Introduction
Neurobiological models of schizophrenia have emphasized a critical role for hippocampal dysfunction (Tamminga et al. Reference Tamminga, Stan and Wagner2010). This was recently supported by a large prospective meta-analysis from the ENIGMA consortium that revealed smaller hippocampus volume in 2028 patients with schizophrenia compared with 2540 healthy controls (van Erp et al. Reference van Erp, Hibar, Rasmussen, Glahn, Pearlson, Andreassen, Agartz, Westlye, Haukvik, Dale, Melle, Hartberg, Gruber, Kraemer, Zilles, Donohoe, Kelly, McDonald, Morris, Cannon, Corvin, Machielsen, Koenders, de Haan, Veltman, Satterthwaite, Wolf, Gur, Gur, Potkin, Mathalon, Mueller, Preda, Macciardi, Ehrlich, Walton, Hass, Calhoun, Bockholt, Sponheim, Shoemaker, van Haren, Pol, Ophoff, Kahn, Roiz-Santiañez, Crespo-Facorro, Wang, Alpert, Jönsson, Dimitrova, Bois, Whalley, McIntosh, Lawrie, Hashimoto, Thompson and Turner2016). Moreover, hippocampal abnormalities have also been identified among individuals with psychotic bipolar disorder (Frey et al. Reference Frey, Andreazza, Nery, Martins, Quevedo, Soares and Kapczinski2007) and those vulnerable to psychosis (Fusar-Poli et al. Reference Fusar-Poli, Perez, Broome, Borgwardt, Placentino, Caverzasi, Cortesi, Veggiotti, Politi, Barale and McGuire2007), suggesting they play a critical role in psychotic phenomenology and do not respect traditional diagnostic boundaries.
Although antipsychotic treatment has been typically linked to subcortical dopaminergic changes, medication effects have also been hypothesized to be secondary to a loss of inhibitory signaling (i.e. hyperactivity) within dopamine systems regulated by the hippocampus (Lodge & Grace, Reference Lodge and Grace2011; Perez & Lodge, Reference Perez and Lodge2014). Additionally, this theory has been used to guide animal models of schizophrenia (Peleg-Raibstein et al. Reference Peleg-Raibstein, Feldon and Meyer2012). Better understanding the relationship between antipsychotic treatment and hippocampal structure/function is critical as this work has important implications for both patients’ ability to flexibly generalize past experiences within new environments (Shohamy et al. Reference Shohamy, Mihalakos, Chin, Thomas, Wagner and Tamminga2010) and declarative memory deficits that have been linked to memories with psychotic content (Tamminga et al. Reference Tamminga, Stan and Wagner2010).
Considerable evidence indicates that antipsychotic treatment may be associated with changes in hippocampal structure and function. Cross-sectional studies reported that antipsychotic dosage was associated with abnormal hippocampal shape (Zierhut et al. Reference Zierhut, Graßmann, Kaufmann, Steiner, Bogerts and Schiltz2013) and less volume (Yang et al. Reference Yang, Wu, Lu, Bai and Gao2015) in patients with schizophrenia. Although the use of longitudinal neuroimaging could more directly facilitate our understanding of how antipsychotics affect hippocampal morphology, findings have been highly inconsistent. Rizos et al. (Reference Rizos, Papathanasiou, Michalopoulou, Laskos, Mazioti, Kastania, Vasilopoulou, Nikolaidou, Margaritis, Papageorgiou and Liappas2014) reported total hippocampal volume reductions in patients following monotherapy with second-generation antipsychotics and Ebdrup et al. (Reference Ebdrup, Skimminge, Rasmussen, Aggernaes, Oranje, Lublin, Baaré and Glenthøj2011) reported hippocampal volume loss among patients treated with quetiapine. This work has been supported by animal studies linking hippocampal volume reduction with olanzapine treatment (Barr et al. Reference Barr, Wu, Wong, Hercher, Töpfer, Boyda, Procyshyn, Honer and Beasley2013). In contrast, some studies reported that atypical antipsychotics are associated with less longitudinal change in hippocampal volume (Koolschijn et al. Reference Koolschijn, van Haren, Cahn, Schnack, Janssen, Klumpers, Hulshoff Pol and Kahn2010), especially when compared with typical antipsychotics such as haloperidol (Mamah et al. Reference Mamah, Harms, Barch, Styner, Lieberman and Wang2012). Other studies, however, reported that longitudinal changes in hippocampal volume did not differ significantly in patients treated with olanzapine v. haloperidol (Panenka et al. Reference Panenka, Khorram, Barr, Smith, Lang, Kopala, Vandorpe and Honer2007; McClure et al. Reference McClure, Styner, Maltbie, Lieberman, Gouttard, Gerig, Shi and Zhu2013).
The hippocampus is a highly heterogeneous structure comprised of multiple subregions that have different afferent and efferent projections relevant to the neurobiology of psychosis and its pharmacological treatment (Tamminga et al. Reference Tamminga, Stan and Wagner2010; Benes, Reference Benes2015). Distinguishing different parts of the hippocampus from cellular features has been difficult to accomplish with the majority of in vivo magnetic resonance (MR) imaging sequences widely used today. Although some protocols have been developed to resolve the internal architecture of the hippocampus, they require higher (e.g. 7 T) field strengths (e.g. Wieshmann et al. Reference Wieshmann, Symms, Mottershead, MacManus, Barker, Tofts, Revesz, Stevens and Shorvon1999; Thomas et al. Reference Thomas, Welch, Niederhauser, Whetsell, Anderson, Gore, Avison and Creasy2008; Kerchner et al. Reference Kerchner, Hess, Hammond-Rosenbluth, Xu, Rabinovici, Kelley, Vigneron, Nelson and Miller2010) not typically available outside of academic centers. In addition, several prior investigations of hippocampal subregions are limited to the body of the hippocampus where they can be most clearly identified (Mueller & Weiner, Reference Mueller and Weiner2009; Mueller et al. Reference Mueller, Laxer, Barakos, Cheong, Garcia and Weiner2009, Reference Mueller, Schuff, Yaffe, Madison, Miller and Weiner2010; Wang et al. Reference Wang, Neylan, Mueller, Lenoci, Truran, Marmar, Weiner and Schuff2010). Other strategies incorporate surface shape analysis by projecting post-mortem atlases onto three-dimensional (3D) renderings of the entire hippocampus (Frisoni et al. Reference Frisoni, Ganzola, Canu, Rüb, Pizzini, Alessandrini and Thompson2008; Xie et al. Reference Xie, Alcantara, Amenta, Fletcher, Martinez, Persianinova, DeCarli and Carmichael2009; Boccardi et al. Reference Boccardi, Ganzola, Rossi, Sabattoli, Laakso, Repo-Tiihonen, Vaurio, Könönen, Aronen, Thompson, Frisoni and Tiihonen2010; Cole et al. Reference Cole, Toga, Hojatkashani, Thompson, Costafreda, Cleare, Williams, Bullmore, Scott, Mitterschiffthaler, Walsh, Donaldson, Mirza, Marquand, Nosarti, McGuffin and Fu2010), and are therefore limited to subregions located on the external surface of the hippocampus. Some studies have ‘unfolded’ the hippocampus to translate its 3D subregions into two-dimensional space (Ekstrom et al. Reference Ekstrom, Bazih, Suthana, Al-Hakim, Ogura, Zeineh, Burggren and Bookheimer2009; Donix et al. Reference Donix, Burggren, Suthana, Siddarth, Ekstrom, Krupa and Bookheimer2010). Hippocampal subregion delineation has also incorporated manual techniques (Mueller et al. Reference Mueller, Stables, Du, Schuff, Truran, Cashdollar and Weiner2007; La Joie et al. Reference La Joie, Fouquet, Mézenge, Landeau, Villain, Mevel, Pélerin, Eustache, Desgranges and Chételat2010; Malykhin et al. Reference Malykhin, Lebel, Coupland, Wilman and Carter2010), automated techniques (Van Leemput et al. Reference Van Leemput, Bakkour, Benner, Wiggins, Wald, Augustinack, Dickerson, Golland and Fischl2009; Bonnici et al. Reference Bonnici, Chadwick, Kumaran, Hassabis, Weiskopf and Maguire2012), or a combination of these two approaches (Yushkevich et al. Reference Yushkevich, Wang, Pluta, Das, Craige, Avants, Weiner and Mueller2010). Such approaches are highly valuable given their potential to accommodate high throughput, but the majority of these techniques did not assess inter-rater and/or intra-rater reliability.
Manual delineation remains the ‘gold standard’ for quantifying the hippocampus and its subregions from MR images, but few studies have successfully implemented such an approach with high reliability using a T1-weighted contrast. We previously reported a reliable set of manual delineation criteria to measure the anterior cornu ammonis (CA) 1–3, posterior CA1–3, subiculum, dentate gyrus (DG)/CA4, entorhinal cortex and fimbria at 3 T with high inter- and intra-rater reliability assessed using both intra-class correlation (ICC) coefficients and Dice indices (Rhindress et al. Reference Rhindress, Ikuta, Wellington, Malhotra and Szeszko2015). In this approach we relied on anatomical rather than cytoarchitectural features where they would be more evident. The goal of the current study was to examine volumetric changes in hippocampal subregion morphology following antipsychotic treatment in first-episode psychosis. Based on prior published work (Barr et al. Reference Barr, Wu, Wong, Hercher, Töpfer, Boyda, Procyshyn, Honer and Beasley2013; Haukvik et al. Reference Haukvik, Westlye, Mørch-Johnsen, Jørgensen, Lange, Dale, Melle, Andreassen and Agartz2015; Kawano et al. Reference Kawano, Sawada, Shimodera, Ogawa, Kariya, Lang, Inoue and Honer2015) and neurobiological models of hippocampal dysfunction (Tamminga et al. Reference Tamminga, Stan and Wagner2010) we hypothesized that antipsychotic treatment would be associated with volume reductions within CA4 and the DG.
Method
All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Subjects
A total of 29 patients experiencing a first-episode of psychosis were recruited from admissions to the in-patient service at The Zucker Hillside Hospital in Glen Oaks, NY and were enrolled in a National Institute of Mental Health-funded double-blind randomized controlled trial comparing aripiprazole, a dopamine and serotonin 5HT1A partial agonist, v. risperidone, a dopamine and serotonin antagonist (R01MH060004). All patients were required to have 2 weeks or fewer of cumulative lifetime exposure to antipsychotics to enter the clinical trial and received a physical examination and laboratory screening to rule out medical causes for their initial psychotic episode. All patient (lifetime) diagnoses were based on the Structured Clinical Interview for DSM-IV Axis I disorders (SCID; First et al. Reference First, Spitzer, Gibbon and Williams1998) supplemented by information from clinicians and, when available, family members. First-episode patients met Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria for schizophrenia (n = 14 for undifferentiated and n = 7 for paranoid), schizophreniform disorder (n = 4), schizo-affective disorder (n = 1) or psychosis not otherwise specified (n = 3).
Patients had, on average, a total lifetime exposure of 6.7 days (s.d. = 8.1 days) of antipsychotic treatment prior to the baseline scan; data were unavailable for one patient. Patients were scanned, on average, within 2.9 days (s.d. = 8.3 days) of entry into the clinical trial. Of the patients, 10 were antipsychotic drug-naïve at the time of the baseline scan. On the date of the baseline scan, nine of the 29 patients in this study received medications other than antipsychotics. Six patients received lorazepam, with dosages ranging from 0.5 to 6 mg. One patient received both lorazepam (1 mg) and benztropine (4 mg). One patient was given loratadine (10 mg), a multivitamin, and a nicotine patch. One patient was given a 15% gel solution of azelaic acid. Mean age at first psychotic symptoms was 19.4 years (s.d. = 2.9 years); data were unavailable for two patients.
Healthy volunteers (n = 29) were recruited to participate in imaging studies being conducted at the North Shore – Long Island Jewish Hospital System (NSLIJHS, now Northwell Health), using advertisements in local newspapers and through word of mouth. All healthy controls were administered the SCID-non-patient edition (SCID-NP; First et al. Reference First, Spitzer, Gibbon and Williams2001) to ensure the absence of an Axis I psychiatric disorder.
In addition, all individuals met the following exclusion criteria: (1) fulfilling DSM-IV criteria for a current substance-induced psychotic disorder, psychotic disorder due to a general medical condition, or mental retardation; (2) having a serious medical condition known to have an impact on the brain; (3) having any medical condition that required treatment with a medication with psychotropic effects; (4) having a risk of suicidal or homicidal behavior; (5) MR imaging contraindications; and (6) pregnancy. All procedures were approved by the NSLIJHS Institutional Review Board and written informed consent was obtained either from study participants or a legal guardian in the case of minors. All minors provided written assent to participate in the study.
Treatment trial antipsychotic titration schedule
Research psychiatrists followed a flexible dosing titration schedule. The initial daily dose for patients in the treatment trial was 5 mg for aripiprazole and 1 mg for risperidone. The dose was initially increased after 3 days of treatment and then further adjustments were made every 1 to 3 weeks until the patient improved, developed side effects that precluded a dose increase, or reached a maximum daily dose of 30 mg of aripiprazole or 6 mg of risperidone. All patients were receiving either aripiprazole or risperidone at the time of the second scan except one patient who discontinued treatment 2 weeks prior. Lorazepam was prescribed for anxiety or agitation. Patients were not allowed to receive antidepressants or mood stabilizers. Cogentin and/or propranolol was prescribed for extrapyramidal symptoms.
Clinical assessments
Patients completed the 18-item Brief Psychiatric Rating Scale – anchored version (BPRS-A; Overall & Gorham, Reference Overall and Gorham1962) and we derived a total score by summing all items. The average BPRS score was 44.2 (s.d. = 9.4) at the time of baseline scanning and 27.7 (s.d. = 7.3) at follow-up.
Chlorpromazine and olanzapine equivalents
To examine the relationship between volumetric changes and antipsychotic exposure over the course of the study we converted each patient's total antipsychotic exposure between the scans into chlorpromazine equivalents using techniques described by Woods (Reference Woods2003). To confirm our results, these values were also converted into olanzapine equivalents using the method recommended by Leucht et al. (Reference Leucht, Samara, Heres, Patel, Furukawa, Cipriani, Geddes and Davis2015).
MR imaging and image processing
All study participants underwent MR imaging examinations at North Shore University Hospital at the baseline and 12-week time points. Images were acquired in the coronal plane using a 3D spoiled gradient sequence (repetition time = 7.5 ms, echo time = 3 ms, matrix = 256 × 256, field of view = 240 mm) on a single 3 T scanner (GE Signa HDx; General Electric, USA), producing 216 contiguous images (slice thickness = 1 mm) through the whole head. All scans were reviewed by a neuroradiologist for gross neuroanatomic abnormalities (e.g. tumors) that would preclude participation in this study. Scans were also reviewed by a member of the research team for quality control. All scans were aligned along the anterior and posterior commissures using previously published software (Ardekani & Bachman, Reference Ardekani and Bachman2009) and resampled into a 512 × 512 matrix yielding voxel dimensions of 0.47 × 0.47 × 1 mm. Measurements were conducted using ITK-SNAP (Yushkevich et al. Reference Yushkevich, Piven, Hazlett, Smith, Ho, Gee and Gerig2006) and completed by a single operator (K.R.) blind to time point and group membership.
Intracranial volume
Intracranial volumes were computed using the FreeSurfer image analysis suite (version 5.1), which is documented and freely available for download online (http://surfer.nmr.mgh.harvard.edu). The technical details for procedures are described in prior publications (e.g. Fischl & Dale, Reference Fischl and Dale2000; Fischl et al. Reference Fischl, van der Kouwe, Destrieux, Halgren, Ségonne, Salat, Busa, Seidman, Goldstein, Kennedy, Caviness, Makris, Rosen and Dale2004; Reuter et al. Reference Reuter, Schmansky, Rosas and Fischl2012). Mean intracranial volume was 1529 cm3 (s.d. = 257) for patients and 1562 cm3 (s.d. = 166) for healthy volunteers.
Hippocampal delineation criteria
The approach for manual delineation of hippocampal subregions and its reliability analyses have been described in detail previously (Rhindress et al. Reference Rhindress, Ikuta, Wellington, Malhotra and Szeszko2015). Briefly, this approach uses anatomical landmarks and differences in the images’ intensities to identify the location of six subregions throughout the hippocampal formation (i.e. anterior CA1–3, posterior CA1–3, subiculum, DG/CA4, entorhinal cortex and fimbria) as illustrated in Figs 1 and 2. Tracing of the subregions was conducted in the coronal plane while simultaneously viewing the images in the axial and sagittal planes. Tracings began on the most posterior slice of the hippocampus and progressed anteriorly. It took approximately 2.5 h for an operator to manually delineate all subregions within a single hippocampus.
Hippocampal subregions delineated on a coronal image in the left hemisphere. DG, Dentate gyrus; CA, cornu ammonis.
Three-dimensional mesh of an individual hippocampal formation produced by ITK-SNAP and viewed: (a) anteriorly (A); (b) posteriorly (P); (c) superiorly (S); (d) inferiorly (I); (e) medially (left hemisphere only); and (f) laterally (left hemisphere only). The color scheme for the subregions is as follows: anterior cornu ammonis (CA) 1–3, red; posterior CA1–3, orange; subiculum, green; dentate gyrus/CA4, blue; entorhinal cortex, yellow; and fimbria, cyan. L, Left; R, right.
Two operators manually traced the hippocampal regions of interest in the right and left hemispheres blind to the other operator's work to establish inter-rater reliability. One operator (K.R.) manually traced these regions of interest again approximately 1 week later to determine intra-rater reliability. Reliability was assessed using the Dice index (Dice, Reference Dice1945), which is defined as: 2(A∩B)/(A + B), that is, the size of intersection (agreement) between two labels (A∩B) divided by their average size (A + B)/2. Dice indices range from 0 (no overlap) to 1 (perfect agreement) and were computed for hippocampal subregions using both inter-rater and intra-rater reliability. In addition, ICC coefficients were computed using the Statistical Package for the Social Sciences (SPSS) using a two-way mixed-effects model based on the ICC value for a single operator. Thus, each subject is assessed by both operators, but the operators are the only ones of interest. ICC coefficients were computed based on the premise that a single operator would conduct the tracings and not that the average of the two operators would be used for the measurements. Inter-rater reliability performed using this approach yielded an average ICC of 0.85 (range 0.71–0.98, median 0.86) and an average Dice index of 0.75 (range 0.70–0.81, median 0.75), whereas the intra-rater reliability metrics yielded an average ICC of 0.92 (range 0.66–0.99, median 0.97) and an average Dice index of 0.85 (range 0.82–0.90, median 0.85).
Handedness
All individuals completed a modified version of the Edinburgh Inventory. The total number of right- and left-hand items was scored and the laterality quotient was computed according to the following formula: (total R – total L)/(total R + total L). This yielded a range from +1.00 (totally dextral) to −1.00 (totally non-dextral). Mean laterality quotient was 0.74 (s.d. = 0.39) for healthy volunteers and 0.71 (s.d. = 0.40) for patients.
Statistical analyses
Analyses were conducted using the Statistical Package for the Social Sciences (SPSS; v. 11.5, USA). We used analysis of covariance to compare the hippocampal subregions with group as a between-subjects factor. As none of the group × hemisphere × time point interactions were statistically significant for any of the individual hippocampal subregions, right and left hemisphere volumes were combined for analysis. The within-subjects factors included six levels of subregion (anterior CA1–3, posterior CA1–3, subiculum, DG/CA4, entorhinal cortex and fimbria) as well as time point (baseline and follow-up). We included age, sex, handedness and intracranial volume as covariates. Due to the sphericity assumption being violated (epsilon = 0.75) we used the Greenhouse–Geisser correction to amend degrees of freedom (df) in the analyses (Mauchly's W = 0.315 for the region × time point interaction with corresponding χ2 = 58, p < 0.001).
We used curve estimation procedures employing both linear and quadratic models to assess the relationship between subregion volume changes and total antipsychotic exposure between the scans. We set α to 0.05 (two-tailed) for all analyses.
Results
Sample demographics are described in Table 1. The two groups did not differ significantly in distributions of age, sex, handedness or intra-cranial volume. As expected, there was a significant difference in education level between the groups (F 1,56 = 5.64, p = 0.02). There was no significant difference in patients being treated with risperidone v. aripiprazole in either chlorpromazine or olanzapine equivalents between the two time points. Descriptive statistics for each of the hippocampal subregions at the time of the baseline and follow-up scans are provided in Table 2 for descriptive purposes only, with individual change scores regarding the subiculum and DG/CA4 illustrated in Fig. 3.
Scatterplots of change scores for patients and healthy volunteers in the subiculum and dentate gyrus/cornu ammonis (CA) 4. Horizontal lines represent mean values.
Sample demographics
s.d., Standard deviation.
Baseline and follow-up volumes (mm3) for patients and healthy volunteers a
s.d., Standard deviation; df, degrees of freedom; A CA1–3, anterior cornu ammonis 1–3; P CA1–3, posterior cornu ammonis 1–3; DG/CA4, dentate gyrus/cornu ammonis 4; EC, entorhinal cortex.
a Subregion volumes were computed by summing right and left hemisphere volumes.
The main finding that distinguished the groups was a significant group × time × subregion interaction (F 3.73,194 = 3.29, p = 0.01). Post-hoc analyses revealed significant group × time interactions for the subiculum (F 1,52 = 4.79, p = 0.03) and DG/CA4 (F 1,52 = 7.10, p = 0.01). Paired-samples t tests were used to compare each hippocampal subregion volume at baseline with its respective volume at 12 weeks. Patients demonstrated a significant increase in subiculum volume (mean paired difference = 32 mm3) as well as a significant volumetric decrease in the DG/CA4 subregion (mean paired difference = −68 mm3). Changes in subiculum and DG/CA4 volume did not correlate significantly with symptom changes over the course of the trial. No significant (p > 0.05) changes in any of the hippocampal subregion volumes were evident between the baseline and 12-week follow-up time points in healthy volunteers.
The percentage difference in volume between time points was calculated for each hippocampal subregion using the following equation: [(12-week volume – baseline volume)/(baseline volume)] × 100. Independent-samples t tests were computed for the volumetric percentage differences of each hippocampal subregion to determine if patients had changes that were significantly different from healthy volunteers between the two time points. Consistent with the paired-samples t tests, patients displayed significantly larger increases in subiculum volumes (t = 2.34, df = 56, p = 0.02) and significantly larger decreases in DG/CA4 volumes (t = −2.49, df = 56, p = 0.02) compared with healthy volunteers.
Mean dosage for patients between the two scans was 16 177 mg (s.d. = 4800 mg) in chlorpromazine and 835 mg (s.d. = 256 mg) in olanzapine equivalents. One patient was excluded from analysis due to being an outlier (>3 s.d.s from the mean) on both chlorpromazine and olanzapine equivalents. Curve estimation did not reveal any significant association between chlorpromazine/olanzapine equivalents and changes in subiculum volume. There was, however, a significant association (R 2 = 0.34) between chlorpromazine equivalents and change in DG/CA4 volume when data were analysed using a quadratic model (F 2,25 = 6.29, p = 0.006; Fig. 4); this effect was not evident with a linear model. These results were replicated when olanzapine equivalents were used in the analysis (R 2 = 0.32, F 2,25 = 5.81, p = 0.008). When the individual with the largest reduction in DG/CA4 volume was removed from the analysis both quadratic models remained statistically significant for the analyses with chlorpromazine and olanzapine equivalents.
Change in dentate gyrus/cornu ammonis (CA) 4 volume and chlorpromazine equivalents.
Discussion
Antipsychotics are widely used in the treatment of schizophrenia and other psychotic disorders. Post-mortem animal data have indicated that antipsychotic treatment is associated with brain volume reductions compared with vehicle-treated control subjects through stereology and the use of ex vivo MR imaging (Vernon et al. Reference Vernon, Natesan, Modo and Kapur2011). There are, however, little in vivo data regarding their potential impact on hippocampal subregion morphology. Moreover, these subregions may be differentially affected by antipsychotic medications as they are known to be heterogeneous in cytoarchitecture and maintain different afferent and efferent connections. In this regard our study provides the first evidence, to our knowledge, of longitudinal in vivo changes within specific hippocampal subregions associated with antipsychotic treatment. Strengths of our study include the use of a longitudinal design, patients studied early in the course of illness prior to extensive antipsychotic treatment, and manual methods for mensuration of hippocampal subregions with high inter-rater/intra-rater reliability assessed using both Dice indices and ICC coefficients. Specifically, while patients demonstrated significant volumetric changes in the DG/CA4 and subiculum over the course of the treatment trial, healthy volunteers demonstrated only minor (non-significant) volumetric changes (i.e. about 0.3%) in these subregions over a comparable 12-week time period, providing strong support for the reliability of our manual mensuration approach.
Few studies have investigated changes in hippocampal subregion morphology in relationship to antipsychotic treatment and, thus, it is difficult to compare our results with prior work. Our findings do converge, however, with cross-sectional in vivo work (Haukvik et al. Reference Haukvik, Westlye, Mørch-Johnsen, Jørgensen, Lange, Dale, Melle, Andreassen and Agartz2015) that used FreeSurfer to identify smaller CA4/DG volumes in a large cohort of 402 previously treated patients with schizophrenia or bipolar spectrum disorders compared with 300 healthy volunteers. Moreover, Kawano et al. (Reference Kawano, Sawada, Shimodera, Ogawa, Kariya, Lang, Inoue and Honer2015) reported that CA4/DG volume was significantly smaller in first-episode, subchronic and chronic patients compared with healthy volunteers. Regarding that study, it may be noteworthy that there was a small, but statistically significant, reduction in left CA4/DG volume that was evident over the first 6 months of illness and that may be consistent with antipsychotic treatment effects. Conversely, both Haukvik et al. (Reference Haukvik, Westlye, Mørch-Johnsen, Jørgensen, Lange, Dale, Melle, Andreassen and Agartz2015) and Kawano et al. (Reference Kawano, Sawada, Shimodera, Ogawa, Kariya, Lang, Inoue and Honer2015) utilized automatic mensuration techniques in these protocols, raising the issue of compatibility of their methods with our own manual approach.
In one neurobiological model of hippocampal pathology in schizophrenia, lower excitatory signaling within the DG has been linked to altered plasticity within CA3 – thereby increasing neuronal activity and concomitant encoding mistakes leading to false memories consisting of psychotic content (Tamminga et al. Reference Tamminga, Stan and Wagner2010). Moreover, simulations (Faghihi & Moustafa, Reference Faghihi and Moustafa2015) and in vivo data (Das et al. Reference Das, Ivleva, Wagner, Stark and Tamminga2014) suggest that abnormalities within the DG are associated with deficits in separation efficiency in schizophrenia, thus further highlighting their functional sequelae and relevance for phenomenology. Post-mortem studies reported molecular changes within the DG in patients with schizophrenia compared with other parts of the hippocampus (Gao et al. Reference Gao, Sakai, Roberts, Conley, Dean and Tamminga2000; Knable et al. Reference Knable, Barci, Webster, Meador-Woodruff and Torrey2004; Allen et al. Reference Allen, Fung and Shannon Weickert2016), which may be consistent with reduced glutamate transmission in this region and the efferent mossy fiber pathway (Tamminga et al. Reference Tamminga, Stan and Wagner2010). Animal studies indicate that chronic olanzapine exposure predicted abnormal hippocampal connectivity, particularly within inhibitory terminals of the DG (Ramos-Miguel et al. Reference Ramos-Miguel, Honer, Boyda, Sawada, Beasley, Procyshyn and Barr2015), and was associated with DG volume reductions in sedentary rats with exercise reversing the majority of this loss (Barr et al. Reference Barr, Wu, Wong, Hercher, Töpfer, Boyda, Procyshyn, Honer and Beasley2013). Thus, the restoration of hippocampal neurogenesis within the DG and associated CA4 subregion could be an important therapeutic target for treating hippocampal dysfunction in psychosis (Newton & Duman, Reference Newton and Duman2007).
Our finding of an inverted U function between chlorpromazine/olanzapine equivalents and changes in DG volume provides the first evidence to our knowledge regarding a non-linear relationship between antipsychotic administration and changes in hippocampal morphology. Our findings are consistent, however, with cognitive studies reporting a non-linear relationship between antipsychotic dosage and functions subserved by the hippocampus, including measures of vigilance (Sakurai et al. Reference Sakurai, Bies, Stroup, Keefe, Rajji, Suzuki, Mamo, Pollock, Watanabe, Mimura and Uchida2013). For example, D2 occupancy above approximately 80% was demonstrated to increase the risk for extrapyramidal side effects and cognitive impairment. In this regard it may be noteworthy that dopamine receptor stimulation and working memory performance follow an inverted U function such that either too little or too much dopamine negatively has an impact on cognitive functioning (Vijayraghavan et al. Reference Vijayraghavan, Wang, Birnbaum, Williams and Arnsten2007). Moreover, several studies indicated that D2-like (Monte-Silva et al. Reference Monte-Silva, Kuo, Thirugnanasambandam, Liebetanz, Paulus and Nitsche2009) and D1 receptor activation (Fresnoza et al. Reference Fresnoza, Paulus, Nitsche and Kuo2014) was associated with a dose-dependent inverted U-shaped impact on neuroplasticity in healthy subjects.
We also identified changes in subiculum volume in patients with psychosis following antipsychotic treatment. It should first be acknowledged that estimates of the hippocampus proper have sometimes included the region of the subicular complex (La Joie et al. Reference La Joie, Fouquet, Mézenge, Landeau, Villain, Mevel, Pélerin, Eustache, Desgranges and Chételat2010; Malykhin et al. Reference Malykhin, Lebel, Coupland, Wilman and Carter2010; Winterburn et al. Reference Winterburn, Pruessner, Chavez, Schira, Lobaugh, Voineskos and Chakravarty2013), and have at other times excluded it (Szeszko et al. Reference Szeszko, Goldberg, Gunduz-Bruce, Ashtari, Robinson, Malhotra, Lencz, Bates, Crandall, Kane and Bilder2003; Wellington et al. Reference Wellington, Bilder, Napolitano and Szeszko2013; Rhindress et al. Reference Rhindress, Ikuta, Wellington, Malhotra and Szeszko2015). This inconsistency occurs because the cellular structure of the subicular complex varies considerably. Namely, some areas of the subicular complex (i.e. the subiculum) possess cytoarchitectural features that resemble those of traditional hippocampus proper subregions such as CA1. Other areas of the subicular complex (e.g. the presubiculum), however, demonstrate cytoarchitecture that is more similar to the entorhinal cortex – an area that is included in the hippocampal formation but is not often considered to be a part of the hippocampus proper (Amaral & Lavenex, Reference Amaral, Lavenex, Andersen, Morris, Amaral, Bliss and O'Keefe2006).
The subiculum, a major output region of the hippocampus, is a potential site of action for antipsychotics and animal data demonstrate an interaction of subicular and dopaminergic inputs to the nucleus accumbens (Greene, Reference Greene1996). Moreover, there are animal data relevant to the phenomenology of schizophrenia including latent inhibition (Meyer & Louilot, Reference Meyer and Louilot2011) and pre-pulse inhibition of the acoustic startle response (Caine et al. Reference Caine, Geyer and Swerdlow1992) that have implicated dysfunction of this region. Using structural MR imaging, Bearden et al. (Reference Bearden, Thompson, Avedissian, Klunder, Nicoletti, Dierschke, Brambilla and Soares2009) reported that previous treatment with atypical antipsychotics in patients with major depression was associated with larger hippocampal volume (including subicular regions) compared with patients not treated with atypicals – suggesting potential neuroprotective effects in this region. Moreover, moderate but significant increases in [3H]AFDX-384 binding sites within the subiculum were associated with 3 months of exposure to olanzapine (Terry et al. Reference Terry, Gearhart, Mahadik, Warsi and Waller2006).
There were no significant changes in regions comprising the CA of the hippocampus; however, prior research has implicated structural changes in these regions in relation to antipsychotic usage in both human and animal studies. Possible differences between the current study and prior work include defining the regions of interest, especially along the medial aspects of the CA subfields using shape analysis. For example, Zierhut et al. (Reference Zierhut, Graßmann, Kaufmann, Steiner, Bogerts and Schiltz2013) found that patients with schizophrenia demonstrated shape deformations corresponding neuroanatomically to the CA1 subfield, which was related to positive symptom severity and antipsychotic dosage. Similarly, Schobel et al. (Reference Schobel, Lewandowski, Corcoran, Moore, Brown, Malaspina and Small2009) reported in vivo abnormal cerebral blood flow increases in the CA1 subfield, which predicted onset of psychosis from the prodrome and correlated with psychotic symptoms. Furthermore, animal studies implicate GABAergic system hypofunctioning in the CA1 subfield (Steiner et al. Reference Steiner, Brisch, Schiltz, Dobrowolny, Mawrin, Krzyżanowska, Bernstein, Jankowski, Braun, Schmitt, Bogerts and Gos2016) and suggest that antipsychotics reduced the number of N-methyl-d-aspartate (NMDA) receptors and associated glutamatergic signaling within the CA1/CA2 brain regions (Krzystanek et al. Reference Krzystanek, Bogus, Pałasz, Krzystanek, Worthington and Wiaderkiewicz2015). In this regard it is noteworthy that treatment with lorazepam could be a mediator/source of bias in examining the relationship between antipsychotic usage and hippocampal volume. Specifically, a key feature of the model proposed by Lodge & Grace (Reference Lodge and Grace2011) involves γ-aminobutyric acid (GABA) signaling. In addition, it has been reported that within the hippocampus a compensatory increase in glutamate develops in association with chronic lorazepam treatment (Bonavita et al. Reference Bonavita, Ferrero, Cereseto, Velardez, Rubio and Wikinski2003).
There were a number of limitations with our study that should be acknowledged. Although the use of manual volumetry remains the ‘gold standard’ for computerized mensuration of the hippocampus, this process is labor-intensive and therefore not practical for large sample sizes. On the other hand, direct comparisons of manual volumetry of the hippocampus with automated approaches have reported inconsistent findings for case–control comparisons (Clerx et al. Reference Clerx, Jacobs, Burgmans, Gronenschild, Uylings, Echávarri, Visser, Verhey and Aalten2013; Arnold et al. Reference Arnold, Ivleva, Gopal, Reddy, Jeon-Slaughter, Sacco, Francis, Tandon, Bidesi, Witte, Poudyal, Pearlson, Sweeney, Clementz, Keshavan and Tamminga2015). Moreover, automated approaches for hippocampal delineation such as FreeSurfer may overestimate volumes, especially among younger individuals, and produce less pronounced right greater than left volume asymmetry (Wenger et al. Reference Wenger, Mårtensson, Noack, Bodammer, Kühn, Schaefer, Heinze, Düzel, Bäckman, Lindenberger and Lövdén2014) that has typically been observed in studies employing manual mensuration (Wellington et al. Reference Wellington, Bilder, Napolitano and Szeszko2013). It should also be acknowledged that the current neurobiological evidence for imaging findings in schizophrenia and associated psychotic disorders has been inconclusive (Weinberger & Radulescu, Reference Weinberger and Radulescu2016) and factors such as smoking, substance use, and medical co-morbidities could potentially influence findings. However, the presence of an inverted U response demonstrates that the simple ‘less is worse’, or ‘more is better’ dichotomy is not supported by the current investigation. Furthermore, limitations of MR imaging must also be acknowledged. For example, there is a possibility that chronic drug use could influence the MR imaging signal (e.g. Cousins et al. Reference Cousins, Aribisala, Nicol Ferrier and Blamire2013). In addition, we acknowledge that chlorpromazine (and olanzapine) equivalence is determined based on clinical dosing and not biological factors (e.g. receptor occupancy). Larger sample sizes would also be useful to investigate the differential effects of antipsychotics on volumetric changes.
In sum, we report in vivo volumetric changes of hippocampal subregions using manual volumetry in patients with first-episode psychosis following 12 weeks of antipsychotic treatment. Changes in DG/CA4 volume demonstrated an inverted U pattern in association with cumulate antipsychotic exposure over the course of the treatment trial. Significant volumetric changes among patients were observed alongside negligible volumetric changes among healthy volunteers, thus providing evidence for the excellent reliability of our mensuration approach. This is especially notable given the presence of well-known limitations with automated measurements of these subregions.
Acknowledgements
This work was supported in part by grants from NARSAD (P.R.S.) and the National Institute of Mental Health to P.R.S. (R01 MH076995), the NSLIJ Research Institute General Clinical Research Center (M01 RR018535), the Advanced Center for Intervention and Services Research (P30 MH090590) and the Center for Intervention Development and Applied Research (P50 MH080173)
Declaration of Interest
None.
