Hypoxia-Ischemia During a Neonatal Sensitive Period Alters Purkinje Neuron Dendritic Complexity Differently in Male and Female Rats

Perinatal hypoxia-ischemia (HI) is a major health issue with no effective therapies beyond head cooling. Notably, male infants are at a greater risk for HI and exhibit more extreme decits than females. Extensive clinical evidence indicates that perinatal HI impacts the developing cerebellum, yet this region has been largely ignored in preclinical models. Using a modied version of the Rice-Vannucci rat model for HI injury at postnatal day 10, we nd reductions in dendritic complexity of Purkinje neurons in males one week later. Females exhibited modest but opposite effects, with slight increases in dendritic complexity, based on Sholl analysis. A custom-made NanoString panel for quantifying mRNAs associated with development, inammation, and sex differences found almost no commonality in the response to HI in males versus females, with males up-regulating genes associated with microglia activity whereas females increased expression of a protective complement protein, but also of enzymes associated with endocannabinoids and prostaglandins. Both sexes exhibited a reduction in the GABA-synthetic enzymes, GAD-65 and GAD-67, after HI, suggesting increased excitotoxicity, but why males suffered more damage to the Purkinje neurons is unknown. system and hormone synthesis were upregulated in females. We then used western blot to measure the impact of HI on protein markers for the predominant cell types and neurotransmitters in the cerebellum. Glutamic acid decarboxylase (GAD) expression was reduced in the anterior vermis, suggesting that perinatal HI reduces GABAergic signaling. Immunohistochemistry for the major cell types in the cerebellum showed no changes in the number or density of Purkinje neurons, astrocytes, or microglia. Implicated pathways and cell types will inform future research into neuropathogenesis induced by perinatal HI. used following: biotinylated and horse were and slides, dehydrated with ascending alcohol concentrations, defatted with xylene, and coverslipped with DPX mounting medium. A design-based unbiased stereological method was performed to quantify microglia and phagocytic cups across the cerebellar vermis. We used StereoInvestigator 10 software (Microbrighteld) interfaced with an Olympus BX51 microscope and an MBF Bioscience 01-MBF-2000R-F-CLR-12 Digital Camera (Color 12 BIT). Counts were performed in lobules I, VIa, and X of the right vermis. The optical fractionator probe method was used to estimate cell and phagocytic cup densities. Both Iba1 + cells and phagocytic cups were counted at 20X magnication. The overall estimated volume of each counting region sampled in the vermis was used to normalize estimated counts to obtain an estimation of the average density of objects of interest (e.g., microglia, phagocytic cups), which was expressed as an estimated number/µm 3 (relative density measurement). A total of three cerebellar sections per animal were used with a physical distance of 225 µm between them to avoid any cell being counted twice during stereological analysis. Individual animals were treated as subjects for statistical analysis (n = 3-5/group/sex).


Introduction
Hypoxia-ischemia (HI) is the most common cause of brain injury in term infants that experience birth complications (e.g., abnormal presentation, placental disruption, umbilical cord compression, etc.) 1 .
Globally, intrapartum-related perinatal HI is responsible for 0.7-1.6 million deaths each year and remains the major cause of infant mortality and long-term morbidity, particularly in communities lacking access to advanced obstetrics 2 . The prognosis following perinatal HI in full-term infants depends on the intensity and duration of the insult. Infants with moderate or severe HI often die or develop lifelong motor, cognitive, and behavioral impairments, whereas those with mild HI often exhibit subtle cognitive and behavioral de cits later in life 3 .
Biological sex is both a risk factor and a prognostic factor for the consequences of perinatal HI. Among term infants, males have a greater incidence of HI 4,5 and poorer recovery and worse long-term outcomes relative to females with equivalent HI insults 6,7 . Despite this clear gender bias in frequency and severity, there remains a signi cant gap in understanding the biological origins of sex differences in outcome following perinatal HI. Greater susceptibility in males is likely mediated by a combination of factors including the organizational effects of androgens produced by the fetal testis during gestation, divergent cell death pathways in females and males, and sex differences in in ammatory signaling at homeostasis and after injury 8- 10 .
Beyond sex differences, there remains an additional, substantial gap in our understanding of the selective vulnerability of different regions of the brain to perinatal HI. The cerebellum has long been thought to be relatively spared from HI injury because it is not directly affected by ischemia due to the preferential shunting of blood ow to the vertebrobasilar system during hypoxia 11,12 . Over the last two decades, however, advances in magnetic resonance imaging have enabled the detection of subtle microstructural features and transformed our understanding of perinatal HI 13,14 . Clinical studies increasingly nd profound yet diffuse injury leading to impaired development of the cerebellum 15,16 . Further, many of the long-term de cits induced by perinatal HI overlap the domain-speci c cognitive and motor de cits associated with cerebellar injury 17 . Despite this robust clinical evidence, the cellular impact of perinatal HI on the cerebellum remains largely under-investigated in animal models, especially with respect to sexrelated differences.
A unique feature of the cerebellum is its prolonged developmental trajectory as it is one of the rst brain structures to differentiate and one of the last to fully mature. This makes it particularly vulnerable to insults during development. A substantial fraction of cerebellar development occurs postnatally in mammals, including neuronal proliferation, neuronal migration, apoptosis, and dendritic spine and synapse formation [18][19][20] . This sensitive developmental period extends into the second postnatal year in humans and occurs during the rst weeks of life in rodents 21,22 . A sensitive period to in ammation in the postnatal cerebellum that corresponds to the time around birth in humans, the time of greatest risk for HI, has been described for the laboratory rat [23][24][25] .
In the present study, the impact of perinatal HI was assessed with attention to whether it would impair cerebellar development in a sex-dependent manner and lead to histopathological abnormalities at the cellular level. A well-established Sprague-Dawley rat model of moderate HI brain injury occurring at postnatal day 10 (PN10), was used to evaluate cerebellar development 7 days post-injury (PN17) in males and females. To analyze the impact of HI on Purkinje neuron branching in more detail, we used a virus expressing eGFP to sparsely label Purkinje neurons. In males only, perinatal HI reduced Purkinje dendritic arbor complexity, such that the dendritic tree branched less frequently. Using a targeted, custom gene-expression array, we found altered expression of genes related to the phagocytic activity of microglia in the male cerebellar vermis, while genes related to the endocannabinoid system and hormone synthesis were upregulated in females. We then used western blot to measure the impact of HI on protein markers for the predominant cell types and neurotransmitters in the cerebellum. Glutamic acid decarboxylase (GAD) expression was reduced in the anterior vermis, suggesting that perinatal HI reduces GABAergic signaling. Immunohistochemistry for the major cell types in the cerebellum showed no changes in the number or density of Purkinje neurons, astrocytes, or microglia. Implicated pathways and cell types will inform future research into neuropathogenesis induced by perinatal HI.

Results
Perinatal HI reduces Purkinje neuron dendritic arborization in males only Central vermis (LVIa): To test the dendritic locations at which HI had the greatest effect on Purkinje neurons, Sholl analysis was performed and compared between groups (See Fig. 1a for representative confocal images of Purkinje neurons). Two cohorts were analyzed, each by separate investigators blinded to treatment and sex (See Supplementary Fig. 1 online) to replicate the data. The data were then combined and analyzed across cohorts. HI signi cantly altered dendritic arborization in Purkinje neurons in a sex-dependent manner (main effect of distance p < 0.0001, sex X treatment interaction p = 0.017, distance X treatment X sex interaction p = 0.001). Sholl analysis showed a clear decrease in the dendritic branching of Purkinje neurons in males after HI compared to same-sex sham controls (Fig. 1b). The highest number of intersections were observed within 50-130 µm from the center of the cell soma. In HI males, the number of dendritic branches was markedly reduced in this intermediate region (80-130 µm from soma; Table 1). Intriguingly, HI in females signi cantly increased proximal dendritic intersections (20-70 µm from soma; Table 1) and decreased distal dendritic intersections (140-150 µm from soma; p = 0.009 and p = 0.025), when compared to female controls (Fig. 1b). HI-induced dendritic alterations were prominent and diverged between sex, with HI males exhibiting marked suppression of Purkinje cell dendritic outgrowth compared to HI females (70-100 and 120 µm from soma; Table 1; Fig. 1c). Lastly, sex differences between sham groups were discordant. Sham males had more dendritic intersections than sham females at 60 µm from the soma (p = 0.004) but fewer dendritic intersections at 140 and 160 µm from the soma (p = 0.031 and p = 0.041, respectively; Fig. 1c). Female pups consistently had a greater average dendrite area and total dendrite area than males (main effects of sex p = 0.029 and p = 0.08, respectively), but there were no main effects of treatment or sex by treatment interactions (Fig. 1d, 1g). The average dendrite length was unaffected by HI or sex (Fig. 1e), but there was a main effect of treatment (p = 0.048) and a treatment by sex interaction (p = 0.014) primarily driven by a male-speci c reduction in total dendrite length following HI (p = 0.0134; Fig. 1h). The average number of dendrite segments was unaffected by HI, sex, or an interaction between the two (Fig. 1f).
Nodular vermis (LX): Sholl analysis revealed HI altered Purkinje cell dendritic development in the nodular vermis (See Fig. 2a for representative confocal images of Purkinje neurons). In males perinatal HI caused a leftward shift in the Sholl curve compared to same-sex control conditions (Fig. 2b), resulting in an increase in the number of proximal dendrites (20-50 µm from soma; Table 2), the Sholl curve peaking markedly closer to the soma (40 vs 100 µm from the soma), and a decrease in dendrites 90-120 µm from the soma ( Table 2). Like the central vermis, female Purkinje neurons in the nodular vermis had more proximal and medial dendrites following HI (20-70 µm from soma; Table 2). Among sham groups (Fig. 2c), males had more medial dendritic intersections than females (90-100 µm from soma; p = 0.013 and p = 0.031). HI males had signi cantly more dendrites at 40 µm from the soma than HI females (p =

NanoString
Using a custom designed NanoString array, we examined the impact of perinatal HI on gene expression associated with in ammation, phagocytosis, steroids/autacoids, masculinization/feminization, neurogenesis/synaptogenesis, neurotransmitters, cell markers, and genomic imprinting in the cerebellar vermis at PN17. HI males and females were compared to same-sex Sham animals to assess HI-related transcriptional changes in the cerebellar vermis (See Supplementary Fig. 2

Western blot and immunohistochemistry
Western blot and immunohistochemistry were used to assess cerebellar protein expression 7 days after injury (Fig. 4b).
Using western blot, we rst investigated the effect of perinatal HI on protein markers for the predominant cell types and neurotransmitter in the cerebellum. Because there is well-known variability in protein expression signatures across the cortex of the cerebellar vermis, analysis was performed on samples composed of the anterior (lobules I-V) and central (lobules VIb-VII) transverse zones to detect regional effects of HI (Fig. 4a). In the anterior vermis of males and females (Table 3)  For enhanced spatial resolution, immunohistochemistry was used to analyze the in uence of perinatal HI on Purkinje neuron number using CALB immunoreactivity at PN17 (Fig. 5). The density of CALB immunopositive Purkinje neurons (Table 4) was greater in males than females in the central (main effect of sex p = 0.007) and nodular (main effect of sex p = 0.045) vermis, but this was not in uenced by HI ( Fig. 5f and 5h). Neither treatment nor sex in uenced Purkinje neuron number in the anterior, central, or nodular vermis (Fig. 5c, 5e, and 5f). There were also no effects of sex or treatment on the density of CALB immunopositive Purkinje neurons in the anterior vermis (Fig. 5d). Table 4 Immunohistochemical analysis of the number of Calbindin immunoreactive soma (CALB + / mm 2 ) and Calbindin optical density (OD; arbitrary units) in cerebellar vermal lobules VIa (LVIa).  There was no effect of HI on the total number of microglia, the number of phagocytic microglia, or the number of phagocytic cups in the anterior, central, or nodular vermis, as determined by immunohistochemistry for IBA1 ( Fig. 6 and Fig. 7).
Expression of GFAP was quanti ed to investigate the effect of HI on astrocyte number and activity (Fig. 8). There were no main effects of sex, HI, or an interaction between the two on astrocyte number or density in the anterior, central, and nodular vermis.
Lastly, we measured the granular level (GL) area in vermal lobules I, VIa, and X in PN17 rats (Fig. 9). No sex differences or effects of HI were found.

Discussion
Until recently, the cerebellum and its vulnerability to perinatal HI have been underappreciated and unexplored. This neglect, in part, was due to the long-held belief the cerebellum functioned only in motor coordination and did not make important contributions to non-motor, cognitive and emotional domains. Using a model of term HI in the rat, we report a diffuse injury to the cerebellum seven days post-injury, with a focused impact on Purkinje neurons that differs in males versus females. When compared to the same-sex control group, male Purkinje neurons had fewer dendrites after HI while female Purkinje neurons tended to have more. To understand the origins of these Purkinje neuron alterations, we explored a targeted list of genes associated with in ammation, steroid hormones, autacoids, neurogenesis, synaptogenesis, and phagocytosis. We found a small number of genes altered in response to HI, with very little overlap between males and females. Western blot quanti cation of proteins to survey gross abnormalities in different cerebellar cell types and their neurotransmitters revealed decreased glutamic acid decarboxylase (GAD) expression, suggesting that a loss of inhibitory regulation may be the source of damage to Purkinje neurons. Immunohistochemistry and volumetric analyses found no evidence of gross changes to microglia, astrocytes, granular neurons, or the number of Purkinje neurons.
Findings from a line of research, initiated a decade ago, determined a latent sensitive period in cerebellar development during the second postnatal week in the rat [23][24][25] . This sensitive period in rats is analogous to full-term birth in humans, a time when males experience greater vulnerability to HI than females [4][5][6][7]26 .
During this time endogenous prostaglandin E2 (PGE2), a lipid best known to mediate in ammation, is elevated in both sexes 23,25 . If PGE2 levels are thrown out of balance (i.e., increased or decreased) by in ammation or injury, the growth of Purkinje dendritic trees is altered, and later social cognitive and social behaviors are impaired in males only 24,25 . Interestingly, mRNA analysis, while not comprehensive, found evidence for facilitation of the prostaglandin synthesis pathway after HI in females. Expression of fatty acid amide hydrolase (Faah), which is the enzyme that produces the most important precursor in prostaglandin synthesis, and prostaglandin receptor subtype EP1 (Ptger1) were upregulated after HI in females only. This is intriguing given that PGE2 stunts the development of Purkinje neuron dendritic tree arborization, and EP1 activation is generally considered a source of neuronal damage following HI in cortical brain regions 25,27 . We did not measure either PGE2 or EP1 protein in this study, but this nding nevertheless highlights the central role of prostaglandin signaling in cerebellar development.
In the term infant, HI is typically associated with diffuse tissue loss, or hypoplasia, and damage to deep gray matter regions including the basal ganglia and thalamus 28 . Until very recently, perinatal HI was mainly characterized as a cerebral injury in the clinical setting. Nevertheless, the injury most often identi ed by MRI in the cerebellum of term infants experiencing HI is a bilateral, symmetrical decrease in hemisphere volume 29 . Though effects of perinatal HI on the cerebellum have been observed in experimental models, results are di cult to interpret due to inconsistencies in the timing of injury. The Rice-Vannucci method (i.e., permanent unilateral carotid ligation + period of systemic hypoxia) is by far the most widely used animal model to replicate HI. Initially developed in adult rats, the model has since been adapted for use in perinatal rodents, with PN7 pups being the current standard. Studies inducing HI in PN7 pups, which is best suited as a model of preterm neonatal injury, report reductions in the number of Purkinje neurons in both the right and left cerebellar hemispheres and abnormalities in Purkinje and granule cell dendrites 30 . It is currently considered that the PN10 Rice-Vannucci method, used here, is a more appropriate model for human infants at term 31 . To determine the effects of term-equivalent HI on Purkinje neuron morphological characteristics we used an AAV to sparsely label individual Purkinje neurons throughout the cerebellar vermis. Purkinje cell dendritic morphology was reconstructed in 3D and revealed that HI in the neonatal rat disrupts the dendritic branching of Purkinje neurons in the cerebellar vermis. Moreover, we nd that the effect of HI on Purkinje neurons differs between the sexes, with males displaying a reduction in dendritic arborization and females displaying modest but signi cant dendritic outgrowth. These differences suggest either that repair mechanisms are initiated earlier in females or that females are protected from HI. Thus, optimal windows for intervention might differ between the sexes.
Considering the timing of HI injury, our results in the cerebellum are consistent with those reported in studies modeling preterm neonatal HI injury. Cerebellar development is delayed relative to the rest of the central nervous system and undergoes signi cant cytoarchitectural changes after birth 19,21 . The second postnatal week in the rat cerebellum, which corresponds to birth in the human, is the most dynamic period of Purkinje neuron dendritic remodeling and a developmental window sensitive to extrinsic factors 32 . During this period, the Purkinje neuron dendritic tree is rapidly expanding and organizing. The Purkinje neuron dendrites, which develop from multiple perisomatic processes, are pruned, leaving only a single primary dendrite that is innervated by a single climbing ber. Simultaneously, Purkinje dendrites undergo tremendous growth and are remodeled from a multiplanar to monoplanar arrangement 21 . The calcium-binding protein Calbindin D28k is expressed by many types of neurons throughout the brain but is disproportionately abundant in Purkinje neurons, along with discrete populations of interneurons in the cerebral cortex and hippocampus, and so it is predominately used as a marker to identify these cell types 37 . Calbindin serves a threefold function as a Ca2+ buffer, transporter, and sensor 38 . While the precise function and regional speci city of Calbindin in Purkinje neurons are not fully elucidated, high expression is likely due to unique electrophysiological properties of Purkinje neurons, such as their high ring rate under resting conditions and characteristic complex spikes 35,39 . Androgens developmentally regulate calbindin as well as several calcium binding proteins and cation transporters, which can increase vulnerability to excitotoxic cell death [40][41][42] . We observed increased levels of calbindin in male Purkinje neurons, with and without HI, suggesting this may be a source of increased sensitivity to altered excitability, thereby increasing male vulnerability.
Cumulative evidence supports a link between abnormal cerebellar development and major neurodevelopmental disorders that present with a strong gender bias, including autism spectrum disorders (ASDs), attention de cit hyperactivity disorder, and developmental dyslexia 43,44 . For example, perinatal cerebellum injury is the leading risk for ASDs, which are diagnosed more frequently in males than females. The two most consistent cerebellar abnormalities found in people with ASDs are Purkinje neuron loss (determined by postmortem histology) and reduced cerebellar volume (determined by MRI) 45 . The latter is believed to result from reductions in Purkinje neuron number and size 46 . Although developmental damage to the cerebellum, particularly Purkinje neurons, is strongly associated with neurodevelopmental disorders, the underlying pathogenesis and resultant behavioral defects, remain poorly understood. Purkinje neurons are the central components of all cerebellar circuits, the rst neurons to populate the cerebellar cortex, and based on their ability to regulate the development of both inhibitory and excitatory neurons, they are considered key modulators of cerebellum development 47,48 . Lurcher mutant mice, which are characterized by postnatal degeneration of cerebellar Purkinje neurons, display ataxia, deterioration of cognitive functions, increased activity, and increased repetitive behaviors 49,50 . In rats, disruption of Purkinje neuron development by in ammation during the second postnatal week leads to stunting of Purkinje neuron dendrites and impairs juvenile social behavior only in males 25 . Early damage to the cerebellum can have broad, enduring consequences on the developmental trajectory of the cerebellum and behavior, but further research is needed to understand this complex relationship.
It remains unclear what role microglia are playing in the progression of cerebellum injury following perinatal HI. Microglia are key contributors to neurodevelopment and mediate distinct neuroimmune responses in the healthy and injured brain. Although much of our knowledge derives from studies in the cerebrum, microglia function in neurodevelopment by promoting cell genesis, controlling cell numbers, synaptic pruning, and apoptosis [51][52][53][54] . In the developing cerebellum, microglia promote apoptosis of immature Purkinje neurons and subsequently clear debris by phagocytosis 55 . In perinatal brain injury, one of the major pathogenic factors is microglia-mediated neuroin ammation, and most studies associate microglia activation detected at initial stages of injury with exacerbation of brain injury 56-59 .
Using immunohistochemical methods we detected no changes in microglia numbers or morphology in the cerebellum one week after HI, which is likely a result of the time between injury and assessment.
Interestingly, our gene expression results hint at a role for microglia in the progression of cerebellar injury after HI, given that several genes associated with microglia activation were increased in males after HI, whereas in females there was an increase in Sirpa, a complement protein known to prevent phagocytosis of stressed cells 60 .
There are limitations to the current study. First is the intrinsic variability of the widely used Rice-Vannucci model of term HI. Procedural times and duration of iso urane exposure, which we standardized and kept to a minimum, are believed to cause inconsistencies in brain injury 61,62 . Second, we did not investigate the behavioral outcomes of perinatal HI injury in adult rodents. Although this model gives rise to welldocumented behavioral phenotypes, including impaired spatial learning and memory, impaired motor control, and cognitive and sensory processing de cits, it would be impossible to ascertain what components of these behaviors rely on the cerebellum as opposed to the more frankly damaged cerebral cortex and hippocampus. These limitations highlight the need for a focal, term-equivalent HI model targeting the cerebellum.
Collectively, the present studies add to a growing body of literature demonstrating that HI insult causes diffuse, bilateral injury in the term-equivalent cerebellum and that Purkinje neurons are particularly susceptible. To our knowledge, this is the rst report of HI effects on Purkinje neuron dendritic branching patterns. A remaining question is the mechanism underlying the effects of HI on Purkinje neuron arborization. The timing of injury coincides with the sensitive period for dendritic remodeling of Purkinje cells, which is modulated by excitatory afferent bers, trophic factors, and hormones. We also report a reduction in GAD signaling in the cerebellar cortex following HI. This, along with the spatio-temporal location of injury, suggests that synaptic inputs to Purkinje neurons may be a target of HI. Further research is clearly needed to explore this and better understand how sex modulates HI in the termequivalent cerebellum.

Methods
Animal husbandry and hypoxia-ischemia procedure Timed-pregnant female Sprague-Dawley rats were obtained from Charles River (Fredrick, MD) and checked daily to determine the day of birth, which was designated as PN1. Litters were culled to 12 pups, when necessary, with equal numbers of males and females when possible, and randomly assigned to experimental conditions on PN1. To model HI in the term-equivalent infant brain, PN10 pups were subjected to a well-established 63,64 , modi ed version of the Rice-Vannucci method 65 . Brie y, pups were placed in a supine position and 3.0% iso urane was delivered via nose cone for 1 min, until anesthesia.
Iso urane was then reduced to a maintenance level of 1.5% for the remainder of the surgery (~4.5 min).
The right internal carotid artery was isolated, ligated twice, and severed between the ligations. The pups were then placed in a large, open jar within a circulating water bath (37°C) for 25 min to recover from anesthesia, as indicated by normal breathing and color. Pups were then returned to the dam for 1 h to permit feeding after which they were placed in temperature-controlled chambers and exposed to 8% O 2 :98% N 2 for 60 min to induce hypoxia. Sham-operated control pups were anesthetized for the average surgery time, incised, and sutured, but did not undergo carotid artery ligation or exposure to hypoxia.

Adeno-associated virus (AAV) injection and Purkinje neuron reconstruction
To achieve sparse labeling of Purkinje neurons at a resolution su cient to complete analytical reconstruction, we used a viral construct purchased from the UPenn Viral Core Gene expression array (NanoString) Male and female rat pups were subjected to HI or Sham surgery on PN10 and euthanized with a lethal dose of sodium pentobarbital (650 mg/kg, intraperitoneally) on PN17 (n=6/group/sex). The cerebellum was divided along the midsagittal plane and the cerebellar vermis was separated from the hemispheres.
For total RNA extraction, the right vermis was incubated in 500 µl of Qiazol (Qiagen #79306) with 1 mm zirconia/silica beads (BioSpec #11107911Oz) and homogenized at 4 o C using a Bullet Blender for 5 min were added to the samples and centrifuged at 12,000 g for 15 min at 4 o C allowing phase separation to occur. The clear layer was removed and mixed with 475 µl of molecular biology grade 70% ethanol and pipetted onto a RNeasy spin column and RNA isolation proceeded according to the manufacturer's protocol (Qiagen #74106/74104). Isolated RNA was eluted in a 40 µl volume and tested using an Agilent 2100 Bioanalyzer to ensure it met speci cations for purity (RNA integrity number ≥ 9) and concentration (≥ 12.5 ng/µl).
Gene expression pro ling of HI response was performed using a custom NanoString nCounter codeset (rat) that we designed to allow for the multiplexed assessment of 171 genes related to in ammation, phagocytosis, steroids/autacoids, masculinization/feminization, neurogenesis/synaptogenesis, neurotransmitters, cell markers, and imprinted genes. RNA was processed on the NanoString nCounter Analysis System by the University of Maryland Baltimore's Institute for Genome Sciences using NanoString's standard consumables and assay procedures 67 . NanoString Reporter Code Count (RCC) les were analyzed using nSolver analysis software (Version 4.0). All samples passed QC parameters.

Western blot
On PN10 male and female rat pups underwent Sham or HI surgery and were euthanized by sodium pentobarbital on PN17, as described above (n=4-6/group/sex). The cerebellum was divided along the midsagittal plane and the cerebellar vermis was separated from the hemispheres. The left and right Membranes were imaged in both the 700 and 800 nm channels using an Odyssey CLx scanner (LI-COR) and quanti ed using ImageStudio software (LI-COR). The proteins of interest were detected as a band with a relative molecular weight of 67 kDa for GAD67, 65 kDa for GAD65, 50 kDa for GFAP, 42 kDa for β-ACTIN, 36 kDa for GAPDH, 28 kDa for Calbindin, and 17 kDa for IBA1. The optical intensity for each protein of interest was normalized to GAPDH and β-ACTIN and the resulting values were averaged for each sample. Each sample was normalized to a standard internal control sample to allow for comparison across membranes. Relative protein quanti cation is expressed as "Normalized Optical Intensity." Immunohistochemistry (IHC), histology, and quanti cation Male and female rats were subjected to HI or Sham surgery on PN10, as described above. On PN17, rats were anesthetized with sodium pentobarbital and perfused with ice-cold PBS, followed by 4% PFA. Brains Vectastain Elite ABC Kit, Vector Laboratories) in TBST for 1 h, and visualized using DAB chromagen (0.05% 3,3′-diaminobenzidine, 0.006% hydrogen peroxide; Sigma-Aldrich D-5905). The DAB reaction was allowed to proceed until completion (~3-5 min). Sections on slides were dehydrated in a series of ascending ethanol, defatted in 100% xylene, and coverslipped with DPX mounting medium (VWR International Inc).
Bright eld images of cerebellar vermal lobules I, VIa and X were captured on a Keyence BZ-X700 microscope using a 20x (0.75 NA) objective and BZ-X Viewer software and were subsequently quanti ed using ImageJ software. The total number of GFAP + cells in the granule and Purkinje layers (GL and PL, respectively) were counted and this number was divided by the combined area of the GL and PL to obtain the density of GFAP + cells. Similarly, the number of CALB + soma within the PL was counted and normalized to the combined area of the PL and molecular layer (ML) to obtain the density of CALB + cells.
In addition, CALB and GFAP relative amounts were analyzed by measuring the integrated optical density (OD) as described in 71,72 . In brief, bright eld images were converted to grayscale, and the Image J Optical Density calibration tool was used to determine the intensity of CALB immunostaining in the PL and GFAP immunostaining in the GL and PL. For each image, the background OD was measured and subtracted from the absolute OD values. Relative ODs were calculated by dividing the optical density by the measured tissue area. All GFAP and CALB measurements were performed in lobules I, VIa, and X of the left and right cerebellar vermis. A total of 3 sections per animal were analyzed and the values were averaged. Individual animals were treated as subjects for statistical analysis (n = 3-7/group/sex). sulfate, and 0.006% hydrogen peroxide; Sigma-Aldrich) until the reaction was complete (~8-10 min).
Finally, sections were exhaustively rinsed in PBS, mounted on silane-coated slides, dehydrated with ascending alcohol concentrations, defatted with xylene, and coverslipped with DPX mounting medium.
A design-based unbiased stereological method was performed to quantify microglia and phagocytic cups across the cerebellar vermis. We used StereoInvestigator 10 software (Microbright eld) interfaced with an Olympus BX51 microscope and an MBF Bioscience 01-MBF-2000R-F-CLR-12 Digital Camera (Color 12 BIT). Counts were performed in lobules I, VIa, and X of the right vermis. The optical fractionator probe method was used to estimate cell and phagocytic cup densities. Both Iba1 + cells and phagocytic cups were counted at 20X magni cation. The overall estimated volume of each counting region sampled in the vermis was used to normalize estimated counts to obtain an estimation of the average density of objects of interest (e.g., microglia, phagocytic cups), which was expressed as an estimated number/µm 3 (relative density measurement). A total of three cerebellar sections per animal were used with a physical distance of 225 µm between them to avoid any cell being counted twice during stereological analysis. Individual animals were treated as subjects for statistical analysis (n = 3-5/group/sex). (LI, LVIa, and LX). Measurements were averaged from the left and right vermes of 3 sections and individual animals were treated as the subjects for statistical analyses. Experimental subjects were categorized into two groups: Sham and HI (n = 3-5/group/sex).

Statistical analysis
Purkinje neuron morphology, western blot, and IHC data are expressed as the mean and SD and follow homogeneity of variance. For Sholl analysis, we rst used a three-way repeated measures ANOVA with treatment, sex, and distance from soma as xed factors. If a signi cant interaction was detected, a post hoc two-way unpaired student's t test was used to compare the number of dendrites for each distance from the soma. For all other morphological measurements and western blot data, we conducted two-way ANOVAs with sex and treatment as xed factors followed by post hoc pairwise comparisons. All statistical tests were computed in GraphPad Prism 7.
NanoString expression data was normalized to the geometric mean of: (1) positive and negative controls included in the NanoString Codeset to adjust for background thresholding and to normalize sample input and differences in hybridization, and (2) the geometric mean of ten reference genes (Alas1, Eef1g, G6pd, Gapdh, Oaz1, Polr2a, Ppia, Rpl19, Sdha, Tbp, and Dlk1) for codeset content normalization. The reference gene list was collated based on a literature review of housekeeping targets that yield best results for quantitative gene expression analysis studies and each reference gene showed minimal variation among the treatment groups. Signi cance was determined by a two-tailed t-test of the log-transformed normalized data within the nSolver Software Package and graphed using GraphPad Prism 7. Signi cance was denoted when p ≤ 0.05. shown as mean ± SD. See Supplementary Fig. 1 for replicate data, which con rms the validity of the changes observed in Purkinje neuron morphology following HI.   sham female. *p < 0.05, **p < 0.01, ***p < 0.001, and #p = 0.07; n = 4-6 animals /group/sex. Data is shown as mean ± SD. Non-signi cant western blot data is shown in Supplementary Fig. 3.

Figure 5
Purkinje neuron analysis in the vermis. (a) Sagittal representation of the vermis stained with Calbindin (4x magni cation). Insets depict higher magni cation of Purkinje neurons in the posterior vermis (20x). (b) Cerebellar lobules I, VIa and X were used for morphometric and optical intensity analysis. (d) In LI, no signi cant differences were found after HI in numbers of CALB+ soma or (d) CALB optical density. (e) In cerebellar LVIa, there were no signi cant differences found after HI in any of the treated groups for numbers of CALB+ somas. (f) However, optical density for CALB+ differences were found between males and females. (g) No signi cant differences were found after HI in the number of CALB+ somas for cerebellar LX, but the (h) optical density analysis showed higher levels of CALB+ in males compared to females. In (c-h) data collapsed across sex are shfown in grayscale. *p < 0.05, **p < 0.01, and #p = 0.08.
Data is shown as mean ± SD.