Interrelated neuropsychological and anatomical evidence of hippocampal pathology in the at-risk mental state

Background Verbal learning and memory deficits are frequent among patients with schizophrenia and correlate with reduced magnetic resonance imaging (MRI) volumes of the hippocampus in these patients. A crucial question is the extent to which interrelated structural-functional deficits of the hippocampus reflect a vulnerability to schizophrenia, as opposed to the disorder per se. Method We combined brain structural measures and the Rey Auditory Verbal Learning Test (RAVLT) to assess hippocampal structure and function in 36 never-medicated individuals suspected to be in early (EPS) or late prodromal states (LPS) of schizophrenia relative to 30 healthy controls. Results Group comparisons revealed bilaterally reduced MRI hippocampal volumes in both EPS and LPS subjects. In LPS subjects but not in EPS subjects, these reductions were correlated with poorer performance in RAVLT delayed recall. Conclusions Our findings suggest progressive and interrelated structural-functional pathology of the hippocampus, as prodromal symptoms and behaviours accumulate, and the level of risk for psychosis increases. Given the inverse correlation of learning and memory deficits with social and vocational functioning in established schizophrenia, our findings substantiate the rationale for developing preventive treatment strategies that maintain cognitive capacities in the at-risk mental state.


Introduction
Recent meta-analyses of magnetic resonance imaging (MRI) studies of the medial temporal lobe (MTL) in patients with first-episode psychosis (Steen et al. 2006 ;Vita et al. 2006 ;Vita & de Peri, 2007) and nonpsychotic first-degree relatives of schizophrenia patients (Boos et al. 2007) document significant decreases in hippocampal volume. The MTL is critical for declarative episodic memory (Scoville & Milner, 1957 ;Squire et al. 2004), and in schizophrenia patients, lower scores on neuropsychological measures of declarative episodic memory correlate with reduced MRI volumes of the hippocampus (Kuroki et al. 2006 ;Nestor et al. 2007). While Kraepelin (1919) and Bleuler (1911) considered memory functions to be relatively preserved in schizophrenia, current concepts of the disorder classify deficient declarative episodic memory among the core symptoms of the clinical phenotype (Aleman et al. 1999 ;Boyer et al. 2007) and emphasize its validity as outcome predictor (Green, 1996 ;Niendam et al. 2006). Observations that declarative episodic memory is impaired in schizophrenia patients in a manner similar to patients with surgical MTL lesions, have even led to proposals of a schizophrenic amnesia (McKenna et al. 1990). Evidence from functional MRI studies converges on abnormal associative encoding of arbitrary information in the hippocampus as a potential underlying pathomechanism (Achim & Lepage, 2005 ;Achim et al. 2007).
While interrelated structural-functional abnormalities of the hippocampus have been identified in schizophrenia, the timing of their emergence and the extent to which they are related to vulnerability to the disorder as opposed to psychotic illness itself is unclear. The prevailing research paradigms to identify vulnerability to schizophrenia are the genetic and the clinical at-risk strategy, with the latter targeting at the identification of prodromal symptoms and behaviours (Cannon, 2005). Operationally, the prodrome or at-risk mental state (ARMS) is defined by duration of time, starting with the onset of decline in the baseline level of functioning and ending with the transition to first-episode psychosis (Yung & McGorry, 1996). The average duration of the ARMS is about 3 years across studies (McGlashan, 1996). Accumulating evidence from cross-sectional and longitudinal MRI studies indicates that the ARMS is associated with smaller hippocampal volumes (Pantelis et al. 2003 ;Borgwardt et al. 2007) as well as emerging learning and memory deficits, particularly in the verbal modality (Brewer et al. 2005 ;Keefe et al. 2006 ;Lencz et al. 2006 ;Eastvold et al. 2007).
As the ARMS can be conceptualized as a continuum of progressive accrual of morbidity, that culminates in first-episode psychosis (Keshavan et al. 2005), hippocampal volume reduction should be paralleled by a progressive worsening of verbal learning and memory, as prodromal symptoms and behaviours accumulate, and the level of risk increases. To test this hypothesis within a cross-sectional study design, we used structural MRI and the Rey Auditory Verbal Learning Test (RAVLT ; Rey, 1941 ; German adaptation by Helmstaedter et al. 2001) to investigate 36 nevermedicated individuals suspected to be in early (EPS) or late prodromal states (LPS) relative to 30 healthy controls.

Subjects
The present work was conducted as part of the early detection and intervention programme of the German Research Network on Schizophrenia (GRNS ; Hafner et al. 2004). Subjects with symptoms suggestive of either EPS or LPS were recruited as previously described (Hafner et al. 2004). In brief, subjects with suggestive complaints were screened by general practitioners, counselling services or secondary healthcare providers using the 17-item Early Recognition Inventory/Interview for the Retrospective Assessment of the Onset of Schizophrenia (ERIraos) checklist, which was created for this purpose (Hafner et al. 2004). Subjects scoring o6 points were referred to the early recognition and intervention centres established at the Departments of Psychiatry, at the Universities of Bonn and Cologne, for detailed assessment with the 110-item ERIraos symptom list (Hafner et al. 2004), which includes items derived from the Bonn Scale for the Assessment of Basic Symptoms (BSABS ; Gross & Huber, 1985) and from the Interview for the Retrospective Assessment of the Onset of Schizophrenia (IRAOS ; Hafner et al. 1992). EPS diagnostic criteria were based either on the presence of basic symptoms with a positive predictive value of >0.7 and a specificity of >0.85 for the transition to firstepisode psychosis (Klosterkotter et al. 2001) and/or a reduction of the Global Assessment of Functioning (GAF) in conjunction with the presence of a firstdegree relative with a psychotic disorder or a history of obstetric complications. While other studies also label individuals to be clinically at ultra-high-risk if they have a genetic risk plus functional decline (McGorry et al. 2002), the additional use of basic symptoms allows prediction of transition to firstepisode psychosis already in an earlier prodromal state (Hafner et al. 2004). In contrast, subjects experiencing attenuated positive symptoms (APS) or brief limited intermittent psychotic symptoms (BLIPS) were considered to be in the LPS, which is consistent with conventional ultra-high-risk criteria used in clinical studies (McGorry et al. 2002). A detailed synopsis of inclusion and exclusion criteria is presented in Table 1.
Based on these criteria, we recruited 36 nevermedicated ultra-high-risk subjects, of whom 20 (56 %) (8 female, 12 male ; mean age 27.3¡5.1 years ; age range 21-45 years) were assigned to the EPS group and 16 (44 %) (8 female, 8 male ; mean age 26.8¡6.2 years ; age range 19-41 years) to the LPS group. The mean longitudinal follow-up examination time was 18¡3 months. During this period, three EPS subjects (15 %) and five LPS subjects (31 %) converted to schizophrenia according to DSM-IV diagnostic criteria. This preliminary conversion rate is in agreement with the predictive validity of EPS and LPS diagnostic criteria reported in prospective studies. According to these studies, EPS criteria predict psychosis onset within 12 months in 19 % of cases and within 64 months in 70 % of cases (Klosterkotter et al. 2001), whereas LPS criteria predict psychosis onset within 12 months in 30-54 % of cases . The healthy control (CTL) group comprised 30 subjects (7 female, 23 male ; mean age 28.2¡6.4 years ; age range 18-40 years) carefully matched for age, education, and sociodemographic area. They were free of personal as well as first-and second-degree family history of DSM-IV Axis I and II disorders as assessed by experienced clinicians (P.F. and S.R.). Subjects were recruited through local advertisement and excluded from participation if they had previous exposure to psychoactive medication, current substance or alcohol abuse or a history of neurological disorder or severe somatic condition. Written informed consent was obtained from all participants prior to the study, and study protocols were approved by the ethics committees of the Medical Faculties of the Universities of Bonn and Cologne.

Psychopathological and neuropsychological instruments
Psychopathological assessment before MRI scanning included the Positive and Negative Syndrome Scale (PANSS), a 30-item rating instrument evaluating the presence/absence and severity of positive, negative, and general psychopathology of schizophrenia (Kay et al. 1987) ; the Montgomery-Å sberg Depression Rating Scale (MADRS), a 10-item semi-structured interview assessing the affective, cognitive, and neurovegetative dimensions of depressive symptomatology (Montgomery & Å sberg, 1979) ; and the GAF, a 100-point numeric scale implemented in DSM-IV Axis V assessment, that provides an index of overall psychological, social, and occupational functioning ( Table 2). Table 1. Synopsis of (A) inclusion criteria and (B) exclusion criteria (Hafner et al. 2004) (A) Inclusion criteria (a) Early prodromal state (EPS) One or more of the following basic symptoms appeared in the last 3 months, several times a week : $ Thought interference or thought perseveration or thought pressure or thought blockage $ Disturbances of receptive language, either heard or read $ Decreased ability to discriminate between ideas and perception, fantasy and true memories $ Unstable ideas of reference (subject-centrism) $ Derealization $ Visual or acoustic perception disturbances and/or Reduction in the Global Assessment of Function score (DSM-IV) of at least 30 points (within the past year) and at least one of the following risk factors : $ First-degree relative with lifetime schizophrenia or schizophrenia-spectrum disorder $ Pre-or perinatal complications (b) Late prodromal state (LPS) Presence of at least one of the following attenuated positive symptoms (APS) within the last 3 months, appearing several times per week for a period of at least one week : $ Ideas of reference $ Odd beliefs or magical thinking $ Unusual perceptual experience $ Odd thinking or speech $ Suspiciousness or paranoid ideation and/or Brief limited intermitted psychotic symptoms (BLIPS), defined as appearance of one of the following symptoms for <1 week (interval between episodes at least 1 week), resolving spontaneously : $ Hallucinations $ Delusions $ Formal thought disorder $ Gross disorganized or catatonic behavior (B) Exclusion criteria APS or BLIPS (early prodromal state) Present or past diagnosis of schizophrenia, schizophreniform, schizoaffective, delusional or bipolar disorder according to DSM-IV Present or past diagnosis of brief psychotic disorder according to DSM-IV with a duration o1 week or within the last 4 weeks regardless of its duration Diagnosis of delirium, dementia, amnestic or other cognitive disorder, mental retardation, psychiatric disorders due to somatic factors or related to psychotropic substances according to DSM-IV Alcohol or drug abuse within the last 3 months prior to inclusion according to DSM-IV Diseases of the central nervous system (inflammatory, traumatic, epileptic, etc.) Aged <18 yr and >36 yr Neuropsychological assessment before MRI scanning included the Verbaler Lern-und Merkfähigkeitstest (VLMT ; Helmstaedter et al. 2001), a German-language analogon of the RAVLT, that consists of five presentations of a 15-word list (Trials 1-5, List A), followed by a free recall trial of a second word list (List B), and a sixth free recall trial of List A (Trial A6, Retention). Delayed recall was examined with a seventh free recall trial after a 30-min interval (Trial A7, Delayed recall). Recognition was tested by asking the respondent to indicate which of 50 words read aloud were from List A and which were not (List A, Recognition). The RAVLT thus provides measures of immediate memory, efficiency of learning, effects of interference, recall following short and long delay periods, and recognition. Schizophrenia patients who have better verbal memory on the RAVLT are likely to perform better on tasks of independent living, underscoring the ecological validity of the RAVLT (Strauss et al. 2006). Verbal IQ (VIQ) was determined with the Mehrfachwahl-Wortschatz-Intelligenztest (MWT-B ; Lehrl, 2005), a German-language analogon of the Spot-the-Word test (STW), that provides a robust estimate of verbal intelligence based upon lexical decisions (Table 2).

MRI protocol
Participants underwent MRI scanning with a 1.5-T MR system (Philips Gryroscan ACS-NT, Philips Medical Systems, Best, The Netherlands) at the Department of Radiology, University of Bonn (EPS subjects, n=11 ; LPS subjects, n=8) or with an identical magnet at the Department of Radiology, University of Cologne (EPS subjects, n=9 ; LPS subjects, n=8). A T1-weighted 3-dimensional turbo gradient echo (TFE) sequence generated 150 contiguous, 1-mm sagittal slices. Imaging parameters were time-to-echo, 1.51 ms ; time-to-repetition, 25 ms ; flip angle, 20x, matrix size, 256r256 ; field of view, 24r24 cm matrix ; acquisition resolution, 1r1.51r1 mm ; reconstructed resolution, 1r1r1 mm. Each MR scanner was calibrated fortnightly using the same proprietary phantom to ensure stability and accuracy of measurements. There were no significant differences in brain structural measures between MR scanners (whole-brain volume, t 64 = x1.164, p=0.249 ; total left hippocampus, t 64 =x1.211, p=0.230 ; total right hippocampus, t 64 =x0.794, p=0.430). MRI data were analysed using Analyze 7.0 (Mayo Clinic, Rochester, MN, USA). A code was used to ensure confidentiality and blind rating of data. Hippocampal volumes were estimated using a manual tracing technique and defined anatomical criteria (Cook et al. 1992). Hippocampal boundaries were defined as posterior (slice with greatest length of continuous fornix) ; medial (open end of the hippocampal fissure posteriorly, uncal fissure in the hippocampal body and medial aspect of ambient gyrus anteriorly) ; lateral (temporal horn of lateral ventricle) ; inferior (white matter inferior to the hippocampus) ; superior (superior border of hippocampus) ; and anterior (alveus was used to differentiate hippocampal head from amygdala). Hippocampal length was estimated as the number of slices per hippocampus. One rater (J.L.) performed all hippocampal tracings included in the analyses. The intra-class correlation for intra-rater reliability (J.L.), determined by blindly retracing 20 randomly selected images, was 0.95 and for inter-rater reliability (J.L. and R.H.) was 0.80 in 10 subjects. Hippocampal volumes were divided by whole-brain volumes to correct (normalize) for inter-subject variation in head size (Free et al. 1995 ; but see also Arndt et al. 1991). Whole-brain volumes were obtained through an in-house developed algorithm (R.T.) following automated tissue segmentation in Statistical Parametric Mapping (SPM5 ; Wellcome Trust Centre for Neuroimaging, London, UK).

Data analysis
Differences in age, sex, VIQ, RAVLT performance, and whole-brain volumes were assessed by a multivariate analysis of variance (ANOVA) defined by a three-level (CTL, EPS, LPS) between-subjects group factor. Differences in hippocampal volumes were assessed by a repeated-measures ANOVA defined by a three-level (CTL, EPS, LPS) between-subjects group factor and a two-level (left and right hippocampal volumes) within-subjects laterality factor, followed by two-tailed two-sample post-hoc t tests to determine the source of significance. As no consensus has evolved on the validity of ratio measures in MRI volumetric studies, group comparisons were calculated on whole-brain volume adjusted (first analysis) and unadjusted hippocampal volumes (second analysis) (Arndt et al. 1991 ;Free et al. 1995). The assigned 0.05 a level was Bonferroni-corrected to account for an inflation of the type I error rate attributable to multiple post-hoc testing. Effect sizes were quantified by calculating the values of Cohen's d. Associations among brain structural, neuropsychological, and psychopathological measures were assessed using Pearson correlations.  (Fig. 1). To ensure that these results were not due to differences in hippocampal length (defined by the number of slices traced per hippocampus), subsidiary analyses were carried out on hippocampal length, covarying for whole-brain volume. However, there were no significant between-group differences in left hippocampal length ( Table 3 presents Pearson correlations between RAVLT subscores and anatomical measures. Within the LPS group, smaller right hippocampal volume correlated significantly with lower RAVLT delayed recall (r= 0.680, p=0.004). No further associations among neuropsychological, psychopathological, and brain structural measures were present. In addition, there was no greater neuropsychological or hippocampal volume deficit in converters (n=8) relative to the majority of subjects (n=28) who remained at-risk at follow-up.

Discussion
This study was designed to examine brain-behaviour relationships in the ARMS, i.e. before the onset of schizophrenia and without the potential confounds of illness chronicity and antipsychotic drug action. Consistent with our a priori hypothesis, anatomical measures revealed 9.6 % reductions of MRI hippocampal volumes across EPS and LPS subjects, with mean weighted effect sizes of 1.26 on the left side and 1.40 on the right side. These results are in line with recent meta-analyses that document 8 % smaller hippocampi in first-episode schizophrenia patients, with effect sizes ranging from x0.06 to 1.21 on the left side and from 0.10 to 1.32 on the right side (Steen et al. 2006 ;Vita et al. 2006 ;Vita & de Peri, 2007). In LPS subjects but not in EPS subjects, a 9.2 % deficit in RAVLT delayed recall was correlated with reduced MRI hippocampal volumes, suggesting progressive and interrelated structural-functional pathology of the hippocampus, as prodromal symptoms and behaviours accumulate, and the level of risk for psychosis increases.
As a neuropsychological measure of verbal learning and memory, the RAVLT is thought to be supported by a widely distributed neural circuitry including frontotemporal subregions (Fletcher et al. 1997). A key function of this circuitry is to bind verbal items into specific episodes that can be later retrieved and consciously recollected (Squire et al. 2004). For the LPS subjects in this study, their specific deficit in RAVLT delayed recall could then reflect disturbances of episodic encoding-and/or retrieval-related frontotemporal functioning. Recent evidence of abnormal hippocampal responses during associative encoding  Left (first column, black symbols) and right (second column, grey symbols) unadjusted hippocampal volumes for each subject group. CTL, Controls ; EPS, ultra-high-risk subjects suspected to be in the early prodromal state ; LPS, ultra-high-risk subjects suspected to be in the late prodromal state. of arbitrary information in first-episode schizophrenia patients (Achim et al. 2007) as well as reports of preserved retrieval-induced forgetting in patients with chronic schizophrenia (Nestor et al. 2007) suggest diminished encoding efficiency in the hippocampus as a potential underlying pathomechanism. However, we cannot rule out that impaired RAVLT delayed recall in LPS subjects is primarily caused by a failure to retrieve, among competing, interfering alternatives, and to bring into consciousness, an intact representation of the original learning episode (Boyer et al. 2007).
Together with related findings in schizophrenia patients (Kuroki et al. 2006 ;Nestor et al. 2007) and non-psychotic first-and second-degree relatives of schizophrenia patients (Lymer et al. 2006), our results propose a link between psychometric measures of verbal learning and memory and the degree of integrity and volume of the hippocampus. However, the structural-functional correlations calculated in this study are based upon rather small subsets of EPS and LPS subjects, and thus the derived empirical relationship is best viewed as preliminary and exploratory. Moreover, we cannot exclude the confounding influence of heterogeneity in MRI acquisition due to the use of two MR scanners.
The observation that hippocampal pathology varies as a function of risk for psychosis may conflict with concepts of schizophrenia as a static encephalopathy. Substantial support for a progressive degenerative pathology comes from longitudinal morphometric studies which revealed an increase in frontotemporal grey-matter volume reduction in ultra-high-risk subjects whose illness proceeded from the ARMS to first-episode psychosis (Pantelis et al. 2003). Moreover, evidence from molecular neuroimaging studies suggests neurochemical abnormalities in ultra-highrisk subjects that accumulate with increasing risk for psychosis (Hurlemann et al. 2008 ; see also Fusar-Poli et al. 2007). Furthermore, there is evidence for continued and interrelated changes in brain structural and bioelectrical indices after onset of schizophrenia (Salisbury et al. 2007). In light of these findings, dynamic models of emerging schizophrenia have been conceptualized, wherein an early (pre-or perinatal) static lesion is thought to synergistically interact with later progressive frontotemporal grey-matter lesion to express the disorder (Pantelis et al. 2003 ;Salisbury et al. 2007). The underlying biological mechanisms are not clear. However, post-mortem histological findings suggest that smaller hippocampi in schizophrenia patients do not necessarily result from decreases in total cell number (Heckers & Konradi, 2002 ;Walker et al. 2002 ;Hurlemann et al. 2005) but could reflect a dendritic/synaptic pathology of the hippocampus (Feinberg, 1983 ;McGlashan & Hoffmann, 2000). While our findings of impaired verbal learning and memory in LPS but not in EPS subjects would be compatible with a dendritic/synaptic hippocampal degeneration near psychosis onset, a progressive and interrelated reduction of hippocampal volume and function within the same individuals over time remains to be documented by future prospective longitudinal studies.
In portraying the ARMS as a period of progressive and interrelated structural-functional hippocampal pathology, our results may contribute to the validation of a diagnostic distinction between EPS and LPS subjects (Hafner et al. 2004) and to the clarification of declarative episodic memory deficits in the ARMS. Such clarification is critical for therapeutic efforts to forestall illness progression and prevent cognitive decline from increasing. As impaired declarative episodic memory in schizophrenia is significantly associated with poorer social and vocational competences (Green, 1996 ;Niendam et al. 2006), cognitive dysfunctioning in the ARMS has become a prime target for the development of preventive treatments. Specifically, cognitive remediation programmes may be able to capitalize on encoding and retrieval strategies to enhance declarative episodic memory in LPS subjects. In summary, our data implicate progressive and interrelated structural-functional pathology of the hippocampus as an index of increased risk for schizophrenia.

Declaration of Interest
None.