P23 Acts as Functional RBP in the Macrophage Inflammation Response

Macrophages exert the primary cellular immune response. Pathogen components like bacterial lipopolysaccharides (LPS) stimulate macrophage migration, phagocytotic activity and cytokine expression. Previously, we identified the poly(A)+ RNA interactome of RAW 264.7 macrophages. Of the 402 RNA-binding proteins (RBPs), 32 were classified as unique in macrophages, including nineteen not reported to interact with nucleic acids before. Remarkably, P23 a HSP90 co-chaperone, also known as cytosolic prostaglandin E2 synthase (PTGES3), exhibited differential poly(A)+ RNA binding in untreated and LPS-induced macrophages. To identify mRNAs bound by P23 and to elucidate potential regulatory RBP functions in macrophages, we immunoprecipitated P23 from cytoplasmic extracts of cross-linked untreated and LPS-induced cells. RNAseq revealed that enrichment of 44 mRNAs was reduced in response to LPS. Kif15 mRNA, which encodes kinesin family member 15 (KIF15), a motor protein implicated in cytoskeletal reorganization and cell mobility was selected for further analysis. Noteworthy, phagocytic activity of LPS-induced macrophages was enhanced by P23 depletion. Specifically, in untreated RAW 264.7 macrophages, decreased P23 results in Kif15 mRNA destabilization, diminished KIF15 expression and accelerated macrophage migration. We show that the unexpected RBP function of P23 contributes to the regulation of macrophage phagocytotic activity and migration.


INTRODUCTION
As primary responding cells of the innate immune system, macrophages recognize pathogen components, like bacterial lipopolysaccharides (LPS) by Toll-like receptor 4 (TLR4). Different branches of TLR4 signaling pathways involve mitogen activated protein kinases (MAPK) and nuclear factor kappa B (NFκB) and ultimately induce inflammatory cytokine and chemokine expression, macrophage migration and phagocytotic activity (Zanotti and Kumar, 2002;Medzhitov and Horng, 2009;Smale, 2012;Vaure and Liu, 2014). Cytokines and chemokines are essential to coordinate cellular responses to infection, but their excessive synthesis disturbs the delicate balance between pro-and anti-inflammatory processes, leading to systemic capillary leakage, tissue destruction, and lethal organ failure (Zanotti and Kumar, 2002;Hotchkiss and Karl, 2003). LPSinduced changes in gene expression (Rutledge et al., 2011;Reynier et al., 2012) require downstream post-transcriptional checkpoints to safeguard faultless immune responses (Carpenter et al., 2014;Kafasla et al., 2014). Emerging evidence highlights the role of RNA-binding proteins (RBPs) in the regulation of both, mRNA turnover and mRNA translation to execute coordinated protein synthesis required for rapid and purposive macrophage functions (Ostareck and Ostareck-Lederer, 2019). Specific RBPs, namely Tristetraprolin (TTP), human antigen R (HUR), T-cell intracellular antigen 1 related protein (TIAR) and heterogeneous ribonucleoprotein K (HNRNPK) are known to modulate the activation and resolution of macrophage immune reactions [reviewed in Ostareck and Ostareck-Lederer (2019)]. Systematic analyses of LPS-controlled RBP-RNA interactions in macrophages revealed first insights into the complex regulatory RNA-protein interactome. Based on their domain composition RBPs exhibit affinities for specific AU-rich or U-rich motifs like TTP (Stoecklin et al., 2008;Kratochvill et al., 2011;Sedlyarov et al., 2016;Tiedje et al., 2016), HUR (Sedlyarov et al., 2016) and TIAR (Kharraz et al., 2016) or pyrimidine-rich sequences, like HNRNPK (Liepelt et al., 2014). They target mRNAs that encode signaling checkpoint proteins, control their stability and timely translation, and confer the regulation of macrophage activity.
In this report, we uncovered an unexpected function of P23 as a post-transcriptional regulator in untreated macrophages. The differential binding of P23 to the mRNA encoding kinesin family member 15 (KIF15) and its impact on Kif15 mRNA stability, suggests a function in the regulation of macrophage migration and phagocytosis. This novel mechanistic insight sheds light on the modulation of the inflammatory response and provides indications for molecular interventions.

Lysate Preparation
Total cell lysate preparation was performed according to Ostareck-Lederer et al. (2002).

Cytoplasmic Extract Preparation
Cytoplasmic extract was prepared as described in Liepelt et al. (2014).

Western Blot Assays
Western blot assays were performed as described (Naarmann et al., 2010) with specific antibodies (Supplementary Material). Images were acquired with the LAS4000 system (GE Healthcare Chicago, IL, United States) and quantified using ImageQuant (GE Healthcare).

RNA Sequencing
RNA samples from inputs or immunoprecipitations with P23 or Luciferase control antibodies (all samples from untreated RAW 264.7 cells or cells treated with LPS for 2 h) were analyzed with an RNA 6000 Nano Kit on a 2100 Bioanalyzer (Agilent Technologies; Santa Clara, CA, United States). Barcoded RNAseq libraries were prepared from either pulled-down or high quality total RNA samples (25 ng/sample) using the Illumina TruSeq total RNA Sample Preparation v2 Kit. Obtained libraries that passed the QC step were combined in equimolar amounts into pools of seven libraries; 8pM solution of each pool was loaded per lane of the Illumina sequencer HiSeq 2000 flowcell and sequenced bi-directionally (each read 50 bases long, altogether 100 bases) with the Illumina v3 SBS chemistry. Sequencing generated >60 million of paired sequence reads/library (∼450 million of paired sequence reads/lane).

Analysis of RNAseq Data
Trimming of low quality reads and adapters was performed using Trimmomatic v.0.33 (Bolger et al., 2014) with the following settings: ILLUMINACLIP:TruSeq3-PE-2.fa:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36. The data quality was evaluated using the FastQC software v0.11.2 before and after trimming. The mouse reference genome (version GRCm38/mm10) and gene annotation (Ensembl release 83) were downloaded from Ensembl. Mapping was performed using Tophat v.2.0.14 (Kim et al., 2013) with standard parameters and the additional annotation file. Quantification of unique read counts per gene was performed using HTSeq-0.6.1 (Anders et al., 2015) with default parameters except for stranded no. Differential expression analysis was performed in R using edgeR (Robinson et al., 2010). Initial assessment of replicates by pairwise distance and principal component analysis (PCA) led to removal of one replicate (LUC 2 h replicate III) as apparent outlier. Comparison of the remaining data yielded 52 RNAs that were specifically enriched in the P23 IP compared to LUC IP at the 0 h time point (adjusted p-value ≤ 0.05, Benjamini Hochberg correction, log 2 -transformed fold change >0) (Benjamini and Hochberg, 1995). In order to assess changes in P23 binding to the 52 RNAs after LPS treatment, means of TPM values from the replicates were used to calculate ratios for P23/LUC at 0 and 2 h LPS treatment (Supplementary Table S1).

RNA Motif Identification
Identified unbound RNA sequences were determined as follows: log2 FC < −0.2 and >0.1 in input 2 h over input 0 h were removed, as well as all transcripts with TPM 0.00 in input 0 and/or 2 h. Furthermore, transcripts without KEGG-ID and/or MGI description were removed. Then, a minimal TPM was set to 10, and the transcripts were filtered for a log2 FC to < −0.1 in inputs 2 h/0 h. Finally, transcripts with log2 FC > 0.00 for P23-0 h vs. LUC-0 h immunoprecipitation were removed. 3′UTR sequences for P23-interacting mRNAs as well as unbound mRNAs were retrieved from the Ensembl genome database (version 92). Sequences were filtered for the presence of a RefSeq-ID. If more than one sequence was identified, the sequence with the best transcript support level was chosen. Discriminative motif enrichment analysis in the P23interacting mRNAs was performed using MEME (Bailey et al., 2009). The 3′UTR sequences of unbound mRNAs were used as background input. The detected motif with the lowest p-value was resubmitted to MAST (Bailey and Gribskov, 1998) to rescreen all sequences for the presence of the motif. In order to assess the sequence composition of the Kif15 3′UTR, we moved a 30-nt sliding window along the Kif15 3′UTR (606nt, NM_010620.1) to count the frequency of U/G-only hexamers.

In vitro Transcription
Linearized plasmids were transcribed using the T3 or T7 polymerase MEGAscript kit (Thermo Fisher Scientific, AM1338 or AM1334, respectively) following manufacturer's protocols.

Expression of Recombinant Proteins
His-P23 was expressed and prepared as described (Liepelt et al., 2016).

Migration Assay
Cells transferred to Transwell chambers (Sigma-Aldrich, Costar, CLS3422) were subjected to starvation (1% FCS) for 24 h before they were exposed to LPS (100 ng/ml) at 10%FSC standard conditions. Following siRNA (Supplementary Material) transfection, cells were allowed to recover for 6 h under standard conditions prior to 16 h FCS starvation and LPS induction. Migrated cells were fixed on cover slips with 4% paraformaldehyde in PBS (Sigma-Aldrich, 33220), embedded in ProLong Gold Antifade Reagent with DAPI (Thermo Scientific, Carlsbad, CA, United States, P36931) and analyzed by fluorescence microscopy.

Immunofluorescence and Fluorescence in situ Hybridization
Immunofluorescence staining was essentially performed as described in De Vries et al. (2013)

Statistical Analysis
Experiments were recapitulated at least three times, statistical analysis was performed with one-way ANOVA, with significance levels defined as * p < 0.05, ** p < 0.01, *** p < 0.001.

Depletion of P23 Enhances Phagocytosis
Whereas expression of P23, which we identified as a new RBP in RAW 264.7 macrophages is not affected by 2 h LPS treatment, its poly(A) + RNA binding activity is reduced in response to LPS (Liepelt et al., 2016). In order to elucidate P23 RBP functions we asked whether the impact of LPS on RNA binding affects the phagocytotic activity of macrophages, a critical function in pathogen clearance (Stuart and Ezekowitz, 2005). To diminish P23 expression, P23-specific and control siRNAs, were transfected in RAW 264.7 cells 4 h prior to 24 h LPS treatment and labelled latex beads supply for 1 h ( Figure 1A). P23 was detected by Western blotting ( Figure 1B) and phagocytosed beads were analyzed and quantified by immunofluorescence microscopy (IF) ( Figures 1C,D). LPS induced phagocytosis was strongly elevated in P23 depleted macrophages compared to the control ( Figures 1C,D). The impact of P23 depletion on macrophage activity along with the declining binding of poly(A) + -RNA by P23 in response to LPS (Liepelt et al., 2016) motivated us to identify P23 target mRNAs and to explore if it exhibits functions in the posttranscriptional regulation of gene expression.

P23 Interacts Specifically With Kinesin 15 mRNA
Initially we set out to identify mRNAs that were differentially bound by P23 in untreated and LPS-induced macrophages. P23 was immunoprecipitated (IP) from cytoplasmic extracts of UV-crosslinked RAW 264.7 untreated macrophages (-LPS) and after 2 h LPS treatment (+LPS). A luciferase antibody served as control (ctrl.) for both conditions (Figure 2A). RNA sequencing (RNAseq) identified 52 RNAs as specifically enriched with P23 compared to the LUC IP at the 0 h time point (adjusted p-value ≤ 0.05, Benjamini Hochberg correction, log 2 -transformed fold change >0) ( Figure 2B). Of those 44 mRNAs possessed Ensembl genome database entries ( Table 1). In response to LPS their enrichment was reduced. The majority of encoded proteins represent transcription factors, cytoskeletal proteins and signaling factors (PANTHER) (Thomas et al., 2003;Mi et al., 2013) (Table 1, Supplementary Table S1). Considering the relevance of cytoskeletal remodeling for macrophage migration and phagocytosis (Harrison and Grinstein, 2002;Rougerie et al., 2013), kinesin mRNAs encoding cytoskeletal proteins KIF15, KIF1B and KIF13A, were selected for functional analysis. All three mRNAs were equally expressed in untreated and 2 h LPS-induced macrophages employed for RIP analysis ( Figure 2C). There binding specificity was validated by P23 IP from cytoplasmic extracts of untreated and LPSinduced macrophages ( Figure 2D). Importantly, specific enrichment of Kif15 mRNA from extracts of untreated cells was significantly diminished in extracts of LPS-induced cells ( Figure 2D). Enrichment of Kif1b mRNA was less pronounced and Kif13a mRNA did not exhibit specific interaction with P23. These results are consistent with the RNAseq analysis demonstrating the most distinct specific binding for Kif15 mRNA (Table 1).
Moreover, immunofluorescence and fluorescence in situ hybridization (IF-FISH) demonstrated an LPS-dependent decline in P23-Kif15 mRNA co-localization ( Figure 2E), consistent with a reduced P23-poly(A) + RNA interaction (Liepelt et al., 2016). These results initially indicated that the modulation of the P23-Kif15 mRNA might contribute to macrophage functional control.
To discern common sequence features of the 44 differentially P23 interacting mRNAs, discriminative motif enrichment analysis was performed for the 5′UTR, 3′UTR and open reading frame sequences (Bailey and Elkan, 1994;Bailey et al., 2009) (Figure 3). The search identified a U/G-rich motif  consisting of U/G hexamers (E-value: 7.0e-044) ( Figure 3A) in the 3'UTR of 38 mRNAs (86%), while a similar 3′UTR motif appeared in less than 30% of unbound control mRNAs ( Figure 3B, Supplementary Table S1). The motif search in the 5′UTR and ORF of the 44 P23 target mRNAs did not retrieve meaningful motifs (data not shown). We investigated RNA folds in the vicinity of the identified U/G-rich motifs that could potentially be involved in P23 binding, but due to the low number of sequences, we cannot draw reliable conclusions whether any such Minimal Free Energy (MFE) fold could potentially be involved in P23 binding (data not shown). The position of all detected motif instances in the 3′UTRs of those P23 target mRNAs with an p-value with less than 0.0001 is shown in the Supplementary Figure S1. Consistent with the P23 IP, a distinct U/G motif mapped to the Kif15 mRNA 3′UTR (E-value 6.7e −18 , p-value 7.74e −21 ) ( Figure 3C, Supplementary Table S1), whereas it aligned to less extent with the Kif1b mRNA 3′UTR (E-value 7.5e −5 , p-value 4.54e −9 ) and was absent from Kif13a mRNA (E-value 5.8e +0 ) (Supplementary Table S1).
In order to examine specific P23 binding to the U/G motif in vitro, we made use of recombinant P23 and 3′biotinylated transcripts coupled to streptavidin beads ( Figure 3D). As control we employed the well characterized ALOX15 mRNA 3'UTR U/C-repeat sequence (DICE) that is devoid of U/Gs, confers HNRNPK binding and mediates translational regulation (Ostareck-Lederer et al., 1994;Ostareck et al., 1997;Moritz et al., 2014). His-P23 strongly bound to U/G motif bearing Kif15 mRNA 3′UTR ( Figure 3D, lane 3), whereas no binding exceeding the empty beads background was detected for the DICE control ( Figure 3D, lanes 2 and 4). These results indicate that the U/G motif constitutes a P23 binding scaffold.

Kif15 mRNA and Protein Are Diminished in Lipopolysaccharides Activated Macrophages
The impact of LPS on the expression of P23 and KIF15 was addressed in time course analyses of both mRNA and protein expression (Figure 4). Efficient RAW 264.7 macrophage activation over 24 h was clearly reflected in the expression kinetics of cytokine mRNAs TNFα and IL1β, which were sharply increased after 2 h and subsequently declined, distinctive for LPS-driven TLR4 signaling ( Figure 4A; Liepelt et al., 2014). An increase of P23 mRNA until 6 h followed by a slight reduction ( Figure 4A), was also observed at the protein level ( Figure 4C). Interestingly, Kif15 mRNA exhibited a sharp drop after 6 h that was reversed until 24 h ( Figure 4A), in contrast to continuously declining KIF15 protein, whereas GAPDH and VINCULIN remained unaffected ( Figure 4C). Actb mRNA did not change significantly ( Figure 4A). The stability of Kif15 mRNA was only slightly diminished in both untreated and 2 h LPSinduced cells, however 6 h LPS treatment resulted in a distinct destabilization (t 1/2 3.5 h) ( Figure 4B, left panel), concurring with Kif15 mRNA reduction at that time point ( Figure 4A). Actb mRNA was only slightly affected ( Figure 4B, right panel). A comparable Kif15 mRNA halflife (t 1/2 5-6 h) was observed in differentiating mouse embryonic stem cells (Sharova et al., 2009). KIF15, also termed Kinesin-12, was originally described as motor protein implicated in mitotic spindle formation (Kashina et al., 1996;Scholey, 2009;Ferenz et al., 2010). Recent studies in neurons indicate functions of KIF15 in the regulation of cell morphology, axonal growth and branching (Buster et al., 2003;Liu et al., 2010), as well as dendrite differentiation and neuronal migration (Lin et al., 2012;Drechsler and Mcainsh, 2016;Feng et al., 2016).
To examine whether LPS-induced macrophage migration (Tajima et al., 2008) is linked to declining KIF15, we established a RAW 264.7 migration assay ( Figure 4D). Clearly, enhanced macrophage migration was significantly progressing until 6 h post LPS addition ( Figure 4D), when Kif15 mRNA stability and The sequence logo represents the information content (column high) and relative frequency of each nucleotide (relative height within the column) at each position of the motif. (B) Frequency of the U/G-rich motif identified by MEME in 44 P23 target mRNAs vs. 117 unbound mRNAs, which were equally expressed but did not exhibit P23 interaction. (C) Genome browser view of the Kif15 3'UTR (606 nt, NM_010620.1) with frequency profile for U/G-only hexamers (top) and zoom-in on the region harbouring the UG-rich motif identified by MEME (bottom). (D) 3'end biotinylated RNAs, Kif15 mRNA 3'UTR or ALOX15 mRNA 3'UTR DICE (ctrl.), were immobilized on streptavidin beads and incubated with His-P23 (n 3). Following elution, P23 was detected by a specific antibody. Statistical analysis was performed with one-way ANOVA, significance levels defined as ** p < 0.01. abundance declined (Figures 4A,B). These observations suggest that LPS entails alterations in Kif15 mRNA stability, which enable a dynamic macrophage migration response. These results are in agreement with the observed increased movement of KIF15depleted astrocytes (Feng et al., 2016).

P23 Depletion Destabilizes Kif15 mRNA and Activates Migration of Untreated Macrophages
The interaction of P23 with Kif15 mRNA in untreated macrophages, which strongly declined following LPS treatment (Table 1; Figures 2D,E), pointed us to a potential impact of P23 on Kif15 mRNA stability. To address this, P23 was depleted from untreated RAW 264.7 macrophages employing two specific siRNAs (#749, #1672) ( Figure 5). Interestingly, KIF15 protein and Kif15 mRNA were specifically reduced as a consequence of diminished P23 expression ( Figures 5A,B). Furthermore, Kif15 mRNA was destabilized when P23 expression was diminished ( Figure 5C), indicating a stabilization by P23 in untreated macrophages. A combination of P23 depletion with LPS treatment did not result in further KIF15 decline (data not shown), probably due to a reduced P23-Kif15 mRNA interaction after 2 h (Table 1; Figures 2D,E). Kif15 and Actb mRNAs were monitored by qPCR, normalized to exogenously added Luc mRNA. Statistical analysis was performed with one-way ANOVA, significance levels defined as ** p < 0.01 and *** p < 0.001.
Since declining KIF15 correlated with enforced macrophage migration ( Figures 4A,C,D), we integrated P23 as well as KIF15 depletion in a migration assay ( Figure 6A). Attenuation of P23 and KIF15 (siRNAs #544, #2481) led to declining KIF15 expression in untreated macrophages ( Figure 6B). Noteworthy, both P23 depletion by siRNAs, as well as the direct siRNA-mediated KIF15 reduction, led to a significantly elevated migration activity ( Figure 6C). Moreover, augmented pseudopodia formation of siRNA-transfected cells corresponds with enhanced macrophage migration ( Figure 6D). Again, a combination of P23 depletion with LPS treatment did not result in further KIF15 decline (data not shown).
The finding that P23 depletion attenuates Kif15 mRNA stability and protein expression ( Figures 5A,B,C, 6B) and furthermore drives untreated macrophage migration ( Figure 6C), supports the hypothesis that the dynamic formation of P23-Kif15 mRNA ribonucleoprotein complexes (mRNPs) contributes to macrophage mobility control.

P23 Interacts With Casein Kinase 2 and Lipopolysaccharides-dependent With HSP90
Besides its function as HSP90 co-chaperone, P23 possesses cytosolic prostaglandin E synthase activity, converting COX-1, but not COX-2-derived prostaglandin PGH2 to PGE2 . In activated cells, casein kinase (CK)2-dependent P23 phosphorylation is linked to increased enzymatic activity and PGE2 production, both facilitated by HSP90 interaction. HSP90 may act as a scaffold that brings P23 and CK2 in close proximity within the P23-CK2-HSP90 complex (Tanioka et al., 2003;Kobayashi et al., 2004;Hara et al., 2010). To address the potential interaction we immunoprecipitated P23 from cytoplasmic extracts of UV-cross-linked untreated RAW 264.7 cells (−LPS) and after 2 h LPS induction (+LPS) with two different P23 antibodies (JJ3 and JJ6, generated against fulllength recombinant P23). A luciferase antibody served as control (ctrl.) ( Figure  7A). Interestingly, JJ6 immunoprecipitated CK2 and HSP90, the latter more effective from LPS treated cells, but did not co-precipitate the Kif15 mRNA. In contrast, efficiently Kif15 mRNA precipitation was demonstrated from untreated cells, which declined after LPS treatment, employing JJ3. However, JJ3 did not coprecipitate CK2 and HSP90. (Figures 7A,B). This indicates that P23 acts as an RBP in untreated cells, while the interaction with HSP90 increases in LPS-stimulated cells. Furthermore, the results may suggest that different structural determinants confer P23-Kif15 mRNA interaction and binding of CK2 and HSP90.

DISCUSSION
Besides its function as a HSP90 co-chaperone (Felts and Toft, 2003), it was shown that P23 possesses cytosolic PTGES3 activity , which is elevated in response to LPS in peritoneal macrophages (Naraba et al., 1998). We identified P23 as a novel RBP that exhibits differential poly(A) + RNA binding in untreated macrophages and cells activated in response to LPS (Liepelt et al., 2016), with so far unknown implications for macrophage function. Interestingly, P23 depletion caused an increase in the phagocytic activity of LPS-induced RAW 264.7 macrophages (Figure 1). In context of the differential poly(A) + RNA interaction (Liepelt et al., 2016) this finding motivated us to explore a potential post-transcriptional impact on specific mRNAs. P23 target mRNA fractions immunoprecipitated from cytoplasmic extracts of cross-linked untreated and LPS activated RAW 264.7 cells (RIP) were analyzed by RNAseq and examined for differential P23 association in untreated and activated macrophages (Figure 2). Annotation of proteins encoded by the identified 44 mRNAs revealed an over-representation of transcription factors, cytoskeletal components and proteins involved in signal transduction ( Table 1, Supplementary  Table S1).
Cytoskeletal proteins are vitally important for macrophage migration and phagocytosis, both connected with extensive cytoskeletal remodeling (Rougerie et al., 2013). The motor protein KIF15 was originally described to participate in spindle formation (Kashina et al., 1996;Scholey, 2009;Ferenz et al., 2010). A diminished interaction of P23 with Kif15 mRNA in response to LPS is evident from co-precipitation and colocalization analyses (Figures 2D,E). The Kif15 mRNA decline between 2 and 6 h LPS treatment ( Figures 4A,B) suggests that in activated macrophages the loss of P23 association triggers the decay of Kif15 mRNA. The reduced amount of Kif15 mRNA that is accompanied by a diminished Kif15 mRNA stability in untreated macrophages, from which P23 is depleted ( Figures  5A,B,C), supports this hypothesis. Furthermore, after 2 h LPS treatment P23 binding to Kif15 mRNA is significantly reduced ( Figure 2D), which could explain why LPS treatment combined with P23 depletion does not cause further KIF15 decline (data not shown).
Discriminative motif enrichment analysis revealed that U/G motifs are specifically enriched in the 3′UTR of P23 target mRNAs, particularly the Kif15 mRNA 3′UTR ( Figures  3A,B,C, Supplementary Table S1), in contrast to unbound mRNAs. The importance of these U/G motifs for specific P23 binding became apparent in vitro RNA-interaction assays where recombinant P23 was bound by the Kif15 mRNA 3′UTR, but not a control RNA consisting of U/C repeats, devoid of U/G motifs ( Figure 3D). Not only for the HSP90 co-chaperone P23 functional poly(A) + RNA interaction was demonstrated (Liepelt et al., 2016), but also for HSPBP1 a HSP70 co-chaperone, which features an in vitro preference for pyrimidine-rich sequences (Mahboubi et al., 2020). Specifically, HSPBP1 is implicated in stress-granule formation control and interacts with related proteins G3BP1, HUR and TIA-1/TIAR (Mahboubi et al., 2020). In addition, the chaperone HSP70 independent of its function, exhibits in vitro affinity for AU-rich elements (ARE) and stabilizes the ARE containing VEGFA mRNA in vivo (Kishor et al., 2017).
Interestingly, a motor protein is encoded by Kif15 mRNA, which is differentially bound by P23 in untreated and LPSactivated macrophages. KIF15 depletion was shown to drive accelerated migration of rat cortical astrocytes (Feng et al., 2016) to contribute to decreased cell proliferation and an extended neuronal axon outgrowth in rodents (Liu et al., 2010). Furthermore, KIF15 mutations accelerate the axonal outgrowth during zebrafish neuronal development and regeneration (Dong et al., 2019). In accordance with the finding that KIF15 diminished conditions lead to an increase in microtubule mobility, these studies point to a function of KIF15 as a brake acting in microtubule transport (Kahn and Baas, 2016;Rao and Baas, 2018;Dong et al., 2019).
Consistently, with decreasing Kif15 mRNA and protein level ( Figures 4A,C) we observed an enhanced migration of activated RAW 264.7 macrophages ( Figure 4D) Therefore we hypothesized that the P23-Kif15 mRNP might contribute to macrophage migration control. P23 or KIF15 depletion from RAW 264.7 cells not exposed to LPS boosted macrophage migration ( Figure 6C) and the formation of podosomal protrusions ( Figure 6D). Similarly HSP70/90-organizing protein HOP/STI1 was found to contribute to neuroblast migration control (Miyakoshi et al., 2017).
Considering the prominent impact of macrophages on innate immune response control, it would be interesting to search for small molecules, which stabilize the P23-Kif15 mRNA interaction to inhibit macrophage migration as recently suggested (Gomez-Rial et al., 2020). Alternatively, the innate macrophages immune response could be primed (Deng et al., 2013) through P23-Kif15 mRNA interaction inhibitors.
P23 immunoprecipitation with two different antibodies, revealed that P23 acts as an RBP in untreated RAW 264.7 macrophages (Figures 2,7). In response to LPS its interaction with HSP90, but not CK2 is enhanced (Figure 7), consistently induction of PTGES3 activity was accompanied by an increase in PTGES3-HSP90 complex formation (Tanioka et al., 2003). In addition, our immunoprecipitation studies suggest that the Kif15 mRNA interaction and binding of CK2 and HSP90 are mediated by different structural determinants. P23 consists of two modules, a compact β-strand CS domain (CHORD-containing proteins and SGT1) providing a HSP90 binding surface and an unstructured C-terminal tail that is necessary for optimal chaperone activity (Weikl et al., 1999;Weaver et al., 2000). For disordered protein regions a RNA chaperone function has been described (Semrad, 2011) and it was shown that the FIGURE 8 | Model of Kif15 mRNA stability regulation by P23. In untreated macrophages P23 bound to 3'UTR U/G motifs stabilizes Kif15 mRNA. Newly synthesized KIF15 protein suppresses macrophage migration. Macrophage LPS induction leads to P23 release and Kif15 mRNA destabilization. Diminished KIF15 expression results in increased macrophage migration, as previously reported for rat cortical astrocytes (Feng et al., 2016). P23 exhibits cytosolic PTGES3 activity, which was demonstrated in LPS-activated peritoneal macrophages (Naraba et al., 1998). Facilitated by HSP90 interaction CK2-catalyzed PTGES3 phosphorylation is linked to increased enzymatic activity and PGE2 production (Tanioka et al., 2003;Kobayashi et al., 2004;Hara et al., 2010). In addition, P23 could also act LPS-dependent as HSP90 co-chaperone, since HSP90 activity is essential for the stabilization and maturation of proteins involved in NFκB activation and inflammation (Chatterjee et al., 2007;Chatterjee et al., 2008;Thangjam et al., 2014). Also in this context, the catalytic activity of CK2 contributes to the modulation of HSP90 chaperone activity (Miyata, 2009). unstructured C-terminal tail of HIV-1 VIF confers RNA binding (Sleiman et al., 2014).
To investigate a possible contribution of P23/PTGES3 to PGE2 production in LPS-treated RAW264.7 cells, we analyzed culture supernatants of untreated and P23-depleted cells in response to LPS (data not shown). The only moderate reduction of PGE2 production, which was determined could be explained by the activity of two additional PGE synthetases besides PTGES3: inducible microsomal mPGES-1 and the constitutively expressed PGES-2 (Jakobsson et al., 1999;Murakami et al., 2000;Tanikawa et al., 2002). While mPGES-1 is functionally coupled to inducible COX2, mPGES-2 functions equally well with both COX-1 and COX-2 (Hara et al., 2010). To answer this question, an in depth enzymatic analysis of the contribution of individual PGE synthetases in macrophages would be required.
Furthermore, P23 HSP90 co-chaperone function could also be affected by LPS as HSP90 activity is essential for the stabilization and maturation of proteins involved in NFκB activation and inflammation (Chatterjee et al., 2007;Chatterjee et al., 2008;Thangjam et al., 2014). Also in this context, the catalytic activity of CK2 contributes to the modulation of HSP90 chaperone activity (Miyata, 2009).
Our new findings related to P23 as an RBP are summarized in Figure 8 in the context of its known cellular activities. From our data it is conceivable that the function of P23 as an RBP in untreated macrophages is superseded by its enzymatic activity or HSP90 cochaperone function in LPS activated macrophages, where its RNA binding activity declines. Future studies will focus on the minimal RNA sequence features required for P23 binding, the identification of its RNA-interaction domain and RNA-binding control.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, GSE159619.