Transcription factor Olig2 is a major downstream effector of histone demethylase Phf8 during oligodendroglial development

The plant homeodomain finger protein Phf8 is a histone demethylase implicated by mutation in mice and humans in neural crest defects and neurodevelopmental disturbances. Considering its widespread expression in cell types of the central nervous system, we set out to determine the role of Phf8 in oligodendroglial cells to clarify whether oligodendroglial defects are a possible contributing factor to Phf8‐dependent neurodevelopmental disorders. Using loss‐ and gain‐of‐function approaches in oligodendroglial cell lines and primary cell cultures, we show that Phf8 promotes the proliferation of rodent oligodendrocyte progenitor cells and impairs their differentiation to oligodendrocytes. Intriguingly, Phf8 has a strong positive impact on Olig2 expression by acting on several regulatory regions of the gene and changing their histone modification profile. Taking the influence of Olig2 levels on oligodendroglial proliferation and differentiation into account, Olig2 likely acts as an important downstream effector of Phf8 in these cells. In line with such an effector function, ectopic Olig2 expression in Phf8‐deficient cells rescues the proliferation defect. Additionally, generation of human oligodendrocytes from induced pluripotent stem cells did not require PHF8 in a system that relies on forced expression of Olig2 during oligodendroglial induction. We conclude that Phf8 may impact nervous system development at least in part through its action in oligodendroglial cells.


| INTRODUCTION
Oligodendrocytes are required for myelin sheath formation, saltatory conduction, and trophic support of axons and represent an essential neural circuit component.As a consequence, defects or altered behavior of oligodendrocytes may contribute to neurodevelopmental disorders such as neuropsychiatric diseases or intellectual disability.
Although neurodevelopmental disorders are traditionally attributed to malfunctions in neurons, the idea of a glial role has gained ground as an important general concept over the last years (de Oliveira Figueiredo et al., 2022;Schmitt et al., 2023;Scuderi & Verkhratsky, 2020).
Interestingly, genes coding for nuclear proteins with functions in transcription, chromatin organization and chromatin modification are overrepresented among the genes associated with neurodevelopmental disorders and intellectual disability (Kleefstra et al., 2014;Kramer & van Bokhoven, 2009).These genes usually exhibit a broad and oftentimes ubiquitous expression so that their mutation is unlikely to be only relevant in neurons.Other cells of the central nervous system (CNS) such as astrocytes and oligodendrocytes may also be affected so that glial dysfunctions may contribute to the resulting disturbances.
One such gene is the plant homeodomain finger protein 8 (Phf8).
It is encoded on the X-chromosome in mice and humans.Phf8 belongs to the Kdm7 subfamily of histone demethylases, and contains within its 1024 amino acids an aminoterminal plant homeodomain finger for recognition of methylated lysines and further to the middle a Jumonji-C domain for lysine demethylation (Fortschegger & Shiekhattar, 2011).Preferred substrates for demethylation are lysine 9 and lysine 27 of histone 3 (H3K9 and H3K27), particularly in the mono and dimethylated state, and lysine 20 of histone 4 (H4K20), mainly in the monomethylated state (Horton et al., 2010;Liu et al., 2010).
Inactivating PHF8 mutations in humans cause Siderius-Hamel syndrome that combines intellectual disability with craniofacial abnormalities such as orofacial clefts (Laumonnier et al., 2005;Siderius et al., 1999).Craniofacial abnormalities were also detected in Phf8-deficient mice together with cognitive impairments and with a resistance to anxiety-and depression-like behavior (Chen et al., 2018;Walsh et al., 2017).These abnormalities were associated with surprisingly mild alterations in neurons.Supporting a broader function of Phf8 beyond neurons in the CNS, a recent study identified additional roles of Phf8 during astroglial differentiation and for astrocyte function (Iacobucci et al., 2021).Phf8 was found to strongly influence the expression of synaptogenic and astrogenic genes including Nfia in astroglial cells (Iacobucci et al., 2021).
Prompted by these studies, we set out to study the role of Phf8 in oligodendroglial cells.Our findings show that the histone demethylase has a strong impact on proliferation and maintains oligodendrocyte progenitor cells (OPC) in their proliferative state.Additionally, Phf8 reduces the capacity of oligodendrocytes to differentiate.Our data furthermore suggest that Phf8 exerts at least part of its effects through the transcription factor Olig2 that serves as an important regulator of oligodendroglial properties throughout their development and as a central hub of the underlying gene regulatory network.

| Cell lines and culture conditions
The rat oligodendroglial CG4 and Oln93 cell lines were cultured as described (Aberle et al., 2022;Louis et al., 1992;Richter-Landsberg & Heinrich, 1996).Primary rat oligodendroglial cells were obtained from mixed glial cultures and kept under proliferating or differentiating conditions as reported (Weider et al., 2018).Transductions of CG4 cells and primary rat oligodendroglia were at a moiety of infection of 1 using retroviruses that contained a green fluorescent protein (GFP) expression cassette or a polycistronic expression cassette for T7-tagged Phf8 and GFP joined by an internal ribosomal entry site (IRES) under control of the artificial CAG promoter, composed of CMV immediate early enhancer and chicken β-actin promoter (Wedel et al., 2020).CG4 cells were transfected using Xfect (Takara Bio), Oln93 cells using polyethylenimine.
Generation and characterization of the SON15 and SON25 human-induced pluripotent stem cell (hiPSC) lines are reported elsewhere (Wihan et al., 2014;Ehrlich et al., 2017).In short hiPSC lines were generated from dermal fibroblasts of a 71 year-old male and a 66 year-old female donor, respectively, and additionally modified by inserting a doxycycline-inducible expression cassette for SOX10, OLIG2, and NKX6.2 (SON cassette, see Figure 8a) into the AAVS1 locus.From these hiPSC lines, neural precursor cells (NPCs) were generated via embryoid bodies and used to produce oligodendrocytes as described (Ehrlich et al., 2017).NPCs were treated for 12 days with doxycycline to induce expression of SOX10, OLIG2, and NKX6.2.
After 2 days, glial development was initiated by switching the cells to glial induction medium containing platelet-derived growth factor A (PDGFA) and low amounts of triiodothyronine (defined as day 0).After 4 more days, cells were once again transferred to PDGFA-free glial differentiation medium containing high levels of triiodothyronine and kept in this medium for 2 weeks before analysis (for a graphical representation, see Figure 8e).

| RNA-sequencing and bioinformatical analysis
Total RNA was prepared from independent Phf8-deficient or -overexpressing CG4 cell clones using the RNeasy Micro Kit (Qiagen).RNA samples were treated with DNase I to remove contaminating DNA and quantified on a Qubit 4.0 Fluorometer (Life Technologies).RNA integrity was confirmed on an Agilent 5300 Fragment Analyzer (Agilent Technologies).Libraries were prepared from 100 ng RNA using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB), multiplexed, and sequenced on an Illumina NovaSeq 6000 instrument with a 2Â 150 bp pair-end configuration (Azenta Life Sciences, Dresden).On average, 60 million reads were generated per sample library and aligned to the Rattus norvegicus rn6 genome.DESeq2 was used to identify genes differentially expressed in samples from CG4 cells with altered Phf8 expression relative to controls.Differentially expressed genes (DEGs) were defined by a mean base count of ≥ 50 gene-specific transcripts per million overall transcripts (TPM), a log2-fold change ≥ ±0.5 and a p-value ≤ .05.Counts were normalized using factor analysis with the RUVseq package (Risso et al., 2014).Gene expression datasets are deposited in GEO under GSE253877 and GSE253878.Principal component analysis (PCA) plots were generated with ClustVis (Metsalu & Vilo, 2015), volcano plots with VolcaNoseR (Goedhart & Luijsterburg, 2020).Both gene ontology (GO) analysis by GOrilla (Eden et al., 2009) and gene set enrichment analysis (GSEA, Broad Institute, http://software.broadinstitute.org/gsea/index.jsp)were employed to identify altered processes and pathways.

| Luciferase assays
For luciferase reporter assays, 500 ng of luciferase reporter plasmids was transfected into wildtype and Phf8-deficient Oln93 cells on standard 24-well culture plates.All transfections were carried out in triplicates and had comparable transfection efficiencies of 25%-30% as judged by the percentage of autofluorescent cells in parallel transfections with 500 ng of pCMV5-GFP.At 48 h post transfection whole cell extracts were prepared and luciferase activities determined (Aprato et al., 2020;Peirano et al., 2000).

| Cell stainings
For crystal violet stainings, cells were seeded in triplicates at a defined density in poly-L-ornithine-coated dishes.One set of cells underwent fixation 2 h after seeding while other identically seeded plates were incubated for 48 h.Following fixation, cells were stained for 30 min with 0.2% crystal violet (Carl Roth) in 20% ethanol.After repeated washing and plate drying, the remaining crystal violet was extracted from the plates with 0.1 mol/L sodium citrate in 50% ethanol and absorbance measured at 550 nm.

| Statistical analysis
To determine whether differences in transcript levels, proliferation, apoptosis, differentiation, genomic binding, or luciferase activities were statistically significant, a two-tailed Student's t-test with Bonferroni-Sidak correction or a one-/two-way Analysis of Variance (ANOVA) with

| Phf8 influences oligodendroglial gene expression
Oligodendroglial development goes along with substantial changes in chromatin structure, including histone methylations (Parras et al., 2020;Selcen et al., 2023).According to RNA-sequencing (RNA-seq) data (www.brainrnaseq.org),Phf8, as one of the main histone demethylases, is not only expressed in astrocytes and neurons, but also in oligodendroglial cells in both mice and humans (Figure 1a,b).In mice, Phf8 expression has been detected in OPCs, newly myelinating and myelinated oligodendrocytes and thus throughout oligodendroglial development (Figure 1a).However, overall levels appear to decrease as differentiation proceeds.In line with these datasets, our own qRT-PCR experiments also detected higher Phf8 levels in cultures of primary rat oligodendroglial cells under proliferating conditions than under prolonged differentiating conditions (Figure 1c).
Phf8 was also expressed in the rat oligodendroglial CG4 cell line, where transcript levels decreased under differentiating conditions in a manner comparable with primary oligodendroglial cells (Figure 1d).

Rat oligodendroglial Oln93 cells likewise expressed substantial amounts of Phf8 (see below).
To investigate the role of Phf8 in oligodendrocytes, we used CRISPR/Cas9-dependent genome editing to inactivate Phf8 in CG4 and Oln93 cells.Targeted regions included the region immediately following the translational start site in exon 1 as well as the region coding for the catalytic JmjC domain in exons 6 or 7 (Figure 1e).Among the obtained clones, four were selected for further studies in both CG4 and Oln93 cells (Phf8ko).In CG4 cells, clones 1, 2, and 4 carried a single basepair insertion in exon 1, whereas clone 3 harbored a 13-basepair deletion in exon 7 (Figure 1f).Oln93 clones carried single basepair insertions in exon 1 (clones 1 and 2) or exon 6 (clone3) or a single basepair deletion in exon 7 (clone 4) of the Phf8 gene (Figure 1g).Both CG4 and Oln93 cells have a female karyotype (Louis et al., 1992;and own data).Clones are thus likely homozygous for the respective Phf8 mutation.All detected deletions were predicted to lead to frameshift mutations and a premature stop in case of mRNA translation.When analyzed for the presence of regular wildtype Phf8 transcripts, CG4 and Oln93 clones exhibited substantial and statistically significant reductions confirming the successful editing event (Figure 1h,i).
To investigate the effects of Phf8 inactivation on global gene expression, we performed RNA-seq analysis on control CG4 cells and knockout clones.PCA plot and sample clustering confirmed the similar expression profiles of Phf8ko clones as well as the substantial differences relative to controls (Figure 2a,b).As evident from pie chart and volcano plot, an approximately similar number of genes was upregulated and downregulated (2083 up and 1723 down, applying a log2-fold change of ≥ ±0.5 and a p-value ≤ .05) in Phf8ko samples (Figure 2c,d).Among the DEGs, 648 exhibited a strong expression (defined as >20 FPKM according to the Brain RNA Seq database; https://brainrnaseq.org) in OPCs, 426 in myelinating oligodendrocytes.Unbiased GSEA pointed to processes related to proliferation and cell cycle as terms enriched in Phf8ko samples (Figure 2e).Separate GO analysis of the up and downregulated genes showed that terms associated with cell cycle, cell division, and DNA replication were preferentially associated with the downregulated genes (Figure 2f).For the upregulated genes, terms linked to lipid metabolism and axon ensheathment were prominent (Figure 2g).The latter did not rank high in the GSEA list as normalized enrichment scores and p-values were lower for the upregulated than for the downregulated genes.Interestingly, Kdm3a and Kdm6b, two histone demethlyases with Phf8-like activities, were among the upregulated genes (log2-fold change of 0.90 and 0.59, respectively).This argues that there is a compensatory upregulation of enzymatically similar histone demethlyases, but that functional redundancy between the enzymes is at best partial.
To gain additional support for the role of Phf8 in CG4 cells, we also generated two polyclonal CG4 cell lines that overexpressed Phf8 after retroviral transduction (Phf8oe lines, Figure 3a) and compared their expression profiles to wildtype CG4 cells.Again, samples from wildtype and Phf8-overexpressing cells clustered separately in PCA plots and were most similar within their groups (Figure 3b,c).Numbers of upregulated and downregulated genes upon Phf8 overexpression were again very similar as judged from volcano plot and listed in the pie chart (Figure 3d,e).However, the number of DEGs corresponded to only 24% of the DEGs obtained after Phf8 inactivation (321 up, 587 down, applying a base mean >100, log2-fold change of ≥ ± 0.5 and adjusted p-value ≤ .05)arguing that loss of Phf8 activity has a much stronger effect on the CG4 expression profile than increase of the already present activity.
According to GO analysis, terms associated to (cellular) differentiation were found preferentially enriched for the downregulated genes after Phf8 overexpression, whereas precursor/progenitor cell proliferation and related terms were prominent for the upregulated genes (Figure 3f,g).We conclude from our RNA-seq experiments on CG4 cells that Phf8-dependent effects on the expression profile point to a positive effect on OPC proliferation and a negative one on oligodendrocyte differentiation.

| Phf8 expression promotes the proliferative OPC state and impedes differentiation
To study whether the observed changes in gene expression translate into altered cellular properties of CG4 cells, we compared the proliferation of wildtype CG4 and Phf8ko cells.Using crystal violet staining, we observed a significant reduction in cell proliferation for all three Phf8ko clones that ranged from 27% to 63% compared with wildtype (Figure 4a).This correlated well with the reduced transcript levels of cell cycle, cell division, and DNA replication genes in the RNA-seq studies.BrdU incorporation was likewise substantially decreased in Phf8ko clones and reached only 71%-79% of wildtype levels (Figure 4b).Apoptosis rates, by contrast, were unaltered and comparable between Phf8ko cells and wildtype CG4 cells (Figure 4c). a very early differentiation marker that is turned on in pre-myelinating oligodendrocytes, whereas MBP is induced later in myelinating oligodendrocytes.After 6 days of differentiation, the difference in O4 was still visible, but no longer statistically significant, whereas the difference in MBP (69% ± 7% for control as compared with 31% ± 2% for Phf8 overexpression) was still substantial (Figure 4h-j).We conclude from these findings that Phf8 expression indeed promotes the proliferative state of OPCs and impedes oligodendrocyte differentiation.
From these experiments, we cannot determine whether the opposite effects on proliferation and differentiation are independent or whether changes in differentiation are secondary to the altered proliferation.

| Olig proteins are targets of Phf8
To understand how Phf8 may exert its effects on oligodendroglial cells on the molecular level and identify potential Phf8 effectors, we studied the expression of the main known oligodendroglial transcription factors in CG4 cells with and without Phf8 (Sock & Wegner, 2021).Many exhibited unaltered or only marginally changed expression levels after Phf8 loss (defined as log2 fold change between + 1 and À 1; Figure 5a).This included Sox10, Ascl1, Nkx2.2, Myrf, Zfp24, and YY1.Only two transcription factors (i.e., Sox4 and Id2) exhibited a significantly increased expression in the absence of Phf8 and five (i.e., Sox9, Sox11, Hes5, Olig1, and Olig2) a significantly decreased expression.Considering that histone demethylation by Phf8 is activating and that Phf8 is usually regarded as a transcriptional co-activator, increased expression after Phf8 loss is likely indirect.
Among the five downregulated transcription factors, Sox9 and Hes5 exhibited very low and biologically negligible expression levels in CG4 cells (see red dots, Figure 5a), and the decrease of Sox11 was counterbalanced by an increase of the functionally redundant Sox4.This drew our attention to Olig2 and Olig1.
qRT-PCR confirmed the downregulation of Olig1 and Olig2 transcripts in Phf8ko cells (Figure 5b,c).Olig proteins are expressed throughout the oligodendroglial lineage (Rowitch et al., 2002).Olig2 has important roles in OPCs and differentiating oligodendrocytes, and is seconded in its function by Olig1.At late stages of differentiation, Olig2 expression becomes downregulated, whereas Olig1 is excluded from the nucleus (Rowitch et al., 2002).Indeed, immunocytochemistry for Olig2 in CG4 cells revealed that the overall Olig2 staining was lower in Phf8ko cells than in wildtype cells and that Olig2 immunoreactivity had disappeared from a significant number of Phf8ko cells altogether (Figure 5d,e).
Given the essential roles of Olig proteins in oligodendroglial development, it seemed reasonable to assume that the Olig proteins may be one of the prominent direct targets through which Phf8 exerts its effects in oligodendroglial cells.In support of this assumption, comparison of the DEGs in Phf8ko cells and genes with Olig2 ChIP-seq (ChIP followed by high throughput sequencing) peaks (Darr et al., 2017) identified 797 Olig2 target genes among the DEGs.The 342 genes downregulated ones in Phf8ko cells were involved in cell cycle, DNA replication and other processes related to proliferation by GO analysis, whereas the 455 upregulated ones were associated to axon ensheathment and myelination (Figure 5f,g).
To further confirm a direct regulatory relationship between Phf8 and the Olig proteins, we searched the literature for previously identified regulatory regions with proven or likely relevance for Olig2 expression (Figure 6a).Olig1 has not been studied in detail, but is localized in the same topologically associated domain in immediate vicinity to Olig2 so that regulatory regions with relevance for Olig2 expression may be relevant for Olig1 as well (Figure 6a).In addition to the promoter region, two regulatory regions are localized in front of the Olig2 gene, approximately À33 and À17 kb upstream of the transcriptional start site and referred to as OLE and 5F7, respectively (Friedli et al., 2010;Oosterveen et al., 2012;Weider et al., 2015).Two additional regulatory regions, named K23 and E1b are present + 9.5 and + 72 kb downstream of the transcriptional start site (Fan et al., 2023;Sun et al., 2006).In contrast to the OLE, 5F7, and E1b enhancers, K23 is active in the neuronal lineage only and thus irrelevant for oligodendroglial expression (Sun et al., 2006).
Using chromatin from CG4 cells ectopically expressing a T7-tagged Phf8, we performed ChIP combined with qRT-PCR (ChIP-PCR) to study whether Phf8 is present on any of the known regulatory regions.ChIP-PCR revealed an enrichment of the OLE and E1b enhancers as well as the Olig2 promoter in the precipitate obtained with an anti-T7-tag antibody relative to control IgG (Figure 6b).In contrast, no enrichment was obtained for the 5F7 and K23 enhancers.Using chromatin from wildtype and Phf8ko CG4 cells, we were furthermore able to show that the levels of monomethylated H3K9 and dimethylated H3K27 were substantially higher on the OLE and E1b enhancers as well as on the Olig2 promoter in Phf8ko as compared to wildtype CG4 cells (Figure 6c,d).
Again, no differences were observed for the 5F7 and K23 enhancers between wildtype and Phf8ko cells.Considering that monomethylated H3K9 and dimethylated H3K27 residues represent substrates for Phf8 and that these repressive histone marks are increased in the absence of Phf8 on regulatory regions that are normally bound by Phf8, it appears likely that Phf8 influences Olig2 expression via these regions.

| Inactivation of PHF8 does not impair the transcription factor-enforced generation of oligodendrocytes from human NPC
To extend our study to human cells, we studied whether PHF8 inactivation would impact the generation of oligodendrocytes from hiPSCs.
We used two independent hiPSC lines named SON15 and SON25 that had been modified to contain a doxycycline-inducible expression cassette for SOX10, OLIG2, and NKX6.2 (Ehrlich et al., 2017) inserted into the AAVS1 safe harbor locus (Figure 8a).These lines underwent CRISPR/Cas9-dependent genome-editing at the PHF8 locus using a guide RNA that targets the beginning of the JmjC domain in exon 6 (Figure 8b).We obtained two clones with hemizygous inactivating mutations in the male SON15 cell line (Figure 8c), and two clones each with heterozygous and homozygous inactivating mutations in the female SON25 cell line (Figure 8d).Each clone and the corresponding hiPSCs were then converted into NPCs and differentiated to oligodendrocytes by doxycycline-induced, forced expression of SOX10, OLIG2, and NKX6.2 during the first 12 days of the differentiation protocol (Figure 8e) (Ehrlich et al., 2017).When analyzed at day 14 or day 18 after the onset of differentiation, a comparable number of the cells had turned on expression of the early differentiation marker O4 in controls and PHF8-mutant clones for both the SON15 and SON25 hiPSC line left).Similarly, we failed to detect differences in the induction of the late differentiation marker MBP (Figure 8f-i, right; for representative picture see Figure 8j).In all PHF8-mutant clones analyzed, not only the number of MBPexpressing cells was comparable with control hiPSCs but also the increase from day 14 to 18.Under the assumption that PHF8 exerts important parts of its function through OLIG2, these results are not surprising as lowered expression of the endogenous OLIG2 upon PHF8 inactivation will likely remain without consequences during differentiation, because OLIG2 is ectopically provided to the cells for the first 12 days following doxycycline induction.If at all, reduced expression of the endogenous OLIG2 may become functionally relevant only during the final phase of differentiation.In that case, normal generation of human oligodendrocytes in the absence of PHF8 may indicate that the protein primarily acts in the earlier progenitor phase and that alterations in differentiation are secondary to the proliferation defect.

| DISCUSSION
Phf8 has previously been shown to be required for learning and memory as well as for proper anxiety response (Chen et al., 2018;Laumonnier et al., 2005;Siderius et al., 1999;Walsh et al., 2017).
Although such functions are usually associated with neurons, Phf8 deficiency has been found to elicit fairly mild neuronal deficiencies and affect astrocytic development and function as well (Chen et al., 2018;Iacobucci et al., 2021;Walsh et al., 2017).Here we show that Phf8 also affects oligodendroglial cells in addition to neurons and astrocytes.Oligodendroglial cells also express additional histone demethylases with overlapping activity such as Kdm3a and Kdm6b.At present we cannot distinguish whether the proliferation-promoting and differentiation-inhibiting functions are independent or not.However, we favor the hypothesis that the effects on differentiation are secondary to the ones on proliferation.For one, Phf8 acts as a demethylase of histone marks that are primarily regarded as inhibitory (Horton et al., 2010;Liu et al., 2010).As a consequence, Phf8 is usually regarded as a transcriptional co-activator.Reduced oligodendroglial proliferation in the absence of Phf8 is compatible with a co-activator function, whereas increased differentiation would rather point to corepressor activity if the effect is a direct one.Additionally, we failed to observe an effect on human oligodendrocyte differentiation when using an experimental strategy that overrides PHF8 requirements during the early phases of NPC differentiation to oligodendrocytes, best corresponding to a progenitor state.Together with the higher expression of Phf8 in OPCs than in myelinating oligodendrocytes and a higher number of OPC-than oligodendrocyte-  Because of the continuous requirement for Olig2 during oligodendroglial development, and the need for higher Olig2 levels in OPCs and lower levels in differentiating oligodendrocytes (Rowitch et al., 2002), it seems reasonable that Phf8 exerts part of its effects by modulating Olig2 expression levels.In support of such an assumption, Olig2 rescued the proliferation defect of Phf8-deficient rat oligodendroglial CG4 cells.Additionally, the generation of human oligodendrocytes from hiPSC-derived NPCs remained unchanged in the absence of PHF8 when OLIG2 was ectopically provided during the initial stages of converting NPCs into oligodendroglial cells before T3-induced differentiation.
While our data strongly support the role of Olig2 and Olig1 as important direct targets and downstream effectors of Phf8, further factors are likely involved in mediating Phf8 effects during oligodendroglial development.In our study, we have focused solely on transcription factors with altered expression after Phf8 loss and have not considered other chromatin-associated molecules.Additionally, we have not looked into the possibility that Phf8 changes transcription factor activity rather than expression by direct interaction or demethylation as recently shown for YY1 in colon and lung cancer cells (Wu et al., 2024).
Interestingly, previous studies provided data that Phf8 mainly works by modifying mTOR signaling in neurons and affecting the expression of astrogenic genes such as Nfia in astrocytes (Chen et al., 2018;Iacobucci et al., 2021).Our identification of Olig2 as a main downstream target of Phf8 therefore implies that Phf8 targets and modes of action are highly variable between cell types.
We think that such a conclusion makes sense because epigenetic landscapes and relevant transcription factors as main determinants of Phf8 recruitment will be very different between cell types.As a consequence, Phf8 targets different, highly specific genomic loci in neurons, astrocytes, and oligodendroglial cells, and thus affects gene expression in a cell type-specific manner.It seems likely that Phf8 influences expression of yet another set of genes in neural crest cells, whose dysregulation upon Phf8 inactivation will then lead to the craniofacial abnormalities observed in Siderius-Hamel patients and Phf8-deficient mice (Chen et al., 2018;Laumonnier et al., 2005;Siderius et al., 1999;Walsh et al., 2017).
Several transcription factors have been previously identified that-like Phf8-promote the proliferative progenitor state of OPCs.Among others, these include Hes5, the SoxC proteins Sox4 and Sox11 and the SoxD proteins Sox5 and Sox6 (Baroti et al., 2016;Kuhlbrodt, Herbarth, Sock, Enderich, et al., 1998;Liu et al., 2006;Potzner et al., 2007;Stolt et al., 2006).It seems reasonable to assume that Phf8 functionally interacts with some of these transcription factors in OPCs.In agreement, bioinformatic analysis identifies potential binding sites for all of these factors within the Olig2 regulatory regions targeted by Phf8.However, ChIP-seq studies are not yet available for confirmation.Therefore, future investigations will have to clarify whether this is indeed the case.
) were separately cloned into the pCAG-SpCas9-GFP-U6-gRNA plasmid (Addgene, Teddington, UK).CG4 and Oln93 cells were transfected, hiPSCs electroporated with the resulting plasmids, purified by fluorescence activated cellsorting and seeded at single cell density in 96-well plates.Homogeneously GFP-expressing clones were analyzed for alterations in the Phf8/PHF8 gene by sequencing the region of the gene targeted by the guide RNA and determining the introduced changes by Inference of CRISPR Edits (ICE) analysis, using Synthego ICE Analysis 2019, v3.0.

1
Expression and inactivation of Phf8 in oligodendroglial cells.(a, b) Summary of Phf8 expression in various cell types of the mouse (a) and human (b) brain.Data are taken from the Brain RNA Seq database (https://brainrnaseq.org).Expression levels are given in fpkm.(f/m) A, (fetal/mature) astrocyte; N, neuron; OPC, oligodendrocyte progenitor cell; (n/m) OL, (newly formed/myelinating) oligodendrocyte.(c, d) Determination of Phf8 transcript levels by qRT-PCR in primary rat oligodendroglial cells (c) and CG4 cells (d) cultured under proliferating (prol) or differentiating conditions for 1, 3, or 6 days (1d, 3d, or 6d).Transcript levels are depicted relative to levels under proliferating conditions and are presented as mean values ± SEM (n = 3-4).(e) Schematic representation of the Phf8 protein with its plant homeodomain finger (P) and the catalytic Jumonji-C (JmjC) domain.Numbers above the bar represent amino acid residues.Arrows below the bar indicate the regions targeted by guide RNAs.Sequences of guide RNAs and their location in exon 1, 6, and 7 are shown in the lower part of the panel.(f, g) Summary of CG4 (f) and Oln93 (g) cell lines (Phf8ko) and the CRISPR/Cas9-induced alterations in the Phf8 gene.Mentioned is the position of the respective insertion (ins) or deletion (del) within the cDNA (c).For single nucleotide changes the inserted or deleted base is given.For larger changes the number of altered nucleotides is listed.(h, i) Comparison of normal Phf8 transcript levels in CG4 (h) and Oln93 (i) Phf8ko lines (white bars) to transcript levels in wildtype cells (set to 1, black bars), shown as mean values ± SEM (n = 3-4).Statistical significance was determined by two-tailed student's ttest with Bonferroni-Sidak correction (* p ≤ .05;** p ≤ .01;***p ≤ .001).

F
I G U R E 2 Changes in CG4 cell expression following Phf8 inactivation.(a) Principal component analysis (PCA) plot of samples from wildtype CG4 cells (WT, black dots) and Phf8ko lines (KO, white dots) used for RNA-seq.(b) Clustering of RNA-seq samples according to their similarity.(c) Pie chart summarizing the number of genes downregulated (blue) or upregulated (red) in Phf8ko lines relative to wildtype CG4 cells.(d) Volcano plot of down and upregulated genes.Olig2 and Olig1 are marked.(e) Gene set enrichment analysis based on the obtained expression profiles for wildtype CG4 cells and Phf8ko lines.(f, g) Gene ontology analysis for the down (f) and up (g) regulated genes (log2-fold change ≥ ±0.5; p-value ≤.05) in the Phf8ko lines.Terms discussed in the text are highlighted in red.For further experiments, we switched to primary cultures of rat oligodendroglial cells and assessed the consequences of Phf8 overexpression upon retroviral transduction.First, we studied the consequences of Phf8 overexpression on BrdU incorporation as a measure of proliferation.In line with our findings in CG4 cells, Phf8 increased the proliferative capacity of primary oligodendroglial cells F I G U R E 3 Changes in CG4 cell expression profile following Phf8 overexpression.(a) Scheme for the generation of CG4 cell lines overexpressing a T7-tagged version of Phf8 (Phf8oe).(b) Principal component analysis (PCA) plot of samples from wildtype CG4 cells (WT, black dots) and Phf8oe lines (white dots) used for RNA-seq.(c) Clustering of RNA-seq samples according to their similarity.(d) Volcano plot of downregulated and upregulated genes.(e) Pie chart summarizing the number of genes downregulated (blue) or upregulated (red) in Phf8oe lines relative to wildtype CG4 cells.(f, g) Gene ontology analysis for the downregulated (f) and upregulated (g) genes (log2-fold change ≥ ±0.5; adjusted p-value ≤.05) in the Phf8oe lines.Terms discussed in the text are highlighted in red.by 37% ± 9% (Figure 4d,e).When differentiation was assessed after 3 days, we observed a strong decrease of cells that had turned on the O4 surface marker (80% ± 2% for control as compared with 48% ± 5% for Phf8 overexpression) or myelin basic protein (MBP; 28% ± 4% for control as compared with 12% ± 4% for Phf8 overexpression) in the Phf8 transduced cultures (Figure 4f,g).O4 represents F I G U R E 4 Impact of Phf8 on properties of rodent oligodendroglial cells.(a) Growth comparison of wildtype CG4 cells (black bars) and Phf8ko lines (white bars) as determined by crystal violet staining (n = 5).Normalization was to the OD 550nm value obtained for freshly seeded cells (dotted line, 1).(b) Quantification of proliferation rates of wildtype CG4 cells and Phf8ko lines from BrdU incorporation assays (n = 3).(c) Determination of apoptosis rates for wildtype CG4 cells and Phf8ko lines according to cleaved caspase 3-stainings (n = 3).(d, e) Immunocytochemical visualization (d) of BrdU incorporation (red in d) in control (ctrl, black bars in e) and Phf8-overexpressing (Phf8, white bars in e) transduced (GFP labeled, green in d) Sox10-positive (pink in d) primary rat oligodendroglial cells (tOGCs) and quantification (e) of the BrdU-positive fraction from five independent experiments, presented as mean ± SEM. (f, g) Quantification of the O4-positive (f) and myelin basic protein (MBP)-positive (g) fraction of control and Phf8-overexpressing transduced primary rat oligodendroglial cells after 3 days of differentiation from three (f) or four (g) independent immunocytochemical stainings.(h-j) Analogous quantification of the O4-positive (h) and MBP-positive (i) fraction of control and Phf8-overexpressing transduced oligodendroglial cells after 6 days of differentiation (j, for representative staining with MBP in red and the oligodendroglial lineage marker Sox10 in pink).The merge in panels (d) and (j) combines GFP and BrdU (d) or GFP and MBP (j) signals.Scale bar: 50 μm.Statistical significance was determined by two-tailed student's t-test with Bonferroni-Sidak correction (* p ≤ .05;** p ≤ .01;*** p ≤ .001).
The promoter and the OLE and E1b enhancers were also amplified from genomic DNA and inserted into luciferase reporter plasmids.Upon transient transfection of the reporter plasmids into wildtype and Phf8ko Oln93 cells (Figure 6e-h), we observed higher activity of the promoter-and E1b-containing plasmids in wildtype Oln93 cells than in Phf8-deficient Oln93 cells (Figure 6f,g).The OLE-containing plasmid was similarly inactive in wildtype and Phf8-deficient Oln93 cells, and the empty reporter plasmid exhibited comparable activity in Phf8-positive and Phf8-negative Oln93 cells (Figure 6e,h).Despite the unexpected behavior of the OLE enhancer in Oln93 cells, the results clearly support the assumption that Olig2 represents a direct target of Phf8 activity in oligodendroglial cells.To investigate whether Olig2 does not only constitute a target but also an important effector of Phf8 function in oligodendroglial cells, we transduced cells from two Phf8ko CG4 clones with an Olig2 and GFP expressing retrovirus and compared proliferation to cells transduced with a retrovirus expressing only GFP.These studies revealed that the presence of Olig2 significantly increased the proliferation rate of transduced Phf8ko cells in both CG4 clones (Figure7a,b).We conclude that Olig2 is capable of rescuing at least the proliferation defect in Phf8-deficient CG4 cells.

F
I G U R E 5 Impact of Phf8 on Olig2 expression in CG4 cells.(a) Expression of the main know transcription factors with relevance for oligodendroglial development.Alterations in expression after Phf8 loss are shown as log2-fold changes (filled green dots, p-value ≤.05; greenrimmed dots, p-value ≥.05) and overall expression levels as base mean (red dots).The area of unchanged or mildly altered expression is colored in light green.(b, c) Comparison of Olig1 (a) and Olig2 (b) transcript levels in wildtype CG4 cells (black bars) and Phf8ko lines (white bars) by qRT-PCR.Transcript levels in wildtype cells were set to 1. Values are presented as mean ± SEM (n = 3-4).(d) Immunocytochemical stainings of wildtype CG4 cells (WT) and Phf8ko lines for Olig2 (upper row).Nuclei were counterstained with DAPI (lower row).(e) Numbers of Olig2-positive wildtype and Phf8ko CG4 cells quantified from immunocytochemical Olig2 stainings (n = 3).(f, g) Gene ontology analysis for the downregulated (f) and upregulated (g) direct Olig2 target genes (log2-fold change ≥ ±0.5; p-value ≤.05) in the Phf8ko lines.Terms mentioned in the text are highlighted in red.Scale bar: 50 μm.Statistical significance was determined by two-tailed student's t-test with Bonferroni-Sidak correction for (a, b) and by two-way ANOVA with Dunnetts correction for (d) (* p ≤ .05;** p ≤ .01;*** p ≤ .001).

F
I G U R E 6 Analysis of Olig2 regulatory regions for the presence of and response to Phf8 in oligodendroglial cells.(a) Regulatory landscape of the Olig2 locus in the rat genome.Regulatory regions are shown as white boxes, Olig2 and Olig1 genes as gray boxes with the direction of transcription indicated by arrows.Numbers above the boxes indicate the distance of regulatory regions to the transcriptional start site of Olig2, terms below the names of previously identified enhancers (OLE, 5F7, K23, E1b) and the promoter (prom).(b) Determination of the amounts of Olig2 regulatory regions precipitated with T7-tag antibody (a-T7, black bars) and IgG control (white bars) from chromatin of CG4 cells expressing a T7-tagged Phf8 by quantitative PCR (qRT-PCR).Data are presented as mean fold enrichment over IgG ± SEM (n = 4).(c, d) Determination of the amounts of Olig2 regulatory regions precipitated with H3K9me1 (c), H3K27me2 (d) and control IgG (c, d) antibodies from chromatin of wildtype (black bars) and Phf8ko CG4 cells (white bars) by quantitative qRT-PCR.Data are presented as mean fold enrichment over IgG (set to 1, indicated by dotted line) ± SEM (n = 4).(e-h) Luciferase reporter assays in wildtype (black bars) and Phf8ko (white bars) Oln93 cells transfected with plasmids carrying the luciferase reporter under control of the OLE enhancer (e), the Olig2 promoter (f) and the E1b enhancer (g).An empty reporter served as control (ctrl, h).Luciferase expression was determined as light units (LU).Data obtained from six to eight independent experiments.Statistical significance was determined by two-tailed student's t-test for (b-d) and one-way ANOVA with Dunnetts correction for (e-h) (*, p ≤ .05;**, p ≤ .01;***, p ≤ .001).

Upon
Phf8 inactivation they are slightly upregulated in CG4 cells but cannot (or not fully) compensate arguing that the histone demethlyases are not completely redundant in their function.Recruitment by different factors to specific genomic regions is one possible reason.Our data clearly show that Phf8 strengthens the expression of genes associated with cell cycle progression, replication, and related processes.As a consequence, Phf8 promotes the proliferation of OPCs and appears involved in maintenance of the proliferative state within the oligodendroglial lineage.At the same time, Phf8 also appears to keep the oligodendroglial capacity for differentiation and myelination in check.
specific genes among DEGs in Phf8ko cells, these considerations favor a predominant role in OPCs.However, clarification of a direct versus indirect role of Phf8 in oligodendrocyte differentiation will have to await future experiments.

F
I G U R E 7 Olig2 impact on proliferation of Phf8-deficient CG4 cells.(a) Immunocytochemical visualization of BrdU incorporation in transduced, GFP-expressing (ctrl) or Olig2 and GFP-expressing (Olig2) Phf8-deficient CG4 cells (Phf8ko1 and Phf8ko3).Total cells are visualized by a nuclear DAPI stain.Red arrows in the GFP panels mark proliferating transduced cells.Scale bar: 50 μm.(b) Quantification of the BrdU-positive fraction of transduced CG4 cells (tCG4) from three to four independent experiments, presented as mean ± SEM.Statistical significance was determined by two-tailed student's t-test with Bonferroni-Sidak correction (* p ≤ .05;** p ≤ .01).Two of the prominent targets of Phf8 identified by RNA-seq analysis are Olig2 and its closely related paralog Olig1.We found (i) that Phf8 is present in vivo at Olig2 regulatory regions, (ii) that Phf8 loss has consequences on the histone modification pattern in these regions, and (iii) that Phf8 activates Olig2 regulatory regions in reporter assays.All these findings support the assumption that Olig2 F I G U R E 8 PHF8 impact on the generation of human oligodendrocytes from hiPSCs.(a) Schematic representation of the SON cassette for the doxycycline (DOX)-inducible expression of SOX10, OLIG2, and NKX6.2 integrated into the AAVS1 locus of the SON15 and SON25 iPSC lines.(b) Schematic representation of the PHF8 protein with its plant homeodomain finger (P) and the catalytic JmjC domain.Numbers above the bar represent amino acid residues.Arrow below the bar indicates the region targeted by the guide RNA.Guide RNA sequence and location in exon 6 are shown in the lower part of the panel.(c, d) Summary of SON15 (c) and SON25 (d) cell lines and their CRISPR/Cas9-induced alterations in the PHF8 gene.Mentioned is the position of the respective insertion (ins) or deletion (del) within the cDNA (c.).For changes of one or two nucleotides the identity of the inserted or deleted bases is given.For larger changes the number of altered nucleotides is listed.(e) Graphical representation of the 20 day differentiation protocol for hiPSC-derived NPCs to oligodendrocytes from day À2 (d-2) until day 18 (d18) with indication of the consecutive steps (incubation in NPC medium, glial induction and glial differentiation medium) and the period of doxycycline treatment.(f-i) Quantified expression of the early oligodendroglial differentiation marker O4 (green) and the late marker myelin basic protein (MBP) (magenta) in wildtype (filled bars) and PHF8-mutant (open bars) SON15 (f, h) and SON25 (g, i) cells after 14 days (f, g) and 18 days (h, i) of oligodendroglial differentiation.(j) Representative picture showing O4 and MBP expression for SON15 wildtype and C1 cells after 18 days of differentiation.Scale bar: 50 μm.No statistical significance was observed by two-way ANOVA in the differentiation capacity of wildtype and PHF8-mutant hiPSCs. is a direct target of Phf8.Considering the localization of the Olig1 gene in the vicinity of Olig2, it appears possible that the same regulatory regions also affect Olig1 expression and confer a comparable Phf8-responsiveness to Olig1.
Nevertheless, we have shown in this study that Phf8 has a substantial impact on oligodendroglial development.As a consequence, changes in oligodendroglial development or function elicited upon Phf8 mutation may impact overall CNS development and circuit function, and contribute to Phf8-induced intellectual disability.