Association between hippocampal subelds and clinical symptoms of rst-episode and drug naive schizophrenia patients during 12 weeks of risperidone treatment

Small hippocampal size, especially in CA1 may be implicated in the pathogenesis and psychopathology of schizophrenia (SCZ). However, does the volume of hippocampal subelds like CA1 in SCZ patients affect response to antipsychotic treatment? In this study, we used risperidone to treat rst-episode and drug naïve (FEDN) SCZ patients for 12 weeks, and then explored the relationship between differences in baseline hippocampal subeld volumes, as well as any changes in the volumes of these hippocampal structures during treatment, and improvement in their psychopathological symptoms. By adopting a state-of the-art automated algorithm, the hippocampal subelds were segmented in 43 FEND SCZ inpatients at baseline and in 26 of them after 12 weeks of risperidone monotherapy, as well as in 30 matched healthy controls. We adopted the Positive and Negative Syndrome Scale (PANSS) to assess psychopathological symptoms in 43 SCZ patients at baseline and in 40 of them at post-treatment. Our results showed that before treatment, schizophrenia patients had no signicant differences in total or subeld hippocampal volumes compared with healthy volunteers (all p>0.05). However, at baseline smaller volumes of the left CA1, right molecular layer of the dentate, left hippocampal tail and left pre-subiculum were signicantly correlated with worse PANSS negative symptom scores, accounting for 41% of its variance. We also found a signicant correlation between a larger left CA1 at baseline and a worse PANSS total score and general psychopathology sub-score at post-treatment (both p<0.05). Furthermore, the left CA1 at baseline was signicantly smaller in responders, who had >50% improvement in PANSS total score, than in non-responders (p<0.05). Our results suggest that FEDN SCZ patients with smaller left CA1 had greater negative psychotic symptoms at baseline and more of them had a >50% improvement in PANSS total score after 12 weeks of risperidone treatment. Thus, smaller left CA1 volume may be a predicator for improvement in psychotic symptoms of FEDN SCZ patients, but hippocampal volumes overall did not differ in size from normal controls and did not change during treatment. Different antipsychotic drugs may also have different effects on hippocampal volumes, since our study used a monotherapy with risperidone at a relatively low to moderate dose of 3~6 mg/day, while the patients in other studies took various types of antipsychotic drugs 9, 27, 29 . A previous animal study reected such medication specic effects such that olanzapine administration caused signicant reduction in total hippocampus, CA1 and DG molecular layer volumes 30 . Thus, these results show reduced volumes of hippocampal subelds in chronically treated SCZ patients, but the results are mixed for FEDN patients, While a recent study reported higher hippocampal subeld volumes in patients with rst-episode antipsychotic naïve SCZ 27 , another more recent study found more hippocampal subeld decits (particularly right hippocampus) in never-treated than treated SCZ patients, suggesting that antipsychotic treatment may reduce shrinkage of hippocampal structures in SCZ 26 . These discrepancies may reect different methods of acquiring the neuroimaging data or segmenting the hippocampal subelds, different course of disease, exposure to antipsychotic treatment, or even different ethnicities. We speculate that the hippocampus and its subelds may be unimpaired at the onset of SCZ, but the hippocampus gradually becomes damaged and shrinks with the development of illness. Contributing factors may be long-term exposure to chronic duration of disease, antipsychotic treatment, smoking, antipsychotic-induced side effects or comorbid diseases. We found that multiple hippocampal subelds including left CA1, right molecular layer, left pre-subiculum and left hippocampal tail were signicantly correlated with PANSS negative symptoms, but few other studies have explored the relationship between clinical symptoms and hippocampal subelds of SCZ patients. A previous study showed that the posterior CA1and left anterior subeld volume decits were correlated with positive symptoms, especially hallucinations and delusions 31 . A subsequent study found that CA1 and CA2/3 subeld volumes


Introduction
Schizophrenia (SCZ) is one of the most common severe psychiatric disorders, with a prevalence rate around 1% worldwide 1 . Its core symptoms include positive symptoms, negative symptoms, and cognitive impairment. Although antipsychotic drugs can signi cantly improve the positive symptoms and some negative symptoms of patients, negative symptoms and cognitive impairment show limited responses 2,3 . The incomplete response to antipsychotics in some SCZ patients is not readily predicted, but Magnetic Resonance Imaging (MRI) technology can provide brain structure correlates for SCZ pharmacotherapy responses. In particular, many studies have demonstrated hippocampal volume reductions in SCZ patients' brains as the most distinguishing difference from healthy controls among both early and chronic patients [4][5][6][7][8][9][10] . Hippocampal abnormalities may be present before the rst episode of psychosis 7,11 , and correlate with SCZ patients' clinical symptoms and cognitive impairment 9,[12][13][14][15][16] .
Anatomically, the formation of the hippocampus is not uniform, but consists of several sub elds with distinct morphology and function, such as the cornu ammonis (CA) 1-4, dentate gyrus (DG), subiculum and presubiculum or entorhinal cortex (EC) 17,18 . Several MRI studies in chronic SCZ identi ed hippocampal sub elds with smaller volumes including the CA1, left CA2/3 and CA4/DG and subiculum 8,9,19 . Moreover, the positive symptoms especially hallucinations and delusions were inversely correlated with the bilateral CA1 and CA2/3 sub eld volumes in chronic SCZ patients 20 . One study showed a signi cant association between smaller pre-subiculum volumes and negative symptoms 7 . However, only a few studies reported on rst episode SCZ patients, and one found lower CA1 and CA2 sub eld volumes in the left hemisphere 21 , while recent studies demonstrated that there was no difference in the volumes of hippocampal sub elds between rst-episode SCZ patients and healthy volunteers 9,22 .
Some previous studies have shown that antipsychotic treatment may affect hippocampal volumes 23,24 with the dose of antipsychotic drugs being inversely associated with total hippocampal and sub eld volumes 25,26 . Another study demonstrated reduced CA1 volume in early SCZ, with this focal atrophy in early illness extending to CA2/3 9 . In contrast, another recent study showed signi cantly larger sub eld volumes in bilateral CA4 and bilateral molecular and granular cell layers of DG in FEDN SCZ patients, but without differences from controls in total hippocampal volume. After 6-week antipsychotic treatment, patients displayed volume decreases in total hippocampus, as well as several sub elds including those previously enlarged sub elds at baseline, which were decreased to the levels of healthy controls 27 . One interpretation is that antipsychotic treatment may correlate with hippocampal brain tissue loss over time 23 .
Based on this consensus of reduced hippocampal CA1 volume in chronic and perhaps FEDN SCZ patients and a potential antipsychotic role in reducing hippocampal brain tissue with greater antipsychotic doses, we examined the correlations between hippocampal sub eld volumes particularly CA1 and clinical symptoms in FEDN SCZ patients during risperidone treatment. The present study had three purposes. 1) Whether the FEDN SCZ patients at baseline differed from controls in their hippocampal sub eld volumes particularly in CA1 and whether those volumes correlated with baseline PANSS scales. 2) Whether baseline CA1 volumes would predict FEDN SCZ patients' responses to risperidone treatment. 3) Whether risperidone treatment would have a signi cant effect on the hippocampal sub elds like CA1; and whether any changes in the hippocampal sub elds would correlate with improvements in psychopathological symptoms.

Baseline comparisons between SCZ patients and healthy volunteers
All patients and healthy controls were right-handed. We did not nd signi cant differences in sex, age, and education between SCZ patients and healthy volunteers (all p>0.05). Also, we did not observe signi cant differences in any of hippocampal sub elds between SCZ patients and healthy volunteers (all p>0.05).
Relationships between psychopathology and the sub eld volumes in SCZ at baseline Using multivariate regression analysis, larger volumes of the right molecular layer (p=0.003), left CA1 (p=0.008), left hippocampal tail (p=0.014) and left pre-subiculum (p=0.019) were signi cantly correlated with worse PANSS negative symptom scores at baseline, accounting for 41% of its variance.

Changes in clinical symptoms and sub eld volumes after treatment
Of these 43 patients, 3 dropped out before 12-week treatment, which we did not include in the statistical analysis. The remaining 40 patients completed the full 12-week trial, and their demographics and clinical parameters are present in Table 1. After 12 weeks of risperidone treatment, all PANSS three subscale scores and total score were signi cantly reduced (all p<0.05; Table 1).
Relationship between changes in clinical outcome and sub elds before and after treatment After 12 weeks of treatment, we obtained the hippocampal sub eld data of 26 patients, because 10 patients refused to receive the second MRI scan and 4 patients were unable to obtain an MRI scanner. After 12 weeks of risperidone treatment, we did not observe any signi cant changes in all hippocampal total and sub eld volumes (all p>0.05; Table 2).
We found a signi cant correlation between a larger left CA1 at baseline and a worse PANSS total score (r=0.35, df=40, p<0.05) and general psychopathology sub-score at post-treatment (r=0.33, df=40, p<0.05) ( Figure 1). After controlling for ICV, sex, age, and age of onset as covariates, these correlations remained signi cant (both p<0.05).
However, these signi cances did not pass Bonferroni correction (both p>0.05). In addition, there were no signi cant correlations between the reduction of PANSS scores and changes in any hippocampal sub eld volumes (all p>0.05).

Discussion
This study found no signi cant hippocampal sub eld volume de cits in FEDN SCZ patients compared to normal controls, but small volumes of some sub elds at baseline were associated with greater negative symptoms. Furthermore, small left CA1 volume at baseline may predict the response to 12 weeks of risperidone treatment, especially for improved general psychopathological symptoms, although no hippocampal sub eld volumes showed signi cant volume changes after treatment.

Hippocampal sub elds in FEDN SCZ patients at baseline
Although many studies showed smaller volumes of the whole hippocampus or its sub elds in SCZ 4, 6, 7, 9 , we did not observe any hippocampal sub eld volume de cits in the patients with FEDN SCZ, which is consistent with a previous study in patients with rst episode SCZ, showing no signi cantly abnormal hippocampal sub elds 8 . Some studies have indicated that a longer course of disease and prior antipsychotic treatment may contribute to volume reductions of hippocampal sub elds in SCZ patients 23,24,26 . One recent study indicated that longer untreated psychosis duration was also correlated with smaller whole hippocampal and some of sub eld volumes 28 . Another study demonstrated that the CA1 volume was selectively decreased in patients in the early to mid-course of schizophrenia. Moreover, their longitudinal analysis demonstrated that the focal atrophy correlated with early disease extended beyond CA1 over time and involved other sub elds, including CA2/3 and GCL 9 .
Another longitudinal study showed that after 12-week antipsychotic treatment for patients with rst-episode psychosis, DG and CA4 volumes were signi cantly reduced 29 . Another recent study showed signi cantly greater volumes of hippocampal sub elds in FEDN SCZ patients at baseline, and after only 6-week treatment with antipsychotics, these patients exhibited volume reductions in total hippocampus and several sub elds 27 . Our results suggest that acute or mid-term antipsychotic treatment may not impact hippocampal structural volume, although reduced hippocampal sub eld volumes seem characteristic of chronic SCZ patients. Different antipsychotic drugs may also have different effects on hippocampal volumes, since our study used a monotherapy with risperidone at a relatively low to moderate dose of 3~6 mg/day, while the patients in other studies took various types of antipsychotic drugs 9,27,29 . A previous animal study re ected such medication speci c effects such that olanzapine administration caused signi cant reduction in total hippocampus, CA1 and DG molecular layer volumes 30 . Thus, these results show reduced volumes of hippocampal sub elds in chronically treated SCZ patients, but the results are mixed for FEDN patients, While a recent study reported higher hippocampal sub eld volumes in patients with rst-episode antipsychotic naïve SCZ 27 , another more recent study found more hippocampal sub eld de cits (particularly right hippocampus) in never-treated than treated SCZ patients, suggesting that antipsychotic treatment may reduce shrinkage of hippocampal structures in SCZ 26 . These discrepancies may re ect different methods of acquiring the neuroimaging data or segmenting the hippocampal sub elds, different course of disease, exposure to antipsychotic treatment, or even different ethnicities. We speculate that the hippocampus and its sub elds may be unimpaired at the onset of SCZ, but the hippocampus gradually becomes damaged and shrinks with the development of illness. Contributing factors may be long-term exposure to chronic duration of disease, antipsychotic treatment, smoking, antipsychotic-induced side effects or comorbid diseases.
We found that multiple hippocampal sub elds including left CA1, right molecular layer, left pre-subiculum and left hippocampal tail were signi cantly correlated with PANSS negative symptoms, but few other studies have explored the relationship between clinical symptoms and hippocampal sub elds of SCZ patients. A previous study showed that the posterior CA1and left anterior sub eld volume de cits were correlated with positive symptoms, especially hallucinations and delusions 31 . A subsequent study found that CA1 and CA2/3 sub eld volumes were negatively associated with PANSS 19 . Another study showed smaller subiculum volumes were associated with negative symptoms 7 . Taken together with our ndings, we propose that structural damage to the hippocampus is not evident at SCZ onset across all patients, but patients with smaller hippocampal sub elds may have more severe negative symptoms, possibly due to abnormal functional activity in these sub elds such as left CA1. Interestingly, a previous study reported that the right hippocampus of patients with rst-episode SCZ had greater resting state activity, which was associated with more negative symptoms, perhaps re ecting the relatively smaller volumes and de cits in functional activity of their left hippocampal sub elds 32 , as we found in the current study. Prolonged SCZ disease may lead to a progressive loss of hippocampal sub eld volumes 8 , as we noted above, but shorter term treatment does not appear to shrink hippocampal volumes in FEDN SCZ patients.
The left CA1 volumes at baseline may predict the response to risperidone treatment for general psychopathological symptoms We found no signi cant hippocampal sub eld de cits in FEDN SCZ patients compared to normal controls, but smaller baseline hippocampal sub eld volumes, especially of CA1 may predict improvement with risperidone treatment, speci cally in general psychopathological symptoms. A recent report revealed that baseline right CA3 volume was positively correlated with improvement in Global Assessment of Functioning (GAF) after 6 weeks of antipsychotic treatment 27 , which is consistent with our current study when contrasting right to left sided hippocampal volumes (left relatively smaller). A previous study also showed that over a 6-year follow-up, a subgroup of SCZ patients exhibited an increase in bilateral hippocampal volumes, who had better results in clinical, cognitive and functional domains 33 . These ndings indicate that the hippocampal volume at baseline may be a potential biomarker for antipsychotic treatment response in schizophrenia. However, hippocampal laterality may be important for the optimal use of this non-invasive biomarker to predict treatment response in early schizophrenia.
In this study, there are several limitations that need to be mentioned. First, although considering the inherent challenges of recruiting FEDN patients, our sample size in this study was still relatively small, so replicated research samples from different races will be important to con rm our results. Second, in our current study, only a single antipsychotic drug risperidone at a relatively stable dose was used. Therefore, the dose correlation must be carefully considered, and whether the results of this study can be generalized to other antipsychotics and other higher doses requires further research. Third, the treatment duration of this study in FEDN schizophrenia is relatively short, and the effect of long-term antipsychotic treatment on the hippocampal sub elds needs to be explored in the future.
In summary, we did not observe obvious baseline hippocampal sub eld de cits in patients with FEDN SCZ; however, patients with larger hippocampal sub elds may have more severe negative symptoms. We found that the baseline smaller left CA1 was signi cantly corrected with the post-treatment PANSS general psychopathology and total scores after 12 weeks of risperidone treatment, suggesting that the smaller baseline left CA1 may predict the response to risperidone treatment. However, we found no signi cant association between changes in hippocampal sub eld volumes and improvement in clinical symptoms of SCZ patients after 12 weeks of treatment. Overall, due to the relatively limited sample size, using a single antipsychotic risperidone at a relatively modest and stable dose, and a relatively short treatment duration, replicated research samples from different ethnicities using different antipsychotic treatments for a long time will be optimal for supporting use this biomarker with antipsychotic treatment and prognosis. . We randomly recruited 30 healthy controls with ages ranging from 16 to 45 years from a community in Beijing during the same period. Under the supervision of a research psychiatrist, trained researchers interviewed them.

Methods
They did not have a personal history of Axis I disease, assessed by the researchers using SCID, and their rstdegree relatives did not have any known history of mental illness. The researchers used SCID for the assessment.
They did not have a personal history of Axis I disease, and their rst-degree relatives did not have any known history of mental illness.
All participants underwent physical examination and laboratory tests. Exclusion criteria for the participants were (1) having current physical diseases; (2) personal and family history of any brain diseases; (3) lifetime alcohol or substance abuse/dependence history except for tobacco; (6) refusal to provide written informed consent.
The Institutional Review Committee of Beijing Huilongguan Hospital approved the research protocol. We obtained the written informed consent of each participant.

Study design and Assessments
Risperidone was used to treat 43 patients for 12 weeks. During the rst week, the risperidone dose was titrated to 3~6 mg per day and remained at these levels throughout the study period. Also, lorazepam or chloral hydrate was used for insomnia, and benzhexol hydrochloride (as needed) for extrapyramidal side effects. There was no other prescribed drug during the study.
By using the Positive and Negative Syndrome Scale (PANSS), two psychiatrists assessed the patient's psychopathological symptoms while they were unaware of the patients' status. They both simultaneously received a training session in using PANSS before this study began. After training, they maintained an inter-rater correlation coe cient more than 0.8 for the PANSS total score during repeated assessments. They rated all patients at baseline and at the end of 12-week treatment. For each patient, the same investigator rated these scales at baseline and at 12 weeks.

Acquisition and pretreatment of MRI data
The GE 3 Tesla MRI scanner was used to acquire the structural T1 weighted scan of each participant. Then the spoiled gradient echo (SPGR) sequence was adopted with these parameters: repetition time (TR) = 6.2 ms, echo time (TE) = 2.8 ms, slice thickness = 1.2 mm, matrix size = 256 × 256 and 142 slices, eld of view (FOV) = 240 mm, and ip angle = 8°. We used FreeSurfer software to perform subcortical segmentation 34 . An automated algorithm was employed to segment hippocampal sub elds 35 (29).
A sub eld segmentation map was established as reported our previous studies 13,35 . In this study, there were 8 hippocampal sub elds including: CA1~CA4, GCL, ML, pre-subiculum, subiculum, and the hippocampal tail 35 .

Statistical Analyses
Initial analysis included all patients and healthy controls. Normality of the data was analyzed by the Kolmogorov-Smirnov test. x 2 test and one-way analysis of variance were performed to compare group differences. Further, after controlling for covariates such as intracranial volume (ICV), age, sex, and education, one-way analysis of covariance (ANCOVA) was employed to compare each hippocampal sub eld between patients and healthy volunteers. Then, using ICV, age, and gender as covariates, a partial correlation was performed between general clinical data, clinical symptoms on PANSS scores, and hippocampal sub eld volumes. Further stepwise multiple regression analysis was conducted to con rm the relationship among clinical variables, clinical symptoms and hippocampal sub elds.
The effect of treatment was assessed using repeated measures multivariate analysis of variance (MANOVA) with the PANSS scores and the hippocampal sub eld volumes as outcome measures. Then we performed partial correlation analysis to explore the relationship between symptom improvement and hippocampal sub eld volume changes, with ICV, sex, age, and age of onset as covariates. We employed Bonferroni correction for multiple tests. Finally, exploratory regression analyses was carried out to investigate whether there were associations between changes in volumes of hippocampal sub elds and improvement in clinical symptoms.
In addition, we used the criteria of a 50% or more improvement in PANSS total score to de ne "responders" or "non-responders". We examined whether risperidone treatment for 12 weeks altered the hippocampal sub eld volumes and whether differences in the hippocampal sub elds occurred between responders and nonresponders.
All statistical analyses were carried out using SPSS version 18.0 (SPSS Inc., Chicago, United States), and the signi cance level was p<0.05 (two-tailed).  # There were no significant differences in hippocampal subfields between schizophrenia patients and healthy cont p>0.05). Figure 1 After controlling for sex, age, education, age of onset and intracranial volume (ICV), the partial correlation analysis showed that the volumes of left cornu ammonis 1 (CA1) at baseline was associated PANSS total score and general psychopathology (both p<0.05) at the end of 10 weeks of risperidone treatment in rst episode patients with schizophrenia.