The NVU plays an essential role in progression of numerous CNS pathologies in the adult brain, such as Alzheimer’s disease (AD), multiple sclerosis (MS), stroke, and ischemia [8, 9]. The BBB in newborns is often considered immature, however a growing body of evidence suggests that while cellular function may differ from the adult, there is early establishment of functional and protective properties to meet the demands of the developing brain [38]. Here, we demonstrate clear evidence of NVU-alterations in the newborn FGR brain, with increased presence of endogenous proteins in the perivascular space and discrete-focal extravasation into the parenchyma. We further demonstrated reduced levels of ZO-1 (tight junction) protein and T-cell infiltration into the FGR brain. These alterations are attributed to the elevated pro-inflammatory profile in the FGR brain and the activated glial morphology reported here. Treatment with a common anti-inflammatory, ibuprofen, significantly decreased glial cell activation, ameliorated elevated cytokine levels, and resulted in normalised glial vessel interaction which was associated with reduced BBB-disruption. Our findings suggest that BBB-disruption in FGR may exacerbate early neuroinflammatory responses and therefore early targeting of inflammatory pathways in the clinical setting may provide a therapeutic approach in protecting the FGR newborn from adverse neurological outcomes.
Mechanisms of action of ibuprofen on the NVU
Ibuprofen’s action on the NVU has not previously been examined in neonatal brain disorders. Previous studies investigating the neuroprotective potential of ibuprofen in compromised newborns have largely focused on neurons, white matter and inflammation in the brain parenchyma [13, 30, 39]. Ibuprofen, through inhibition of COX 1 and 2 activity, has various critical functions in the brain, through the reduction in the production of prostaglandins (PGs) and modulation of pro-inflammatory pathways. PGs and pro-inflammatory cytokines TNFα and IL-1β have major actions on blood vessels whereby TNFα and IL-1β affect vascular permeability and increase immune cell infiltration. The dosage of ibuprofen used in the current study significantly reduced the pro-inflammatory profile in the FGR brain as previously reported [13]. This reduction in inflammation in the brain microenvironment following ibuprofen treatment coincided with reduced disruption to the NVU composition and reduced infiltration of immune cells.
In a rat spinal cord injury model, ibuprofen significantly reduced lanthanum (an electron dense tracer) infiltration into endothelial cells and in the basal lamina, but extravasation of lanthanum in vesicles was mainly absent [40]. Ibuprofen also prevented the disturbances in the blood-spinal cord barrier permeability, edema formation, spinal cord blood flow changes, and cell reaction. In contrast, ibuprofen treatment amplified rather than decreased microvascular damage associated with seizures [41]. In this rat seizure study, ibuprofen administration increased the expression of TNFα and IL-1β in microvessels with aggravation of edema and microbleeds; however cerebral tissue was protected from inflammation following treatment [41].
FGR newborn pig display altered vascular integrity and glial morphology
The neurovascular unit (NVU) is composed of neurons, glial cells (astrocytes, microglia, and oligodendroglia), vascular endothelial cells, and the basement membrane. The NVU plays a critical role in regulating cerebral homeostasis through maintenance of BBB integrity and cerebral blood flow [9, 14]. Studies of induced-FGR in the preterm and term lamb, through single umbilical artery ligation, indicate altered BBB-permeability which may be associated with brain pathology [17, 42]. Here we employed the use of MRI with gadolinium contrast agent to ascertain whether BBB-disruption could be identified in vivo in naturally occurring FGR piglets using a common clinical approach [43]. Studies in neurological disorders such as MS, which is characterised by inflammatory lesions of the white matter, have reported the formation of new lesions preceding BBB-disruption as detected by gadolinium enhancement [44]. However very few studies examine gadolinium for BBB-disruption in newborns. A neonatal stroke model showed minimal gadolinium enhancement 24h after reperfusion, even with microscopic evidence of BBB-disruption [45]. We detected in vivo alterations in MRI contrast enhancement in the FGR brain supporting the postmortem microscopic investigations of changes in vascular integrity. The T1 signal was increased (indicative of tissue gadolinium), while T2 was also increased (vascular gadolinium) even though vascular density was reduced. We observed a 30% decrease in collagen IV, the most abundant basement membrane protein, which provides support and adhesion for basal vascular endothelial cells (CD34) [46, 47] for which we report a 40% decrease in the parietal cortex of the FGR brain. The primary cells involved in production of collagen IV are endothelial cells, astrocytes, and pericytes[48–51]. Thus, a loss or dysfunction of these cell types (discussed below) may contribute to loss of collagen IV. We also observed truncation and discontinuity of labelling of basement membrane and endothelial cells in the FGR brains with a significant reduction in proliferating cells. Studies by Castillo-Mendez et al.,[17, 18] report a reduction in vascular density in the white matter of FGR lambs using the non-collagenous glycoprotein laminin as a marker of the basement membrane. Similarly, they also demonstrate a decrease in proliferating blood vessels in the brain at 24h postnatal age in the FGR lamb [18], supporting vascular disruption in the FGR brain.
Inflammatory state alters glial morphology and interactions with brain microvessels in FGR brain
Glial cell changes have been extensively characterised following brain trauma. Changes in morphology and phenotype result in an altered profile of surface markers and expression of an array of inflammatory mediators [12, 13, 52, 53]. In FGR brain, we have shown that juxtavascular microglia and astrocytes demonstrate classical thickening and retraction of processes, and increased expression of pro-inflammatory mediators. Key upstream inflammatory initiators NF-κΒ, IL-1β and TNFα were expressed in juxtavascular microglia and astrocytes that interact with the NVU of FGR brains. These factors exhibit a multitude of innate and adaptive immune functions which are critical mediators in brain pathology and brain development [54–56]. Excess production of these inflammatory cytokines in both acute and chronic environments is associated with disease pathogenesis and brain dysfunction [57].
Microglia display dynamic motility behaviours following tissue damage, and preferentially associate with microvessels [58, 59]. Juxtavascular microglia have reported to be key players in BBB disruption following stroke [15], where inhibition of microglial activation after stroke promotes BBB integrity and viability [60]. A recent induced-systemic inflammation mouse study reported microglia have dual effects on BBB integrity [16]. Microglia have been shown to respond to inflammation by migrating towards and accumulating around cerebral blood vessels before any detectable changes in BBB permeability are evident. The initial juxtavascular microglial contact with endothelial cells protected the integrity of the BBB; however, as inflammation progressed, a more activated microglial phenotype dominated, resulting in phagocytosis of astrocytic end-feet and impairment of the BBB [16]. A mouse stroke study suggests phagocytic microglia may disrupt the BBB by engulfing endothelial cells [15]. Microglia were activated soon after reperfusion with expansion of their cellular processes towards blood vessels. All activated microglia in the damaged region were associated with, and engulfed cerebral blood vessels within 24h and concurrently broke down the BBB [15]. We observed an increase in activated juxtavascular microglia, hypertrophic microglia appeared to engulf some blood vessels in the FGR brain. The activated microglia were also in close contact with astrocytic end-feet and therefore may be acting in a phagocytic manner towards these and other cell types.
Utilising 3D reconstruction of electron microscopic images, Mathiisen et al., (2010) demonstrated almost complete ensheathment of brain microvessels by astrocytic end-feet under non-pathological conditions [61]. The interaction of astrocyte end-feet and brain microvessels is a key component in maintaining BBB-integrity, with studies demonstrating that under pathological conditions end-feet retract from vessels, exacerbating injury and altering BBB-permeability [62–64]. In the FGR brain, we report activation of juxtavascular astrocytes was associated with decreased end-feet ensheathment of the vasculature and may be a key mediator of BBB permeability. FGR sheep studies concur with this alteration to BBB, reporting a decrease in astrocyte end-feet coverage of blood vessels within white matter regions 24h after preterm and term birth [17, 42]. Further, in our current study, the astrocytes in contact with blood vessels in the FGR brain displayed enlarged end-feet indicative of change in functionality. These astrocytes may be no longer functioning as a support cell, rather are in a phenotypic reactive state with the ability to release inflammatory cytokines.
Activation of microglia and astrocytes results in the initiation of numerous inflammatory pathways in the order to resolve tissue damage. However, sustained release of cytokines and chemokines can exacerbate the inflammatory cycle contributing to NVU instability. The pro-inflammatory cytokines IL-1β and TNFα are well established as regulators of neuroinflammation but there is evidence to support them as key inducers of BBB dysfunction [65]. Data from several studies suggest microglial derived TNFα and IL-1β play a major role in BBB disruption [66–68]. Microglia activated in response to epileptic seizures release IL-1β which has been shown to downregulate the tight junction protein Cldn5 in endothelial cells resulting in disruption to BBB function [66]. TNFα binds to endothelial cells, leading to a downregulation of tight junctions and subsequent increase in BBB permeability in many central and peripheral inflammatory conditions [69, 70]. Microglial secreted TNFα has also been recognised as a key cytokine leading to endothelial cell programmed necrosis after stroke [68] [71]. These findings support our observations of TNFα and IL-1β expression in juxtavascular microglia in FGR brains and corresponding alterations to BBB permeability. CXCL10 is expressed in multiple CNS pathologies and strongly linked with alterations in BBB-permeability, although the mechanism through which it acts is unclear [72–74]. Further, in the current study, juxtavascular astrocytes showed strong TNFα expression in FGR brains. We also observed robust labelling of NF-κΒ in both astrocytes and microglia in the FGR newborn brain while only moderate labelling was noted in the other groups investigated. This prototypical pro-inflammatory pathway is likely triggered by the increased expression of IL-1β and TNFα [75, 76]. NF-κB may be essential in promoting microglial activation and TNFα release, which mediates endothelial programmed necrosis and accelerates BBB disruption after stroke injury[68]. Activation of this pathway may exacerbate and prolong the pro-inflammatory profile in the FGR brain with studies demonstrating inhibition of NF-κΒ signaling is associated with reduced inflammation and brain injury in rodent models of neuropathology [77–79].
Administration of ibuprofen significantly reduces parenchymal glial activation in FGR newborns [13]. In our current study, ibuprofen treated FGR animals showed a reduction in the number of activated juxtavascular microglia as well as parenchymal microglia with a return to the glial morphology observed in normally grown animals. Juxtavascular astrocytes in the treated FGR brains also returned to a resting morphology that resulted in more consistent end-feet coverage of microvessels. These findings coincided with a reduction in several inflammatory cytokines. This positive effect on the glial cells of the NVU in the FGR brain has also been shown with other anti-inflammatory treatments. Maternal melatonin treatment for the FGR fetal sheep resulted in improved astrocytic end-feet coverage of blood vessels and reduced BBB disruption [18]. It is therefore plausible that reducing the ongoing proinflammatory cycle in the FGR brain could allow the brain to function in a healthy way whereby the astrocytes and microglia provide protection and support to the BBB rather than disruption to the BBB.
The use of endogenous proteins (IgG and albumin) are well-established for detection of altered BBB-permeability [29, 80, 81]. A clear limitation of this approach is that they cannot be used to determine progression or duration of the ‘leak’ as these proteins readily diffuse once in the brain parenchyma [82]. The increase in perivascular albumin labelling we demonstrate in FGR animals may explain the observed increase in T2 vascular gadolinium. This MRI technique is unable to distinguish between vascular and perivascular concentrations of gadolinium, therefore the T2 concentrations we report may largely be confined to the perivascular space. Regions displaying BBB-disruption were associated with glial alterations as described above. We also observed neuronal and glial uptake of endogenous proteins in the FGR brain. In models of hypoxia-ischemia, neuronal uptake of IgG was associated with neuronal degeneration and reported to result in perturbed neuronal function and survival, and therefore may contribute to the progression of neuropathology [29, 83].
In the normally grown and ibuprofen treated FGR piglets, a reduction in the presence of perivascular albumin was observed, along with infrequent focal IgG extravasation in the cortex and white matter with strong astrocytic end-feet interaction around vessels. Further, the astrocytic processes appear to form a boundary around the focal leaks, suggesting containment of the leak or support of the vessels. This may be due to the reduced pro-inflammatory state of the brain whereby the astrocytes are able to function as a supportive cell rather than as a reactive phenotype. In agreement with our findings, melatonin administration to the FGR lamb reduced albumin extravasation in the brain and conferred protection on the BBB [18]. Umbilical cord blood-derived stem cell treatment to the FGR lamb also resulted in decreased albumin extravasation in brain when assessed at 24h postnatal age, with a concomitant reduction in brain injury [84]. Whether this protection to the BBB is directly via an anti-inflammatory effect of treatment is unknown. However, mouse studies suggest this may indeed be plausible [16, 60]. Minocycline treatment in a mouse model of systemic inflammation maintained BBB integrity by directly inhibiting microglial activation [16]. Further, inhibition of microglia in a mouse stroke functional microglia knockout model not only reduced extravasation of MRI contrast agent in the brain parenchyma but also resulted in a reduction in lesion size [15].
BBB-permeability is associated with infiltration of peripheral immune cells
To our knowledge, this is the first report of T-cell infiltration into the newborn FGR brain. T-cell infiltration has been reported to promote neuroinflammation and cognitive decline in animal and human neuropathological studies [85, 86]. Mild suppression of immune responses in the FGR pig at postnatal day 24 have been demonstrated with a decrease in peripheral lymphocytes, more specifically CD3+CD4+ T-cells [87]. Combined with our findings of early CD3+ cell infiltration into the brain parenchyma, FGR newborns likely present a compromised immune response. Characterisation of changes in both peripheral and CNS immune cell populations is necessary to determine the contribution of immune dysfunction to the altered brain pathology we and others have reported in FGR newborns. We have not yet identified whether these T-cells belong to cytotoxic (CD8+) or T-helper (CD4+) populations, which would enable a better understanding of whether the CD3+ cell infiltration is a rescue phenotype or a population that contributes to neuropathology in the FGR brain.
The pro-inflammatory profile observed in FGR brain included expression of a number of chemokines and cytokines involved with chemotaxis of myeloid and lymphoid cells in juxtavascular glia. Studies have reported juxtavascular astrogliosis coincides with alterations in BBB integrity, and suggest that BBB disruption precedes overt CNS infiltration by immune cells [88]. These studies propose that immune cell infiltration may therefore be a direct consequence of glial activation [88]. There is evidence of activated juxtavascular microglia accumulating in sites of lymphocyte infiltration in a bacterial endotoxin lipopolysaccharide postnatal rat model [89]. Activated microglia are known to release chemoattractants for lymphocytes in addition to pro-inflammatory mediators that stimulate astrocyte activation. We observed elevated expression of CCL2 (MCP-1) which is known to contribute to site-specific infiltration of lymphocytes into the brain [90]. These findings demonstrate the interplay between glial activation, NVU dysfunction and the drive to promote recruitment of peripheral infiltrates, which if left unchecked may result in sustained and deleterious effects on the newborn FGR brain.
We observed labelling of the tight junction protein Cldn-1 in the FGR brain, with less labelling detected in the cortex and white matter in the normal brain [91]. Claudin-1 expression was co-localised with GFAP-positive astrocytes throughout the parenchyma as well as in juxtavascular astrocytes. The pro-inflammatory cytokine IL-1β has been shown to induce astrocytic expression of Cldn-1 which acts to corral activated T-cells in the brain [92]. Activated T-cells in turn release metalloproteinases (MMPs) 3, 7, and 9, to degrade Cldn-1 protein. The authors propose a subtle ongoing struggle between activated astrocytes and T-cells attempting infiltration of the brain parenchyma. The expression of Cldn-1 in astrocytes may therefore be a compensatory mechanism to provide a ‘secondary’ barrier to infiltration of peripheral immune cells into the CNS. By dampening the pro-inflammatory environment, with ibuprofen, we propose that glial cells return to their resting morphological state, allowing normal function and reduced energy demands. This normalised glial interaction at the NVU provides a ‘tighter’ more functional BBB, which minimises leak of serum proteins and peripheral cell infiltration into the brain. Together these ameliorated changes provide the newborn brain a healthier environment for early postnatal brain development, which in turn is likely to improve long-term neurodevelopmental outcomes.
Ibuprofen treatment reduces apoptosis in the brain parenchyma and at the NVU
There was an evident increase in the number of Casp3+/GFAP+ juxtavascular astrocytes in the FGR brain. A study in human AD patients showed the presence of degenerating astrocytes as labelled with an antibody specific to GFAP-caspase cleavage product [93]. These degenerating astrocytes co-localised with caspase-3 antibodies and were found in proximity to the neurovasculature. Further, a significantly higher number of caspase-3 positive cells are observed on GLUT-1 positive endothelial cells in FGR lambs with a concurrent decrease in the astrocytic end-feet coverage of blood vessels[18]. Together with our observations, these findings suggest activation of caspase-3 pathways and cleavage of GFAP (cytoskeletal proteins) may contribute to decreased astrocyte end-feet interaction with brain microvessels and subsequently result in BBB-permeability. Following ibuprofen treatment, we observed a significant decrease in Casp3-positive cells in the FGR brain. Ibuprofen has been shown to have a direct effect on apoptosis [94]; therefore the positive impact of ibuprofen treatment on the NVU may not only be due to its anti-inflammatory effect, but also its anti-apoptotic actions. A similar effect has been shown following melatonin treatment in the FGR lamb. Maternal melatonin treatment significantly reduced the number of caspase-3 positive cells associated with blood vessels in the white matter of FGR lamb brains [18]. It is plausible that these treatments provide neuroprotection through multiple actions that target apoptosis and inflammation in the FGR brain.
Decreased expression of tight junction protein ZO-1 in FGR brain
Tight junction proteins are key regulators of BBB-integrity, with numerous studies demonstrating that loss or mutations in TJs are associated with BBB-disruption. Depletion of Cldn-5 induces BBB-disruption in knockout mice [95] and in cultured human brain endothelial cells[96], while OCLN knockout resulted in growth retardation but no significant alterations in BBB-function [97]. Our investigations of these key TJ proteins found no change in Cldn-5 and OCLN, however the cytosolic TJ protein, ZO-1 was significantly decreased in FGR brain. The absence of disruption to Cldn-5 and OCLN protein levels may explain why overt BBB-disruption was not observed in FGR brains [98, 99]. However, in neonatal HI piglet with extensive brain injury and BBB disruption (IgG extravasation), no changes were observed at the mRNA level of these TJs, and only loss of ZO1 and OCLN protein in the parietal cortex of HI animals with seizures [29].
While expression of TJs were generally maintained in the FGR brain, interactions between these key regulatory proteins may be affected. Under pro-inflammatory endothelial conditions, Cldn-1 is expressed at the endothelial cells and reported to disrupt the interaction between Cldn-5 and ZO-1 which in turn reduces BBB-integrity [100]. While we report relatively normal expression levels of TJ proteins in FGR, it is plausible that Cldn-1 expression at the microvasculature may contribute to altered TJ protein interactions and subsequently result in reduced BBB-integrity.
It is also possible the TJ proteins may have undergone redistribution. Models of stroke have demonstrated movement of ZO-1 from the membrane to cytoplasm during reperfusion [101]. ZO-1 loss and TJ redistribution in traumatic brain injury has also been shown to be associated with IL-1β [102]. As we have demonstrated a significant increase in IL-1β, but also a loss in ZO-1, redistribution of TJs in the FGR brain cannot be discounted. The present study used whole protein preparations and did not examine the subcellular fractions of brain tissue; further investigation is required to ascertain whether TJ proteins are significantly redistributed in FGR brains. Studies demonstrating significant alterations in TJ protein expression generally show overt BBB-disruption [101]. Given the mild BBB-disruption reported in the present study and relatively normal expression of total protein for each TJ, our findings would indicate they are not the key mediator involved in the disruption observed. These findings indicate only subtle structural alterations to the vasculature and TJs of FGR brain suggestive that the BBB-permeability observed is likely due to altered function or interaction rather than loss of NVU components.