Neurodevelopmental Perspectives on Schizophrenia

Daniel R. Weinberger

INTRODUCTION

A dramatic conceptual shift in thinking about the neurobiology of schizophrenia has taken place since the previous volume in this series. After many decades of speculation that schizophrenia occurs because of cerebral pathological events that happen or are expressed around early adult life, it has recently become de rigeur to refer to schizophrenia as a neurodevelopmental disorder in which the primary cerebral insult or pathological process occurs during brain development long before the illness is clinically manifest (14, 20, 52, 71).

The reasons for this shift are primarily threefold: scientific data confirming adult onset pathological cerebral changes have remained elusive; replicable evidence implicating cortical maldevelopment has emerged; and neurobiological models that may explain the relationship between such maldevelopment and the clinical features of the illness exist. This chapter reviews several issues underlying this conceptual reorientation of schizophrenia (see Molecular and Cellular Mechanisms of Brain Development and The Development of Brain and Behavior for background).

An association between putative abnormalities in intrauterine development and schizophrenia has been reported throughout this century. The evidence ranges from highly circumstantial and weak (e.g., slight overrepresentation of minor physical anomalies) to compelling (e.g., replicated cytoarchitectural anomalies). When viewed in total, the evidence might be interpreted to weave a coherent story of developmental abnormalities; however, individual findings show many inconsistencies and methodological problems. The most important research questions at present are whether evidence of cytoarchitectural deviations can be more widely replicated and whether lingering methodological uncertainties can be resolved. If these questions are answered positively, most of the research controversies and theoretical speculation considered below (e.g., obstetrical complications, pruning defects) will be irrelevant, because a "smoking gun" will have been found.

Because abnormal intrauterine events might be expected to affect the development of extracerebral tissues, reports of somatic morphological variations are of potential relevance. The spectrum of potential minor physical anomalies (MPA) and their relationship to abnormal brain development is not clearly defined. In the schizophrenia literature, there is no consistent pattern of associations, and often the samples have to be broken down post hoc into putative subgroups for any associations to be found (68). Many different anomalies have been reported, including high palate, low set ears, webbed digits, and variations in limb length and angle, finger print patterns and ridge counts. Proponents of the relevance of MPAs have stressed that many of these abnormalities are seen as a result of second-trimester insults, timing that is consistent with a putative defect in neuronal migration. They also stress that MPAs are seen in other neurodevelopmental disorders such as intrauterine viral encephalopathies and in other psychiatric disorders of presumed developmental origin (e.g., autism) (68).

However, the circumstantial nature of these arguments does not obscure problems with the fundamental data base. The true frequency of MPAs in patients with schizophrenia is not known, as large unselected samples of patients and well-matched controls have not been studied. Whether all the morphological characteristics reported are actually pathological is also uncertain. Moreover, few of the studies have used the same methods of assessment, and it is uncertain if any of the studies were truly blind comparisons. The studies that report increases in MPAs tend to lump them all together as if each signified the same thing, when, in fact, their relative frequencies seem not to correlate with each other (68). Because the search for MPAs has not been a centerpiece of schizophrenia research, it is doubtful that investigators who failed to find MPAs in their samples reported the negative results. Another uncertainty is that the timing of events that cause MPAs is not well established, and considerable variability exists. In light of these questions and other contradictory data [e.g., no difference in MPAs within pairs of discordant monozygotic twins (68), no consistent abnormalities in birth weights of schizophrenic births compared with healthy sibling births (48!popup(ch110), and no relationship of MPAs to birth weight or other clinical signs of intrauterine adversity (48, 68)], the relevance of MPAs to understanding potential cerebral maldevelopment in schizophrenia is weak.

If the brains of patients with schizophrenia have not developed normally, it might be expected that some evidence of subtle abnormalities of neural function would be apparent during their childhood before they become clinically ill. Several lines of circumstantial data support this possibility. Studies of premorbid neuropsychological test performance and school achievement have tended to report that individuals who later manifest schizophrenia did worse than their healthy siblings. Reports of no difference have also appeared (68). Four series of reports of high-risk subjects, that is children with at least one schizophrenic parent, have found various abnormalities of motor function (31, 51), autonomic responsivity (51), and attention (6, 27). The abnormalities vary somewhat from one study to another and are not linked with schizophrenia per se. Fish et al. (31) have provided a detailed description of what they refer to as a pandysmaturational defect in high-risk children, consisting of gross abnormalities of gait, posture, muscle tone, and reflexes in early childhood. It is uncertain how representative these relatively dramatic findings are of most children who subsequently manifest the illness. Moreover, the findings disappear to some degree over time, further complicating their interpretation. Finally, Walker and colleagues (70) recently reported a novel study of home movies of families having a child who later developed schizophrenia. In a blind comparison of affected and unaffected siblings in four sibships, they reported clear differences in bimanual dexterity, in gait, and in other gross motor functions that allowed them to invariably identify the affected family member even in the first few years of life. To clinicians, the severity of the observations of Fish and the severity and consistency of the observations of Walker et al. may seem surprising. Clearly, further replication is needed. Nevertheless, the results of these studies, although open to other interpretations, are consistent with the possibility of brain-maldevelopment.

Beginning in 1976 with an in vivo study of cerebral ventricular size using computerized axial tomography (CAT) scanning (63), literally hundreds of reports have appeared of subtle variations in cerebral anatomy associated with schizophrenia. The CAT data base established that as a group patients with schizophrenia have slightly enlarged ventricles and wider cortical fissures and sulci (63). These findings suggested a nonfocal reduction in cerebral and probably cortical volume. Moreover, the data did not support the widely expressed assumption that morphological abnormalities characterized only a subgroup of patients. Instead, they seemed to describe a continuum of pathological change (22, 63) that characterized the majority of the patient population. Recently, using magnetic resonance imaging (MRI), which produces images of much greater contrast and resolution, the CAT data have been confirmed and refined. Evidence of cortical volume loss, most frequently replicated in but not confined to mesial temporal cortex in the area of the rostral hippocampus, has been a relatively consistent finding (74). Again, the data do not implicate only a subgroup of patients (66, 74).

In some studies, evidence of pathological changes appears to favor the left side of the brain. This has been especially true for the size of the ventricles (20) and for the volume of the superior temporal gyrus (64). Reports of relatively greater reductions in other regions of the left temporal cortex have also appeared (66). Indeed, when asymmetric findings have been reported, they usually involve the left hemisphere. This has prompted some to suggest that this lateralization tendency is consistent with a putative delay in the development of the left hemisphere during the second trimester, leaving it more vulnerable to adverse events that might otherwise affect the brain diffusely (20). This is an interesting hypothesis, but it depends on a number of assumptions and probably overinterprets the morphometric literature. Whether the slight delay in the appearance of surface gyri means that the left hemisphere is developing slower or that it is more vulnerable to injury or vulnerable for a longer period is unknown. Moreover, most of the morphometric studies, even those that report unilateral findings (20, 64, 66), have observed bilateral changes as well.

Although in vivo evidence of subtle morphometric deviations might implicate a neuropathological process as being related to schizophrenia, it does not by itself implicate a neurodevelopmental one. However, beginning with the results of the second CAT study of schizophrenia, the correlative data have not been consistent with what would be expected of a degenerative condition of adult onset and have been consistent with what would be expected of a neurodevelopmental one. In particular, in the majority of cross-sectional studies, ventricular size has been found not to correlate with duration of illness (63), as would have been expected if the neuropathological process responsible for ventricular enlargement advanced as the illness progressed. More recently, lack of progression has been confirmed by prospective studies that have followed the same individuals for up to 10 years from the first psychotic episode (38). Thus, it appears that, for most patients, the pathological process responsible for the in vivo morphometric changes associated with schizophrenia has arrested at least by the time their clinical illness is diagnosed.

Circumstantial and correlational evidence that the pathological process might have arrested early in life if not during early development also emerged from in vivo imaging studies. Ventricular enlargement has been found in most studies of patients at the onset of the clinical illness in early adulthood (15, 24, 63) and reduced hippocampal volume has been observed in at least one first-break study (15). This suggests that these findings do not develop during the early phases of illness and probably predate the onset of the illness. Several, although not all, studies have found an unexpected correlation between ventricular enlargement in patients with schizophrenia and poor premorbid social and educational adjustment during early childhood (26, 40, 63). This finding further suggests the possibility that the morbid pathology had occurred early in life and manifested itself in different ways at different times of life.

Data from postmortem morphometric studies of brains of patients who died with schizophrenia are generally consistent with the conclusions from the in vivo studies (63). In general, volumetric assessments of cerebral ventricular size (20, 55), of various cortical regions, the hippocampal formation (including the parahippocampal cortex), and of various periventricular subcortical nucleii have found differences between patients and normals (3, 39, 55, 63). Likewise, neuronal counts have been reported to be reduced in patients in selected cortical and periventricular regions (12, 20, 29). These studies, like the in vivo studies, implicate a fairly widespread neuropathological process. It should be noted, however, that there have also been negative reports (35), and the reasons for the inconsistencies are unclear.

One observation that has been consistent is lack of gliosis. In fact, it was the absence of gliosis that prompted many classical neuropathologists during the first half of the twentieth century to dismiss reports of neuropathological changes in schizophrenic brain specimens. Recent studies, whether in neocortex (60, 63), hippocampus (60), or parahippocampal cortex (29), have not found evidence of either acute or chronic gliosis. Because proliferation of glial cells is seen in most degenerative brain conditions and encephalopathies that arise after birth, this negative result would seem more consistent with neuropathological events that predate the responsivity of glial cells to injury, which is before the third trimester of gestation.

It is also interesting to note that, as in the in vivo morphometry data, correlational results are not consistent with a neuropathological process that progresses once the illness has manifested. Ventricular size and cortical volume have not been correlated with duration of illness (20, 55). Pakkenberg (55) reported, moreover, that ventricular size did not correlate with cognitive deficits noted in the hospital records at the time of death, but instead, correlated with those noted at the time of onset of the illness. This curious finding is more consistent with an underlying pathological defect that set the neurobiological stage for an illness rather than one that evolved over the course of the illness.

The possibility that normally lateralized aspects of the brain might be anomalous in patients with schizophrenia has come from several experimental directions. Studies of lateralized cerebral function, such as handedness, dichotic listening asymmetries, and lateralized cognitive tasks, have suggested that patients with schizophrenia may be less completely lateralized than normal individuals (20). If these functional asymmetries are related to mechanisms of the development of normal anatomical asymmetries, the findings may have implications for abnormal cerebral development in schizophrenia. Since the times when many of the well-characterized normal anatomical asymmetries originate are known; the time of developmental disruption might be inferred.

Although some in vivo imaging studies have reported diminished asymmetry of the widths of the occipital lobes, at least as many have failed to replicate this finding (63). Such inconsistency appears to characterize the majority of the other literature on cerebral anatomical asymmetries in schizophrenia. For example, some groups have reported less asymmetry of normally asymmetric perisylvian structures, such as the sylvian fissure (28) and the planum temporale (61), but other groups have failed to replicate these findings (8, 44). Based on a finding of asymmetry of the ventricles seen in two-dimensional x-ray films of postmortem brain specimens, Crow et al. (20) went so far as to hypothesize that schizophrenia is caused by a gene that controls the normal development of temporal lobe asymmetries. A test of this hypothesis in a sample of discordant monozygotic twins, where the abnormalities of asymmetry would be expected in the unaffected twin as well as in the affected twin, has been negative (8).

In summary, notwithstanding the uncertainties about lateralization, the data from morphometric studies, in general, have a considerable degree of internal consistency and lead to the interpretation of neurodevelopmental deviance, most likely occurring before the end of the second trimester. The strongest feature of the morphometric data is its replicability. Moreover, the weight of the evidence does not fit into the model of adult-onset brain disorders. Unfortunately, measurements of the size of various cerebral structures and counts of cell numbers are not conclusive evidence of any neuropathological condition, developmental or otherwise, and could conceivably be accounted for by other explanations (e.g., nonpathological volume changes on in vivo scanning due to dehydration or coincidental neuronal loss or other artifacts unrelated to illness in postmortem studies, which usually involve elderly subjects). The possibility that the development of normal anatomical asymmetries is disrupted in patients with schizophrenia also is inconclusive and does not necessarily implicate a neuropathological condition. The most conceptually unimpeachable evidence comes from qualitative studies of cytoarchitecture.

The laminar patterns of neurons in cortex, the orientation of neurons, and their internal relationships are fixed during the second trimester of birth. It is generally assumed that such relationships and patterns do not change during life, even if cells are lost or if secondary pathological conditions arise. If this assumption is correct, and this is not certain, then abnormalities of cytoarchitecture would strongly implicate pathological development.

Kovelman and Sheibel reported in 1984 (43) that the orientation of hippocampal pyramidal cells in nissl stained sections of left hemispheres from ten patients with chronic schizophrenia was abnormal in most of the specimens as compared with eight normal controls. Conrad et al. (19) subsequently reported the identical finding in the right hemispheres of mostly the same subjects. They interpreted their findings as consistent with a defect in neural migration, arguing that disorientation reflected a fault of neuronal settling into their target sites. This finding has not been independently replicated, and negative studies have been reported (4, 18).

In contrast to the data on pyramidal cell orientation, the data on laminar organization in neocortex and limbic cortex are more consistent, although still far from ideal. Jakob and Beckman (37) made a potentially landmark observation in the entorhinal cortex. In nissl stained sections of the brains of 64 patients with the diagnosis of schizophrenia and 10 controls, they reported that in the majority of ill cases there were cytoarchitectural anomalies of laminar organization. Specifically, they described attenuation of cellularity in superficial layers I and II, incomplete clustering of neurons into normal glomerular structures in layer II, and the inclusion of such clusters in deeper layers where they are not normally found. They studied the rostral entorhinal cortex in the region of the amygdala and pes hippocampus. Interestingly, this is similar to the area of mesial temporal cortex where the most consistent morphometric abnormalities have been found both in vivo and postmortem (3, 15, 29, 39, 66). They interpreted their findings as the result of a failure of cortical development, probably an arrest of migration, whereby relatively recent generation neurons destined for the superficial cortical lamina were held up in deeper layers. Perhaps the most compelling aspect of their observations is that these findings are difficult if not impossible to attribute to an insult that occurred to an already developed brain. The three major uncertainties about their findings are as follows: (a) Are they artifacts of localization within the entorhinal cortex? (b) Are they related to schizophrenia per se? (c) Are they replicable? Other problems with the study included that it was not blind, that the controls were neurologically impaired in nine of the cases, and that the patient population was probably atypical (e.g., mean age of illness onset was 36).

Normal entorhinal cortex anatomy is characterized by remarkable regional variability. In fact, as one moves caudal in the entorhinal cortex, the normal appearance looks increasingly like what Jakob and Beckman reported in schizophrenia. Therefore, it is critical that patients and controls be very carefully examined in the same cytoarchitectonic areas. The possibility that what Jakob and Beckman observed may not be related to schizophrenia per se also must be considered. They subsequently reported the identical abnormalities in four patients with bipolar disorder, although two of these patients had originally been diagnosed with schizophrenia (10).

In spite of these questions, the basic findings of Jakob and Beckman have been independently replicated using the same methods and further supported by other recent data from different approaches to cortical cytoarchitecture. Arnold et al. (5) studied nissl-stained sections of six brains of patients with schizophrenia from the Yakovlev Collection and 16 controls. They observed essentially the same abnormalities as described by Jakob and Beckman and in addition reported anomalous mesial temporal sulci in their specimens. They felt that all six cases were abnormal, but in five the abnormalities were dramatic and unequivocal. None of their controls had similar findings. The authors acknowledged the importance of location within entorhinal cortex and attempted to control for this. Moreover, they reported that as they moved caudally, the differences between patients and controls disappeared, an observation that again corresponds to reports from the morphometric literature. The authors believed that their findings indicated an abnormality of entorhinal cortex development, probably a migration failure, that would render normal neocortical-hippocampal communication impossible. In addition to its small sample size, this study suffered from one other obvious problem. The entire patient sample had undergone prefrontal leukotomies. To control for the possible effect of this, they had included three controls who underwent leukotomy for nonpsychiatric indications (e.g., chronic pain). Although the potential confound of leukotomy cannot be conclusively ruled out by this control group, it is difficult to imagine how the pattern of changes could be explained by this procedure.

Akbarian and colleagues (1) took a different approach to studying cytoarchitecture, but their results suggest the same defect as do those of Jakob and Beckman and Arnold et al. Using a histochemical stain for cortical neurons that express the enzyme, nicotinamide adenine dinucleotide diaphorase (NADPH-diaphorase), a neuronal population said to be remnants of the embryological subplate zone, they studied the superior frontal gyrus region of the dorsolateral prefrontal cortex in the brains of five patients with schizophrenia and five controls matched for age, gender, and postmortem interval before fixation. They found reduced numbers of these neurons in superficial cortical layers I, II, and III and increased numbers in deep layers, especially in subcortical white matter, the putative vestigial subplate neurons. In essence, they observed a qualitative shift in the representation of NADPHdiaphorase positive neurons, as if the younger neurons destined to migrate last from the subplate zone got held up and never made it to their superficial cortical targets. The interpretations of the findings are remarkably similar to those of Jakob and Beckman and Arnold et al.

It is also possible that the underlying defect reflected in the findings of Akbarian and colleagues is the same as that of yet another recent report, a study of nissl sections of frontal cortex by Benes et al. (12). They found decreased numbers of small presumably g-aminobutyric acid-ergic (GABA-ergic) neurons in prefrontal cortex of patients with schizophrenia and larger numbers of pyramidal cells in deeper layers. This finding also might suggest a failure of completion of the normal inside-out migratory gradient. The potential coherence of these two studies may be underscored by the fact that NADPH-diaphorase positive neurons appear to be GABA-ergic. Unfortunately, there are some inconsistencies. Subsequent cell counts of small presumably GABA-ergic neurons in nissl-stained sections of the samples of Akbarian et al. could not directly confirm the finding of Benes et al. of a reduction in the small neuron population (16).

In a subsequent study of the NADPH-diaphorase neurons of the temporal lobe, including the lateral temporal neocortex and the mesial limbic cortex, Akbarian and colleagues (2) extended their abnormal findings to this region, suggesting a more widespread cortical developmental defect. This also would be consistent with the morphometric data. However, this second study presented some new inconsistencies. Although they did find gradient abnormalities in the hippocampus and lateral temporal neocortex, the entorhinal cortex was normal. It is conceivable that the abnormalities of the entorhinal cortex observed by Jakob and Beckman and by Arnold et al. did not involve the subset of neurons that express NADPH-diaphorase. This might be consistent with the latter neurons being primarily GABA-ergic and the layer II entorhinal cortex neurons being primarily glutamatergic. On the other hand, the involvement of NADPH-diaphorase neurons in each of the other cortical areas examined by Akbarian et al. make this explanation seem a bit strained. It also should be noted that the small sample size of the Akbarian et al. studies necessitated a matched-pair statistical analysis. It is unclear whether the findings would have held up if the matching had been done differently. Clearly, further studies of this type are needed before these uncertainties can be resolved.

In summary, if one looks at the research data base about brain abnormalities in schizophrenia as a whole, the most coherent impression is of subtle multifocal or diffuse anatomical deviations that predate the onset of the illness and are static, that are most consistent with the notion of a developmental defect, and that may implicate a failure of second-trimester neuronal migration leading to cortical maldevelopment. The most potentially incriminating evidence comes from studies of cortical cytoarchitecture. The possibility of cortical maldevelopment in the second trimester is also critical to a discussion of the additional speculation about neurodevelopmental factors and models that follows.

The possibility that obstetrical complications (OC) could contribute to or even cause schizophrenia has been the subject of a surprisingly large number of investigations over the past three decades. The literature on this issue is difficult to interpret, as the same methods are rarely used, the same findings are rarely reported, and the implications of the data are rarely critically addressed (68). Nevertheless, the weight of the evidence suggests that there is a statistical association between OCs and schizophrenia (68). An overview of the recent literature in which obstetrical histories of patients were compared with that of their unaffected siblings is shown in Table 1 and Table 2. Sibship studies are particularly important because uncertainty about the validity of maternal recall is less problematic. Although the twin studies tend to be less conclusive, perhaps because twin pregnancies are often complicated, the sibling data suggest an association.

The controversial aspects of the OC literature concern the specific nature of the complications themselves [e.g., prenatal or delivery, (48)], their neuropathological correlations, and their overall significance. The neuropathological implications that have been attributed to OCs are difficult to reconcile with the neuropathological findings associated with schizophrenia. Murray et al. (52) and Mednick et al. (49) have suggested that OCs result in periventricular hemorrhages, hypoxic-ischemic injury, and ultimately abnormalities of pruning, cell death, and developmental connectivity. McNeil (48) also suggests that delivery complications result in brain damage by virtue of hypoxic-ischemic injury. Such injuries are typically characterized by gliosis and could not account for the cytoarchitectural changes described above, which presumably occur at least 3 months before delivery. If cytoarchitectural abnormalities and lack of gliosis are a neuropathological signature of schizophrenia, it is virtually certain that these occur independent of delivery events. Other inconsistencies in the OC literature have to do with the overall pathogenic significance of the findings. Mednick and colleagues have proposed that OCs are related to genetic risk for schizophrenia and that somehow the gene for schizophrenic liability increases the neuropathological effects of OCs, including of anesthesia used during delivery (17, 49). This complex hypothesis emerged as a result of correlational data from arbitrarily defined patient subgroups, when in fact, these investigators did not find an absolute increase of OCs in their entire sample of patients with schizophrenia (56) or in their entire high-risk population (17). Moreover, the interpretation of OCs being related to increased genetic risk is exactly the opposite interpretation of Murray et al. (52), who argued from their data that OCs are especially relevant to nongenetic forms of schizophrenia.

A sober perspective on the OC literature and the pathogenic implications of OCs was offered by Goodman (34) in an enlightening critique of the subject. He pointed out that even if one accepts the frequency data, OCs increase the risk of schizophrenia by, at most, 1%. Moreover, OCs are much more common in certain environments, but the frequency of schizophrenia does not parallel this geographical and cultural distribution. Thus, OCs appear to be poor predictors of schizophrenia. Finally, he suggested that a more likely scenario for a relationship between OCs and schizophrenia was that the latter caused the former, and not the other way around. It has become increasingly apparent, as first proposed by Freud in reference to cerebral palsy, that preexisting fetal abnormalities predispose to OCs. This synthesis of the OC literature also would be more compatible with the neuropathological data.

Another potential cause of developmental injury that has occasionally been considered as a pathogenic factor in schizophrenia is prenatal viral exposure. Recently, interest in this potential etiology has mushroomed as a result of a remarkable result reported by Mednick and colleagues (50) in 1988. They examined the hospital admission records of 1781 adult individuals in Finland, some of whom were born around the time of the Helsinki influenza A-2 epidemic of 1957 (the index cases) and others who were born before the epidemic (control cases). They found that both males and females who had been in their second trimester of gestation during the height of the epidemic had a significantly higher percentage of subsequent admission diagnoses of schizophrenia than either the control cases or the index cases exposed during other trimesters. The implications were that influenza itself or a related phenomena (e.g., fever) interfered with second-trimester brain development and that such interference was an etiological risk factor for schizophrenia. This study has spawned at least six other studies (summarized in Table 3), opening a new area of research controversy. While there have been positive reports supporting the findings of Mednick and colleagues, negative reports are virtually as frequent, and methodological uncertainties cloud the interpretation of the results.

Kendell and Kemp (42) compared hospital admission records throughout Scotland of individuals who had been in utero during the influenza A epidemics of 1918–1919 and 1957. Although they did find an increased risk of schizophrenia for those in the second trimester in 1957 in a sample from Edinburgh, the overall Scottish admission data revealed no such effect. These authors tended to dismiss the Edinburgh data and emphasized their negative findings. Moreover, they have argued that a reanalysis of the data of Mednick et al. using absolute numbers of patients with schizophrenia rather than proportions of such patients relative to other diagnoses is also negative (41).

Subsequent studies have only added to the controversy. Torrey et al. (66) looked at a large birth cohort in the United States and found no association of increased risk with the U.S. influenza A-2 epidemic of 1957. Barr et al. (7) attempted to replicate the study of Mednick et al. (50) in a Danish sample. Regrettably, they chose a different approach to data collection and analysis. They collected hospital admission data about all patients in Denmark with a diagnosis of schizophrenia born between 1911 and 1951, divided them into three groups based on the relative frequency of influenza infections in the general population during their birth month as recorded in public health infectious disease records, and analyzed the birth-rate data as a deviation score from the expected schizophrenia birth rate for a particular month. This complex analysis was justified as an attempt to control for spurious associations that might result from seasonal variations in influenza exposure and in schizophrenia birth rates. The authors reported that in the high seasonal influenza exposure subgroup, those exposed during the sixth month of gestation had the highest rates of schizophrenia. Unfortunately, the data also appear to show that the other two groups have lower than expected schizophrenia birthrates during the same month (i.e., the sixth) of gestation. Moreover, it appears, although this analysis is not reported, that unless the tripartite approach to subgrouping is used, there is no overall association between schizophrenia birth rate and exposure to influenza A-2. Indeed, the authors acknowledge that even in their positive subgroup, the association between schizophrenia and influenza is weak, accounting for at most 4% of the variance.

Another example of the confusing nature of this literature is the study by O'Callaghan et al. (53). These investigators compared admission diagnoses in eight health regions of England and Wales with birth records around the time of the 1957 influenza epidemic. They reported that for patients in utero during the second trimester (especially the fifth month) "the number of births of individuals who later developed schizophrenia was 88% higher" than expected. In fact, what they actually found was that the overall monthly distribution of affected births (i.e., subsequent schizophrenia) was different in the exposure and the control years. Because they did not perform post hoc tests on individual months, they did not statistically determine which month(s) accounted for the overall distribution differences. Indeed, their data also show a peak at around 3 months and a trough at around 8 months, both of which may have been important factors in the overall distribution analysis. In a related study, Sham et al. (62) examined the numbers of first-time admissions with a diagnosis of schizophrenia throughout England and Wales from 1970 to 1979 and compared them to the number of deaths from influenza during the period 1930 to 1969. Arguing that the latter was a relative index of likelihood of influenza exposure in utero for the former, they found a statistical relationship using a complex model-fitting paradigm of the two frequency distributions, with greatest correspondence during the third and seventh months of gestation. The data suggested, however, that at most 1% to 2% of schizophrenic births could be explained by this relationship.

An important limitation of each of the perinatal exposure studies is that none of them documented actual maternal, let alone, intrauterine infection. In the only study to attempt this, Crow and Done (21) investigated psychiatric admissions of individuals born around the 1957 epidemic who had been enrolled in a perinatal injury and child development research project in England. They identified 945 individuals whose mothers had been diagnosed during their second trimester of pregnancy as having influenza. They did not find an increased risk of subsequent schizophrenia in the offspring of these mothers.

In summary, the perinatal viral exposure literature is provocative but inconclusive. The reasons for the inconsistencies are uncertain, and it is doubtful that they will be resolved in the near future. Even if the inconsistencies can be resolved and the positive results prevail, exposure to influenza will account for at most a small minority of cases. Moreover, although the positive results may add circumstantially to the notion that second-trimester maldevelopment increases the risk of schizophrenia, the mechanisms by which this happens will be difficult to establish from epidemiological studies of potential viral exposure.

Because of the possibility that neural development may be disrupted by a number of adverse environmental events in addition to viral exposure, other causes have been considered. For example, in a recent study of the potential impact of maternal malnutrition, Susser and Lin (67) studied the effects of starvation in occupied Holland during World War II and found that first- but not second-trimester starvation was associated with increased risk of schizophrenic births. Because the implications of this finding for the neuropathological changes associated with schizophrenia are unclear, if not contradictory, this study is difficult to integrate with the rest of the neurodevelopmental data base. Also, this study did not control for the implications of social class both on access to food and on risk for schizophrenia.

In light of the widely accepted data that genetic factors convey susceptibility to schizophrenia, it is not surprising that there has been speculation about genetic factors that may affect brain development in schizophrenia. Approximately 30% of the genome is expressed in brain, and many genes are turned on and off during discrete phases of brain development; therefore, there are many potential candidates. No existing data link schizophrenia with a defect in any known gene related to brain development. Nevertheless, Murray et al. (52) have hypothesized that the fundamental neuropathological deviations in the schizophrenic brain arise because of a primary genetic defect in at least a substantial subgroup of patients. Mednick et al. (49) have hypothesized that a genetic defect predisposes the schizophrenic brain to being adversely affected by intrauterine or perinatal environmental events. These hypotheses and undoubtedly many others that will be advanced must await the discovery and practicability of scientific methods to test them.

The possibility that schizophrenia is related to an abnormality of early brain development poses yet another interesting challenge, for the clinical expression of the illness is delayed typically for about two decades after birth. If the neurological abnormality is present at birth, why is the illness itself not manifested earlier in life and what accounts for its predictable clinical expression in early adulthood? Speculation about the answers to these questions has come primarily from two perspectives: (a) the possibility of an additional pathological process occurring around the time of onset of the clinical illness; and (b) an interaction between a static developmental defect and normal developmental programs or events that occur in early adult life.

As the foremost proponent of the first perspective, Feinberg (30) focused on the age of onset of schizophrenia as a clue to neurodevelopmental abnormalities that might explain the illness. He posited that schizophrenia is caused by a defect in adolescent synaptic reorganization, because either "too many, too few, or the wrong synapses are eliminated." In effect, he argues for a second pathological process, a specific pathology of synaptic elimination not necessarily related to possible maldevelopment in utero. His hypothesis does not take into account the replicable neuropathological data base (most of which did not exist at the time of his original proposal), and he does not address the biological mechanisms that might be responsible for this putative disorder of synaptic elimination. In light of the neuropathological data base that implicates maldevelopment in utero, this hypothesis would require the unlikely scenario of a second primary pathology. Another problem with this hypothesis is that it is unclear how one could directly test it, especially because it accommodates all potential variations (i.e., too much, too little, or the "wrong" pruning).

An alternative scenario that might be considered is that maldevelopment in utero sets the stage for secondary synaptic disorganization that has its greatest neurobiological and clinical impact in adolescence. This integration of the neuropathological data with the Feinberg hypothesis would tend to regard irregularities of synaptic pruning as epiphenomena. It would be consistent with the notion that neuronal circuitry that is anomalous from early in development may have particularly profound implications for eventual connectivity. In other words, perhaps primary migratory or other developmental defects lead to the creation of abnormal circuits that compete successfully for survival, whereas certain normal circuits either do not form or are structurally disadvantaged, so that they cannot avoid elimination.

Other mechanisms for delayed onset that emphasize a new process going wrong around the time of clinical onset have been proposed. These include abnormalities of myelination (11), of neuronal sprouting (65), and of adverse effects of stress-related neural transmission (14). Each of these involves a variation on the theme of another abnormality taking place in early adult life. In essence, they are dual pathology hypotheses, either positing that maldevelopment in utero is not sufficient pathology, or is coincidental, or that it is only one of two relatively independent pathologies that characterize the illness.

The second perspective maintains that it might be possible to accommodate both maldevelopment in utero and delayed clinical onset without positing an additional abnormal process in adolescence. This perspective involves an interaction between cortical maldevelopment in utero and normal developmental events that occur much later (71). This view rests on several assumptions: that the clinical implications of a developmental defect vary with the maturational state of the brain; that the neural systems disrupted by the defect in early brain development in schizophrenia are normally late-maturing neural systems; and that a defect in the function of these neural systems will not be reliably apparent until their normal time of functional maturation. In other words, it is posited that certain neural systems are destined from early development to malfunction in a manner that accounts for the illness, but until a certain state of postnatal brain development they either do not malfunction to a clinically significant degree, or their malfunctioning can be compensated for by other systems. The first of these assumptions has been repeatedly validated in developmental neurobiology. Indeed, a fundamental principle of the clinical impact of developmental neuropathology is that in general early brain damage is apparent early and tends to become less so over time. The young brain has a greater capacity for functional compensation than does the old brain. It is also a fundamental principle of pediatric neurology that, in some cases, congenital brain damage can have delayed or varying clinical effects if the neural systems involved are neurologically immature at birth (71).

In the case of schizophrenia, these fundamental principles appear to be violated, in that the impact of putative early damage is less apparent early and more apparent late. In this respect, the other two assumptions of this perspective are much more speculative. It is not known whether the principle of clinical effects being delayed until the affected neural systems reach functional maturity applies to those neural systems implicated in schizophrenia. More data are needed about the neural systems that develop abnormally in schizophrenia and about their normal course of functional maturation. Nevertheless, in the absence of such data, the following speculation seems appropriate. The neuropathological data base about schizophrenia has highlighted subtle maldevelopment of the cerebral cortex. Correlative data from studies of cortical function in patients with schizophrenia, including neuropsychological testing results (32) and studies of cortical physiology using functional brain-imaging techniques (13), indicate that cortical dysfunction is a prominent characteristic of the illness and that prefrontal–temporal functional connectivity is especially impaired. Even if cortical maldevelopment is widespread, the functional neural systems that appear to be particularly relevant to the clinical characteristics of schizophrenia are those involved in prefrontal–temporal connectivity (13, 72). This pattern is consistent with the developmental neuropathology reviewed above. If the function of systems that subserve such connectivity matures late, as a number of lines of evidence suggest, then this would fit the model of this perspective about delayed onset. The molecular events that account for the functional maturation of these systems are probably complex and may involve stabilization of synapses, leveling off of the growth of dendritic arbors, and other processes related to the refinement of cortical connectivity, all of which seem to plateau in early adult life.

These alternative perspectives on mechanisms of delayed onset, although differing on the question of whether neurodevelopmental processes of adolescence are abnormal, share an emphasis on cortical connectivity being abnormal, as do the in vivo imaging, neuropsychological, and postmortem data. This raises an additional problem for the explanatory power of neurodevelopmental models of schizophrenia, in that the diagnostic symptoms of the illness, that is, hallucinations and delusions, have not been classically imputed to cortical dysfunction. Moreover, it is unclear how this apparent inconsistency could be resolved by the added complexity of the neurodevelopmental frame of reference. Potential insights into these issues have come from studies of neurological illnesses associated with developmental neuropathology and psychosis and from animal models of delayed effects of perinatal injury.

Hallucinations and delusions are not unique to schizophrenia. They are encountered in a number of neurological conditions, involving many areas of the brain. It has been pointed out that one of the better predictors of whether a neurological condition presents with psychosis is the age at which it presents (72). Of those disorders that may have psychopathology as a prominent symptom, psychosis is much more likely to be manifest in late adolescence and early adulthood than at other times of life. This is true even if the neuropathological changes do not vary with age. In other words, simply disrupting the neural systems related to psychosis is not necessarily sufficient to cause the syndrome. The brain needs to be at a certain state of development for the maximum likelihood of psychosis.

Metachromatic leukodystrophy (MLD) is an informative example of this age association and also of the potential importance of functional "dysconnection" of cortical regions. Hyde and colleagues (36) have demonstrated that when MLD presents between the ages of 13 and 30, it presents in the majority of cases as a schizophrenialike illness. Moreover, the clinical presentation is probably more similar to schizophrenia than is seen in any other neurological disease. Patients have disorganized thinking, act bizarrely, have complex delusions, and when hallucinated, invariably have complex, Schneiderian-type auditory hallucinations. The condition is often misdiagnosed as schizophrenia, sometimes for years, before neurological symptoms appear. Interestingly, MLD is not a disease of mesial temporal cortex or of frontal cortex. It is a pure connectivity disorder in that the neuropathological changes involve white matter connections. In its early neuropathological stages, when it is most likely to present with psychosis, the changes are especially prominent in subprefrontal white matter. This suggests that a neural dysfunction with a high valence for producing psychotic symptoms is failure of some aspects of prefrontal connectivity, analogous functionally to what has been implicated in schizophrenia.

In the case of MLD, however, this functional "dysconnection" does not appear to be enough. When MLD presents outside of this critical age range, it almost never presents with psychosis, even though the location of the neuropathology is not age dependent. In other words, the involvement of critical neural systems is not by itself sufficient for the expression of psychosis. An age-related factor that appears to be independent of the illness is also required. Because this age factor transends specific illness boundaries, it is probably a function of normal postnatal brain maturation. Thus, the example of MLD supports the theoretical perspective that psychosis may reflect a cortical defect that interacts with developmental programs normally linked to a late adolescent brain.

The analogy of MLD suggests that a putative defect in intracortical connectivity occurring at a critical period in postnatal brain maturation may clinically manifest as psychosis. It does not, however, suggest that a defect in such connectivity existing at birth could remain clinically silent until late adolescence. Studies in animals have supported this possibility.

Goldman (33) showed that a perinatal ablation of the dorsolateral prefrontal cortex did not impair performance on delayed response tasks in animals before adolescence as a similar lesion placed in adult animals did. However, when animals with perinatal lesions of this cortical region reached adolescence, their performance on these tests actually became impaired. She suggested that prior to puberty other brain regions (e.g., the caudate) took responsibility for this behavior, but by adolescence, the brain was developmentally committed to use the cortex for this activity. In other words, the other neural strategies or systems were no longer available or were incapable of performing normally in the context of a more developed brain. She also showed that this delay in presentation of the cognitive deficit was a characteristic of some neural systems and not others. In animals with ablations of orbital frontal cortex, a phylogenetically older cortical region, deficits were manifest early in life, and if anything, they improved as the animals grew older.

Recently, Beauregard et al. (9) reported preliminary data suggesting that an analogous delay in the appearance of certain cognitive deficits can be seen after perinatal hippocampal ablations. In rhesus monkeys with selective ablations of the rostral hippocampus, performance on delayed nonmatching to sample tests, tests that are sensitive to hippocampal dysfunction in adult monkeys, are only mildly impaired until after puberty when they become more profoundly so.

These studies indicate that perinatal injury to prefrontal and limbic cortices, two regions that are strongly implicated in schizophrenia, can have minimal impact on cognitive function before puberty and clinically significant impact afterward. This suggests that the apparent deterioration in cognitive function seen in patients with schizophrenia around the time of onset of their illness (32) could conceivably be the result of early developmental brain injury by itself. However, these studies do not suggest that psychotic symptoms, symptoms that tend to respond to antipsychotic drugs presumably through the blockade of dopamine receptors, also could suddenly emerge during early adulthood as a result of early developmental brain injury by itself. Recent studies in the rat, however, have made it possible to conceive of this as well.

In a series of studies, Lipska and colleagues (45, 46) have shown that perinatal excitotoxic damage of the ventral hippocampus of the rat represents an interesting model of delayed emergence of hyperdopaminergic behaviors. These animals show no evidence of abnormalities of mesolimbic or nigrostriatally mediated behaviors until after puberty, whereupon they become hyperresponsive to dopaminergic drugs and to a variety of experiential stresses. They also manifest the delayed emergence of other abnormalities associated with schizophrenia, such as abnormal prepulse inhibition of startle. Moreover, in some respects they grow up to look like animals with adult prefrontal lesions, suggesting that the neonatal hippocampal lesion has affected prefrontal development as well. Their abnormal dopaminergic behaviors are ameliorated by antipsychotic drugs.

This provocative animal model illustrates that early developmental damage to cortical systems that have been implicated in schizophrenia may also have delayed effects on the regulation of brain dopamine systems. These delayed effects may have something to do with the maturation of these dopamine systems. They might also reflect the maturation of experience-based cortical systems which are programmed to take responsibility in early adulthood for regulating subcortical dopamine systems when they are needed. It seems theoretically possible that if these cortical systems developed abnormally, they may be unable to appropriately regulate subcortical dopamine systems when it is their time to do so.

The research summarized in this chapter tends, on the whole, toward the interpretation that individuals who manifest schizophrenia in early adult life have suffered some form of subtle cerebral maldevelopment in utero. This assumption is based on imperfect studies and inconclusive data. Nevertheless, there are a number of converging lines of evidence. The morphometric anatomical data from in vivo and postmortem investigations are difficult to attribute to a neuropathological process of adult life. The static nature of the structural findings, the correlations with early life adaptation, and the absence of gliosis are much more consistent with the possibility of a developmental anomaly. By far, the most important anatomical data concerns the evidence of cytoarchitectural disorganization of the cortex. These findings are virtually pathognomonic of a defect in cortical development occurring during the second trimester of gestation. If these data can be replicated in methodologically unimpeachable studies, a "smoking gun" will have been identified. Further efforts in this regard should be at the top of the list of priorities in schizophrenia research.

The other issues in thinking about schizophrenia as a neurodevelopmental disorder pivot on the validity of the cytoarchitectural findings. If these postmortem results are valid, then the etiology is one that affects brain development during this critical period. The etiology would clearly not be birth complications. Moreover, the question of why the clinical manifestations of such congenital damage are not present in recognizable form until early adulthood also becomes increasingly important, because answering this question may hold clues to the mechanisms of clinical compensation and decompensation. The scenario stressed in this chapter involves an interaction of cortical maldevelopment with normal programs of postnatal functional development of critical intracortical neural systems. The systems that appear to be especially relevant to the cortical functional impairments of schizophrenia involve prefrontal and limbic cortices and their connectivity. Such highly evolved, late maturing systems may be especially important in coping with the vicissitudes of independent psychosocial functioning and may be critical as well for the stress-related management of subcortical dopamine activity. As recently demonstrated in a new wave of heuristically meaningful animal models, it is conceivable that a congenital defect in such systems would remain submerged until early adult life and then fail to properly regulate critical secondary systems (e.g., subcortical dopamine activity) in the context of environmental stress. It is further conceivable that genetically determined variations in intracortical connectivity or in responsivity of the limbic dopamine system, which would not be clinically significant by themselves, could become clinically devastating when combined with the manner of cortical maldevelopment implicated above.

published 2000