Chapter 8. Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis

348

TABLE 8-1



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



Major Neuropathological Varieties of

Neonatal Hypoxic-Ischemic

Encephalopathy



Selective neuronal necrosis

Parasagittal cerebral injury

Periventricular leukomalacia

Focal (and multifocal) ischemic brain necrosis, stroke



necrosis) is not a prominent feature of hypoxic-ischemic

encephalopathy in the human newborn.14-18 In

one study, clearly increased intracranial pressure

(i.e., >10 mm Hg) was observed in only 22% of

32 asphyxiated term newborns, it did not compromise

cerebral perfusion pressure, it reached a maximum at 36

to 72 hours, and it correlated with computed tomography evidence for early brain necrosis.17 In a second systematic study, intracranial pressure reached a maximum

at a mean age of 29 hours and was not correlated with

clinical or electroencephalographic evidence for neurological deterioration.16 Changes in cerebral perfusion

pressure most often reflected decreases in arterial

blood pressure rather than increases in intracranial pressure.16 Moreover, administration of mannitol in a single

dose to asphyxiated infants in a controlled study had no

beneficial clinical effect.18 These observations support

the conclusion that early brain edema is a consequence of,

rather than a causative factor in, hypoxic-ischemic brain

injury (Table 8-2).14-18

Experimental Aspects in Perinatal Animals

Intrauterine Partial Asphyxia in the Fetal Monkey.

The notion that brain swelling is an important early feature with perinatal hypoxic-ischemic insults and the

cause of subsequent tissue necrosis is based on studies

with term fetal monkeys by Myers and co-workers.19-25

In these experiments, in association with ‘‘prolonged

partial asphyxia’’ of the term fetal monkey (produced

by a variety of procedures that impair placental gas

exchange, e.g., maternal hypotension, maternal hypoxemia, and umbilical cord compression), a pattern of

cerebral injury characterized by necrosis and edema

was observed. The topography of the necrosis was typical of the parasagittal cerebral injury observed in the

asphyxiated human term infant (see later discussion).



TABLE 8-2



Evidence Against Early Brain Edema

as a Causative Factor in HypoxicIschemic Brain Injury*



Intracranial pressure higher than 10 mm Hg is uncommon in

asphyxiated term infants.

When intracranial pressure greater than 10 mm Hg occurs,

timing is relatively late (i.e., 24 to 72 hours).

Marked decreases in cerebral perfusion pressure are

uncommon, and decreases in cerebral perfusion pressure that do occur are usually caused by decreases in

blood pressure rather than by increases in intracranial

pressure.

*See text for references.



Associated with the cerebral injury in the monkeys

was brain swelling, defined primarily by gyral flattening.

The edema was considered to be intracellular on the

basis of electron microscopic observations in related

experiments.25 In similar experiments, statistically significant changes in brain water content could not be

demonstrated.24 On balance, the brain swelling in

these experiments appears most likely to be secondary

to the pronounced tissue injury (with cytotoxic edema),

rather than a primary event leading to the injury. This

possibility would be compatible with conclusions

derived from human pathological material (see earlier

discussion).

Intrauterine Partial Asphyxia in the Fetal Lamb.

Studies of the fetal lamb subjected to intrauterine partial asphyxia do not support the notion of brain edema,

at least of the vasogenic variety, as an important consequence of acute hypoxic-ischemic brain injury in this

model.26,27 Extravascular plasma volume was quantitated by the iodine-125–labeled albumin method with

asphyxia and was found not to be significantly

increased in cerebrum, brain stem, or cerebellum.26

Moreover, the postasphyxial delayed cerebral hypoperfusion observed in this model occurred in the absence

of brain edema.27 In later studies of near-term fetal

lambs subjected to hypoxia-ischemia, findings indicative of cytotoxic edema correlated with documented neuronal injury were obtained, a correlation consistent

with the observations in human infants (see earlier).28

Hypoxic-Ischemic Insult in the Neonatal Rat and

Piglet. Careful morphological, physiological, and biochemical studies of the neonatal rat and piglet subjected to a combination of ischemia (carotid ligation)

and hypoxemia also fail to support the notion of brain

edema as a primary or injury-causing result of hypoxicischemic insult.29-38 In the studies of neonatal rats,

although the water content of brain increased, a close

correlation was defined between the degree of tissue

necrosis and the increase in brain water. No sign of

transtentorial or cerebellar herniation was observed,

unlike in adult animals similarly studied. No inverse correlation of cerebral blood flow (CBF) and brain water

content could be identified over the 6 days following

the hypoxic-ischemic insult. Moreover, administration

of four doses of mannitol over 2 days following the

insult did not ameliorate the incidence, distribution,

or severity of the extensive tissue injury, despite reduction in the increase in brain water content in the

hypoxic-ischemic hemisphere.30 Additionally, the spatial relationships between this increase in brain volume

and the tissue injury did not suggest that the apparent

edema caused or contributed to the cerebral injury.33

The conclusion is that the brain ‘‘edema’’ is a ‘‘consequence rather than a cause of major ischemic damage

in the immature animal.’’29-31,33-35,37

Selective Neuronal Necrosis

Selective neuronal necrosis is the most common variety

of injury observed in neonatal hypoxic-ischemic



Chapter 8



Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis



encephalopathy. The term refers to necrosis of neurons

in a characteristic, although often widespread, distribution. Neuronal necrosis often coexists with other distinctive manifestations of neonatal hypoxic-ischemic

encephalopathy (see later sections), and in fact it is

very unusual to observe one of the other varieties of

neonatal hypoxic-ischemic encephalopathy without

some degree of selective neuronal injury as well. The

topography of the neuronal injury depends in considerable part on the severity and temporal characteristics

of the insult and on the gestational age of the infant.

Three basic patterns derived primarily from correlative

clinical and brain imaging findings, and observed best

in term infants, can be distinguished (Table 8-3).

Diffuse neuronal injury occurs with very severe and

very prolonged insults in both term and premature

infants. A cerebral cortical–deep nuclear neuronal predominance occurs in primarily term infants with moderate

to severe, relatively prolonged insults. The deep

nuclear involvement includes basal ganglia (especially

putamen) and thalamus. Deep nuclear–brain stem neuronal predominance occurs in primarily term infants with

severe, relatively abrupt insults. Two other patterns,

pontosubicular neuronal injury and cerebellar injury,

occur particularly in premature infants with a still-tobe-defined temporal pattern of insult (see later discussion), but these patterns are usually accompanied by

other features of selective neuronal injury and are discussed in this overall context. Thus, overall five patterns are described. In the discussion that follows,

I review the cellular aspects of selective neuronal

injury, the regions of predilection, and the current concepts of pathogenesis.

Cellular Aspects

As the name implies, the neuron is the primary site of

injury.4-6,8 Experimental studies indicate that the

first observable change in the neuron is cytoplasmic

vacuolation, caused by mitochondrial swelling,

occurring within 5 to 30 minutes after the onset of

hypoxia.39-42 In contrast to the rapid onset of neuronal

changes in tissue cultures of neonatal mouse cerebellum exposed to hypoxia, no structural alteration was



TABLE 8-3



349



observed in astrocytes.42 However, as discussed later,

studies of a variety of developing models suggest that

differentiating oligodendrocytes exhibit approximately

the same sensitivity to glucose and oxygen deprivation

as do neurons. On balance, the data suggest that in

the immature and mature brain, the order of vulnerability is neuron ! oligodendroglia > astrocyte > microglia. In the context of the present discussion, the

neuron is the cellular element most vulnerable to

hypoxia-ischemia.

The temporal features of neuronal and related changes

in neonatal human brain have been well documented.38,43,44 The major changes seen by classic light microscopy occur after 24 to 36 hours and are characterized

by marked eosinophilia of neuronal cytoplasm, loss of

Nissl substance (endoplasmic reticulum), condensation

(pyknosis) or fragmentation (karyorrhexis) of nuclei,

and breakdown of nuclear and plasma membranes,

often with observable cell swelling (Fig. 8-1). Two factors alter the ability to identify such neuronal changes

early after perinatal asphyxia: (1) the gestational age of

the infant and (2) the nature of the survival period.

Thus, recognition of neuronal changes in premature

infants is difficult because of the close packing of

immature cortical neurons and their relative lack of

Nissl substance. Moreover, the brain of any infant

who has been maintained on a respirator for several

days, with compromised ventilation or perfusion, may

have undergone enough autolysis to obscure early cellular changes. When these factors are not taken into

account, the presence and magnitude of neuronal

injury may be misjudged and may lead to spurious conclusions about the nature of the neuropathology.

The early neuronal changes are followed in several

days by overt signs of cell necrosis (Fig. 8-2). Associated

with this cell necrosis is the appearance of microglial

cells and, by 3 to 5 days after the insult, hypertrophic

astrocytes. Foamy macrophages consume the necrotic



Major Patterns of Selective Neuronal

Injury and Characteristics of Usual

Insult in Term Newborns



Pattern*



Usual Insult



Diffuse

Cerebral cortex–

deep nuclear{

Deep nuclear{–

brain stem



Very severe, very prolonged

Moderate to severe, prolonged

Severe, abrupt



*The patterns reflect areas of predominant neuronal injury; considerable overlap is common. Two additional patterns of selective neuronal necrosis (i.e., pontosubicular and cerebellar), which occur

predominantly in premature newborns (see text), are not listed

here because the temporal characteristics of the insult are

unknown.

{

Deep nuclear: basal ganglia (especially putamen) and thalamus.



Figure 8-1 Ischemic neuronal injury (arrowheads) within the hippocampus of an asphyxiated term infant. Note shrinkage of cytoplasm

and pyknotic nuclei with irregular nuclear membranes, loss of nucleoli,

and nuclear hyperchromasia. (Hematoxylin and eosin stain, Â500.) In

typical hematoxylin and eosin sections, the shrunken neuron is eosinophilic, and thus this appearance is the classic ‘‘red, dead’’ neuron of

ischemic injury. (From Clancy RR, Sladky JT, Rorke LB: Hypoxicischemic spinal cord injury following perinatal asphyxia, Ann Neurol

25:185-189, 1989.)



350



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



Figure 8-2 Selective neuronal necrosis. Note the ‘‘encrusted’’ necrotic neuron in the center of the figure. Most of the other cells in this

brain stem nucleus are reactive astrocytes and microglial cells.

(Courtesy of Dr. Margaret G. Norman.)



debris, and a glial mat forms over the next several

weeks. Severe lesions may result in cavity formation,

especially in the cerebral cortex.3,5-8

Apoptotic as well as necrotic cell death is observed in

hypoxic-ischemic disease in human infants, as in neonatal animal models.8,43,45-52 In one study of neuronal

injury after ‘‘birth asphyxia,’’ the mean fractions of apoptotic and necrotic cells in cerebral cortex were 8.3%

and 20.8%, respectively.43 In a study of the neonatal

piglet subjected to hypoxia-ischemia, apoptotic neuronal death predominated among immature neurons and

necrotic cell death predominated among mature neurons.45 A similar susceptibility of immature neurons to

apoptosis has been shown in N-methyl-D-aspartate

(NMDA)–treated neurons in culture.50 In one specific

form of human neonatal injury, pontosubicular necrosis (see later), the predominant form of cell death

appears to be apoptosis.47,53

Regional Aspects

As noted earlier, three major regional patterns of selective neuronal necrosis can be delineated in the human

newborn, especially the term infant (see Table 8-3). In

diffuse disease, certain neurons at essentially all levels of

the neuraxis are affected. In predominantly cerebral–deep

nuclear disease, the prominent involvement is of cerebral neocortex, hippocampus, and basal ganglia-thalamus. In predominantly deep nuclear–brain stem disease,

basal ganglia–thalamus–brain stem is the topography. A

fourth pattern, more commonly observed in the preterm infant, pontosubicular necrosis, is characterized by

involvement of neurons of the base of the pons and

the subiculum of the hippocampus (see later). A fifth

pattern, observed particularly in the small premature

infant but to a different degree in the term infant,

involves the cerebellum (see later). Given that overlap

among these groups is the rule rather than the exception,

I discuss diffuse disease first, because all the vulnerable

groups are involved.



Diffuse Neuronal Injury. The major sites of predilection for diffuse neuronal necrosis in the term

and preterm newborn infant are shown in

Table 8-4.2-6,8,43,47,54-75

Cerebral Cortex. Neurons of the cerebral cortex in

the term infant are particularly vulnerable, most notably the hippocampus (pyramidal cells) among the cerebral cortical regions. Sommer’s sector (and contiguous

areas) in the term newborn and the subiculum of the

hippocampus in the premature newborn (see later discussion) are especially prone to injury (see Table 8-4).

With more severe injury in the term infant, the better

differentiated neurons of the calcarine (visual) cortex

and of the precentral and postcentral cortices (i.e., perirolandic cortex) may be injured. In very severe injury,

diffuse involvement of cerebral cortex occurs. Neurons

in deeper cortical layers and, particularly, in the depths

of sulci are especially affected. A role for patterns of

blood flow in the determination of the topography is

apparent from the more severe neuronal injury consistently observed in border zones between the major

cerebral arteries, especially in the posterior cerebrum,

and in depths of sulci. Perhaps reflecting the relative

immaturity of cerebral cortical neurons in premature

infants, involvement of cerebral cortex is uncommon,75

particularly in comparison with neurons of deep

nuclear structures and brain stem (see later).

However, sophisticated brain imaging studies of premature infants at term equivalent age and later in childhood show

impressive abnormalities of cerebral cortex (see Chapter 9).

Thus, diminutions of cerebral cortical volumes



TABLE 8-4



Sites of Predilection for the Diffuse

Form of Hypoxic-Ischemic Selective

Neuronal Injury in Premature and

Term Newborns*



Brain Region

Cerebral neocortex

Hippocampus

Sommer’s sector

Subiculum

Deep nuclear

structures

Caudate-putamen

Globus pallidus

Thalamus

Brain stem

Cranial nerve nuclei

Pons (ventral)

Inferior olivary nuclei

Cerebellum

Purkinje cells

Granule cells (internal,

external)

Spinal cord

Anterior horn cells (alone)

Anterior horn cells and contiguous cells (? infarction)

+, common; ±, less common.

*See text for references.



Premature



Term

Newborn

+

+



+

+

+

+



+

+

+



+

+

+



+

+

+



±



+

±

±



±



Chapter 8



Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis



and gyral development have been documented. The

disturbances may reflect abnormalities of cerebral cortical development and may be related to concomitant

cerebral white matter injury (see later). The important

point in this context is that the abnormalities of cerebral cortex do not appear to reflect direct cortical neuronal necrosis, at least as evidenced by histological

criteria.

Deep Nuclear Structures. Involvement of deep

nuclear structures, principally thalamus and basal ganglia,

is particularly characteristic of hypoxic-ischemic neuronal injury in both preterm and term newborns. With

diffuse disease, thalamus is particularly vulnerable.

As discussed later, a particular pattern of injury in

term newborns involves a combination of affection

of neurons of thalamus, basal ganglia, and brain

stem, with relative sparing of cerebral cortical neurons.

Hypothalamic neurons and those of the lateral

geniculate nuclei also are especially vulnerable. In

preterm newborns, involvement of deep nuclear structures is a major form of gray matter injury.75 Injury

to thalamus and basal ganglia was apparent in 40%

to 50% of one series of 41 premature infants studied

at autopsy (see later).75 Of the basal ganglia, neurons

of the caudate, putamen, and globus pallidus are often

injured, in both term and premature newborns (see

Table 8-4). Neurons of the putamen (and head of

the caudate nucleus) are somewhat more likely to

be affected in the term infant, whereas neurons of

the globus pallidus are more likely to be affected

in the premature infant.5,6 This distinction is subtle,

however. Neuronal injury to basal ganglia usually

is accompanied by thalamic neuronal injury.

Indeed, the combination of putaminal and thalamic neuronal injury in my experience is typical of

neonatal hypoxic-ischemic disease, especially in the

term infant.

Brain Stem. Particularly characteristic of hypoxicischemic encephalopathy in the newborn is involvement of the brain stem.4,6-8,54-67,69-72,74-88 In general,

hypoxic-ischemic injury to the brain stem in the term

newborn tends to be more or less restricted to neurons.

With premature infants, although neurons are involved

primarily, injury may be so marked as to result in cystic

necrosis.65 As discussed later, involvement of neurons

of brain stem may occur in combination with basal

ganglia and thalamic involvement.

In midbrain, the inferior colliculus stands out in

terms of vulnerability. This finding is in keeping with

the studies of Ranck, Windle, Faro89,90 of asphyxiated

fetal monkeys, particularly with total asphyxia.

Neuronal injury is also found frequently in the neurons

of the oculomotor and trochlear nuclei, substantia

nigra, and reticular formation.

In pons, particularly frequently involved are the

motor nuclei of the fifth and seventh cranial nerves,

the reticular formation, the dorsal cochlear nuclei,

and the pontine nuclei. Striking involvement of the

nuclei in ventral pons and of the neurons of the subicular portion of the hippocampus in some cases led

Friede6 to the term pontosubicular neuronal necrosis. This

pattern of injury is discussed later.



351



In medulla, particularly vulnerable are the dorsal

nuclei of the vagus, nucleus ambiguus (ninth and

tenth cranial nerves), inferior olivary nuclei, and the

cuneate and gracilis nuclei. Involvement of neurons

of the inferior olivary nuclei is the single most

common brain stem neuronal lesion in both term and

preterm infants.8,72,75 In one series of 41 premature

infants studied at autopsy, fully 90% had evidence of

inferior olivary injury.75 Important clinical correlates of

many of these brain stem lesions are discussed in

Chapter 9

Cerebellum. The cerebellum is especially vulnerable

to hypoxic-ischemic neuronal injury, and the Purkinje

cells in the term infant and the granule cell neurons (of

both the internal and external granule cell layers) in

both the term and premature infant are the most vulnerable cerebellar neurons (see Table 8-4). Neurons of

the vermis may be especially easily injured in the term

infant.91,91a Neurons of the dentate nucleus (and other

roof nuclei) are also somewhat susceptible to injury,

more so in the preterm newborn than at later ages. In

the term infant, a subsequent disturbance of cerebellar

growth, especially involving the vermis, has been

observed by magnetic resonance imaging (MRI).91,91a

In this setting, frequent concomitant injury to thalamus and basal ganglia also raises the possibility of

transsynaptic effects.91,91a,92 Involvement of the cerebellum, especially cerebellar hemispheres, and subsequently impaired cerebellar growth are particular

features of very premature infants and are sufficiently

distinctive to be discussed as a separate form of selective neuronal necrosis (see later).

Spinal Cord. Affection of anterior horn cells by

hypoxic-ischemic injury has been identified.5,93,94

This involvement is accompanied clinically by hypotonia and weakness and electrophysiologically by signs of

anterior horn cell disturbance and may underlie at least

some cases of so-called atonic cerebral palsy (see

Chapter 9). The neuronal injury occurs in typical

form in the term infant and is similar cytopathologically

to that observed in other regions. When present in the

premature infant, the lesion, as with hypoxic-ischemic

injury to brain stem, often also involves contiguous

cellular elements that may have the histological appearance of infarction and may be accompanied by

hemorrhage.93

Cerebral–Deep Nuclear Neuronal Injury. Although

systematic data are difficult to gather, MRI studies

of ‘‘asphyxiated’’ term infants suggest that approximately 35% to 85% exhibit predominantly cerebral–

deep nuclear neuronal involvement.95-101 Among

neurons of cerebral cortex, those in the parasagittal

areas of perirolandic cortex are especially likely to

be affected. Involvement of the hippocampus and of

other neocortical areas was described earlier. The

most common additional neuronal lesion affects

basal ganglia, especially putamen, and thalamus. The

pathogenesis appears usually to involve a moderate

or moderate-to-severe insult that evolves in a gradual

manner (i.e., a ‘‘prolonged, partial’’ insult; see

‘‘Pathogenesis’’).



352



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



Deep Nuclear–Brain Stem Neuronal Injury.

Although involvement of neurons of basal ganglia

and thalamus occurs in approximately two thirds of

‘‘asphyxiated’’ term infants, in approximately 15% to

20% of infants with hypoxic-ischemic disease, involvement of deep nuclear structures (i.e., basal ganglia, thalamus, and tegmentum of brain stem) is the predominant

lesion.6,70,71,79,85,102-104 Until the advent of MRI,

detection of this deep gray matter predominance in

the living infant had not been accomplished readily,

and thus the relative frequency of this pattern of neuronal injury was not recognized. However, studies

based on MRI and careful clinical pathological correlations have delineated this pattern as a distinct

entity.70,71,85 The topography of the neuropathology

is illustrated in Figure 8-3.

At least some cases of this predominantly deep gray

matter form of selective neuronal injury may evolve to

status marmoratus, a disorder of basal ganglia and thalamus not seen in its complete form until the latter part

of the first year of life, despite the perinatal timing of

the insult. The basic initiating role of hypoxia-ischemia

is demonstrated not only by clinical data in human

infants (see later discussion) but also by the reproduction of the lesion in the newborn rat subjected to

hypoxic-ischemic insult,105,106 as well as in the term

fetal monkey subjected to intrauterine asphyxia.19

Status marmoratus has three major features:

neuronal loss, gliosis, and hypermyelination.5-7,107

Hypermyelination is the characteristic feature of the

lesion; this term refers to an apparent increase in

amount and an abnormal distribution of myelinated

fibers within the affected nuclear structures, especially

the putamen (Fig. 8-4). The hypermyelination has been

noted at as early as 8 months of life.6 The abnormal

myelin pattern provides a marbled appearance to the basal

ganglia, hence the term status marmoratus or etat marbre.

´

´

Previous observations by light microscopy had led to

the suggestion that the many myelinated fibers in status

marmoratus were axons, and the idea that such apparent overgrowth was a result of aberrant myelination of

nerve fibers was accepted for many years. However,

electron microscopic techniques were used to show

that the abnormal myelinated fibers, at least in part,

are astrocytic processes.108 It appears that the very

young brain, at the time of normal myelination, may

myelinate fibers that are not axonal in origin. Thus, this

distinctive response to injury appears to depend on the

time of occurrence of the insult as well as the locus of the

injury. Nevertheless, the proportion of infants with

hypoxic-ischemic involvement of basal ganglia and

thalamus who develop status marmoratus and the

determinants for the occurrence of this relatively

specific pathological response to injury versus that of

only gliosis and atrophy remain to be determined.

Concerning the sequela of gliosis and atrophy alone,

a reasonable speculation is that injury that is so

severe as to eliminate oligodendrocytes as well as

neurons may prevent the occurrence of the typical

hypermyelination of status marmoratus. As discussed

later (see ‘‘Pathogenesis’’), the hypoxic-ischemic insult

associated with the occurrence of predominant



Pons



Basal ganglia, thalamus



Medulla



A



Midbrain

Spine



B

Figure 8-3 Neuropathology of the deep nuclear–brain stem form of

selective neuronal necrosis. A, Schematic depiction of the topography

of the lesions in a typical case of a term newborn subjected to severe,

terminal asphyxia. The dark areas indicate nuclei with neuronal loss,

and the diagonally striped areas represent regions of marked gliosis.

B, Holzer stain of the pons for gliosis in a typical case. The tegmentum

is atrophied and deeply stained because of gliosis; the base of the

pons is nearly normal. (From Natsume J, Watanabe K, Kuno K,

Hayakawa F, et al: Clinical, neurophysiologic, and neuropathological

features of an infant with brain damage of total asphyxia type

(Myers), Pediatr Neurol 13:61-64, 1995.)



involvement of deep gray matter structures typically is

severe and abrupt in evolution (i.e., an ‘‘acute, total’’

insult).

Pontosubicular Neuronal Necrosis. Pontosubicular

neuronal necrosis is a fourth type of selective neuronal

injury with predominant involvement of neurons of the

basis pontis (i.e., not the tegmentum, as described earlier) and the subiculum of hippocampus.47,66,67,74,78,8083,109-111 Among all types of selective neuronal injury,

pontosubicular neuronal necrosis is by far the least

common. The lesion is characteristic of the premature



Chapter 8



Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis



353



infant but occurs in infants up to 1 to 2 months beyond

term (Table 8-5).66,67,78,80-83,109,112 A strong association exists with periventricular leukomalacia (PVL).

Although the disorder is characterized principally by

affection of neurons of ventral pons and of subiculum

of hippocampus, neuronal death in the fascia dentata of

the hippocampus was observed in 60% of cases in one

series.67 In several neuropathological studies that

included electron microscopy, labeling of oligonucleosomal fragments, and detection of Fas receptor and

activated caspase-3, the neuronal death in pontosubicular neuronal necrosis appeared to be predominantly

apoptotic.47,53,110,111,113



A



Cerebellar Injury. Cerebellar injury is particularly characteristic of premature infants, especially those of extremely low

birth weights, and it is sufficiently distinctive to be considered a fifth type of selective neuronal necrosis (Table

8-6). This injury has occurred primarily in premature

infants with serious respiratory disease, although single

major hypoxic-ischemic insults typical of asphyxiated

term infants have not been present. Cerebellar disturbance has been identified most often by the finding by

MRI of bilateral, generally symmetric decreases of cerebellar hemispheric volumes at term equivalent or later

in infancy or childhood.114-126 Although a few cases

have been focal and asymmetric, suggestive of infarction, nearly all abnormalities have consisted of bilateral

and symmetric diminutions in size. The possibility of a

trophic disturbance, perhaps related to supratentorial

white matter injury, is suggested by a strong association

with cerebral white matter injury (see later).122,124,125

Moreover, the high frequency of injury to neurons of

brain stem cerebellar relay nuclei (see earlier) also raises

the possibility of impaired trophic interactions at the

transsynaptic level.

Pathogenesis

Cerebral Ischemia, Impaired Cerebrovascular

Autoregulation, and Pressure-Passive Cerebral

Circulation. Cerebral ischemia, with deprivation of

oxygen and glucose, followed by reperfusion and the

cascade of metabolic events described in Chapter 6, is

the likely pathogenetic sequence in selective neuronal

necrosis (Table 8-7). Although the causative relationship between cerebral ischemia and both selective

neuronal necrosis and parasagittal cerebral injury (see

later) has been established in several excellent perinatal

animal models (see previous section), studies in human

infants also provide excellent support for the role of



TABLE 8-5



B

Figure 8-4 Status marmoratus. Coronal sections of cerebrum from

two infants who died several years after the perinatal insult.

A, Formalin fixed and unstained; note the marbled appearance of the

caudate nucleus and putamen. B, Stained for myelin; note the blackstaining myelin, particularly in the putamen. (Courtesy of Dr. E. P.

Richardson, Jr.)



Pontosubicular Neuronal Necrosis in

the Premature Newborn*



Pons and subiculum common sites of neuronal injury in premature newborns

Strong association with periventricular leukomalacia

Clinically associated with hypoxia-ischemia, hypocarbia, and

hyperoxia (reproducible in newborn rat by hyperoxia

alone)

*See text for references.



TABLE 8-6



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



Cerebellar Injury in the Premature

Newborn



Cerebellar injury, especially involving the cerebellar hemispheres, is generally bilateral and symmetric, especially

in very small premature infants.

Microscopic features include cerebellar neuronal injury, cerebellar white matter necrosis/gliosis, and neuronal injury

to brain stem relay nuclei to cerebellum.

Diminished cerebellar volume is the most common structural sequela.

Cerebellar injury is often associated with supratentorial

injury, especially to cerebral white matter, and with

relay nuclei in brain stem, both principally hypoxicischemic.

The cause may reflect a combination of transsynaptic trophic disturbances and direct neuronal injury.



diminished CBF secondary to systemic hypotension.

Thus, as discussed in detail in Chapter 6, because of

the impaired vascular autoregulation in asphyxiated

infants, CBF becomes passively related to arterial

blood pressure (Fig. 8-5; see Table 8-7). The impaired

vascular autoregulation has been documented in the

hours to days after the insult and, by extrapolation

from experimental data (see Chapter 6), is presumed

to begin during the insult, when hypotension is most

pronounced. This situation makes the infant exquisitely vulnerable to the diminutions in blood pressure

characteristic of severe asphyxia, and those regions

most vulnerable are in the distribution of selective neuronal necrosis as well as the watershed (i.e., parasagittal) distributions of the cerebrum; see later. The data of

Pryds and Greisen and their co-workers127,128 clearly

show that those asphyxiated term infants with impaired

autoregulation have the poorest neurological outcome

(see Fig. 8-5). The causes of the pressure-passive circulation

in the asphyxiated newborn could relate to the following factors: (1) the hypoxemia or hypercarbia, or both,

of the primary asphyxial insult; (2) the postasphyxial

impairment in vascular reactivity observed in experimental models of perinatal asphyxia and presumably

related to the effect of one or more of the vasodilatory

compounds that accumulate secondary to ischemiareperfusion (see Chapter 6); (3) an ‘‘immature’’



TABLE 8-7



Selective Neuronal Necrosis:

Pathogenesis



Cerebral ischemia

Impaired cerebrovascular autoregulation with pressurepassive cerebral circulation

Systemic hypotension

Regional vascular factors

Regional metabolic factors

Regional distribution of excitatory (glutamate) receptors

on neurons*

Factors related to the hypoxic-ischemic insult

Severity and temporal characteristics

Preceding/concomitant infection/inflammation

*Single most important factor for determining regional distribution of

selective neuronal necrosis.



CBF-MABP reactivity

(% change CBF/mmHg change MABP)



354



8

6

4

2

0

–2

–4



Asphyxiated—

Asphyxiated—

death or severe moderate to severe Asphyxiated— Nonasphyxiated—

brain injury later brain injury later normal outcome normal outcome



A



B

C

Groups A,B,C,D — term infants



D



Figure 8-5 Cerebrovascular autoregulation. This is expressed as

reactivity of cerebral blood flow (CBF) to changes in mean arterial

blood pressure (MABP), in a group of 19 asphyxiated term newborns

and 12 control infants. Note the normal reactivity in the control infants

(D) and the loss of reactivity in the infants with the poorest outcomes

(A and B). (Data from Pryds O, Greisen G, Lou H, Friis-Hansen B:

Vasoparalysis associated with brain damage in asphyxiated term

infants, J Pediatr 117:119-125, 1990.)



autoregulatory system with limited capacity for reactivity

because of the deficient arteriolar muscular lining of

penetrating cerebral arteries and arterioles in the third

trimester129,130; (4) an autoregulatory system with a

lower limit so close to the range of ‘‘normal’’ blood

pressure that even slight hypotension places CBF on

the down slope of the curve (see later discussion of

PVL); or (5) a combination of these factors. Whatever

the mechanisms, the clinical implications are enormous. Falls in arterial blood pressure lead to decreases

in CBF and injury to certain vulnerable brain cells (i.e.,

neurons in the distribution of selective neuronal necrosis), and regions (i.e., parasagittal cerebrum; see later).

Regional Vascular Factors. The reasons for the selective vulnerability of neuronal groups in the central nervous system have become increasingly clear. Regional

vascular factors certainly can play a role because neuronal injury is more marked in vascular border zones

(e.g., depths of sulci and parasagittal cerebral cortex;

see Table 8-7). Moreover, the relationship of pontosubicular necrosis with hypocarbia and hyperoxemia

(often following hypoxic-ischemic insult) suggests a

role for cerebral vasoconstriction in pathogenesis of

this specific type of neuronal necrosis of the premature

infant (see Table 8-5).6,66,112 However, the finding that

most of selective neuronal injury is not in strictly vascular distributions suggests that other factors are operative. For example, the rapid neuronal maturation in

the pons and subiculum during the time of occurrence

of this lesion suggests that vulnerability of this region

may relate in part to the simultaneous occurrence of

neuronal differentiation, the insults that impair brain

blood flow, and a propensity of these neurons to

undergo apoptosis.47,109-111

Regional Metabolic Factors. Regional metabolic factors

must play a central role (see Table 8-7). Those factors

that lead to hypoxic cell death in experimental systems



Chapter 8



Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis



(see Chapter 6) raise the possibilities of regional differences in anaerobic glycolytic capacity, energy requirements, lactate accumulation, mitochondrial function,

calcium (Ca2+) influx, nitric oxide synthesis, and free

radical formation and scavenging capacity. For example, the high metabolic rate and energy use of deep

gray matter may render these neurons particularly

vulnerable to the severe, abrupt ischemic insults

that lead to particular injury to these neuronal

structures (see later). Similarly, a particular regional

vulnerability of mitochondrial cytochrome oxidase

activity to hypoxic-ischemic insult may play a role in

determining the regional vulnerability of certain neurons to injury.131 However, little is known about these

issues in human brain.

Regional Distribution of Excitatory (Glutamate)

Receptors. The regional distribution of glutamate receptors, particularly of the N-methyl-D-aspartate (NMDA)

and alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic

acid (AMPA) types, now appears to be the single most

important determinant of the distribution of selective

neuronal injury (see Table 8-7). As discussed in detail

in Chapter 6, the topography of hypoxic-ischemic neuronal death in vivo is similar to the topography of glutamate synapses; the particular vulnerability of certain

neuronal groups in the perinatal period correlates with

a transient, maturation-dependent density of glutamate

receptors; extracellular glutamate increases dramatically at such receptors with hypoxia-ischemia; and

hypoxic-ischemic neuronal death in vivo can be prevented by administration of blockers of the NMDA

receptor-channel complex and, to a considerable

extent also, of non-NMDA receptors, especially Ca2+permeable AMPA receptors (see Chapter 6). The

demonstrations that the molecular mechanisms by

which activation of the glutamate receptors leads to

cell death operate over hours after termination of the

insult and that prevention or amelioration of such excitotoxic injury can be effected by glutamate receptor

blockers also administered after termination of the

insult have profound and obvious clinical implications

(see Chapters 6 and 9).

The importance of the regional distribution of glutamate receptors in the determination of regional selectivity of neuronal injury is particularly apparent in basal

ganglia. Thus, first, it is clear that a transient, dense

glutamatergic innervation of the basal ganglia occurs

in the perinatal period, both in the rat (see Chapter 6)

and in the human.132-136 Second, the development of

vulnerability of striatum to hypoxic-ischemic injury

parallels the expression of glutamate receptors and

the vulnerability to direct injections of glutamate (see

Chapter 6). Third, extracellular glutamate levels have

been shown to rise in perinatal models of hypoxicischemic striatal injury (see Chapter 6). Fourth, a

highly effective blocker of perinatal hypoxic-ischemic

striatal injury is MK-801, a specific blocker of the

NMDA receptor-channel complex (see Chapter 6).

Fifth, a specific hierarchy of potency of glutamate

receptor agonists exists for the production of striatal

injury, and this potency parallels the expression of



355



receptor subtypes and the inhibitory capabilities of specific receptor antagonists (see Chapter 6).

As noted earlier, neuronal injury in the brain stem

often accompanies such injury in basal ganglia.

Notably, studies of the developmental profiles of glutamate receptor subtype binding in the human brain

stem have shown transient elevations in inferior olive

and basis pontis of NMDA and kainate receptors in

early infancy.72,137

Perhaps related to the role of glutamate receptors of

the NMDA type in pathogenesis of selective striatal

neuronal injury is a relative sparing of the reduced form

of nicotinamide-adenine dinucleotide phosphate (NADPH)–

diaphorase neurons in hypoxia-ischemia.138,139 NADPH

diaphorase has been shown to be identical to nitric

oxide synthase, and generation of nitric oxide has

been shown to be one mechanism whereby activation

of NMDA receptors (expressed by NADPH-diaphorase

neurons) leads to striatal neuronal death (see also

Chapter 6). Because nitric oxide synthase is activated

by Ca2+, and Ca2+ influx follows activation of the

NMDA receptor (see Chapter 6), the data suggest a

sequence of NMDA receptor activation, activation of

nitric oxide synthase, generation of nitric oxide, and

diffusion of nitric oxide (a highly reactive molecule

that can generate free radicals) to adjacent neurons,

and free radical–mediated cell death. The peak period

of vulnerability of striatal neurons in the immature rat

corresponds to the peak periods of sparing of NADPHdiaphorase neurons and of hypoxic-ischemic vulnerability.138,139 The reason for the relative sparing of

NADPH-diaphorase neurons remains unclear, but

this sparing may contribute importantly to perinatal

striatal neuronal death. Currently, it is not known

whether a similar sparing of nitric oxide–synthesizing

neurons contributes to selective neuronal injury in

cerebral cortical and other areas vulnerable in hypoxia-ischemia in the neonatal human. However, the

potential importance for neuronal nitric oxide synthase

in mediation of such hypoxic-ischemic neuronal injury

is suggested by the results of a study of the development of neuronal expression of the enzyme in human

brain.140 The striking findings were a higher density of

nitric oxide synthase–positive neurons in late fetal

human brain than in adult brain and a concentration

of such neurons in areas known to be injured in selective neuronal necrosis (i.e., deeper layers of cerebral

cortex, striatum, and brain stem tegmentum).

Studies of glutamate receptors in developing human

cerebral cortex suggest that a transient expression of

Ca2+-permeable AMPA receptors and NMDA receptors occurs around the time of term birth

(Fig. 8-6).141 Thus, the theme is similar to that for

basal ganglia and brain stem (i.e., a maturation-dependent exuberant expression of Ca2+-permeable glutamate receptors becomes lethal to neurons with

excessive activation, as occurs with cerebral ischemia).

Factors Related to the Hypoxic-Ischemic Insult.

Factors related to the severity and the temporal characteristics of the insult appear to be of particular importance

in determining the major pattern of selective neuronal



356



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



injury in the newborn (see Table 8-3). Severe and prolonged insults result in diffuse and marked neuronal

necrosis, involving the many levels of the neuraxis

described earlier as the diffuse pattern of injury. The

cerebral–deep nuclear pattern of neuronal injury appears

to be related to insults that are less severe and prolonged, often termed partial, prolonged asphyxia. The

deep nuclear–brain stem pattern of injury to basal ganglia–thalamus–brain stem has been described in human

infants with a severe, abrupt event, often termed total

asphyxia (see earlier discussion). It is postulated that the

severe, abrupt event prevents the operation of major

adaptive mechanisms normally operative with asphyxial

events (see Chapter 6). The most important of these

may be the diversion of blood from the cerebral hemispheres to the ‘‘vital’’ deep nuclear structures. Because

the latter have high rates of energy use (and also a high

content of glutamate receptors), these nuclei are

particularly likely to be injured. In the more prolonged

and less severe insults, the diversion of blood to deep

nuclear structures occurs at least to a degree, and the

cerebral regions are more likely to be affected. Studies

in the near-term fetal lamb indicate that the severe

terminal insult that results in injury to deep nuclear

structures especially may be likely to occur after brief,

repeated hypoxic-ischemic insults first cause a cumulative deleterious effect on cardiovascular function

that presumably then can result in a severe late

insult.28,142-144

As discussed in detail in Chapter 6, experimental

data demonstrate the potentiation of hypoxic-ischemic

insults by preceding or concomitant infection/inflammation.

Thus, hypoxic-ischemic insults not severe enough to

cause injury alone can be rendered seriously injurious

if the fetus or infant is exposed to inflammatory factors



associated with intrauterine or postnatal infection. This

phenomenon could underlie, at least in part, the accentuated risk of apparent hypoxic-ischemic brain injury

observed in infants who sustain their insults in association with chorioamnionitis or who have elevated levels

of cytokines in blood or cerebrospinal fluid (CSF).145155 However, most term infants who are exposed to

chorioamnionitis have an uncomplicated neonatal

course and neurological outcome, and histological

chorioamnionitis does not appear to increase the risk

of adverse outcome even in infants with hypoxicischemic encephalopathy.156 The presence or absence

of a fetal inflammatory response may be most critical, and

the determinants of such a response in the presence of

chorioamnionitis require clarification.



Parasagittal Cerebral Injury

Cellular and Regional Aspects

Parasagittal cerebral injury refers to a lesion of the cerebral cortex and subcortical white matter with a characteristic distribution (i.e., parasagittal, superomedial

aspects of the cerebral convexities) (Figs. 8-7 and

8-8). The injury is bilateral and, although usually symmetrical, may be more striking in one hemisphere than

the other. The posterior aspect of the cerebral hemispheres, especially the parietal-occipital regions, is

more impressively affected than the anterior aspect.

The term watershed infarct has been used to describe

the lesion and to emphasize its ischemic nature (see

later discussion). I prefer the more descriptive term

parasagittal cerebral injury.

Parasagittal cerebral injury is characterized by necrosis of the cortex and the immediately subjacent white



Figure 8-6 Transient expression of calcium-permeable (GluR2-deficient) alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid (AMPA)

receptors in cerebral cortex and cerebral white matter (WM) in developing human brain. Note the transient dense expression of GluR2-deficient

AMPA receptors in cortical pyramidal neurons around the time of term. N-Methyl-D-aspartate receptors show a similar time course (data not shown).

The changes in cerebral white matter are discussed later in relation to periventricular leukomalacia (PVL). (From Talos DM, Follett PL, Folkerth RD,

Fishman RE, et al: Developmental regulation of alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid receptor subunit expression in forebrain

and relationship to regional susceptibility to hypoxic/ischemic injury. II. Human cerebral white matter and cortex, J Comp Neurol 497:61-77, 2006.)



Chapter 8



Hypoxic-Ischemic Encephalopathy: Neuropathology and Pathogenesis



Figure 8-7 Parasagittal cerebral injury, coronal view. Schematic diagram of the distribution of the injury, which is indicated by symmetrical

black areas in the superomedial aspects of cerebrum.



matter; neuronal elements are most severely affected.

Although usually nonhemorrhagic, the areas of infarction may be hemorrhagic. In particularly severe cases,



357



the necrosis extends to a large proportion of the lateral

cerebral convexity (see Fig. 8-8B), especially in the parietal-occipital regions, the most vulnerable regions of

the cerebrum. The precise pathological evolution of

parasagittal cerebral injury in the newborn is not

known, but atrophic gyri or ulegyria, or both, are the

chronic neuropathological correlates. Parasagittal cerebral injury is characteristic of the full-term infant with

perinatal asphyxia; it is unlikely, in fact, that cerebral

necrosis in the parasagittal distribution occurs in the

premature infant to a major degree for reasons outlined

in the section ‘‘Periventricular Leukomalacia.’’

Parasagittal cerebral injury has been well documented in classic neuropathological writings (e.g., the

work of Friede,157 Courville,158and Norman and colleagues159), particularly those concerned with older

survivors with cerebral palsy. However, the lesion has

been more difficult to define as an isolated entity in neuropathological studies of infants dying in the neonatal

period, although examples are apparent (see Fig. 8-8).

I believe that the difficulty in pathological identification

of the discrete lesion in the neonatal period relates to

the severe nature of the cases in newborns who die. Thus,

the neuropathological findings are most often diffuse



A



B

Figure 8-8 Parasagittal cerebral injury. A, Coronal section of cerebrum in an asphyxiated, full-term infant who died on the third postnatal day.

Areas of necrosis of cerebral cortex and subcortical white matter in the parasagittal regions are marked by arrowheads. B, Lateral view of cerebral

convexity of a 6-month-old infant who had experienced severe perinatal asphyxia. Note the cortical atrophy in parasagittal distribution (compare with

Fig. 8-9). (B, Courtesy of Dr. Alan Hill.)



358

TABLE 8-8



UNIT III



HYPOXIC-ISCHEMIC ENCEPHALOPATHY



Parasagittal Cerebral Injury:

Pathogenesis



Anterior cerebral

artery



Cerebral ischemia

Impaired cerebrovascular autoregulation with pressurepassive cerebral circulation

Systemic hypotension

Parasagittal vascular factors

Arterial border zones and end zones

Excitatory (glutamate) receptors on neurons (and premyelinating oligodendrocytes)



Posterior

cerebral

artery



Middle

cerebral

artery



and severe, very frequently complicated by autolytic

changes related to survival for many hours or days on

life support. These diffuse changes obscure elemental

lesions, such as parasagittal cerebral injury, which,

however, is identifiable in those less severely affected

infants who survive. In keeping with this explanation is

the observation of a high frequency in asphyxiated term

newborns (%90% of whom survive) of parasagittal cerebral injury identifiable by radionuclide brain scanning,160,161 positron emission tomography,162 and

MRI98,163,164 (see the section in Chapter 9 on diagnosis). Indeed, in an MRI study of 173 term infants with

neonatal encephalopathy, fully 45% (n = 78) had watershed injury as the predominant lesion.164 The computed tomography scan, still often used in evaluation

of such infants, is not particularly sensitive for detection of this lesion because the axial images frequently

fail to detect the superficial cortical-subcortical lesions

of parasagittal injury. The advent of MRI scanning,

with coronal and lateral views, has proved more effective in identification of parasagittal cerebral injury in

vivo and has provided further documentation of its frequency (see Chapter 9 ).

Pathogenesis

The pathogenesis of parasagittal cerebral injury relates

principally to a disturbance in cerebral perfusion. The

two factors underlying the propensity of the parasagittal region to ischemic injury relate to parasagittal vascular anatomical factors and cerebral ischemia with a

pressure-passive state of the cerebral circulation (Table

8-8). The reasons that one infant with ischemia may

develop primarily parasagittal cerebral injury and

another may have the various patterns of selective neuronal necrosis are not entirely clear. As discussed in the

section ‘‘Selective Neuronal Necrosis,’’ the severity and

the temporal characteristics of the insult are likely to be

very important. Indeed, some degree of concomitant

selective neuronal injury, particularly involving basal

ganglia and thalamus, is common in my experience.

Cerebral Ischemia, Impaired Cerebrovascular

Autoregulation, and Pressure-Passive Cerebral

Circulation. The importance of the asphyxial and

postasphyxial impairment of cerebrovascular autoregulation (see Chapter 6) in the genesis of cerebral ischemia with associated systemic circulatory failure is

apparent for parasagittal cerebral injury, as described



Basilar artery

Vertebral artery



Figure 8-9 Parasagittal cerebral injury, lateral view. Schematic diagram of cerebral convexity, lateral view, showing distribution of major

cerebral arteries. The distribution of injury, shown by the line-marked

area, is in the border zones and end fields of these arteries.



earlier for selective neuronal necrosis (see earlier

discussion).

Parasagittal Vascular Anatomical Factors. The

likely areas of greatest ischemia relate to parasagittal

vascular anatomical factors (see Table 8-8). Thus, the

areas of necrosis in parasagittal cerebral injury are in

the border zones between the end fields of the major

cerebral arteries (Fig. 8-9).165,166 These border zones

are the brain regions most susceptible to a fall in cerebral perfusion pressure. Meyer,165 who defined the

characteristic topography of the cerebral lesions in 30

infants in the 1950s, related the injury to systemic hypotension. This watershed concept is based on the

analogy with an irrigation system supplying a series of

fields with water and emphasizes the vulnerability of

the ‘‘last fields’’ when the head of pressure falls.167,168

Experimental support for this concept initially was provided in the monkey by Brierley and co-workers,169,170

who produced rapid, profound systemic hypotension

and prevented hypoxemia when respiratory failure

developed. Typical watershed lesions were produced

in the cerebral cortex (and cerebellum) and were

ascribed to the sharply reduced CBF. As my colleagues

and I160,162 observed in asphyxiated human infants,

more marked injury was demonstrated in the posterior

cerebrum in the experimental animals, as well as in the

adult human.166,169,170 The more marked injury in

posterior cerebrum presumably relates to the finding

that this region represents the watershed of all three

major cerebral vessels (see Fig. 8-9).

The border zone concept has received ample additional experimental support in several developing

animal models. Parasagittal cerebral cortical-subcortical injury has been documented in the perinatal

monkey, sheep, rabbit, and mouse subjected to a variety of insults complicated by hypotension and