Phf15—a novel transcriptional repressor regulating inflammation in mouse microglia

Aim Excessive microglial inflammation has emerged as a key player in mediating the effects of aging and neurodegeneration on brain dysfunction. Thus, there is great interest in discovering transcriptional repressors that can control this process. We aimed to examine whether Phf15—one of the top differentially expressed genes in microglia during aging in humans—could regulate transcription of pro-inflammatory mediators in microglia. Methods RT-qPCR was used to assess Phf15 mRNA expression in mouse brain during aging. Loss-of-function (shRNA-mediated knockdown (KD) and CRISPR/Cas9-mediated knockout (KO) of Phf15) and gain-of-function (retroviral overexpression (OE) of murine Phf15 cDNA) studies in a murine microglial cell line (SIM-A9) followed by immune activation with lipopolysaccharide (LPS) were used to determine the effect of Phf15 on pro-inflammatory factor (Tnfα, Il-1β, Nos2) mRNA expression. RNA-sequencing was used to determine global transcriptional changes after Phf15 knockout under basal conditions and after LPS stimulation. Results Phf15 expression increases in mouse brain during aging, similar to humans. KD, KO and OE studies determined that Phf15 represses mRNA expression levels of pro-inflammatory mediators such as Tnfα, Il-1β and Nos2. Global transcriptional changes after Phf15 KO showed that Phf15 specifically represses genes related to the antiviral (type I interferon) response and cytokine production in microglia. Conclusion We provide the first evidence that Phf15 is an important transcriptional repressor of microglial inflammation, regulating the antiviral response and pro-inflammatory cytokine production. Importantly, Phf15 regulates both basal and signal-dependent activation and controls the magnitude and duration of the microglial inflammatory response.


66
Studies have also highlighted an important role for chromatin modifications in the transcriptional 67 control of inflammatory gene expression [19,20] . A recent study by Soreq et al. [21] , which compared 68 transcriptional profiles of different brain cell types and regions throughout healthy human aging found 69 microglial gene expression profiles as being one of the most predictive markers of biological age in the 70 brain [21] . The same study identified a relatively unknown gene, PHD finger protein 15 (PHF15) among the 71 top 25 differentially expressed genes in microglia during aging. Work in embryonic stem cells, and 72 sequence and structural similarity to other members of the PHF family, indicate that PHF15 is a putative 73 chromatin-mediated gene regulator [22,23] .

74
Given that aging skews microglia towards a pro-inflammatory phenotype, and that PHF15 was 75 found to be highly upregulated during non-pathological aging, we sought to determine whether Phf15 76 might regulate microglial inflammatory function. We found that Phf15 strongly represses pro-inflammatory 77 gene expression, regulating both basal and signal-dependent activation and modulating the magnitude and 78 duration of the mouse microglial inflammatory response. Importantly, Phf15 seems to regulate pro-79 inflammatory and Interferon type I (IFN-I)-dependent gene expression. Increased IFN-I tone and pro-80 inflammatory cytokine expression are both hallmarks of the aging brain [24][25][26][27][28] . Our findings suggest that 81 Phf15 is an important novel repressor of microglial inflammatory function that might work to counteract 82 age-induced inflammation in the healthy, aging brain.

87
Adult male C57Bl6/J mice were purchased from The Jackson Laboratory and maintained on a 12:12-h 88 light-dark cycle (lights on at 0700 hours) with ad libitum access to food and water and aged for ~2.5, ~14 89 or ~20 months. All animal care and procedures were approved by the University of California, Berkeley 90 Animal Care and Use Committee.

156
(mm10) using Spliced Transcripts Alignment to a Reference (STAR) aligner [31] . Count data was analyzed 157 with Hypergeometric Optimization of Motif EnRichment (HOMER) software for next-generation 158 sequencing analysis (http://homer.ucsd.edu/homer/ngs/index.html) which uses the R/Bioconductor package 159 adjusted p values and log2fold expression changes between Phf15 KO and control conditions for each genes and less than -1.5 for downregulated genes). Too few downregulated genes (< 200) passed the more 165 stringent adjusted p < 0.01 cutoff for robust downstream biological function analysis, so the adjusted p 166 value threshold was lowered to p < 0.05. Results were visualized using the R package 167 EnhancedVolcano [33] . Lists of upregulated and downregulated genes were input into Metascape [34] , a gene 168 annotation and analysis tool, to determine enriched biological themes within the gene lists.

RESULTS 183
Aging increases Phf15 expression in mouse brain.

184
To investigate whether Phf15 increases in mouse brains similar to humans [21] , we measured Phf15 mRNA 185 expression in mouse frontal cortical brain areas across age. We were interested in frontal cortical regions 186 because of their involvement in mediating various aspects of cognitive function and because they are 187 selectively affected in several aging-related neurodegenerative conditions like PD, AD and frontotemporal 188 dementia (FTD) [35,36] .

204
cognitive and neurological deficits [37] , 2) Aged individuals show increased systemic levels of LPS in the 205 bloodstream [38] which are associated with increased inflammation and microglial activation [39] and 3) In 206 humans, TLR4 activation is linked to age-related pathologies like PD and AD [40][41][42] . Thus, LPS serves as a

293
Differential gene expression analysis after 6 hours of LPS stimulation in KO versus control cells 294 revealed 576 up-regulated genes (log2 fold change > 1.5 and p adj < 0.01) and 322 down-regulated genes 295 (log2 fold change < -1.5 and p adj < 0.05) ( Figure 6A). Interestingly, by 6 hours after LPS administration, 296 some of the most enriched biological process categories in KO cells were related to "cytokine secretion" 297 and "immunoregulatory interaction" (Figure 6B and C), denoting a strong increase in magnitude of 298 expression of genes involved in regulating the secretion of pro-inflammatory mediators. The downregulated 299 genes at 6 hours after LPS stimulation in KO cells relative to control again displayed more variability, but 300 did show decreases in biological process categories related to "regulation of defense response" and 301 "cytokine production", indicating negative regulation of these processes in Phf15 KO cells compared to 302 control (Supplementary Figure 9A).

303
Motif enrichment analysis for TF binding sites enriched in the promoters of upregulated genes at 304 the 6-hour time point revealed consensus sequences for AP-1, a key regulator of microglia reactivity in 305 inflammation [49] (Figure 6D). Motif enrichment for the set of downregulated genes revealed ISRE, such as 306 IRF1 and IRF3 motifs (Supplementary Figure 9B), supporting the observation that there is a negative 307 "regulation of defense response" by 6 hours post stimulation. It is interesting to note that a functional 308 transition from cytokine production to cytokine secretion seems to occur in the 6 hour period after LPS 309 activation.

310
Taken together, our RNA-seq results confirm that Phf15 is a repressor of microglial inflammatory 311 gene expression, regulating the antiviral response -specifically, IFN-I-dependent responses -as well as 312 processes related to pro-inflammatory cytokine production and release.

314
DISCUSSION 315 Our results show that Phf15 inhibits microglial expression of pro-inflammatory mediators under basal and 316 signal-dependent activation, regulating both the magnitude and duration of the inflammatory response.

317
Genetic deletion of PhfF15 in a microglial cell line followed by stimulation with LPS lead to an 318 exaggerated pro-inflammatory response with increased production of Tnfα, Il-1β and Nos2 over a time 319 course of 24 hours. Importantly, levels of pro-inflammatory factors remained elevated at 24 hours 320 demonstrating a sustained and prolonged response. Consistent with our LPS stimulation of TLR4 results, 321 similar results were obtained after TLR9 and TLR3 activation confirming that Phf15 is a general negative 322 regulator and controls both the MyD88 and TRIF downstream signal transduction pathways rapidly rise in experimental models of PD and are highly toxic to dopaminergic neurons [12,13,50] . Similarly, 328 high levels of TNFα are a hallmark of PD in humans [51][52][53] . Additionally, both TNFα and IL-1β are 329 involved in maintaining proper synaptic plasticity at physiological levels [54,55] and overproduction of these 330 cytokines can result in neuronal death via excitoxicity and cognitive dysfunction [56,57] .

331
Our studies further demonstrate that Phf15 can regulate both basal and signal-dependent

362
But how might Phf15 be involved in regulating transcriptional repression of the inflammatory 363 response? PHF15 was first described in embryonic stem cells as an E3 ligase that directly targets Lysine-364 specific demethylase 1 (LSD1, Kdm1a) -a key demethylase of histone 3 lysine 4 -for degradation [22] . 365 LSD1 has been identified as a member of the CoREST co-repressor complex [60,61] which is required for 366 transcriptional repression of inflammation in microglia [8] . We therefore initially hypothesized that increased 367 levels of Phf15 upon aging might lead to decreased levels of LSD1 and increased microglial inflammatory 368 output. Our results, however, demonstrate that Phf15 itself inhibits microglial inflammatory function, thus, 369 its purported mechanism for inhibition is likely not via degradation of LSD1.

370
Interestingly, the global transcriptional changes caused by Phf15 deletion are highly similar to 371 age-associated transcriptional changes in microglia that have been previously reported [9,62,63] . In particular, 372 a study by Deczkowska et al. [64] , found "immune system process" and specifically "response to virus" 373 Similarly, a recent study from the Tabula Muris Consortium which produced a single-cell transcriptomic 380 atlas of 23 tissues and organs across the Mus musculus life span, confirmed that microglia in the aged 381 (P540 and P720) brain are enriched for IFN-I-responsive genes and upregulate a similar set of genes 382 including Ifit3, Irf7, Isg15, Oasl2, Ifitm3, and Rtp4 [65] .The genes upregulated by the interferon-responsive 383 microglia clusters in both these studies are highly similar to those upregulated in our Phf15 KO cells under 384 basal conditions (see Figure 5A and C). Because ISGs can modulate inflammation [24] , it is possible that 385 interferon-responsive microglia could play a role in contributing to the inflammatory signature found in the 386 aged brain. Interestingly, among the set of downregulated genes in Phf15 KO cells at baseline and 6 hours 387 after LPS stimulation, is Myocyte Enhancer Factor 2C (Mef2C). Mef2C is an important checkpoint 388 inhibitor that restrains microglial activation in response to pro-inflammatory insults and is lost in brain 389 aging via IFN-I mediated downregulation [64,66] . Thus, an increase in Phf15 expression in microglia during 390 healthy aging could putatively work to counteract not only microglial activation but increased IFN-I in the 391 aged brain as well.

392
Notably, a recent study by Readhead et al.,[67] found that several virus species are commonly 393 present in the aged human brain. Among them, human herpesvirus 6A and 7 (HHV-6A and HHV-7) were 394 highly upregulated in the brain of AD patients and were found to modulate host genes associated with AD 395 risk, for example, Amyloid precursor protein (APP) processing. APP is the precursor molecule whose 396 proteolysis forms amyloid-β (Aβ) and formation of Aβ plaques has long been thought of as the driving 397 force behind Alzheimer's disease [62] . Aβ has more recently been found to have antimicrobial properties, 398 conferring increased resistance against infection from both bacteria and viruses [63] . APP is among the

404
Importantly, Phf15 also serves to repress baseline inflammatory output in the absence of immune 405 activation. Putatively, increases in Phf15 during healthy aging could help counteract brain inflammation 406 and protect brain health.

407
Future studies will determine the mechanism of action of Phf15. For example, the identity of its 408 binding partner proteins, its genome-wide binding sites and associated histone marks to determine the 409 specific gene regulatory regions it interacts with (e.g. active enhancers or promoters). Additionally, studies