Cerebral White Matter Injury of the Premature Infant—More Common Than You Think

IMPORTANCE OF CEREBRAL WHITE MATTER INJURY

Brain injury in the premature infant consists of multiple lesions, principally germinal matrix-intraventricular hemorrhage, posthemorrhagic hydrocephalus, and periventricular leukomalacia (PVL). The last of these now appears to be the most important determinant of the neurologic morbidity observed in survivors of birth weight <1500 g. Indeed, of these very low birth weight infants, ∼10% later exhibit cerebral palsy, and ∼50%, cognitive and behavioral deficits.^1–5 The focal necrotic lesions of PVL deep in the cerebral white matter correlate well with the cerebral palsy, whereas the cognitive/behavioral deficits may relate to more diffuse white matter injury observed with PVL (see below). The nature of the relationship between the diffuse white matter injury and the cognitive/behavioral deficits is complex and not entirely understood (see below). The current report of the Hammersmith group by Counsell et al,⁶ published elsewhere in this issue, addresses the frequency and magnetic resonance imaging (MRI) characteristics of this diffuse white matter abnormality.

PREDOMINANCE OF NONCYSTIC CEREBRAL WHITE INJURY

The study of Counsell et al⁶ stimulates further a necessary change in the widely held concept of PVL as principally focal necrotic lesions in the periventricular white matter with subsequent cyst formation. This concept is based on the earlier classic neuropathological description in 1962 of the focal necrotic lesions by Banker and Larroche.⁷ That landmark work led to the appropriate descriptive term for focal softening in the periventricular white matter, ie, PVL. Indeed, innumerable later in vivo clinical and epidemiologic studies of PVL published in the last decade or so have used the ultrasonographic finding of focal periventricular echolucency as the hallmark of PVL (see ref. 1 for review). However, focal necrotic lesions evolving to cysts, readily identified by cranial ultrasonography, are no longer the principal feature of white matter injury in premature infants. Cystic PVL identified by brain imaging is a very uncommon finding in modern neonatal intensive care units. Indeed, in a recent study of 96 consecutive preterm infants by MRI, the findings of cystic PVL were present in only 4.⁸ By contrast, fully 34 had MRI findings of noncystic white matter injury.⁸ A smaller series of 32 infants studied by MRI detected no examples of cystic white matter injury, whereas 79% of those imaged at term had MRI findings of noncystic white matter injury.⁹ Thus, the current concept of PVL must include not only the focal necrotic component deep in periventricular white matter that usually evolves to cyst formation but particularly also a more diffuse noncystic component in central cerebral white matter.

The diffuse cerebral white matter injury has been the subject of increased scrutiny recently. For many years, studies of PVL based on conventional neuropathology have noted central cerebral white matter diffuse astrocytosis, often with abnormal glial cells.^1,10 However, only recently has this diffuse lesion been clarified by more sophisticated anatomic techniques (see below). Thus, perhaps in describing cerebral white matter injury in current small premature infants, use of the term, “PVL,” which accurately describes the focal softening, would be better replaced by the term “cerebral leukoencephalopathy,” to encompass both the focal component and the more diffuse component. Indeed, over 30 years ago Gilles and Murphy used the term “perinatal telencephalic leukoencephalopathy.”¹¹ For this review, however, rather than using a new term, “cerebral leukoencephalopathy,” I will retain PVL as the encompassing term, to include either the focal component or the diffuse component or both.

NEW INSIGHTS FROM MRI/DIFFUSION-WEIGHTED MRI (DWI)

In marked contrast to the paucity of the finding of cystic PVL in premature infants in modern neonatal intensive care units has been the high frequency of the MRI demonstration, especially by the Hammersmith group, of diffuse white matter abnormality.^6,9,12 The abnormality consists principally of MRI signal change, often accompanied by ventricular dilation at term. The frequency of the abnormality on MRI increases as a function of postnatal age. In an earlier report in Pediatrics, the finding of diffuse excessive high signal intensities (DEHSI) on T₂-weighted MRI of premature infants (median gestational age, 27 weeks) increased from 21% in the first postnatal week, to 53% in the next several weeks, to 79% at term equivalent.⁹ This striking result raised the important question of whether this frequent MRI finding represents a biological abnormality, a normal developmental change, or an imaging artifact.

Counsell et al⁶ addressed the question of the nature and significance of DEHSI by using DWI to quantitate at term equivalent the apparent diffusion coefficient (ADC) in cerebral white matter of 50 selected premature infants (median gestational age, 29 weeks).⁶ ADC is a measure of the overall diffusion of water in the tissue and has been shown to decline in normal cerebral white matter of premature infants as they approach term.^13,14 By conventional MRI, the 50 infants at term had evidence for normal white matter (n = 13), DEHSI (n = 23), or “overt white matter lesions” (n = 11). Thus, 34, or 68%, of the infants had MRI evidence for white matter abnormality, and because 8 of the 11 with “overt” lesions also had DEHSI, 31 of the 50, or fully 62% exhibited DEHSI.⁶ Notably, when compared with infants with normal white matter, infants either with DEHSI alone or with “overt white matter lesions” had increased values for ADC in cerebral white matter. Moreover, the values for ADC were similarly increased in infants with DEHSI alone and with “overt white matter lesions.” Thus, the finding of DEHSI by conventional MRI was associated with a quantifiable abnormality in a measure of overall water diffusion, and the abnormal values were similar to the values of more immature cerebral white matter.

KEY QUESTIONS RAISED BY MRI/DWI

What is the implication of the high ADC values in cerebral white matter in the presence of DEHSI at term equivalent? Some clue to the answer relates to the likely mechanism(s) for the normal decline in ADC values in the period from 28 to 40 weeks postconceptional age. This decline occurs during a period in which premyelinating oligodendrocytes, ie, preoligodendrocytes, are abundant in human cerebral white matter.¹⁵ Moreover, these cells are beginning to ensheath cerebral white matter axons¹⁶ in preparation for myelination, which occurs primarily postterm in the human cerebrum.¹⁷ It is likely that these oligodendroglial changes contribute importantly to the decreases in the extracellular space and water content in cerebral white matter and thus the decline in ADC. Measurements of relative anisotropy, ie, an MRI measure of preferred directionality of diffusion, increase during this time interval, consistent with the notion of ensheathment of axons and consequent restriction of diffusion perpendicular but not parallel to the axons.¹⁴ Changes in axon size, axonal membranes, and intracellular axonal constituents likely accompany these premyelination events and could contribute to the decline in ADC and the rise in relative anisotropy.^18–22

In this context, a reasonable hypothesis is that the failure of ADC to decline fully from the higher levels of the small premature infant to the lower levels of the term infant in the presence of DEHSI (or “overt white matter lesions”) relates principally to prior injury or destruction of preoligodendrocytes and subsequent failure of their development and ensheathment of axons. Consistent with this hypothesis is the earlier neuropathologic observation by Gilles of “acutely damaged glia” in cerebral white matter of premature infants with PVL.¹⁰ Could these glial cells be preoligodendrocytes? Also supportive of the hypothesis is the finding by various imaging modalities of hypomyelination and ventricular dilation in follow-up studies of premature infants (see ref. 1 for review). Could these findings relate to the diffuse preoligodendrocyte loss and subsequent failure of myelination? Answers to these key questions required a modern neuroanatomic/neuropathologic approach to the study of the cerebral white matter of human premature infants with and without white matter injury. Such a study has just been completed, as discussed briefly in the next section.

NEW NEUROPATHOLOGICAL INSIGHTS

Using modern immunocytochemical techniques and double labeling to study autopsy brain tissue from 17 infants with PVL and 28 controls, Haynes et al²³ recently made 5 key observations. First, the abnormality in PVL was confined to the cerebral white matter—cerebral cortex appeared to be spared. Thus, the recent finding by volumetric MRI of decreased cerebral cortical gray matter volume at term in premature infants with PVL, some with diffuse injury exclusively, appears not to relate to primary injury to neurons.²⁴ This key abnormality of cerebral cortical development probably is a consequence of the diffuse injury in the cerebral white matter, involving principally preoligodendrocytes, although involvement of axons or subplate neurons or both could contribute. Second, the injury to cerebral white matter was regionalized, with focal necroses, when present, localized in deep periventricular white matter, and less severe, more cell-specific injury present more diffusely in central white matter. These regional characteristics are consistent with, although not proof of, the presence of vascular end-zones/border zones, more marked in periventricular white matter, and less marked more diffusely in central white matter (see ref. 1 for review). Third, in the diffuse injury there was preferential death of preoligodendrocytes, identified by specific immunocytochemical markers and previously shown to be the dominant cell in the oligodendroglial lineage in the cerebral white matter during the period of particular predilection for PVL.¹⁵ This observation identifies the preoligodendrocyte as the key cellular target in the diffuse white matter injury. Death of these cells likely accounts for the subsequent failure of white matter development, identified by MRI/DWI at term by the Hammersmith group,⁶ and of myelination, identified by subsequent conventional imaging by many others (see ref. 1 for review). Fourth, the diffuse white matter injury contained a marked prominence of activated microglia, identified by a specific immunocytochemical marker, as well as astrocytosis. The prominent activated microglia raise strongly the possibility of a role for these cells in the causation of the diffuse injury to preoligodendrocytes (see below). Fifth, specific markers identified striking evidence in preoligodendrocytes for lipid peroxidation and protein nitration in the diffuse white matter injury. This finding suggests that the mode of killing of these cells is attack by reactive oxygen species (ROS) and reactive nitrogen species (RNS). Consistent with these direct observations of brain, a recent study of premature infants found elevated neonatal levels of markers of lipid peroxidation and oxidative protein products in cerebrospinal fluid of those who had PVL documented at term by MRI.²⁵ These several observations have major implications concerning pathogenesis, as discussed in the next section.

IMPLICATIONS FOR PATHOGENESIS

A major role for ROS and RNS in the pathogenesis of white matter injury in premature infants is consistent with the concept that ischemia/reperfusion plays a central pathogenetic role.²⁶ Thus, ischemia/reperfusion is well-known to result in the formation of ROS and RNS which then lead to cell death (see ref. 26 for review). A critical role for ischemia in pathogenesis of both the focal and diffuse components of PVL is suggested by the presence of vascular end zones and border zones in cerebral white matter, the minimal levels of cerebral blood flow to normal cerebral white matter, the relationship of clinical conditions complicated by ischemia (eg, severe hypotension, severe congenital heart disease, extracorporeal membrane oxygenation, severe hypocarbia, impaired cerebrovascular autoregulation) to PVL, and the regionalization of the degree of white matter injury described earlier (see refs. 1 and 26 for review). The development in recent years of more than a dozen models of white matter injury based on diminished cerebral perfusion in immature sheep, rats, mice, rabbits, and dogs further supports the likely role of cerebral ischemia in pathogenesis (see ref. 27 for review.)^27–33

The prominence of activated microglia in the diffuse component of PVL suggests that these cells may be involved in the generation of the ROS and RNS found in the human lesion. Microglia have been shown to be activated by ischemia and to remain activated for weeks following the insult.^29,34,35 Activated microglia have been shown to release ROS and RNS that then result in cell death.^36–38 The abundant reactive astrocytes in the diffuse lesion could contribute to the formation of RNS, because these cells, like microglia, contain inducible nitric oxide synthase. The scenario of release of ROS and RNS by microglia seems likely to result in death of preoligodendrocytes in the diffuse cerebral white matter injury. Recent studies of preoligodendrocytes of the same maturational stage as those that populate the cerebral white matter of the human premature infant¹⁵ show that these cells are exquisitely vulnerable to attack by ROS³⁹ and RNS.^40–42 Interestingly, mature oligodendrocytes capable of myelin production and found primarily postterm in the human infant are resistant to such attack.^39,40 This maturation-dependent vulnerability of preoligodendrocytes to ROS and RNS appears to be related to such factors as deficient antioxidant defenses and acquisition of iron for differentiation.^26,39,40 A recently discovered role for glutamate in the killing of preoligodendrocytes with hypoxia-ischemia may also be mediated by ROS. Thus, it has been shown that non-N-methyl-d-aspartate receptors are present on preoligodendrocytes, result in free radical-mediated death of these cells when activated in vitro, and in vivo lead to death of preoligodendrocytes when activated by hypoxia-ischemia in an immature animal model of diffuse white matter injury.^43–51

The central pathogenetic theme of killing of vulnerable preoligodendrocytes by ROS and RNS, at least in part generated by activated microglia, is consistent with evidence relating maternal intrauterine infection/inflammation and PVL. Thus, a series of clinical, neuropathologic, and experimental studies suggest that at least a subset of cases of PVL is related to maternal intrauterine infection/inflammation (see refs. 26 and 27 for review). Experimental studies support the important involvement of microglia in the white matter injury (see ref. 27 for review.)^52,53 Microglia activated by the molecular products of microorganisms have been shown to release ROS and RNS to lead to cell death.^36,54–57 A recent report shows that the key RNS produced by activation of microglia is peroxynitrite, the deadly radical produced from nitric oxide and superoxide anion.⁵⁷ Activation of microglia in the context of infection is postulated to occur by way of specific cell surface receptors, ie, Toll-like receptors (TLRs), that respond to specific molecular motifs shared by the products of multiple microorganisms.^58,59 Because similar molecular motifs are shared by many microbial products, the relatively small number of specific TLRs are the basis for an immediate response to many different organisms, ie, the mechanism of innate immunity. Relevance of this system to the link between maternal intrauterine infection and PVL are the recent demonstrations that brain microglia contain TLR4, the specific receptor for lipopolysaccharide, the key molecular product of many Gram-negative organisms, and that when activated by lipopolysaccharide, these microglia secrete diffusible products that are highly toxic to preoligodendrocytes.⁶⁰ These products may be, in considerable part, ROS and RNS. Thus, a link appears established between infection and the previously shown vulnerability of preoligodendrocytes, the cellular target in the diffuse component of PVL, to ROS and RNS.

POTENTIAL DELETERIOUS INTERACTION OF ISCHEMIA AND INFECTION

The fact that both ischemia and infection may lead to activation of microglia and generation of ROS and RNS, to which preoligodendrocytes are exquisitely vulnerable, raises the possibility that these 2 insults could potentiate each other. This notion received direct support in recent experiments with a model of hypoxia-ischemia in the developing rat.⁶¹ Thus, in this model a degree of hypoxia-ischemia insufficient to result in overt brain injury caused massive brain injury when the insult was preceded by exposure to lipopolysaccharide. This critical observation raises the possibility that the infant previously exposed to intrauterine infection could be vulnerable to the occurrence of preoligodendrocyte killing by modest ischemic insult(s) not sufficient to cause injury alone. The sometimes modest but frequent declines in cerebral perfusion, detected by near-infrared spectroscopy in premature infants with impaired cerebrovascular autoregulation,⁶² might constitute such insults. Such infants have been shown to be at high risk for the occurrence of white matter injury.⁶² Moreover, it is noteworthy in this context that the neuropathologic and MRI findings of white matter injury in premature infants increase in frequency with increasing duration of survival.^1,9 Although this observation could relate to the cumulative molecular effects of multiple small ischemic insults per se, the possibility now should be considered that at least in some infants prior infection primed the white matter for the subsequent insults.

CONCLUSIONS

The MRI work of the Hammersmith group suggests that diffuse white matter injury with subsequently impaired white matter development is extremely common in small premature infants.⁶ The imaging data may be the structural consequence of the recent neuropathological findings of selective and regionalized involvement of the cerebral white matter of the human infant, the identification of the preoligodendrocyte as the key cellular target, the involvement of activated microglia in the diffuse involvement, and the key role of ROS and RNS in the process of cell death.²³ The MRI and neuropathological observations direct attention to potential interventions for prevention during a time window that may be relatively long. Such interventions could include prevention of modest ischemic insults, especially in the infant with prior exposure to infection, agents to decrease production of ROS/RNS, antioxidants, other compounds to scavenge ROS/RNS, glutamate antagonists, and agents to prevent microglial activation by the products of infection.