protein levels, indicating the absence of a feedback loop (Supplementary Fig. S3B). Given that MYC is a key downstream effector of KRAS signaling, for which oncogenic mutations are drivers in approximately 30% of NSCLC, we hypothesized that KRAS would contribute to CKAP2L transcriptional regulation. Indeed, downregulation of KRAS by siRNA (Fig. 2E; Supplementary Fig. S3C) or inhibition of RAS signaling with trametinib (Supplementary Fig. S3D) decreased CKAP2L mRNA levels to a similar extent of MYC downregulation. In support of a KRAS-dependent expression, overexpression of exogenous KRAS significantly rescued CKAP2L mRNA levels that were reduced by KRAS silencing (Fig. 2F). Overall, our findings corroborate a role for KRAS and MYC in the regulation of CKAP2L expression in NSCLC.

We next examined the functional requirement for CKAP2L in human NSCLC cell lines. CKAP2L depletion by two distinct siRNAs (Supplementary Fig. S4A) or shRNA (Supplementary Fig. S4B) significantly reduced cell proliferation (Fig. 3A; Supplementary Fig. S4C), colony formation (Fig. 3B) and 3D cell invasion (Fig. 3C) in comparison with control cells. Cell-cycle analysis upon CKAP2L depletion revealed a significant increase of cells in the G₁ phase, accompanied by a reduction in the S-phase in comparison with control cells (Fig. 3D). Interestingly, CKAP2L silencing also induced apoptotic cell death, which occurred more prominently in H1299 cells in comparison with A549 cells or normal human bronchial epithelial cells (HBEC), in agreement with their differential expression of CKAP2L (Fig. 3E; Supplementary Fig. S4D).

To further assess the impact of CKAP2L depletion in vivo, we established a lung orthotopic model whereby H1299 cells, stably expressing an shRNA targeting CKAP2L or a nontargeting control (Supplementary Fig. S4B), were injected into the left thorax of NOD/SCID mice. Analysis of the tumor samples revealed that CKAP2L depletion did not affect levels of the nuclear antigen Ki-67 (Supplementary Fig. S4E), which is a marker of cell proliferation that is expressed in all phases of the cell cycle except G₀, but caused a significant increase in apoptotic cell death, as assessed by cleaved caspase-3 (Fig. 3F), and significantly extended overall survival of tumor-bearing mice in comparison with mice injected with control cells (Fig. 3G). Collectively, our data demonstrate that CKAP2L depletion impaired tumor growth in vitro and in vivo.

As an initial step toward understanding of CKAP2L oncogenic function, we performed immunoprecipitation of endogenous CKAP2L followed by mass spectrometry to identify potential binding partners. Interestingly, among the numerous proteins interacting with CKAP2L (Supplementary Table S2), we found key components of the transcriptional machinery: RPB1 and RPB2, which are the two largest subunits of the RNA polymerase II (RNA Pol II). Interaction between endogenous or exogenous CKAP2L and RPB1 was further validated by coimmunoprecipitation experiments (Fig. 4A and B). To evaluate the effect of CKAP2L downregulation on global transcription, we measured the incorporation of the uridine analogue 5-ethynyl uridine (EU) into nascent RNA. Intriguingly, RNA synthesis was modestly reduced after CKAP2L downregulation (Fig. 4C), suggesting that CKAP2L might contribute to RPB1-mediated gene transcription.

To specifically identify genes exhibiting reduced transcription, we profiled the genome-wide distribution of RPB1 in CKAP2L-deficient cells by performing ChIP followed by deep sequencing (ChIP-seq). Notably, we observed a global reduction of RPB1 binding at proximal–promoter sites (≤1 kb) and increased binding within distal intragenic regions in CKAP2L-depleted cells compared with control cells (Supplementary Fig. S5A). Differential binding analysis followed by GSEA indicated that CKAP2L depletion resulted in reduced binding of RNA Pol II at genes involved in the mitotic cell cycle, such as multiple cyclin-dependent kinases and cyclins, E2F1 signaling but also microtubule cytoskeleton organization and cell projection organization (Fig. 4D; Supplementary Table S3). Interestingly, several genes encoding proteins involved in chromosome segregation and spindle assembly, such as SKA1, CENPE, NEK2, CKAP5, PLK1 among others (Supplementary Table S3), were identified.

Regulation of the transcription cycle by the RNA Pol II occurs in three stages: initiation, elongation, and termination. After transcription initiation, the RNA Pol II undergoes pausing at the promoter–proximal region, followed by its release into productive transcription elongation (25). Thus, the notion of Pol II traveling ratio (TR), which identifies the ratio of Pol II bound at the promoter–proximal region as opposed to the gene body, has been previously introduced and used to assess RNA Pol II transition between the phases of initiation and elongation (26), which can impact transcription rates. Essentially, higher TR values indicate a lower efficiency of Pol II transition from initiation to elongation, therefore a reduction in productive transcription elongation.

We evaluated the effect of CKAP2L depletion on the TR of RNA Pol II for those genes identified in the enrichment analysis. Intriguingly, we observed an increase in the TR upon CKAP2L downregulation in comparison with control cells (Fig. 4E; Supplementary Fig. S5B), indicative of a decreased transition of RNA Pol II into the elongation phase. MYC has been previously shown to regulate transcriptional pause release of RNA Pol II (27) and, more recently, to also control the assembly of processive Pol II elongation complexes (28). Notably, analysis of complex formation between CKAP2L and RNA Pol II was not affected by MYC depletion (Supplementary Fig. S5C), thus indicating that such interaction occurs independently of MYC. Taken together, these results propose a novel and important role for CKAP2L in the regulation of RNA Pol II transcriptional elongation for a subclass of genes involved in cell-cycle regulation and mitotic microtubule cytoskeleton organization.

To corroborate and further explore the function of CKAP2L in the regulation of gene expression, we performed RNA-seq analysis of CKAP2L-depleted cells and identified over 3,400 genes that were significantly differentially expressed (Supplementary Fig. S6A). Interestingly, gene ontology (GO) analysis showed that significantly downregulated genes were enriched in GO terms involved in biological processes such as nuclear division, chromosome segregation, and microtubule organization (Fig. 5A and B). Furthermore, GSEA identified signatures for G₂–M checkpoint, E2F targets, mitotic spindle, and epithelial–mesenchymal transition (Fig. 5C).

Given the well-recognized role of the family of E2F transcription factors in the regulation of cell-cycle progression (29) and in genomic stability (30, 31), we further assessed the effect of CKAP2L downregulation on the expression levels of the activator proteins (E2F1, E2F2, and E2F3) and found significantly reduced expression levels of E2F1 and E2F2 as well as decreased levels of some of their established targets (Fig. 5D and E). Furthermore, the transcriptome analysis identified downregulation of the antiapoptotic genes MCL-1 and BCL2 in CKAP2L-depleted cells, which was confirmed by qPCR analysis (Supplementary Fig. S6B).

Together, these data confirmed our previous results obtained by ChIP-seq analysis, revealing a major role for CKAP2L in the expression of genes involved in cell cycle and microtubule organization and highlighting the regulation of biological processes, such as chromosome segregation and spindle formation that are key aspects of CIN and tumorigenesis.

A hallmark for productive transcription elongation is represented by phosphorylation of RNA Pol II at serine 2 (Ser2) within its carboxyl terminal domain (32). We observed a decrease in the levels of Ser2 phosphorylation upon CKAP2L depletion while phosphorylation levels of serine 5, indicative of transcription initiation, were only marginally altered (Fig. 6A).

It is well established that the cyclin-dependent kinase CDK9, which along with cyclin T forms the positive transcription elongation factor b (P-TEFb), phosphorylates the RNA Pol II at Ser2 (33) to promote the transition into productive elongation. Therefore, we assessed the effect of CKAP2L downregulation on CDK9 protein levels and observed that they were not perturbated upon loss of CKAP2L depletion (Fig. 6B), indicating that regulation of RNA Pol II Ser2 phosphorylation by CKAP2L might be independent of P-TEFb recruitment. Accordingly, a synergistic effect was observed upon CKAP2L depletion in combination with CDK9 inhibition by alvocidib treatment, which resulted in a significant inhibition of cell proliferation (Fig. 6C) and increased cell death (Fig. 6D).

Ongoing changes to the genome of cancer cells, spanning from mutations to chromosome gain or loss, have been identified as hallmarks that generate genetic diversity and accounts for the molecular complexity and heterogeneity of the tumor.

It is now clear that genome instability, prevalently in the form of CIN, is inherent to the majority of human cancers and it is instrumental for tumor progression by favoring the accumulation of advantageous genotypes and, in turn, phenotypes. Therefore, targeting the molecular mechanisms leading to CIN may represent a valuable therapeutic opportunity for many cancer types, including NSCLC, for which there is still an urgent need of more effective approaches.

Here, we present evidence that the cytoskeleton-associated protein 2-like (CKAP2L), which we show as being overexpressed in human lung adenocarcinoma samples with high CIN index, exerts oncogenic functions and promotes lung tumor progression both in vitro and in vivo.

CKAP2L has been previously described to localize at microtubules of the spindle pole (34) similarly to Radmis (9), its mouse orthologous. Interestingly, the downregulation of Radmis or CKAP2L has been shown to result in abnormal monopolar spindle formation and defective chromosome segregation (9, 35). Likewise, Radmis overexpression was reported to induce mitotic spindle defects, suggesting that fine-tuning of its expression levels might be required for optimal spindle formation and cell division in noncancerous cells (9).

Importantly, our studies expand on the previously reported observations on the function of CKAP2L and add novel mechanistic insights. Indeed, we support a role for CKAP2L that goes beyond simply being a structural component of the spindle pole and demonstrate that it is instead a key orchestrator of cell division and microtubule–cytoskeleton dynamics as well of tumor progression.

This newly discovered function of CKAP2L occurs via direct regulation of transcription elongation through the interaction with the catalytic subunit of the RNA Pol II. Indeed, CKAP2L depletion led to a decreased transition of RNA Pol II into the elongation phase for genes associated with mitotic cell division, including DNA replication and repair, spindle organization, centrosome cycle, chromosome organization, and segregation, which are all documented key aspects of CIN and often dysregulated in cancer. In particular, multiple cyclin-dependent kinases as well as the transcription factors E2F1 and E2F2, and several of their target genes with well-established roles in cell-cycle control and mitotic progression, including AURKB, PLK1, and CDC20 (32, 36), were identified as being regulated by CKAP2L, and many of them are targets for the development of inhibitors in cancer therapy (37). Therefore, we suggest that CKAP2L, by directly binding to RNA Pol II and, indirectly through E2F signaling, modulates the expression of several genes implicated in the cell cycle to sustain and promote tumor progression.

We additionally present evidence that CKAP2L gene expression is regulated by RAS and MYC signaling, two key oncogenic factors in NSCLC that have been reported to promote CIN through induction of replication stress, mitotic errors, and spindle assembly defects (38, 39). Therefore, this raises the possibility that CKAP2L might mediate, at least in part, the role of RAS and MYC in inducing CIN and promoting tumorigenesis, although further studies will be required to experimentally validate our hypothesis.

Our data also indicate that CKAP2L affects phosphorylation levels of the RNA Pol II at Ser2 in the CTD domain, which is a key event for the transition into productive transcription elongation and known to be mediated by P-TEFb containing the cyclin-dependent kinase 9 (CDK9). However, our mass spectrometry analysis did not identify CDK9 as interacting with CKAP2L and, importantly, the silencing of CKAP2L did not affect CDK9 protein levels. Furthermore, inhibition of CDK9 with alvocidib in combination with CKAP2L downregulation led to markedly reduced levels of cell proliferation and an increase in apoptotic cell death in comparison with single treatments, indicating that the two proteins operate in parallel independent pathways. Taken together, these results indicate that it is plausible that CKAP2L-mediated regulation of RNA Pol II Ser2 phosphorylation does not involve CDK9 activity, implying that other potential mechanisms may occur. Indeed, CDK12 and CDK13 have also been reported to contribute to RNA Pol II CTD phosphorylation and affect transcription elongation rates (40, 41). Moreover, our mass spectrometry analysis revealed that CKAP2L interacts with the Negative elongation factor A (NELFA), which also plays a key role in regulating the elongation of transcription by RNA Pol II (42, 43). Despite the precise molecular events behind CKAP2L-mediated regulation of transcription elongation remain to be elucidated, our studies have highlighted the intriguing opportunity of exposing cancer to transcriptional vulnerabilities by combining CKAP2L and CDK9 inhibition. Interestingly, a recently described inhibitor of CDK9, AZD4573, was shown to rapidly induce apoptosis in hematologic cancers mainly via depletion of MCL-1 (44), which was also observed upon CKAP2L depletion and could therefore account for the synergistic effect of CDK9 and CKAP2L inhibition.

Overall, our results suggest that targeting CKAP2L could serve as a promising approach to tackle CIN-induced tumor heterogeneity, which is at the base of tumor progression and resistance to therapies, and successfully challenge transcriptional addiction in NSCLC.

M. Fassan reports grants from QED Therapeutics; grants and personal fees from Astellas Pharma; and personal fees from GlaxoSmithKline, Tesaro, Diaceutics, and Lilly outside the submitted work. No disclosures were reported by the other authors.

T. Monteverde: Conceptualization, data curation, formal analysis, validation, methodology, writing–original draft. S. Sahoo: Formal analysis. M. La Montagna: Validation. P. Magee: Validation, methodology. L. Shi: Investigation. D. Lee: Software, formal analysis. R. Sellers: Software, formal analysis. A.R. Baker: Software, formal analysis. H.S. Leong: Formal analysis. M. Fassan: Formal analysis, investigation, methodology. M. Garofalo: Conceptualization, supervision, funding acquisition, writing–review and editing.

The authors are grateful to members of the CRUK Biological Resources Unit for support with the in vivo studies and to H. Woodhouse from CRUK Visualisation, Irradiation & Analysis for assistance with high-content imaging. They also thank the Molecular Biology and Histology core facilities in the CRUK Manchester Institute for their help. Research samples were obtained from Manchester Cancer Research Centre (MCRC) Biobank, United Kingdom. The role of the MCRC Biobank is to distribute samples and, therefore, cannot endorse studies performed or the interpretation of results. Work in the M. Garofalo lab is funded by CRUK core grant (C5759/A20971) and Lung Cancer Centre (C5759/A20465).

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.