Brain Development after Neonatal Intermittent Hyperoxia-Hypoxia in the Rat Studied by Longitudinal MRI and Immunohistochemistry

Background Neonatal intermittent hyperoxia-hypoxia (IHH) is involved in the pathogenesis of retinopathy of prematurity. Whether similar oxygen fluctuations will create pathological changes in the grey and white matter of the brain is unknown. Methods From birth until postnatal day 14 (P14), two litters (total n = 22) were reared in IHH: hyperoxia (50% O2) interrupted by three consecutive two-minute episodes of hypoxia (12% O2) every sixth hour. Controls (n = 8) were reared in room-air (20.9% O2). Longitudinal MRI (Diffusion Tensor Imaging and T2-mapping) was performed on P14 and P28 and retinal and brain tissue were examined for histopathological changes. Long-term neurodevelopment was assessed on P20 and P27. Results Mean, radial and axial diffusivity were higher in white matter of IHH versus controls at P14 (p < 0.04), while fractional anisotropy (FA) was lower in the hippocampal fimbria and tended to be lower in corpus callosum (p = 0.08) and external capsule (p = 0.05). White matter diffusivity in IHH was similar to controls at P28. Higher cortical vessel density (p = 0.005) was observed at P14. Cortical and thalamic T2-relaxation time and mean diffusivity were higher in the IHH group at P14 (p ≤ 0.03), and albumin leakage was present at P28. Rats in the IHH group ran for a longer time on a Rotarod than the control group (p ≤ 0.005). Pups with lower bodyweight had more severe MRI alterations and albumin leakage. Conclusion IHH led to subtle reversible changes in brain white matter diffusivity, grey matter water content and vascular density. However, alterations in blood-brain barrier permeability may point to long-term effects. The changes seen after IHH exposure were more severe in animals with lower bodyweight and future studies should aim at exploring possible interactions between IHH and growth restriction.


Introduction
Preterm born infants may be subject to pathological conditions in several organ systems like retinopathy of prematurity (ROP) and preterm encephalopathy. The severity of ROP is a predictor of neurodevelopmental outcome [1] and parallels postnatal head growth deficit [2]. The retina has been called a "window to the brain" since it is an extension of brain tissue [3], and it is conceivable that environmental insults during critical periods in development may contribute to both ROP and preterm encephalopathy. Exposure to unphysiological levels of oxygen may be such an environmental insult. In the retina it is the fluctuations and range between high (hyperoxic) and low (hypoxic) levels of oxygen that create ROP rather than either exposure alone [4,5], and even small fluctuations in oxygen concentration around a normoxic mean induce pathological changes [6]. Hyperoxia due to treatment combined with hypoxia due to apnoea is common in the sick premature child, and clinicians question whether such unphysiological oxygen levels will also affect long-term brain development [7]. In animal models this has been explored after neonatal hypoxia [8][9][10] or hyperoxia [11,12]. However, brain development following neonatal fluctuating oxygen levels, an exposure that would more closely mimic a clinical setting, is largely unexplored.
In ROP the coordinated formation of the neurovascular unit comprising neurons, endothelial cells, pericytes and astrocytes is disrupted [13]. Obliteration of immature vessels by hyperoxia leads to tissue hypoxia that in turn stimulates uncontrolled neovascularization, microbleedings, and eventually injury to the rapidly growing neural tissue of the retina [14]. The neurovascular unit may also be a target in the brain of preterm born children, a population where long-term pathological changes in grey and white matter are abundant [15,16]. Periventricular white matter is poorly vascularized due to arterial end-zones, and disturbances of vascular supply is likely involved in the development of periventricular leukomalacia (PVL) [17]. Furthermore, paucity of endothelial pericyte [18] and astrocyte coverage [19] in the germinal matrix of preterm born infants, a predilection site for intraventricular haemorrhage, indicates fragile vasculature. Vascularization of both the retina [20] and brain [21] occur in parallel to the brain growth spurt from gestational week 20 until term in humans and from birth until postnatal day 14 (P14) in the rat, a species that at birth is comparable to a very preterm human infant [22]. Both intermittent and continuous neonatal hypoxia reduce longterm myelination and induce angiogenesis in the rat brain [10,23]. Furthermore, neonatal hyperoxia causes microvascular degeneration in grey matter [24] and maturation-dependent cell death in white matter [25,26]. It is conceivable that, both in the retina and brain, fluctuations between high and low oxygen levels disturb the coordinated formation of the neurovascular unit in a period of major growth and differentiation in both organs. Therefore, exposure of neonatal rats to a profile of intermittent hyperoxia-hypoxia (IHH) that cause ROP may induce a parallel disruption of neural and vascular tissue in the brain with ensuing white and grey matter injury.
To investigate this hypothesis we conducted a study of longterm brain development in rats exposed to IHH from birth until P14. Outcome was evaluated with multimodal in vivo magnetic resonance imaging (MRI) combined with neurodevelopmental testing and detailed immunohistochemical studies of bloodbrain barrier (BBB) permeability and vascular density.

Ethics Statement
Experiments were conducted in accordance with Guidelines set by the Norwegian Ethics Committee for Animal Research and approved by the responsible governmental authority (Forsøksdyrutvalget, permit no: 3748).

Oxygen Profiling
An A84 Oxycycler (Biospherix Ltd., Lacona, NY) was used to program a profile of IHH that create ROP in neonatal rats [27]. Continuous hyperoxia (50% O 2 ) was interrupted every sixth hour by three consecutive two-minute long episodes of hypoxia (12% O 2 ), each ten minutes apart ( Figure 1). The ramp time between hyperoxia and hypoxia was approximately two minutes. Bedding was changed during a 50% cycle on day seven. The chamber was then opened for less than three minutes. Inside the chamber, oxygen was controlled via a sensor and a computerized feed-back system where nitrogen was used to lower and oxygen was used to raise the oxygen concentration. CO 2 , temperature and relative humidity was continuously monitored and kept within physiological levels via the use of Soda lime (Anmedic AB, Stockholm, Sweden), ventilation and a built-in fan.

Experiment Groups
Time-mated Sprague-Dawley rats were purchased from Scanbur AS (Nittedal, Norway). Within four hours after delivery of the last pup, litters with dam were placed in specialized controlled oxygen chambers in their cages (A-30524, Biospherix Ltd.). To ensure a uniform age at start of exposure whole litters were exposed from birth. Two litters (A: n = 13 and B: n = 9) were exposed to IHH while one litter (controls: n = 8) was exposed to room-air (20.9% O 2 ) from P0 to P14. Animals were kept on a 12:12 hours light: dark cycle and had food and water ad libitum. Half the animals were euthanized on P14 and the other half on P28 after MRI and neurodevelopmental testing.

Magnetic Resonance Imaging
MRI was performed on P14 (IHH: n = 22; control: n = 8) and P28 (IHH: n = 12; control: n = 4) using a 7T magnet (Biospec 70/20 AS, Bruker Biospin MRI, Ettlingen, Germany) with water-cooled (BGA-12, 400 mT/m) gradients. All animals were imaged longitudinally until euthanization. A volume resonator was used for radiofrequency (RF) transmission and actively decoupled head surface coils were used for RF reception. During scanning the anaesthetized (2% isoflurane in 30% O 2 /70% N 2 ) pups lay prone in a dedicated water-heated bed (Bruker Biospin MRI) and the head of every animal was fixated in the same position with tooth-bar, nose-mask and polystyrene. Temperature and respiration were monitored during the scanning procedure. On P14 and P28 coronal T 2maps were obtained with a turbo spin-echo (RARE) sequence:

Magnetic Resonance Image Analyses
In-house developed software was used for calculating T 2maps (MATLAB ver. R2010a, Math Works Inc., Natick, MA), and ImageJ (ver. 1.42q, National Institute of Health, Bethesda, MD) was used for image analyses. T 2 -maps were calculated by fitting a mono-exponential model to the signal intensity of the images with different TE-values. Regions of interest (ROI) were drawn in the parasagittal cortex, thalamus, hippocampus, putamen and corpus callosum in two slices corresponding to -3.25 mm and -1.5 mm from the bregma [28] ( Figure S1). For brain volume measurements the areas of both hemispheres were drawn manually on the T 2 -maps in each image slice and the total brain volumes were calculated and compared between groups. DTI-analyses were performed with the tools of the FMRIB software library (FSL ver. 4.1.4, Oxford Centre for Functional MRI of the Brain, Oxford, UK; www.fmrib.ox.ac.uk/ fsl). Images were pre-processed to reduce artefacts due to motion and eddy current distortions by affine transformation and co-registration of the diffusion encoded images to the b0 images. Brains were segmented out using the Brain Extraction Tool before FDT ver. 2.0 (FMRIB's Diffusion Toolbox; both part of FSL) was used to fit a voxelwise diffusion tensor model to the diffusion image data. Maps for the fractional anisotropy (FA), mean, radial and axial diffusivity were created. To limit partial volume effects, ROI were manually drawn in the centre of white matter structures at all relevant image slices on the FA maps and combined to five volumes of interest: corpus callosum (body), external capsule, internal capsule, hippocampal fimbriae and white matter (comprising all of the aforementioned areas, see Figure S1). Mean FA, radial, axial and mean diffusivity were calculated in each of these volumes of interest in each animal. Due to low signal-to-noise-ratio (SNR) in the deeper brain structures in the diffusion-weighted images on P28, masks were only drawn in the body of the corpus callosum where SNR was adequate for robust calculations of DTI-metrics at this time-point (IHH: n = 6, control: n = 4).

Rotarod
On P20 and P27, rats were tested on a Rotarod (Bioseb, Vitrolles, France) in four consecutive trials with increasing rotation from 0 to 40 rpm within two minutes. The time the animals ran on the rotating rod was recorded.

Histology
Rats were euthanized by an overdose of pentobarbital (300 mg/kg; Vétoquinol, Lure Cedex, France). For fixation an intracardial perfusion of 4% paraformaldehyde (Fluka Chemie AG, Buchs, Switzerland) in phosphate-buffered saline (Oxoid Limited, Hampshire, UK) was given. Eyes were enucleated and the orbita were inspected for gross abnormalities. Thereafter the retina of the left eye was dissected and whole-mounted onto Super Frost object glass (Thermo Fisher Scientific Inc., Waltham, MA) while the right eye and the brain were paraffinembedded. Only a few retinas were successfully wholemounted in each group (IHH: n = 2; control: n = 1). Coronal brain sections corresponding to -3.25 mm from the bregma [28] and retinal cross-sections (4 μm) were cut and stained with Haematoxylin & Eosin (H&E; Cell Path Ltd., Newtown, UK and Sigma-Aldrich Inc., St. Louis, MO) and examined for morphological changes.

Immunohistochemistry
Brain sections were incubated with anti-rat-albumin (1:16000) (Nordic Immunology, Eindhoven, the Netherlands) for albumin leakage, Biotinylated Griffonia (Bandeiraea) Figure 1. Experimental profile of intermittent hyperoxia-hypoxia. Every sixth hour of hyperoxia (50% O 2 ) was interrupted by a cluster of three consecutive episodes of hypoxia (12% O 2 ) ten minutes apart, each of two minutes. This profile was applied continuously from within four hours after birth until postnatal day 14.
Simplicifolia Lectin I Isolectin B4 (1 μg/ml) (Vector Laboratories Inc., Burlingame, CA) for vascular density, anti-glial fibrillary acid protein (GFAP; 1:200) (Cymbus Biotechnology, Southampton, UK) for astrogliosis and anti-ED1-FITC (1:400) (Serotec, Raleigh, NC) for cluster of differentiation 68 (CD68) positive activated microglia. Sections were then incubated with labelled polymer HRP anti-rabbit (Dako A/S, Glostrup, Denmark), streptavidin-conjugated FITC (Vector Laboratories Inc.), horse-anti-mouse-biotin (Vector Laboratories Inc.) and rat-anti-FITC-biotin (Roche, Basel, Switzerland). Visualization was performed using a Vectastain ABC kit (Vector Laboratories Inc.) and a Diaminobenzidine (DAB) kit (Vector Laboratories Inc.). Images of brain sections were captured with a MIRAX MIDI system (Carl Zeiss MicroImaging GmbH, Jena, Germany) and examined. The presence (score = 1) or not (score = 0) of increased albumin immunoreactivity in the neuropil in the hippocampus, cortex, thalamus and hypothalamus was scored. Vessel density was quantified in three areas of the cortex (parietal and temporal), thalamus, hippocampus (CA2, CA3 and dentate gyrus) and periventricular white matter at x400 magnification using ImageJ and normalized relative to control. The result was averaged for each animal and the mean was compared between groups. GFAP and CD68-stained sections were evaluated qualitatively for changes between groups. Following incubation with Isolectin B4 (12.5 μg/ml) overnight and thereafter with streptavidin-conjugated FITC, retinal wholemounts were analyzed qualitatively for neovascularization and maturation of the retinal vascular bed.

Statistical Analysis
To test for differences in mean between experiment groups and between time-points in the same individual, paired and unpaired two-tailed Student's t-tests were applied, respectively. Between-litter differences were analyzed with one-way ANOVA followed by post-hoc tests with Bonferroni correction for multiple comparisons. A p < 0.05 was chosen as level of significance. Data are presented as mean ± 95% confidence interval. Analyses were performed in SPSS ver. 19.0 (SPSS Inc., Chicago, IL).

Retinal Changes
At P14 orbital macroscopic bleedings were observed in four out of ten IHH animals and in none of the controls. Furthermore, haemorrhage in the ganglion cell layer and/or inner and outer nuclear layer were seen in five out of nine animals in the IHH group, often in all three layers and on several locations (Figure 2 a-d). Similar bleedings were observed in one control animal in one location. At P28 areas of neovascularization were observed in retinal wholemounts of both IHH animals ( Figure 2 g-h) but not in the control (Figure 2 e-f), and the control retinal vasculature appeared more remodelled compared to IHH.

Histopathology
No pathological changes were seen in the H&E brain sections with regards to neuropil, neurons, glial cells, vessels or leptomeninges. There were especially no extravasations of erythrocytes, inflammation or eosinophilic degeneration in neurons observed.

Immunohistochemistry
A generalized positive immunoreaction for albumin was seen in both IHH and controls in the neuropil at P14 and in the hypothalamus, a circumventricular organ void of BBB [29], at P28 (Figure 3 a & c). However, in the IHH group focal perivascular albumin leakage above control levels was observed in the neuropil at P28 (Figure 3 e & f). Although albumin leakage was present in most animals in both litters in the IHH group (A: 6/7 animals versus B: 3/5 animals), litter A had multiple larger areas (Figure 3 e), while those of litter B were singular smaller areas similar to those observed in control animals. Cortical vascular density of the IHH animals was higher than controls at P14 (p = 0.005, see Figure 4) and tended to be higher in the thalamus (p = 0.07). At P28 vascular density was similar to controls. GFAP staining was not elevated in the IHH versus controls, nor was there increased activation of microglia at neither time-point (data not shown).

T 2 Maps
Visual evaluation of T 2 -maps did not reveal differences in grey or white matter. However, at P14 the measured T 2relaxation time was higher in the IHH group than in controls in cortex (p = 0.01) and thalamus (p = 0.03) ( Table 1), but similar on P28. In both groups T 2 -relaxation time decreased from P14 to P28 (p < 0.003). At P14 litter A had a higher mean T 2 relaxation time than litter B in all areas (p < 0.006) and than controls in cortex and thalamus (p < 0.001, Figure 5). There were no differences in T 2 -relaxation time in corpus callosum at P14 between experiment groups (IHH: 78.4 ± 0.7 vs. controls: 78.3 ± 4.0, p = 0.7) and P28 (IHH: 55.6 ± 1.3 vs. controls: 54.6 ± 3.6, p = 0.5).

Diffusion Tensor Imaging
At P14, mean, radial and axial diffusivity were higher in the IHH than in the control group in all areas of white matter (p < 0.04) except axial diffusivity in the external capsule (p = 0.09) and in hippocampal fimbriae (p = 0.08). Fractional anisotropy (FA) was lower in the hippocampal fimbriae (p = 0.001) and tended to be lower in the external capsule (p = 0.05) and corpus callosum (p = 0.08, see Figure 6) in the IHH group. At P28 there was no difference in white matter diffusivity of the corpus callosum between IHH and controls, but there was a significant increase in FA (p = 0.03) and decrease in radial (p = 0.008) and mean (p = 0.009) diffusivity from P14 to P28 in the IHH group (Table 2). Litter A had a lower mean FA in all of white matter, corpus callosum and hippocampal fimbria than B and controls (p < 0.02, Figure 7). Furthermore, radial diffusivity in A was higher than that of B and controls in all areas (p < 0.01). Mean diffusivity for litter A was higher than litter B in the internal capsule (p = 0.001) and in all areas compared to controls (p < 0.005). Axial diffusivity in the internal capsule was higher in litter A than controls (p = 0.04). At P28 there were no differences in DTI parameters between litters (data not shown). Grey matter mean diffusivity in cortex and thalamus were higher at P14 in IHH than in controls (Table 1). Litter A had a higher mean diffusivity than controls in all white matter areas and than litter B in thalamus, while litter B was not significantly different from controls in any area. At P28 there were no differences in mean diffusivity in cortex between groups (IHH: 0.00101 ± 0.00003 mm 2 /s vs. controls: 0.00098 ± 0.00002 mm 2 /s, p = 0.1) but litter A (0.00104 ± 0.00005 mm 2 /s) had a higher mean diffusivity than controls (p = 0.04). Litter B (0.00099 ± 0.00004 mm 2 /s) was not significantly different from either litter A or controls.

RotaRod
The IHH group ran for a longer time on the RotaRod than controls at both P20 (p = 0.003) and P27 (p = 0.005, Figure 8).

White Matter
During the first postnatal weeks the white matter of rats undergoes maturational changes, such as increase in axon calibre, axon packing, oligodendrocyte coverage and myelination, which affects the free water diffusion mainly by hindering water diffusion perpendicular to the white matter axons [30]. DTI is an excellent method to study directional free water diffusion and thereby in vivo white matter microstructural development and myelination [31]. In our study we used DTI to evaluate white matter and found higher mean, axial and radial diffusivity together with a lower FA among IHH animals than controls at P14. The high mean diffusivity can be explained by increased water diffusivity in all directions reflected in the concurrent increase in both axial and radial diffusivity. Such an overall increase in water diffusivity would occur in extracellular (vasogenic) edema. However, an increase in extracellular water content would also increase T 2 in white matter, which we did not find. Therefore the higher mean diffusivity possibly reflects a less hindered water diffusion that may indicate a less mature white matter, since mean diffusivity is known to decrease with white matter maturation. This decrease in mean diffusivity during brain development is probably caused by increasing complexity of the white matter tissue structures and the reduction in brain water content that occur with increasing age (reviewed by Hüppi et al. [32]). Similarly, our finding of higher radial diffusivity and low FA in white matter structures of IHH animals may also indicate less mature white matter, since  A normalization of white matter diffusivity was observed from P14 to P28 following IHH exposure, and a similar normalization has earlier been reported following neonatal hypoxia-ischemia [33]. Albeit DTI data were limited at P28, we speculate that alterations in diffusivity after neonatal IHH may be related to white matter maturational delay rather than to permanent injury. Such an interpretation is supported by studies reporting that neonatal intermittent hypoxia delay the maturation of oligodendrocytes rather than injure axons permanently [34] and that neonatal hyperoxic exposure causes a similar maturational delay [35]. Delayed white matter maturation with subsequent hypomyelination is also observed in preterm infants with diffuse white matter injury [36]. However, our DTI findings contrast reports of permanent white matter diffusivity alterations after neonatal hypoxia [37] and hyperoxia [12]. Our profile of IHH had lower hyperoxia (50% versus 80%) and shorter hypoxia (three two-minute episodes every sixth hour versus ten days continuous exposure) than the aforementioned studies. This less severe exposure may have caused transient white matter affection with a preserved ability to catch-up maturation rather than a permanent injury, which may explain the observed normalization of white matter diffusivity after IHH exposure over time.

Grey Matter
Higher grey matter T 2 -relaxation time as well as increased mean diffusivity following IHH indicates increased grey matter water content, and alterations in angiogenesis and BBB-formation may explain the increased vascular density and albumin leakage long-term. The angiogenic effect of hypoxia [38] has been extensively studied in the adult, but not the neonatal brain [39]. On the other hand, hyperoxia cause microvascular degeneration in the neonatal brain [24] possibly inducing tissue hypoxia. By exposing the neonatal brain to a combination of high and low oxygen levels as in IHH, such vasoobliteration by hyperoxia may have potentiated the angiogenic effect of the hypoxic episodes. Moreover, the normalization of vascular density in the IHH group at P28 may possibly be mediated via apoptosis of the newly-formed endothelium [23,40]. However, the concomitant long-term albumin leakage indicates that BBB-integrity was permanently altered even though vessel density returned to normal levels.
The decrease in T 2 -relaxation time in normal brain tissue during maturation of the brain coincides with a reduction in free extracellular water as neurons mature [41,42]. Although the BBB is not fully developed at neonatal age [43], most of the functional barrier properties i.e. glucose and albumin transporters are [29]. Thus the relatively high brain water content and high concentration of albumin in the cerebrospinal fluid at this age [29] may explain the observed albuminimmunoreactivity in the neuropil at postnatal day 14. The suggested presence of a transcellular route for albumin transport [44] is supported by visible albumin in intracellular vesicles in the choroid plexus cells at both P14 and P28 (Figure 3 b & d).

Rotarod
The IHH animals ran longer on the Rotarod than controls, contrary to what would be expected if IHH exposure had given motor function deficits. It is unlikely that this was caused by an improved motor function after IHH exposure. However, during consecutive trials several animals in the control group were able to turn around and apparently willingly jump off the rotating rod, thereby resulting in shorter time spent on the rod. None of the IHH animals showed this behaviour. Neonatal hyperoxia impair memory in rats [11] and although behavioural tests were not performed in the present study, we speculate that differences in memory may explain the observed longer time spent on a Rotarod after IHH exposure.

Retina
It should be noted that classic avascular zones with subsequent prominent neovascularization such as those seen in models for ROP with 24 hours exposure of alternating hyperand hypoxia [4] have not been observed after IHH [27], or after fluctuating oxygen levels around a normoxic mean [6]. Nevertheless, in line with earlier reports [27], the highly clinically relevant profile of neonatal IHH in our study created retinopathy seen as macro-and microscopic haemorrhages.

Implications for Further Research
The strength of our study is the longitudinal design with MRI, immunohistochemistry and neurodevelopmental testing, enabling the comparison of in vivo grey and white matter changes to detailed histopathological characterization and long-term outcome. However, the exploration of brain development after neonatal IHH has raised several questions: (I) Does intrauterine and postnatal growth influence susceptibility to injury from IHH? Rat pups born in larger litters have a lower bodyweight at birth and in the whole suckling period [45], indicating that intrauterine growth determines postnatal body weight gain. Even though no intervention was applied to induce growth restriction in our study, pups from the most numerous litter with the lowest bodyweight and brain volume had the most severe alterations in many parameters of brain development. This finding could have been caused by confounding maturational differences due to age with IHH  Nevertheless, lower body weight may lead to later achievement of developmental milestones [45], even though the higher brain volume/bodyweight ratio in the litter with the lowest bodyweight at P14 indicated that brain volume was relatively spared from a possible growth restriction. Such a maturational delay may have contributed to the differences observed in white matter diffusivity and in T 2 -relaxation time since, as mentioned earlier, changes in these parameters are part of normal brain development. However, the albumin leakage over the BBB was most pronounced in pups with the lowest bodyweight and was a pathological finding that could not have been related to maturation. Therefore, albeit the limited number of litters, this  study may imply that growth restriction, even within the normal range [46], may sensitize the brain to adverse events such as IHH. Indeed, birth weight at term within the normal range correlates with longitudinal brain volume development [47]. Furthermore, being born small-for-gestational-age (SGA) deteriorates neurodevelopmental outcome [48] and delays white matter maturation at term-equivalent age in preterm babies [49]. To explore if an interaction between growth and IHH exists, studies where growth restriction is a controlled condition should be performed.
(II) What is the role of key cellular players in the formation of BBB integrity like pericytes [50] and mediators of angiogenesis like vascular endothelial growth factor (VEGF) [38] in the alterations observed after IHH? Endothelial coverage of pericytes closes the time-window when immature vessels are vulnerable to obliteration by hyperoxia [51], and hypoxia causes the same cells to migrate away from vessels, further destabilizing the endothelium [52]. Furthermore, although angiogenesis in tissue hypoxia is most likely beneficial, its mediator VEGF increases vessel permeability [53]. Future studies should probe whether alterations in pericyte development and VEGF secretion form a causal basis for the observations of increased vascular density, higher T 2 -relaxation time and long-term BBB leakage reported in the present study.

Conclusion
IHH led to subtle changes in brain white matter diffusivity, grey matter water content and vascular density. However, alterations in BBB permeability may indicate permanent longterm effects. The formation of the neurovascular unit may represent a common site of injury in both the retina and the brain. Furthermore, alterations after IHH exposure were more severe in lower-weight animals, and future studies should aim at exploring whether an interaction between inhibited postnatal growth and IHH exists. Figure S1. Regions of Interest in T 2 and FA-maps. (a) T 2map corresponding to -1.5 mm from the bregma with regions of interest (ROI) marked in putamen (blue). (b) T 2 -map corresponding to -3.25 mm from the bregma with ROI in parasagittal cortex (green), hippocampus (red) and thalamus (pink). (c) Colour-encoded directional FA-maps from postnatal day 14 (c) and 28 (d) with ROI in corpus callosum (blue), hippocampal fimbriae (green), internal capsule (red) and external capsule (pink). The ROI in corpus callosum encompassed three slices anterior and three posterior to the one shown. The ROI in hippocampal fimbriae, external and internal capsule encompassed two posterior to the one shown. (EPS)