Schizophrenia is a clinically heterogeneous disorder in which patients exhibit a broad range of deficits and symptom severity. A core feature of the illness is a fundamental impairment in the ability to filter sensory information (
1,
2), and the complex cognitive deficits in attention, information processing, learning, and memory that also characterize schizophrenia may arise from these primary deficits in sensory processing, or “gating.” Intact gating processes allow healthy individuals to filter out irrelevant stimuli and appropriately allocate attentional resources. However, impaired gating in schizophrenia patients may lead to sensory overload, distorted perceptions, and cognitive fragmentation (
3). Misinterpretations of a cascade of irrelevant stimuli may further result in some of the observed symptoms associated with schizophrenia, such as hallucinations, delusions, disorganized speech and thought, and social awkwardness (
1,
2,
4–
6).
Prepulse inhibition (PPI) and P50 gating are neurophysiological endophenotypes for schizophrenia that reflect inhibitory abnormalities in central gating processes. PPI is an index of sensorimotor gating measured as a reduction in the acoustic startle response magnitude that occurs when the intense startling pulse stimulus is preceded 30–300 ms by a weak prepulse (
7–
9). The P50 wave is a midlatency auditory evoked potential that exhibits reduced amplitude when a second (test) click is presented 500 ms after an initial (conditioning) click in a paired-click paradigm (
2). The PPI and P50 deficits observed in schizophrenia patients extend to their clinically unaffected relatives, as well as to schizotypal individuals and adolescents at high risk for schizophrenia, indicating that these deficits are present across the schizophrenia spectrum and may be biomarkers for genetic risk for the disorder (
1,
6,
10–
21). Although PPI and P50 measures are conceptually linked to gating processes, evidence directly connecting these measures to sensory overload, distorted perceptions, or cognitive fragmentation is currently lacking. Evidence does suggest, however, that these two measures detect distinct aspects of pathophysiology that may be characteristic of different subgroups of patients (
22–
26).
The Consortium on the Genetics of Schizophrenia (COGS) explores neurophysiological and neurocognitive endophenotypes as a means to understand the neurobiological processes underlying schizophrenia. We have previously reported significant heritability for 12 primary endophenotypes in the COGS Family Study (COGS-1). These analyses revealed estimates of heritability for PPI and P50 gating that were significant but modest compared with earlier family studies of these measures (
27–
32). COGS-1 focused on the recruitment of discordant sib pairs and intact families available for extensive testing in order to maximize the potential for phenotypic contrasts between and within families. We speculated that this ascertainment strategy might have also created a downward pressure on the estimated heritability of certain endophenotypes through the recruitment of probands with nonfamilial (i.e., sporadic) forms of schizophrenia (
33–
35). The heritability of PPI and P50 measures in particular may have been affected, as they reflect primary gating deficits in schizophrenia. Here we report the evaluation of the heritability of PPI and P50 gating separately in families with a positive family history for schizophrenia and related disorders derived from multiple affected family members (multiply affected) compared with families containing only the proband affected with schizophrenia and no other family members with any major psychiatric illness (singleton). These heritability estimates were compared with those generated from the entire COGS-1 sample to test the hypothesis that family history substantially and specifically affects the heritability of gating measures.
Results
As shown in
Table 2, the availability of diagnosis information based on the Family Interview for Genetic Studies allowed for an assessment of the presence or absence of family history of schizophrenia or related disorders (
35). Estimates of heritability for PPI and P50 gating were substantially increased in the 97 multiply affected families with a positive family history (47% and 36%, respectively) compared with the entire sample of 296 COGS-1 families (29% and 20%, respectively). By contrast, no evidence for significant heritability was observed for either PPI or P50 gating in the 96 singleton families. For comparison, we evaluated pulse-alone startle magnitude as a measure of initial startle reactivity and observed reasonably consistent heritability estimates between the subsets of families stratified by family history (54% for multiply affected compared with 60% for singleton) and the entire sample (62%). We similarly evaluated P50 S1 amplitude, which, although significantly heritable in all subsets, demonstrated a dramatic increase in heritability in the multiply affected subset (80% compared with 35% and 39% in the singleton subset and in the entire sample, respectively). Conversely, P50 S2 produced reasonably consistent estimates (i.e., within standard error) between the subsets of families stratified by family history and the entire sample, although it was just below the threshold for significance in the positive family history cohort. Finally, we evaluated the three primary cognitive endophenotypes from COGS-1 (
26) and observed consistent heritability estimates (i.e., within standard error) among the subsets of families stratified by family history and the entire sample.
We next compared the schizophrenia probands with positive or negative family history for differences on these gating measures, as well as symptom severity (
Table 3). Although differences in PPI and P50 gating deficits were not observed between probands derived from multiply affected versus singleton families, significant clinical differences were observed. Probands with a positive family history had significantly greater clinical severity, as evidenced by lower functioning assessed by the GAF (p=0.011), and had substantially higher global negative symptom scores, as assessed by the SANS (p<0.001). These probands also scored significantly higher across all individual SANS subscales, with particular elevations noted for avolition and anhedonia (p<0.001). Although no significant difference was noted for the global SAPS score, the thought disorder subscale was significantly higher in probands with a positive family history as well (p=0.003). Not surprisingly, these probands also revealed slightly lower levels of education on average than their family history negative counterparts (13.4 versus 14.1 years, p=0.023).
Discussion
Substantial heritability of both PPI and P50 gating was observed for multiply affected families with a positive family history. Conversely, neither PPI nor P50 gating displayed any evidence of heritability in the subset of singleton families, in which the probands likely represent sporadic cases of schizophrenia. Pulse-alone startle magnitude showed consistent heritability estimates across all families. While P50 S1 also demonstrated significant heritability across all families, a dramatic increase in S1 heritability was observed for the positive family history subset. This may suggest that both sensory reactivity and gating are more heritable in schizophrenia families with a positive family history. Alternatively, the heritability of P50 gating in the multiply affected families may be driven, at least in part, by the increased heritability of P50 S1 in this subset, as these two measures are highly correlated (r=0.7, p<0.001). To our knowledge, this is the first report of differential heritability of PPI and P50 gating in schizophrenia mediated by family history. However, previous studies have suggested the potential impact of positive family history on P50 gating, reporting greater deficits for subjects with both schizotypal personality disorder and a first-degree relative with schizophrenia compared with healthy subjects and those at risk based solely on clinical criteria (
12,
13).
Other studies of PPI in both schizophrenia families and healthy twins have shown heritability estimates in the range of 45%−50% (
29–
31), which are consistent with the heritability estimate of 47% obtained here for the COGS-1 multiply affected families. The heritability of PPI in the entire COGS-1 sample was lower, at 29% (
26), which is likely due to the combined influence of the stronger genetic transmission of genes underlying PPI in the multiply affected families and to the apparent lack of genetic transmission in the singleton families on the heritability of this endophenotype. Prior heritability estimates for pulse-alone startle magnitude range from 67% to 70% in schizophrenia families and healthy twins, respectively (
29,
30), and show reasonable consistency with the estimates obtained for the entire COGS-1 sample (62%), as well as across the subsets stratified by family history (54%−60%). Taken together, these data suggest that variation in startle magnitude may be generally and highly heritable in humans, and that the inhibition observed in the prepulse paradigm is of particular relevance to schizophrenia, with a substantial genetic component.
P50 gating measured as the difference between the conditioning and test amplitudes has been shown to have robust psychometric properties (e.g., reliability) (
47) and a heritability of 41%−46% in healthy twins (
32). While the estimated heritability of P50 gating was substantially lower (20%) in the entire COGS-1 cohort (
26), it rose to 36% in the multiply affected families, demonstrating greater consistency with previous estimates derived from healthy subjects. This estimate is also consistent with the estimate of 33% from the Bipolar and Schizophrenia Network on Intermediate Phenotypes (BSNIP) study of schizophrenia and bipolar families (
48). Prior assessments of P50 S1 and S2 in healthy twins have revealed heritability estimates of 58%−61% and 52%−56%, respectively (
32). Heritability estimates of S1 and S2 in the entire COGS-1 sample were lower, at 39% and 27%, respectively (
26), consistent with those observed in BSNIP (43% and 23%, respectively) (
48). However, we observed a dramatic increase in the heritability of S1 for the multiply affected subset (80%), while the heritability of this measure in the singleton subset remained consistent with the complete sample (35%). By contrast, the heritability of S2 remained fairly constant between subsets. Some studies have suggested that an increased P50 difference score, which is indicative of poor gating, is related more to a diminished response to the conditioning stimulus (S1) in schizophrenia patients rather than to deficient gating of the response to the test stimulus (S2) (
49–
51). Collectively, these data suggest that P50 gating is a generally heritable human trait with a substantial genetic component, that individual variability in gating is largely driven by S1, and that both measures specifically detect deficits in schizophrenia patients that appear to segregate with illness in multiply affected families.
The observed heritability for PPI and P50 gating in healthy twins (
27–
29,
32) may at first seem at odds with the lack of heritability in the singleton families. While the singleton probands themselves would be expected to show deficits, their unaffected relatives would instead reveal a pattern of normal variation in these endophenotypes. Thus, it may be that the deficits in the probands from the singleton families do not aggregate in a consistent manner with the normal variation in their unaffected relatives, which may obscure the heritability of these measures in this subset.
PPI and P50 gating deficits represent schizophrenia-linked biomarkers that can even be identified in subsyndromal populations (
10–
13). Our results provide further evidence to suggest that gating deficits do in fact represent a core feature of schizophrenia, as PPI and P50 gating deficits were consistent between probands regardless of family history. However, schizophrenia probands differed notably in terms of clinical severity based on family history, with significantly increased negative symptom scores and thought disorder observed for probands with a positive family history. Thus, gating deficits appear to be consistent among schizophrenia patients, regardless of family history, but they co-occur with increased clinical severity in patients with a positive family history.
The primary limitation of this study is that the smaller sample sizes of the subsets of the families stratified by family history reduced power to detect heritability below 20%. While this may have affected the heritability estimates of PPI and P50 gating in the singleton families, there was no suggestion of heritability for these two endophenotypes at all in this subset. Additionally, PPI and P50 gating are at least partially normalized by the use of atypical antipsychotic medications in schizophrenia patients (
7,
9,
52–
54). As approximately 85% of all schizophrenia probands were taking atypical antipsychotics at the time of endophenotype testing, this “normalization” effect of atypical antipsychotics complicates the interpretation of heritability data for these endophenotypes in the context of the COGS-1. While the use of atypical antipsychotics may be in part responsible for the lower than expected heritability estimates of PPI and P50 in the complete COGS-1 sample, this effect is independent of family history. The distributions of medication use were comparable between the family history subgroups, with 85% of patients from multiplex families and 87% of patients from singleton families taking atypical antipsychotics at the time of testing. Of the remaining patients, 11% and 10%, respectively, were taking typical antipsychotics, with a minority of subjects not medicated at the time of testing. Thus, the impact of family history on PPI and P50 gating heritability is not confounded by differences in medication use, and neither are the differences in clinical symptom severity that appear to be correlated with family history.
It is noteworthy that the pattern observed here for PPI and P50 sensory registration and gating of substantially higher heritability in the context of increased genetic vulnerability was not seen for the cognitive endophenotypes assessed by COGS-1. This may suggest that the cognitive endophenotypes reflect variation in normal cognitive functioning in the general population in addition to cognitive deficits in schizophrenia, thereby masking the impact of family history. Alternatively, these results may further demonstrate that cognitive endophenotypes are highly polygenic, even in the context of schizophrenia, with many genes contributing small effects to the overall genetic vulnerability (
55). Moreover, the portion of the variation in PPI and P50 measures that relates specifically to schizophrenia may be due to greater gene effects observed only in multiply affected families, as was demonstrated in a multigenerational family study of PPI in schizophrenia (
31). This latter point is also consistent with the suggestion that cognitive deficits are secondary to primary deficits in early sensory information processing. Overall, these results extend to the genetic realm the work of Callaway and Naghdi (
56), who distinguished automatic information processing measures (i.e., PPI, P50 gating) from effortful, controlled processing (i.e., cognition). The neurobiology of automatic responses may be more specific, resulting in a relatively simpler genetic architecture, whereas the controlled processes may be much more complex, with a more widely distributed neural circuit basis and resultantly complex genetic architecture. Indeed, the neural circuit regulation of PPI appears to be relatively constrained within forebrain cortico-striato-pallido-thalamic circuits and their pontine projections (
57–
59), while the neural circuit regulation contributing to learning and memory is viewed to be more widely distributed across cortical networks. Although these results demonstrate a clear impact of family history on the heritability of PPI and P50 gating in schizophrenia, more work in this area is needed to identify the causal genes. Furthermore, because all schizophrenia probands demonstrated PPI and P50 gating deficits, regardless of family history, schizophrenia patients of presumably sporadic origin may harbor de novo mutations in the same genes or pathways that underlie the heritability signal for these deficits in multiplex schizophrenia families. Therefore, this familial versus nonfamilial stratification approach to endophenotypes may be useful for resolving the genetic architecture of schizophrenia and may be applicable across many domains of function (
60).