Bmi1 deficient neural stem cells have increased Integrin dependent adhesion to self-secreted matrix

Biochimica et Biophysica Acta 1790 (2009) 351–360

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / b b a g e n

Bmi1 deficient neural stem cells have increased Integrin dependent adhesion to self-secreted matrix Sophia W.M. Bruggeman 1, Danielle Hulsman, Maarten van Lohuizen ⁎ The Netherlands Cancer Institute, Division of Molecular Genetics, Plesmanlaan 121, 1066 CX Amsterdam, the Netherlands

a r t i c l e

i n f o

Article history: Received 2 October 2008 Received in revised form 25 February 2009 Accepted 2 March 2009 Available online 16 March 2009 Keywords: Polycomb group Bmi1 Neural stem cells Integrins Adhesion Brain cancer

a b s t r a c t Background: Neural cells deficient for Polycomb group (PcG) protein Bmi1 are impaired in the formation and differentiation of high grade glioma, an incurable cancer of the brain. It was shown that mechanisms involved in cell adhesion and migration were specifically affected in these tumors. Methods: Using biochemical and cell biological approaches, we investigated the adhesive capacities of Bmi1; Ink4a/Arf deficient primary neural stem cells (NSCs). Results: Bmi1;Ink4a/Arf deficient NSCs have altered expression of Collagen-related genes, secrete increased amounts of extracellular matrix, and exhibit enhanced cell–matrix binding through the Beta-1 Integrin receptor. These traits are independent from the well described role of Bmi1 as repressor of the Ink4a/Arf tumor suppressor locus. Conclusion: In addition to proliferative processes, Bmi1 controls the adhesive capacities of primary NSCs by modulating extracellular matrix secretion. General significance: Since PcG protein Bmi1 is important for both normal development and tumorigenesis, it is vital to understand the complete network in which this protein acts. Whereas it is clear that control of Ink4a/Arf is a major Bmi1 function, there is evidence that other downstream mechanisms exist. Hence, our novel finding that Bmi1 also governs cell adhesion significantly contributes to our understanding of the PcG proteins. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Whereas the widespread neurogenesis taking place in the embryo rapidly ceases after birth [1], two distinct brain regions continuously generate new cells throughout adult life: the subgranular and the subventricular zones (SVZ) [2]. Their gross organization is strikingly similar: slowly dividing glial cells in close proximity of blood vessels and basal laminae give rise to progenitors, which migrate significant distances away from the germinal zone and eventually differentiate into mature cells. In case of brain injury, SVZ cells may also invade the lesion in an attempt to repopulate the damaged area [3,4]. It is not well understood which molecular pathways mediate these activities, however one can assume that signalling molecules as well as contacts between cells and adhesion to matrix molecules play instructive roles. Interestingly, most stem cell niches are distinctly organized in terms of extracellular matrix (ECM) interactions [5–7]. For instance, a specific layered expression pattern of ECM molecules has been described for the SVZ that is partially conserved in neurospheres, clonal clusters of cells formed from cultured SVZ neural stem cells

⁎ Corresponding author. Tel.: +31 20 5122030; fax: +31 20 5122011. E-mail address: [email protected] (M. van Lohuizen). 1 Present address: MRC Centre for Developmental Neurobiology, Guy's Hospital Campus, King's College, London SE1 1EL, UK. 0304-4165/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.bbagen.2009.03.009

(NSCs) [8–10]. It is conceivable that this organized local environment contributes to the determination and migration of the cells of the NSC lineage. Intriguingly, extensive migration is also a hallmark of high grade glioma, a malignant type of brain cancer believed to originate from SVZ stem cells or their relatives [11,12]. It is tempting to speculate that these tumor cells use similar mechanisms for migration as SVZ cells. If this be the case, it would be most informative to understand which genes govern binding to as well as detachment from the SVZ niche. A potential player in this process is the Polycomb group (PcG) protein and epigenetic repressor Bmi1, which has recently been reported to affect the adhesive properties of glioma cells [13]. A well known target of this protein is the Ink4a/Arf tumor suppressor locus [14]. Previous studies have extensively investigated the relationship between Bmi1 and Ink4a/Arf in stem cells [15–18]. However, it is clear that not all phenotypes observed in Bmi1 deficient brain can be explained by Ink4a/Arf deregulation, prompting a search for Ink4a/Arf independent Bmi1 functions [15]. We now report that mouse NSCs deficient for Bmi1 strongly adhere to a Collagen-like extracellular matrix compound secreted by these cells in culture, independently from Ink4a/Arf. Control NSCs that normally do not adhere also bind to the matrix deposited by Bmi1 deficient cells. This binding is mediated by the Beta-1 Integrin receptor that is known to stimulate NSC proliferation and survival

352

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

[9,19]. This provides Bmi1 with a dual function in stem cells: the first is to safeguard stem cell proliferation by means of controlling the expression of Ink4a/Arf, and the second being modulation of Integrin– matrix binding of NSCs through an alternative pathway. We speculate that both functions may play a role in gliomagenesis. 2. Materials and methods 2.1. Cell culture SVZ derived NSCs were isolated from FVB mice as described before [13,15]. Embryonic NSCs were derived from the forebrain of E12.5 embryos. They were routinely propagated on poly-Ornithine and Laminin coated plastic dishes as described before [13,20]. For retroviral transductions, cells were incubated with supernatant from transfected Phoenix producer cells for approximately 6 h on 3 consecutive days, and selected with 4 μg/ml puromycin or 10 μg/ml blasticidin [13]. For knockdown experiments, constructs used were pRetrosuper GFPi or Bmi1i [13]. To label NSCs, constructs used were pMSCV puromycin GFP-IRES-Luciferase (kind gift of Dr. A. Berns) and pMSCV blasticidin (both described in [13]). For some experiments, wells were coated overnight with polyL-ornithine (15 μg/ml, Sigma) followed by 3 h coating with Laminin (5 μg/ml, Sigma), Fibronectin (150 μg/ml, Biomedical Technologies), Collagen (2 mg/ml, Roche) or Gelatin (0.15%, Sigma). 2.2. Adhesion assays For enzymatic detachment of cells, 300,000 NSCs were seeded in triplicate into uncoated 6 wells plates. The following day, cells were treated for 4–10 min with increasing concentrations of Hyaluronidase (Sigma) or Collagenase A (Roche). The amount of remaining attached cells was determined using Crystal Violet staining [15]. EDTA and Beta-1 Integrin blocking experiments were done as follows: 30,000 NSCs were seeded in duplicate in a final volume of 100 μl onto either non-coated, Laminin or Collagen coated 96 wells plates in the absence or presence of increasing concentrations EDTA in Hank's balanced salt solution (Invitrogen, Ca2+ and Mg2+ free); or in the absence or presence of 10 μg/ml Beta-1 Integrin blocking antibody (hamster-anti-mouse Ha2/5, Pharmingen) in NSC medium. Cells were allowed to attach for 3 h and directly fixed with formalin. The amount of attached cells was determined using Crystal Violet staining. The co-culture experiment was performed as follows: Ink4a/Arf−/− NSCs were labeled with GFP and the Puromycin resistance gene (IA-GFP/Puro), Bmi1−/−;Ink4a/Arf−/− NSCs were labeled with the Blasticidin resistance gene (BIA-Blast). Onto uncoated 6 wells plates, 100,000 IA-GFP/Puro NSCs, 100,000 BIA-Blast NSCs, a mixture of these cells (total of 200,000 cells), or 300,000 BIA-Blast NSCs, were seeded. Approximately 48 h after plating, the 300,000 BIA-Blast NSCs were removed either with 4 μg/ml puromycin or 20 mM EDTA/PBS in the presence of protease inhibitors (Roche). Wells were washed with PBS, and replated with 100,000 IA-GFP/Puro NSCs. The IA-GFP/Puro and BIA-Blast mixtures were treated with 4 μg/ml puromycin approximately 36 h after plating. Wells were scored for adhering cells and GFP signal. 2.3. Flow cytometry Cells were removed from the plate by applying 20 mM EDTA/ PBS for 5 min at 4 °C and kept on ice during the entire procedure. They were washed once with 0.5% BSA/PBS (FACS buffer). 500,000 cells were incubated with primary antibody in FACS buffer for 1 h. Following two wash steps, they were incubated with secondary antibody for 1 h and subsequently analyzed on a FACSCalibur (Becton Dickinson) using Cell Quest Pro software. Primary

antibodies were hamster-anti-mouse CD49a (Alpha-1 Integrin, Ha31/8), anti-CD49b (Alpha-2 Integrin, HMα2), and anti-AlphaV (H9.2B8, all Pharmingen); rat-anti-mouse Alpha-6 (GoH3 supernatant, gift from Dr. Sonnenberg), and rat-anti-mouse Beta-1 (MAB1997, Chemicon). Concentrations were 20 μg/ml except for GoH3 which was diluted 1:1. Secondary antibodies were donkeyanti-rat FITC (Rockland) and goat-anti-hamster Cy5 (Jackson Immunoresearch) diluted 1:250. The level of apoptotic cells was determined using the Annexin V-FITC Apoptosis Detection kit according to the manufacturer's protocol for flow cytometrical analysis (Abcam). 2.4. Quantitative real time PCR, oligo-microarrays and pathway-focussed arrays RNA was extracted either directly from Ink4a/Arf−/− versus Bmi1−/−; Ink4a/Arf−/− NSCs, or from Ink4a/Arf−/− NSCs transduced with pRetrosuper GFPi versus Bmi1i using Trizol reagent (Invitrogen). qRT-PCR was described before [13,15]. Primers used were Bmi1-1: sense 5′-TGGAGACCAGCAAGTATTGTCCTA-3′, antisense 5′-CTTATGTTCAGGAGTGGTCTGGTTTT-3′ and Bmi1-2: sense 5′-AAGAAGAGATTTT-TATGCAGCTCACC-3′, antisense 5′-GAGCCATTGGCAGCATCAG-3′. Alpha-1 Integrin-1 sense: 5′AGCAGCAGCAACCGGAAAC-3′, antisense 5′-GGAGCCCGTCTTTGGATATCT3′, Alpha-1 Integrin-2 sense 5′-TCATCTTGTGGAAACCAACTTTCA-3′, antisense 5′-GATTTAAGCTGGAAAAATGTGCTTT-3′; Procollagen type III (alpha1)-1 sense 5′-CAGGAGCAC-GCGGTGAA-3′, antisense 5′TTTGCCATCTTCGCCCTTAG-3′, Procollagen type III(alpha1)-2 sense 5′TTGGAATTGCAGGGCTAACTG-3′, antisense 5′-CATTATGGCCACTGGCTCCT3′; Procollagen type V(Alpha2)-1 sense 5′-GCCCGCACTTGTGATGATTT-3′, antisense 5′-TCACCGCTCTGCTTTGTGG-3′, Procollagen type V(Alpha2)-2 sense 5′-GTGGAAAGGCTGGTGATCAACT-3′, antisense 5′-GAAGCATGTGTGTCAGGTTCAGAT-3′; Pcolce1-1 sense 5′-TTCCCCAACCTCTACCCCC-3′, antisense 5′-CACCGTAATTGTCCAGATGCA-3′, Pcolce1-2 sense 5′-CCCTCAAACCAGGTGATCATG-3′, antisense 5′-CAGGCTCCACATCAAACTTCC-3′, Pcolce1-3 sense 5′-ATTTTGCGGAGACAAGGCC-3′, antisense 5′-GAGCTCGTTCCCTTCAGAAGAG-3′; Emilin1-1 sense 5′TCCGTACGTTGTGACTGCGT-3′, antisense 5′-GTCAAGAGGAGATGCCTCGG3′, Emilin1-2 sense 5′-CAGCCCCACTGTTCCCG-3′, antisense 5′GCGAGGGCGAAGAAAGCT-3′. Oligo-microarray analysis and Ingenuity software-based ontological clustering were done as described previously [13]. Complete expression data is available from the EBI ArrayExpress database (http://www.ebi.ac.uk/microarray-as/ae/), EBI-ID: E-NCMF-22. Lists of significant outliers are available upon request. Pathway focussed oligoarrays (OligoGEArray Mouse ECM and Adhesion molecules microarray OMM-013-4, SuperArray) and qRTPCR arrays (Mouse ECM and Adhesion molecules RT2Profiles TM PCR array APM-013a-2, SuperArray) were performed according to the manufacturer's protocols. 3. Results and discussion 3.1. Bmi1 deficient NSCs have increased Integrin-mediated matrix adhesion SVZ derived NSCs grow as floating clusters of stem and progenitor cells termed ‘neurospheres’ when cultured under non-adherent conditions [10]. We have shown previously that when cultured under such neurosphere-forming conditions, NSCs deficient for both Bmi1 and Ink4a/Arf (Bmi1−/−;Ink4a/Arf−/− NSCs) can be readily propagated for several passages as neurospheres that are indistinguishable from Ink4a/Arf−/− control spheres in terms of self-renewal or differentiation capacity [15]. Remarkably, we now observe that when grown on normal tissue culture-treated dishes, Bmi1−/−;Ink4a/ Arf−/− NSCs do not form neurospheres, but instead adhere to the dish and form a monolayer (Fig. 1A, left panels). When plated onto a

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

353

Fig. 1. Bmi1−/−;Ink4a/Arf−/− NSCs have increased adhesive capacities. (A) (Bmi1−/−;)Ink4a/Arf−/− NSCs were plated onto different substrates. Control cells (upper panels) only firmly adhere to Laminin coatings whereas Bmi1 deficient cells adhere to all substrates including standard tissue culture plastic. (B) Bmi1 mRNA levels after treatment of Ink4a/ Arf−/− NSCs with a short hairpin against GFP (shGFP) or Bmi1 (shBmi1) measured with qRT-PCR. (C) Ink4a/Arf−/− NSCs treated with shBmi1 adopt an adherent growth mode on uncoated plates. (D) Bmi1−/− embryonic NSCs (Embryonic day E12.5) attach more efficiently to Gelatin than control cells.

Laminin substrate, Ink4a/Arf−/− control NSCs could be induced to grow in a similar fashion (Fig. 1A, right panels). Onto either Fibronectin, Gelatin or Collagen, control cells can adhere albeit less tightly (Fig. 1A, middle and data not shown), whereas Bmi1−/−; Ink4a/Arf−/− NSCs formed stable monolayers under all conditions. Importantly, we could mimic this phenotype by removing Bmi1 acutely from Ink4a/Arf−/− or Arf−/− NSCs using an shRNA targeting Bmi1 (Fig. 1B, C and data not shown). This demonstrates that the adherent growth mode does not reflect any indirect effects of Bmi1 absence, such as compensation during development. To exclude the possibility that the absence of the Ink4a/Arf locus is involved in this process, we also studied the behaviour of primary Bmi1−/− cells in an Ink4a/Arf proficient background. These cells cannot be maintained for longer periods of time as they undergo a premature growth arrest due to Ink4a/Arf upregulation [15]. However we observed that on a Gelatin coating, the few Bmi1−/− neurospheres that were generated spread out more efficiently than the wild type control spheres (Fig. 1D). Next, we wanted to know which membrane proteins were mediating Bmi1−/−;Ink4a/Arf−/− NSC adhesion. One major group of ECM binding receptors that we tested for involvement was the Integrin family. Integrins are heterodimeric proteins consisting of an alpha and a beta subunit [21]. Once localized to the membrane, they can change from an inactive to an active conformation upon certain stimuli (reviewed in [22,23]). Subsequent higher order-clustering may further enhance the affinity of these receptors for their substrates (reviewed in [24]). Using flow cytometry, we tested and confirmed the presence of the Alpha-1, -2, -6, -V and Beta-1 receptors on control and Bmi1 deficient NSCs (Fig. 2A, Supplemental Fig. 1, see also Fig. 6A). To adopt the active

ligand-binding form, Integrins (like some other receptors) must engage with a divalent cation. To test whether the Integrin receptors could be involved in the binding of Bmi1 deficient cells, we exploited this feature by removing all positive ions from the medium using EDTA as chelator. We found that with increasing concentrations EDTA, attachment of Bmi1−/−;Ink4a/Arf−/− NSCs was inhibited (Fig. 2B, left panel). This was not the result of an increase in cell death (Supplemental Fig. 2). Notably, when cells were offered a substrate, adhesion of both Bmi1 deficient and control cells was strongly reduced, suggesting that under these circumstances all NSCs may use their Integrins for attachment (Fig. 2B, Laminin and Collagen panels). We want to point out that for these and all of the following experiments, we used primary NSCs that were propagated as adherent cell cultures. We have shown before that both the Bmi1 deficient and control cultures are highly enriched for multipotent NSCs. Furthermore, these cultures closely resemble each other in terms of stem cell and differentiation marker expression [13]. Hence, we can assume that differences found are directly due to Bmi1 absence and not to confounding factors like heterogeneity of the cell cultures. Integrin receptors and ligands play diverse roles in brain development and cancer. Their spatial and temporal expressions suggest an instructing function in the positioning of the cortical layers and other structures [25]. Accordingly, deletion of the Beta-1 Integrin or some of its Alpha-subunit binding partners leads to severe brain defects [26–30]. As the Beta-1 Integrin is the most widely used beta subunit and also known to be involved in adult NSC proliferation, migration and survival [9,19,31], we subsequently investigated whether the specific blocking of this protein could prevent attachment of Bmi1 deficient cells. Indeed, we observed that the application of a Beta-1

354

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

Fig. 2. Bmi1−/−;Ink4a/Arf−/− NSCs use Integrin receptors to attach. (A) Schematic overview of Integrins expressed on (Bmi1−/−;)Ink4a/Arf−/− [(B)IA] NSCs as determined using flow cytometry. (B) (Bmi1−/−;)Ink4a/Arf−/− NSC binding to uncoated, Laminin coated, or Collagen coated plates in the presence of increasing concentrations of EDTA was determined using Crystal Violet staining. Two independently derived cell lines from each genotype are shown. (C) (Bmi1−/−;)Ink4a/Arf−/− NSC binding to uncoated, Laminin coated, or Collagen coated plates in the presence or absence of an isotype control antibody or a blocking antibody against the Beta-1 Integrin was determined using Crystal Violet staining. (D) Phase-contrast microscopy photographs of (Bmi1−/−;)Ink4a/Arf−/− NSCs show that binding of Bmi1−/−;Ink4a/Arf−/− NSCs is strongly inhibited in the presence of a Beta-1 Integrin blocking antibody but not a control antibody.

blocking antibody but not an isotype control antibody inhibited binding of Bmi1−/−;Ink4a/Arf−/− NSCs, which now instead started to grow as neurospheres (Fig. 2C left panel, Fig. 2D). Similar to the EDTA experiments, on protein coated plates, binding of control NSCs was also prevented in the presence of the blocking antibody (Fig. 2C, middle and right panels). It should be noted though that some residual binding occurred. This may be due to technical limitations of the assay, but it is also possible that other receptors play a role in the binding in addition to the Beta-1 Integrin. 3.2. Bmi1 deficient NSCs generate extracellular matrix used for cell attachment These experiments suggested that Bmi1 deficient and control NSCs do not intrinsically differ in their capacity to bind to the matrix using the Beta-1 Integrin, as long as the proper substrate is present. This led us to hypothesize that the increased binding of Bmi1 deficient cells is

secondary to the presence of one or more ECM molecules that are not, or less abundantly, present in control cultures. To explore this possibility, we transduced control Ink4a/Arf−/− cells with a GFP and Puromycin-resistance gene expressing retrovirus (referred to as IA-GFP/Puro), and the Bmi1−/−;Ink4a/Arf−/− cells with a Blasticidinresistance gene expressing virus (BIA-Blast), allowing us to distinguish between the two cell types. When these NSCs were grown on standard tissue culture-treated plastic, they took on the expected growth mode: the IA-GFP/Puro cell line formed neurospheres while the BIA-Blast cells grew in a monolayer (Fig. 3A). The IA-GFP/Puro NSCs could be visualized with UV light verifying GFP expression. Surprisingly, when we mixed the two cell populations, the IA-GFP/ Puro NSCs readily contributed to the monolayer, showing that they can attach to the matrix if they are in the presence of Bmi1 deficient cells (Fig. 3B). To test whether this control cell-binding is truly the result of attachment to matrix deposited by Bmi1 deficient cells and not of some cell–cell interaction, we removed a three day old

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

355

Fig. 3. Bmi1−/−;Ink4a/Arf−/− NSCs secrete a matrix component that is used for attachment. (A) Ink4a/Arf−/− NSCs were transduced with GFP and Puromycin resistance gene containing retrovirus (IA-GFP/Puro), Bmi1−/−;Ink4a/Arf−/− NSCs were transduced with Blasticidin resistance gene containing retrovirus (BIA Blast). They were plated onto uncoated plates. Upper panels represent phase-contrast photographs demonstrating the floating growth mode for control cells and attached growth mode for the Bmi1 deficient cells. Lower panels show GFP staining as revealed using UV light. (B) Control IA-GFP/Puro cells grow adherent when mixed with BIA-Blast cells. (C) When IA-GFP/Puro NSCs are replated, they continue to grow as neurospheres on uncoated plates (left panels), but form attached monolayers when they are plated onto plates where secreted matrix from BIA-Blast cells is present (right panels). (D) When BIA-Blast cells are removed from adherently growing mixtures of IA-GFP/Puro–BIA-Blast cultures using Puromycin treatment, IA-GFP/Puro cells continue to grow as attached monolayers.

monolayer of pure BIA-Blast cells either with EDTA (Fig. 3C) or Puromycin treatment (not shown) and subsequently plated IA-GFP/ Puro NSCs onto the remaining dish. Supporting our hypothesis, in both cases the IA-GFP/Puro cells attached and formed monolayers. A similar result was obtained when we removed BIA-Blast cells from a co-culture with Puromycin (Fig. 3D). As incubation of Ink4a/Arf−/− cells with Bmi1−/−;Ink4a/Arf−/− conditioned medium had no effect on adhesion (not shown), together these data strongly suggest that the Bmi1−/−;Ink4a/Arf−/− NSCs have secreted a matrix attached to the plate that evokes neural stem cell attachment. Notably, the attachment is reversible as Ink4a/Arf−/− cells grown as a monolayer on this matrix return to the neurosphere growth mode when replated onto normal plastic (not shown). 3.3. Identifying Ink4a/Arf-independent Bmi1 targets We then attempted to unravel the mechanism relaying Bmi1 controlled adhesion. Since Bmi1 belongs to a transcription factor family that binds at multiple sites throughout the genome, we

reasoned that its absence likely inflicts deregulation of a number of genes [32–36]. To study gene expression patterns on a genome wide scale, we cultured (Bmi1−/−);Ink4a/Arf−/− NSCs as monolayers on either Laminin or Fibronectin coated plates and hybridized their RNA onto oligo-microarrays (n = 2 experiments, each confirmed with a dye swap experiment). These experiments gave an average of 594 outliers (complete dataset available from the EBI ArrayExpress database, see Materials and Methods) from which approximately half was upregulated (Supplemental Fig. 3). Plotting the two data sets in a ratio scatter plot revealed a nice correlation (Fig. 4A), and the common outliers clustered perfectly suggesting consistent deregulation of Bmi1 target genes on different substrates (Fig. 4A, C). This was somewhat surprising since control cells bind more efficiently to Laminin than to Fibronectin, suggesting possible differences in the expression of Fibronectin (binding) genes. However, as the microarray data already suggested, we could not find such differences in the expression of either Fibronectin or its receptors in an independent experiment (Supplemental Fig. 4, see also Supplemental Fig. 1B).

356

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

Fig. 4. Altered gene expression in Bmi1−/−;Ink4a/Arf−/− NSCs. (A) Oligo-microarray analysis was performed on Ink4a/Arf−/− versus Bmi1−/−;Ink4a/Arf−/− NSCs growing either on Laminin (LM) or on Fibronectin (FN). Plotted are the outliers from the Laminin arrays versus the Fibronectin arrays. (B) Ontological analysis demonstrating the most significantly altered ontological clusters in the absence of Bmi1 for the Laminin (L) and Fibronectin (F) arrays. (C) Clustering of common outliers from the Laminin (LM) and Fibronectin (FN) arrays. Red indicates upregulated in the absence of Bmi1, green indicates downregulated. (D) List of the common top outliers from the Laminin and Fibronectin arrays.

Among the most induced common targets was a substantial set of Homeobox-containing genes that are known PcG targets (Fig. 4D) [37]. Ontological clustering of the oligoarray data using Ingenuity Pathway Analysis software revealed a number of functionally related gene groups that were significantly enriched for in Bmi1 deficient

NSCs. The top ten (Fig. 4B) included proliferation and development associated groups (‘Cellular Growth and Proliferation’; ‘Cellular Development’; Connective Tissue/Nervous System/Skeletal/Muscular Tissue Development and Function; Organ/Tissue Development) in agreement with earlier reports on PcG gene function. Interestingly,

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

also the group ‘Cellular Movement’ was significantly enriched for, which might reflect the role of Bmi1 in adhesion as this group contains genes involved in processes like cell movement, migration, and chemotaxis. Of note, this ontological group was also shown to be deregulated in Bmi1 deficient brain tumors [13]. Moreover, a recent study showed that acute removal of Bmi1 from sarcoma or medulloblastoma cell lines also affects the expression of adhesion related genes, as well as their adhesive capacities [38]. To follow up on this, we set out to more specifically analyze expression of genes involved in ECM adhesion. Hereto, we employed pathway-focused oligoarrays (GEArrays) and quantitative real time PCR (qRT-PCR) arrays that only contained probes for ECM-related genes (Fig. 5A, n = 4 experiments in total). We compared either gene expression of Ink4a/Arf−/− versus Bmi1−/−;Ink4a/Arf−/− NSCs, or

357

Ink4a/Arf−/− NSCs treated with a control shRNA (shGFP) versus an shRNA against Bmi1 (shBmi1), in order to reduce background due to compensation mechanisms or unspecific side effects or RNAi treatment, respectively (Fig. 5B). We combined the common outliers of the separate experiments and generated a list of genes that consistently demonstrated expression levels of at least 120% of the control (Fig. 5C). It should be noted though that overall, the changes in gene expression were relatively small, except for Emilin1 and Tenascin. 3.4. Novel Bmi1 targets are associated with Collagen metabolism Our efforts to identify novel Bmi1 targets using large-scale approaches provided us with a number of genes with functions linked to matrix binding or assembly that were differentially expressed in

Fig. 5. Pathway focussed arrays reveal ECM and adhesion-related genes altered in the absence of Bmi1. (A) Either Ink4a/Arf−/− NSCs treated with short hairpins against GFP (shGFP) versus Bmi1 (shBmi1), or Ink4a/Arf−/− versus Bmi1−/−;Ink4a/Arf−/− NSCs, were used to specifically study ECM and adhesion-related gene expression in qRT-PCR arrays or in Hybridization GEArrays (oligoarrays). (B) GEArray oligoarrays were hybridized with biotinylated cRNA from (Bmi1−/−;)Ink4a/Arf−/− NSCs (upper panels) and Ink4a/Arf−/− NSCs treated with shGFP or shBmi1 (lower panels). Using enhanced chemiluminescence, photographic films were exposed. The intensity of the spots was determined using GEArray software. (C) Summary of outliers (as defined by genes having expression levels of at least 120% compared to controls in 3 or more experiments) induced in the absence of Bmi1 in both the qRT-PCR and oligoarrays. BIA = Bmi1−/−;Ink4a/Arf−/− NSCs.

358

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

Bmi1 deficient NSCs. However, several of them could not be confirmed on new, independently generated biological replicas indicating a certain level of noise in both the large scale and pathway-focused array experiments (data not shown). Fortunately, we were also able to robustly verify some targets on which we have focused in Fig. 6. When performing flow cytometrical analyses to establish the Integrin expression pattern (Fig. 2A), we noticed an increase in the expression level of the Alpha-1 Integrin (Itga1) on the Bmi1−/−;Ink4a/Arf−/− cells compared to Ink4a/Arf−/− or wild type controls (Fig. 6A). This

change could also be observed at the mRNA level (Fig. 6B). Together with the Beta-1 Integrin subunit, this protein forms the major Collagen receptor VLA1. However, preliminary results suggest that blocking of the Alpha-1 Integrin does not prevent NSC binding, indicating that if this dimer plays a role in NSC attachment, it is not an exclusive one (S.B., unpublished observations). Nonetheless, we obtained more data pointing in the direction of altered Collagen metabolism in Bmi1 deficient cells. We found increased mRNA levels for Collagen-3(alpha1), Collagen-4(alpha1), Collagen-4(alpha2),

Fig. 6. Putative novel Ink4a/Arf-independent Bmi1 targets. (A) Flow cytometrical analysis for Alpha-1 Integrin expression in wild type (upper right panel), Ink4a/Arf−/− (lower left panel) or Bmi1−/−;Ink4a/Arf−/− (lower right panel) NSCs. Note a dramatic increase in the percentage of Alpha-1 Integrin positive cells in the Bmi1−/−;Ink4a/Arf−/− NSCs (12.1%–13.5% positive in control cultures versus 84.6% in Bmi1 deficient cells). (B) Alpha-1 Integrin mRNA levels are induced in the absence of Bmi1 as determined by qRT-PCR. Loading controls are given between brackets. IA = Ink4a/Arf−/− NSCs, BIA = Bmi1−/−;Ink4a/Arf−/− NSCs, shGFP = short hairpin against GFP in Ink4a/Arf−/− NSCs, shBmi1 = short hairpin against Bmi1. (C) Enzymatic detachment of Bmi1−/−;Ink4a/Arf−/− NSCs by Hyaluronidase of Collagenase A. Hyaluronidase treatment does not lead to a reduction in the number of attached cells in time (grey lines), whereas cells treated with increasing amounts of Collagenase A detach from the plate in a dose and time-dependent manner (black lines). Measurements were made after 4 min (4′) and 10 min (19′) of treatment. (D) mRNA levels of putative novel Bmi1 targets Collagen3(Alpha1), Collagen5(Alpha2) and Pcolce in control versus Bmi1−/−;Ink4a/Arf−/− NSCs as determined by qRT-PCR. (E) mRNA levels of putative novel Bmi1 target Emilin1 as determined by qRT-PCR.

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

Collagen-5(alpha2) chains, and for Pcolce (Procollagen Proteinase Enhancer), a protein catalyzing Collagen metabolism (Fig. 6D and data not shown). Additional evidence suggesting a Collagen-like protein to be involved in the binding of Bmi1 deficient NSCs came from an experiment in which we tried to abrogate cell attachment with enzymes cleaving specific ECM molecules (Fig. 6C). We could demonstrate that even at very low concentrations, a bacterial Collagenase was very efficient at diminishing ECM binding of Bmi1−/−;Ink4a/Arf−/− cells whereas Hyaluronidase, a proteinase with no cleavage activity towards Collagens, was not. Interestingly Emilin1, another candidate target we confirmed with qRT-PCR (Fig. 6E), is an extracellular matrix protein that structurally closely resembles the Collagens and has been shown to be bound primarily by the Beta-1 Integrin [39–41]. Lastly, a role for Bmi1 in Collagen metabolism is further supported by a recent finding that acute removal of Bmi1 from medulloblastoma cell lines leads to increased Collagen expression in xenografts [42]. The question remains what the precise function of PcG mediated control of adhesion is and how that control is relayed within the cell. During development, constant crosstalk between brain cells and ECM is essential for the proper execution of organogenesis (reviewed in [1,43]). The aberrant arborisation of Bmi1 null cerebellar neurons might be an example of defective axon-pathfinding due to misguidance along matrix molecules [15]. If true, we can anticipate the existence of additional abnormalities associated with adhesion-defects in the Bmi1 null brain. Close examination of cortical layering or adult neuroblast migration might reveal such defects. Unfortunately at this time, we can only speculate which signal transduction pathways are involved in PcG controlled adhesion. One possibility is enhanced TGFβ superfamily signalling. TGFβ-like proteins have been reported of inducing expression of secreted ECM molecules as well as Integrin receptors, which fits nicely with our observations [44,45]. Furthermore, recent binding and expression studies indicate a role for PcG in TGFβ signalling [33,34,46]. However, the existence of a direct link between PcG, TGFβ and cell adhesion remains to be demonstrated. Altogether, the finding that Bmi1 controls Integrin mediated adhesion to self-secreted matrix has important implications for understanding PcG function during embryogenesis and normal adult homeostasis. But it might also explain part of the impact that PcG proteins have on tumorigenesis, as the progression of cancer heavily relies on altered adhesion and migration through the stroma in addition to increased proliferation and resistance against cell death. This may be especially important in cancers with a strong migratory character such as glioma, which for this reason are particularly difficult to eradicate. Acknowledgements We would like to thank Drs. Charles ffrench-Constant, Peter Hall, Laurien Ulfman, and Arnoud Sonnenberg for reagents and critical discussions. We also thank Dr. Ron Kerkhoven, Ellen Tanger, the NKI Microarray facility, FACS facility and Microscopy facility for assistance with experiments. This work is supported with grants from the Dutch Cancer Society/KWF and the Nederlandse Hersenstichting to S.B. and M.v.L. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.bbagen.2009.03.009. References [1] S.L. Donovan, M.A. Dyer, Regulation of proliferation during central nervous system development, Semin. Cell. Dev. Biol. 16 (2005) 407–412. [2] A. Alvarez-Buylla, D.A. Lim, For the long run: maintaining germinal niches in the adult brain, Neuron 41 (2004) 683–686.

359

[3] G.E. Goings, V. Sahni, F.G. Szele, Migration patterns of subventricular zone cells in adult mice change after cerebral cortex injury, Brain Res. 996 (2004) 213–226. [4] J. Macas, C. Nern, K.H. Plate, S. Momma, Increased generation of neuronal progenitors after ischemic injury in the aged adult human forebrain, J. Neurosci. 26 (2006) 13114–13119. [5] E. Fuchs, T. Tumbar, G. Guasch, Socializing with the neighbors: stem cells and their niche, Cell 116 (2004) 769–778. [6] M.J. Murphy, A. Wilson, A. Trumpp, More than just proliferation: Myc function in stem cells, Trends Cell Biol. 15 (2005) 128–137. [7] S.K. Nilsson, H.M. Johnston, G.A. Whitty, B. Williams, R.J. Webb, D.T. Denhardt, I. Bertoncello, L.J. Bendall, P.J. Simmons, D.N. Haylock, Osteopontin, a key component of the hematopoietic stem cell niche and regulator of primitive hematopoietic progenitor cells, Blood 106 (2005) 1232–1239. [8] L.S. Campos, Beta1 integrins and neural stem cells: making sense of the extracellular environment, Bioessays 27 (2005) 698–707. [9] L.S. Campos, D.P. Leone, J.B. Relvas, C. Brakebusch, R. Fassler, U. Suter, C. FfrenchConstant, Beta1 integrins activate a MAPK signalling pathway in neural stem cells that contributes to their maintenance, Development 131 (2004) 3433–3444. [10] B.A. Reynolds, S. Weiss, Clonal and population analyses demonstrate that an EGF-responsive mammalian embryonic CNS precursor is a stem cell, Dev. Biol. 175 (1996) 1–13. [11] N. Sanai, A. Alvarez-Buylla, M.S. Berger, Neural stem cells and the origin of gliomas, N. Engl. J. Med. 353 (2005) 811–822. [12] Y. Zhu, F. Guignard, D. Zhao, L. Liu, D.K. Burns, R.P. Mason, A. Messing, L.F. Parada, Early inactivation of p53 tumor suppressor gene cooperating with NF1 loss induces malignant astrocytoma, Cancer Cell 8 (2005) 119–130. [13] S.W. Bruggeman, D. Hulsman, E. Tanger, T. Buckle, M. Blom, J. Zevenhoven, O. van Tellingen, M. van Lohuizen, Bmi1 controls tumor development in an Ink4a/Arfindependent manner in a mouse model for glioma, Cancer Cell 12 (2007) 328–341. [14] J.J. Jacobs, K. Kieboom, S. Marino, R.A. Depinho, M. van Lohuizen, The oncogene and Polycomb-group gene bmi-1 regulates cell proliferation and senescence through the ink4a locus, Nature 397 (1999) 164–168. [15] S.W. Bruggeman, M.E. Valk-Lingbeek, P.P. van der Stoop, J.J. Jacobs, K. Kieboom, E. Tanger, D. Hulsman, C. Leung, Y. Arsenijevic, S. Marino, M. van Lohuizen, Ink4a and Arf differentially affect cell proliferation and neural stem cell self-renewal in Bmi1-deficient mice, Genes Dev. 19 (2005) 1438–1443. [16] A.V. Molofsky, S. He, M. Bydon, S.J. Morrison, R. Pardal, Bmi-1 promotes neural stem cell self-renewal and neural development but not mouse growth and survival by repressing the p16Ink4a and p19Arf senescence pathways, Genes Dev. 19 (2005) 1432–1437. [17] A.V. Molofsky, R. Pardal, T. Iwashita, I.K. Park, M.F. Clarke, S.J. Morrison, Bmi-1 dependence distinguishes neural stem cell self-renewal from progenitor proliferation, Nature 425 (2003) 962–967. [18] I.K. Park, D. Qian, M. Kiel, M.W. Becker, M. Pihalja, I.L. Weissman, S.J. Morrison, M.F. Clarke, Bmi-1 is required for maintenance of adult self-renewing haematopoietic stem cells, Nature 423 (2003) 302–305. [19] D.P. Leone, J.B. Relvas, L.S. Campos, S. Hemmi, C. Brakebusch, R. Fassler, C. FfrenchConstant, U. Suter, Regulation of neural progenitor proliferation and survival by beta1 integrins, J. Cell Sci. 118 (2005) 2589–2599. [20] L. Conti, S.M. Pollard, T. Gorba, E. Reitano, M. Toselli, G. Biella, Y. Sun, S. Sanzone, Q.L. Ying, E. Cattaneo, A. Smith, Niche-independent symmetrical self-renewal of a mammalian tissue stem cell, PLoS Biol. 3 (2005) e283. [21] D.O. Clegg, K.L. Wingerd, S.T. Hikita, E.C. Tolhurst, Integrins in the development, function and dysfunction of the nervous system, Front. Biosci. 8 (2003) d723–d750. [22] T.A. Springer, J.H. Wang, The three-dimensional structure of integrins and their ligands, and conformational regulation of cell adhesion, Adv. Protein. Chem. 68 (2004) 29–63. [23] J.A. Askari, P.A. Buckley, A.P. Mould, M.J. Humphries, Linking integrin conformation to function, J. Cell. Sci. 122 (2009) 165–170. [24] T.S. Gomez, D.D. Billadeau, T cell activation and the cytoskeleton: you can't have one without the other, Adv. Immunol. 97 (2008) 1–64. [25] R.S. Schmid, E.S. Anton, Role of integrins in the development of the cerebral cortex, Cereb. Cortex 13 (2003) 219–224. [26] E.S. Anton, J.A. Kreidberg, P. Rakic, Distinct functions of alpha3 and alpha(v) integrin receptors in neuronal migration and laminar organization of the cerebral cortex, Neuron 22 (1999) 277–289. [27] B.L. Bader, H. Rayburn, D. Crowley, R.O. Hynes, Extensive vasculogenesis, angiogenesis, and organogenesis precede lethality in mice lacking all alpha v integrins, Cell 95 (1998) 507–519. [28] E. Georges-Labouesse, N. Messaddeq, G. Yehia, L. Cadalbert, A. Dierich, M. Le Meur, Absence of integrin alpha 6 leads to epidermolysis bullosa and neonatal death in mice, Nat. Genet. 13 (1996) 370–373. [29] D. Graus-Porta, S. Blaess, M. Senften, A. Littlewood-Evans, C. Damsky, Z. Huang, P. Orban, R. Klein, J.C. Schittny, U. Muller, Beta1-class integrins regulate the development of laminae and folia in the cerebral and cerebellar cortex, Neuron 31 (2001) 367–379. [30] J.A. Kreidberg, M.J. Donovan, S.L. Goldstein, H. Rennke, K. Shepherd, R.C. Jones, R. Jaenisch, Alpha 3 beta 1 integrin has a crucial role in kidney and lung organogenesis, Development 122 (1996) 3537. [31] T.S. Jacques, J.B. Relvas, S. Nishimura, R. Pytela, G.M. Edwards, C.H. Streuli, C. Ffrench-Constant, Neural precursor cell chain migration and division are regulated through different beta1 integrins, Development 125 (1998) 3167–3177. [32] L.A. Boyer, K. Plath, J. Zeitlinger, T. Brambrink, L.A. Medeiros, T.I. Lee, S.S. Levine, M. Wernig, A. Tajonar, M.K. Ray, G.W. Bell, A.P. Otte, M. Vidal, D.K. Gifford, R.A. Young, R. Jaenisch, Polycomb complexes repress developmental regulators in murine embryonic stem cells, Nature 441 (2006) 349–353.

360

S.W.M. Bruggeman et al. / Biochimica et Biophysica Acta 1790 (2009) 351–360

[33] A.P. Bracken, N. Dietrich, D. Pasini, K.H. Hansen, K. Helin, Genome-wide mapping of Polycomb target genes unravels their roles in cell fate transitions, Genes Dev. 20 (2006) 1123–1136. [34] T.I. Lee, R.G. Jenner, L.A. Boyer, M.G. Guenther, S.S. Levine, R.M. Kumar, B. Chevalier, S.E. Johnstone, M.F. Cole, K. Isono, H. Koseki, T. Fuchikami, K. Abe, H.L. Murray, J.P. Zucker, B. Yuan, G.W. Bell, E. Herbolsheimer, N.M. Hannett, K. Sun, D.T. Odom, A.P. Otte, T.L. Volkert, D.P. Bartel, D.A. Melton, D.K. Gifford, R. Jaenisch, R.A. Young, Control of developmental regulators by Polycomb in human embryonic stem cells, Cell 125 (2006) 301–313. [35] N. Negre, J. Hennetin, L.V. Sun, S. Lavrov, M. Bellis, K.P. White, G. Cavalli, Chromosomal distribution of PcG proteins during Drosophila development, PLoS Biol. 4 (2006) e170. [36] B. Tolhuis, E. de Wit, I. Muijrers, H. Teunissen, W. Talhout, B. van Steensel, M. van Lohuizen, Genome-wide profiling of PRC1 and PRC2 Polycomb chromatin binding in Drosophila melanogaster, Nat. Genet. 38 (2006) 694–699. [37] N.M. van der Lugt, M. Alkema, A. Berns, J. Deschamps, The Polycomb-group homolog Bmi-1 is a regulator of murine Hox gene expression, Mech. Dev. 58 (1996) 153–164. [38] D. Douglas, J.H. Hsu, L. Hung, A. Cooper, D. Abdueva, J. van Doorninck, G. Peng, H. Shimada, T.J. Triche, E.R. Lawlor, BMI-1 promotes ewing sarcoma tumorigenicity independent of CDKN2A repression, Cancer Res. 68 (2008) 6507–6515. [39] G.M. Bressan, D. Daga-Gordini, A. Colombatti, I. Castellani, V. Marigo, D. Volpin,

[40] [41]

[42]

[43] [44]

[45] [46]

Emilin, a component of elastic fibers preferentially located at the elastinmicrofibrils interface, J. Cell Biol. 121 (1993) 201–212. A. Colombatti, R. Doliana, S. Bot, A. Canton, M. Mongiat, G. Mungiguerra, S. ParonCilli, P. Spessotto, The EMILIN protein family, Matrix Biol. 19 (2000) 289–301. P. Spessotto, M. Cervi, M.T. Mucignat, G. Mungiguerra, I. Sartoretto, R. Doliana, A. Colombatti, beta 1 Integrin-dependent cell adhesion to EMILIN-1 is mediated by the gC1q domain, J. Biol. Chem. 278 (2003) 6160–6167. D. Wiederschain, L. Chen, B. Johnson, K. Bettano, D. Jackson, J. Taraszka, Y.K. Wang, M.D. Jones, M. Morrissey, J. Deeds, R. Mosher, P. Fordjour, C. Lengauer, J.D. Benson, Contribution of polycomb homologues Bmi-1 and Mel-18 to medulloblastoma pathogenesis, Mol. Cell. Biol. 27 (2007) 4968–4979. M.E. Hatten, New directions in neuronal migration, Science 297 (2002) 1660–1663. J. Heino, R.A. Ignotz, M.E. Hemler, C. Crouse, J. Massague, Regulation of cell adhesion receptors by transforming growth factor-beta. Concomitant regulation of integrins that share a common beta 1 subunit, J. Biol. Chem. 264 (1989) 380–388. F. Verrecchia, A. Mauviel, D. Farge, Transforming growth factor-beta signaling through the Smad proteins: role in systemic sclerosis, Autoimmun. Rev. 5 (2006) 563–569. P. van der Stoop, E.A. Boutsma, D. Hulsman, S. Noback, M. Heimerikx, R.M. Kerkhoven, J.W. Voncken, L.F. Wessels, M. van Lohuizen, Ubiquitin E3 ligase Ring1b/Rnf2 of polycomb repressive complex 1 contributes to stable maintenance of mouse embryonic stem cells, PLoS ONE 3 (2008) e2235.