Main

Motor paralysis and cognitive deficits are major complications of premature birth in humans. PVL and PHI are the brain lesions most frequently observed in infants born between 25 and 32 wk of gestation (1). Neither the distribution nor the extent of these destructive lesions account fully for the neurologic impairments observed in preterm infants with PVL or PHI (2). Furthermore, premature infants born before 32 wk of gestational age can develop neurologic deficits in the absence of detectable perinatal brain lesions (3). In aggregate, these data suggest that mechanisms distinct from or secondary to axonal damage in the periventricular white matter contribute to the pathophysiology of spectrum of neurologic disorders observed in preterm infants. Several hypotheses, which are not mutually exclusive, have been formulated regarding these mechanisms.

In severe PVL and PHI, the interruption of projection and associative fibers by the lesions clearly has adverse effects (4). PVL and PHI are associated with death or degeneration of axotomized neurons and with formation of abnormal intracortical circuits (5). Conceivably, extensive PVL and PHI may involve the subplate neurons, which play important roles in cortical neuritic organization and in the development of cortical associative and projective connections (6, 7).

White matter lesions also induce death of oligodendrocytes with subsequent myelin defects (8). We previously demonstrated that, after completion of neuritic genesis, the late murine germinative zone produces most of the astrocytic precursors destined for the upper half of the neocortex (9). Studies on glial genesis in human fetuses (10, 11) suggest that the germinative zone of the human fetus and premature newborn produces glial cells that migrate to the cortex after the end of neuritic migration. In premature newborns with PVL or PHI, migration of astrocytic precursors destined ultimately for the upper layers of the neocortex is disrupted by the white matter lesions. Appropriate astroglial equipment is a requisite for neuron survival and differentiation, as demonstrated by Delanay et al. (12), who showed that selective ablation of astrocytes in newborn mouse cerebellum dramatically impairs the growth and maturation of the cerebellar cortex.

Crawford and Hobbs (2) have proposed that motor paralysis in preterm infants may be caused by selective degeneration of corticospinal tracts consecutive to a deficiency in an unidentified trophic factor specific for corticospinal neurons. The developing brain produces various classes of growth factors that differ in their cell specificity [for a review, see Loughlin and Fallon (13)]. On the other hand, the maternal circulation provides not only nutrients for the developing fetus but also hormones and growth factors such as insulin (14), thyroid hormones (15), granulocyte colony-stimulating factor (16), transforming growth factor-β (17), and VIP (18). Transplacental transfer of several of these factors is probably limited to the early stages of brain development, and the fetus produces some of these factors at later stages of development. However, premature delivery may abruptly cut off the maternal supply of molecules that are not yet produced in sufficient amounts by the fetus.

VIP is present in the human maternal and fetal circulation throughout gestation (19) and can cross the rodent placenta (18). Furthermore, VIP plays a key role in promoting neocortical astrocyte genesis by the late murine germinative zone (20). If these data from animal models apply to humans, premature delivery may deprive the human newborn of maternal VIP, leading to neocortical astrocyte depletion with potential deleterious consequences for neuritic survival and differentiation.

The above data and hypotheses prompted us to study in the murine neocortex the effects on neuritic survival and differentiation of VIP blockade during astrocyte production by the late germinative zone.

METHODS

Drug administration and brain histology.

Swiss mice were housed in groups and fed laboratory chow and water ad libitum. Experimental protocols were approved by our institutional review board and complied with guidelines of the Institut National de la Santé et de la Recherche Médicale (INSERM). Pregnant mice were injected intraperitoneally twice daily (0800–0900 h and 1800–1900 h) on E17 and E18 with 200 μL of PBS and 50 μg of VA; controls received PBS alone. VA is a neurotensin-VIP hybrid and a specific VIP antagonist with no agonist activity (21–23). Previous studies demonstrated 30% neuritic loss in neural cultures exposed to VA (22). Injection of VA into pregnant mice at the early stages of brain development induced severe microcephaly (24). Administration of VA to neonatal animals caused damage to cerebral cortical neurons (25) and delayed the acquisition of developmental milestones (26). As previously mentioned, prenatal administration of VA inhibited neocortical astrocyte genesis (20). Cotreatment with neurotensin did not influence the biologic activity associated with VA (26).

Animals treated in utero with VA or PBS were killed by decapitation on P2, P5, P12, or P45 (adults). Several litters of mouse pups of both sexes were used for the experiments. The sex ratios of animals examined at adulthood were similar in the two experimental groups. After sacrifice, the brains were dissected out, weighed, and either fixed in formalin (cresyl violet staining) or immediately frozen at −80°C (immunohistochemistry). Coronal serial sections were used for cresyl violet staining (paraffin sections, 15 μm thick) and for antigen detection (cryostat sections, 20 μm thick). To avoid regional variations, analyses were performed on the somatosensory cortex. Neuronal layering and morphology were studied in both experimental groups at P5, P12, and P45 (n = 4–8 in both experimental groups, at each studied age). As an evaluation of total cell density, cresyl violet-stained nuclei were counted in eight nonadjacent 0.0625-mm2 fields (four brains in each group) in the paramedian neocortex of P45 control pups and P45 VA-treated pups.

Immunohistochemistry and lectin staining.

After fixation with methanol and acetone at −20°C, cryostat sections were reacted with polyclonal rabbit antiserum to GFAP (Dako, Glosstrupp, Denmark) or mouse MAb directed against MAP-1 (Sigma Chemical Co., St. Louis, MO), MAP-2 (Sigma), MAP-5 (Sigma), NF 68 kD (Sigma), NF 160 kD (Sigma), or synaptophysin (Boehringer Mannheim, Meylan France), diluted 1/500, 1/200, 1/1000, 1/500, 1/100, 1/20, or 1/50, respectively, in PBS containing 1% normal goat (GFAP) or horse (MAb) serum. Sections were incubated for 1 h at room temperature. Avidin-biotin horseradish peroxidase kits (Vector, Burlingame, CA) were used as directed to detect rabbit or mouse antibodies. Diaminobenzidine (Sigma) served as the chromogen.

GFAP-positive cells were counted in 0.025-mm2 areas located in the paramedian upper (layers I to IV) neocortex (on P5, P12, P20, and P45) and the paramedian deep (layers V and VI) neocortex (on P5). At each age studied, four to 12 animals were included in each experimental group, and three to five sections were examined per animal. Sections from control and VA-treated animals were processed simultaneously.

To avoid regional and experimental variations in labeling intensity in nonquantitative immunohistochemical analyses, sections from VA-treated and control groups including comparable anatomic regions were treated simultaneously. Qualitative analysis was performed by two investigators working independently and was focused on the somatosensory neocortex and underlying white matter. For each animal, several sections were immunoreacted in successive experiments. On the basis of an analysis of pairs of sections including similar anatomic regions from VA- and PBS-treated animals, staining differences between the two experimental groups were scored independently by the two observers as doubtful, mild, or obvious. Only obvious differences consistently observed throughout the experimental material by both investigators were considered significant. This procedure minimized the potential subjective bias inherent in qualitative analysis of MAP, NF, and synaptophysin stains. Eight to 16 nonadjacent sections were examined for each antibody on P12 or P45 in both experimental groups.

For the study of brain vessels, cryostat sections from the frontoparietal cortex were incubated overnight at room temperature with biotinylated Griffonea simplicifolia lectin I isolectin B4 (Vector), diluted 1/400 in PBS. The probe was revealed using an avidin-biotin horseradish peroxidase kit (Vector). Eight nonadjacent sections from both experimental groups (four brains in each group) were analyzed on P5, P12, and P45. Qualitative analysis of lectin staining was performed as described above for MAP, NF, and synaptophysin immunodetection.

Cell death (fragmented DNA).

Cell death was detected using an in situ cell death detection kit (Boehringer Mannheim). After methanol-acetone fixation, cryostat sections were treated for 20 min at 37°C with 20 mg/mL proteinase K and incubated for 2 min on ice with 0.1% Triton X-100. DNA strand breaks were identified by labeling, for 60 min at 37°C, of free 3′-OH termini with terminal deoxynucleotidyl transferase (TdT) and fluorescein-labeled nucleotides. Incorporated nucleotides were detected using an anti-fluorescein antibody conjugated with alkaline phosphatase, using nitroblue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate toluidonium salt as the substrates. TUNEL-positive cells were counted in a 0.01-mm2 area in the deep neocortex (layers V and VI) of the paramedian somatosensory area. On P2, P5, and P12, 14–20 nonadjacent fields from four to eight brains were studied in each group. To further characterize the dying cells, other sections from P2 pups were first reacted for GFAP, MAP-1, or MAP-2 immunohistochemistry as described above, then treated for TUNEL detection. To increase the contrast between chromogens, Vector VIP peroxidase substrate (Vector) was used instead of diaminobenzidine.

Dot blot analysis of synaptophysin and NMDA-R1 receptor subunitmRNA.

To confirm data obtained with anti-synaptophysin antibody, dot blot analysis was performed to quantitate mRNA for synaptophysin (presynaptic component) and NMDA R1 receptor subunit (postsynaptic component) in VA-treated and control adult cerebral hemispheres. On P45, animals (control group, 11 females and 12 males; VA-treated group, 10 females and 11 males) were killed by decapitation, and the cerebral hemispheres were promptly frozen at −80°C until analysis. Tissues were homogenized, and total RNA was extracted using the Bio/RNA-Xcell kit (Bio/Gene Limited, Kimbolton Cambs, UK). Each sample (30 μg total RNA/well) was transferred in triplicate to positively charged nylon membranes (Boehringer Mannheim) to permit separate hybridization with the three different probes (see below). UV-cross-linked membranes were incubated overnight at 37°C in the presence of 5 μL/15 mL of a DIG-dUTP-labeled oligoprobe diluted in DIG Easy Hyb (Boehringer Mannheim). Oligoprobes (synaptophysin, ACT CGA ATT CAA CTT CGA TGT TGA GGG CAC TCT CCG TCT TGT TG; NMDA R1, GTA GTG GCG TTG AGC TGT ATC TTC CAA GAG CCG TGT CGC TTA TT; cyclophilin, TGC CAT CCA GCC ATT CAG TCT TGG CAG TGC AGA TAA AAA ACT GG) were end-labeled with DIG-dUTP using the DIG oligonucleotide tailing kit (Boehringer Mannheim). Probe specificity was confirmed by Northern blot analysis of total adult mouse brain mRNA, which showed, for each probe, a single band of the expected molecular size (not shown). After hybridization, membranes were washed at 42°C for 30 min in 2× SSC containing 0.2% SDS and 50% formamide. Subsequent washes and detection of DIG-dUTP with anti-DIG-alkaline phosphatase complex were performed using the DIG wash and block buffer set and the DIG nucleic acid detection kit (Boehringer Mannheim), respectively. After incubation with CSPD (Boehringer Mannheim) at 37°C for 10 min, membranes were exposed against x-ray films (Kodak) for 1 h to detect chemiluminescent signals. The OD of x-ray films was measured using a photometer and, for each sample, the NMDA R1/cyclophilin and synaptophysin/cyclophilin ratios of OD were calculated.

Glucose consumption.

A femoral artery and vein were catheterized with polyethylene tubing under light halothane anesthesia. Both catheters were threaded under the skin, up to the back of the hindpaw to allow free access to the catheters without disturbing mouse movements. The animals were allowed to recover from surgery in their home cages for 17–24 h before the onset of the experiment. Autoradiographic experiments were performed on six control and six VA-exposed P55 mice. Rates of cerebral glucose utilization were measured by the 2DG technique (27) adapted to small animals (28). The 2DG (4.625 MBq/kg, specific activity 1.65–2.04 GBq/mmol, Isotopchim, Ganagobie-Perhuis, France) was injected as an i.v. pulse to freely moving mice. Timed arterial blood samples were drawn during the subsequent 45 min for the measurement of plasma glucose and 2DG concentrations. Approximately 45 min after the injection of the tracer, the animals were killed by decapitation. The brains were rapidly removed, frozen, and cut into 20-μm coronal sections. Sections were autoradiographed on Amersham Biomax MR film along with [14C]methylmethacrylate standards calibrated for their carbon-14 concentration in brain tissue. The autoradiographs were then digitized and analyzed by densitometry using an image-processing system (Biocom 500, Les Ulis, France). The localization of specific nuclei was assessed on adjacent sections stained with cresyl violet. Local cerebral metabolic rate for glucose (LCMRglcs) were calculated according to the operational equation of the 2DG method (27). Mean arterial blood pressure, blood pH, PO2, PCO2, and hematocrit were recorded throughout the experimental procedure and were identical in both experimental groups (data not shown).

Histologic analysis of maternal tissues and placentas.

Pregnant mothers treated as described above with PBS (n = 3) or VA (n = 3) were killed at E18, 5 h after the last intraperitoneal injection. Maternal heart, gut, kidneys, and liver, as well as placentas, were rapidly fixed in formalin and embedded in paraffin. Ten-micrometer-thick sections were stained with Masson's trichrome staining.

Statistical analysis.

VA and control quantitative values were compared at matching ages using an unpaired t test. Results were expressed as mean ± SD.

RESULTS

Fertility, litter size, and pup survival (<2% of postnatal mortality in live-born pups) were similar in the VA-treated and control groups. Although no specific test was performed, VA-treated animals did not display any markedly abnormal behavior as compared with controls.

Brain growth and cytoarchitecture.

Neocortical cytoarchitecture was similar in both experimental groups at all ages studied (Fig. 1, A and B). Cell density measured on P45 in the paramedian frontoparietal cortex was not modified by prenatal VA treatment (99.7 ± 5.5 cells/0.0625 mm2 in controls versus 100.4 ± 7.7 cells/0.0625 mm2 in the VA group). When compared with age-matched controls, a moderate reduction in brain weight was observed in P12 pups prenatally treated with VA (Fig. 1C), whereas brain weight was unchanged in P5 and P45 VA-treated pups.

Figure 1
figure 1

Effects of prenatal VA treatment on neocortical cytoarchitecture and brain weight. A, P45 control frontoparietal neopallium. B, Similar brain area of an age-matched animal treated in utero with VA. Cresyl violet–stained coronal sections. Bar = 40 μm. C, Brain weight of controls (closed bars) and VA-treated animals (open bars) during postnatal development. Values shown are mean ± SD. **p < 0.001 vs age-matched controls.

Full size image

Cell death.

In P2 and P5 controls, several neocortical nuclei were stained by the TUNEL technique, which labels dying cells containing fragmented DNA (Fig. 2, A and C). Most of these labeled cells were in the deep neocortex (layers V and VI) and underlying white matter, and a few were seen in the upper neocortex (layers I to IV). No labeled dying cells were detected in P12 control neocortex. In prenatally VA-treated animals, cell death was significantly increased on P2 and P5, when numerous dying cells were seen in the deep neocortex (layers V and VI) and underlying white matter (Fig. 2, B and C). As in the controls, no labeling was observed on P12 in VA-treated animals.

Figure 2
figure 2

Typical effects of prenatal VA treatment on cell death in the somatosensory neopallium on P5 in control (A) and VA-treated (B) animals. Arrowheads point to examples of nuclei labeled by the TUNEL technique. Bar = 40 μm. C, Quantitative analysis of TUNEL-positive cell density in the somatosensory deep cortical layers of controls (closed bars) and VA-treated animals (open bars) during postnatal development. Values shown are mean ± SD. ***p < 0.001 vs age-matched controls.

Full size image

Double-labeling experiments showed that >95% of TUNEL-positive cells were also stained with anti-MAP-1 or anti-MAP-2 antibodies (Fig. 3) but not with anti-GFAP antiserum.

Figure 3
figure 3

VA pretreatment induces neuritic death. MAP-1 (A) or MAP-2 (B, C) (red deposit in the perikaryon; peroxidase detection) and TUNEL (purple deposit in the nucleus; alkaline phosphatase detection) double labeling in the deep cortical layers. Arrowhead points to endogenous alkaline phosphatase activity present in blood cells). Bar = 10 μm.

Full size image

Neurite genesis and synaptogenesis.

MAP and NF stains were used as axonal and dendritic markers of the neocortex. On P12 and P45, VA treatment did not modify the intensity or distribution of immunostaining with antibodies directed against MAP-1 or NF 68 kD (data not shown). In contrast, P12 and P45 VA-treated animals exhibited stronger NF 160 kD staining throughout the whole neocortical plate (Fig. 4, A and B) and increased MAP-2 and MAP-5 immunostaining in the upper neocortex (layers I to IV) and in the molecular layer (layer I), respectively (Fig. 4, C–F).

Figure 4
figure 4

Typical effects of prenatal VA treatment (controls, A, C, E; VA-treated animals, B, D, F) on NF 160 kD [A, B : note the more intense labeling in the white matter (arrows) and the increased density of heavily labeled structures (arrowheads) in the VA-treated cortical plate], MAP-2 [C, D : note the increased density of heavily labeled structures (arrowheads) in the VA-treated cortical plate], and MAP-5 [E, F : note the increased density of labeled structures (arrowheads) in the VA-treated molecular layer] immunostaining in the somatosensory cortical plate on P45. Bar = 40 μm.

Full size image

Synaptophysin was used as a marker of synaptic density. On P12 and at adulthood, synaptophysin immunostaining was increased throughout the neocortical plate in VA-treated animals, being most marked in the molecular layer (Fig. 5, A and B). To confirm these data by quantitative measurements, dot blot analysis of mRNA for synaptophysin and for NMDA R1 receptor subunit was performed on adult brains. Results showed that VA treatment significantly increased both synaptophysin and NMDA R1 mRNA (Fig. 5, C and D).

Figure 5
figure 5

Effects of prenatal VA treatment on neocortical synaptogenesis. Immunodetection (arrowheads point to examples of labeled structures) of synaptophysin in the somatosensory molecular layer of control (A) and VA-treated (B) adult animals: note the increased density of labeled structures in B. These data are representative of the results consistently obtained in the current study. Bar = 40 μm. C, Typical dot blot analysis of synaptophysin, NMDA R1, and cyclophilin mRNA extracted from VA-treated (1, 3) and control (2, 4) adult brains. D, Quantitative analysis of synaptophysin/cyclophilin mRNA and NMDA R1/cyclophilin mRNA ratios in control adults (closed bars) and VA-treated adults (open bars). Values shown are mean ± SD. ***p < 0.001 vs age-matched controls.

Full size image

Astrocyte genesis.

As previously described (20), VA treatment induced a dramatic postnatal depletion of GFAP-positive cells in the upper neocortex (layers I to IV;Table 1). The decrease in astrocyte density was observed on P5 and P12 but not on P45. Astrocyte counts in the deep neocortex (layers V and VI) were not modified by VA treatment. On P5, most neocortical astrocytes were bipolar or exhibited only a few ramified processes in VA-treated brains, contrasting with the numerous ramified processes seen in the controls (Fig. 6).

Table 1 Quantitative analysis of GFAP-labeled cell density in the frontoparietal cortical plate of control and VA-treated animals during postnatal development Values are mean ± SD;t test comparison vs age-matched controls. Abbreviations: UCL, upper cortical layers (I to IV); DCL, deep cortical layers (V and VI). (Data derived from reference (20).
Full size table
Figure 6
figure 6

Typical effects of prenatal VA treatment (A, control;B, VA-treated animal) on P5 astrocyte genesis in the upper layers of the somatosensory neocortex, showing an increased density of labeled cells in B. GFAP immunostaining (arrowheads point to examples of labeled cells). Bar = 20 μm.

Full size image

Angiogenesis.

Griffonea simplicifolia lectin I isolectin B4 was used to label neopallial vessels. The distribution, density, and morphology of labeled vessels was similar in both experimental groups on P5 and in adults. In contrast, on P12, VA treatment induced a moderate but reproducible increase in blood vessel branching in the upper neocortical layers (Fig. 7).

Figure 7
figure 7

Representative effects of prenatal VA treatment on P12 neocortical vessels. Griffonea simplicifolia lectin I isolectin B4 staining (arrowheads point to examples of labeled structures) in the upper layers of the somatosensory neocortex in control (A) and VA-treated (B) animals: note the increased density of vascular ramifications in B. Bar = 40 μm.

Full size image

Glucose consumption.

In adult control mice, glucose utilization rates were within the normal range and were similar to previously reported values (29, 30). Prenatal VA treatment failed to change glucose utilization rates in cortical areas or white matter (Table 2).

Table 2 Quantitative analysis of glucose consumption in P55 neocortex Values, expressed as μmol/100 g/min, represent mean ± SD of 5 animals in each group;t test comparison vs controls. Abbreviations as in Table 1.
Full size table

Analysis of maternal organs and placentas.

Histologic exam of maternal tissues (kidney, heart, liver, and gut) and placentas from animals treated with VA did not show any sign of cell death, inflammation, or vascular modification when compared with control organs (data not shown).

DISCUSSION

Our most salient finding is that prenatal VIP blockade during the critical period of neocortical astrocyte genesis induced delayed astrocyte genesis, altered neonatal neuritic death, and prolonged overexpression of some axonal-dendritic and synaptic markers.

Although we cannot fully exclude an indirect effect of VA treatment through hemodynamic, ischemic, or possibly toxic effects on maternal systems, several lines of evidence indicate that VA was directly acting on the embryonic brains. E17–E18 murine neopallium displays VIP binding sites (20). Radiolabeled VA has been found to reach the brains of E15 to E17 rat embryos after either intraperitoneal (0.005% reached the embryonic brain) or intrauterine (0.01% reached he embryonic brain) treatment of the mother (Gozes I, Liling G, Reshef A, unpublished data). Similarly, radiolabeled VIP administered to pregnant rats or mice was found intact in the embryonic brains (18). Prenatal administration of VA between E9 and E11, but not at later stages, reduced brain and body weights of the embryos (24): this marked developmental sensitivity argues against VA producing indirect effects on the embryo through the maternal systems. Furthermore, pregnant mothers treated with VA did not show any sign of toxicity [(20) and current study], and placentas of treated embryos did not reveal ischemic changes. Finally, administration of VA to newborn pups resulted in an inhibition of neocortical astrocyte genesis, mimicking the effects observed after VA administration to pregnant mothers (31).

Increased cell death in the deep cortical layers, as detected using the TUNEL technique, reflects neuritic loss, because astrocyte counts were modified in the upper layers but not in the deep neocortical layers. Double labeling with the TUNEL method and immunohistochemistry for GFAP or MAP confirmed that most of the dying cells displayed a neuritic phenotype. Astrocytes promote neuritic survival by a variety of mechanisms. The poor astrocyte differentiation induced by VA treatment may reflect immature astroglial function responsible for insufficient neuritic support. Furthermore, VIP is a secretagogue for astrocytes, which release neurotrophic factors including protease nexin I (32), activity-dependent neurotrophic factor (33), brain-derived neurotrophic factor (34), and cytokines (35). VA treatment may have impaired astroglial release of these neurotrophic agents. On the other hand, VIP binding sites have also been described on neocortical neurons around birth (36), suggesting that early postmigratory neurons may be directly responsive to VIP or VIP-related molecules. A large body of evidence [for a review, see Voyvodic (37)] points to programmed cell death as a critical developmental step in normal brain development. Indeed, programmed cell death serves multiple functions including reduced competition for growth factors, elimination of abnormal cells, and promotion of neuritic diversity by elimination of initially produced redundant cells. The exact consequences on mature brain function of exaggerated neocortical cell death during the neonatal period remains to be determined by behavioral studies.

The moderate and transient, but significant, reduction in brain weight observed on P12 in VA-treated animals probably reflected the increased neuritic cell death and reduced astrocyte density in the neocortex. The subsequent normalization of brain weight could be caused by changes in extracellular matrix and water content and by the observed catch-up of astroglial production. This normalization of astrocyte density in the cortical plate within a few postnatal weeks has been previously observed in other models of prenatal inhibition of cortical astrocyte genesis (9, 38). As an alternate hypothesis to an increased neuritic death, the enhanced density of TUNEL-positive cells in VA-treated pups could reflect a prolonged clearance time of dying cells caused by the depression of astrocytic function. In this context, fluctuations of brain weight in VA-treated animals would be secondary to changes in astrocytic density rather than to modifications of neuritic cells.

The combination of data obtained by immunohistochemistry and dot blot analysis in our study strongly supports the hypothesis that prenatal blockade of VIP or VIP-related molecules induces up-regulation of some neuritic and synaptic molecules. The enhanced MAP-5 and NF 160 kD stainings and the unchanged pattern of other neuritic markers suggest an abnormal neuritic differentiation rather than a modified number of neuritic processes. The increased density of synaptophysin and NMDA receptors may reflect either an increased density of synapses or an increased density of synaptic molecules per synapse. The abnormalities in synaptic and neuritic labeling may have been caused by several mechanisms:1) the predominance of neuritic and synaptic alterations in the superficial neocortical layers where astrocyte depletion occurred after VA treatment suggests that the VA-induced alterations in astrocyte genesis contributed to these observed neuritic abnormalities; similar findings were reported in the cerebellum after astrocyte depletion by genetic manipulation (12);2) alterations in synaptic and neuritic equipment may be an adaptive mechanism used by the developing brain to cope with the excessive neuritic death occurring in the neonatal cortex; similar compensatory mechanisms affecting synapses and neurites have been reported in other models of neuritic cell death (39–42); and 3) a direct effect of VA on developing axons and synapses cannot be excluded inasmuch as VIP modulates in vitro differentiation of sympathetic neurons (43) and neuroblastoma cells (44).

Synaptic transmission is generally considered the main source of brain glucose consumption (45). The absence of modifications in brain glucose consumption in our VA-treated animals may give rise to a number of explanations:1) as suggested above, VA treatment may have produced not an increase in synapse density, but rather an increase in the density of synaptic molecules within each synapse, an effect that would not change synaptic energy requirements;2) an increase in synapse density would not necessarily lead to a detectable increase in glucose consumption under basal conditions, and in this regard a study of glucose consumption in a dynamic behavioral paradigm would perhaps allow the detection of metabolic differences between the two experimental groups; and 3) additional synapses in VA-treated adults may not all be functional or recruited simultaneously, and therefore may not influence total glucose consumption. Interestingly, early formed synapses with NMDA receptors seem to account for the electrophysiologically “silent synapses” observed in adult animals (46). The normal brain glucose consumption in the adult animals treated prenatally with VA is consistent with the absence of vascularization changes in these animals compared with the controls. In contrast, the moderate increase in vessel branching observed in P12 superficial cortical layers after VA treatment coincided with the emergence of abnormal cortical neuritic and synaptic patterns and may reflect a need for an increase in blood supply to build additional synapses and abnormal neurites.

VIP shares considerable amino acid sequence homology and a number of receptors (called VPAC1 and VPAC2) with PACAP [for a review, see Arimura (47)]. Recent data demonstrating that developmental VIP effects were not replicated by PACAP strongly suggest the existence of a specific VIP receptor (or receptor subtype) not recognized by PACAP (24, 48, 49). We previously showed that VA-induced inhibition of neocortical astrocyte genesis was mediated by the common VPAC2 receptor (20). The data from the current study do not allow us to determine which types of VIP receptors were involved in increased cell death and abnormal neuritic differentiation. Either VIP or PACAP, or both, may have mediated the physiologic effects antagonized by VA treatment.

As previously mentioned, premature delivery is potentially responsible for an abrupt loss of the maternal supply of molecules that are not yet produced in sufficient amounts by the fetus. The potential resulting deficiency in trophic factors may be further exacerbated by the malnutrition and hypercatabolism generally observed in sick premature newborns. VIP is present in human maternal and fetal circulation throughout gestation (19). Interestingly, the VIP concentration in the human umbilical vessels has been shown to be about 2.5 times greater than in maternal venous blood, suggesting that VIP is selectively concentrated in the uterine circulation (50). Furthermore, radiolabeled VIP injected to rats and mice during early pregnancy is found intact in the brain of embryos, suggesting that transplacental transfer of maternal VIP occurs during gestation (18). Although no data are available on the production of VIP by the human fetal brain, only small amounts of VIP are produced in monkey, rat, and mouse brains at stages corresponding to premature delivery in humans (25, 51, 52). No data are currently available on PACAP circulating concentrations or the ability of PACAP to cross the placenta. PACAP is first detected in rat upper neocortical layers on E18 and its concentration increases during the first 2 postnatal wk [(53), Gressens P, Hill JM, unpublished data]. No studies of PACAP production in human fetal brains are available. Additional investigations will have to determine the relative importance of neural cells producing VIP and PACAP versus circulating VIP, and further work is required to correlate the risk of neurologic impairment with circulating VIP levels in premature infants.

In conclusion, we found that prenatal blockade of VIP or PACAP during a critical period of neocortical development decreased neuritic survival and induced long-term overexpression of some markers of neuritic differentiation. As a working hypothesis, we propose that VA-induced abnormal neocortical genesis involves several steps:1) VA delays neocortical astrocyte genesis;2) the neonatal depression of astrocytic function leads to abnormal neuritic cell death in the early postnatal period; and 3) although the mechanism remains unclear, the combination of VIP blockade, impaired astrocytic function, and altered cell death induces the overexpression of several neuritic and synaptic markers in the adult neocortex. Further behavioral studies are needed to determine the functional impact of these abnormalities of neocortical ontogenesis. Finally, we suggest that the mouse model used in our study may provide new working hypotheses for explaining some of the neurodevelopmental impairments observed in human premature infants.