Abstract
Purpose.:
To characterize concurrent retinal vessel pathologies reminiscent to retinopathy of prematurity (ROP) in a rat model of periventricular leukomalacia (PVL), in order to identify uniform damage pathways in both organs, the eye and the brain.
Methods.:
Ischemia was induced in Long Evans rat pups on postnatal day 6 (P6) with unilateral (left side) carotid ligation (UCL) followed by exposure to different oxygen concentrations. Four different groups were studied: group A, hypoxia/ischemia (UCL + 6% O2, 1 hour); group B, hyperoxia (80% O2, 24 hours); group C, hypoxia/ischemia + hyperoxia (UCL + 6% O2, 1 hour + 80% O2, 24 hours); and group D, normoxia. In groups A and C, both retinae were examined separately (left retina, group A [A-L], right retina, group A [A-R]; left retina, group C [C-L], right retina, group C [C-R]). Morphologic analysis of vessel development based on flatmounts and cryosections was performed at P11 and P21. Quantitative (q)PCR was performed at P7, P11, and P21 (VEGF-A164, HIF-1α, EpoR, TNFα, iNOS, BMP-9, and IGF-1).
Results.:
On flatmounts, distinct retardation in deeper vascular plexus development was observed, most prominent in A-L and C-L. Retinae of groups A-L and C-L displayed reduced capillary-free zones and an increased number of branching points at P11. Quantitative PCR analysis showed significantly different expression profiles of IGF-1 in A-L and B compared with D over the time course of the experiment.
Conclusions.:
This is the first report on concurring damage to the retina that was evaluated in a rat model of white matter injury in the developing brain. The relatively mild damage to the retinal vessel system may represent the basis for a model of moderate forms of ROP and to study vascular remodeling.
In developed countries, approximately 10% of all infants are born preterm, and growing technical facilities and theoretical knowledge lead to a decrease in mortality even of those with birth weights lower than 1000 g.
1,2
Due to the immaturity of their organ systems, very premature infants are at high risk for serious typical health problems like retinopathy of prematurity (ROP), intraventricular hemorrhage (IVH), periventricular leukomalacia (PVL), bronchopulmonary dysplasia (BPD), and necrotizing enterocolitis (NEC), which often mean a lifelong physical restriction for the children.
1,3,4
Increased oxygen availability after birth and use of supplemental oxygen leading to oxidative stress can promote several of these injuries. Although clinicians are aware of the danger and toxicity of oxygen in this vulnerable group of patients,
3 the optimal oxygenation of very preterm infants is still unknown,
4–6 as recently published data of large randomized multicenter trials are controversial in this regard.
7,8
Infants born with less than 28 weeks of gestation and/or those with birth weights < 1000 g have an increased risk of developing treatment-requiring ROP, which, for example, affects approximately 400 to 600 infants per year in Germany.
9 One pathogenic key factor is a dramatic increase of oxygen partial pressure after birth (95–100 mm Hg) compared with intrauterine conditions, where the oxygen tissue tension is 25 to 35 mm Hg.
10 Several growth factors, such as vascular endothelial growth factor (VEGF) or insulin-like growth factor 1 (IFG-1), which are highly influenced by oxygenation and play important roles in physiological organ and tissue development, are differently expressed and cause a change in the fragile homeostasis around birth.
11,12
Retinopathy of prematurity can be separated into two phases, in which most importantly the VEGF expression levels are different.
6 The first phase is characterized by a stop of retinal vessel growth because of environmental relative hyperoxia and downregulation of VEGF. The incomplete maturation of the vasculature will subsequently cause relative hypoxia in unvascularized peripheral retina and upregulation of VEGF expression above normal levels. This can cause exaggerated vessel growth at the peripheral border of the vasculature, the development of a rim, and subsequent outgrowth of vessels into the vitreous, which may lead to serious visual impairment or blindness through retinal detachment in severe cases.
13 However, the vast majority of ROP cases regress spontaneously and do not require treatment.
A wealth of knowledge about mechanisms of angiogenesis and in particular retinal vessel development has been gained from the classic murine oxygen-induced retinopathy (OIR) model that was developed by Smith and colleagues.
14–19 This model also serves as the standard model for ROP, even though clinical signs in these animals only partly mimic the situation in preterm infants. Mice are kept from P5 to P12 in 75% oxygen atmosphere, followed by normoxic conditions. During and after the hyperoxic situation, vessel growth stops due to relative hyperoxia, and the avascular area in the central retina reaches its maximum at P14. Few days after return to room air, the animals develop peripheral neurovascular tufts additionally to the central avascular regions. The group of Penn and colleagues adopted a different approach by establishing a rat model of OIR by altering the oxygen concentration in the atmosphere in 24-hour cycles between 10% and 50% from birth until P14 followed by room air inhalation from P14 to P18.
20 Peripheral avascular regions appear most prominent at P14 and peripheral hypervascularization at P18.
21–23
The brain disorder PVL is caused by oxygen imbalances and ischemia and evokes motoric, visual,
24 and mental
25 disabilities in later life.
26,27 The disease is characterized by an increased sensitivity of immature oligodendrocytes to hypo- and hyperoxia,
28–30 which results in apoptosis of those developing brain cells. As a consequence, insufficient myelinization of nerve fibers is a major hallmark of the so-called “white matter damage.”
31,32
A model for PVL has existed in rats for more than a decade, and is induced by unilateral ligation of the common carotid artery (UCL) at P6 and exposure to hypoxia thereafter for 1 hour.
31–33 The white matter damage is typically characterized by scoring the presence of myelin basic protein 1 (MBP-1), that is, the presence of oligodendrocytes in the periventricular region.
28–30 Compared with the classic OIR models in mice and rats, the PVL model is induced by ischemia (UCL) and a short period of altered oxygen levels, but these insults are sufficient to induce oligodendrocyte cell death.
Even though both pathologic entities, ROP and PVL, develop regularly in preterm infants and are both related to altered oxygen conditions and hypoxia within the first week of life, any possible correlations between both diseases have been only sparsely examined. It has been shown long ago that very preterm infants with severe ROP are at high risk of also developing brain damage.
34 This observation was confirmed recently in a study including 173 children with severe ROP, who were at higher risk than their peers without ROP to have Bayley scales of mental and psychomotor development of 2 to 3 SD below the expected mean.
35 The authors ascribed this increased risk either to shared risk factors between ROP and brain lesions or, less likely, to anesthetic neurotoxicity associated with ocular examinations. Ng and colleagues
36 reported in 1989 a small series of six patients that suffered from ROP and PVL, in which they postulated that since cerebral and ocular blood supply in premature infants arises both from the internal carotid artery and episodes of hypoperfusion are frequent in premature neonates, cerebral ischemia and retinal ischemia can happen concomitantly, thus causing PVL and ROP simultaneously. Huang and colleagues
37 reported recently that 33 out of 195 infants with ROP of any stage did suffer from PVL (17%), but could not correlate the presence of PVL with the severity of ROP or the necessity of treatment. Yang et al.
38 determined that even 26% of ROP patients in their study also suffered from PVL, which led to restricted visual outcome of these children after 7 years of observation.
In order to shed more light on the correlation of cerebral and retinal vessel abnormalities under ischemic and altering oxygen conditions, we aimed at studying retinal vessel abnormalities in the standard rat model of PVL. Even though the period of altered oxygenation in this model is rather short, we hypothesized that ischemia/hypoxia alone or in combination with transient hyperoxia may provoke vessel changes in the retina. We found significant alterations of the normal development of the retinal vasculature reminiscent of certain features of ROP. The long-term goal is to establish potential biomarkers that are significant for the onset of both retinal vessel abnormalities and PVL symptoms, first in the rat PVL model and subsequently in human preterm infants.
Rat pups underwent surgery at P6. Animals were anesthetized with fentanyl (0.05 μg/10g BW, Fentanyl 0.1 mg; Ratiopharm, Ulm, Germany); medetomidine (1.5 μg/10g BW, Domitor 1 mg/mL; Janssen, Neuss, Germany); and midazolam (20 μg/10g BW, Dormicum V 5 mg/5 mL; Roche, Grenzach-Wyhlen, Germany) mixture. An incision (1 cm) was made on the left side of the animals' necks. The left common carotid artery was separated from the vagus nerve and the jugular vein and ligated using electrocauterization. Antagonization of the drugs was accomplished with naloxone 1.2 μg/10 g BW, (Naloxon Inresa 0.4 mg; Inresa, Freiburg, Germany); atipamezole 7.5 μg/10 g (Antisedan 5 mg/mL; Janssen); and flumazenil 2 μg/10 g BW (Flumazenil Inresa, 0.5 mg/5 mL; Inresa), leading to full recovery after 5 minutes. For pain inhibition, animals received metamizole natrium 1 mg/10 g BW (Novaminsulfon 1 g/2 mL; Ratiopharm).
At P11 and P21, rat pups were killed by perfusion with 4% paraformaldehyde (PFA). Eyes were enucleated and received additional fixation in 4% PFA for 2 hours at room temperature. Neuroretina was isolated and stained with isolectin B4 (Life Technologies, Darmstadt, Germany) for 2 hours.
Fluorescence microscopy was performed with a digital microscope, viewer, and software (Keyence Biozero fluorescence microscope, Keyence Biozero II Viewer, and Biozero II Analyzer Software; Keyence, Neu-Isenburg, Germany). In total, 32 eyes were examined at P11 (×4 A-L, ×4 A-R, ×8 B, ×4 C-L, ×4 C-R, ×8 D) and 28 eyes on P21 (×4 A-L, ×4 A-R, ×8 B, ×3 C-L, ×3 C-R, ×6 D).
Expansion of the deep vascular plexus was measured at P11 and P21 and indicated in relation to total retinal area. Diameters of big retinal arteries were measured at P11 and P21 in central retina and midperiphery (halfway between center and ora serrata). Only arteries with diameters larger than 20 μm were included. Capillary-free zones (CFZ) were determined at P11 along 300 μm length of big retinal arteries in the midperipheral retina and set in relation to the corresponding vessel diameter. We used a modified version of a published method
16 for counting branching points (BPs)/intercapillary junctions. The number of BPs was determined per millimeter squared on vascular front of P11 superficial vascular plexus in at least four fields per retina. One junction was counted where three capillaries met, two junctions were counted where four capillaries met, and so on.
Statistical analysis was performed using the R program for statistical computing
43 (version 3.1.1). Immunohistologic data at P11 and P21 were analyzed using a linear-mixed model that included main effects for—as well as possible interaction between—the hypoxia/ischemia and hyperoxia treatments. A nested factor, indicating which side (left/right) the measurement was taken on, was included for UCL groups due to the stronger effect of UCL on the animal's ligated left side. The mixed model incorporated the intrasubject correlation between measurements taken on the same animal (left/right side) into the analysis as well. As an exception, beta regression for measurements made on closed intervals
44 was used to analyze the data for deep plexus expansion at P21, since these values were close to or reached the maximum of 100% in several groups.
For analyses regarding the changes in artery diameter from P11 to P21 and gene expression from P7 to P11 and P21, a mixed model was used to compare each of the five treatment groups (A-L, A-R, B, C-L, and C-R) to the control group D, while accounting for intrasubject correlation.
Statistical significance was assumed at P < 0.05.
The present study describes for the first time retinal vessel abnormalities in a standard rat model of PVL, which can very likely be used as biomarkers in further studies on the correlation of retinal vascular changes and cerebral damage. The observed abnormalities include a retarded growth of the deep vascular plexus, increased numbers of local vascular pathologies, increased artery diameters, as well as alterations in the size of capillary-free zones and elevated numbers of branching points. It remains to be seen whether one of these markers can be further developed into a clinically relevant tool in premature infants. Regardless of any potential clinical outcome, the standard PVL model in rats represents a valuable model for analyzing the importance of ischemic insults to the vessel maturation in the retina.
The intervention groups can be sorted into two groups depending on the degree of vessel alteration. Moderate vessel modifications were observed in groups A-L and C-L (UCL ipsilateral), and mild vessel modifications in the contralateral retinae A-R and C-R, as well as in group B. The stronger phenotype in the retina ipsilateral to the UCL is likely due to the transitory ischemia, even though complete absence of perfusion is prevented by collateral vessels.
45
The generally milder phenotype in eyes contralateral to UCL was expected. A similar observation was made in brain hemispheres following UCL and correlated with the death of nerve cells in the ipsilateral hemisphere, while cells in the contralateral hemisphere were almost not affected.
46 Retinae from group B showed prominent retardation in vessel maturation that can be directly correlated with the transitory hyperoxia (for significance levels, see
Fig. 6e). Similar experiments to examine the microvasculature of the neonatal rat brain demonstrated that hyperoxia alone has a degenerative effect on brain endothelial cells due to upregulation of NOS-3.
47
We did not observe significant differences between A-L and C-L, which indicates that the major trigger for the manifestation of vessel pathologies/retardation of vessel maturation is indeed the ULC-induced ischemia in combination with subsequent hypoxia. Hyperoxia by itself does not seem to have a large effect following UCL. This would favor the hypothesis that ischemia is a common pathologic feature in both PVL and ROP, as it was postulated by Ng and colleagues.
36 It is, however, not enough to induce dramatic vascular changes observed in classic OIR models.
As stated above, compared with the classic OIR models,
14,23 vessel abnormalities in our ligation model are rather mild. Considering the data from the classic retinal ischemia model, this result is not surprising.
48,49 In the former model, permanent ligation of the pterygopalatine artery (PPA) and external carotid artery (ECA) was performed, which—with respect to retinal ischemia—outlines a more effective method than the ligation of the common carotid artery (CCA). This is due to vessel anastomoses between ECA and internal carotid artery (ICA), and concomitant retrograde blood flow from ECA to ICA in case of CCA ligation. This leads to complete stop of blood flow within the first hours after ligation, and recovery of vascularization occurs partially until day 5 after surgery. The mild to moderate effects on the retinal vascularization of a permanent ligation of the CCA in the PVL model are probably related to the transitory character of the ischemia due to the above-mentioned anastomoses.
In terms of retinal vascularization, the PVL model is a mixture of a retina ischemia model and models with altering oxygen treatment regimens like in the ROP/OIR models. It was originally developed to induce maximum damage to immature oligodendrocytes in the brain. We show for the first time that changes in the retinal vessel system can also be observed, albeit at a slight to moderate level. As stated above, hyperoxia is only a minor contributor to the observed changes, and indeed, oxygen exposition time in the PVL model is much shorter (24 hours) than in the OIR mouse model (Smith,
14 5 days) or the ROP rat model (Penn,
20 14 days). Exposure time and fluctuation of oxygen levels are crucial for the manifestation of vessel pathologies, while constant high and low oxygen concentrations are not. Lower “hyperoxia” concentrations lead to peripheral avascular regions.
23,50,51 In addition, retinal vessels of mice and rats do react differently to altering oxygen exposition. In particular, rats that were exposed to the OIR mouse oxygen regime showed a much milder phenotype.
52
Concerning groups A and C, the deep vascular plexus size at P11 is reduced in the left eye but not the right eye, showing the negative effect of ischemia in combination with hypoxia on the plexus outgrowth (significance at
P < 0.001; see
Fig. 4e). This demonstrates that ischemia itself could play a role in the pathology but may not be sufficient to trigger dramatic vessel alterations.
Diameters of big retinal arteries are enlarged in the central retina at P21 in all groups exposed to ischemia/hypoxia (highly significant [
P < 0.001] for both left and right eyes; see
Fig. 5i), leading to an opposing growth of vessel diameters between P11 and P21 in these groups (increase in diameter) compared with control group D (decrease in vessel diameter). The physiological reduction of the vessel diameter in group D results from the different demand of oxygen in the remodeling retina compared to the initial high demand during plexus outgrowth.
53 In contrast, ischemia/hypoxia seems to prolong the need for high oxygen levels, which may indicate retarded retinal remodeling. Furthermore, the physiological reduction of the vessel diameter between central and midperipheral retina at P11 is attenuated in A-L and C-L and takes place only until P21, again indicating a retarded vessel remodeling.
Based on data concerning the vessel diameters, it seems that ischemia and hypoxia in A-L indeed lead to reduced oxygen diffusion through big retinal arteries in the early stages and, therefore, to reduced CFZ areas (
Fig. 6b). In contrast, the CFZ area and the CFZ/VD quotient are increased in B (
Figs. 6b,
6e), reflecting the high oxygen tissue pressure. This effect is very likely the cause of physiological values for CFZ/VD quotient in group C (ischemia, hypoxia, and hyperoxia), where the hyperoxia neutralizes the hypoxia effect.
The increased number of BP in A-L and C-L at P11 again indicates retardation in vascular remodeling in these retinae because of ischemia/hypoxia (highly significant,
P < 0.001; see
Fig. 6e). Increased numbers of BP at a given time point are a typical sign of delayed vascular remodeling mechanisms, as the maturation status of a vessel system is expressed by the number of BP.
16 Alteration processes in the vasculature lead to apoptosis of endothelial cells and therefore reduction of the number of intercapillary junctions (branching points).
16,41,54
Local vessel pathologies, like neovascular tufts, which were repeatedly observed exclusively in A-L and C-L, are very likely related to the induced manipulations. However, the insult of ischemia and the subsequent hyperoxia might not have been strong enough to produce these signs more often, as it is the case in classic OIR models.
Gene expression analyses were performed at P7, P11, and P21 in order to cover the entire experimental period. All of the seven mRNA targets were chosen because of their known involvement in the regulation of angiogenesis, in the development of ROP, or in inflammatory mechanisms potentially leading to retinal vascular changes.
Even though HIF-1α has been shown to be the master regulator of hypoxia-related gene expression,
55 and was typically found to be upregulated during the proliferative phase in the OIR mouse model,
56 overexpression of this factor was not observed at any of the investigated time points. Interestingly, VEGF mRNA levels dropped immediately in those tissues that were exposed to 24-hour hyperoxia. This, on one hand, demonstrates the influence of hyperoxia in this model but is surprising given the unchanged HIF-1α levels at P7. The reason may be that HIF-1α expression is differently regulated at time points, which we do not cover, or that VEGF expression here is independent of HIF-1α activity, a mechanism proposed to be effective in some forms of cancer.
57
Insulin-like growth factor 1, a known factor to interact with VEGF during normal angiogenesis as well as during the development of ROP,
11,12 was strongly upregulated early (P7) in groups A and B, but not in group C, indicating that the induced ischemia/hypoxia or the isolated hyperoxia, but not the combination of both, triggers IGF-1 expression in this model. Over the time course of the experiment, IGF-1 levels dropped in A-L and B to levels below that of control animals, showing a significantly different evolution compared to control group D. It was shown that low levels of IGF-1 are associated with low levels of VEGF during the initial hyperoxic phase in the development of ROP but increase during the proliferative phase similar to VEGF expression.
6,11,12 This is not the case in our model, where VEGF levels dropped early and increased over time in groups B and C, while they were normal in group A. This dissociation of the expression profiles for the two genes in our model could not be explained so far.
Inflammatory targets TNFα and iNOS have been shown to be dramatically upregulated in the proliferative phase of the murine OIR model,
56 but only iNOS was upregulated at P7 in the PVL rat model. The inducible NOS has even been shown to be connected to increase HIF-1α activity in the OIR model.
58 This correlation was not observed here. While upregulation in the classic OIR model was effective throughout the proliferative period until P21, this was not observed in our study for both inflammatory markers.
Expression of BMP-9, a marker shown to be important during vessel maturation in the retina, was not altered throughout the experiment.
In conclusion, the ischemia, which was induced through a UCL together with a subsequent hypoxia phase or in combination with hyperoxia, is suitable to induce moderate changes/pathologies of the developing retinal vessel system in rats. Since most ROP patients do not develop severe forms of the disease, this model may have a high reference to the clinical situation and can as such very well serve as a disease model for mild to moderate forms of ROP. Furthermore, most alterations can be quantified and therefore be used as biomarkers in studies for the development of clinical diagnostic markers as well as experimental treatment studies.
In addition, since most of the pathologic features go along with typical signs of vascular remodeling, the model may also serve for the research of this physiological mechanism.
Supported by a grant from the Von Behring-Röntgen Foundation (58-0038).
Disclosure: J. Steck, None; C. Blueml, None; S. Kampmann, None; B. Greene, None; R.F. Maier, None; S. Arnhold, None; B. Gerstner, None; K. Stieger, None; B. Lorenz, None