Sage Journals: Discover world-class research

Abstract

Malignant melanomas often arise from nevi, which result from initial oncogene-induced hyperproliferation of melanocytes that are maintained in a CDKN2A/p16-mediated senescent state. Thus, genes that can bypass this senescence barrier are likely to contribute to melanoma development. We have performed a gain-of-function screen of 17,030 lentivirally expressed human open reading frames (ORFs) in a melanoma cell line containing an inducible p16 construct to identify such genes. Genes known to bypass p16-induced senescence arrest, including the human papilloma virus 18 E7 gene (HPV18E7), and genes such as the p16-binding CDK6 with expected functions, as well as panel of novel genes, were identified, including high-mobility group box (HMGB) proteins. A number of these were further validated in two other models of p16-induced senescence. Tissue immunohistochemistry demonstrated higher levels of CDK6 in primary melanomas compared with normal skin and nevi. Reduction of CDK6 levels drove melanoma cells expressing functional p16 into senescence, demonstrating its contribution to bypass senescence.

Keywords

senescence p16 high-content imaging overexpression screening CDK6

Introduction

The CDKN2A/INK4A gene, which encodes the p16 protein, is one of the most commonly mutated tumor suppressors.¹ It is deleted, mutated, or epigenetically silenced in variety of tumor types, including melanomas, and germline CDKN2A mutations are observed in melanoma-prone families.¹ The p16 protein is thought to suppress tumorigenesis through its ability to induce cellular senescence. p16 binds to and inhibits the activity of the CDK4/6–cyclin D complex, which normally phosphorylates and inactivates the retinoblastoma tumor suppressor (RB) protein. Thus, inactivation of p16 or other components of the p16-CDK4-RB1 pathway allows E2F1-dependent transcription and cell cycle progression. Indeed, dysfunction of this pathway is a common feature in melanomas.¹

Senescence can be triggered by DNA damage, reduction in telomere length during aging, and expression of oncogenes that drive excessive mitogenic signaling.² Senescence also plays a role in tissue remodeling during active repair and early development. Senescent cells are commonly characterized by enlarged morphology, absence of the Ki-67 proliferation marker, and presence of senescence-associated β-galactosidase (SA-β-gal) activity.² The best described in vivo example of oncogene-induced senescence is the melanocytic nevus or mole.³ Melanocytic nevi are benign, clonal lesions of cutaneous melanocytes that most often harbor activating mutations in BRAF or NRAS mitogenic signaling proteins.¹ Increased p16 expression, as seen in nevi, is sufficient to drive senescence in melanoma cell lines with functional RB1.⁴

Between 20% and 50% melanomas arise from preexisting nevi.⁵ Thus, nevi represent a reservoir of premalignant cells, although the transformation rate is low. Early stage melanomas with a high proliferative activity commonly retain p16 expression, which is generally lost only in late-stage disease.¹ This suggests that there must be mechanisms for bypassing the p16-induced senescence arrest during early stages of melanoma development. Several genes have been identified that are capable of bypassing a senescence barrier. Activated Wnt and Hedgehog pathway signaling and overexpressed TBX2 and BMI1 all act by suppressing p16 expression.^6,7 Overexpression of ZBTB7A/LRF or MYC⁸ upregulates cyclin E expression and consequently cyclin E/CDK2 activity; this, in turn, promotes cell cycle progression by bypassing the p16-CDK4/6-RB arrest point. However, there is little evidence for ZBTB7A or PDK1 overexpression in primary melanoma, and while MYC gene amplification has been found in primary melanomas, MYC expression was elevated in only a small proportion of these.⁹ This suggests that there are as yet unidentified factors that can drive melanoma development past the p16-induced senescence seen in a nevus.

High-throughput functional genome-wide screens are a powerful, unbiased approach that can be used to elucidate novel potentially oncogenic mechanisms. Loss-of-function screens have been used to identify additional genes required for maintenance of cellular senescence.^7,10 However, potential oncogenic drivers of senescence bypass are more likely to be revealed using gain-of-function genome screening. Therefore, we chose this approach to identify genes that can trigger bypass of p16-induced senescence in a melanoma cell line. Here, we used a recently constructed lentiviral expression library comprising 17, 030 human open reading frames (ORFs)¹¹ to look for genes that can overcome senescence in a melanoma cell line containing an inducible p16 expression construct.⁴

Materials and Methods

Unless otherwise noted, all regents were obtained from Sigma-Aldrich (Sydney, Australia).

Cells and Culture Conditions

WMM1175 isopropyl β-D-1-thiogalactopyranoside (IPTG)–inducible p16 wild-type (WT) clone B17 cells⁴ were a kind gift from Dr. Therese Becker (Westmead Millennium Institute, Westmead, Australia) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% Serum Supreme calf serum (Lonza, Waverley, Australia), 4 mM L-glutamine, 1 mM sodium pyruvate, and 10 mM HEPES. To induce p16 expression, IPTG (4 mM) (Amresco, Solon, OH) was added to culture medium. Additional melanoma-derived cell lines were cultured as described previously.¹²

Viral Supernatant Library and Plasmids

Lentiviral ORF expression library consisting of viral supernatants arrayed in 96-well plates (n = 205) was obtained from and screened at the ARVEC facility at UQ Diamantina Institute. Lentiviral expression constructs were generated in pLV411G (synonym pLVEIG, accession KF486506.1), a Gateway destination vector that allows EF1α promoter-driven coexpression of an upstream ORF and a downstream green fluorescent protein (GFP), separated by an intervening IRES sequence. Control wells on each plate were four wells containing negative control supernatant derived from plv411 (empty expression plasmid, no ORF and no ccdB gene), two “mock” wells containing viral particles without the expression plasmids, and two positive control wells expressing mutant CDK4^R24C. The positive control expression constructs were derived from pcDNA3-HA-Cdk4^R24C (kind gift from Dr. William Hahn, Dana Farber Cancer Institute, Boston, MA). The CDK4^R24C fragment with a C-terminal FLAG tag was PCR amplified and cloned into pENTR/D-TOPO according to the manufacturer’s directions (Life Technologies, Mulgrave, Australia). The resultant pENTR-CDK4^R24C-FLAG entry clone was sequence verified and then recombined with pLV411G as described previously.¹¹

To generate additional plates for hit validation, expression clones were rearrayed from bacterial glycerol stocks, DNA isolated, and fresh virus generated by packaging in HEK293T cells as described previously.¹¹

High-Throughput Transductions and Plate Processing for Imaging

Cell seeding and medium and fixative bulk dispensing were performed using a WellMate microplate dispenser (Thermo Fisher Scientific, Hudson, NH). An ELx405 Microplate Washer (BioTek Instruments, Winooski, VT) was used for postfixation phosphate-buffered saline (PBS) washes. All other liquid handling steps were performed using the SciClone ALH3000 robotic workstations (Caliper Life Sciences, Hopkinton, MA).

WMM1175 p16 WT clone B17 cells were seeded into ViewPlate96-Black plates (PerkinElmer, Waltham, MA) at 1000 cells per well in 120 µL/well and incubated in the Liconic (Liechtenstein) robotic multiwell plate tissue culture incubator. The following day, media were removed. For this and all subsequent aspirate steps, aspirate tip height was adjusted so that there was a residual volume of 20 µL/well. Plates containing lentiviral supernatants were thawed and hexadimethrine bromide (polybrene) added at 12 µg/mL. This supernatant was then added to cells (30 µL/well), incubated for 2 h, and then topped up with 100 µL per well of medium. The next day medium was aspirated and replaced with 180 µL per well medium with or without 4.4 mM IPTG (Amresco). After an additional 5-day incubation, media were aspirated and replaced with 180 µL/well 3.7% formaldehyde in PBS. After a 15-min incubation at room temperature (RT), wells were washed by streaming 700 µL/well PBS. Unless they were processed immediately, fixed cells were stored at 4 °C in 75 µL/well PBS.

Subsequently cells were permeabilized in 0.05% (w/v) Triton-X-100 in PBS for 15 min, washed in PBS, and then incubated in blocking buffer (1.5% bovine serum albumin [Amresco] and 0.1% Saponin in PBS) for at least 1 h. Blocking buffer was aspirated (residual volume 20 µL) and primary antibodies added in 10 µL/well at 3× concentration in blocking buffer. Plates were incubated for minimum 1 h at RT or overnight at 4 °C. Wells were washed in PBS and then incubated in secondary fluorescently tagged antibodies as above. After another PBS wash, plates were stained with 300 nM 4′,6-diamidino-2-phenylindole (DAPI) in PBS for 1.5 h. Wells were washed and the cells imaged in 75 µL/well PBS.

High-Content Imaging

All antibodies were diluted in blocking buffer as indicated. For the primary screen and initial validation experiment, primary antibodies used were mouse monoclonal anti–Ki-67 (1:600, clone MIB-1; Dako, Campbellfield, Australia) and rabbit polyclonal anti–α-tubulin (1:1500, ab18251; Abcam, Cambridge, MA). Secondary probe-conjugated antibodies used in the screen were goat Cy3-anti-mouse-IgG (1:1500) and goat Cy5-anti-rabbit-IgG (1:1500) (Life Technologies). For subsequent validation experiments, additional antibodies used were as follows: rabbit anti-pRB Ser780 antibody (1:1000; Cell Signaling), rabbit polyclonal anti–Ki-67 (1:1000, ab15580; Abcam), and mouse monoclonal anti-p16 (1:800, clone G175-405; BD Biosciences, San Jose, CA) for primary and goat Cy3-anti-rabbit-IgG (1:1500; Life Technologies) and goat Alexa Fluor 647–anti-mouse-IgG (1:1500; Life Techno-logies) for secondary, as indicated in the figure legends. Plates were scanned using the ArrayScan VTI HCS Reader (Thermo Scientific, Rockford, IL) coupled to a Twister II Plate Handler (PerkinElmer, Waltham, MA). Fluorescent images were captured using a 10× objective and the XF2046 (400-485-558-640QBDR) quadband dichroic and excitation/emission filter set (Omega, Brattleboro, VT). Images were acquired and processed using the Cellomics CellHealthProfiling.v3 algorithm (Thermo Scientific, Rockford, IL, USA) to obtain numerical parameters in four filter channels: blue, green, red, and far red. For the primary screen and initial validation experiments, channels were assigned as follows: (1) blue, DAPI (for setting nuclear mask and collecting cell counts and nuclear DNA content); (2) green, GFP (for transduced cell selection); (3) red, Cy3 (for Ki-67 detection within the nuclear mask and proliferating cell selection); and (4) far red, Cy5 (for α-tubulin detection and setting the cytoplasmic area mask). For the subsequent validation experiments, channels 3 and/or 4 were reassigned to alternative antibodies and/or quantification area masks (e.g., nuclear, cytoplasmic), as indicated in the results.

Data Analysis and Hit Selection

Obtained object parameters were processed using the R system for statistical computation and graphics (http://www.r-project.org/). Data from all four imaging channels for each cell were linked to each object (i.e., nucleus identified by setting a threshold-based mask in the DAPI channel). The screen was performed in five batches of plates, which were stained and imaged separately so that there was considerable variation in signal intensity. To correct for batch-wise differences in signal intensity and thus apply common thresholds for assigning object GFP and Ki-67 status and compare relative DNA content across the whole primary screen, intensity values from these three channels were adjusted with a batch-specific correction factor such that the modes of the density histograms for each of these parameters from all five batches were aligned (Suppl. Fig. S4).

Each of the 205 screened plates contained four negative control wells, two positive control wells, and two mock-transduced wells (treated with supernatant with empty viral particles), one of which was not stained with primary antibodies. The two mock wells were used to set fluorescence intensity thresholds for assigning GFP and Ki-67 status, respectively, to individual nuclei. For quantitation of GFP expression and Ki-67 staining, average nuclear fluorescence intensities were used. For determination of relative DNA content and assigning cell cycle phase, total integrated intensities from staining with DAPI were used. The thresholds were batch-adjusted postscreening using density plots (Suppl. Fig. S3). Individual object data were summarized for each well by calculating the proportion of transduced (GFP positive), proliferating (Ki-67 positive) cells prior to calculation of plate median and median absolute deviation (MAD)–based z scores (Suppl. Fig. S3). To avoid bias due to small sample size, only wells with a minimum of 100 GFP-positive cells were used for plate median calculation. Data from hit validation plates where replicates were available on independent plates were compared using analysis of variance (ANOVA) and post hoc Tukey’s honestly significant difference (HSD) test.

Details of the other methods used are provided in the Supplementary Material.

Results

Immunohistochemical staining of a panel of 57 nevi and 34 primary melanomas showed that 51 of 57 nevi and 29 of 34 melanomas were positive for p16 staining and that overall staining for p16 was very similar in both groups (Suppl. Fig. S1). The average proportion of Ki-67–positive cells in nevi was 1%, compared with 9.2% in melanomas, with the latter showing a broad spread of Ki-67 positivity from 2% to 50%. These data demonstrate that although p16 continues to be expressed in primary melanoma, there must be genes that promote bypass of p16-induced senescence to a proliferative state in primary melanomas, indicated by the increase Ki-67 positivity, without affecting p16 expression.

To identify such genes, an overexpression screening approach was adopted, using a clonal isolate of the melanoma-derived WMM1175 cell line (which has homozygous deletion of endogenous CDKN2A) containing an IPTG-inducible wild-type p16 construct. These cells have functional RB and undergo senescence upon IPTG-induced expression of p16. This is independent of p53 and does not involve the DNA damage response often associated with senescence.⁴ The induction of p16 was verified by immunoblotting (Suppl. Fig. S2a) and resulted in a senescent phenotype confirmed by increased SA-β-gal activity and HP1β staining foci, both markers of senescence⁴ ( Fig. 1a , c ). This was accompanied by reduced proliferative potential, demonstrated by a decreased percentage of Ki-67–stained cells ( Fig. 1b ). The reduction in proliferative potential was further confirmed by the reduced capacity for colony formation after IPTG exposure (Suppl. Fig. S2b). Expression of p16 also increased cytoplasmic area measured using α-tubulin immunofluorescence ( Fig. 1b and Suppl. Fig. S2c). Based on these phenotypes, we developed the screening protocol outlined in Figure 2a . Briefly, cells were seeded and robotically transduced using a lentiviral library comprising 17,033 arrayed clones of annotated human ORFs that are coexpressed with GFP, under the control of the human EF1α promoter.¹¹ Transduced cells were treated with IPTG to induce p16 one day posttransduction and fixed and processed for imaging 5 days later. Four-channel high-content imaging assays ( Fig. 2b ) were then performed to identify potential senescence bypass genes. Nuclear staining with DAPI was used to identify individual cells and assign cell cycle phase, while GFP fluorescence was used to identify transduced cells. Ki-67 and α-tubulin immunofluorescence were used to assess the senescent phenotype, with loss of nuclear Ki-67 and increased cell area defined by α-tubulin staining. Empty pLV411 vector provided a negative control, while the positive control was a vector expressing CDK4^R24C, a mutated form of CDK4 insensitive to inhibition by p16.¹ The decrease in the proportion of Ki-67–positive cells and increase in cytoplasmic area did not occur in wells treated with the positive control virus ( Fig. 2a , CDK4^R24C +IPTG). In wells transduced with vector only, there was an 80% reduction in Ki-67–positive cells and up to an eightfold increase in cytoplasmic area in the p16-induced cells compared with uninduced wells. Good separation between positive and negative assay controls was confirmed by calculating the Z′ factor for each parameter; this yielded values of 0.702 and 0.869 for percentage of Ki-67–positive cells and cytoplasmic area, respectively (Suppl. Fig. S2d).

Figure 1.

(a) Proportion of SA-β-gal–positive cells before and after isopropyl β-D-1-thiogalactopyranoside (IPTG) addition. The bar graph shows the mean and standard deviation from three biological replicates with 200 cells counted per replicate. (b) WM1175 with and without 5-day IPTG addition. Cells were stained for DNA (4′,6-diamidino-2-phenylindole [DAPI]), Ki-67, and p16 to show the induction of p16 expression. The bar graph shows the percentage of Ki-67–positive cells from four replicate wells. (c) Cells were stained for DNA (DAPI), the senescence marker HP1β, and α-tubulin to define cell size. The bar graphs represents cell size measured by pixel size of the α-tubulin staining area for each cells. (Microscope field width = 660.5 µm in all images.)

Figure 2.

(a) Diagram illustrating primary screen workflow: WMM1175 cells containing inducible p16 were seeded and transduced with arrayed lentiviral supernatant. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added to induce p16 expression. After a 5-day incubation, plates were fixed; stained with 4′,6-diamidino-2-phenylindole (DAPI), anti–Ki-67, and anti–α-tubulin antibodies; and then imaged. (b) Pseudocolored overlay and corresponding inset images illustrating the four fluorescence channels scored as follows: (1) DAPI quantifying of DNA content and nuclear mask setting for cell counts, (2) green fluorescent protein (GFP)–transduced cell selection, (3) anti–Ki-67 antibody proliferating cells detection, and (4) anti–α-tubulin antibody cytoplasmic mask setting around the nucleus for cell area measurement. (c) Assay controls illustrating reduction of cell number (i.e., nuclei, DNA) and proportion of Ki-67–positive nuclei and increase in cell size (α-tubulin) in IPTG-treated cells transduced with vector (screen-negative control) compared with untreated vector controls and lack of effect in CDK4^R24C-transduced cells (screen-positive control). (Microscope field width = 660.5 µm in all images except enlarged incepts in b.)

Data for valid wells (defined as having >100 GFP-positive cells) from the 202 plates that passed technical quality control are presented in Supplemental Table S1. To overcome batch-wise differences in signal intensity and apply common thresholds across the whole primary screen, data from individual channels were processed and robust z scores calculated as described in the Supplementary Material and Supplemental Figure S3. A scatterplot of Ki-67 score versus cytoplasmic area for the entire screen is presented in Figure 3a . The CDK4^R24C controls showed high proliferation and low cytoplasmic area scores, clustering away from the sample, and negative control wells. As expected, most of these had low proliferation and high area scores, so we selected 145 candidate hits that clustered with the CDK4^R24C controls for further screening. These had Ki-67 z scores >2.5 and/or more than 1000 cells, cytoplasmic area <2500 pixels, and a combined proportion of nuclei in S, G2, and M (S/G2/M) phases of cell cycle above 0.25. Given that this was a single-pass screen, the primary selection criteria were inclusive rather than stringent.

Figure 3.

Scatterplot of variables used for hit selection; Ki-67 robust z score (y-axis), cytoplasm area (pixels, x-axis), and combined proportion of nuclei in S, G2, and M phases of the cell cycle (point size), representing data for all wells analyzed in the primary screen. Sample wells (blue) were considered putative hits if they grouped with CDK4^R24C (red) and away from vector (yellow) wells.

To confirm these candidate hits, we performed a secondary screen in triplicate. Sequence-verified hit clones were rearrayed from bacterial glycerol stocks and fresh virus generated. Validated hits were selected by comparison of Ki-67 scores (Suppl. Table S2). IPTG induction of p16 was confirmed by immunofluorescence (Suppl. Fig. S4), indicating that the hits were not interfering with p16 expression. Mean Ki-67 values for sample wells were compared with control wells using ANOVA and Tukey’s HSD test. The identities of 54 validated hits (significantly different from the vector control, p < 0.05) representing 50 unique genes are presented in Table 1 , along with summary statistics for observed Ki-67 values. They include 47 human ORFs, two pseudogenes (HMGB1P1, MGC42157), and one human papillomavirus protein (HPV18E7). The latter represented an internal validation of the screen since E7 expression inhibits RB1 function and thereby bypasses p16 inhibition of CDK4/6. Among the confirmed hits, there were six cell cycle–associated genes (CCNE1, CDK6, HSPA2, E4F1, RBM38, OSM), eight zinc-binding transcription factors (MICALL2, EBF3, PHF23, KLF15, E4F1, CXXC5, MMP13, ZFP161), six RNA-processing genes (RBM3, TRA2A, CIRBP, RBM38, HNRNPC, HNRNPR), and strikingly seven clones represented by four genes (HMGA1, HMGB1, HMGB2, HMGB3) and one pseudogene (HMGB1P1) belonging to the high-mobility group (HMG) family. In addition, one of the strongest hits from the screen, HMGB2, partially decreased the formation of DAPI-stained DNA foci, which are regarded as alternative markers of senescence (Suppl. Fig. S5).

Table 1.

Confirmed Hits Grouped by Functional Category.

Group	Gene (Clones)	Difference^a (%Ki-67+)	p (Adjusted)
HMG group	HMGB2 (2)	63	8.77E-15
	HMGB1 (2)	56	8.77E-15
	HMGB3	33	9.88E-15
	HMGA1	29	2.1E-13
	HMGB1P1	23	1.1E-09
Cell cycle	CDK4_R24C	65	8.77E-15
	CDK6	43	8.77E-15
	CCNE1	31	6.99E-13
	E4F1	22	4.37E-12
	HSPA2	22	1.15E-11
	RBM3	17	0.000117
	OSM	19	0.007408
RNA binding	CIRBP	45	8.77E-15
	HNRNPC (3)	34	8.77E-15
	TRA2A (2)	33	1.03E-14
	HNRNPR	21	4.24E-11
	RBM38	17	0.000109
Membrane	TMEM45A	37	8.77E-15
	C20orf141	30	7.76E-14
	TMEM45B	27	4.58E-13
	TSPAN3	22	5.59E-12
	GPR137C	23	3.69E-09
	VKORC1	17	5.26E-07
	PEMT	19	3.89E-06
	TMEM168	16	6.44E-06
	AP2A2	14	0.000387
	PAQR3	13	0.002162
	TMEM185A	15	0.004738
	ABCA8	12	0.009884
Transcription	PHF23	33	9.44E-15
	MICALL2	24	8.25E-13
	ZFP161	25	7.77E-11
	NFIC	16	2.51E-06
	CXXC5	19	6.79E-06
	KLF15	17	0.000105
	MMP13	14	0.000522
	EBF3	16	0.000554
Other	HPV18E7	40	8.18E-14
	ATOH1	26	7.06E-13
	HSP90AB1	18	4.32E-08
	CXCL11	17	1.09E-06
	IFI16	15	0.004871
	GPX4	12	0.006374
	KIAA0141	14	0.007197

HMG, high-mobility group.

Results of analysis of variance and Tukey’s honestly significantly difference test of high-content image data presented as the mean difference compared with values observed for vector-treated samples in Ki-67 proportion and associated adjusted p values.

Hit genes were also assessed in two other p16-inducible models of senescence derived from breast cancer–derived MDA-MB-231 cells and melanoma-derived SKMel13,¹² respectively. In each of these cases, overexpression of the control genes CDK4^R24C and our hit HPV E7 drove a strong bypass of the p16-induced proliferative arrest ( Fig. 4 ). Cyclin E overexpression also strongly drove senescence bypass in the SKMel13 model ( Fig. 4d ), as in the WMM1175 model. Two genes (PHF23, RBM38) consistently increased proliferation in both the MDA-MB-231 and the WMM1175 models, albeit to a lesser degree than the positive controls ( Fig. 4a , d ). CXXC5 and ZPF161 each produced a weaker bypass in one of the model lines, while CDK6 just failed to reach significance in SKMel13. HNRNPC, HMGA1, and HMGB1 had no obvious activity in these models. A subset of genes from the initial hit list, KIAA0141, KLF15, and GPR137C, and the strongest hit from the screen, HMGB2, failed to increase proliferation but significantly (p < 0.001) reversed the increased cell size associated with p16-induced senescence in MDA-MB-231 cells ( Fig. 4c ). KLF15 affected cell size in both uninduced and p16-induced conditions. Thus, at least six hits were validated in at least one other model of p16-induced senescence, and a further four significantly reduced the increased cell size that is a hallmark of senescence. The robust effect of HMGB proteins and CDK6 in bypassing p16-induced senescence arrest in the WMM1175 but weaker effects in the other models suggests that cell line–intrinsic features, such as the relative levels of induced p16 expression and HMGB and CDK6 expression, can influence the outcome.

Figure 4.

MDA-MB-231 cells with inducible p16 were transduced with the indicated lentivirus and induced to express p16 by addition of doxycycline the next day. After 5 days, cells were fixed and assessed for Ki-67 (b), with α-tubulin costaining to assess cytoplasmic area (c) using high-content imaging. The data are from four wells and represent the mean and SEM. *p < 0.05, **p < 0.005 difference from the empty vector control. (a) The combined data are shown. (d) SKMel13 cells with inducible p16 were transduced with the indicated lentivirus and induced to express p16 by addition of isopropyl β-D-1-thiogalactopyranoside (IPTG) the next day. After 5 days, cells were fixed and assessed for Ki-67 using high-content imaging. The data are from three wells and represent the mean and SEM.

The primary target of p16 inhibition of CDK4/cyclin D that results in G1 phase arrest and senescence is RB, particularly the phosphorylation of RB on Ser780 (pRB), a CDK4/cyclin D phosphorylation site. We examined the effect of selected hits on RB phosphorylation. Three genes, HMGB2, ZFP161, and KLF15, increased Ki-67 staining and pRB levels, suggesting that they may also operate upstream of CDK4/cyclin D ( Fig. 5a ). Cells expressing HPVE7, which induces RB degradation,¹³ had reduced levels of pRB, as expected ( Fig. 5a ). None of the other hit genes analyzed had similarly reduced level of pRB, indicating that degradation of RB was not a mechanism of senescence bypass for these genes.

Figure 5.

(a) WMM1175 cells transduced with the indicated genes at 5 days after p16 induction. (R24C = CDK4^R24C). The horizontal line in each histogram defines the cutoff for significance over vector-only (pLV) control. Wells without p16 induction are indicated (–I). Cells transduced with indicated hit genes were induced to express p16 for 5 days, then fixed and costained for pRB Ser780 and Ki-67. Analysis of the proportion of cells in each transduction stained for pRB Ser780 and Ki-67. The data are mean and SD of minimum four wells per gene. –I, no IPTG controls. (b) Immunohistochemical assessment of CDK6 levels in tissue microarrays of benign nevi and melanoma. Signal intensity and proportions of cells stained were assigned a consensus H-score. (c) Melanoma cell lines D22, D28, and MM383 were transduced with empty vector or CDK4^R24C expressing lentivirus, then transfected with lipofection reagent alone (untransfected), with nontargeting or CDK6 small interfering RNA. Cells were fixed and stained for Ki-67 three days after transfection. The bars represent the mean and SEM from four wells. The level of CDK6 expression and p16 status of each cell line is indicated (Wt, wild type; Meth, hyper-methylated; Mut, mutant). *p > 0.05.

CDK6 was one of the top hits in the screen. The level of CDK6 protein was assessed on the set of lesions as in Figure 1 . A subset of melanomas with strongly increased CDK6 levels compared with normal skin or nevi ( Fig. 5b ). Immunoblot analysis of cell extracts from a panel of melanoma cell lines revealed MM383, SKMel28, and D22 had elevated levels of CDK6 (Suppl. Fig. S6). Since CDK6 is overexpressed in a subset of melanomas, altered expression of CDK6 and HMGB2 may contribute to senescence bypass in human melanoma. To assess this, we performed small interfering RNA (siRNA) depletion of CDK6 in D22, D28, and MM383 melanoma cell lines, which differ in levels of endogenous CDK6 and p16 (achieving >85% transfection in each cell line and 60%–85% CDK6 depletion; Suppl. Fig. S7b,c). CDK6 depletion in D22 resulted in a dramatic loss of Ki-67 staining and an equivalent reduction in EdU-positive cells, indicating a loss of proliferation associated with a 40% decline in cell numbers ( Fig. 5c and Suppl. Fig. S7a,b). The reduction in Ki-67 staining and cell number was effectively rescued by expression of the CDK4^R24C mutant, suggesting that the effect observed was dependent on endogenous p16 expression. The lack of effect in the MM383 cell line, which also has elevated CDK6 levels, could be explained by the fact that it expresses a mutant p16. CDK6 depletion had no effect on the D28 cell line, which has normal low levels of endogenous CDK6 ( Fig. 5c ).

Discussion

Here we have reported a genome-wide overexpression screen to identify novel drivers of bypass of p16-induced senescence. The identification of HPV18 E7 validated the screen, while the ability of cyclin E overexpression to overcome the p16-induced arrest in multiple models is further evidence that the screen was capable of identifying potential bypass genes. Cyclin E was identified as an essential target of both ZBTB7A and MYC to overcome oncogene-induced senescence arrest.⁸

Among the novel hits, the incidence of HMG family members was striking and is discussed further below. Other strong hit genes have previously been reported to either interfere with senescence or at least be involved in regulation of a G1 phase cell cycle arrest. Cold-inducible RNA-binding protein (CIRBP) can bypass replicative senescence without affecting p16 expression and is upregulated in a range of cancers but not in melanoma.¹⁴ TMEM45A has been reported to be overexpressed in ovarian cancers, and its depletion resulted in G1 phase accumulation.¹⁵ HSPA2, a member of HSP70 family, is overexpressed in breast cancers although rarely in melanoma, and its depletion drives a senescence-like phenotype.¹⁶ ZNF161 (ZBTB14) is related to ZBTB7A and may act in a similar manner to bypass senescence arrest by upregulating cyclin E expression.⁸ Oncostatin M (OSM) has a range of effects, dependent on the cellular context. It can drive a senescence phenotype by upregulating STAT3 activity in the context of a normal cell, but in MYC-overexpressing cells, OSM promotes a transformed phenotype.¹⁷ TRA2A is a nuclear RNA binding protein phosphorylated by CDK4/6, suggesting a role in regulating G1 phase progression.¹⁸ In addition to these genes, we have identified four novel genes with modest senescence bypass activity in all three models assessed here. PHF23 contains a plant homeodomain (PHD) motif that binds H3K4me3 residues and is present in an oncogenic fusion with NPU98 in leukemia.¹⁹ RBM38 is an RNA binding protein that regulates the translation of p53.²⁰ CXXC4 and CXXC5 are inhibitors of WNT signaling through direct binding to the Disheveled protein, which in turn drives degradation of β-catenin.²¹ Little is known about how these genes contribute to senescence bypass during transformation from nevus to melanoma.

HMGA and HMGB gene family members are frequently upregulated in cancer, which often correlates with poor prognosis.²² HMGA proteins have been demonstrated to have an essential role in Ras-induced senescence, with a major role in establishing senescence-associated heterochromatin foci.²³ HMGA1 has been reported to bind RB directly and inhibit binding of HDAC1 to RB, thereby enhancing E2F activity.²⁴ Surprisingly, despite the reported direct effect of HMGA1 binding to RB on its activity, HMGA1 was a weak bypass gene in the WMM175 model and failed to bypass senescence in either of the other models tested. The remarkable finding that the HMG proteins B1, B2, and B3 were all strong hits in our screen suggests that p16-induced senescence bypass is an activity shared by this family of proteins, at least when overexpressed. HMGB1 is secreted from senescent cells and regulates the senescence-associated secretory phenotype.²⁵ Evidence from the literature provides support for potential roles for HMGB proteins in senescence. HMGB2 has been linked to senescence through altering the β-catenin pathway activity by enhancing the β-catenin/Lef transcriptional program.²⁶ HMGB1 and HMGB2 also modify p53- and p73-dependent transcriptional activity.²⁷ Secreted HMGB1 acts as a cytokine, signaling through RAGE and TLR4 receptors to activate MAPK and NF-κB pathways.²⁸ However we found that neither culture supernatant from HMGB2-overexpressing WMM1175 cells nor purified HMGB1 were capable of bypassing p16-induced senescence in WMM1175 cells (unpublished observations). This argues that HMGB proteins are unlikely to act as cytokines to bypass senescence.

HMGB2, along with two other hits from our study, HNRNPC and HSP90AB1, were all strongly downregulated in cells driven into senescence.²⁹ In addition, CXXC4 and CXXC5 are inhibitors of WNT signaling, which act by directly binding to the Disheveled protein and driving degradation of β-catenin.²¹ WNT signaling is a common feature of nevi,³⁰ and thus inhibition of WNT signaling could contribute to overcoming senescence. Thus, it is possible that multiple hits from this screen use similar signaling pathways to bypass senescence.

We have functionally validated one of the hits from the screen, CDK6. Although CDK6 was highly expressed in a subset of melanomas, CDK6 and wild-type p16 are very rarely coexpressed at a high level. This was detected in only one melanoma cell line examined, D22. In MM383, CDK6 is overexpressed, but CDKN2A is methylated and not expressed. Thus, other mechanisms may supplant the requirement for high CDK6 as tumors progress. The finding that CDK6, but not CDK4, overexpression effectively bypassed senescence may reflect the relative level of CDK4 and CDK6 expression achieved in the screen, as previous studies have reported that CDK4 also exhibited this activity.³¹ Compared with CDK6, CDK4 is more commonly upregulated in melanoma, although both occur. CDK6 overexpression drove weak bypass of p16-induced senescence arrest in MB-MDA-231 cells and failed in the SK-Mel-13 cell model, possibly reflecting the level of CDK6 overexpression and p16 induction in these models. Nevertheless, the fact that depletion of CDK6 in an overexpressing melanoma cell line restored p16-induced arrest demonstrated that CDK6 overexpression contributes to senescence bypass in a melanoma context.

In summary, we have described here a high-throughput screening approach using 17,030 human open reading frames for identifying genuine genes that can bypass p16-induced senescence in the WMM1175 cell line when overexpressed. The relevance of this screening approach was supported by the detection of HPV18E7, which is known to abolish p16-induced senescence. In addition, CDK6, which is known to be involved in the retinoblastoma pathway, was identified. Knockdown of CDK6 in a p16-wild-type and CDK6-overexpressing melanoma cell line reinstated proliferative arrest. Our screening approach also identified unexpected genes, notably the HMGB family of proteins, as potential p16-induced senescence bypass genes. These results suggest that the ability of overexpressed HMGB proteins to bypass p16-induced senescence is a class effect of this protein family.

Footnotes

Acknowledgements

We thank Dr. Therese Becker and Prof. Helen Rizos for the WMM1175 cell line, Dr. Alexey Bazarov for the MDA-MB-231-TetR-CMV/TO-p16 cell line, Prof. H. Peter Soyer for advice on histology, and Prof. Nick Hayward and Prof. Dorothy Bennett for discussions on this work.

Supplementary material is available online with this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This work was supported by grants from the Worldwide Cancer Research (formerly AICR), the University of Queensland, and NHMRC Australia. W.J.L. was an NHMRC C. J. Martin Fellow, and B.G. was an NHMRC Senior Research Fellow.

References

de Snoo

F. A.

Hayward

N. K.

Cutaneous Melanoma Susceptibility and Progression Genes. Cancer Lett. 2005, 230, 153–186.

Munoz-Espin

Serrano

Cellular Senescence: From Physiology to Pathology. Nat. Rev. Mol. Cell Biol. 2014, 15, 482–496.

Gray-Schopfer

V. C.

Cheong

S. C.

Chong

. Cellular Senescence in Naevi and Immortalisation in Melanoma: A Role for p16? Br. J. Cancer 2006, 95, 496–505.

Haferkamp

Becker

T. M.

Scurr

L. L.

. p16INK4a-Induced Senescence Is Disabled by Melanoma-Associated Mutations. Aging Cell 2008, 7, 733–745.

Goldstein

A. M.

Tucker

M. A.

Dysplastic Nevi and Melanoma. Cancer Epidemiol. Biomarkers Prev. 2013, 22, 528–532.

Delmas

Beermann

Martinozzi

. Beta-Catenin Induces Immortalization of Melanocytes by Suppressing p16INK4a Expression and Cooperates with N-Ras in Melanoma Development. Genes Dev. 2007, 21, 2923–2935.

Jacobs

J. J.

Keblusek

Robanus-Maandag

. Senescence Bypass Screen Identifies TBX2, Which Represses Cdkn2a (p19(ARF)) and Is Amplified in a Subset of Human Breast Cancers. Nat. Genet. 2000, 26, 291–299.

Vredeveld

L. C.

Rowland

B. D.

Douma

. Functional Identification of LRF as an Oncogene That Bypasses RASV12-Induced Senescence via Upregulation of CYCLIN E. Carcinogenesis 2010, 31, 201–207.

Kraehn

G. M.

Utikal

Udart

. Extra c-myc Oncogene Copies in High Risk Cutaneous Malignant Melanoma and Melanoma Metastases. Br. J. Cancer 2001, 84, 72–79.

10.

Kaplon

Homig-Holzel

Gao

. Near-Genomewide RNAi Screening for Regulators of BRAF(V600E)-Induced Senescence Identifies RASEF, a Gene Epigenetically Silenced in Melanoma. Pigment Cell Melanoma Res. 2014, 27, 640–652.

11.

Skalamera

Dahmer

Purdon

A. S.

. Generation of a Genome Scale Lentiviral Vector Library for EF1alpha Promoter-Driven Expression of Human ORFs and Identification of Human Genes Affecting Viral Titer. PLoS One 2012, 7, e51733.

12.

Burgess

Ruefli

Beamish

. Histone Deacetylase Inhibitors Specifically Kill Nonproliferating Tumour Cells. Oncogene 2004, 23, 6693–6701.

13.

Munger

Basile

J. R.

Duensing

. Biological Activities and Molecular Targets of the Human Papillomavirus E7 Oncoprotein. Oncogene 2001, 20, 7888–7898.

14.

Artero-Castro

Callejas

F. B.

Castellvi

. Cold-Inducible RNA-Binding Protein Bypasses Replicative Senescence in Primary Cells through Extracellular Signal-Regulated Kinase 1 and 2 Activation. Mol. Cell Biol. 2009, 29, 1855–1868.

15.

Guo

Chen

Luo

. Inhibition of TMEM45A Suppresses Proliferation, Induces Cell Cycle Arrest and Reduces Cell Invasion in Human Ovarian Cancer Cells. Oncol. Rep. 2015, 33, 3124–3130.

16.

Rohde

Daugaard

Jensen

M. H.

. Members of the Heat-Shock Protein 70 Family Promote Cancer Cell Growth by Distinct Mechanisms. Genes Dev. 2005, 19, 570–582.

17.

Jin

Wang

C. L.

. PIM-1 Modulates Cellular Senescence and Links IL-6 Signaling to Heterochromatin Formation. Aging Cell 2014, 13, 879–889.

18.

Anders

Hydbring

. A Systematic Screen for CDK4/6 Substrates Links FOXM1 Phosphorylation to Senescence Suppression in Cancer Cells. Cancer Cell 2011, 20, 620–634.

19.

Gough

S. M.

Lee

Yang

. NUP98-PHF23 Is a Chromatin-Modifying Oncoprotein That Causes a Wide Array of Leukemias Sensitive to Inhibition of PHD Histone Reader Function. Cancer Discov. 2014, 4, 564–577.

20.

Avery-Kiejda

K. A.

Bowden

N. A.

Croft

A. J.

. P53 in Human Melanoma Fails to Regulate Target Genes Associated with Apoptosis and the Cell Cycle and May Contribute to Proliferation. BMC Cancer 2011, 11, 203.

21.

Sun

Wang

. Enhancer of Zeste Homolog 2 Activates wnt Signaling through Downregulating CXXC Finger Protein 4. Cell Death Dis. 2013, 4, e776.

22.

Takeda

Izumi

Kitada

. The Combination of a Nuclear HMGB1-Positive and HMGB2-Negative Expression Is Potentially Associated with a Shortened Survival in Patients with Pancreatic Ductal Adenocarcinoma. Tumour Biol. 2014, 35, 10555–10569.

23.

Narita

Krizhanovsky

. A Novel Role for High-Mobility Group a Proteins in Cellular Senescence and Heterochromatin Formation. Cell 2006, 126, 503–514.

24.

Esposito

Pierantoni

G. M.

Battista

. Interaction between HMGA1 and Retinoblastoma Protein Is Required for Adipocyte Differentiation. J. Biol. Chem. 2009, 284, 25993–26004.

25.

Davalos

A. R.

Kawahara

Malhotra

G. K.

. p53-Dependent Release of Alarmin HMGB1 Is a Central Mediator of Senescent Phenotypes. J. Cell Biol. 2013, 201, 613–629.

26.

Taniguchi

Carames

Kawakami

. Chromatin Protein HMGB2 Regulates Articular Cartilage Surface Maintenance via Beta-Catenin Pathway. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 16817–16822.

27.

Stros

Ozaki

Bacikova

. HMGB1 and HMGB2 Cell-Specifically Down-Regulate the p53- and p73-Dependent Sequence-Specific Transactivation from the Human Bax Gene Promoter. J. Biol. Chem. 2002, 277, 7157–7164.

28.

Kang

Zhang

Zeh

H. J.

III . HMGB1 in Cancer: Good, Bad, or Both? Clin. Cancer Res. 2013, 19, 4046–4057.

29.

Biniossek

M. L.

Lechel

Rudolph

K. L.

. Quantitative Proteomic Profiling of Tumor Cell Response to Telomere Dysfunction Using Isotope-Coded Protein Labeling (ICPL) Reveals Interaction Network of Candidate Senescence Markers. J. Proteomics 2013, 91, 515–535.

30.

Pawlikowski

J. S.

McBryan

van Tuyn

. Wnt Signaling Potentiates Nevogenesis. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 16009–16014.

31.

Ruas

Gregory

Jones

. CDK4 and CDK6 Delay Senescence by Kinase-Dependent and p16INK4a-Independent Mechanisms. Mol. Cell Biol. 2007, 27, 4273–4282.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.82 MB

2.61 MB

0.03 MB