Oligodendrocytes remodel the genomic fabrics of functional pathways in astrocytes

We profiled the transcriptomes of primary mouse cortical astrocytes cultured alone or co-cultured with immortalized precursor oligodendrocytes. The experimental set-up (insert systems) prevented formation of gap junction channels but allowed free exchange of the two culture media. The study complements our previously published reports that the genomic fabrics of major functional pathways in oligodendrocytes are substantially remodeled by the proximity of non-touching astrocytes. Here, we present new analysis indicating that the transcriptomic landscape of astrocytes likewise changes significantly in the proximity of non-touching oligodendrocytes. The research was stimulated by the reported transcriptomic similarity between the brains of Cx43KO and Cx32KO mice, both substantially different from that of the Cx36KO mice. Since the three connexins are expressed in different cell types (Cx43 in astrocytes, Cx32 in oligodendrocytes and Cx36 in neurons), altogether these findings support the idea of a “panglial transcriptomic syncytium” in the mouse brain. Going further, our results suggest that integration in a heterocellular tissue modulates not only the expression profile but also the expression control and networking of the genes in each cell phenotype.


Introduction
The main glial cells, astrocytes and oligodendrocytes, together with ependymal cells and microglia, form what has been called the "silent majority" of the brain cells. They were considered "silent" because they do not generate action potential as neurons do. Nonetheless, they do signal among themselves by exchanging small (< 1kD) molecules via gap junction channels and release of gliotransmitters that bind the membrane receptors of the cells in the neighborhood. Astrocyte and oligodendrocyte interactions, among themselves and with each other, form the panglial network linking these glial cells throughout the brain. The panglial network provides metabolic support for neuronal activity, thereby impacting both constitutive brain functions such as sleep but also dynamic activities that include learning and cognition [1][2][3].
Gap junctions provide a two-way route for interaction between oligodendrocytes and astrocytes [4][5][6], where specific connexin proteins form the intercellular channels in the different cells. Astrocytes express Cx43 (Gja1) and Cx30 (Gjb6), whereas oligodendrocytes express Cx32 (Gjb1) and Cx47 (Gja12), and the two cell types communicate through heterotypic connexin pairing. None of these connexins expressed in astrocytes or oligodendrocytes is found in brain neurons (that culture 6 well insert systems (www.fishersci.ca), with cortical astrocytes in all plates. Oli-neu cells were placed in all the six inserts of the first system and only the culture medium used for Oli-neu cells (but without any cells present) in all the six inserts of the second system. Owing to astrocytes adhering to the bottom of the companion plate and Oli-neu cells confined to the insert, formation of hetero-cellular gap junction channels between astrocytes and oligodendrocytes [32] was prevented.
However, the astrocytes were exposed to the molecules released by the oligodendrocytes diffusing through the 0.4 µm pores of the inserts. The astrocytes were collected in separate labeled Eppendorf vials after 10 days in the Falcon systems. Four vials with well-developed astrocytes were selected from each of the two Falcon systems. Thus, we had four samples (labeled INS) with astrocytes co-cultured with Oli-neu cells and four samples (labeled CTR) with astrocytes in Oli-neu culture medium alone.
Microarray: Total RNA was extracted as previously described [35] with Qiagen RNeasy minikits separately from each of the eight selected vials from the two Falcon systems. RNA concentration before and after reverse transcription in the presence of Cy3/Cy5 dUTP was determined with NanoDrop ND 2000 Spectrophotometer and quality with Agilent RNA 6000 Nano kit in an Agilent 2100 Bioanalyzer. 825ng of differently (Cy3/Cy5) labeled biological replicas were hybridized 17h at 65°C with Agilent G2519F unrestricted AMADID Release GE 4x44k 60mer two-color mouse gene expression microarrays using the "multiple yellow" strategy. The chip (4 microarrays) was scanned with an Agilent G2539A dual laser scanner at 5μm resolution in 20-bit scan mode (>10 5 dynamic range) and primary analysis performed with (Agilent) Feature Extraction 11.6 software.
All corrupted spots or with foreground fluorescence less than twice background fluorescence in any of the eight samples were eliminated from the analysis. Data were normalized using our standard algorithm alternating intra-and inter-array normalization to the median of the background-subtracted fluorescence. Spots probing the same transcript were grouped into redundancy groups. Agilent mouse 4x44k microarrays used in this study hybridizes 30,175 distinct transcripts, out of which 22,657 are probed by single spots. The largest redundancy groups (13 spots) probed the genes: Abcc5 (ATP-binding cassette, sub-family C (CFTR/MRP), member 5), Cpne4 (copine IV), Csf1 (estrogen receptor 1 colony stimulating factor 1), Esr1 (estrogen receptor 1), Mapk1 (mitogen-activated protein kinase 1), Oprm1 (opioid receptor, mu 1), P2rx3 (purinergic receptor P2X, ligand-gated ion channel, 3) and Socs2 (suppressor of cytokine signaling 2).
Data Analysis: Profiling four biological replicas of each condition produces with adequate statistical power three independent measures for each transcript: i) average expression level, ii) variability of transcript abundance and iii) expression coordination with each other transcript [36].
The rarely used analysis of expression variability provides information about the degree to which homeostatic mechanisms limit range of transcript abundance and the analysis of expression coordination allows assessment of interactions within gene networks that underlie functional pathways. We report here the astrocyte genes that were up-or down-regulated, exhibited stricter or looser expression control and were differently networked when the oligodendrocytes are close by.
These pathways were selected for the following reasons: -Calcium signaling (hereafter denoted by CAS) is evolutionary the oldest, yet most common way by which a wide diversity of cells communicates to each other [19]. Change in calcium signaling is a major modulator of the glia cell behavior [39].
-Chemokine signaling (CS) between astrocytes and oligodendrocytes, most likely the main crosstalk in our experiment, is important for glial development and stimulating regeneration and repair [41].
-The thyroid hormone signaling pathway (TH) was chosen because astrocytes are thought to be the main regulator of thyroid hormone in the brain and T3 is a main driver of oligodendrocyte maturation [42,43].
-NOD-like receptor signaling pathway (NOD) was chosen because of its role in cognition, anxiety and activation of the hypothalamic-pituitary-adrenal axis [44].
-The actin cytoskeleton (AC), is an elaborate cytoplasmic protein structure central in determining cell and organ size and morphology, intracellular transport and cell division [45].
-Autophagy (AU) is a major degradation pathway, essential in maintaining astrocyte function [46] -Cell-cycle (CC) is expected to be one of the most dependent pathway on the cellular environment.
-The circadian rhythm (CR) -increasing evidence indicates that astrocytes are very important players in the regulation of circadian rhythms [47].
-Even though the experimental set up did not allow formation of hetero-cellular gap junction channels, the very proximity of the oligodendrocytes might have an effect on the expression level and networking of the gap junction (GJ) pathway in astrocytes.  hits that result from an arbitrary fixed cut-off (such as 1.5x) [48].

Relative Expression
Pathway regulation: The regulation of a given pathway was analyzed from the perspective of both percent of genes that were significantly regulated (using the above criterion of the absolute individual gene fold-change cut-off) and the Weighted Pathway Regulation [36]: In (3), pi is the p-value of the heteroscedastic (two tails, unequal variance) t-test of the means equality in the two conditions.
REVs in "condition" CTR or INS Pearson correlation coefficient of genes i and j log expression levels The top gene (highest GCH) was termed the Gene Master Regulator and is expected to be the most influential for the preservation of the cell phenotype,

Overview
Raw and processed gene expression data were deposited in the publicly available website https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE109035. In total, 18,891 unigenes were adequately quantified in all 8 samples, out of which almost 20% were significantly regulated in astrocytes cultured in the proximity of oligodendrocytes (INS) compared to astrocytes cultured alone (CTR). Alterations of the expression level are presented as percentages of the up-and down-regulated genes, and as the Weighted Pathway Regulation (WPR). The individual gene cut-off (CUT) ranged from 1.06x for TGF-beta activated kinase 1/MAP3K7 binding protein 1 (Tab1) from NOD, up-regulated by 1.25x, to 3.61x for guanylate cyclase 1, soluble, alpha 2 (Gucy1a2) from GJ, significantly upregulated by 35.94x. Using the uniform 1.5x absolute fold-change would result in 4% false hits and neglect 6% significant regulations as a consequence of sample variability.
Table S1 in the Appendix lists the genes whose >1.5x absolute fold-change did not meet the individual CUT criterion ("false hits" for the uniform 1.5x fold-change cut-off). Tables S2 and S3 in the Appendix list the genes considered as significantly up-and down-regulated according to our criterion, although for some of them was below the 1.5x traditionally used as a cut-off.
For the entire transcriptome, the percentage of down-regulated genes is balanced by that of the up-regulated (9.63% down-vs. 9.89% up-regulated, down/up ratio = 0.97). However, the Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 6 April 2020 doi:10.20944/preprints202004.0053.v1 percentages of the down-and up-regulated genes are quite dissimilar tor individual pathways, the most notable example being the cell cycle pathway where the ratio of down to up regulated genes is 31x. The bias indicates that oligodendrocyte proximity had a profound inhibitory effect on this pathway in astrocytes.
Note in Fig. 1(b) that REV was higher in co-cultured astrocytes for all gene groups except those encoding the CS and GJ pathways and was statistically significant for ALL, AC, AU, CAS, PA and TH groups. Figure 1(c) indicates statistically significantly higher median GCH of the synaptic receptors, gap junction, autophagy and chemokine signaling. GCH values were significantly lower for the circadian rhythm pathways, while the differences of the other pathways were not statistically significant.  rhythm pathways. Graphs below pathways present the expression ratios (together with the respective individual gene fold-change cut-offs, both negative for down-regulation) of significantly altered genes. Each figure also shows the GCH scores of the pathway genes in the two conditions.  Fig. 1(c), considering the entire pathway, the median GCH increased by over 17% in INS with respect to CTR.

Significantly regulated cell-cycle genes
In Fig. 3, Anapc2 (anaphase promoting complex subunit 2), Mad2l2 (mitotic arrest deficient-like 2) and Tgfb1 (transforming growth factor, beta 1) were the only significantly upregulated genes (however for both the fold-change was less than 1.5x).

Significantly regulated genes responsible for the circadian rhythm
In the circadian rhythm pathway ( Figure S2 in the Appendix) we found as significantly up-regulated only F-box and leucine-rich repeat protein 13 (Fbxl13) and as down-regulated: cryptochrome 1 photolyase-like (Cry1), period circadian clock 2 (Per2), protein kinase, AMP-activated, alpha 2 catalytic subunit (Prkaa2).

Regulation of signaling pathways
Figures 4-6 present the regulation of Ca 2+ -, NOD-like receptor and thyroid hormone signaling pathways. The graphs below each pathway show the expression ratios and the individual cut-offs for the pathway genes that can be considered as significantly regulated, and the GCH scores of the regulated and other important genes. Where appropriate, we have shown also the genes whose variance was so high that even though they exceeded the traditional 1.5x in gene expression ratio, the differences were not statistically significant ("false hits").

Oligodendrocyte proximity remodels the integration of astrocytes with neighboring, synaptically coupled neurons
As illustrated in Figure

Cellular environment remodels gene networks
We found that, in addition to regulating numerous individual genes, oligodendrocytes proximity had a major impact on the gene networks. in the oligodendrocytes cultured alone (Oli -Ast) or with non-touching astrocytes in the neighborhood (Oli + Ast). Gja1 coordination partners are shown for astrocytes but not for oligodendrocytes which do not express Cx43.

Discussion
In two previously published papers, we have shown that astrocyte-conditioned medium is a major regulator of gene expression in oligodendrocytes even in the absence of cytosol-to-cytosol communication via gap junction channels connecting these two cell types [29,30]. In the present study, we analyzed whether oligodendrocyte-conditioned medium changes significantly the astrocyte transcriptome. Together with our studies on brains of Cx43KO, Cx32KO and Cx36 mice [7][8][9][10][11], these data on astrocytes and oligodendrocytes cultured alone and co-cultured with each-other show the transcriptomic integration of the brain glia.
Our experimental results show that glial integration persists even in the absence of direct astrocyte-oligodendrocyte communication via gap junction channels The regulation of genes that modulate the neurotransmission suggests that neurons may be also part of the transcriptomic integration. Until now, no gap junction channel was found between neurons and astrocytes or oligodendrocytes in the mouse brain. However, formation of functional neuron-glia gap junction channels has so far been identified between somata of peripheral sensory ganglia neurons and surrounding satellite glial cells [54]. There are also reports of glia-neuronal gap junctions in C.
elegans [55]. Nevertheless, integration of neurons with glial cells can be carried out by molecules released by one cell type binding membrane receptors of other cell type as reported by many authors (e.g. ). It can be also acquired via transfer of exosomes [59].
Microarray data were analyzed from several, complementary perspectives, considering all three independent expression features that can be determined from studies incorporating four biological replicas. The study was empowered by advanced analytical approaches. For instance, as listed in Supplementary Table S1, our method eliminated the "false hits" (absolute fold-change over 1.5x but below CUT). However, we have determined that some of the "false negatives" were actually significantly regulated (absolute fold-change less than 1.5x but over CUT), listed in Supplementary Tables S2 and S3. Moreover, the traditional percentage of significantly regulated genes was complemented with the Weighted Pathway Regulation (WPR). WPR weighs the contribution of each regulated gene with respect to a cut-off tailored for it by considering the technical noise and biological variability. Of note in Fig. 1a is that, from the perspective of the WPR score, the gap junction (GJ) pathway was more affected than the others (WPR = 43.3), indicating that Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 6 April 2020 doi:10.20944/preprints202004.0053.v1 Oli-neu proximity exerted major impact on hetero-cellular communication even in the absence of cell-to-cell contact.
In the case of CC, the much larger percentage of down-regulated genes suggests a slow-down of the mitotic progression of astrocytes in the vicinity of oligodendrocytes that we have observed but have not yet quantified. In several other transcriptomic studies we found an overall reduction of gene REV in cells and tissues collected from subjects with various diseases (epilepsy [60], experimental autoimmune encephalomyelitis [61], kidney cancer [62]) compared to healthy counterparts. REV was also significantly lower in tissues from animals subjected to various stresses (microgravity [63], chronic hypoxia ) or genetic manipulations (e.g. [7,33]. The consistency of these findings suggest that the overall higher expression variability seen in astrocytes co-cultured with Oli-neu cells indicates that the proximity of oligodendrocytes provide an environment that is closer to that of their physiological state than when astrocytes are cultured alone. proteins [67,68]. The much higher ranking of Pdcd7 (GCH = 48.13 in INS astrocytes compared to GCH = 2.14 in CTR astrocytes), highlights the importance of controlling the astrocyte proliferation [69] during brain development and formation of the normal cellular environment (that contains also precursor oligodendrocytes). The prominence of Tubb3 in the GJ pathway of CTR astrocytes confirms its role in cytoskeletal adaptation of microtubules in astrocytes cultured without any other cells [70], while that of Sos1 in astrocytes co-cultured with oligodendrocytes shows its prevalence with regard to cell proliferation and viability [71]. Interestingly, all three cell-cycle genes from the Ccnh.Cdk7.Pea15a complex [72]  is promoted to the top position in this pathway (GCHINS = 12.01), while the other two genes of the complex stayed closed to their initial place: 1.45 (Cdk7), 1.65 (Pea15a)). The high position of Ccnh in an environment closer to that of the brain can explain why the Ex8+49T>C variant of this gene is associated with increased risk of glioma [73].

Relative Expression
We found that several members of the gap junction pathway related genes (including Gjb2, encoding Cx26) were differentially expressed in astrocytes in the presence of Oli-neu cells ( Figure 2).
However, to our surprise, the expression of the main connexin gene, Gja1, was not affected, although the astrocytes should have sensed the presence of the oligodendrocytes in the neighborhood. This result is surprising because the oligodendrocytes responded to the presence of astrocytes by increasing the expression of Gjc2 (encoding Cx47), the Cx43 partner in heterocellular gap junction channels [74], by 9.70x. We determined the up-regulation of Gjc2 by re-analyzing the previously reported microarray data on Oli-neu with the same set up [29,30]. It should be noted that increased expression levels for gap junction proteins are not always matched with changes in coupling. There is variation in astrocyte-astrocyte coupling as well as astrocyte-oligodendrocyte coupling. Intra-astrocyte gap junctions (a.k.a reflexive or autaptic gap junctions) seems to be common but has barely been studied. Given the preceding considerations and that Cx43 expression in astrocytes is very high in comparison to most other brain proteins, we predict that the non-significantly changed Cx43 expression is compatible with increased oligodendrocyte Cx47 (and presumably resulting in higher astrocyte-oligodendrocyte coupling). Nevertheless, the overexpression of Gjb2 in astrocytes can also be related to the up-regulation of Gjc2 in oligodendrocytes, confirming the potential coupling of the encoded connexins [75].
Interestingly, by regulating the astrocyte transcriptome, as illustrated in Fig.7, the proximity of oligodendrocytes modulates also the bidirectional interactions of astrocytes with the inter-neuronal synapses [76] that are essential for the development of cortical circuits [77].
The coordination patterns of Panx1 illustrate the degree to which gene networks can be remodeled in response to environmental factors. Panx1 has mostly synergistic coordination with actin cytoskeleton genes in astrocytes cultured alone, whereas it is antagonistically coordinated with these genes when cocultured with oligodendrocytes (Fig.8a). Conversely, oligodendrocyte actin genes are largely independent of Panx1 when cultured alone but become synergistically coordinated with Panx1 when cultured with astrocytes (Fig.8d). The opposite consequences of coculture on Panx1-coordinated cytoskeleton related genes in astrocytes and oligodendrocytes might reflect fundamental divergence in maturation or/and response to inflammatory stimuli.
A distinct type of network remodeling is exemplified by the mostly antagonistic coordinations of Gja1 with circadian rhythm genes in cultured alone astrocytes, with almost no coordination for Panx1. In coculture, antagonistic and synergistic coordinations with Gja1 are approximately equal, while each coordination is matched by a similar coordination with Panx1. Thus, these two maxi channel proteins act independently on circadian rhythm genes of astrocytes alone but become cooperative and may compensate for one another when in the presence of oligodendrocytes.
Interestingly, we reported that the coordination patterns of Gja1 and Panx1 with the whole brain transcriptome of wild type mice were highly (90.8%) similar [9]. A recent study [83] indicated that astrocytes can drive circadian related gene cycling in neurons of the suprachiasmiatic nucleus. Astrocyte gap junctions have been shown to affect and be affected by wakefulness [84]. Our results show that co-culturing substantially increased the coordination of Panx1 with circadian rhythm genes in the astrocytes but has practically no consequences for the oligodendrocytes.
The perivascular end-feet of astrocytes have been implicated in the uptake of thyroid hormones, especially thyroxine (T4) which is the vastly predominant, but mostly inactive form of thyroid hormone in the blood. After T4 is transported through brain endothelial cells it is thought to be mainly brought into astrocytes by transporters with varying specificity for T4. Once taken into astrocytes, T4 is converted to the highly active form triiodothyronine (T3) by iodothyronine deiodinase type 2 (DIO2) with DIO2 mRNA expression largely exclusive to astrocytes as reviewed in [85]. Taken together, the preceding aspects of thyroid handling point to astrocytes as key uptake and distribution cells in the brain. Additionally, oligodendrocyte maturation and key myelin production gene expression is strongly regulated (promoted) by T3 [86,87] Therefore, thyroid hormone handling pathway is important to oligodendrocytes and their precursors leading us to examine our data for changes astrocyte thyroid pathway genes in response to the presence of non-contacting Oli-neu cells. The presence of Oli-neu cells decreased expression of Slc16a10 (Mct10) but did not produce a change in the other two transporters expressed in astrocytes: Slo1c1 (OATP1C1) and Slc16a2 (MCT8). More interestingly, Dio2 was upregulated in the presence of Oli-neu cells. Tbc1d4 is upregulated in astrocytes cultured in the presence of Oli-neu cells and this GTPase increases surface expression of glucose transporters [88]. These results may indicate a shift towards an astrocyte phenotype that may produce activated thyroid hormone and bring in additional glucose. It is interesting to speculate that the presence of Oli-neu cells produces gene expression changes in astrocytes that would support myelination through increased production of activated T3 and support of the high metabolic demands of myelinating oligodendrocytes.

Conclusions
The major caveat of this study is that the oli-neu cells are not the natural but the immortalized precursor oligodendrocytes. However, our results clearly indicate that the cellular environment plays an integrative role by modulating the expression level, the expression control and the Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 6 April 2020 doi:10.20944/preprints202004.0053.v1 expression coordination of the genes in each and every cell phenotype of a heterogeneous tissue like brain. Because of this, one should be very cautious when extending the observations from homocellular cultures to the behavior of the same cell phenotypes within a heterocellular tissue.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1. Table S1: Genes whose >1.5x absolute fold-change did not meet the individual CUT criterion. Red/green background of the expression ratio indicates not significant (false) up-/down-regulation. Table S2: Genes considered as significantly up-regulated although their absolute fold-change was below the traditional 1.5x. Table S3: Genes considered as significantly down-regulated although their absolute fold-change was below the traditional 1.5x. Figure S1: Presence of non-touching precursor oligodendrocytes regulate the actin cytoskeleton (AC) pathway. Figure