Prenatal overexpression of platelet‐derived growth factor receptor A results in central nervous system hypomyelination

Abstract Background Platelet‐derived growth factor (PDGF) signaling, through the ligand PDGF‐A and its receptor PDGFRA, is important for the growth and maintenance of oligodendrocyte progenitor cells (OPCs) in the central nervous system (CNS). PDGFRA signaling is downregulated prior to OPC differentiation into mature myelinating oligodendrocytes. By contrast, PDGFRA is often genetically amplified or mutated in many types of gliomas, including diffuse midline glioma (DMG) where OPCs are considered the most likely cell‐of‐origin. The cellular and molecular changes that occur in OPCs in response to unregulated PDGFRA expression, however, are not known. Methods Here, we created a conditional knock‐in (KI) mouse that overexpresses wild type (WT) human PDGFRA (hPDGFRA) in prenatal Olig2‐expressing progenitors, and examined in vivo cellular and molecular consequences. Results The KI mice exhibited stunted growth, ataxia, and a severe loss of myelination in the brain and spinal cord. When combined with the loss of p53, a tumor suppressor gene whose activity is decreased in DMG, the KI mice failed to develop tumors but still exhibited hypomyelination. RNA‐sequencing analysis revealed decreased myelination gene signatures, indicating a defect in oligodendroglial development. Mice overexpressing PDGFRA in prenatal GFAP‐expressing progenitors, which give rise to a broader lineage of cells than Olig2‐progenitors, also developed myelination defects. Conclusion Our results suggest that embryonic overexpression of hPDGFRA in Olig2‐ or GFAP‐progenitors is deleterious to OPC development and leads to CNS hypomyelination.

The transcription factor Nkx2.2 directly binds the promoter region of PDGFRA and inhibits its expression during OPC differentiation (Zhu et al., 2014). MicroRNAs miR-219 and miR-338 are upregulated during OPC differentiation and repress PDGFRA transcripts by binding the 3′ untranslated region (Dugas et al., 2010;Zhao et al., 2010). Progressive loss of the methyltransferase PRMT5 facilitates ubiquitination and degradation of membrane PDGFRA thereby promoting OPC differentiation (Calabretta et al., 2018).
PDGFRA signaling, on the other hand, is frequently elevated in adult and pediatric gliomas (Mackay et al., 2017), including the rare and universally fatal diffuse midline glioma (DMG), which is predominantly found in pediatric patients (Buczkowicz & Hawkins, 2015). Amplifications of the wild type (WT) PDGFRA gene locus have been found in approximately 12% of adult and 8-39% of pediatric cases, and activating mutations in approximately 13% of adult and 4-9% of pediatric cases, respectively (Brennan et al., 2013;Buczkowicz & Hawkins, 2015;Paugh et al., 2011;Puget et al., 2012). Moreover, emerging evidence suggests that neonatal OPCs are a likely cell-of-origin for DMG (Filbin et al., 2018;Lindquist et al., 2016;Nagaraja et al., 2017). Despite the importance of PDGFRA signaling for OPC development, and its dysregulation in gliomas, the cellular and molecular consequences of amplified PDGFRA signaling in OPCs are unknown.

Generation of transgenic mice
The hPDGFRA KI mice were generated using the Rosa-CAG-LSL-  Figure 1a). The linearized vector was inserted between exons 1 and 2 of the Gt(ROSA)26Sor locus via electroporation of (129 × 1/SvJ x 129S1/Sv)F1-Kitl+-derived R1 embryonic stem cells (Nagy et al., 1993). Homologous recombination events were detected by PCR, and correct rearrangement of the locus was verified by Southern blot analysis. Correctly targeted stem cells were injected into recipient blastocysts following standard procedures (Nagy, 2003

In vitro studies
RCAS viruses were generated by transfecting retroviral plasmids into DF1 chicken fibroblasts (ATCC) as previously described (Barton et al., 2013 (Cordero et al., 2017). Seventy-two hours after plating, cells were serum-starved overnight and treated with 100 nM purified PDGF-AA ligand (Sigma) for 5 min, trypsinized, and collected by centrifugation at 1500× g for 5 min for western blot analysis.

RNA sequencing (RNA-seq) and analysis
For RNA-seq, BS was isolated from P18/P19 mice from Olig2-Cre; p53 fl/fl ; hPDGFRA fl/+ and age-matched Olig2-Cre; p53 fl/fl control mice and snap frozen. RNA was extracted using RNeasy Mini Kit (QIA-GEN) and quantified using Cytation5 Image Reader (BIOTEK). 50 ng of RNA was submitted to the Northwestern University Sequencing Core facility for analysis. For bioinformatics analysis, paired-end fastq files were imported into Galaxy (Afgan et al., 2016), aligned to the mm10 genome using RNA-STAR, and aligned reads were counted using HTSeq-count with the Ensembl mm10 transcriptome GTF file as the feature file. The following HTSeq-count parameters were used: stranded = no, mode = union, minimum alignment quality = 10, map nonunique or ambiguous reads = none. HTSeq-count files were imported into R (https://www.r-project.org/https://www. r-project.org/), genes with < 10 reads base mean were removed, and differential expression analysis was performed with the DESeq2 package (Love et al., 2014) using the DESeqDataSetFromHTSeqCount function with default settings. DESeq2 analyses were run comparing hPDGFRA KI and control mice. Gene set enrichment analysis (GSEA) was run using genes ranked according to the Wald statistic with the following parameters: permutations = 1000, enrichment statistic = classic, max size = 500, min size = 20, normalization mode = meandiv. For the analysis in Table 2, custom GSEA gene lists were generated from , by taking the top 500 genes positively expressed in astrocytes versus neurons, OPCs, microglia, newly formed OLs, myelinating OLs, or endothelial cells. Genes with minimum FPKM (fragments per kilobase per million reads) less than 5 in astrocytes were excluded.
GSEA was run using the parameters described above to compare control and KI mice based on the custom gene lists. RNAseq data has been submitted to GEO (GSE181899) and are publicly-available.

Experimental design and statistical analysis
Three to eleven mice of each genotype were used per experiment, with roughly equal numbers of mutant and control mice, and males and females. IHC data was quantified by analyzing one whole BS section and six SC sections spanning C4 to T6 per mouse. Unpaired, two-tailed Student t-test was used to determine statistical significance between experimental and control groups using GraphPad Prism. Kaplan-Meier survival curves were analyzed using log-rank (Mantel-Cox) test. p < .05 was considered statistically significant. The n and p values for each experiment are stated in the text and figure legends.

Generation and validation of KI mice overexpressing hPDGFRA in prenatal Olig2-progenitors
We targeted hPDGFRA to prenatal OPCs using Olig2-Cre mice. We inserted the WT hPDGFRA cDNA into the Rosa26 locus, allowing for the stable expression of a single transgene (Bouabe & Okkenhaug, 2013). A stop codon flanked by loxP sites is an immediate upstream of hPDGFRA, which is under the control of the CAG promoter (Figure 1a). The Rosa26 locus lacks the endogenous hPDGFRA regulatory elements, thus preventing its transcriptional (Zhu et al., 2014) and translational (Dugas et al., 2010;Zhao et al., 2010)  ± 0.3651 days, n = 10, p = .0001, log-rank (Mantel-Cox) test). All the hPDGFRA KI mice (10/10) exhibited ataxia, and hindlimb and tail tremors by P18 or P19, and all the mice that survived until P21 (5/10) also exhibited hindlimb paralysis. Examination of body weights revealed stunted growth in P18 and P19 KI mice when compared with control ( Figure 1c, control, 6.8523 ± 0.4576 g, n = 13, KI, 5.4088 ± 0.4024 g, n = 8, p = .0431, two-tailed unpaired Student t-test). The brain weights, however, were not significantly different (Figure 1d We verified that PDGFRA is indeed overexpressed in the hPDGFRA KI mice using IHC and western blot analyses. We found increased PDGFRA expression in P18 KI mouse brain when compared to control

PDGFRA overexpression in prenatal Olig2-progenitors results in CNS hypomyelination
The tremors and hindlimb paralysis exhibited by hPDGFRA KI mice are reminiscent of symptoms commonly seen in mouse models of  (Calver et al., 1998;Dugas et al., 2010;Fruttiger et al., 1999;Suzuki et al., 2012;Zhu et al., 2014). Therefore, we investigated whether hPDGFRA KI mice exhibit myelin deficiencies. IHC analysis revealed a drastic loss of MBP expression in the brain (Figure 3a

PDGFRA overexpression decreases oligodendroglial and myelination gene signatures
To identify the molecular changes accompanying hypomyelination in mice overexpressing hPDGFRA, we performed RNA-seq analysis of BS tissue derived from P19 Olig2-Cre; p53 fl/fl ; hPDGFRA fl/+ and Olig2-Cre; p53 fl/fl mice. Using a false discovery rate (FDR) of ≤0.05, we identified >5000 significantly differentially expressed genes in the hPDGFRA KI mice, >2600 of which were significantly downregulated. Importantly, several OL lineage markers were downregulated in the KI mice as shown in Table 1, consistent with the myelination defects described in Figure (Table 1). Since astrocytic markers are few and not highly specific to astrocytes, we performed GSEA using gene lists comparing astrocytes with other cell types in the brain . This analysis revealed positive enrichment of astrocytic signatures relative to several other cell types in the KI mice (Table 2). We also observed approximately a three-fold increase in PDGFRA transcript levels in the KI mice compared to control (KI, 8337, n = 4, control, 2907, n = 4, p = .0004, FDR adjusted). We also aligned the FASTQ

Hypomyelination in hPDGFRA KI mice is independent of p53 loss
The above studies were carried out in hPDGFRA KI and control mice that also lack expression of the tumor suppressor p53 because of our initial investigations of glioma formation in these mice. We were interested in examining whether p53 loss was necessary for the development of myelination defects. We generated Olig2-Cre; hPDGFRA fl/+ (KI) mice and compared its phenotype with control hPDGFRA fl/+ mice, both in p53 WT backgrounds. The KI mice exhibited ataxia, tremors, and hindlimb paralysis and were euthanized by P21 (Figure 6a, control, mean survival undefined, n = 9, KI, 20 ± 0.6071, n = 11, p = .0010, log-rank (Mantel-Cox) test). The KI mice were significantly stunted in growth (control, 9.1933± 0.4735, n = 6, KI, 7.1738± 0.2230, though their brains appeared slightly larger than control (control, 0.4314 ± 0.0104, n = 6, KI, 0.5088 ± 0.0041, n = 8, p = .0011, two-tailed unpaired Student t-test). We confirmed increased expression of PDGFRA in the BS ( Figure 6b) and SC (Figure 6c) of P20-P21 KI mice relative to control. No gross abnormalities were detected in the brain tissue of KI mice, using H&E staining (not shown), but hemorrhaging was seen in SC tissue, similar to the phenotype observed in   Table 3). This is similar to the loss of myelination markers seen in Olig2-Cre; p53 fl/fl ; hPDGFRA fl/+ mice (Figures 3   and 5). MBP and CNPase expression levels were not significantly different between control and KI at P0, but progressively declined with postnatal development in the KI, and were significantly lower than the control at P19-P21 in both the BS and the SC (Table 3). We also observed a similar progressive reduction in Olig2-expressing cells in the KI BS and SC versus control (Figure 7a,b and Table 3). Thus, hPDGFRA overexpression in prenatal Olig2-expressing progenitors results in hypomyelination in the brain and SC, with or without p53 loss.  Table 3 3.

PDGFRA overexpression in prenatal GFAP progenitors leads to hypomyelination
To investigate whether overexpression of hPDGFRA in a wider compartment of cells than OPCs gives rise to hypomyelination, we crossed hPDGFRA fl/fl mice with GFAP-Cre mice. GFAP-progenitors are not specific to OPCs, instead give rise to a more diverse lineage of cells including neurons and glia (Semerci & Maletic-Savatic, 2016).

DISCUSSION
PDGF signaling promotes OPC maintenance and proliferation, but inhibits OPC differentiation into myelinating OLs (Butt, Hornby, Ibrahim, et al., 1997;Ellison & de Vellis, 1994;Hall et al., 1996;Hart et al., 1989;Pringle et al., 1989;Richardson et al., 1988). PDGFRA gene copy number variations and amplifying mutations are found in 4-39% of pediatric DMG, whose cell-of-origin is thought to be OPCs (Filbin et al., 2018;Lindquist et al., 2016;Nagaraja et al., 2017). However, the in vivo consequences of increased PDGFRA expression in OPCs has not been described before. In this study, we analyzed the consequences of overexpression of WT hPDGFRA in prenatal Olig2-expressing OPCs. The mutant mice exhibited severe hypomyelination in the brain and SC. Overexpressing hPDGFRA more globally using GFAP-Cre also resulted in CNS myelination defects. hPDGFRA overexpression with or without p53 loss, however, did not result in tumors. Our results suggest that prenatal hPDGFRA overexpression in glial progenitors leads to defective oligodendroglial development and hypomyelination in the CNS.
We overexpressed hPDGFRA in embryonic Olig2-and the more ubiquitously expressed GFAP-progenitors (Zhuo et al., 2001) using the Rosa26 locus that lacks the endogenous regulatory elements of hPDGFRA. We show that the hPDGFRA KI mice exhibit severe hypomyelination in the CNS (Figures 3, 7, and 8), which explains the stunted growth, ataxia, and early mortality seen in the mutant mice. We observed 100% mortality in the hPDGFRA KI mice driven by Olig2-Cre or GFAP-Cre by weaning age (Figures 1, 6, and 8). Importantly, total PDGFRA levels were significantly higher in the mutant brain and SC  Table 1). Consistent with that observation, we found decreased expression of the mature oligodendroglial markers MBP, CNPase, and ASPA in the hPDGFRA KI mouse brain and SC (Figures 3,   7, 8, and Table 3). Defects in myelination were also apparent in LFB stained images of the brain (Figures 3 and 8). Importantly, expression of Olig2 and Sox10, markers of OPCs, appeared normal at P0 and P10 but were significantly reduced by weaning age in the KI brain and SC suggesting a defect in OPC development (Figures 5,7,and Table 3).
The reduction in OPCs could be either due to increased apoptosis as reported for mice overexpressing PDGF-A (Calver et al., 1998) and/or due to differentiation into other lineages such as astrocytes, which is supported by our RNA-seq data (Tables 1 and 2). In fact, loss of Olig2 in NG2+ OPCs has been shown to increase differentiation into astrocytes at the expense of OLs (Zuo et al., 2018), and Olig2 deletion converts gliomas from proneural to astrocytic signatures (Lu et al., 2016).
Myelination defects were seen in the hPDGFRA KI mice both in the presence and the absence of p53, indicating that the mutational status of p53 does not impact hypomyelination.
Our results are consistent with previous reports suggesting that PDGFRA downregulation is important for OPC differentiation. Conditional deletion of PDGFRA in Olig1+ OPCs leads to decreased OPC proliferation but premature OPC differentiation resulting in hypomyelination, consistent with PDGFRA's dual role in promoting OPC proliferation but inhibiting OPC differentiation (Zhu et al., 2014).
Overexpressing the PDGFRA inhibitory transcription factor Nkx2.2 also results in precocious OPC differentiation (Zhu et al., 2014). On the other hand, disrupting the generation of PDGFRA inhibitory factors, such as microRNAs, affects OPC differentiation and myelination (Dugas et al., 2010). Mice that overexpress the PDGF-A ligand exhibit a transient increase in the embryonic OPC population that returns to baseline levels in postnatal mice (Calver et al., 1998). Both the hPDGFRA KI mice, described here, and PDGF-A KO mice (Fruttiger et al., 1999) exhibit hypomyelination, which suggests a developmental balancing act where too little or too much of PDGF-A/PDGFRA signaling is deleterious to OL development. Of note, hypomyelination was not reported in mice overexpressing PDGF-A (Calver et al., 1998).
The authors reported increased OLs in late embryonic and neonatal mutant mice when compared to control, due to increased OPC proliferation, but the OLs return to baseline levels by P6. The lack of myelination defects with PDGF-A overexpression could be because PDGFRA downregulation and, therefore, normal OPC differentiation into OLs, remains unaffected in these mice. The authors also did not look past P6. In our study, MBP and CNPase levels are not significantly different between control and KI mice at P0 or P10, but only at P19 (Table 3). In addition, unlike the PDGF-A overexpressing mice that show increased OPCs (PDGFRA+) at embryonic and neonatal stages relative to control, our PDGFRA KI mice do not show increased OPCs (Olig2+) at P0, possibly because PDGF-A levels are limiting (Calver et al., 1998) and excess PDGFRA does not result in increased OPC proliferation.
PDGFRA expression in the developing mouse CNS is high until approximately P14, corresponding temporally with proliferation and migration of glial progenitors, and then decreases in the adult brain (Yeh et al., 1993). Thus, the progressive myelination defects seen in hPDGFRA KI mice (  (Furusho et al., 2012). Further, the continued expression of myelination regulatory genes is important for maintaining sheath integrity post-initiation, and myelin regulatory factor ablation in mature OLs results in delayed CNS myelination (Koenning et al., 2012). The exact mechanism by which PDGFRA overexpression impacts OL development, including effects on OPC proliferation, differentiation, apoptosis, and myelin sheath maintenance, and PDGFRA's interplay with other growth factors and myelin regulatory genes will be evaluated in future studies.
A surprising finding of our study is that PDGFRA overexpression in embryonic glial progenitors, with or without p53 loss, did not result in tumors. Possible reasons for the lack of gliomas are discussed below. First, Olig2-Cre and GFAP-Cre transgenes are activated early in embryonic development (∼ E13.5) Zhuo et al., 2001). Sustained high levels of PDGFRA expression driven by these transgenes, in multiple cell types, could induce neurodevelopmental perturbations preventing gliomagenesis in our models. Second, PDGFRA amplification may not be a precipitating event in tumorigenesis, supported by the observation that activating PDGFRA mutations but not WT PDGFRA initiate gliomas in the context of p53 loss in mice (Paugh et al., 2013). The tumorigenic impact of WT PDGFRA may also be context-dependent as PDGFRA overexpression with Ink4a/Arf tumor suppressor deletion promotes gliomagenesis (Liu et al., 2011).
PDGF-A and PDGF-B, although rarely amplified in the human disease (Paugh et al., 2010;Puget et al., 2012;Zarghooni et al., 2010), consistently induce gliomas in mouse models (Hambardzumyan et al., 2009;Misuraca et al., 2015;Ozawa et al., 2014), even with no accompanying oncogenic mutations (Becher et al., 2010). This discrepancy could be due to the non-cell autonomous activation of PDGF signaling in endothelial or stromal cells by PDGF ligands, and/or differences in the timing of pathway activation (prenatal vs. neonatal mice). Third, if PDGFRA alterations are a late event in human tumors, they may require earlier founding events to exert tumorigenic effects. Thus, further investigations are needed to determine the potential tumorigenic impact of WT PDGFRA in the context of alternative genetic alterations, cells-of-origin, and/or timing of genetic perturbations.
A potential caveat of the Olig2-Cre driven hPDGFRA KI mouse model is that Olig2-progenitors in the ventral neuroepithelium first give rise to motor neurons (MNs) (∼ E10-E13), followed by OPCs (Fu et al., 2002;Takebayashi et al., 2002;Wu et al., 2006), raising the possibility that the hypomyelination in the KI mice is a result of MN abnormalities. We consider this possibility unlikely, given that inspection of H&E stained images did not show gross abnormalities in the MN distribution in the SC. Due to the ubiquity of transgene expression with GFAP-Cre, and the fact that Olig2-progenitors also give rise to MNs, future studies could utilize PDGFRA-Cre mice to overexpress the PDGFRA transgene specifically in OPCs (Fruttiger et al., 1999), and match the spatiotemporal expression patterns of endogenous PDGFRA. It is also possible that prenatal overexpression of PDGFRA, regardless of the promoter used, is deleterious for neurodevelopment.
Therefore, an inducible system such as CreER, or CNPase-Cre, could help activate PDGFRA later in development to examine myelin dysregulation.