Main

The interrelationships among maternal (prenatal) infection, preterm birth, and neonatal brain damage are incompletely understood. In this review we offer one scheme and suggest how maternal IUI, neonatal IVH, PVL, and cerebral palsy might be related. We place special emphasis on the relation between IVH and PVL, which are traditionally viewed as two distinct disease entities, and present evidence for the contrary.

In the epidemiology section, we discuss the observations that IUI, IVH, PVL, and cerebral palsy are much more common among infants born near the end of the second trimester than among their gestationally older peers. In the cytokine section, we review the recent clinical and basic science literature to support our hypothesis that a web of mediators of inflammatory responses accounts for many of these relationships.

EPIDEMIOLOGY

Preterm birth is defined as delivery before the completion of 37 wk of gestation(1). In this article, however, we focus on the increased risk of brain damage among the very preterm, i.e. those born before 32 wk of gestation, and/or those with a VLBW (<1500 g). In the United States, 1-2% of liveborn babies fall into these categories(2, 3).

Infection. Preterm labor and prelabor rupture of membranes, but not preeclampsia/pregnancy-induced hypertension, appear to be related to maternal infection(4–7). For example, chorioamnionitis is associated with a 3-fold increased risk of delivery before term with intact membranes and a 4-fold increased risk of preterm birth due to prelabor rupture of membranes(8).

Bacterial vaginosis at 23-26 wk of gestation appears to increase a woman's risk of preterm delivery by 50%(9). A randomized trial comparing women with bacterial vaginosis, who received antibiotic treatment, with untreated peers resulted in a 27% decrease of preterm delivery(10).

We prefer the use of the term “intrauterine infection” (IUI) to the widely used “intraamniotic infection” to emphasize that the amniotic fluid need not be involved in upper genital tract infection(11), and to include the possibility of prepregnancy origin of the infection(12). When referring to the work of others, however, we tend to use the term used in the original article.

IVH. IVH occurs in up to 40% of infants below 35 wk or <1500 g as compared with 3-4% of babies with a gestational age >37 wk(13). About 5% of infants with apparently isolated IVH later develop cerebral palsy(14).

White matter damage. Neonatal white matter damage, which has been subdivided into periventricular hemorrhagic infarction and PVL on the basis of neuropathologic findings(15), is presumably more important than IVH in predicting long-term developmental disability(16). We tend to restrict ourselves to the description of hyper- and hypoechoic images demonstrable by ultrasound in vivo. We do so because the widely used ultrasound patterns of periventricular hemorrhagic infarction and PVL account for only a portion of neonatal white matter damage observed in large populations of infants (our unpublished observation). About 60% of infants with PVL, defined sonographically as a“hypoechoic” or “echolucent” image in the cerebral white matter, later develop cerebral palsy(14, 16).

IUI and IVH. Although gestational age is known to be a major predictor of IVH occurrence(13), the exact mechanisms leading to the occurrence of blood in the baby's ventricles are still unknown. Preterm infants whose amnion is acutely inflamed are at a 3-4-fold greater risk than are their peers of developing IVH(17–19). Receipt of prophylactic antibiotic by the gravida after prelabor rupture of membranes has been inconsistently followed by reduced risk of IVH in her offspring(20, 21).

Our own findings indicate that, among VLBW infants, those whose placenta had any inflammation were at a 2.6-fold increased risk of IVH (26%versus 12%) (Table 1). An additional intriguing finding is that this increased risk is most prominent in the subgroup of infants born after a short duration of ruptured membranes (data not shown). This suggests that a preexisting infection of the mother before birth is more important in predicting IVH risk than the perinatal ascending infection that often accompanies “long” duration of ruptured fetal membranes (our unpublished observation).

Table 1 Risk (%), odds ratio (OR), and 95% confidence interval (95% CI) of any IVH, PVL, or ventriculomegaly (VMEG) among VLBW infants who were exposed (n = 546) and not exposed (n =598) to intrauterine infection (IUI: any membrane inflammation, chorionic vasculitis, acute umbilical vasculitis, and/or amnion epithelial necrosis)
Full size table

This relationship between IUI and IVH might be due to associations with a common confounder variable. Perhaps the most obvious candidates for confounding our perception of the IUI/IVH relationship are low gestational age, labor, and vaginal delivery. However, when we adjust for these potential confounders, the IUI-IVH relationship persists.

IUI and PVL. In our sample, those infants whose placental membranes were inflamed were at 60-70% increased risk of PVL and ventriculomegaly (Table 1). These findings are in agreement with the results of others. A positive placenta culture or chorioamnionitis was present in the fully 54% of infants <36 wk of gestation who developed echolucencies of the white matter within 3 d after birth(22). VLBW infants with periventricular echodensities, which often precede PVL, were exposed to chorioamnionitis 2.6 times more often than their controls without echodensities(23). Ultrasonographic evidence of “cystic PVL” was present in 9% of a sample of 753 preterm infants born between 24 and 32 wk of gestation, whereas this percentage was 19% in the subset of babies whose mothers had IUI and even higher (22%) among those whose mothers had IUI and premature rupture of membranes(24). The finding that acute appendicitis is associated with both preterm labor and IVH or PVL(25) suggests that even extrauterine, intraabdominal infections can be part of the cascade of events linking infection, labor, and neonatal brain injury.

IVH and PVL. Among preterm or VLBW infants, IVH is associated with an increased risk of PVL. In a cohort of 124 neonates with a birth weight<1250 g, intraparenchymal echodensities in the periventricular region were found only among those infants who also had IVH(26). Grade III or IV IVH was six times more common among infants who had signs of periventricular white matter damage by 3 d after birth, than among gestational age peers who had no white matter abnormalities(22). Among infants with an abnormal periventricular echogenicity, those who had IVH were three times more likely to develop so-called periventricular“cysts” than those who did not have IVH(27). Fully 75% of the brains of infants who died after an in vivo diagnosis of IVH had PVL on histologic postmortem examination(28). In our sample, the prevalence of PVL was 1-2% among infants without IVH, 5-fold higher among infants with IVH only, and 24 times higher among those who had both IVH and ventriculomegaly(Table 2). Once an infant has IVH, her risk of PVL is elevated to the same extent regardless of whether or not her placental membranes were inflamed.

Table 2 Risk (%), odds ratio (OR), and 95% confidence interval (95% CI) of any PVL after occurrence of IVH, or IVH and ventriculomegaly (VMEG) among VLBW infants
Full size table

PVL and cerebral palsy. Children with spastic cerebral palsy who die have neuropathologic changes in the white matter consistent with PVL in the neonatal period (i.e. prominent gliosis and paucity of cerebral white matter)(29). Between 60 and 100% of preterm infants with PVL subsequently develop cerebral palsy(14, 16). Later in childhood, approximately 90% of those born preterm who develop spastic cerebral palsy have magnetic resonance images of the white matter viewed as sequelae of neonatal PVL(30, 31).

IUI and cerebral palsy. To support the scenario laid out in the preceding paragraphs, we need documentation that, among children born preterm, those who develop cerebral palsy are more likely than their peers to have been exposed to maternal IUI. Such an association was found in two recent case-control studies. Chorioamnionitis was diagnosed in the mothers of 17% of preterm infants who subsequently developed cerebral palsy in contrast to only 3% among mothers of preterm controls (odds ratio 4.2, 95% confidence interval 1.4-12)(32). In the other study of VLBW infants, those whose mother had chorionitis were at increased risk of cerebral palsy (odds ratio 4.3; 95% confidence interval 1.1-13)(33).

CYTOKINES

Major advances in our understanding of the biology of preterm birth have been made by the study of the cytokine network of mediators of uterine contractions. Although some reviewers of the topic do not believe in the importance of maternal infection as an initiator of preterm delivery(34), others find support for the hypothesis that intraamniotic infection(6, 35, 36) or, even earlier, IUI(12) precede or accompany an appreciable proportion of preterm births.

Adinolfi(37) proposed that cytokines produced by the immune system during the course of maternal infection are harmful to the developing brain of the unborn infant. Leviton(38) extended this hypothesis, suggesting that cytokines such as TNF-α produced in response to IUI contribute to both preterm birth and periventricular white matter damage. Indeed, women who were found to have increased levels of TNF-α or IL-6 in cervicovaginal swabs were at a 3-fold increased risk of delivering preterm(39). Compared with women who gave birth at term, those who delivered at or before 34 wk of gestation were much more likely to have elevated amniotic fluid levels of TNF-α, IL-6, and prostaglandin E2, and to have histologic chorioamnionitis(40). Most recently, IL-6 levels were significantly elevated in the amniotic fluid of women whose infants later developed IVH or PVL(41). IL-6 levels were also elevated in the umbilical cord blood of neonates who developed a persistent echodensity or echolucency(42).

The somewhat simplistic, but helpful model of the “maternofetal unit” may be an approach to the problem of how IUI, preterm birth, IVH, and PVL may be linked. The uterus, the fetal circulation, and the fetal brain can be viewed as three compartments with the placenta representing the boundary between the mother and fetus, and the fetus' BBB as the boundary between the systemic circulation and the brain (Fig. 1).

Figure 1
figure 1

Hypothetical pathways describe how proinflammatory cytokines gain access to the fetal brain after being produced during maternal IUI and how these cytokines might contribute to the occurrence of IVH and white matter damage. The uterus, fetal circulation, and fetal brain can be viewed as three compartments separated by boundaries represented by the placenta (here: operationally including the membranes) and the BBB. The symbol♦ denotes a location of cytokine production. Pathways: Cytokines originate in the uterus/amniotic fluid (1) and/or the placenta(2)(6, 11, 35, 36, 40, 44–48, 61–68). The bidirectional arrow denotes that the primary source of intraamniotic cytokines may vary, i.e. membranes, placental tissue, or fetal sources (see below pathway 4). Some proinflammatory cytokines are produced by cells in the fetal circulation(3), presumably in response to amniotic/placental cytokines(49–56, 74). These cytokines might then have access to the fetal brain(1-3) or the amniotic fluid (by fetal urine) (4). Once in the brain either by crossing the intact BBB(70–73) or by altering its permeability(84–86), cytokines might cause white matter damage directly (5) or via their causing IVH (6), or they might induce further cytokine production by microglia and astrocytes(7)(57–60, 78, 128). These brain-derived cytokines might then in turn damage germinal matrix endothelium (8) or developing white matter (9)(37, 38, 41, 42, 127). The pathways are neither mutually exclusive nor exhaustive.

Full size image

When proposing a scenario of how cytokines may be the mediators between the mother's infection and brain damage in her newborn, one has to find support for the following hypotheses: 1) cytokines are present during infection in all three compartments (uterus, fetal circulation, and brain); 2) cytokines are present in, or can cross, the boundaries (placenta and BBB) between the compartments; 3) cytokines contribute to the occurrence of IVH, and 4) cytokines are involved in the induction of the white matter damage usually described as PVL. In the remainder of this review we emphasize the possible role of the proinflammatory cytokines IL-1β, TNF-α, and IL-6.

Cytokines are present during infection in all three compartments(i.e. uterus, fetal circulation, and brain). IL-6 first appears in amniotic fluid during the second trimester, and increases at term (with the onset of labor) as do TNF and IL-1, suggesting a role for these cytokines in normal labor(43). The role of the cytokine cascade in the induction of preterm labor has been discussed in detail by others(6, 36, 44).

TNF-α and/or IL-6 have been found elevated in amniotic fluid of pregnant women with chorioamnionitis(40, 45–47), even among women with intact membranes(48), but only IL-6 was elevated in both maternal and fetal blood(47). IL-6 levels are higher in the cord blood of term newborns whose mothers had clinical(49) and histologic chorioamnionitis(50) than in their unexposed peers. After stimulation with lipopolysaccharide and concanavalin A(51) or IL-1(52), both neonatal and adult mononuclear cells produce IL-6. The blood serum concentration of IL-6 is higher in infants with confirmed prenatal infections than in those with infections acquired after birth(53), and elevated serum TNF-α levels are associated with Gram-negative sepsis in both term(54) and preterm newborns(55). Recently, IL-6 blood levels of ≥100 pg/mL before the 12th h of life were proposed as a better marker of neonatal infection than the usually obtained C-reactive protein(56).

Finally, proinflammatory cytokines and their respective receptors are expressed in the brain (for review, seeRefs. 57 and 58). Lipopolysaccharide-stimulated microglia produce large amounts of IL-1β, TNF-α, and IL-6. In murine models, these effects can be blocked by pentoxifylline (blocks TNF-α and IL-1) and dexamethasone(blocks TNF-α and IL-6)(59). IL-1β-production by microglia is also induced by TNF-α and IL-1β itself. In turn, IL-1β induces the production by fetal astrocytes of IL-1β, TNF-α, and IL-6. Thus, as suggested by Lee et al.(60), IL-1 seems to play a key role as a mediator between fetal microglia and astrocytes in the presence of lipopolysaccharide.

Cytokines are present in or may cross the boundaries (placenta and BBB) between the compartments. Although placental cells produce IL-6 and other proinflammatory cytokines(61–64), the evidence that these cytokines cross the placenta remains unclear(65). On the other hand, cells in the placental villi and chorion [for an overview, see Mitchell et al.(66)] and umbilical veins(67) are also a source of proinflammatory cytokines, thereby allowing them access to the fetal circulation without requiring maternally derived cytokines to cross the placenta. The amounts of proinflammatory cytokines produced by cultured fetal amniochorionic membranes in response to lipopolysaccharide challenge are comparable with those found in the amniotic fluid after microbial invasion(68), suggesting intraplacental production rather than cross-placental transport (Fig. 1) as a relevant mode of cytokine occurrence.

In the fetus, the BBB is much less effective than in children and adults(69). Nevertheless, IL-1α(70), IL-1β(71), IL-6(72), and TNF-α(73) cross the intact adult murine BBB. More important for our scenario, however, are the cytokine-BBB interactions that might lead to BBB damage (see below).

Cytokines may contribute to the occurrence of IVH. IVH is thought to originate mainly from the blood vessels in the germinal matrix. During fetal development, these vessels grow rapidly and may be prone to a variety of damaging mechanisms, which may, in part, be mediated by cytokines.

Access. Cytokines are present in the fetal and neonatal circulation(42, 49, 50, 53–56, 74) and by this means might damage the endothelium of germinal matrix vessels. A second way cytokines can reach the germinal matrix region is via cerebrospinal fluid. IL-1β and TNF-α have been detected in the cerebrospinal fluid of newborns with Gram-negative meningitis(75), and IL-6 concentrations are elevated in the cerebrospinal fluid of children with bacterial meningitis(76). Choroid plexus ependymal cells are capable of TNF-α production in vivo and in vitro after endotoxin exposure(77). Once in the cerebrospinal fluid, cytokines might attack the developing germinal matrix vessels from their outside.

A third way cytokines can gain access to germinal matrix blood vessels is from adjacent microglia and/or astrocytes(60, 78). Astrocytes are capable of producing cytokines and are in close connection with endothelial cells in forming the BBB. Parenchymal microglia, also capable of producing cytokines, extend their processes into the ependymal layer of the third ventricle(79).

Mechanisms. Usually, the BBB protects the brain parenchyma from exposure to protein leakage from blood vessels. The BBB is characterized by tight junctions between the endothelial cells of most brain blood vessels, except in some areas close to the third ventricle, the so-called circumventricular organs(80, 81). However, early in gestation the endothelium of the germinal matrix has neither fenestrations(such as that of the circumventricular organs) nor tight junctions, but instead has so-called “strap junctions”(82). This strap-junction ependyma is replaced at about mid-gestation by a“mature looking ependymal layer”(83). This period of physiologic BBB change could be a time of increased vulnerability of the fetus' BBB to cytokine-induced alterations.

In adult rats, intracisternal administration of IL-1β and TNF-α result in a reversible, dose-dependent increase of BBB permeability(84), as do intracerebral injection of TNF-α(85) and systemic administration of IL-2 and IL-6(86). The BBB does not develop in transgenic mice overexpressing IL-6(87). In an observational study on human patients, however, intrathecal TNF-α levels, but not those of IL-1β, correlated with a measure of BBB damage(88). On the other hand, when administered systemically via intracarotid injection to rats, TNF-α resulted in an increase of the BBB(86).

Indomethacin appears to reduce the risk of IVH(89). One explanation for this is that indomethacin enhances maturation(90). Another is that it reduces (or prevents) post BBB injury permeability increases(91). A third is that indomethacin abolishes the increased BBB permeability induced by proinflammatory cytokines(92).

One role of cytokines is the mediation of intravascular leukocyte adhesion and transendothelial migration. TNF-α-treated astrocytes express the mRNA for adhesion molecules E-selectin, vascular cell adhesion molecule 1, and intercellular adhesion molecule 1(93), which“might guide inflammatory leukocytes into and through the brain parenchyma”(94). Another cytokine presumably involved in the chemoattraction of mononuclear leukocytes in the CNS during inflammation is astrocyte-produced interferon-γ-inducible protein 10(95), a chemokine whose expression is induced by TNF-α(96) and which has been shown to inhibit angiogenesis(97, 98). Most important for our scheme employing cytokines to explain the occurrence of IVH during maternal infection is that this process of adhesion of leukocytes to the endothelium may be followed by endothelial cell damage(99, 100). If TNF-α is able to induce hemorrhagic necrosis in highly vascularized and rapidly growing mouse tumors (one of the first functions observed for this cytokine which gave it its name)(101), could it similarly damage the fetal germinal matrix? In another context, brain hemorrhages occurring during the course of cerebral malaria have readily been related to TNF-α mediated cerebrovascular damage(102).

Cytokines can have profound effects on the coagulation system(103). TNF interacts with IL-1 in inducing procoagulant activity in the vascular endothelium(104), prostacyclin synthesis(105), and expression of plasminogen-activator-inhibitor, the main inhibitor of fibrinolysis(106–108). These observations relate to the clinical finding that IL-1β, IL-6, and TNF-α measurements were significantly elevated in adult patients with disseminated intravascular coagulation(109). Blood platelets are activated, and their aggregation is enhanced by IL-6(110), whereas TNF-α induces platelet-activating factor release(111). Platelet-activating factor can lead to vasoconstriction(112) as might IL-6 during subarachnoid hemorrhage(113), which can occur as a complication of CNS infection(114).

In summary, cytokines are able to mediate BBB alteration or even breakdown, intravascular cell adhesion, coagulation and/or thrombosis, and vasoconstriction in the presence of infection. These mechanisms may in turn lead to endothelial damage and subsequent hemorrhage from germinal matrix vessels into the ventricles of the preterm newborn.

Cytokines may be involved in the induction of white matter damage. IVH might contribute to white matter damage via ventriculomegaly, which compresses periventricular capillaries, thereby leading to ischemia in the white matter(115). Another mechanism might be periventricular hemorrhagic venous infarction, although its“pathogenesis... is not entirely established”(15). Furthermore, both IVH and the multifocal or focal hemorrhages in the white matter may be expressions of a diffuse hemorrhagic tendency, as seen in our sample (F. Gilles, J. Golden, and R. Rudelli, personal observation).

How may cytokines in the intraventricular blood contribute to periventricular white matter damage? First, because an ependymal layer does not line the lateral ventricles until about 22 wk of gestation(116), there might be no physiologic barrier protecting the periventricular parenchyma from potentially harmful cytokines in the cerebrospinal fluid of extremely preterm babies.

Second, potentially harmful molecules, either produced in the ventricle(e.g. in the choroid plexus) or present in the intraventricular hemorrhage, might enter the periventricular parenchyma from the ventricular lumen by passing the intact ependyma on the trans-cellular level. One of the tasks of the normal fetal ependyma, beside many others, is the exchange of fluid, ions, and small molecules between the intraventricular space and the surrounding tissue(116). We still do not know if intraventricular cytokines pass through the ventricular ependyma to reach the periventricular white matter. However, because cytokines are able to cross the intact BBB(70–73), and the BBB shares some structural characteristics with the ventricular ependyma (e.g. tight junctions), it seems plausible that cytokines can pass through the intact ventricular ependyma once it is established.

Third, hemorrhagic cerebrospinal fluid may leave the ventricle after interruption of ependymal continuity. The integrity of the ventricular ependyma may be interrupted by mechanical trauma (e.g. stretching in posthemorrhagic ventricular dilation), infarction, or inflammation(117). The resulting atrophy of the ependymal cells and discontinuity of the ependymal lining (gaps) may expose the parenchyma to“cerebrospinal fluid edema”(118) which, in our scenario, may carry blood and cytokines. What may be found in the subventricular brain parenchyma in cases of ependymal injury are an abundance of glial cells and, in the presence of infection, macrophages(117). Macrophages and glial cells are not only producers of cytokines, but are also histologic characteristics of neonatal white matter damage(119).

Fourth, the inflammatory response mounted by the ependyma after IVH(120) might include the production of TNF-α(77) and glial fibrillary acidic protein(116, 121–123). Glial fibrillary acidic protein is expressed by glial cells and in the white matter of infants with histologic evidence of PVL(124).

Once in the parenchyma, how may cytokines cause brain tissue damage? In the kitten, intraperitoneal injection of lipopolysaccharide, a strong stimulator of cytokine production, was followed by astrogliosis, cystic necrosis, and deposits of eosinophilic or mineralized debris in the white matter(125). Now there is evidence that cytokines may be mediators of neuronal and white matter tissue injury. Microinjections of IL-1, IL-2, IL-6, and TNF-α into neonatal murine cortex enhance the extensive astroglial reactivity that follows a scissors stab wound(126). Recently, microglial expression of TNF-α immunoreactivity was found in the white matter of infants with PVL associated lesions twice as commonly as in the white matter of those without such lesions(127). This observation alone does not explain how tissue destruction may occur, but supports the findings of others that fetal microglia are capable of TNF-α release and TNF-α mRNA expression(60, 128). Cytokines such as TNF-α (or IL-6) might play a role in damaging developing white matter by 1) leading to hypotension, 2) inducing blood vessel obstruction via intravascular coagulation, 3) adversively influencing oligodendrocytes, astrocytes, and myelin, or 4) by inducing the production of other cytokines such as platelet activating factor, which in turn has cell damaging properties [for review, see Leviton(38)].

Multivariability. Although we have emphasized the cytotoxic effects of cytokines, some relationships between cytokines and other mediators of brain damage in the fetus and newborn deserve mention. IL-1 release by human fetal microglia leads to IL-6 and TNF production by astrocytes(60). Fetal astrocytes also produce nitric oxide upon stimulation with IL-1, and this process is enhanced by addition of interferon-γ or TNF-α(129). Maximal injury to human fetal brain cell cultures containing neurons and glial cells occurred after the addition of IL-1β plus TNF-α(130). This synergism of TNF-α and IL-1β has also been observed in animal models of septic shock(131) and can be partially blocked by a nitric oxide synthase inhibitor (45% reduction of damage) and by N- methyl-D-aspartate receptor antagonists (55%)(130). This gives rise to the speculation that cytokines in the developing brain play a role in the mediation of neurotoxicity by nitric oxide(59, 129, 132) and excitatory amino acids (for review, seeRefs. 133 and 134).

In the newborn rat model of so-called “hypoxia-ischemia,” IL-1 and IL-6 mRNA are expressed in brain tissue(135). Human umbilical vein endothelial monolayers produce significantly more IL-1 (α and β) and IL-6 when exposed to hypoxia/reoxygenation than when incubated in room air(66). The ability of superoxide dismutase and glutathione peroxidase to prevent this effect suggests a role for oxygen-derived free radicals in the synthesis of cytokines in this setting.

We have only addressed mechanisms involving cytokines. Most certainly, cytokines are only one piece in the multivariable puzzle of neonatal white matter damage. However, this does not diminish their importance.

CONCLUSION

The epidemiologic evidence of an association between IUI in the mother and both IVH and white matter damage in her newborn, at least in part, can be explained by the biologic effects of proinflammatory cytokines. Further etiologic studies employing the cytokine network are needed to develop strategies intended to prevent some of the disabilities associated with very preterm delivery. Such strategies should emphasize prevention and treatment of IUI to reduce the risk of very preterm delivery(10, 12), although anticytokine strategies deserve consideration(136–138).