Co-regulation of senescence-associated genes by oncogenic homeobox proteins and polycomb repressive complexes

Abstract

Cellular senescence is a stable cell cycle arrest that can be induced by stresses such as telomere shortening, oncogene activation or DNA damage. Senescence is a potent anticancer barrier that needs to be circumvented during tumorigenesis. The cell cycle regulator p16^INK4a is a key effector upregulated during senescence. Polycomb repressive complexes (PRCs) play a crucial role in silencing the INK4/ARF locus, which encodes for p16^INK4a, but the mechanisms by which PRCs are recruited to this locus as well as to other targets remain poorly understood. Recently we discovered the ability of the homeobox proteins HLX1 (H2.0-like homeobox 1) and HOXA9 (Homeobox A9) to bypass senescence. We showed that HLX1 and HOXA9 recruit PRCs to repress INK4a, which constitutes a key mechanism explaining their effects on senescence. Here we provide evidence for the regulation of additional senescence-associated PRC target genes by HLX1 and HOXA9. As both HLX1 and HOXA9 are oncogenes implicated in leukemogenesis, we discuss the implications that the collaboration between Homeobox proteins and PRCs has for senescence and cancer.

Introduction

Cellular senescence was originally described as a state of stable cell cycle arrest that accompanies the exhaustion of proliferative potential in cultured primary mammalian cells. Whereas this so-called replicative senescence is triggered by telomere erosion, other stimuli such as activated oncogenes, oxidative stress, DNA damage and enforced expression of pluripotency-associated factors can trigger premature senescence. Senescent cells remain metabolically active but undergo a cell cycle arrest in G₁ and are characterized by a flat and enlarged morphology, senescence-associated β-galactosidase activity (due to increased lysosome numbers), and the presence of senescence-associated heterochromatin foci (SAHF).¹ Senescent cells also secrete a plethora of extracellular proteins and soluble factors, referred to as the senescence-associated secretory phenotype (SASP), which have multiple and potentially opposing functions. This secretome includes extracellular proteases, matrix components (such as MMPs), growth factors, proinflammatory cytokines (IL6, IL1α, IL1β), and chemokines (IL8, GROα, MCP1).² Cellular senescence is relevant in multiple physiological and pathological contexts. Importantly, oncogene-induced senescence (OIS) acts as a potent cell-intrinsic tumor suppressor mechanism inhibiting tumor progression. On the other hand, cellular senescence also limits the self-renewal potential of stem cells and has been implicated in age-related disorders.¹ A related observation is that senescence impairs the efficiency of reprogramming of somatic cells to induced pluripotent stem cells (iPSCs).^3,4

Cellular senescence is a genetically driven program implemented by the activation of the p53 and RB tumor suppressor pathways. The INK4/ARF locus (comprising the CDKN2A and CDKN2B genes) has the potential to regulate both pathways, as it encodes for p15^INK4b and p16^INK4a, two cyclin-dependent kinase inhibitors controlling RB phosphorylation, and the unrelated protein ARF that activates p53.⁵ In primary cells, gene expression of the INK4/ARF locus is kept under control by Polycomb repressive complexes 1 (PRC1) and 2 (PRC2).^6,7 PRC2 interacts with histone deacetylases (HDAC)⁸ and catalyzes histone H3 lysine 27 trimethylation (H3K27me3). This epigenetic mark is recognized by PRC1,⁹ which catalyzes histone H2A lysine 119 monoubiquitination.^10,11 Upregulation of many PRC target genes is observed during senescence, while aberrant silencing of PRC target genes is frequent in tumorigenesis.¹² However, the question of how PRCs are recruited to their target genes is still a matter of investigation.^9,13 Recent data have provided evidence for the role of specific transcription factors^14,15 and long non-coding RNAs¹⁶ in PRC recruitment.

We recently identified HLX1 (H2.0-like homeobox 1) as a suppressor of cellular senescence in a genetic screen for transcription factors increasing cellular lifespan of human fibroblasts.¹⁷ HLX1 is part of the Homeobox family of transcription factors, which have important roles in developmental patterning.¹⁸ We observed that HLX1 overexpression both extends replicative lifespan and blunts premature senescence induced by oncogenic RAS. Conversely, HLX1 knockdown induces premature senescence. In order to understand the molecular mechanisms underlying HLX1 function, we used proteomics to monitor protein expression in response to HLX1 knockdown. Stable isotope labeling with amino acids in cell culture (SILAC) and subsequent functional experiments identified p16^INK4a as the key target mediating HLX1 effects. We observed that HLX1 directly associates with the INK4a promoter and represses INK4a expression by recruiting HDAC1 and PRC2. An siRNA screen identified that 6 other Homeobox proteins (DLX3, HOXA9, HOXB13, HOXC13, HOXD3 and HOXD8) are also able to repress p16^INK4a. We showed that expression of HOXA9 (homeobox A9), used as an example among these factors, is sufficient to delay senescence by recruiting PRCs and HDACs to repress INK4a. Altogether our work added Homeobox proteins to the list of factors (which includes TWIST1,¹⁵ Zfp277,¹⁴ and the lincRNA ANRIL^19,20) recruiting PRCs to repress INK4a. In addition, gene set enrichment analysis (GSEA) suggested that HLX1 not only regulated INK4a but also a broader subset of PRC targets.¹⁷ Here, we identify senescence-associated PRC targets other than INK4a that are also regulated by HLX1 and HOXA9, and we discuss the broad implications that the interplay between Homeobox proteins and PRCs could have in senescence and cancer.

Results and Discussion

We recently demonstrated that p16^INK4a is a key mediator explaining the effects of HLX1 on senescence. A ~2-fold increase in p16^INK4a protein level was detected by SILAC upon HLX1 knockdown in IMR-90 primary human fibroblasts.¹⁷ qRT-PCR confirmed the regulation of p16^INK4a by HLX1. Depletion of INK4a expression rescued the cell cycle arrest triggered by HLX1 knockdown almost completely, indicating that INK4a is a key mediator of the effects of HLX1 on senescence. However, the knockdown of HLX1 still caused a small decline in BrdU incorporation in cells lacking p16^INK4a expression, suggesting that HLX1 could also control senescence by additional mechanisms. As HLX1 is a transcription factor, we looked for other genes that HLX1 could regulate in this context.

A look to the proteomics data identified additional senescence-associated proteins whose expression was altered upon HLX1 knockdown. Levels of matrix metalloproteinase-1 (MMP1), plasminogen activator inhibitor-1 (PAI1), matricellular protein CCN1 and β-galactosidase (GLB1) were found upregulated, while CSIG (cellular senescence-inhibited gene) was downregulated. We performed qRT-PCR to test whether HLX1 could regulate MMP1 expression. We observed that knockdown of HLX1 resulted in an increase in MMP1 mRNA, suggesting that MMP1 regulation is an early response to HLX1 depletion (Fig. 1). A similar induction of MMP1 was observed upon HOXA9 knockdown (Fig. 1). Interestingly, MMP1 has been described as a Polycomb target gene,²¹ and this was confirmed by knocking down the Polycomb protein CBX7 (Fig. 1). These results indicate that the Homeobox proteins HLX1 and HOXA9 regulate MMP1 together with PRCs. Microarray analysis showed that knockdown of HLX1 or CBX7 in IMR-90 cells also upregulated the mRNA levels of PAI1 and CCN1, in accordance with the proteomics observations (ref. 17 and data not shown). Moreover, expression of interleukin-8 (IL8), a pro-inflammatory chemokine part of the SASP,²² which is also a polycomb target,²¹ was induced upon knockdown of HLX1 or CBX7, as observed by microarray and validated by qRT-PCR.¹⁷ MMP1, PAI1, CCN1 and IL8 are all part of the senescence secretome.² PAI1,²³ CCN1²⁴ and IL8²² have all been linked with senescence reinforcement, therefore suggesting that the transcriptional regulation of SASP components by HLX1 and PRCs could contribute to explain HLX1 effects on senescence.

Figure 1. Knockdown of Homeobox genes induces MMP1 expression. MMP1 mRNA level was measured by qRT-PCR in IMR-90 primary human fibroblasts transfected with All stars scrambled siRNA (scrambled) or siRNAs targeting HLX1 (siHLX1), HOXA9 (siHOXA9) or CBX7 (siCBX7).

Despite a more prominent role for p16^INK4a, the cell cycle regulator p15^INK4b (encoded by the CDKN2B gene at the INK4/ARF locus) is equally induced during senescence and its upregulation is functionally significant.²⁵ Although initially we did not detect p15^INK4b in our proteomic studies, microarray analysis highlighted INK4b as one of the genes most significantly induced upon knockdown of HLX1 or CBX7 in IMR-90 cells. Both INK4a and INK4b are subject to PRC-mediated transcriptional repression,²¹ as was confirmed by knocking down CBX7 (Fig. 2A). In addition, knockdown of HLX1 resulted in INK4b upregulation (Fig. 2A), while HLX1 overexpression in IMR-90 cells inhibited INK4b expression (Fig. 2B). Similar observations were made for HOXA9 (Fig. 2A and B). We also tested the effect of MEOX2 (mesenchyme homeobox 2), a Homeobox gene found aberrantly methylated in non-small cell lung carcinoma and previously shown to activate INK4a and cause senescence.²⁶ Consistent with its effect on INK4a and in contrast to that observed with HLX1 and HOXA9, MEOX2 overexpression induced INK4b (Fig. 2B). Altogether these data suggest that both INK4a and INK4b are regulated by homeobox proteins, and that INK4b could be another mediator of the effects of these proteins on senescence. Chromatin immunoprecipitation (ChIP) experiments will be needed to formally prove that HLX1 and HOXA9 directly regulate INK4b and components of the SASP, and that this regulation involves PRC recruitment. In the case of MEOX2, how it activates gene expression is not known. One possible mechanism is the recruitment of CBP/p300 as it has been shown for other Homeobox proteins,²⁷ but a scenario in which there is competition between different Homeobox proteins for binding to DNA or recruiting different co-factors could also be possible.

Figure 2. Regulation of INK4b by Homeobox proteins. INK4b mRNA level was measured by qRT-PCR in IMR-90 primary human fibroblasts transfected with All stars scrambled siRNA (scrambled) or siRNAs targeting HLX1 (siHLX1), HOXA9 (siHOXA9), or CBX7 (siCBX7) (A) or infected with pBabe-HLX1, pMSCV-FLAG-HOXA9, pBabe-HA-MEOX2, or pBabe (vector) (B).

Altered expression of Homeobox genes is often observed in human cancers and contributes to their progression.¹⁸ HLX1 and HOXA9 are frequently overexpressed in patients with acute myeloid leukemia (AML) and high levels of these proteins correlate with reduced survival. HLX1 and HOXA9 regulate normal hematopoiesis and also promote the pathogenesis of AML.^28,29 Transcriptional regulation of PAK1 (p21-activated kinase 1) and BTG1 (B cell translocation gene 1), two genes implicated in the control of cell cycle and malignant proliferation, have been proposed as mechanisms underlying how HLX1 controls leukemogenesis.²⁸ Interestingly, gene set enrichment analysis (GSEA) of HLX1 knockdown in mouse URE leukemic cells showed enrichment in HDAC and PRC1 targets.²⁸ By analyzing this data set, we found that the expression of HDAC1 target genes³⁰ and PRC2 target genes³¹ were also associated with HLX1 depletion,¹⁷ suggesting that HLX1 regulates a subset of PRC target genes in leukemia. In the case of HOXA9, we recently showed that its role in leukemogenesis can be explained in part by its ability to circumvent senescence via INK4a repression.²⁹ Here, we further analyzed HOXA9-driven gene expression in leukemic cells. Inactivation of HOXA9 by 4-OHT withdrawal in HOXA9:ER-transformed myeloblastic cells²⁷ led to the upregulation of genes associated with senescence,³² as well as HDAC1 and HDAC2 targets³⁰ and EZH2 targets³¹ (Fig. 3). In an independent data set derived from t(9;11) MOLM-14 AML cells,³³ we observed the enrichment of PRC2 targets, SUZ12 targets and genes decorated by H3K27me3 marks³⁴ upon HOXA9 knockdown (data not shown). This is reminiscent of what was observed upon HLX1 knockdown in IMR-90 primary human fibroblasts.¹⁷ These results suggest that HOXA9 does not only regulate INK4a, but also additional PRC target genes in the context of leukemia.

Figure 3. HOXA9 regulates PRC target genes in leukemic cells. Gene sets associated with senescence,³² HDAC1, and HDAC2 targets³⁰ and EZH2 targets³¹ are enriched in the transcriptional profile of HOXA9:ER-transformed myeloblastic cells after 4-OHT withdrawal.²⁷ NES, normalized enrichment score; FDR, false discovery rate.

Which are the molecular mechanisms underlying the oncogenic activity of Homeobox proteins is still not fully understood. We propose that the ability of some Homeobox proteins, such as HLX1 and HOXA9, to regulate gene expression through the recruitment of PRC and their ability to delay or circumvent senescence contribute to their oncogenic potential. In our recent work¹⁷ we identified 5 additional homeobox proteins (DLX3, HOXB13, HOXC13, HOXD3 and HOXD8) with similar ability to repress p16^INK4a. Among them, HOXB13 is a known oncogene involved in prostate cancer.³⁵ It will be important to investigate whether these as well as other Homeobox proteins can also regulate senescence and play a role in PRC-mediated transcriptional regulation. Investigating the expression of Homeobox proteins at different stages of tumor progression will contribute to understanding the physiological relevance of the proposed mechanisms. The redundancy of Homeobox proteins as well as their capacity to cooperate should also be explored. For example, we found that HLX1 and HOXA9 are able to heterodimerize.¹⁷ Identification of interacting partners of Homeobox proteins will help to explain how they work at the molecular level. Understanding how alterations of the chromatin status and gene expression lead to malignancy is crucial for the development of new targeted therapy approaches. In this context, strategies to inhibit the expression or activity of HLX1 and HOXA9 could be exploited to revert PRC-mediated transcriptional repression and reactivate loci that have been silenced in cancer.

PRC have a crucial role in the maintenance of pluripotency by inhibiting the expression of lineage-specific transcription factors.³⁶ The INK4-ARF locus is repressed in induced pluripotent stem (iPS) cells and embryonic stem (ES) cells but displays the epigenetic marks of a bivalent chromatin domain, able to be reactivated after differentiation.⁴ It is well established that PRCs play a crucial role in regulating pluripotency in ES cells and iPS cells. Several Homeobox binding motifs are found significantly associated with K27me3-silenced genes in ES cells.³⁷ Whether HLX1 and HOXA9 or other Homeobox proteins can help to maintain pluripotency by recruiting PRCs, either in physiological circumstances or upon enforced expression, needs to be investigated. Senescence has been proposed as a barrier for somatic cell reprogramming.³⁸ Inhibition of different senescence effectors, INK4a among them, significantly improves the efficiency of reprogramming to pluripotency.^3,4 In this context it will be also interesting to investigate the effect of HLX1, HOXA9 and other Homeobox proteins on the efficiency of generation of iPSCs. PRC1 components such as BMI1 sustain the self-renewal capacity of normal and leukemic stem cells.³⁹ INK4a expression gradually increases in stem cell compartments upon aging, limiting the proliferative and regenerative capacity of progenitor cells.⁴⁰ Whether Homeobox proteins such as HLX1 and HOXA9 are implicated as well in the regulation of stem cell self-renewal through PRC-mediated INK4a repression will need to be investigated.

Altogether, the work that we published recently and the additional data presented here highlight PRC recruitment by Homeobox proteins as a novel mechanism for gene regulation, with relevance for senescence, cancer and potentially development. Although INK4a is key to explain the effects of HLX1 and HOXA9 on senescence, data presented here suggest that regulation of INK4b and several components of the SASP also play an important part (summarized in Fig. 4). ChIP-Seq approaches concomitant with mRNA profiling will allow the identification of additional homeobox target genes and suggest cellular functions in which Homeobox proteins could be implicated, possibly in close collaboration with Polycomb repressive complexes.

Figure 4. Functional collaboration between Homeobox proteins and Polycomb repressive complexes in the regulation of cellular senescence. Schematic diagram depicting the proposed model: Homeobox (HOX) proteins like HLX1 and HOXA9 repress the INK4/ARF locus and several genes encoding secreted factors by recruiting Polycomb repressive complex 2 (PRC2). Expression of PRC members decreases upon senescence. Accumulation of products of the INK4/ARF locus and of secreted proteins triggers cell cycle arrest and a senescence-associated secretory phenotype (SASP). During cancer deregulation of PRC or abnormal expression of HLX1 or HOXA9 can contribute to cancel OIS.

Materials and Methods

Cell culture, siRNA transfection and retroviral infection

IMR-90 primary human fibroblasts were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and antibiotics. IMR-90s were reverse transfected with 30 nM siRNA (Qiagen) in 6-well plates using 3.5% solution of Hiperfect transfection reagent (Qiagen). All Stars scrambled siRNA (as negative control), siHLX1 (Hs_HLX_2, SI00439299, target sequence AAGGTTTGAGATTCAGAAGTA), siHOXA9 (Hs_HOXA9_6, SI04239165, target sequence CTACCTAGGGTTTATGCTTAA), and siCBX7 (Hs_CBX7_7, SI04155312, target sequence CGCCGTGGAGAGCATCCGGAA) were used. IMR-90s were retrovirally infected with pBabe-HLX1, pMSCV-FLAG-HOXA9,¹⁷ pBabe-HA-MEOX2 (a kind gift of G. Peters) or empty vector using standard methods.

Quantitative RT-PCR analysis

Total RNA was prepared with the RNeasy Mini kit (Qiagen) and reverse transcribed by SuperScript II (Invitrogen). Real-time quantitative PCR was performed on an Opticon 2 (Biorad) using TaqMan Universal PCR Master Mix (Applied Biosystems) or SYBR Green PCR Master Mix (Applied Biosystems). Taqman Gene Expression Assays (Applied Biosystems) were used to measure MMP1 expression (Hs00899658_m1) and normalize it to TBP (4333769F). Primer sets were used to measure INK4b expression (forward primer: GAATGCGCGAGGAGAACAAG, reverse primer: CCATCATCATGACCTGGATCG) and normalize it to RPS14 (forward primer: TCACCGCCCTACACATCAAACT, reverse primer: CTGCGAGTGCTGTCAGAGG).

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was performed as described previously¹⁷ to examine the association between gene sets and gene expression upon HOXA9 inactivation²⁷ or knockdown.³³ A gene set was considered significantly enriched when the nominal P value was <0.05 and the false discovery rate (FDR) Q value was <0.25.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.