Intrauterine Growth Restriction Followed by Oxygen Support Uniquely Interferes with Genetic Regulators of Myelination

Abstract Intrauterine growth restriction (IUGR) and oxygen exposure in isolation and combination adversely affect the developing brain, putting infants at risk for neurodevelopmental disability including cerebral palsy (CP). Rodent models of IUGR and postnatal hyperoxia have demonstrated oligodendroglial (OL) injury with subsequent white matter injury (WMI) and motor dysfunction. Here, we investigate transcriptomic dysregulation in IUGR with and without hyperoxia exposure to account for the abnormal brain structure and function previously documented. We performed RNA sequencing and analysis using a mouse model of IUGR and found that IUGR, hyperoxia, and the combination of IUGR with hyperoxia (IUGR/hyperoxia) produced distinct changes in gene expression. IUGR in isolation demonstrated the fewest differentially expressed genes (DEGs) compared with control. In contrast, we detected several gene alterations in IUGR/hyperoxia; genes involved in myelination were strikingly downregulated. We also identified changes to specific regulators including TCF7L2, BDNF, SOX2, and DGCR8, through ingenuity pathway analysis (IPA), that may contribute to impaired myelination in IUGR/hyperoxia. Our findings show that IUGR with hyperoxia induces unique transcriptional changes in the developing brain. These indicate mechanisms for increased risk for WMI in IUGR infants exposed to oxygen and suggest potential therapeutic targets to improve motor outcomes.


Introduction
White matter injury (WMI) following in utero hypoxic-ischemic events, stroke and prematurity is well 4 these infants are often admitted to the neonatal intensive care unit and exposed to supraphysiologic oxygen.
Rodent studies have found that hyperoxia exposure alone results in damage to the developing white matter Thromboxane A2 (TXA2), is a vasoconstrictor overly expressed in mothers whose pregnancies are complicated 141 by hypertension, cigarette smoking, and poorly controlled diabetes (McAdam et al., 2005;Hayakawa et al., 2006; 142 Fung et al., 2011;Gibbins et al., 2018). Infusion of TXA2-analog U-46619 has been demonstrated to result in 143 placental vasculature reduction, suggesting placental vascular insufficiency, similar to human placental 144 pathology resulting in IUGR (Gibbins et al., 2018). This model does not require invasive surgery and it is 145 physiologically relevant to human IUGR pregnancies.

146
Micro-osmotic Alzet pumps (model 1007D, 0.5 ml/h) (DURECT Corporation) were implanted into gravid mice 147 at 12.5 days post-coitus, correlating with the last trimester of mouse pregnancy. Pumps were inserted into a 148 subcutaneous pocket created in the hip space. The pumps contained either the thromboxane A2 (TXA2)-analog 149 U-46619 (Cayman Chemical) dissolved in 0.5% ethanol or 0.5% ethanol (vehicle) which was continuously 150 infused at 2000 ng/h throughout the remainder of pregnancy (Fung et al., 2011). Previous model characterization 151 has shown that plasma 11-dehydrothromboxane B2 levels were similar between the vehicle and U-46619 152 exposed fetuses, providing evidence that U-46619 did not cross the placenta to affect the pups directly (Fung et 6 receiving TXA2-analog and weighing less than 1.266g, <10 th percentile for weight based on sham pup weights, 155 were assigned to the IUGR group. Using this cut-off, approximately one-third of TXA2-analog pups were defined 156 as small for gestational age (SGA), which is similar to the incidence of human SGA infants, born to mothers with 0.5% ethanol and weighing greater than 1.266g (>10 th percentile) were assigned to the vehicle group. All pups 159 were cross-fostered to unmanipulated mouse dams post-delivery to minimize the surgical effects of pump 160 insertion in the birth dams.

162
Postnatal hyperoxia exposure 163 Litters of vehicle and IUGR pups were placed in either 75% oxygen (hyperoxia) in a Plexiglass chamber 164 (Biospherix) or 21% oxygen (room air) within 24 hours after birth for 14 days (Aslam et al., 2009;Lee et al., 165 2014). Exposure to hyperoxia was continuous, with brief interruption only for animal care (<10 minutes/day). The 166 concentration of oxygen was maintained with an oxygen controller (ProOx, Biospherix). Ventilation within the 167 chamber was adjusted to remove CO2 such that it did not exceed 0.5%. A hygro-thermometer was used in the 168 chamber to monitor temperature and humidity. Temperature in the chamber did not exceed 23°C and humidity 169 level was maintained using dishes of desiccant in the bottom of the chamber. A foster dam was placed in the 170 hyperoxia chamber with each vehicle or IUGR litter, and foster dams were rotated from hyperoxia to room air 171 every 24-48 hours to prevent excessive oxygen toxicity to the adult animals. Litters were removed from the 172 hyperoxia chamber at 14 days and euthanized for tissue collection.

199
For sub-analysis, gene expression results were also categorized by brain cell type using a transcriptome 200 database created by the Barres lab at Stanford University available at http://web.stanford.edu/group/ 201 barres_lab/brain_rnaseq.html (Zhang et al., 2014). RNA sequencing of purified neurons, astrocytes, microglia, 202 endothelial cells, pericytes, and various maturation states of oligodendrocytes from mouse cortex were used to 203 generate this high-resolution transcriptome database of > 22,000 genes.

204
Co-expression analysis was performed using the CEMiTool package in Bioconductor using variance 205 stabilizing transformation and an FDR cutoff of 0.05 (Russo et al., 2018). Ingenuity Pathway Analysis (IPA)

206
(Qiagen) was used to identify significant biological pathways from the RNA-seq data sets (Kramer et al., 2014).

207
A list of detected genes and detected proteins was used as the data input, using a q < 0.05 cutoff for the gene 8 pathway and p < 0.1 cutoff for the protein pathway analyses, such that only significant genes/proteins were 209 considered for significant pathways. The "User dataset" option was chosen to use each individual detected 210 gene/protein data set as the "reference set" for which to generate significant pathways. Pathways from the 211 "diseases and biological functions" category were used for comparison analyses. Fisher's t-test of p < 0.05 was 212 used to determine statistical significance of a pathway.

254
To test if IUGR and postnatal hyperoxia alters expression of genes associated with oligodendrogliogenesis 255 or myelination, we performed RNA-seq in a pilot experiment using total RNA pooled from a small number of P14 256 brains. P14 was chosen as it is the midpoint of myelination in rodents and allowed for sufficient postnatal 257 hyperoxia exposure. We performed differential expression analysis (DEA) and compared three experimental 258 groups against control: IUGR, hyperoxia, and IUGR/hyperoxia ( Figure 1A To determine if expression changes lead to WMI in the three groups, we evaluated DEGs specifically 267 expressed by oligodendroglia (OLDEGs) in the pooled data set. DEGs were categorized by brain cell type using 268 a publicly available transcriptome database (Zhang et al., 2014). We found that IUGR, hyperoxia, and 269 IUGR/hyperoxia demonstrated distinct patterns ( Figure 3A). IUGR yielded 6 OLDEGs, hyperoxia yielded 113,  (Table 2).

Hyperoxia with and without IUGR decreases myelin gene network expression 281
We repeated the RNA-seq using a larger sample size of individual (non-pooled) samples. Focusing on the 282 IUGR/hyperoxia data, as our pooled samples had shown the greatest effect on WMI in this group, we again 283 found distinct gene expression compared to control ( Figure 4A). While DEA identifies a large number of genes 284 that differ between groups, it does not give information on interconnections between DEGs. To address this, we 285 performed unsupervised gene co-expression analysis on the non-pooled RNA sequencing data with Co-

286
Expression Molecules identification Tool (CEMiTool) (Russo et al., 2018). CemiTool generated 5 modules highly 287 correlated with the data set ( Figure 4B). Module 4 (M4) was significantly enriched with 134 genes that were 288 identified to be related to myelination by the hub genes: MoBP, Plp1, Gsn, and Mog ( Figure 4B). Notably, activity 11 in M4 was lower in both hyperoxia and IUGR/hyperoxia ( Figure 4C and 4D) as demonstrated by statistically 290 significant adjusted p-values and normalized enrichment scores (Table 3). These findings support the results 291 from the DEA performed on our pooled samples and add further evidence that specific myelin genes are 292 differentially expressed following these exposures (Chang et al., 2018).

294
Unique gene regulators identified in Hyperoxia and IUGR/hyperoxia.

295
We applied Ingenuity Pathway Analysis (IPA) to our non-pooled dataset to determine potential gene To evaluate the impact sex has on WMI in IUGR and hyperoxia, qRT-PCR data for MoBP, Plp1, Mog, and 321 Cnp from P14 hemispheres was separately analyzed by sex. There were no significant differences found 322 between sexes in the control group. Nor were there significant differences between sexes found in IUGR, but 323 there was a trend toward decreased myelin gene expression in females compared to males for all 4 genes 324 ( Figure 8A). Hyperoxia showed decreased myelin gene expression in females compared to males that was 325 statistically significant for Plp1 (p=0.01); the other 3 genes showed a trend toward decreased expression in 326 females compared to males ( Figure 8B). In contrast, IUGR/hyperoxia showed statistically significant decreases 327 in Plp1 (p=0.0006) and Cnp (p=0.0076) in males compared to females ( Figure 8C).

337
hyperoxia, and IUGR/hyperoxia differed in the number and type of affected genes with minimal overlap in DEGs 338 between each group. As all three study groups have previously shown altered myelination to some degree 339 (Chang et al., 2018), it was somewhat surprising to find minimal overlap in differential gene expression.

340
Additionally, the directionality of DEGs differed between groups, with prominent downregulation of genes in 13 matter development likely occurs through distinct mechanisms and cellular interactions in each perturbation to 343 the brain.
Distinct transcriptomic changes between groups were also seen in sub-analysis of oligodendrocyte specific downregulation of OLDEGs in IUGR/hyperoxia. The predicted type of OL cell that was most affected also differed 348 between groups. In hyperoxia, DEGs specific to newly formed/myelinating OLs were found to be the most

358
Of note, the majority of the significant DEGs in our differential expression analysis (DEA) had fold changes 359 <2 or >-2, a cut-off routinely used in RNA sequencing studies. One explanation, is that the brain contains a 360 multitude of different cell types and the effect on oligodendrocytes, primarily perturbed in white matter injury,

381
While co-expression networks are able to identify correlations, indicating which genes are active 382 simultaneously and likely biologically related, they do not provide information about causality or distinguish 383 between regulatory/regulated genes (van Dam et al., 2018). Therefore, we next used IPA to identify potential 384 upstream regulators and provide insight into the mechanism for WMI in the different exposure groups. In both 385 hyperoxia and IUGR/hyperoxia, a pronounced downregulation of TCF7L2 signaling was seen. Downstream 386 target molecules of TCF7L2 in our dataset included major myelin genes MoBP, Cnp, Mog, Plp1, and Mbp.

387
TCF7L2 is a transcription factor specifically expressed in OLs during the time window that is critical for myelin 388 formation, during the transition from OPCs to mature myelinating OLs (Fu et al., 2012;Lurbke et al., 2013; 389 Hammond et al., 2015). It acts as a co-activator of β-catenin and is part of the canonical Wnt/β-catenin pathway, 390 a well-known signaling pathway involved in neurogenesis and OL maturation (Gaesser and Fyffe-Maricich, 391 2016). Inhibition of TCF7L2 in hyperoxia and IUGR/hyperoxia can therefore explain the specific downregulation 392 we found in newly formed and myelinating OLs. Dysregulation of the Wnt pathway has been implicated in other 15 prematurity (Fancy et al., 2009;Back, 2017) and thus it is unsurprising that it may be involved in WMI secondary to hyperoxia and IUGR/hyperoxia. and mTOR, which are known to be important in normal myelination, made up 3 out of the 4 upstream regulators identified in the hyperoxia group as inhibited. The transcription factor SRY-box 2 (SOX2) has 400 been shown to be involved in OL proliferation and differentiation during postnatal brain myelination (Hoffmann 401 et al., 2014). SOX2 also plays a role in CNS remyelination after injury and acts by recruiting adult OPCs (Zhao 402 et al., 2015). The mTOR/Akt pathway is a signaling pathway known to be integral in many aspects of OL

408
In contrast to IUGR and hyperoxia alone, a much larger number of upstream regulators were identified in large white matter tracts, however, this was not observed. Instead, the most significant downregulated myelin explanation is that IUGR/hyperoxia affects the descending motor tracts, including the corticospinal tract, that run

470
In one study, the cellular functions related to energy metabolism, stress, response, and maturation due to 471 oxidative stress were shown to be more pronounced in male versus female-derived OPCs that were exposed to 472 high oxygen for 24 hours (Sunny et al., 2020). In contrast, we demonstrate downregulation of myelin genes in 18 animals were exposed to oxygen over a longer time period of 14 days. The combination of IUGR with hyperoxia 475 resulted in significant downregulation of myelin genes in males compared to females. This is consistent with the 476 male sex being a risk factor for worse neurodevelopment outcomes in other types of perinatal brain injury.

477
Overall, these findings highlight the complex nature of perinatal brain injury. They also underscore the 478 detrimental effect that oxygen exposure can have on the developing white matter of IUGR infants. Hyperoxia is 479 well known to be implicated in the pathogenesis of bronchopulmonary dysplasia and retinopathy of prematurity 480 (Saugstad, 2001;Weinberger et al., 2002). Our study now adds to the increasing evidence that hyperoxia 481 negatively influences brain maturation and development and results in WMI (Felderhoff-Mueser et al., 2004; 482 Felderhoff-Mueser et al., 2005;Gerstner et al., 2008;Sunny et al., 2020). Additionally, our study importantly,