has become more complex (Lambe, Krimer, & Goldman-Rakic, 2000). The possibility that schizophrenia involves molecular and functional abnormalities of information flow in these circuits suggests that such abnormalities may converge on the dynamic process of brain maturation during adolescence and increase the risk of a psychotic episode in predisposed individuals.

Despite over a century of research, we have only a limited understanding of what causes schizophrenia and related psychotic disorders. Early studies of the biological basis of schizophrenia relied mostly on either postmortem studies of brains of people with this illness or brain imaging studies typically of older patients with chronic schizophrenia, many of whom were treated with medications. It was therefore difficult to know to what extent the observed changes were the results of aging, illness chronicity, or medication effects. One can avoid such difficulties by conducting studies of individuals in the early phases of schizophrenia (Keshavan & Schooler, 1992). First, these studies allow us to clarify which of the biological processes may be unique to the illness and which ones might be a result of medications or of persistent illness. Second, first-episode studies allow us to longitudinally evaluate the course of the brain changes, and how such changes can help us predict outcome with treatment. Follow-up studies suggest that less than half of early psychosis patients go on to develop a chronic form of schizophrenia with poor level of functioning and intellectual deficits (Harrison et al., 2001). An understanding of which patients may have such an outcome will greatly help treatment decisions early in the illness. Finally, not all who have features of the prodromal phases of the illness go on to develop the psychotic illness (Yung et al., 2003). Studies of the prodromal and early course of psychotic disorders provide an opportunity to elucidate the neurobiological processes responsible for the transition from the prodromal to psychotic phase of the illness.

Several conceptual models of the biology and causation of schizophrenia have been recently suggested, and serve to guide research into the early phase of this illness. One view, which dates back to the late 1980s, is the so-called early neurodevelopmental model (Murray & Lewis, 1987; Weinberger, 1987). This model posits abnormalities early during brain development (perhaps at or before birth) as mediating the failure of brain functions in adolescence and early adulthood. Several lines of evidence, such as an increased rate of birth complications, minor physical and neurological abnormalities, and subtle behavioral difficulties in children who later developed schizophrenia, support this view. However, many nonaffected persons in the population also have these problems; their presence cannot inform us with confidence whether or not schizophrenia will develop later in life. The fact that the symptoms typically begin in adolescence or early adulthood suggests that the illness may be related to some biological changes related to adolescence occurring around or prior to the onset of psychosis. Childhood is characterized by proliferation of synapses and dendrites, and normal adolescence is characterized by elimination or pruning of unnecessary synapses in the brain, a process that serves to make nerve cell transmission more efficient (Huttenlocher et al., 1982). This process could go wrong, and an excessive pruning before or around the onset phase of illness (Feinberg, 1982b; Keshavan, Anderson, & Pettegrew, 1994) has been thought to mediate the emergence of psychosis in adolescence or early adulthood. Our understanding of the underlying neurobiology of this phase of illness remains poor, however. Another view is that active biological changes could occur after the onset of illness, during the commonly lengthy period of untreated psychosis. This model proposes progressive neurodegenerative changes (Lieberman, Perkins, et al., 2001). It is possible that all three processes are involved in schizophrenia (Keshavan & Hogarty, 1999); additionally, environmental factors such as drug misuse (Addington & Addington, 1998) and psychosocial stress (Erickson, Beiser, Iacono, Fleming, & Lin, 1989) may trigger the onset and influence the course of schizophrenia. Careful studies of the early phase of schizophrenia can shed light on these

apparently contrasting models. The three proposed pathophysiological models might reflect different critical periods for prevention and therapeutic intervention.

The most basic question in the genetics of schizophrenia is whether the disorder aggregates (or “runs”) in families. Technically, familial aggregation means that a close relative of an individual with a disorder is at increased risk for that disorder, compared to a matched individual chosen at random from the general population. Twenty-six early family studies, conducted prior to 1980 and lacking modern diagnostic procedures and appropriate controls, consistently showed that first-degree relatives of schizophrenia patients had a risk for schizophrenia that was roughly 10 times greater than would be expected in the general population (Kendler, 2000). Since 1980, 11 major family studies of schizophrenia have been reported that used blind diagnoses, control groups, personal interviews, and operationalized diagnostic criteria. The level of agreement in results is impressive. Every study showed that the risk of schizophrenia was higher in first-degree relatives of schizophrenic patients than in matched controls. The mean risk for schizophrenia in these 11 studies was 0.5% in relatives of controls and 5.9% in the relatives of schizophrenics. Modern studies suggest that, on average, parents, siblings, and offspring of individuals with schizophrenia have a risk of illness about 12 times greater than that of the general population, a figure close to that found in the earlier studies.

Recently, results of the first methodologically rigorous family study of child-onset schizophrenia have been reported. Compared to parents of matched normal controls and children with ADHD, parents of childhood-onset schizophrenia had an over 10-fold increased risk for schizophrenia. This finding supports the hypothesis of etiologic continuity between childhood-and adult-onset schizophrenia (Asarnow, Tompson, & Goldstein, 2001).

Resemblance among relatives can be due to either shared or family environment (nurture), to genes (nature), or to both. A major goal in psychiatric genetics is to determine the degree to which familial aggregation for a disorder such as schizophrenia results from environmental or genetic mechanisms. Although sophisticated analysis of family data can begin to make this discrimination, nearly all of our knowledge about this problem in schizophrenia comes from twin and adoption studies.

Twin studies are based on the assumption that “identical,” or monozygotic (MZ), and “fraternal,” or dizygotic (DZ), twins share a common environment to approximately the same degree. However, MZ twins are genetically identical, whereas DZ twins (like full siblings) share on average only half of their genes. Results are available from 13 major twin studies of schizophrenia published from 1928 to 1998 (Kendler, 2000). Although modest differences are seen across studies, overall, the agreement is impressive. Across all studies, the average concordance rate for schizophrenia in MZ twins is 55.8% and in DZ twins, 13.5%. When statistical models are applied to these data to estimate heritability (the proportion of variance in liability in the population that is due to genetic factors), the average across all 13 studies is 72%. This figure, which is higher than that found for most common biomedical disorders, means that, on average, genetic factors are considerably more important than environmental factors in affecting the risk for schizophrenia.

Adoption studies can clarify the role of genetic

and environmental factors in the transmission of schizophrenia by studying two kinds of rare but informative relationships: (1) individuals who are genetically related but do not share their rearing environment, and (2) individuals who share their rearing environment but are not genetically related. Three studies conducted in Oregon, Denmark, and Finland all found significantly greater risk for schizophrenia or schizophrenia-spectrum disorders in the adopted-away offspring of schizophrenic parents than that for the adopted-away offspring of matched control mothers. The second major adoption strategy used for studying schizophrenia begins with ill adoptees rather than with ill parents and compares rates of schizophrenia between groups of biologic parents and groups of adoptive parents. In two studies from Denmark using this strategy, the only group with elevated rates of schizophrenia and schizophrenia-spectrum disorders were the biological relatives of the schizophrenic adoptees (Kety et al., 1994).

Twin and adoption studies provide strong and consistent evidence that genetic factors play a major role in the familial aggregation of schizophrenia. Although not reviewed here, evidence for a role for nongenetic familial factors is much less clear. Some studies suggest they may contribute modestly to risk for schizophrenia, but most studies find no evidence for significant nongenetic familial factors for schizophrenia.

Since the earliest genetic studies of schizophrenia, a major focus of such work has been to clarify more precisely the nature of the psychiatric syndromes that occur in excess in relatives of schizophrenic patients. To summarize a large body of evidence, relatives of schizophrenia patients are at increased risk for not only schizophrenia but also schizophrenia-like personality disorders (best captured by the DSM-IV categories of schizotypal and paranoid personality disorder) and other psychotic disorders (Kendler, 2000). However, there is good evidence that relatives of schizophrenia patients are not at in creased risk for other disorders, such as anxiety disorders and alcoholism. The most active debate in this area is the relationship between schizophrenia and mood disorders. Most evidence suggests little if any genetic relationship between these two major groups of disorders, but some research does suggest a relationship particularly between schizophrenia and major depression.

The evidence that other disorders in addition to schizophrenia occur at greater frequency in the close relatives of individuals with schizophrenia has led to the concept of the schizophrenia-spectrum—a group of disorders that all bear a genetic relationship with classic or core schizophrenia.

Given the evidence that genetic factors play an important role in the etiology of schizophrenia, a major focus of recent work has been to apply the increasingly powerful tools of human molecular genetics to localize and identify the specific genes that predispose to schizophrenia. Two strategies have been employed in this effort: linkage and association. The goal of linkage studies is to identify areas of the human genome that are shared more frequently than would be expected by relatives who are affected. If such areas can be reliably identified, then these regions may contain one or more specific genes that influence the liability to schizophrenia. The method of linkage analysis has been extremely successful in identifying the location of genes for simple, usually rare medical genetic disorders (termed “Mendelian” disorders) in which there is a one-to-one relationship between having the defective gene and having the disorder. This method, however, has had more mixed results when applied to disorders such as schizophrenia that are genetically “complex.” Such complex disorders are likely to be the result of multiple genes, none of which have a very large impact on risk, interacting with a range of environmental risk factors.

Eighteen genome scans for schizophrenia

have been published between 1994 and 2002. None of these scans has revealed evidence for a single gene with a large impact on risk for schizophrenia. Indeed, these results suggest that the existence of a single susceptibility locus that accounts for a large majority of the genetic variance for schizophrenia can now be effectively ruled out.

The most pressing scientific issue in the interpretation of linkage studies of schizophrenia has been whether there is agreement at above-chance levels across studies on which individual regions of the genome contain susceptibility genes for schizophrenia. Until recently, the across-study agreement had not been very impressive.

Two recent findings have increased our confidence that linkage studies of schizophrenia may be producing reliable results. First, in a large-scale study of families containing two or more cases of schizophrenia, conducted in Ireland, the sample was divided, prior to analysis, into three random subsets (Straub, MacLean, et al., 2002). When a genome scan was performed on these three subsets, three of the four regions that most prominently displayed evidence for linkage (on chromosomes 5q, 6p, and 8p) were replicated across all three subsets. Interestingly, one region, on chromosome 10p, was not replicated even within the same study. Probably more important, Levinson and collaborators were able to obtain raw data from nearly all major published genome scans of schizophrenia to perform a meta-analysis—a statistical method for rigorously combining data across multiple samples (Lewis et al., 2003). Ten regions produced nominally significant results including 2q, 5q, 6p, 22q, and 8p. The authors concluded: “There is greater consistency of linkage results across studies than had been previously recognized. The results suggest that some or all of these regions contain loci that increase susceptibility to schizophrenia in diverse populations.”

The evidence for replicated linkages in schizophrenia represents an important step toward the ultimate goal of identifying susceptibility genes and characterizing their biologic effects. Because the human genome contains within its 23 pairs of chromosomes over three billion nucleotides (i.e., “letters” in the genetic alphabet) and 30,000 genes (i.e., protein-encoding units), it is a large territory to explore. Linkage is a strategy to narrow the search and to provide a map of where the treasure (i.e., the genes) may lie. The linkage results in schizophrenia so far have highlighted several regions of the genome for a more thorough search. Association (also called linkage disequilibrium) is the next critical step in this search for the treasure. Linkage represents a relationship between regions of the genome shared by family members who also share the phenotype of interest—here, schizophrenia. It provides a low-resolution map because family members share relatively large regions of any chromosome. Association, however, represents a relationship between specific alleles (i.e., specific variation in a gene or in a genetic marker) and illness in unrelated individuals. It provides a high-resolution map because unrelated people share relatively little genetic information. For a given allele to be found more frequently in unrelated individuals with a similar disease than it is in the general population, the probability that this specific allele is a causative factor in the disease is enhanced. If the frequency of a specific allele (i.e., a specific genetic variation) is greater in a sample of unrelated individuals who have the diagnosis of schizophrenia than it is in a control population, the allele is said to be associated with schizophrenia. This association represents one of three possibilities: the allele is a causative mutation related to the etiology of the disease; the allele is a genetic variation that is physically close to the true causative mutation (i.e., in “linkage disequilibrium” with the true mutation); or the association is a spurious relationship reflecting population characteristics not related to the phenotype of interest. This latter possibility is often referred to as a population stratification artifact, meaning that differences in allele frequencies between the cases and control samples are not because of disease but because of systematic genetic differences between the comparison populations.

Association has become the strategy of choice

for fine mapping of susceptibility loci and for preliminary testing of whether specific genes are susceptibility genes for schizophrenia. The strategy involves identifying variations (“polymorphisms”) in a gene of interest and then performing a laboratory analysis of the DNA samples to “type” each variation in each individual and determine its frequency in the study populations. Genetic sequence variations are common in the human genome and public databases have been established to catalog them. The most abundant sequence variations are single nucleotide polymorphisms (SNPs), which represent a substitution in one DNA base. Common SNPs occur at a frequency of approximately one in every 1,000 DNA bases in the genome and over two million SNPs have been identified. While SNPs are relatively common, most SNPs within genes either do not change the amino acid code or are in noncoding regions of genes (“introns”) and are thus not likely to have an impact on gene function.

Early association studies in schizophrenia focused on genes based on their known function and the possibility that variations in their function might relate to the pathogenesis of the disease. These so-called functional candidate gene studies had no a priori probability of genetic association. A number of studies compared frequencies of variations in genes related to popular neurochemical hypotheses about schizophrenia, such as the dopamine and glutamate hypotheses, in individuals with schizophrenia with those in control samples. In almost every instance the results were mixed, with some positive but mostly negative reports. Many of the positive studies were compromised by potential population stratification artifacts. However, because the effect on risk of any given variation in any candidate gene (e.g., a dopamine or glutamate receptor gene) is likely to be small (less than a twofold increase in risk), most studies have been underpowered to establish association or to rule it out.

Recent association studies have been much more promising, primarily because of the linkage results. Using the linkage map regions as a priori entry points into the human genetic sequence databases, genes have been identified in each of the major linkage regions that appear to represent at least some of the basis for the linkage results. Moreover, confirmation of association in independent samples have appeared, which combined with the linkage results comprise convergent evidence for the validity of these genetic associations. In the August and October 2002 issues of the American Journal of Human Genetics, the first two articles appeared that claimed to identify susceptibility genes for schizophrenia, starting with traditional linkage followed by fine association mapping. Both of these were in chromosomal regions previously identified by multiple linkage groups: dysbindin (DTNBP1) on chromosome 6p22.3 (Straub, Jiang, et al., 2002) and neuregulin 1 (NRG1) on chromosome 8p-p21 (Stefansson et al., 2002). Both groups identified the genes in these regions from public databases and then found variations (SNPs) within the genes that could be tested via an association analysis. In both studies, the statistical signals were strong and unlikely to occur by chance. In the January 2003 issue of the same journal, two further articles were published, replicating, in independent population data sets also from Europe, association to variations in the same genes (Schwab et al., 2003; Stefansson et al., 2003). In the December 2002 issue of the same journal, authors of a study on a large population sample from Israel reported very strong statistical association to SNPs in the gene for catechol-O-methyltransferase (COMT), which was mapped to the region of 22q that had been identified as a susceptibility locus in several linkage studies (Shifman et al., 2002). Positive association to variation in COMT had also been reported in earlier studies in samples from China, Japan, France, and the United States (Egan et al., 2001). Starting with the linkage region on chromosome 13q34, a group from France discovered a novel gene, called G72, and reported in two population samples association between variations in this gene and schizophrenia (Chumakov et al., 2002). The SNP variations in G72 have recently been reported to be associated with bipolar disorder as well.

In addition to these reports based on relatively strong linkage regions, several other promising associations have emerged from genes found in weaker linkage regions. For example, a weak