Isolation and characterization of mammalian cells expressing the Arf promoter during eye development

Abstract

Although many researchers have successfully uncovered novel functions of the tumor suppressor p19^Arf utilizing various types of cultured cancer cells and immortalized fibroblasts, these systems do not accurately reflect the endogenous environment in which Arf is developmentally expressed. We addressed this by isolating perivascular cells from the primary vitreous of the mouse eye. These cells represent a rare cell type that normally expresses the p19^Arf tumor suppressor in a non-pathological, developmental context. We utilized fluorescence activated cell sorting to purify the cells by virtue of a GFP reporter driven by the native Arf promoter, and characterized their morphology and gene expression pattern. We further examined the effects of reintroduction of Arf in the PVCs to verify expected downstream effectors of p19^Arf as well as uncover novel functions as a regulator of vasculogenesis. This methodology and cell culture model should serve as a useful tool to examine p19^Arf biology.

Introduction

Over the last two decades, our understanding of the regulation of the tumor suppressor, p19^Arf, has grown, as has our understanding of its tumor suppression, primarily by controlling p53. The original model supported the notion that p19^Arf induction antagonizes Mdm2, the negative regulator of p53; the stabilized p53 promotes the progression of a transcriptional program that, ultimately, arrests cell proliferation, facilitates DNA damage responses, or promotes apoptosis (1-4). Ample evidence indicates that p19^Arf also acts independently of p53. This was first suggested by the differences in the types and latency of spontaneous tumors in animals lacking either p53 or Arf and in mice lacking both (5,6). Established p53-independent functions of p19^Arf include regulation of rRNA biogenesis via its interaction with NPM/B23 (7,8); inhibition of transactivation by c-Myc (9,10) and E2F1 (11); sumoylation of Mdm2, NPM and others (12,13); as well as inhibition of NF_κB (14). While these studies have been vital for ascribing functions to the tumor suppressor, their significance is sometimes subject to question because the work was not accomplished in cells that normally express Arf. Indeed, one of the clearest in vivo examples of a p53-independent role for p19^Arf resides in its capacity to block Pdgfrβ expression and proliferation of perivascular cells of the developing eye to foster involution of the underlying hyaloid vessels (15-18).

Despite wide-spread importance of perivascular cells to support vascular integrity, robust Arf expression is only observed in the perivascular cells flanking the hyaloid vessels and the internal umbilical artery (19), both of which represent vascular beds that are not necessary beyond embryo development. In the absence of p19^Arf, Pdgfrβ accumulates in the perivascular cells and leads to hyperplasia of cells in the vitreous space. This results in a retrolental mass and causes catastrophic secondary effects on the lens and retina, leaving the animals blind (17,20). Even though it is clear that the developmental function of p19^Arf is imperative to the animal, little is known about the particular perivascular cells that normally express the tumor suppressor gene.

In order to better understand p19^Arf function during development, we isolated cells that normally activate the Arf promoter from the vitreous compartment of the eye. We were able to purify cells endogenously expressing the Arf promoter by fluorescence activated cell sorting (FACS) and examine them in culture. In this report, we describe the isolation and in vitro culture of Arf expressing primary vitreous cells (PVCs) and compare them to previously established cell culture models for studying Arf function. By global transcriptome analysis, we gain clearer insight into the identity of the PVCs and further demonstrate the utility of these cells by examining expected and novel molecular changes upon reintroduction of Arf. The availability and use of the Arf^GFP/GFP PVCs holds great potential for better understanding the role of p19^Arf in mammalian development and how these functions are abrogated in human ocular and cardiovascular disease as well as tumorigenesis.

Materials and Methods

Animals

Mice in which Arf exon 1β is replaced by a reporter gene encoding green fluorescence protein (GFP) (21) were maintained in a mixed C57BL/6 × 129/Sv genetic background. Primary MEFs from Arf^lacZ/lacZ mice were derived as previously described (17). Animal studies were accomplished at the University of Texas Southwestern Medical Center, with approval of the animal care and use committees.

Eyes were isolated from Arf^GFP/GFP mice, euthanized and decapitated at postnatal days (P) 0 - 4. The eyelid was incised using a No. 11, straight surgical blade (Feather Safety Razor Co.) to expose the eye. While holding the eyelid open, the scalpel blade was used to transect extra-ocular muscles and other connective tissue between the globe and the bony orbit. Small angled forceps (Fine Science Tools) were inserted between the orbit and globe, grasping the optic nerve/ophthalmic vessels firmly and gently lifting out the intact eye. Enucleated eyes were submerged in ice-cold PBS and stabilized under PBS by holding the optic nerve stub. Small spring scissors (Fine Science Tools) were used to cut along the circumference of the eye at the equator. The cornea/anterior part of the sclera were lifted off, leaving the optic cup and lens together under PBS. The retina was then removed in piecemeal fashion, leaving the lens with attached retrolental mass.

Cell Isolation and Culture

The lens/retrolental mass tissue from 60 individual eyes were pooled in a 1.5 mL microcentrifuge tube and digested in M2 media with 300 μg/mL hyaluronidase and 1 mg/mL collagenase (all from Sigma-Aldrich) at 37° C for 15 minutes. The tissue was briefly triturated and further incubated at 37° C for 10 minutes. Digested material (including undigested lenses) was centrifuged, washed with D-MEM with 20% FBS, and then resuspended in D-MEM/20% FBS with penicillin/streptomycin. Resuspended cells (including PVCs) were passed through a 35 μM filter into polystyrene tube for FACS. GFP-positive PVCs were collected using the MoFlo (Dakocytomation) cell sorter. Sorted PVCs were plated (6,000 cells/well) in a 96-well plate with Pericyte Medium (PM) (ScienCell). Cells were passed (1:4) using trypsin/EDTA every 3 days.

Arf^lacZ/lacZ MEFs (19), 10T1/2 fibroblasts and PVCs were used for RNA-Seq analysis. Briefly, cells were plated at a density of 1 × 10⁶ cells/ 10 cm plate and cultivated in PM until ∼80% confluence, at which point cells were harvested for RNA extraction.

Whole Transcriptome Sequencing (RNA-Seq)

Total RNA was isolated using the miRNeasy mini kit (Qiagen) and treated with RNase free DNaseI to remove genomic DNA (Qiagen). RNA integrity and purity was determined using the Bioanalyser Pico Chip (Agilent), assuring that each sample had a RIN score of 10. RNA (1 μg) from two biological replicates of each cell type was fragmented in the UT Southwestern Next-Generation Sequencing core, converted to cDNA, and amplified by PCR according to the Illumina RNA-Seq protocol (Illumina, Inc. San Diego, CA). The Illumina HiSeq 2000 (San Diego, CA) instrument was used to generate 50 bp single-end sequence reads. RNA-Seq read quality was evaluated in the core using the Illumina purity filter and distribution of base quality scores at each cycle.

Sequence reads for each sample were aligned to the UCSC mm10 version of the mouse reference genome assembly using Bowtie 2.1.0 (22) and the splicing-aware aligner TopHat 2.0.8 (23). The alignment allows only uniquely aligned reads and up to two mismatches per read. All other parameters were set to the default values. The quality of the RNA-Seq data was evaluated by FastQC (v0.10.1) and a series of Perl (v5.16.1) and R (v3.0.1) scripts. Normalized gene expression values expressed as fragments per kilobase of exon per million fragments mapped (FPKM) were determined using Cufflinks 2.0.2 (24) with default settings, which reports the mean of the maximum likelihood estimates from each of three replicates processed independently.

Western blotting

Protein expression was examined by Western-blotting according to a standard procedure. The following antibodies were used: anti-p19^Arf (Ab80, Abcam, 1:1000), anti-p21 (Sc-756, Santa Cruz, 1:1000), anti-p53 (Sc-6243, Santa Cruz, 1:1000), anti-MDM2 (Sc-965, Santa Cruz 1:1000) and anti-Hsc70 (Sc-1059, Santa Cruz, 1:5,000).

Quantitative RT-PCR (qRT-PCR)

Total RNA was extracted from PVCs using the miRNeasy mini kit (Qiagen). For qRT-PCR, 1 μg of total RNA was reverse transcribed using NCode miRNA First-Strand Synthesis (Invitrogen) and KAPA SYBR Green Master Mix (KAPA). qRT-PCR was performed in a 96-well plate using BioRad instrument. The PCR program consisted of 20 sec at 95 C, followed by 40 cycles of 95 C for 15s and 60 C for 20s. Primer quality was analyzed by dissociation curves. The expression of mir34abc and Pdgfrα, Pdgfrβ and αSMA was normalized to U6 and Gapdh, respectively.

Results and Discussion

Currently, there is very little known about the cells that normally express Arf, motivating us to generate a cell culture model that accurately reflects the unique environment in which Arf is expressed developmentally. We decided to pursue this by taking advantage of Arf^GFP/GFP mice in which GFP replaces exon 1β of the endogenous Arf locus, rendering the mice Arf null. In this context, a retrolental mass persists in the primary vitreous space in which GFP (+), Arf-expressing cells, in addition to other cell types including endothelial cells, are present (Figure 1A). It is important to note that the retrolental mass is not evident in WT or GFP/+ animals postnatally, making it unfeasible to derive these cells under “wild type” conditions. We isolated the retrolental tissue from Arf^GFP/GFP mice and retrieved a total of 38,000 GFP-positive cells (averaging 633 cells/eye), which represented 2-3% of the total population (Figure 1B). We collected the Arf^GFP/GFP Primary Vitreous Cells (PVCs) for in vitro culture and expansion. At confluence, the PVCs form a monolayer and adopt a fibroblast-like morphology with some variation in GFP expression (Figure 1C). We continued to expand the cells in culture and observed that GFP expression persists through passage 15 (data not shown). At passage 5 and sub-confluence, the PVCs are elongated and spindle-like with long cytoplasmic processes. The cells continue to express GFP while WT MEF cells cultured in tandem do not (Figure 1D). Although many laboratories, including our own, have successfully utilized classically immortalized fibroblasts and cancer cell models to gain valuable insight into some of the developmental and tumor suppressive functions of Arf, these systems are imperfect in recapitulating the non-pathological environment in which Arf is expressed. The PVCs represent, for the first time, a cell culture model in which the Arf promoter is endogenously turned on during development allowing us and other researchers to explore the capacity of p19^Arf to control vascular remodeling and mural cell proliferation in the context of a cell that normally expresses this promoter. These cells will also be useful in clarifying the complex regulation of the Arf promoter.

Figure 1. Isolation and expansion of Arf^GFP/GFP PVCs.

(A) Representative photomicrograph of enucleated mouse eyes from P0-P4 Arf^GFP/GFP mice. Phase contrast image (left), GFP (right). The retrolental mass (*) is behind the lens (L) and expresses GFP. (R= retina) (B) Purification of PVCs by fluorescence activated cell sorting (FACS) for GFP. Representative images of cultured PVCs at passage 0 (C) and passage 5 (D). WT MEF (wild type mouse embryonic fibroblasts) cells at passage 5 (D). Phase contrast image (left), fluorescence image showing GFP expression (right). Arrows indicate heterogeneous GFP expression between high expressing (black) and low expressing (white) cells.

Because we have not unequivocally established the identity of the Arf-expressing cells, we sought to capture the global gene expression profile of the Arf^GFP/GFP PVCs in comparison to other cell culture models that have been previously used to study Arf biology. Like the PVCs, Arf^lacZ/lacZ MEFs do not express a functional p19^Arf protein, while 10T1/2 cells, a widely used pericyte model, carry a biallelic deletion of Arf (19,21)(and unpublished data). When cultured in Pericyte Medium, all three cell lines resemble fibroblasts in their morphology (Figure 2A). To define these cells by gene expression, we cultured the Arf^GFP/GFP PVCs, Arf^lacZ/lacZ MEFs and 10T1/2 cells, extracted total RNA and performed high-throughput RNA-sequencing. We observed 85.3, 86.7 and 83.8 million sequence reads for Arf^lacZ/lacZ MEFs, 10T1/2 and PVCs, respectively; after applying a series of computational tools (see Methods), we were left with 71.6, 81.9 and 73.5 million reads that were successfully mapped to the mouse reference genome. We examined how all 10,704 genes expressed in PVCs partitioned between Arf^lacZ/lacZ MEFs and 10T1/2 cells. We found 970 expressed genes in PVCs that were also expressed in either Arf^lacZ/lacZ MEFs or 10T1/2 cells; 769 genes were expressed in Arf^lacZ/lacZ MEFs, while 201 were found in 10T1/2 cells (binomial test, P= 2.8×10^-79) (Figure 2B). Based on this analysis, we conclude that on a genome-wide scale, the Arf^GFP/GFP PVCs are more closely related to the Arf^lacZ/lacZ MEFs than the 10T1/2 cells.

Figure 2. PVCs express perivascular genes.

(A) Representative phase contrast image of 10T1/2, Arf^lacZ/lacZ MEFs and Arf^GFP/GFP PVCs in culture. (B) Venn diagram comparison of all genes expressed in the three cell lines showing unique and overlapping transcripts. (C) Expression of a subset of Tgf-beta pathway genes (FPKM) (D) Expression of cell type specific markers demonstrates that the PVCs express perivascular and fibroblast genes, not endothelial or retinal genes.

The fact that the MEFs and the PVCs share 86% similarity in gene expression reaffirmed our previous studies utilizing Arf^lacZ/lacZ MEFs to establish a pathway beginning from Arf induction to characterization of its downstream effects required for eye development. We have previously established that Tgfβ signaling drives Arf expression in Arf^lacZ/lacZ MEFS and in vivo resulting in p19^Arf mediated down-regulation of Pdgfrβ (19,25). In order to understand if the Arf^GFP/GFP PVCs were similar to the Arf^lacZ/lacZ MEFs in this regard, we looked for the expression of all Tgf-beta pathway genes as defined by KEGG and performed a hierarchical clustering (26,27). We found the components of the Tgfβ pathway that we have so far defined as important for Arf regulation, including Smad2/3, Sp1 and Cebpβ to be expressed in all three cell lines (Figure 2B) (28). Further, based on all Tgfβ pathway genes, we found that the PVCs clustered more closely to the Arf^lacZ/lacZ MEFs than the 10T1/2 cells (supplemental).

To explicitly establish our previous finding that the Arf-expressing cells of the primary vitreous are perivascular, we examined the expression of known markers that identify vascular/mural cells, as well as fibroblasts, endothelial cells and retinal cells. As we have previously shown, the Arf^GFP/GFP PVCs express the transmembrane cell surface protein Pdgfrβ (17,18,29). We also observed expression of, Angpt1, which is critical for angiogenesis and vessel maturation as well as α-SMA, a bona fide perivascular cell marker (30). Vimentin, a cytoskeletal component associated with mesenchymal cells, was also highly expressed in all three cell lines, further establishing that these cells are perivascular (Figure 2D) (31). We found several markers of fibroblasts such as S100a4, Col1a1 and Ph4b to be expressed in all three cell types, while Fap, a marker of differentiated fibroblasts, was only present in the Arf^lacZ/lacZ MEFs (Figure 2D) (32). Further, we observed the lack of expression of endothelial specific genes, Pecam, Cdh5 and vWf, demonstrating that Arf is not expressed in the endothelial cell population that coexists with the PVCs in the developing eye (Figure 2D) (33). Finally, to ensure that we did not contaminate our cell prep with neighboring retinal tissue, we checked for the expression of known retinal defining transcription factors: Six3, Otx2, Nr2e3, Nrl and Crx (34). None of these transcription factors were expressed in either cell line (Figure 2D). Based on this gene expression signature and the morphology of the cells, we assert that our Arf^GFP/GFP PVCs are perivascular cells.

While we observed that 86% of all the genes expressed in the Arf^GFP/GFP PVCs were also expressed in the Arf^lacZ/lacZ MEFs, we found 323 genes representing 3% of the total genes expressed in PVCs to be mutually exclusive from those genes expressed in MEFs and 10T1/2 cells (Figure 2B). To understand how the PVCs are distinct from the other cultured cells, we subjected the dissimilar set of genes to Gene Ontology (GO) pathway analysis. KEGG pathway and GO terms were collected from the Molecular Signatures Database (35). We highlighted several pathways that were enriched in the Arf^GFP/GFP PVCs (p > 0.05) (Figure 3A, extensive list supplemental). Given their neural crest origin, it was not unexpected that we observed that the PVC only genes were enriched for the term Nervous System Development (25). Of interest, we found that the term Anatomical Structure Morphogenesis was enriched with 9 genes expressed in the Arf^GFP/GFP PVCs (-log p value = 2.84) including Pax6 and Eya2, both of which are important for eye development (Figure 3A) (36). Only the Arf^GFP/GFP PVCs were significantly enriched for genes in the Tgfβ Receptor Signaling Pathway, including Gdf15, Eng and Lefty1; perhaps suggesting that the PVCs more aptly reflect a signaling environment to endogenously regulate Arf (Figure 3A and supplemental). We also found the terms Cell Proliferation, Cellular Localization and Cell-cell Signaling to be enriched in the PVC only gene set, reflecting the dynamic environment of the vitreous compartment as well as the requirement of p19^Arf to blunt the proliferation of these cells.

Figure 3. Gene set enrichment analysis of differentially expressed PVC genes identifies EMT related pathways.

(A) Gene Ontology (GO) term enrichment analysis for the PVC only genes. Indicated are a subset of enriched pathways listed in order of significance (q>0.05). (B) KEGG Pathway analysis of microRNA target genes expressed in PVCs. microRNA target prediction was performed using Diana miRPath v2.0 Top ten pathways are listed in order of significance (q>0.05). (C) EMT marker gene expression in the PVCs (FPKM). Insert shows phase contrast and fluorescence overlay of Wnt1-cre, tdTomato e14.5 mouse eye. PVCs are derived from the neural crest (arrow) while retinal pigment epithelium (RPE) is not (arrowhead, inset).

We recently identified a previously unrecognized role for p19^Arf during development in its capacity to regulate microRNA expression independently of p53 (37). With this in mind, we sought to understand functional pathways targeted by microRNAs expressed in the PVCs. We found of all 10,704 genes expressed, 1.8% or 186 represented small non-coding RNA genes, 80 of which are defined as microRNAs based on annotation mm10 from the UCSC Genome Browser (data not shown)(38). In order to understand the function of the microRNAs, we analyzed all microRNAS expressed in the Arf^GFP/GFP PVCs by DIANA-miRPath v2.0, a web-based server for microRNA target prediction and pathway analysis (39). The most significantly enriched pathway targeted by microRNAs expressed in the Arf^GFP/GFP PVCs was ECM-receptor interaction with 10 microRNAs expressed that target 21 different genes in this pathway (-log p value = 13.8) (Figure 3B). Of interest, Tgf-beta signaling was also enriched in both the Arf^GFP/GFP PVCs and Arf^lacZ/lacZ MEFs (Figure 3B and data not shown). The Arf^GFP/GFP PVCs expressed 12 microRNAs targeting 38 genes within this pathway (-log p value = 1.48) (Figure 3B). Tgf-beta regulated microRNAs are known to target genes that promote angiogenesis and components of the epithelial to mesenchymal transition (EMT) program (40). In this regard, we also found the terms Focal adhesion, Pathways in cancer and Transcriptional misregulation in cancer to have significant enrichment of genes targeted by the repertoire of microRNAs expressed in the PVCs.

In line with the idea that these cells are derived from the neural crest, evidenced by lineage-tracing experiments using Wnt1-Cre, Rosa26-LSL-tdTomato and Wnt1-Cre, Arf^fl/fl mouse models, and thus have undergone EMT, we were prompted to examine the expression of EMT associated genes in the PVCs (Figure 3C) (25). We found high expression of known mesenchymal marker genes (Cdh2) as well as transcription factors such as Twist1, Zeb1/2 and Snai1 that are required for EMT (40). In contrast, the classical epithelial marker Cdh1 (E-cadherin), was not expressed. Because p19^Arf is turned on after the cells have migrated, an intriguing hypothesis posits that p19^Arf expression in these cells negatively regulates the EMT program by inhibiting their proliferation and migration. Our in vitro model of the Arf^GFP/GFP PVCs will be ideal for investigating how p19^Arf controls aspects of the EMT program.

While characterizing these cells is critical, the real utility in this model stems from wanting to understand Arf activity in a cell that normally expresses the gene during development. To address this, we ectopically expressed Arf-RFP by retroviral transduction in the Arf^GFP/GFP PVCs (Figure 4A). Upon ectopic Arf expression, we observed activation of the p53 pathway as detected by expression of downstream target effectors that have been previously described, p53, MDM2 and p21 as compared to the RFP control (Figure 4A) (3). Moreover, we have recently showed that the expression of the miR-34 family is dependent upon Arf status in cultured cells and in vivo. In our analysis, we observed that ectopic Arf expression could upregulate all three members of the miR-34 family, miR-34abc in MEFs triple negative for p53, MDM2 and Arf (TKO MEFs). Furthermore, shRNA knockdown of p19^Arf in p53-/- MEFs, reduced the expression of miR-34abc (37). In the Arf^GFP/GFP PVCs, we observed that ectopic Arf expression induced miR-34a and decreased the expression of miR-34b. miR-34c was not affected by p19^Arf in these cells (Figure 4C). Finally, because we observed that the PVCs had undergone EMT and expressed mesenchymal genes, we became interested in how re-expression of Arf affected vascular gene expression. As we have shown previously, Pdgfrβ is down-regulated in response to overexpression of Arf (17,18). Pdgfrα mRNA was not affected by Arf, while αSMA was significantly decreased upon reintroduction of Arf suggesting that it may play a role in regulating vascular gene expression. These cells will be a useful tool in clarifying how p19^Arf affects vascular smooth muscle biology.

Figure 4. Ectopic p19^Arf expression in PVCs activates the p53 pathway.

(A) Retroviral transduction of PVCs with MSCV-RFP (top panel) or MSCV-Arf (bottom panel). (B) Representative western blot showing induction of p21, p53 and MDM2 upon ectopic p19^Arf expression (C) Quantitative analysis by qRT-PCR of miR-34abc family in PVCs upon retroviral transduction of Arf. microRNA expression is normalized to U6 and represented as relative to RFP control (D) Quantitative analysis by qRT-PCR of Pdgfrα, Pdgfrβ and αSMA in PVCs transduced with RFP or Arf. Expression is relative to RFP control and normalized to Gapdh.

We believe that the PVCs will represent an important model for studying Arf, particularly for studies focused on how the Arf promoter is activated and the functional consequence of p19^Arf expression in these cells. Furthermore, heterotypic interactions between perivascular and endothelial cells help to drive vascular stabilization and remodeling. Indeed, the molecular mechanism by which p19^Arf regulates vascular/mural cell identity and proliferation, as well as the contribution of the human CDKN2A locus and intergenic 9p21 region to cardiovascular disease risk remain unclear. In our model, loss of Arf in perivascular cells seems to derail the developmentally-timed regression of the underlying endothelial cells of the hyaloid vasculature. Given that Arf expression in normal cells is largely limited to perivascular cells embracing two vascular structures that become essentially functionless in the postnatal period, these cells could be particularly valuable tools to study perivascular-endothelial cell interactions. Additionally, by examining microRNA and protein changes that occur when Arf is expressed in these cells, we will be able to better understand the full repertoire of p19^Arf dependent changes that drive vascular involution.

Supplementary Material

SF1

Supplemental Figure 1. Gene expression of all Tgf-beta pathway genes as defined by KEGG in 10T1/2, Arf^lacZ/lacZ MEFs and Arf^GFP/GFP PVCs.

SF2

Supplemental Figure 2. Gene Ontology term enrichment for differentially expressed genes in 10T1/2, Arf^lacZ/lacZ MEFs and Arf^GFP/GFP PVCs. Significance is noted as –log q value.

Method Summary.

Here we describe the isolation of cells that express the Arf promoter from the primary vitreous space of the developing eye. Using mice in which exon 1β of the Arf gene is replaced with Gfp, we purified hyperplastic cells by FACS from the primary vitreous and expanded them in culture. These primary vitreous cells (PVCs) represent the first cell type in which the Arf gene product, p19^Arf, normally acts to block their aberrant accumulation in vivo, and this report is the first description of these cultured cells.