Effect of Neuroprotective Magnesium Sulfate Treatment on Brain Transcription Response to Hypoxia Ischemia in Neonate Mice

MgSO4 is widely used in the prevention of preterm neurological disabilities but its modes of action remain poorly established. We used a co-hybridization approach using the transcriptome in 5-day old mice treated with a single dose of MgSO4 (600 mg/kg), and/or exposed to hypoxia-ischemia (HI). The transcription of hundreds of genes was altered in all the groups. MgSO4 mainly produced repressions culminating 6 h after injection. Bio-statistical analysis revealed the repression of synaptogenesis and axonal development. The putative targets of MgSO4 were Mnk1 and Frm1. A behavioral study of adults did not detect lasting effects of neonatal MgSO4 and precluded NMDA-receptor-mediated side effects. The effects of MgSO4 plus HI exceeded the sum of the effects of separate treatments. MgSO4 prior to HI reduced inflammation and the innate immune response probably as a result of cytokine inhibition (Ccl2, Ifng, interleukins). Conversely, MgSO4 had little effect on HI-induced transcription by RNA-polymerase II. De novo MgSO4-HI affected mitochondrial function through the repression of genes of oxidative phosphorylation and many NAD-dehydrogenases. It also likely reduced protein translation by the repression of many ribosomal proteins, essentially located in synapses. All these effects appeared under the putative regulatory MgSO4 induction of the mTORC2 Rictor coding gene. Lasting effects through Sirt1 and Frm1 could account for this epigenetic footprint.


Introduction
Magnesium appears to be a potent tool for the prevention of cerebral palsy and resulting disabilities when administered as MgSO 4 to women in danger of preterm delivery before 33 completed weeks of gestation (GWs) [1][2][3][4]. Historically, MgSO 4 used as a tocolytic did not prove to have efficiency in retarding delivery, but retrospectively it has shown neuroprotection effects at low doses [5,6]. The Mg 2+ ion has innumerable biological targets e.g., as co-factor in nucleic acid biology, protein synthesis, enzymatic systems, or proteinprotein interactions [7]. Its putative effects resulting in fetal neuroprotection are therefore very numerous, and its actual modes of action remain elusive. Given at neuroprotective (low) doses, MgSO 4 has no deleterious side effects, which is a requirement for prophylactic treatment. However, owing to its large spectrum of putative effects, it appears necessary to strengthen the evidences of its harmlessness at neuroprotection doses, since high doses of MgSO 4 have adverse effects at the cellular and whole-organism levels [8,9].
Pediatricians have detected diverse effects of prenatal MgSO 4 in newborn presentation and behavior. Reports on cardiorespiratory function were somewhat controversial, showing no effect or depressant effect of antenatal magnesium [10,11]. It reduces heart rate but stabilizes blood pressure and reduces cerebral blood flow [10,12]. Overall, prenatal MgSO 4 was shown to decrease the need to resort to intubation in very preterm births [13]. Finally, although MgSO 4 supply has not been assessed as protective against bronchopulmonary dysplasia, hypomagnesemia was described as a risk factor [14].
Magnesium is mainly restrained in intracellular stores and circulating levels after bolus injection rapidly decrease due to near saturation of renal reabsorption pumps in physiological magnesemia conditions [15]. Antenatal magnesemia is high in the preterm and bolus MgSO 4 resulted in a peak observed at delivery followed by a moderate hypermagnesemia during the first 3 days of life in very preterm infants exposed in utero [16,17]. In neonatal mice, the clearance of Mg 2+ appeared even faster since no hyper-magnesemia remained 12 h after a 600-mg/kg bolus injection [18]. Thus, the presence of MgSO 4 in the mother and fetus organisms after a low bolus administration is fleeting at the blood level, whereas lasting neuroprotection is expected and in fact observed in the subsequent hours and days [15,19]. It is possible that intracellular stores are rapidly sequestrated, although this requires active and saturable mechanisms. One may therefore consider the possibility that transient Mg 2+ elevation initiates some processes that predispose the fetus against prospective noxious experiences, such as preterm birth and related consequences, somewhat as may proceed a pre-conditioning. Indeed, in animal studies, Mg 2+ supplementation prevents maternal inflammation and offspring cerebral injuries [20][21][22].
It is likely that prenatal Mg 2+ supplementation exerts neuroprotection via systemic and central effects. It has been suggested that direct effects at the fetal brain level occur through calcium antagonism and/or glutamate N-methyl-D-aspartate type glutamate receptor (NMDA) blockade [23]. Inflammation in the mother or consecutive hypoxia-ischemia is a major trigger of neonatal brain lesions and magnesium exerts anti-inflammatory activity in human or animal neonates [24,25]. MgSO 4 pretreatment provided neuroprotection in the Rice and Vannucci experimental paradigm modeling hypoxia-ischemia (HI), and in the excitotoxic model of ibotenic acid injection [26,27]. In these models, the effects of MgSO 4 do not account for maternal-mediated effects, suggesting direct effects in the neonate brain and improved delayed behavior [18,28]. In the 5 day-old mouse brain, HI provoked an early onset of coordinated gene inductions (repressions) 3 h to 12 h after insult, affecting transcription regulation at the RNA-polymerase-II complex, inflammation, cell death, and angiogenesis. In parallel, convergent gene repression affected many genes encoding for cholesterol metabolism enzymes and synaptic actors [29]. The causal factors of long-term deleterious consequences among these regulations are likely to be explored in the field of inflammation as well as its interaction with development [30][31][32].
Magnesium may affect a plethora of factors and interferes with HI [7,33,34]. In light of our previous description of the strong effects of HI on the transcription pattern in P5 mice, the Rice and Vannucci model appeared to be a simple approach for an overview of mechanisms recruited in sole treatment with MgSO 4 , and of its modulation of HIeffects [29]. Neonatal mice at 5 days post-birth (P5) are usually considered a suitable model to explore brain development and injuries of human preterm born between 28 and 34 GW [35,36]. A further demonstration that the insult sensitivity of P5 and P10 mouse brains accurately mimicked preterm neonates and term infant brains was obtained through magnetic resonance imaging and histology after HI using the Rice and Vannucci model. In fact, P5 mice only showed white matter hypersignal and microglial activation. Moreover, age-dependent memory or social behavior defects were observed in adults [28,37].
The period of special white matter vulnerability is the recommended therapeutic window of MgSO 4 use (before 33 GW) for preterm prophylactic neuroprotection [38,39].
This study had dual goals-(i) the description of the proper effects of MgSO 4 single bolus injection at P5 in mice, investigated at the transcription level in short delays and at the behavior level in grown-up mice, (ii) the identification of the MgSO 4 interference with HIinduced transcription effects and with spontaneous developmental transcription evolution.

Results
Transcriptome analyses were performed on the basis of three independent experiments. Experiment 1 compared the effects of single MgSO 4 injection to phosphate buffered saline (PBS) injection, Experiment 2 compared the effects of MgSO 4 or PBS injection prior to HI, and Experiment 3 compared the effects of HI compared to control mice ( Figure 1). The comparison of the 3 experiments was performed using the combination of two-color and one-color analyses of hybridization data detailed in Section 4.3 ( Figure 2). fact, P5 mice only showed white matter hypersignal and microglial activation. Moreover, age-dependent memory or social behavior defects were observed in adults [28,37].
The period of special white matter vulnerability is the recommended therapeutic window of MgSO4 use (before 33 GW) for preterm prophylactic neuroprotection [38,39]. This study had dual goals-(i) the description of the proper effects of MgSO4 single bolus injection at P5 in mice, investigated at the transcription level in short delays and at the behavior level in grown-up mice, (ii) the identification of the MgSO4 interference with HIinduced transcription effects and with spontaneous developmental transcription evolution.

Results
Transcriptome analyses were performed on the basis of three independent experiments. Experiment 1 compared the effects of single MgSO4 injection to phosphate buffered saline (PBS) injection, Experiment 2 compared the effects of MgSO4 or PBS injection prior to HI, and Experiment 3 compared the effects of HI compared to control mice ( Figure 1). The comparison of the 3 experiments was performed using the combination of two-color and one-color analyses of hybridization data detailed in Section 4.3 ( Figure 2).   MgSO4 modulation of HI response to HI in P5 brain. All three experiments were analyzed independently (first step analyses in two-color co-hybridization experiments), providing lists of genes noted ①, ②, and ③. The genes detected in experiment 1 provided the List A of proper effects of MgSO4 single injection, to be submitted to pathway and gene ontology analyses. Coincident observations in ① and ② were extracted and split into two sub-series (④ and ⑥). List B constitutes the effects of MgSO4 seen only in the HI context (List B1), MgSO4 reversion of HI effects (List B2), and MgSO4 amplification of HI effects (List B3), combining two-color and one-color analyses. The subseries of List A (④)-containing genes showing induction after MgSO4 injection in both control (①) and HI-exposed (②) animals regrouping the genes affected by MgSO4 that were unchanged after HI-was subtracted from List B. List C contains genes affected by HI in control as well as in MgSO4-pretreated mice. Black and Red arrows indicate two-color and one color insertions; blue arrows indicate subtractions in destination boxes.

Effects of Single-Injection MgSO4 on Transcription
A total of 1411 genes exhibited significant alteration of expression, at least at one time point from 90 min to 24 h after injection. Two-fold more repressions (n = 973) than inductions (n = 447) were recorded. Nine genes appeared in the induction and repression lists; with seven at the 6 h time point, revealed by separate probes, suggesting alternative splicing (Atp1a3, Aut2, Celf4, Kalrn, Psd3, Tenm4, Tln2), and two exhibiting successive opposite variation (Dmtn, Septin3) ( Table S1A). The time course study revealed a wave of transcription inductions and repressions peaking after 6 h and including 1322 genes (93.7% of the total list in 24 h) ( Figure 3A). The median fold change (FC) amplitude was significantly lower for inductions (1.68; Q1 = 1.64−Q3 = 1.80) than for repression (−1.82; Q1 = −2.02−Q3 = −1.65; p < 0.0001; according to the Mann-Whitney test) ( Figure S1A). of HI response to HI in P5 brain. All three experiments were analyzed independently (first step analyses in two-color co-hybridization experiments), providing lists of genes noted 1 , 2 , and 3 . The genes detected in experiment 1 provided the List A of proper effects of MgSO 4 single injection, to be submitted to pathway and gene ontology analyses. Coincident observations in 1 and 2 were extracted and split into two sub-series ( 4 and 6 ). List B constitutes the effects of MgSO 4 seen only in the HI context (List B 1 ), MgSO 4 reversion of HI effects (List B 2 ), and MgSO 4 amplification of HI effects (List B 3 ), combining two-color and one-color analyses. The subseries of List A ( 4 )-containing genes showing induction after MgSO 4 injection in both control ( 1 ) and HI-exposed ( 2 ) animals regrouping the genes affected by MgSO 4 that were unchanged after HI-was subtracted from List B. List C contains genes affected by HI in control as well as in MgSO 4 -pretreated mice. Black and Red arrows indicate two-color and one color insertions; blue arrows indicate subtractions in destination boxes.

Up Keywords
The highest enrichments of UP_keywords (seKws) specific to list A were noted for "synapse," "neurogenesis," and "cell junction" (Table 1). Looking at neurogenesis one could note many genes coding transcription factors, membrane associated scaffolding and cue proteins, neurotrophin transmission associated proteins. The cell junction list included all seKws associated to synapses and had 20% of genes in common with the cytoskeleton list. Enrichment of these seKws, also exhibiting very low false discovery rate (FDR) values, clearly indicated that MgSO4 targeted the transcription in neurons, inducing the repression of membrane proteins and especially of synaptic proteins. Many glutamatergic synapse elements exhibited high amplitude repression (Grin1, Shank3, Dlgap3, Dlgap1, and Slc17a7 coding the Vglut1 transporter). In addition, Grina3, Dlg2, Gria1, and Grik5, also appeared to be repressed, but with less amplitude, whereas Grik3 was induced. Repression of Ache and the muscarinic receptor M1 genes were also recorded (see details in Table S1B).

Up Keywords
The highest enrichments of UP_keywords (seKws) specific to list A were noted for "synapse," "neurogenesis," and "cell junction" (Table 1). Looking at neurogenesis one could note many genes coding transcription factors, membrane associated scaffolding and cue proteins, neurotrophin transmission associated proteins. The cell junction list included all seKws associated to synapses and had 20% of genes in common with the cytoskeleton list. Enrichment of these seKws, also exhibiting very low false discovery rate (FDR) values, clearly indicated that MgSO 4 targeted the transcription in neurons, inducing the repression of membrane proteins and especially of synaptic proteins. Many glutamatergic synapse elements exhibited high amplitude repression (Grin1, Shank3, Dlgap3, Dlgap1, and Slc17a7 coding the Vglut1 transporter). In addition, Grina3, Dlg2, Gria1, and Grik5, also appeared to be repressed, but with less amplitude, whereas Grik3 was induced. Repression of Ache and the muscarinic receptor M1 genes were also recorded (see details in Table S1B).

GO Analysis
DAVID ® Gene Ontology identified 44 seGO-terms; 10 biological processes (BP), 24 cell components (CC), and 10 molecular functions (MF) subclasses. Five seGOterms were related to neurons, synapses nervous system development, and postsynaptic density, exhibited the highest significant p-values and showed among the highest number of genes involved (Table 1). With a lower significance seGO-terms also focused on cell junctions, neuron projection or ion binding, and with decreasing incidence on, kinasephosphorylation activities, the cytoskeleton, the regulation of transcription, cell signaling, and the Golgi apparatus ( Figure 3B, Supplementary Table S1B). Those genes associated with seGO-terms mainly underwent repressions.

Pathway Analyses KEGG Pathways
DAVID ® analysis allowed the extraction of 45 significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (seKpaths) at the Expression Analysis Systematic Explorer (EASE) p-value threshold < 5 × 10 −2 (and 14 others out of scope) (Supplementary Table S1C). Of note the lowest p-value recorded were associated to axon guidance (mmu04360, p = 1.66 × 10 −5 in Bonferroni test), thyroid hormone signaling, neurotrophin signaling, cAMP signaling, and pathways in cancer (including genes involved in extracellular matrix-cell regulation and insulin signaling ( Table 1). Less significant seKPaths were related to neurotransmission systems (glutamatergic synapse, GABAergic synapse, retrograde endocannabinoid signaling, cholinergic synapse, and many peptide signaling and membrane transduction pathways ( Figure 3C, Supplementary Table S1C).

Ingenuity Pathway Analysis (IPA) Pathways
Pathway extraction using IPA was based on a strictly different approach and allowed a much broader view as the strategy did not depend on previous validations, as for the KEGG pathways, which are relatively limited in mice. IPA pathways were extracted at all five time points on the corrected log ratio of the 1492 probes affected by MgSO 4 injection at least at one time point. A total of 13 significantly enriched IPA pathways (seIPaths) exhibited Fisher's Exact Test p-value < 1 × 10 −3 , of which 11 had z-score < −2 (Table 1, Figure 3C).
Se-IPath enrichments were noted early (1 h 30 min) and exceeded what could be expected from the gene kinetics record ( Figure 3D). The seIPaths exhibiting the most significant p-values, together with more specific genes included were synaptogenesis, reelin signaling in neurons, and axonal guidance (Table 1). Only one seIPath had a positive p-value (PTEN signaling) indicating an activation of the pathway, although this was exclusively based on gene repressions (Table S1D).

IPA ® Upstream Regulators
Putative upstream regulators (pURs) refers to key regulators, on the basis of convergences between reports in the literature of any relationships between this gene/protein expression/activity and the enrichment of these anticipated effects in an experimental set of data. A z-score, significant when >2 (or <−2), indicates the sense of the regulation. When a pUR was not affected at its transcription level, one must conclude that the experimental condition had downstream effects on its translation or on the function of the coded protein. In the following text, pURs are identified using protein symbols, whereas gene symbols (in italics) refer to transcription observations. Furthermore, the terms induction and repression are devoted to transcriptions observation, whereas the effects of pUR are called activation and inhibition.
Only two pURs (Mnk1 and Frm-1) showed very highly significant p value (<1 × 10 −7 ), although MgSO 4 did not affect their transcription. Upstream analysis revealed inhibition of Mnk1 and activation of Fmr-1, but the putative effects of MgSO 4 on downstream effects converged toward essentially the same canonical pathways, i.e., the inhibition of synaptogenesis, reelin-, and glutamate receptor signaling ( Figure 4, Table S1E). The mechanism of the interaction of Mg 2+ with Mnk1 and Fmr-1 remains elusive. Among the 30 genes putatively downstream to Mnk1 in the Mg group, three genes affected in HI showed reversed effects in MgHI (Map1b, Mapre2, Vim), 26 genes affected by Mg no longer appeared in MgHI, and the four genes remaining exhibited inverted regulation (Bag3, Cplx2, Trim32, and Ttc3). As a polyribosome-associated mRNA-binding protein with specific activity in brain development and synapse plasticity, Fmr-1 is an indicator of Mg tropism toward protein translation in neurons. Sirt1 was another, although less significant pUR, with lasting occurrence, a positive z-score, and p < 1 × 10 −4 at 24 h, which is involved in histone de-acetylation (Table S1E).

Synthesis
MgSO4 injection mainly provoked gene repressions, which predominated even more significantly in the coordinated effects revealed by seKPaths, seIPaths, and pUR, indicating a convergent expression of activity toward two groups of functions ( Figure 3B-D, Table S1).
Both seKPaths and seIPaths were related to nervous system development (axon guidance, thyroid hormone signaling, synaptogenesis signaling, reelin signaling, netrin signaling). pUR analysis pointed out Mknk1 and Fmr1 as MgSO4 targets for the cooperative inhibition of synaptogenesis, reelin, and glutamate signaling. Repressions largely predominate in these pathways, suggesting a transient brake in nervous system development ( Table 1). pUR analysis suggested Frm1 as an MgSO4 target involved in the inhibition of synaptogenesis as well as implicating Sirtuin-1 as a potential site for Mg epigenetic activities.
A large group of seIPaths showed effects of MgSO4 effects on chemical signaling by hormones (insulin, corticoliberin, renin-angiotensin, GnRH) and neurotransmission modulators (NO, Ephrin, opioids, endocannabinoids). Transduction systems were also affected (signaling by Rho GTPase and CREB). All were slowed down (z-scores < −2) at several time points indicating reduced activity.
MgSO4 affected the expression of 299 developmentally regulated genes, mainly anticipating ontogenetic increases, and to a lesser extent ontogenetic decreases of expression As a polyribosome-associated mRNA-binding protein with specific activity in brain development and synapse plasticity, Fmr-1 is an indicator of Mg tropism toward protein translation in neurons. Sirt1 was another, although less significant pUR, with lasting occurrence, a positive z-score, and p < 1 × 10 −4 at 24 h, which is involved in histone de-acetylation (Table S1E).

Synthesis
MgSO 4 injection mainly provoked gene repressions, which predominated even more significantly in the coordinated effects revealed by seKPaths, seIPaths, and pUR, indicating a convergent expression of activity toward two groups of functions ( Figure 3B-D, Table S1).
Both seKPaths and seIPaths were related to nervous system development (axon guidance, thyroid hormone signaling, synaptogenesis signaling, reelin signaling, netrin signaling). pUR analysis pointed out Mknk1 and Fmr1 as MgSO 4 targets for the cooperative inhibition of synaptogenesis, reelin, and glutamate signaling. Repressions largely predominate in these pathways, suggesting a transient brake in nervous system development (Table 1). pUR analysis suggested Frm1 as an MgSO 4 target involved in the inhibition of synaptogenesis as well as implicating Sirtuin-1 as a potential site for Mg epigenetic activities.
A large group of seIPaths showed effects of MgSO 4 effects on chemical signaling by hormones (insulin, corticoliberin, renin-angiotensin, GnRH) and neurotransmission modulators (NO, Ephrin, opioids, endocannabinoids). Transduction systems were also affected (signaling by Rho GTPase and CREB). All were slowed down (z-scores < −2) at several time points indicating reduced activity.
MgSO 4 affected the expression of 299 developmentally regulated genes, mainly anticipating ontogenetic increases, and to a lesser extent ontogenetic decreases of expression (Supplementary data). Very few MgSO 4 effects were antagonistic to ontogeny (Supplementary Figure S2).

Long-Term Behavioral Effects of Neonatal MgSO 4 Single Injection
Regarding the effects of MgSO 4 single injection on the expression of synaptic genes and spontaneous ontogenic evolution, there may be a concern that it could affect the brain development and result in lasting behavioral alterations. MgSO 4 effects were examined in sex-separated groups in several tests, in comparison with positive control animals treated at P5 with MK-801 or ketamine. Detailed effects and statistics are given in Tables S4-S7.

Spontaneous Locomotor Activity
The activity estimated by the running distance in the free wheel showed a significant effect of sex (females were more active than males) in the two-way ANOVA. The Bonferroni post-test showed an increased activity in females treated by ketamine ( Figure 5A). No differences between acutely treated neonates (BPS, MK-801, or MgSO 4 ) nor in Repeat-edly treated animals (PBS or MgSO 4 ) were observed. It is noteworthy that repeated injection, which required daily handling in pups, had a lasting reduction effect on activity, although the nature of the treatment had no effect ( Figure S3A,B, Table S4). (Supplementary data). Very few MgSO4 effects were antagonistic to ontogeny (Supplementary Figure S2).

Long-Term Behavioral Effects of Neonatal MgSO4 Single Injection
Regarding the effects of MgSO4 single injection on the expression of synaptic genes and spontaneous ontogenic evolution, there may be a concern that it could affect the brain development and result in lasting behavioral alterations. MgSO4 effects were examined in sex-separated groups in several tests, in comparison with positive control animals treated at P5 with MK-801 or ketamine. Detailed effects and statistics are given in Tables S4-S7.

Spontaneous Locomotor Activity
The activity estimated by the running distance in the free wheel showed a significant effect of sex (females were more active than males) in the two-way ANOVA. The Bonferroni post-test showed an increased activity in females treated by ketamine ( Figure 5A). No differences between acutely treated neonates (BPS, MK-801, or MgSO4) nor in Repeatedly treated animals (PBS or MgSO4) were observed. It is noteworthy that repeated injection, which required daily handling in pups, had a lasting reduction effect on activity, although the nature of the treatment had no effect ( Figure S3A,B, Table S4).

Open Field Animal Spontaneous Activity
Open field exploration was recorded for 30 min in the open field and was analyzed in acute or repeated treatment groups using two-way ANOVA (sex/treatment). Sex differences were observed but no effects of treatments on the distance run through the open field or on the exploration time of the central area ( Figure S3C,D, Table S5).

MK-801 Evoked Hyperlocomotion
MK-801 acute injection at the end of 30 min exploration induced locomotor activation in adults in all neonatally treated groups. It induced running activity and stereotypies but reduced vertical exploration ( Figure S4, Table S6). The post-test revealed that neonatal MK-801 increased vertical exploration in males (p = 0.0116) and had a tendency to produce the same effect in females (p = 0.0921) ( Figure 5B). An interaction of sex with treatment was observed in stereotypies. The post-test revealed an amplification of stereotypies induced by acute MK-801 in males (p = 0.0018) whereas it had a tendency to reduce them in females ( Figure 5C).

Social Interaction
MgSO 4 had no influence on adult behavior nor did MK-801 ( Figure 5D,E). Global activity evaluated on the basis of total distance run or rearing, did not reveal differences related to acute nor repeated treatments although sex differences were observed (not shown). The time spent in the contact zone was selected as a raw index of social interaction and did not show an effect of treatment (Table S7). Tenuous treatment effects were detected using the comparison of inter-zone crossings to evaluate the activity induced by the apparition of the congener. Treatments had no effect on raw indexes according to twoway ANOVA (on sex and treatment), in acute or repeated treatment groups whereas sex effects were constant (Supplementary Table S7). The distribution of crossings did not differ between groups during habituation whereas it revealed a significant difference in females during the interaction period ( Figure 5D,E, Supplementary Table S7). Neonatal ketamine affected activity provoked by an unknown congener in females (Chi 2 = 31.15 df3 , p = 0.0003). The reduced number of entries evoked by the congener did not necessarily reflect a poor interest for the congener since time spent in the contact zone was not reduced in these ketamine-treated females (Supplementary Table S7)). Thus, the experimental paradigm appeared to be discriminant, and did not show any effect of MgSO 4 on this behavior, even after repeated neonatal administrations.

Synthesis
MK-801 or ketamine, used as positive controls, induced sex-dependent effects in the different behavior tests. This selectivity of MK-801 and ketamine among tests and sex reinforces the choice of the tests as they appeared to be relevant for distinct behavioral traits. In addition, they showed ketamine effects that were independent of glutamate NMDA receptor subtype blocking, since they were distinct from the effects of MK-801. The question of long-term effects of single neonatal ketamine exposure is once again addressed. Of note, neonatal acute or even repeated MgSO 4 neonatal single injection did not alter any of this series of independent behavioral traits ( Figures S3 and S4).

General Observations
The modulation by MgSO 4 of the effects of HI did not result from the differences observed in experiment 2; therefore, we must take into account the effects of HI unaffected by MgSO 4 and reciprocally the MgSO 4 effects observed after HI ( Figure 2).
The MgHI effects did not appear as a superposition of the separate effects of MgSO 4 and of HI. The major proportion of regulated genes was specific to one of the three experimental conditions (two-color analyses of experiments 1-3). The list of MgHI vs.
HI effects (list 2 ) showed only 145 genes in common with that of Mg effects (List 1 ), and 89 genes in common with that of HI effects (List 3 ) (20.1% and 12.8% of 2 , respectively). A similar observation was made when comparing MgHI vs. Ctrl (List 7 ) in one-color analysis ( Figure 6A).
The modulation by MgSO4 of the effects of HI did not result from the differences observed in experiment 2; therefore, we must take into account the effects of HI unaffected by MgSO4 and reciprocally the MgSO4 effects observed after HI ( Figure 2).
The MgHI effects did not appear as a superposition of the separate effects of MgSO4 and of HI. The major proportion of regulated genes was specific to one of the three experimental conditions (two-color analyses of experiments 1-3). The list of MgHI vs. HI effects (list ②) showed only 145 genes in common with that of Mg effects (List ①), and 89 genes in common with that of HI effects (List ③) (20.1% and 12.8% of ②, respectively). A similar observation was made when comparing MgHI vs. Ctrl (List ⑦) in one-color analysis ( Figure 6A).  MgSO 4 pretreatment induced more negative regulations than enhancement of expressions due to HI in the P5 mouse brain (513 vs. 180 in list 2 ). It modified the transcription response as early as 3 h after HI, i.e., before showing proper transcription effects (peak at 6 h), indicating that before magnesium had transcriptional effects, it affected effector systems (e.g., enzyme activities), which in turn largely modified the whole transcription response to HI. Much fewer genes continued to have an altered expression 12 h after MgHI, and most showed a reduction of an effect already seen at 3 h, indicating the short-term biological response ( Figure 6B). MgSO 4 also evoked de novo HI-inductions (n = 545) and repressions (n = 943) revealed by one-color analyses (List B 1 ), confirming that MgSO 4 rather than preventing the HI response, oriented the HI transcription response in a new direction ( Figure 6C).
The assembly of MgSO 4 effects in HI (List B) results from the summation of de novo inductions (repressions) compared to HI (List B 1 ), as well as reversions (List B 2 ), and amplifications of HI effects (List B 3 ). From these the effects of MgSO 4 that were unchanged in the MgHI groups (list A 1 ) must be subtracted. Thus, the analysis was complex, as the different lists resulted from different kinds of data extraction. List B describes MgSO 4 modulation of the HI transcription response. A total of 1964 genes in MgHI (842 inductions, 1091 repressions and 31 biphasic evolution or putatively spliced) showed effects that differed from HI effects. Repressions predominated (55.6%), much more than in HI effects in list 3 (30.7%), but somewhat less than in List A of MgSO 4 proper effects (68.5%) ( Figure S5A). The comparison of List B with HI effects (List 3 ) revealed a minority of genes regulated in both conditions (171 on 723 in 3 ), and 1488 genes affected de novo ( Figure S5B).
A small proportion of HI effects (18.2%) were unaffected by MgSO 4 pretreatment (List C) ( Figure 6C). Even based on the restrained gene series presented in list C, David ® analysis showed the persistence of major HI effects on BP seGOterms (response to cytokine, transcription from RNA polymerase II promoter, cytokine chemotaxis, DNA-templated transcription, positive regulation of gene expression, and negative regulation of neuron apoptotic process) (see below).
Four seKws were enriched from both List A and List B, referring to basic mechanisms (phosphoproteins, alternative splicing, Ubl-conjugation and transcription regulation). These few seKws included hundreds of genes in A or B, but about 10% of these genes were common to Lists A and B, 7-8% of which even behaved in opposite direction. Many of the seKws from List B were also enriched, due to de novo effects (List B 1 ) and consisted of repressions (Table S2B).
Reversion by MgSO4 of HI effects (Lis B 2 ) exhibited seKws in the fields of the major effects of HI; i.e., inflammation, immunity, and regulation of transcription by RNApolymerase II. Whether these effects underlie neuroprotection by MgSO 4 requires further studies.
Altogether, these observations indicate that MgSO 4 did not so much change the nature of the HI transcription response, but rather it profoundly modified tissue adaptation to injury. GO Analysis DAVID ® extraction of seGO-terms from List B identified eight CC-seGO-terms and two MF-seGO-terms ( Table 2). The terms "mitochondrion" and "ribosome" exhibited high counts and the lowest p-values (<1 × 10 −20 ), although their enrichment were modest (Table S2C). Most of these terms appeared from list B1 of de novo effects, indicating that MgSO 4 pretreatment allowed HI to affect mitochondria and ribosome related genes. The question of whether effects at these sites could account for neuroprotection by MgSO 4 would be worth investigating. Nevertheless, and even if they were scarce, enriched BP and MF seGO-terms do not indicate major modification in cell functions.

IPA Biostatistical Analyses IPA Pathways in MgHI Mice
IPA ® analysis of List B could not be performed as this list resulted from lists obtained using two-color and one-color data, in distinct Genespring ® standardizations. However, separated approaches toward the MgSO 4 interference with HI transcription were carried out in two-color MgHI vs. HI (List 2 ) and one-color MgHI vs. Ctrl (List 7 ) analyses ( Figure 7).
Only one seIPath (reelin signaling in neurons) reached the 1 × 10 −4 p-value from List 2 at 3 h (Supplementary Table S2E). Several pathways (i.e., synaptogenesis signaling pathways) had p-values below 1 × 10 −2 and significant z-scores (<−2), although they included modest gene numbers. Of note reelin and synaptogenesis signaling pathways were also enriched by the effects of MgSO 4 alone. No seIPaths were extracted from list 2 at the 12-h time point.
IPA ® analysis of MgSO 4 effects on HI response evaluated based on one color analysis (List 7 ) revealed only one seIPath with p < 1 × 10 −4 (the sirtuin signaling pathway), although with no clear indication of regulation sense (z = 1). Less significant, although clearly inhibited, was EIF2 signaling which included 26 genes (p = 1.82 × 10 −4 , z = −3.051). Repression (more than 75%) affected 15 ribosomal proteins, indicating an inhibition of protein translation in MgHI conditions. This inhibition of translation appeared de novo after HI and MgSO 4 pretreatment since 28 of these genes plus four others, associated to EIF4 signaling, were not affected in the absence of pretreatment (HI in list 3 ). Other de novo pathways were detected, i.e., the lasting repression of glutathione redox reactions I, and melatonin-and CDK5-signaling (Table S2E).

IPA ® Upstream Regulators in MgHI Conditions
The identification of pUR in the experimental condition combining MgSO 4 treatment and HI would not reveal a target, as would be the case in single pharmacological conditions. Rather, it would indicate putative hubs affected by the combined effects of the two stimuli which would not represent a simple summation of effects. No pUR could be extracted from the two-color analyses of MgHI vs. HI effects at 3 h or at 12 h, possibly because of small sizes of the lists. pUR extraction from one-color analysis of MgHI vs. Ctrl (List 7 ) identified three pURs at highly significant p values < 1 × 10 −4 ; two pURs were activated (Rictor and Creb-1) and one was inhibited (Tfrc) ( Figure 8A, Table S2E).

IPA ® Upstream regulators in MgHI Conditions
The identification of pUR in the experimental condition combining MgSO4 treatment and HI would not reveal a target, as would be the case in single pharmacological conditions. Rather, it would indicate putative hubs affected by the combined effects of the two stimuli which would not represent a simple summation of effects. No pUR could be extracted from the two-color analyses of MgHI vs. HI effects at 3 h or at 12 h, possibly because of small sizes of the lists. pUR extraction from one-color analysis of MgHI vs. Ctrl (List ⑦) identified three pURs at highly significant p values < 1 × 10 −4 ; two pURs were activated (Rictor and Creb-1) and one was inhibited (Tfrc) ( Figure 8A, Table S2E).

Tracking of Magnesium Targets
We compared, seIPaths in nine separate conditions-MgSO4 effects at 1 h 30 min, 3 h, 6 h and 12 h, HI effects at 3 h and 12 h; MgHI vs. HI effects at 3 h and 12 h (List ②) and MgHI vs. Ctrl effects using the one-color approach (list ⑦) (Figure 7).

Effects of MgSO4 Injection
Fmr1 and Mnk1 were two pURs among the effects of MgSO4 (Figure 4 and Figure  S6). In Mg, 23 of the 30 genes under the Mnk1putative upstream inhibitor were specifically associated to this pUR. They were not significant pUR in HI, but Fmr1 and Mnk1 showed

Tracking of Magnesium Targets
We compared, seIPaths in nine separate conditions-MgSO 4 effects at 1 h 30 min, 3 h, 6 h and 12 h, HI effects at 3 h and 12 h; MgHI vs. HI effects at 3 h and 12 h (List 2 ) and MgHI vs. Ctrl effects using the one-color approach (list 7 ) (Figure 7).

Effects of MgSO 4 Injection
Fmr1 and Mnk1 were two pURs among the effects of MgSO 4 (Figure 4 and Figure S6). In Mg, 23 of the 30 genes under the Mnk1putative upstream inhibitor were specifically associated to this pUR. They were not significant pUR in HI, but Fmr1 and Mnk1 showed inverted z-score and nearly significant p-values in the one-color MgHI analysis. Reciprocal activation of Fmr-1 as a pUR in Mg appeared far less significant in MgHI at 3 h and was possibly reversed in one-color analysis (Tables S1E and S2E). Owing to its recurrent and highly significant enrichment, it appears to be a highly likely target on which MgSO 4 had an inhibitory effect, downstream to transcription. However, as almost no pUR downstream gene were affected in HI, the regulation on Mnk1 and Fmr1 by MgSO 4 has probably no direct effect on neuroprotection.
Casr, Sirt1, and IL4 appeared to be lowly significant pURs in mice treated solely with MgSO 4 . All three were highly significant potential hubs in HI but these putative activities did not remain in MgSO 4 -pretreated animals (Tables S1E and S2E,F).

pURs Extracted from the Panel of Genes Affected after HI
IPA ® extraction allowed us to identify 48 pURs with p-values below 1 × 10 −4 (Table S2F). A high number of pURs in HI had disappeared in MgHI group. As the majority of pURs in HI were related to inflammation, acute reaction and transcription one could expect MgSO 4 effects to point in these directions, although we could not identify specific targets at this point. Only three of these 48 pURs appeared in the Mg group (Crem, Casr and Sirt1), and did not remain in MgHI. However, it seems unlikely that these effectors are MgSO 4 targets since the majority of their downstream-targeted genes had expression patterns that were unchanged in MgHI, compared to HI conditions (31/54) (Table S2E,F).

Effects of Combined MgHI Conditions
Many seIPaths appeared to be affected with different kinetics, in opposite sense in Mg or HI groups (Figures S8 and S9). Among them, several showed Mg-like or HI-like effects in MgHI, whereas others disappeared. The inversion of HI activation in MgHI was noted for oxidative phosphorylation, ephrin receptor signaling, and B-cell receptor signaling (Figure 7). Abolitions of HI activation were noted for mTOR signaling, hepatic fibrosis signaling, and ILK signaling. The highly significant inhibition of the super-pathway of cholesterol biosynthesis observed 12 h after HI also disappeared in MgHI.
Several apparent discordances appeared between MgHI vs. HI effects and MgHI vs. Ctrl effects, likely due to the different timing of evaluation, and the occurrence of transient responses (Figure 7). Indeed, the rapid bio-availability of Mg 2+ and its rapid clearance may have short-term effects that do not last or are overwhelmed by HI effects. Thus, comparing pURs in the nine lists, we could classify the observations in four groups. Only one gene/protein (Rictor) appeared as a de novo pUR in MgHI. Three pURs exhibited opposite z-scores in the Mg and HI groups, resulting in annihilation or HI reversion effects in MgHI (Casr, Ppara and Fmr1). In addition, larger series with less clear patterns regrouped pUR in HI for which MgSO 4 prevented (n = 68) or did not affect (n = 48) their significance in MgHI (Table S2E). These series contain many genes involved in effectors in inflammation, cell death, transcription, and DNA repair, indicating that MgSO 4 in fact reduced these effects, although the analysis pointed out few specific putative Mg targets for these effects. Five entities, however, retained our attention-Rictor, Infg, Il4, Ppara, and Sirt1.
Rictor expression was unaffected in all conditions. mTORC2, coded by Rictor appeared to be a de novo pUR, based on de novo gene regulations in MgHI conditions. It had the lowest p-value (4.70 × 10 −6 ), a high z-score (4.885), and the highest number of downstream targeted genes (n = 37 genes) ( Figure 8A). The majority of these genes (32) participated in few canonical pathways-mitochondrial dysfunction; oxidative phosphorylation, protein ubiquitination, mTOR signaling and regulation of elongation by Eif2 and Eif4. These functions clearly confirmed the aforementioned seKws, seGO-terms, KEGG, and IPA pathways observations, indicating that in MgHI condition, protein translation and oxidative metabolism were inhibited through the repression of ribosomal proteins, ubiquinone oxydoreductases, cyclooxygenases, and proteasome proteins ( Figure 8A).
Sirtuin1, appeared very early (1 h 30 min) and showed lasting activation of pURs in MgSO 4 treatment based on the regulation of 32 putative downstream genes, mainly repressions. An opposite activating pattern was observed 12 h after HI. Sirtuin1 did not appear to be a pUR in the MgHI one-color list, but the Sirtuin signaling pathway was a significant pathway extracted from the de novo list 7 . These observations indicate that MgSO 4 influenced the expression of many genes downstream to Sirtuin1, possibly at the protein level, resulting in a change in HI induction in pretreated animals ( Figure 8B).
Infg was the most significant pUR in HI conditions at 3 and 12 h as a potential activation factor (z > 4.4). Although a high z-score was measured in the MgHI one-color analysis, it did not remain a significant pUR ( Figure 9A). A majority (n = 30) of the 42 genes downstream of Ifng in HI at 12 h were no longer affected in MgHI conditions. Many other factors involved in inflammation processes showed similar patterns (Table S2E). pUR comparison did not allow to extract a unique target for MgSO 4 in lowering the HIinduced inflammation process, but the global tendency allowed us to conclude that MgSO 4 prevented inflammatory effects of HI in the P5 mouse brain as a possible cause of its neuroprotection effects. Sirtuin1, appeared very early (1 h 30 min) and showed lasting activation of pURs in MgSO4 treatment based on the regulation of 32 putative downstream genes, mainly repressions. An opposite activating pattern was observed 12 h after HI. Sirtuin1 did not appear to be a pUR in the MgHI one-color list, but the Sirtuin signaling pathway was a significant pathway extracted from the de novo list ⑦. These observations indicate that MgSO4 influenced the expression of many genes downstream to Sirtuin1, possibly at the protein level, resulting in a change in HI induction in pretreated animals ( Figure 8B).
Infg was the most significant pUR in HI conditions at 3 and 12 h as a potential activation factor (z > 4.4). Although a high z-score was measured in the MgHI one-color analysis, it did not remain a significant pUR ( Figure 9A). A majority (n = 30) of the 42 genes downstream of Ifng in HI at 12 h were no longer affected in MgHI conditions. Many other factors involved in inflammation processes showed similar patterns (Table S2E). pUR comparison did not allow to extract a unique target for MgSO4 in lowering the HI-induced inflammation process, but the global tendency allowed us to conclude that MgSO4 prevented inflammatory effects of HI in the P5 mouse brain as a possible cause of its neuroprotection effects.  Il4 appeared to be an inducer in HI (z = 2.0333 h after HI), and a possible inhibitory pUR of MgSO 4 effects (z < −2 from 3 to 12 h), but did not reach significance in MgHI conditions (11 of 16 genes downstream to Il4 affected in HI were unaffected in MgHI) ( Figure 9B). The Il4 gene did not show transcription regulation in any experimental.
Ppara appeared to be a lowly significant activating pUR in HI at 12 h (p = 4.02 × 10 −4 , z = 2.302), and did not appeared significant in MgHI at 3 h (p > 5 × 10 −2 , z = −1.938) ( Figure 9C). It retained our attention however, as it may be a hub for the inhibition of the cholesterol biosynthesis pathway in HI at 12 h. (Figure 9C). Two of these 5 genes (Idi1 and Msmo1) were activated in MgHI, while Mvd, Mvk, and Nsdhl were unaffected. These divergent effects preclude MgSO 4 effect on Ppara.

HI Effects Insensitive to MgSO 4
seGO-term extraction from List C enabled us to evaluate the putative effects of HI that were insensitive to MgSO 4 pretreatment. Gene inductions, predominant in HI, are related to transcription by RNA-polymerase-II, inflammation and angiogenesis as previously described [29]. Reciprocally, the repression of genes encoding cholesterol biosynthesis enzymes was observed (Idi1, Mvk and Mvd, Msmo1 and Nsdhl), although the significance threshold was not reached. A minority of genes affected by HI in list 3 appeared to be unaffected in List C-166/499 inductions (33%) and 66/224 repressions (29%). As far as gene ontology is valid for partial lists, seGO-terms related to inflammation were absent in C, with the exception of the response to cytokine and monocyte chemotaxis. Otherwise the seGO-terms related to RNA polymerase II identified from List 3 remained in List C (Table S8).

Discussion
Neonate mice at P5 show forebrain development and sensitivity to insults which was somewhat representative of human brain development and vulnerability around 28-32 GWs, and may be used as experimental surrogates [24,[35][36][37]. However, one must keep in mind that putative MgSO 4 neuroprotection in humans also results from activities in the mother, at the placental level, and that the purpose of the present study was restricted to MgSO 4 effects on the neonate. Acute single MgSO 4 600-mg/kg administration in P5 mice evoked transient transcription effects, mainly repressions, for a short period culminating 6 h after injection. The gene ontology and pathways extraction pointed out the effects of MgSO 4 on synapses and brain development ( Figure 10). The administration of 600-mg/kg MgSO 4 in P5 mice induces a transient peak concentration 30 min after administration. Blood concentration was 1.5 mM after 6 h, a concentration detected for longer periods in preterm born from mother having received 4-6 g bolus plus 1-2 g/h maintenance [18,40]. In mice pups, the concentration returned to basal level within 24 h, while it often remained above 1 mM in human, leading us to consider the 600-mg/kg bolus as a low MgSO 4 dose [17,40]. Under these conditions MgSO 4 did not induce apoptosis [28].
Concerning the transcription effects of MgSO 4 alone, one could note the convergence toward synapse differentiation and function in either, seKws, seGO-terms and pathways identified in Mg groups. Two highly significant and durable pURs were identified as potential targets of Mg 2+ : Mnk1 and Frm-1. These two proteins have activity in protein translation. Mnk1 requires Mg 2+ binding on a specific binding loop to achieve maximum activity [41]. Mnk1 inhibition as a pUR nevertheless appeared to be the most significant and durable putative target of MgSO 4 , inducing the actual repression of many downstream genes. Fmr1 transcription was not affected in any experimental group but the synaptic functional regulator Fmr-1 also appeared to be an activating pUR site for MgSO 4 , although the literature does not report an Mg 2+ cation link to protein activity. Functional relationships between (i) MgSO 4 effects through Mnk1 and Fmr-1, and (ii) subsequent synaptogenesis appear to be reliably supported by our finding, but the modes of action remained undetermined. As a polyribosome-associated mRNA-binding protein, Fmr-1 indicates a Mg 2+ putative action on protein translation that appeared more clearly in the MgHI group. undetermined. As a polyribosome-associated mRNA-binding protein, Fmr-1 indicates a Mg 2+ putative action on protein translation that appeared more clearly in the MgHI group. Although transcription effects were very transient, it is possible that this interference with such a crucial development process could have long-term consequences, as reported for several pharmacological agents eventually used in neonatal pediatrics (corticoids anesthetics, NSAIDs), or in animal models (MK-801, ketamine) [42][43][44][45][46]. We assessed several behavioral responses in adults exposed to neuroprotective dose of MgSO4 compared with drugs with known effects at P5. The battery of tests was chosen to mitigate the simple spontaneous response to a new environment and more integrated drug-or congenerevoked responses. This battery of tests in fact allowed for the discrimination of non-convergent behavioral parameters, since positive controls injected with MK-801 or ketamine had separate sex-and test-specific effects. MgSO4 acute neonatal treatment did not modify adult behavior in these tests. Although it is difficult, from a statistical point of view, to demonstrate an absence of effect, these data argue against a putative deleterious effect of MgSO4 at a neuroprotective dose, in particular through its interaction with the NMDAtype glutamate receptor. Previous studies have reported that high doses of MgSO4 may impair vascular development, then nervous parenchyma [8]. The use of repeated MgSO4 administration was performed to assess this possibility, in conditions designed to mimic the prolonged administration that eventually occurs in pregnant women at risk of preterm delivery. In this respect, it is interesting to note the absence of lasting effects of repeated MgSO4. Although transcription effects were very transient, it is possible that this interference with such a crucial development process could have long-term consequences, as reported for several pharmacological agents eventually used in neonatal pediatrics (corticoids anesthetics, NSAIDs), or in animal models (MK-801, ketamine) [42][43][44][45][46]. We assessed several behavioral responses in adults exposed to neuroprotective dose of MgSO 4 compared with drugs with known effects at P5. The battery of tests was chosen to mitigate the simple spontaneous response to a new environment and more integrated drug-or congener-evoked responses. This battery of tests in fact allowed for the discrimination of non-convergent behavioral parameters, since positive controls injected with MK-801 or ketamine had separate sex-and test-specific effects. MgSO 4 acute neonatal treatment did not modify adult behavior in these tests. Although it is difficult, from a statistical point of view, to demonstrate an absence of effect, these data argue against a putative deleterious effect of MgSO 4 at a neuroprotective dose, in particular through its interaction with the NMDA-type glutamate receptor. Previous studies have reported that high doses of MgSO 4 may impair vascular development, then nervous parenchyma [8]. The use of repeated MgSO 4 administration was performed to assess this possibility, in conditions designed to mimic the prolonged administration that eventually occurs in pregnant women at risk of preterm delivery. In this respect, it is interesting to note the absence of lasting effects of repeated MgSO 4 .
The present data on MgSO 4 effects showed transient transcription modulation apparently devoid of delayed deleterious effects and putatively capable of exerting brain protection. The study of MgSO 4 in neonate mice confirmed its harmlessness, which has also been largely reported in humans. This does not preclude other effects of MgSO 4 , i.e., on the highly vulnerable white matter around 30 GWs (MagNUM) study; [47]. In the Mag-NUM study, white matter analysis by means of magnetic resonance imaging revealed the induction by MgSO 4 of microstructure development, a finding coherent with the present observation of MgSO 4 transcription inductions/repressions orthologous to spontaneous development in 264 genes.
MgSO 4 is not only safe, it is also a potent prevention agent of motor, behavior, and cognitive deficiencies in preterm subjects [6,48], although its mode of action is not fully understood.
Another concern about the use of MgSO 4 as prophylactic a neuroprotection in fragile populations was the frequent statement that it acts as an NMDA channel blocker [49]. This belief was often based on a very dated description of NMDA function and also to the fact that MgSO 4 was neuroprotective in NMDA mediated neurotoxicity [18,35,50]. This possibility is frightening owing to the very deleterious effects in the developing brain of NMDA blockers such as MK-801 in animals [43], and with the relationship of NMDA hypoactivity or genetic defects with schizophrenia [51]. Administration of Mg 2+ reduced the release of Mg 2+ from NMDA channel, but it does not act as an antagonist. Mg block lifting is dependent on depolarized membrane potential but not on glutamate and glycine (Dserine) binding [52]. Under depolarization Mg 2+ loses intra-channel binding high affinity, making it unlikely that overload would reduce NMDA excessive activation. Moreover, the expression of NMDA-R subunits in 30-GW human brains have shown Mg 2+ blocking sensitivity [53]. In vitro studies showed that in different neuron populations, low doses of glutamate-induced pre-conditioning toward high glutamate excitotoxicity can be mimicked using Mg 2+ supplementation, indicating a clearly more complex effect of magnesium on NMDA receptors than channel blocking [52]. At the transcription level, MgSO 4 repressed Grin1 and Grin3a in Ctrl and reversed Grin2b induction after HI (Tables S1 and S2). These effects together with the reduction by Mg 2+ of glutamate release would underlie its antiglutamatergic toxicity more likely than NMDA blockade [54]. The present behavioral observations showed that MgSO 4 effects did not mimic the effect of the NMDA specific blocker MK-801. In addition, transcriptome IPA ® analysis never proposed genes coding NMDA receptor subunits as pURs, even at low significance levels before filtering. The statistical proof of an absence of effect is always a difficult challenge. Thus, a demonstration that MgSO 4 does not affect NMDA transmission in the 30-GW fetus is probably elusive; however MgSO 4 may be considered a safe prevention tool in the neonatal context, and is very likely devoid of NMDA-interfering side effects. On the contrary, the incidental observation of ketamine's interference with basal activity and social behavior renews the question of its use (or abuse) in the newborn [55]. The clearly distinct effects of ketamine compared to MK-801 ruled out ketamine's interaction with NMDA in these effects and lead us to consider ketamine for broader potential neurotoxic effects.
A major concern in very preterm care is the high incidence of brain hemorrhages. Meta-analyses of prophylactic antenatal MgSO 4 exposure did not conclude that it led to intraventricular hemorrhage (IVH) prevention, although tendencies were reported [6]. Relatedly, MgSO 4 slightly reduced cerebellar hemorrhage [1]. In an original model of preterm hemorrhage, we observed a 20% reduction of hemorrhage occurrence after MgSO 4 600 mg/kg administration in P5 mice (not shown), in line with therapeutic trends [1,56]. As hemorrhages largely depend on brain vessel immaturity in humans, it appears likely that the deleterious vascular effects of MgSO 4 reported for high dosages did not occur at neuroprotection low doses [8,57].
Neuroprotection in fetuses using low MgSO 4 doses given to women at risk of preterm delivery before completed 33 GWs is now recognized and approved as the gold standard in many countries [4,58]. Its action reported in preterm exposed to MgSO 4 in the last in utero periods showed respiratory and hemodynamic stabilization, but MgSO 4 also showed direct neuroprotection of the neural tissue. Combined transcriptomics and metabolomics approaches showed a type of MgSO 4 preconditioning effect on mitochondria, resulting in the sparing of high-energy phosphate and the reduction of succinate, involved in ROS production and inflammation [33].
Moving to MgSO 4 s prevention of HI effects in the neonate mouse, a remarkable observation was that the combination of MgSO 4 plus HI did not elicit a summation of the effects observed in separate conditions. MgSO 4 alone had its maximum transcription effects after 6 h although it affected HI responses as early as 3 h after insult, reorientating the responses in an original direction. This effect of Mg 2+ was likely the result of nontranscriptional effects; i.e., at protein activity levels, before Mg 2+ induced transcription modulation occurred ( Figure 6). Although HI responses largely differed in terms of identity of the genes affected in Ctrl or in MgSO 4 -treated animals, they converged toward the same functions; inflammation, innate immunity and mRNA transcription by RNA-polymerase II complex. The reduced amplitude of effects was, however, in line with expected antiinflammatory MgSO 4 effects [25]. Anti-inflammatory effects of MgSO 4 at the transcription level in an HI context toward Ccl2 Cxcl1, Ccl3, and Csf1 coincided with the reported effects at the protein level on early inflammation response factors (MCP-1,GRO/KC,MIP-1α and M-CSF) described in rat [33]. Transcriptomics allowed us to extend the MgSO 4 action field to Tlr3,4,9, Il1α,β, Tnf, NfκB, and Ifng-mediated pathways. Relatedly, the absence of transcription effects on Il4 and Il10 transcription was in accordance with the observations at the protein level. Ifng appeared as a pUR in HI at the highest significance but did not appear in MgHI, together with many pro-inflammation genes (Stat3, Tlr4, Ccr2, Cx3cl1, Chuk, Nf кb, Ccl2, Stat1, and Saa1). Among these genes, only Ccl2 and Stat1 were affected at the transcription level, also indicating that MgSO 4 essentially prevented HI-induced inflammation at post-transcription levels. Ifng also had a putative impact on innate immune cells since most putative downstream effectors were related to chemokine and, macrophages-immune cell communication pathways ( Figure 9A).
In preterm subjects, increased Il-4 levels in the cerebrospinal fluid contributed to a cytokinic neuro-inflammatory profile [59]. In HI mice, Il-4 appeared as a pUR, suggesting an early regulatory function on the inflammation process. In Mg groups it showed a tendency to aggregate Il-4 putative downstream targets, but no clear inhibition by MgSO 4 of HI effects indicate that Il-4 was a target of Mg. Il-4 s anti-inflammation function is in part due to its favoring the M2a macrophage/microglial phenotype in neonates [60,61]. In fact, an Mg induction of M2a phenotype has been reported as enhancing Il4, Il10, Bmp2, and Vegf expression [62]. In our hands, MgSO 4 likely had post-transcription interaction with Il-4, but examination of Il-4 downstream microglial markers of repair (M2A) or the immuno-modulatory (M2B) phenotypes did not indicate convergent Mg effects [61]. Therefore, Mg 2+ 's putative anti-inflammatory effects were rather due to the inhibition of the pro-inflammatory pathways than by inducing an anti-inflammatory differentiation of microglia.
De novo effects were detected in MgHI, with high p-values for mitochondria and to a lesser extent for ribosome-related indexes. Magnesium reduced the transcription of genes involved in energy metabolism, protein translation, and protein degradation through ubiquitination. mTORC2 (coded by Rictor) and Sirt1 should be considered inhibitory hubs for oxidative phosphorylation and protein translation and degradation in MgHI mice. Rictor codes mTORC2, the rapamycine insensitive companion of mTOR which integrates many cell-survival-linked functions and requires Mg 2+ for its activation [63]. MgSO 4 previous to the Rice and Vannucci procedure in rats at P7 showed preserved mitochondrial functions after HI suggesting a pre-conditioning mechanism [33]. In our hands, putative mTORC2 MgSO 4 activation may have resulted in the repression of 33 genes and 1 induction, of which 11 genes were coding factors of mitochondrion membrane respiration chain complex 1. These genes were distinct from those identified in rats at P7, although at a different time point, but the observations converge toward mitochondrial targeting by MgSO 4 [33]. In addition, mTORC2 appeared to be a potential downstream hub by which Mg 2+ affects protein translation. The repression of 12 genes coding ribosomal proteins putatively occurred downstream of mTORC2 activation, half of which were coding proteins further localized at the synapse (Rpsa, Rps1, Rps15, Rps18, Rpl7, Rpl10). Of note, none of the genes coding these proteins was affected in Mg groups in which synaptic tropism of Mg effects was noted, once again indicating a de novo mechanism recruited in MgHI.
Many MgSO 4 pretreatment effects could be detected or suspected in the absence of up-coming HI, such as inhibitory effects on signaling pathways dependent on GPCR, a well-known target of Mg 2+ in terms of membrane trafficking, and transduction effects. The rapid diffusion of the divalent cation and interaction at these sites support the view of rapid MgSO 4 interference with HI, and sustain the findings of upstream regulators as putative Mg targets, even when unaffected at their own transcription level.

Animals
National marine research Institute (NMRI) mice, purchased from Janvier Labs (Le Genest-Saint-Isle France), were housed sex-separated at a controlled temperature (21 • C) with access to food and water ad libitum with a 12-h light/dark cycle. Reproduction was performed locally by putting females in male cages for one night to cause birth at definite days. The day of birth was noted as day 1. All nursing and experimental procedures were performed according to the recommendations of the European Communities council directives (2010/63/UE) and French national legislation. Protocols received the agreement of the French Ministry of Higher Education and Research (#01680.02/2014) and were performed by authorized experimenters with efforts to minimize animal numbers and suffering. Animal numbers are given in Table S3.
Transcriptome studies were carried out in four groups of NMRI mice aged 5 days (P5). The Control group (Ctrl) underwent no intervention. The Mg group of animals received an intra-peritoneal MgSO 4 bolus injection (3.33 µL/g) at 600 mg/kg of body weight as previously reported [28]. Brains were collected at 5 time points-1.5 h, 3 h, 6 h, 12 h, and 24 h.
The HI group in which animals were exposed to hypoxia-ischemia as previously described [28]. Briefly, isoflurane anesthetized pups had a right carotid ligature, and were returned to the mother. Then, after 60-120 min of recovery from anesthesia and surgery, the pups were exposed to 8% O 2 for 40 min in a humidified and thermostatic chamber at 36 • C [28,37]. The end of hypoxia was considered t 0 for the 3-h and 12-h post HI periods.
The MgHI group of animals received MgSO 4 pretreatment one hour before being exposed to hypoxia-ischemia [28].
Behavioral studies were performed in a total of 204 mice treated at P5 by a single injection of saline buffer (PBS), MgSO 4 (600 mg/kg), or positive control pharmacological agents-MK-801 (1 mg/kg) or ketamine (40 mg/kg), known to induce long-term effects in the behavior paradigm evaluated (details in Supplementary file). Repeated daily injections of PBS or MgSO 4 (600 mg/kg) from day 5 to day 9 were performed in separate groups. Each parameter is considered as mean ± standard error to the mean.

RNA Sample Preparation and Microarray Hybridization
Brains were quickly removed under isoflurane anesthesia, sectioned on medial line and ipsilateral hemispheres and were immediately frozen in liquid nitrogen. Pool samples for the microarray study were made with 5 µg of RNA extracted from 6 ipsilateral hemispheres of sex-matched injured or control mice in the different experimental groups and time points. Mice pups at each time point were taken from at least two different litters. A total of 288 pups were used for the transcriptome study (details in supplementary file).
RNA extraction was performed in individual hemispheres. Brain tissues were thawed in 350 µL of tissue lyser ® (Qiagen, Courtaboeuf, France) and homogenized using ceramic beads (1.4 mm Ozyme, Montigny le Bretonneaux) for 20 s at 50 Hz. Total RNA extraction was done using the Nucleospin RNA plus kit ® according to the manufacturer's recommendations (Macherey-Nagel, Hoerdt, France, www.servilab.fr (accessed on 15 April 2021)) and stored at −80 • C until use. The quality and quantity of isolated RNAs were assessed using the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and the Nanodrop (Thermo Scientific, Wilmington, NC, USA). Labeled cRNA was synthesized from 100 ng total RNA using the Quick Amp Labeling Kit according to the manufacturer's recommendations (Agilent Technologies, Les Ulis, France). A total of 825 ng of labeled cRNA (Cy3 for control samples and Cy5 for test samples) was co-hybridized for 17 h at 65 • C on Whole Mouse Genome Oligo 4 × 44 K microarrays (G4845A, Agilent Technologies, Les Ulis, France). Raw hybridization data were extracted using an Agilent DNA microarray scanner G2565CA (Agilent Technologies), and normalized (using LOWESS method) by Feature extraction software (v.10.5.1.1), then were transferred to Genespring ® (GX v.14.9.1 software, Agilent Technologies, Santa Clara, CA, USA) for data processing and data mining.
Hybridization data were filtered on spot quality (uniformity) detection level (above 2-fold background in at least 2 samples out of three in one-color analyses) and statistically selected using GeneSpring ® (GX 12.6 software, Agilent Technologies, Les Ulis, France), as previously described in detail [29,64]. Only probes referring to validated genes (Official Gene Symbol) were taken into account. Probes for cRIKEN sequences were excluded from analysis. In the case of multiple probes for one gene, the probe exhibiting the highest fold change was used for subsequent analyses.

Strategy of Transcriptomic Analysis
The study was designed to extract the effects of MgSO 4 alone (List A) and MgSO 4 interference with HI effects (List B) at the transcriptional level. Transcription responses were measured using a two-color strategy in three independent experiments. The signal from any probe was recorded when the p-value in the t-test against 0 was below 5 × 10 −2 . The extraction of regulated probes was performed at each time point separately (Figure 1). The effect of MgSO 4 pretreatment on the transcription response to HI was evaluated by comparison with HI-exposed brains at 2 time points after HI. The choice of 3 h and 12 h post-insult time points was based on previous data, in order to detect early onset and peak responses in P5 brains [29]. • Experiment 3 (HI vs. Ctrl): The effect of HI was evaluated by comparison to control brains at 3 h and 12 h time points to ensure reproducibility with the previous study and to provide reference effects of HI at P5 [29].
In all three experiments, the two-color analysis identified significant inductions and repressions. Thus, the effects on genes affected both in 1 and 2 , or in 2 and 3 must take into account the direction of variation that might differ among common entities (Figure 2).
Comparisons of expression levels on one fluorochrome can also be extracted from separated arrays after normalization using the Genespring ® one-color protocol and allow broader comparisons of conditions. One-color analyses were performed to compare expression levels in the Mg group to (i) respective levels in the MgHI group, in order to exclude MgSO 4 pretreatment effects that did not interfere with HI, and (ii) MgHI to Ctrl, as detailed below. o List A 1 grouped the fraction of List A of MgSO 4 effects observed in the same direction in the two-color protocol in both Ctrl and HI mice, indicating HIevoked transcription responses that were unaffected by MgSO 4 ( 4 in Figure 2). The effects observed at least at one time point, in any sense of variation, at 3 and 12 h were merged in order to be handled as Mg or MgHI responses. • List B reports the effects of MgSO4 pretreatment on transcription responses to HI. It was obtained by concatenation of three sub-lists (B1, B2, and B3) describing three classes of effects of MgSO4: (i) the effects of HI revealed after MgSO4 exposure (List B1), (ii) the reversion by MgSO4 of HI transcription effects (List B2), and (iii) the amplification by MgSO4 pretreatment of HI transcription effects (List B3). The elaboration of these lists required crossed comparisons of two-color and one-color analyses corrections and the subtraction of genes in List A1 (Figure 2).
o List B1 (concatenation of ⑤ + ⑥ + ⑦−④−⑪) contains the genes affected after HIexposed MgSO4-pretreated animals in ② but not after HI in Ctrl animals in twocolor analyses in ③ (⑤ in Figure 2), plus genes exhibiting opposite effects in ① and ② (⑥ in Figure 2). B1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in ① and ③) and MgHI values in ② (⑦ in Figure 2). The later series could reveal genes unaffected by HI (in ③). Finally, unchanged MgSO4 and HI effects in combined conditions (A1 and ⑪) were subtracted.
o List B2 (concatenation of ⑧ + ⑨−④) contains the genes affected in opposite senses by HI in Ctrl or in MgSO4 pretreated animals in two-color analysis (⑧ in Figure  2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI (⑨ in Figure 2), minus A1 (④). o List B3 (concatenation of ⑩−④) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO4 pretreated animals (⑩ in Figure 2) minus A1 (④).
• List C contained genes exhibiting transcription regulation in the same sense in MgSO4-treated or non-treated mice in ③ and ⑦ (⑪ in Figure 2) and represented genes affected by HI but not affected by MgSO4 pretreatment.

Data Mining
Gene lists A and B, generated as detailed in Figure 2 were submitted to DAVID ® (v 6.8) software [65,66]. Enrichment of seKws, seGO-terms, on biological processes (GOTERM_BP_ALL; BP), cell components (GOTERM_CC_ALL; CC) and molecular functions (GOTERM_MF_ALL; MF), and seKpaths were researched using DAVID ® software. Keywords and GO-terms enrichment was extracted on the 1411 genes of List A, and the 1958 genes of List B filtered on gene number (N) ≥10, at the Bonferroni p-value threshold <1 × 10 −4 , or coincidence of p-value according to EASE < 1 × 10 −4 with FE >2 [67]. Clustering of seGO-terms was performed using the online REVIGO software (http://revigo.irb.hr/ ) contains the genes affected after HI-exposed MgSO 4 -pretreated animals in 2 but not after HI in Ctrl animals in two-color analyses in 3 ( 5 in Figure 2), plus genes exhibiting opposite effects in 1 and 2 ( 6 in Figure 2). B 1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in 1 and 3 ) and MgHI values in 2 ( 7 in Figure 2). The later series could reveal genes unaffected by HI (in 3 ). Finally, unchanged MgSO 4 and HI effects in combined conditions (A 1 and repression of separate probes. This was due to biphasic regulation (1 item) or detection of variations on separate probes (8 items). These ambiguous items were placed in induced or repressed gene lists for further analyses. o List A1 grouped the fraction of List A of MgSO4 effects observed in the same direction in the two-color protocol in both Ctrl and HI mice, indicating HIevoked transcription responses that were unaffected by MgSO4 (④ in Figure 2). The effects observed at least at one time point, in any sense of variation, at 3 and 12 h were merged in order to be handled as Mg or MgHI responses.

•
List B reports the effects of MgSO4 pretreatment on transcription responses to HI. It was obtained by concatenation of three sub-lists (B1, B2, and B3) describing three classes of effects of MgSO4: (i) the effects of HI revealed after MgSO4 exposure (List B1), (ii) the reversion by MgSO4 of HI transcription effects (List B2), and (iii) the amplification by MgSO4 pretreatment of HI transcription effects (List B3). The elaboration of these lists required crossed comparisons of two-color and one-color analyses corrections and the subtraction of genes in List A1 (Figure 2).
o List B1 (concatenation of ⑤ + ⑥ + ⑦−④−⑪) contains the genes affected after HIexposed MgSO4-pretreated animals in ② but not after HI in Ctrl animals in twocolor analyses in ③ (⑤ in Figure 2), plus genes exhibiting opposite effects in ① and ② (⑥ in Figure 2). B1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in ① and ③) and MgHI values in ② (⑦ in Figure 2). The later series could reveal genes unaffected by HI (in ③). Finally, unchanged MgSO4 and HI effects in combined conditions (A1 and ⑪) were subtracted.
o List B2 (concatenation of ⑧ + ⑨−④) contains the genes affected in opposite senses by HI in Ctrl or in MgSO4 pretreated animals in two-color analysis (⑧ in Figure  2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI (⑨ in Figure 2), minus A1 (④). o List B3 (concatenation of ⑩−④) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO4 pretreated animals (⑩ in Figure 2) minus A1 (④).

•
List C contained genes exhibiting transcription regulation in the same sense in MgSO4-treated or non-treated mice in ③ and ⑦ (⑪ in Figure 2) and represented genes affected by HI but not affected by MgSO4 pretreatment.

Data Mining
Gene lists A and B, generated as detailed in Figure 2 were submitted to DAVID ® (v 6.8) software [65,66]. Enrichment of seKws, seGO-terms, on biological processes (GOTERM_BP_ALL; BP), cell components (GOTERM_CC_ALL; CC) and molecular functions (GOTERM_MF_ALL; MF), and seKpaths were researched using DAVID ® software. Keywords and GO-terms enrichment was extracted on the 1411 genes of List A, and the 1958 genes of List B filtered on gene number (N) ≥10, at the Bonferroni p-value threshold <1 × 10 −4 , or coincidence of p-value according to EASE < 1 × 10 −4 with FE >2 [67]. Clustering of seGO-terms was performed using the online REVIGO software (http://revigo.irb.hr/ ) were subtracted. o List B 2 (concatenation of 8 + 9 − 4 ) contains the genes affected in opposite senses by HI in Ctrl or in MgSO 4 pretreated animals in two-color analysis ( 8 in Figure 2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI ( 9 in Figure 2), minus A 1 ( 4 ). o List B 3 (concatenation of pression levels in the Mg group to (i) respective levels in the MgHI group, in order to exclude MgSO4 pretreatment effects that did not interfere with HI, and (ii) MgHI to Ctrl, as detailed below.
• List A, extracted from the two-color analysis of experiment 1, described the transcription effects of MgSO4 single injection. The effects of MgSO4 were evaluated at 5 time points. All entities exhibiting either mean change ≥2 fold and/or p-values ≤ 0.05 t test against 0, according to Benjamini-Hochberg correction were selected at each time point in both directions and recorded in List A. Few genes showed induction and repression of separate probes. This was due to biphasic regulation (1 item) or detection of variations on separate probes (8 items). These ambiguous items were placed in induced or repressed gene lists for further analyses.
o List A1 grouped the fraction of List A of MgSO4 effects observed in the same direction in the two-color protocol in both Ctrl and HI mice, indicating HIevoked transcription responses that were unaffected by MgSO4 (④ in Figure 2). The effects observed at least at one time point, in any sense of variation, at 3 and 12 h were merged in order to be handled as Mg or MgHI responses. o List B1 (concatenation of ⑤ + ⑥ + ⑦−④−⑪) contains the genes affected after HIexposed MgSO4-pretreated animals in ② but not after HI in Ctrl animals in twocolor analyses in ③ (⑤ in Figure 2), plus genes exhibiting opposite effects in ① and ② (⑥ in Figure 2). B1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in ① and ③) and MgHI values in ② (⑦ in Figure 2). The later series could reveal genes unaffected by HI (in ③). Finally, unchanged MgSO4 and HI effects in combined conditions (A1 and ⑪) were subtracted.
o List B2 (concatenation of ⑧ + ⑨−④) contains the genes affected in opposite senses by HI in Ctrl or in MgSO4 pretreated animals in two-color analysis (⑧ in Figure  2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI (⑨ in Figure 2), minus A1 (④). o List B3 (concatenation of ⑩−④) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO4 pretreated animals (⑩ in Figure 2) minus A1 (④).
• List C contained genes exhibiting transcription regulation in the same sense in MgSO4-treated or non-treated mice in ③ and ⑦ (⑪ in Figure 2) and represented genes affected by HI but not affected by MgSO4 pretreatment.

Data Mining
Gene lists A and B, generated as detailed in Figure 2 were submitted to DAVID ® (v 6.8) software [65,66]. Enrichment of seKws, seGO-terms, on biological processes (GOTERM_BP_ALL; BP), cell components (GOTERM_CC_ALL; CC) and molecular functions (GOTERM_MF_ALL; MF), and seKpaths were researched using DAVID ® software. Keywords and GO-terms enrichment was extracted on the 1411 genes of List A, and the 1958 genes of List B filtered on gene number (N) ≥10, at the Bonferroni p-value threshold <1 × 10 −4 , or coincidence of p-value according to EASE < 1 × 10 −4 with FE >2 [67]. Clustering of seGO-terms was performed using the online REVIGO software (http://revigo.irb.hr/ − 4 ) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO 4 pretreated animals ( broader comparisons of conditions. One-color analyses were performed to compare expression levels in the Mg group to (i) respective levels in the MgHI group, in order to exclude MgSO4 pretreatment effects that did not interfere with HI, and (ii) MgHI to Ctrl, as detailed below.

•
List A, extracted from the two-color analysis of experiment 1, described the transcription effects of MgSO4 single injection. The effects of MgSO4 were evaluated at 5 time points. All entities exhibiting either mean change ≥2 fold and/or p-values ≤ 0.05 t test against 0, according to Benjamini-Hochberg correction were selected at each time point in both directions and recorded in List A. Few genes showed induction and repression of separate probes. This was due to biphasic regulation (1 item) or detection of variations on separate probes (8 items). These ambiguous items were placed in induced or repressed gene lists for further analyses. o List A1 grouped the fraction of List A of MgSO4 effects observed in the same direction in the two-color protocol in both Ctrl and HI mice, indicating HIevoked transcription responses that were unaffected by MgSO4 (④ in Figure 2). The effects observed at least at one time point, in any sense of variation, at 3 and 12 h were merged in order to be handled as Mg or MgHI responses.

•
List B reports the effects of MgSO4 pretreatment on transcription responses to HI. It was obtained by concatenation of three sub-lists (B1, B2, and B3) describing three classes of effects of MgSO4: (i) the effects of HI revealed after MgSO4 exposure (List B1), (ii) the reversion by MgSO4 of HI transcription effects (List B2), and (iii) the amplification by MgSO4 pretreatment of HI transcription effects (List B3). The elaboration of these lists required crossed comparisons of two-color and one-color analyses corrections and the subtraction of genes in List A1 ( Figure 2).
o List B1 (concatenation of ⑤ + ⑥ + ⑦−④−⑪) contains the genes affected after HIexposed MgSO4-pretreated animals in ② but not after HI in Ctrl animals in twocolor analyses in ③ (⑤ in Figure 2), plus genes exhibiting opposite effects in ① and ② (⑥ in Figure 2). B1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in ① and ③) and MgHI values in ② (⑦ in Figure 2). The later series could reveal genes unaffected by HI (in ③). Finally, unchanged MgSO4 and HI effects in combined conditions (A1 and ⑪) were subtracted. o List B2 (concatenation of ⑧ + ⑨−④) contains the genes affected in opposite senses by HI in Ctrl or in MgSO4 pretreated animals in two-color analysis (⑧ in Figure  2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI (⑨ in Figure 2), minus A1 (④). o List B3 (concatenation of ⑩−④) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO4 pretreated animals (⑩ in Figure 2) minus A1 (④).

•
List C contained genes exhibiting transcription regulation in the same sense in MgSO4-treated or non-treated mice in ③ and ⑦ (⑪ in Figure 2) and represented genes affected by HI but not affected by MgSO4 pretreatment.

Data Mining
Gene lists A and B, generated as detailed in Figure 2 were submitted to DAVID ® (v 6.8) software [65,66]. Enrichment of seKws, seGO-terms, on biological processes (GOTERM_BP_ALL; BP), cell components (GOTERM_CC_ALL; CC) and molecular functions (GOTERM_MF_ALL; MF), and seKpaths were researched using DAVID ® software. Keywords and GO-terms enrichment was extracted on the 1411 genes of List A, and the 1958 genes of List B filtered on gene number (N) ≥10, at the Bonferroni p-value threshold <1 × 10 −4 , or coincidence of p-value according to EASE < 1 × 10 −4 with FE >2 [67]. Clustering of seGO-terms was performed using the online REVIGO software (http://revigo.irb.hr/ in Figure 2) minus A 1 ( 4 ).

•
List C contained genes exhibiting transcription regulation in the same sense in MgSO 4treated or non-treated mice in 3 and 7 ( broader comparisons of conditions. One-color analyses were performed to compare expression levels in the Mg group to (i) respective levels in the MgHI group, in order to exclude MgSO4 pretreatment effects that did not interfere with HI, and (ii) MgHI to Ctrl, as detailed below.

•
List A, extracted from the two-color analysis of experiment 1, described the transcription effects of MgSO4 single injection. The effects of MgSO4 were evaluated at 5 time points. All entities exhibiting either mean change ≥2 fold and/or p-values ≤ 0.05 t test against 0, according to Benjamini-Hochberg correction were selected at each time point in both directions and recorded in List A. Few genes showed induction and repression of separate probes. This was due to biphasic regulation (1 item) or detection of variations on separate probes (8 items). These ambiguous items were placed in induced or repressed gene lists for further analyses. o List A1 grouped the fraction of List A of MgSO4 effects observed in the same direction in the two-color protocol in both Ctrl and HI mice, indicating HIevoked transcription responses that were unaffected by MgSO4 (④ in Figure 2). The effects observed at least at one time point, in any sense of variation, at 3 and 12 h were merged in order to be handled as Mg or MgHI responses.

•
List B reports the effects of MgSO4 pretreatment on transcription responses to HI. It was obtained by concatenation of three sub-lists (B1, B2, and B3) describing three classes of effects of MgSO4: (i) the effects of HI revealed after MgSO4 exposure (List B1), (ii) the reversion by MgSO4 of HI transcription effects (List B2), and (iii) the amplification by MgSO4 pretreatment of HI transcription effects (List B3). The elaboration of these lists required crossed comparisons of two-color and one-color analyses corrections and the subtraction of genes in List A1 ( Figure 2).
o List B1 (concatenation of ⑤ + ⑥ + ⑦−④−⑪) contains the genes affected after HIexposed MgSO4-pretreated animals in ② but not after HI in Ctrl animals in twocolor analyses in ③ (⑤ in Figure 2), plus genes exhibiting opposite effects in ① and ② (⑥ in Figure 2). B1 was supplemented by genes exhibiting significant t test differences in one-color analysis between Ctrl values (in ① and ③) and MgHI values in ② (⑦ in Figure 2). The later series could reveal genes unaffected by HI (in ③). Finally, unchanged MgSO4 and HI effects in combined conditions (A1 and ⑪) were subtracted.
o List B2 (concatenation of ⑧ + ⑨−④) contains the genes affected in opposite senses by HI in Ctrl or in MgSO4 pretreated animals in two-color analysis (⑧ in Figure  2), plus the genes exhibiting no difference in one-color analysis between Ctrl and MgHI (⑨ in Figure 2), minus A1 (④). o List B3 (concatenation of ⑩−④) contains the genes affected by HI in the same sense after HI in Ctrl or in MgSO4 pretreated animals (⑩ in Figure 2) minus A1 (④).

•
List C contained genes exhibiting transcription regulation in the same sense in MgSO4-treated or non-treated mice in ③ and ⑦ (⑪ in Figure 2) and represented genes affected by HI but not affected by MgSO4 pretreatment.

Data Mining
Gene lists A and B, generated as detailed in Figure 2 were submitted to DAVID ® (v 6.8) software [65,66]. Enrichment of seKws, seGO-terms, on biological processes (GOTERM_BP_ALL; BP), cell components (GOTERM_CC_ALL; CC) and molecular functions (GOTERM_MF_ALL; MF), and seKpaths were researched using DAVID ® software. Keywords and GO-terms enrichment was extracted on the 1411 genes of List A, and the 1958 genes of List B filtered on gene number (N) ≥10, at the Bonferroni p-value threshold <1 × 10 −4 , or coincidence of p-value according to EASE < 1 × 10 −4 with FE >2 [67]. Clustering of seGO-terms was performed using the online REVIGO software (http://revigo.irb.hr/ in Figure 2) and represented genes affected by HI but not affected by MgSO 4 pretreatment.
Pathway investigation was completed by the literature-based Ingenuity Pathway Analysis (IPA ® ). IPA ® used human gene nomenclatures (showing slight differences with mice). IPA analysis provided a z-score, indicating the sense of regulation (activation or inhibition) of significantly enriched pathways (seIPaths). We recorded the canonical pathways with N ≥ 10 and right-tailed Fisher's Exact Test p-values < 0.01 and |z-score| >2, at least at a one-time point after HI or MgSO 4 treatment. Pathway list were compared and classified using hierarchical clustering based on p-value and z-scores (selection set at p-values < 1 × 10 −4 and |z-score| > 2), using IPA ® , on the basis of comparison of MgSO 4 effects at 4 different time points; HI effects at 3 h and 12 h and MgHI vs. HI effects at 3 h and 12 h, in lists determined by the Genespring ® normalized Cy ratio. IPA analysis did not indicate an FDR.
Interference of MgSO 4 injection with development in control and in HI-exposed mice has been explored using previously described spontaneous variations of expressions before and after P5 (details in Supplementary Materials) [37], deposited in the NCBI Gene Expression Omnibus; GSE144456).
A complement experiment of two-color analysis of MgHI and Ctrl on 3 replicates at 3 h and one at 12 h after HI was performed to consolidate one color analysis (Data in Supplementary Materials).

Behavioral Studies
Three different behavioral tests were performed to assess long-term putative effects at adulthood of neonatal treatments. Six experimental groups were studied. Four groups received a single injection at P5 of saline (PBS), MgSO 4 , (600 mg/kg), MK-801(1 mg/kg), or ketamine (40 mg/kg), and two groups were treated daily for P5 to P9 with PBS or MgSO 4 . Behavior was tested at adulthood to evaluate spontaneous activity, acute MK-801 (375 µg/kg) evoked activity, and social behavior. Spontaneous activity was assessed in a free-access running wheel in the home cage for 3 days and in an open field for 30 min. In the open field, activity changes induced by acute MK-801 subcutaneous injection were recorded for 150 min separately [69]. Social interaction with an unknown congener was assessed for 6 min in a two-compartment device after a 3-min habituation period of the test animal, as previously described [37].

Conclusions
MgSO 4 had a wide spectrum of activities slowing-down transcription, which were largely oriented toward synaptic development in P5 neonatal mice. These effects were transient and did not affect further behavioral activities in grownups. The transcription pattern response to MgHI appeared very different to the arithmetic summation of separate effects of MgSO 4 and HI. Rather, it appeared that MgSO 4 priming reoriented HI response in an original direction, from which neuroprotection may result. Administered before HI, MgSO 4 reduced inflammation and innate immune responses. MgSO 4 's slowing effects on transcription by RNA-polymerase II or on protein translation by Eif2,4 may result from the orientation of energy metabolism toward low-energy-cost high phosphate pathways. These effects possibly result from mitochondrial targeting and the reduction of protein translation but not from microglial induction toward regulatory functions.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the French "Ministère de l'Education Nationale de l'Enseignement Supérieur et de la Recherche" (protocol code 01680.02 on 13 October 2014).

Informed Consent Statement: Not applicable.
Data Availability Statement: Data supporting the reported results can be found in the NCBI Gene-Expression-Omnibus repository; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE144 455 (accessed on 24 January 2020)