Introduction
Identifying novel pharmacological interventions to reduce symptom severity and prevent transition to psychosis in individuals at clinical high-risk for psychosis (CHR-P) is a significant unmet clinical need (Bosnjak Kuharic et al., Reference Bosnjak Kuharic, Kekin, Hew, Rojnic Kuzman and Puljak2019; Fusar-Poli, de Pablo, Correll, et al., Reference Fusar-Poli, Salazar de Pablo, Correll, Meyer-Lindenberg, Millan, Borgwardt, Galderisi, Bechdolf, Pfennig, Kessing, van Amelsvoort, Nieman, Domschke, Krebs, Koutsouleris, McGuire, Do and Arango2020). Current neurobiological theories of psychosis development identify the hippocampus as a central hub of pathophysiology (Heckers & Konradi, Reference Heckers and Konradi2015; Knight, McCutcheon, Dwir, et al., Reference Knight, McCutcheon, Dwir, Grace, O’Daly, McGuire and Modinos2022; Lieberman, Girgis, Brucato, et al., Reference Lieberman, Girgis, Brucato, Moore, Provenzano, Kegeles, Javitt, Kantrowitz, Wall, Corcoran, Schobel and Small2018) and a promising pharmacological target (Uliana, Lisboa, Gomes, & Grace, Reference Uliana, Lisboa, Gomes and Grace2024). Several neuroimaging studies in individuals at CHR-P have identified increased hippocampal cerebral blood flow/volume compared to healthy controls (HC) (Allen, Chaddock, Egerton, et al., Reference Allen, Chaddock, Egerton, Howes, Bonoldi, Zelaya, Bhattacharyya, Murray and McGuire2016; Allen, Azis, Modinos, et al., Reference Allen, Azis, Modinos, Bossong, Bonoldi, Samson, Quinn, Kempton, Howes, Stone, Calem, Perez, Bhattacharayya, Broome, Grace, Zelaya and McGuire2018; Provenzano, Guo, Wall, et al., Reference Provenzano, Guo, Wall, Feng, Sigmon, Brucato, First, Rothman, Girgis, Lieberman and Small2020). The CA1 subfield is proposed to be the origin of hippocampal dysfunction in the CHR-P state, in terms of volume loss (Ho, Holt, Cheung, et al., Reference Ho, Holt, Cheung, Iglesias, Goh, Wang, Lim, de Souza, Poh, See, Adcock, Wood, Chee, Lee and Zhou2017) and hyperactivity (Schobel, Chaudhury, Khan, et al., Reference Schobel, Chaudhury, Khan, Paniagua, Styner, Asllani, Inbar, Corcoran, Lieberman, Moore and Small2013; Schobel, Lewandowski, Corcoran, et al., Reference Schobel, Lewandowski, Corcoran, Moore, Brown, Malaspina and Small2009), which then spreads to the subiculum following psychosis onset (Schobel et al., Reference Schobel, Chaudhury, Khan, Paniagua, Styner, Asllani, Inbar, Corcoran, Lieberman, Moore and Small2013). The CA1 and subiculum have a high number of glutamatergic efferent projections (Qiu et al., Reference Qiu, Hu, Huang, Gao, Wang, Wang, Ren, Shi, Chen, Wang, Wang, Han, Liang, Liu, Liu, Deng, Chen, Zhan, Chen and Xu2024), and anterior projections innervate a cortico-limbic-striatal circuit encompassing the nucleus accumbens (NAc) of the striatum, amygdala, and the ventromedial prefrontal cortex (vmPFC) (Grace, Reference Grace2016). These regions are highly interconnected (Alexander, DeLong, & Strick, Reference Alexander, DeLong and Strick1986; Grace, Reference Grace2016; Haber, Reference Haber2003; Haber, Reference Haber2016; Haber & Fudge, Reference Haber and Fudge1997; Harrison, Guell, Klein-Flügge, & Barry, Reference Harrison, Guell, Klein-Flügge and Barry2021; Kahn & Shohamy, Reference Kahn and Shohamy2013; Lodge & Grace, Reference Lodge and Grace2006; Sah, Faber, Lopez De Armentia, & Power, Reference Sah, Faber, Lopez De Armentia and Power2003) and are associated with positive, negative, and cognitive symptoms of schizophrenia, respectively (Floresco, Todd, & Grace, Reference Floresco, Todd and Grace2001; Ghoshal & Conn, Reference Ghoshal and Conn2015; Grace, Reference Grace2010). Therefore, hippocampal dysfunction preceding the onset of psychosis may disrupt downstream cortico-limbic-striatal regions, contributing to circuit dysfunction and the emergence of psychosis (Grace, Reference Grace2016).
Circuit dysfunction can be investigated in terms of the functional connectivity (FC) between brain regions measured using resting-state functional magnetic resonance imaging (rs-fMRI) (Power, Schlaggar, & Petersen, Reference Power, Schlaggar and Petersen2014). rs-fMRI studies have identified altered hippocampal FC with the cortico-limbic-striatal circuit in individuals with a first episode of psychosis or chronic schizophrenia compared to HC. More specifically, these studies reported lower hippocampal FC with the striatum (Edmiston, Song, Chang, et al., Reference Edmiston, Song, Chang, Yin, Zhou, Zhou, Jiang, Wei, Xu, Tang and Wang2020; Gangadin et al., Reference Gangadin, Cahn, Scheewe, Hulshoff Pol and Bossong2021; Knöchel, Stäblein, Storchak, et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Kraguljac, White, Hadley, et al., Reference Kraguljac, White, Hadley, Hadley, Ver Hoef, Davis and Lahti2016; Nelson, Kraguljac, Maximo, et al., Reference Nelson, Kraguljac, Maximo, Briend, Armstrong, Ver Hoef, Johnson and Lahti2022; Sarpal, Robinson, Lencz, et al., Reference Sarpal, Robinson, Lencz, Argyelan, Ikuta, Karlsgodt, Gallego, Kane, Szeszko and Malhotra2015; Song, Yang, Chang, et al., Reference Song, Yang, Chang, Wei, Yin, Zhu, Zhou, Zhou, Jiang, Wu, Kong, Xu, Wang and Tang2022) and vmPFC (Blessing et al., Reference Blessing, Murty, Zeng, Wang, Davachi and Goff2020; Cen, Xu, Yang, et al., Reference Cen, Xu, Yang, Mei, Chen, Zhuo, Xiang, Song, Wang, Guo, Wang, Jiang, Xu, Li and Liu2020; Duan, Gan, Yang, et al., Reference Duan, Gan, Yang, Cheng, Gao, Shi, Zhu, Liang and Zhao2015; Fan, Tan, Yang, et al., Reference Fan, Tan, Yang, Tan, Zhao, Chen, Li, Song, Wang, Jin, Zhou, Milham, Zou and Zuo2013; Knöchel et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Kraguljac et al., Reference Kraguljac, White, Hadley, Hadley, Ver Hoef, Davis and Lahti2016; Liu, Li, Liu, et al., Reference Liu, Li, Liu, Yan, Wang, Sun, Fan, Song, Xu, Chen, Chen, Wang, Guo, Wan, Lv, Yang, Li, Lu, Yan and Liu2020; Nelson et al., Reference Nelson, Kraguljac, Maximo, Briend, Armstrong, Ver Hoef, Johnson and Lahti2022; Qiu, Lu, Zhou, et al., Reference Qiu, Lu, Zhou, Yan, Du, Zhang, Xie and Zhang2021; Samudra, Ivleva, Hubbard, et al., Reference Samudra, Ivleva, Hubbard, Rypma, Sweeney, Clementz, Keshavan, Pearlson and Tamminga2015; Song et al., Reference Song, Yang, Chang, Wei, Yin, Zhu, Zhou, Zhou, Jiang, Wu, Kong, Xu, Wang and Tang2022; Wang, Yin, Sun, et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021; Xue, Chen, Wei, et al., Reference Xue, Chen, Wei, Chen, Han, Wang, Zhang, Song and Cheng2023; Zhou, Shu, Liu, et al., Reference Zhou, Shu, Liu, Song, Hao, Liu, Yu, Liu and Jiang2008), and either lower (Tian, Meng, Yan, et al., Reference Tian, Meng, Yan, Zhao, Liu, Yan, Han, Yuan, Wang, Yue, Zhang, Li, Zhu, He and Zhang2011; Xue et al., Reference Xue, Chen, Wei, Chen, Han, Wang, Zhang, Song and Cheng2023), higher (Walther, Lefebvre, Conring, et al., Reference Walther, Lefebvre, Conring, Gangl, Nadesalingam, Alexaki, Wüthrich, Rüter, Viher, Federspiel, Wiest and Stegmayer2022), or unaltered (Walther et al., Reference Walther, Lefebvre, Conring, Gangl, Nadesalingam, Alexaki, Wüthrich, Rüter, Viher, Federspiel, Wiest and Stegmayer2022) hippocampal FC to the amygdala. The pattern is less clear in subclinical psychosis spectrum individuals (although there are far fewer studies): lower hippocampal-striatal FC has been shown in healthy individuals with high schizotypy traits (Kozhuharova, Saviola, Diaconescu, & Allen, Reference Kozhuharova, Saviola, Diaconescu and Allen2021; Waltmann, O’Daly, Egerton, et al., Reference Waltmann, O’Daly, Egerton, McMullen, Kumari, Barker, Williams and Modinos2019), while both lower (Edmiston et al., Reference Edmiston, Song, Chang, Yin, Zhou, Zhou, Jiang, Wei, Xu, Tang and Wang2020; Liu et al., Reference Liu, Li, Liu, Yan, Wang, Sun, Fan, Song, Xu, Chen, Chen, Wang, Guo, Wan, Lv, Yang, Li, Lu, Yan and Liu2020) and normal (Aberizk, Sefik, Addington, et al., Reference Aberizk, Sefik, Addington, Anticevic, Bearden, Cadenhead, Cannon, Cornblatt, Keshavan, Mathalon, Perkins, Stone, Tsuang, Woods and Walker2023; Allen, Hird, Orlov, et al., Reference Allen, Hird, Orlov, Modinos, Bossong, Antoniades, Sampson, Azis, Howes, Stone, Perez, Broome, Grace and McGuire2021; Wang et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021) hippocampal-striatal and hippocampal-PFC FC have been observed in individuals at CHR-P compared to HC. To our knowledge, no studies in CHR-P individuals have investigated hippocampal-amygdala FC, or FC alterations from specific hippocampal subfields to the cortico-limbic-striatal circuit. Given that hippocampal dysfunction may be localised to the CA1 subfield in the CHR-P stage (Schobel et al., Reference Schobel, Lewandowski, Corcoran, Moore, Brown, Malaspina and Small2009), alterations in FC may not be present across the whole hippocampus.
GABAergic dysfunction has been proposed as a key mechanism underlying hippocampal hyperactivity in psychosis (Heckers & Konradi, Reference Heckers and Konradi2015). Studies in rats exposed to the mitotoxin methylazoxymethanol acetate (MAM) during neurodevelopment showed that reduced PV+ interneuron number in the hippocampus was associated with an increased firing rate of local excitatory neurons and excitatory/inhibitory imbalance (Lodge & Grace, Reference Lodge and Grace2007). This hyperactivity is found to drive functional alterations of downstream regions in MAM-treated rats, evidenced by experiments where chemical (Lodge & Grace, Reference Lodge and Grace2007) or pharmacological inactivation of the hippocampus (with a nonspecific GABAA-enhancing benzodiazepine (Perez, McCoy, Prevot, et al., Reference Perez, McCoy, Prevot, Mian, Carreno, Frazer, Cook, Sibille and Lodge2023) or an α5-GABAA specific compound (Gill et al., Reference Gill, Lodge, Cook, Aras and Grace2011; Perez et al., Reference Perez, McCoy, Prevot, Mian, Carreno, Frazer, Cook, Sibille and Lodge2023)) normalised midbrain dopaminergic neuron firing. Furthermore, this mechanism is proposed to underlie the findings that repeated peripubertal diazepam administration in MAM-treated rats prevented the emergence of schizophrenia-related neurophysiological and behavioural phenotypes in adulthood. Such phenotypes included prevention of midbrain dopamine hyperactivity and hyperlocomotion response to amphetamine (positive symptoms), amygdala hyperactivity (negative symptoms), and PFC dysfunction (cognitive symptoms) (Du & Grace, Reference Du and Grace2013; Du & Grace, Reference Du and Grace2016; Du & Grace, Reference Du and Grace2016).
This preclinical evidence suggests that GABA-enhancing compounds may be an effective strategy for psychosis prevention by downregulating hippocampal hyperactivity and normalising downstream circuit dysfunction. In healthy individuals, prior rs-fMRI studies using an acute, non-sedating dose of a GABA-enhancing compound report increases in FC under benzodiazepine (or other GABA-enhancing drugs, e.g., Z-drugs such as zopiclone/zolpidem) compared to placebo across the hippocampal-amygdala-PFC circuit (Licata et al., Reference Licata, Nickerson, Lowen, Trksak, Maclean and Lukas2013), the default mode network (Flodin, Gospic, Petrovic, & Fransson, Reference Flodin, Gospic, Petrovic and Fransson2012; Frölich, White, Kraguljac, & Lahti, Reference Frölich, White, Kraguljac and Lahti2020), and a wider brain network including visual, auditory, sensorimotor, and prefrontal regions (Blanco-Hinojo, Pujol, Macià, et al., Reference Blanco-Hinojo, Pujol, Macià, Martínez-Vilavella, Martín-Santos, Pérez-Sola and Deus2021). In CHR-P individuals, we recently demonstrated that an acute, non-sedating dose of diazepam normalised elevated hippocampal and subfield cerebral blood flow to levels seen in healthy controls (Livingston et al., Reference Livingston, Kiemes, Devenyi, Knight, Lukow, Jelen, Reilly, Dima, Nettis, Casetta, Agyekum, Zelaya, Spencer, De Micheli, Fusar-Poli, Grace, Williams, McGuire, Egerton and Modinos2024). However, whether this is accompanied by a normalisation of the FC between the hippocampus and downstream cortico-limbic-striatal regions was not known.
Therefore, the current study examined the effects of an acute dose of diazepam versus placebo on FC between the hippocampus and this cortico-limbic-striatal circuit in the same cohort of CHR-P individuals (Livingston et al., Reference Livingston, Kiemes, Devenyi, Knight, Lukow, Jelen, Reilly, Dima, Nettis, Casetta, Agyekum, Zelaya, Spencer, De Micheli, Fusar-Poli, Grace, Williams, McGuire, Egerton and Modinos2024). Each condition was also compared to HC data collected on the same scanner. We focussed on the CA1 subfield as a seed, given its proposed role in psychosis development at the CHR-P stage (Lieberman et al., Reference Lieberman, Girgis, Brucato, Moore, Provenzano, Kegeles, Javitt, Kantrowitz, Wall, Corcoran, Schobel and Small2018) and the substantial number of anatomical connections to output regions of interest (NAc, amygdala, and vmPFC (Aggleton, Reference Aggleton2012; Rosene & Van Hoesen, Reference Rosene and Van Hoesen1977)). On the basis of previous findings in hippocampal FC across the psychosis spectrum (Alexander et al., Reference Alexander, DeLong and Strick1986; Floresco et al., Reference Floresco, Todd and Grace2001; Gangadin et al., Reference Gangadin, Cahn, Scheewe, Hulshoff Pol and Bossong2021; Ghoshal & Conn, Reference Ghoshal and Conn2015; Grace, Reference Grace2010; Haber, Reference Haber2003; Haber, Reference Haber2016; Haber & Fudge, Reference Haber and Fudge1997; Harrison et al., Reference Harrison, Guell, Klein-Flügge and Barry2021; Kahn & Shohamy, Reference Kahn and Shohamy2013; Knöchel et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Nelson et al., Reference Nelson, Kraguljac, Maximo, Briend, Armstrong, Ver Hoef, Johnson and Lahti2022; Power et al., Reference Power, Schlaggar and Petersen2014; Qiu et al., Reference Qiu, Lu, Zhou, Yan, Du, Zhang, Xie and Zhang2021; Sah et al., Reference Sah, Faber, Lopez De Armentia and Power2003; Samudra et al., Reference Samudra, Ivleva, Hubbard, Rypma, Sweeney, Clementz, Keshavan, Pearlson and Tamminga2015; Song et al., Reference Song, Yang, Chang, Wei, Yin, Zhu, Zhou, Zhou, Jiang, Wu, Kong, Xu, Wang and Tang2022; Wang et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021; Xue et al., Reference Xue, Chen, Wei, Chen, Han, Wang, Zhang, Song and Cheng2023; Zhou et al., Reference Zhou, Shu, Liu, Song, Hao, Liu, Yu, Liu and Jiang2008), we hypothesised that individuals at CHR-P (in the placebo condition) would display lower CA1-NAc and CA1-vmPFC FC and altered CA1-amygdala FC compared to HC. Based on prior benzodiazepine challenge rs-fMRI studies in healthy individuals (Du & Grace, Reference Du and Grace2013; Gill et al., Reference Gill, Lodge, Cook, Aras and Grace2011; Lodge & Grace, Reference Lodge and Grace2007; Perez et al., Reference Perez, McCoy, Prevot, Mian, Carreno, Frazer, Cook, Sibille and Lodge2023), we hypothesised that a single dose of diazepam would increase CA1 FC within this circuit, to the extent that it would no longer differ from HC. For completeness, the following supplementary analyses were included: (i) using the anterior hippocampus as a seed (given it is specifically the anterior portion of the CA1 implicated in psychosis development (Schobel et al., Reference Schobel, Lewandowski, Corcoran, Moore, Brown, Malaspina and Small2009; Schobel et al., Reference Schobel, Chaudhury, Khan, Paniagua, Styner, Asllani, Inbar, Corcoran, Lieberman, Moore and Small2013)) and (ii) exploring broader effects of diazepam on CA1/anterior hippocampus FC with the rest of the brain.
Methods
Study design, participants, and procedure
This experimental medicine study was conducted at King’s College London. The study received ethical approval from the National Health Service UK Research Ethics Committee (18/LO/0618), and each participant gave written informed consent. While the study received ethical clearance as ‘not a Clinical Trial of an Investigational Medicinal Product’ by the EU directive 2001/20/EC, it was registered on clinicaltrials.gov (NCT06190483). Full study details, including inclusion/exclusion criteria, can be found in our recent publication describing the hippocampal cerebral blood flow findings in the same participants (Livingston et al., Reference Livingston, Kiemes, Devenyi, Knight, Lukow, Jelen, Reilly, Dima, Nettis, Casetta, Agyekum, Zelaya, Spencer, De Micheli, Fusar-Poli, Grace, Williams, McGuire, Egerton and Modinos2024). Briefly, this study used a randomised, double-blind, placebo-controlled, crossover design, whereby 24 antipsychotic-naïve individuals at CHR-P underwent MRI scanning on two occasions, once following a single oral dose of diazepam (5 mg) and once following an oral placebo (50 mg ascorbic acid). The diazepam/placebo capsule was administered 60 min before MRI scanning, and there was a minimum 3-week washout period between scans. Data from a group of 22 HC from a prior study (PSYAUD17/25) acquired with the same MRI scanner, scanning sequences, and acquisition parameters were used as a comparison group (Modinos, Egerton, McMullen, et al., Reference Modinos, Egerton, McMullen, McLaughlin, Kumari, Barker, Williams and Zelaya2018).
MRI acquisition
MRI data were acquired on a General Electric MR750 3.0 T MR scanner with an 8-channel head coil at the Centre for Neuroimaging Sciences, KCL. A 3D T1-weighted scan was acquired using a SPGR sequence and rs-fMRI data was acquired using a multi-echo echo planar imaging sequence (full acquisition details in Supplementary Methods). During the rs-fMRI scan, participants were instructed to remain awake with their eyes open, while a fixation cross was displayed in the centre of the screen.
Neuroimaging data processing
Preprocessing
The structural and rs-fMRI data were preprocessed using fMRIPrep (version 23.1.3) (Esteban, Markiewicz, Blair, et al., Reference Esteban, Markiewicz, Blair, Moodie, Isik, Erramuzpe, Kent, Goncalves, DuPre, Snyder, Oya, Ghosh, Wright, Durnez, Poldrack and Gorgolewski2019), SPM12 (Friston, Reference Friston, Ashburner, Kiebel and Nichols2007), CONN (Whitfield-Gabrieli & Nieto-Castanon, Reference Whitfield-Gabrieli and Nieto-Castanon2012), and FSL (Woolrich, Jbabdi, Patenaude, et al., Reference Woolrich, Jbabdi, Patenaude, Chappell, Makni, Behrens, Beckmann, Jenkinson and Smith2009). Structural images from both sessions were corrected for intensity non-uniformity using N4, skull-stripped, segmented, and averaged across sessions to generate a singular participant structural image which was then normalised to MNI space (1mm3 resolution) (Esteban et al., Reference Esteban, Markiewicz, Blair, Moodie, Isik, Erramuzpe, Kent, Goncalves, DuPre, Snyder, Oya, Ghosh, Wright, Durnez, Poldrack and Gorgolewski2019). For the rs-fMRI data, volume re-alignment and slice-timing correction parameters were calculated using the first echo and applied to all echoes (Esteban et al., Reference Esteban, Markiewicz, Blair, Moodie, Isik, Erramuzpe, Kent, Goncalves, DuPre, Snyder, Oya, Ghosh, Wright, Durnez, Poldrack and Gorgolewski2019). Participants were excluded if they moved >3 mm on any translation/rotation parameter or had a mean framewise displacement of >0.5 mm, as advised by prior methodological investigations (Power et al., Reference Power, Barnes, Snyder, Schlaggar and Petersen2012). The three echoes in native space underwent TE-dependent ICA-based denoising and were optimally combined using T2* weighted averaging via TEDANA (Kundu et al., Reference Kundu, Voon, Balchandani, Lombardo, Poser and Bandettini2017), before being normalised to MNI space (2mm3 resolution) with transformations generated during fMRIPrep (see supplementary materials for full boiler plate) (Esteban et al., Reference Esteban, Markiewicz, Blair, Moodie, Isik, Erramuzpe, Kent, Goncalves, DuPre, Snyder, Oya, Ghosh, Wright, Durnez, Poldrack and Gorgolewski2019). The denoised, optimally combined, normalised functional data was then spatially smoothed in SPM12 (Friston, Reference Friston, Ashburner, Kiebel and Nichols2007) with a 6 mm FWHM Gaussian kernel, and further denoised by removing white matter and CSF signal using the first five components of aCompCor, despiking, scrubbing, and band-pass filtering (0.008–0.09 Hz) in CONN (Whitfield-Gabrieli & Nieto-Castanon, Reference Whitfield-Gabrieli and Nieto-Castanon2012).
Generation of seed and region-of-interest masks
Hippocampal and subfield seed masks were generated for each participant from their preprocessed structural scan collected during their first scanning visit using the MAGeT Brain (multiple automatically generated templates of different brains) toolbox (Pipitone, Park, Winterburn, et al., Reference Pipitone, Park, Winterburn, Lett, Lerch, Pruessner, Lepage, Voineskos and Chakravarty2014) (see previous publication for further details (Livingston et al., Reference Livingston, Kiemes, Devenyi, Knight, Lukow, Jelen, Reilly, Dima, Nettis, Casetta, Agyekum, Zelaya, Spencer, De Micheli, Fusar-Poli, Grace, Williams, McGuire, Egerton and Modinos2024)). Using all participants’ CA1 segmentations, study-specific left and right CA1 masks were generated by using majority vote (ANTs/2.5.0; Figure 1). ROI masks for the cortico-limbic-striatal circuit (NAc, amygdala, and vmPFC) were derived from Neurosynth (https://www.neurosynth.org/) using the search terms ‘nucleus accumbens’, ‘amygdala’, and ‘vmPFC’ (uniformity tests). The resulting images were thresholded, binarised, and dilated (MINC toolkit; https://bic-mni.github.io/).
Within-group CA1-to-voxel functional connectivity. CA1-to-voxel functional connectivity networks averaged across each group independently (healthy controls, CHR-P placebo, and CHR-P diazepam) for the left and right CA1 subfield using study-specific masks (Z > 2.3, P FWE < 0.05). CHR-P, ‘clinical high-risk for psychosis’.
Neuroimaging data analysis
To control for the number of models, FDR correction was performed on all FWE-corrected second-level analyses described below.
First- and second-level analysis
To generate participant-level seed-to-voxel Z-maps, the mean functional time series was extracted from the left and right CA1 and used in first-level analysis models as regressors of interest in FSL. These first-level seed-to-voxel Z-maps were then entered into second-level analysis models using FLAME-1 (FMRIB’s Local Analysis of Mixed Effects) (Woolrich et al., Reference Woolrich, Jbabdi, Patenaude, Chappell, Makni, Behrens, Beckmann, Jenkinson and Smith2009), which employs Bayesian modelling and a weighted least-squares approach to perform a mixed-effects analysis. FLAME-1 was chosen as mixed-effects modelling is optimal for within-subject designs (i.e., CHR-P diazepam vs. placebo) to account for within-subject correlations, and FLAME-1 is able to estimate different variances for different groups of subjects within a model, which is advantageous for unpaired two-sample comparison (i.e., CHR-P vs. HC) (Sabaroedin, Tiego, & Fornito, Reference Sabaroedin, Tiego and Fornito2023). All models below use an FWE-corrected (P < 0.05) threshold of Z > 2.3. This threshold with FLAME-1 models has been shown to produce FWE rates lower than 5%, and is therefore similar to traditional FSL ordinary least square analyses using a threshold of Z > 3.1 (Eklund, Nichols, & Knutsson, Reference Eklund, Nichols and Knutsson2016).
Within-group CA1 resting-state FC analyses
Before comparing differences between groups/conditions, we first validated within-group resting-state FC networks for the CA1 to the whole brain to ensure they matched expected networks (one-sample contrast for each group independently) (Ezama et al., Reference Ezama, Hernández-Cabrera, Seoane, Pereda and Janssen2021).
Group and condition seed-to-ROI analyses
To investigate the effect of group (CHR-P placebo/diazepam vs. HC) and condition (CHR-P diazepam vs. placebo) on FC differences between CA1 and cortico-limbic-striatal circuit regions, we conducted seed-to-ROI analysis. Second-level models were run per seed-to-ROI per hemisphere for each group/condition comparison using a small volume adjustment approach by applying a pre-threshold ROI mask generated independently from Neurosynth as described above. Models were run both contralaterally (e.g., left CA1 to the right amygdala) and ipsilaterally (e.g., left CA1 to the left amygdala), as disruptions to both have been found across the psychosis spectrum within this circuit (Sabaroedin et al., Reference Sabaroedin, Tiego and Fornito2023). Voxel-level thresholding was used (Z > 2.3) for inference, which was FWE-corrected (P < 0.05) for multiple comparisons. Again, this threshold has been demonstrated to be quite conservative when using voxel-level inference in FLAME-1 models (Eklund et al., Reference Eklund, Nichols and Knutsson2016). For CHR-P placebo/diazepam versus HC models, age (mean centred) and sex were added in as covariates of no interest. For CHR-P diazepam versus placebo, change in pre-post scan fatigue score from the Bodily Symptoms Scale (Zuardi, Cosme, Graeff, & Guimarães, Reference Zuardi, Cosme, Graeff and Guimarães1993) for each condition was included as a covariate of no interest to control for drug effects of sedation/fatigue.
Supplementary/exploratory analyses
For completeness, supplementary analyses explored the effect of group/condition on FC between (1) anterior hippocampus and the cortico-limbic-striatal circuit (seed-to-ROI), and (2) hippocampal seeds (CA1 and anterior hippocampus) and the rest of the brain on a voxel-wise basis. Anterior hippocampal masks were generated by masking the study-specific averaged whole hippocampus segmentation with a hippocampus head mask derived from the Allen human reference atlas (Allen Human Reference Atlas - 3D, 2020), then thresholded, binarized, and dilated. Identical second-level models were run as described above, and for the seed-to-voxel analyses an inclusive grey matter mask was used during the pre-threshold masking.
Results
Demographics and clinical assessments
Following data quality checks, six CHR-P participants were excluded (n = 2 missing rs-fMRI data, n = 1 poor data quality, n = 3 excessive motion), along with 2 HC participants (n = 1 missing rs-fMRI data, n = 1 poor data quality). This resulted in a final sample of 18 CHR-P and 20 HC for analyses. Participant details can be found in Table 1. There were significant group differences in IQ and ethnicity in our sample. The IQ difference was driven by an above-average mean IQ in the HC group (mean ± SD = 122.9 ± 13.9, mean IQ in the general population for 20–29 years old = 100 (Silva, Reference Silva2008)), while the CHR-P group had an average mean IQ (mean ± SD = 96.9 ± 22.1). The difference in ethnicity was driven by a high proportion of white ethnicity in the HC group (70%) compared to the CHR-P group (55%). There were no significant differences in head motion parameters or change between pre- and post-scan Bodily Symptom Scale (Zuardi et al., Reference Zuardi, Cosme, Graeff and Guimarães1993) scores between the placebo and diazepam conditions.
Participant demographic information, clinical characteristics, head motion parameters, and fatigue scores
Note: The significant (i.e., <0.05) p values are shown in bold.
Abbreviations: CAARMS, ‘comprehensive assessment of at-risk mental states’; CHR-P, ‘clinical high-risk for psychosis’; HC, ‘healthy control’; IQ, ‘intelligent quotient’; WAIS, ‘Weschler adult intelligence scale’.
Resting-state functional connectivity
Within-group CA1 resting-state FC
Within each group/condition, as expected (Ezama et al., Reference Ezama, Hernández-Cabrera, Seoane, Pereda and Janssen2021), the CA1 showed significant FC with the rest of the hippocampus, extending to the temporal lobe, amygdala, precuneus, posterior cingulate cortex, mPFC, and parieto-occipital regions (Z > 2.3, P FWE < 0.05; Figure 1).
CA1-to-ROI
Compared to HC, individuals at CHR-P in the placebo condition showed significantly lower FC between the left CA1 and the right NAc (Figure 2A and Table 2), and between the right CA1 and the left NAc, left amygdala, and left vmPFC (Figure 2B and Table 2). Additionally, the right CA1 showed higher FC to the right amygdala (Figure 2B and Table 2). In the diazepam condition, these differences observed in the placebo condition compared to HC were ameliorated (no significant difference), apart from the right CA1 to left and right amygdala, which still showed significantly lower and higher FC compared to HC, respectively (Figure 2B and Table 2). We observed a significant drug effect on CA1-vmPFC FC, where diazepam (compared to placebo) significantly increased the FC strength from the left CA1 to left vmPFC and right CA1 to bilateral vmPFC (Figure 2 and Table 2).
Region-of-interest functional connectivity results for the CA1. Parameter estimates of functional connectivity strength between left (a) and right (b) CA1 and output regions (nucleus accumbens, amygdala, and ventromedial prefrontal cortex) displayed for healthy controls and individuals at clinical high-risk for psychosis (in the placebo and diazepam conditions) at peak coordinate of significant effect of group/condition (Z > 2.3, P FWE < 0.05). CA1 (green), amygdala (red), nucleus accumbens (yellow), and vmPFC (purple) are visualised on the brain using masks. CHR-P, ‘clinical high-risk for psychosis’; Amy, ‘amygdala’; NAc, ‘nucleus accumbens’; vmPFC, ‘ventromedial prefrontal cortex’; *** < 0.001; * < 0.05, ns, ‘not significant’.
Summary statistics of region-of-interest functional connectivity results for the CA1
Note: The significant (i.e., <0.05) p values are shown in bold.
Abbreviations: CHR-P, ‘clinical high-risk for psychosis’; FDR, ‘false discovery rate’; HC, ‘healthy control’; NAc, ‘nucleus accumbens’; ROI, ‘region-of-interest’; vmPFC, ‘ventromedial prefrontal cortex’.
Supplementary/exploratory analyses
At the whole-brain level, compared to HC, individuals at CHR-P in the placebo condition showed significantly higher FC between the right CA1 and a right medial temporal network, including the hippocampus, insula, and inferior/medial temporal gyri (Figure 3A and Supplementary Table S1). Conversely, lower FC was observed between the right CA1 and a left medial temporal network that extended to include key regions of the default mode network (bilateral mPFC, anterior cingulate cortex, and posterior cingulate cortex). In the diazepam condition, higher FC between right CA1 and a right medial temporal lobe network was also observed compared to HC, and additionally extended to parieto-occipital regions such as the angular gyrus (Figure 3B and Supplementary Table S1). When comparing CHR-P diazepam versus placebo conditions directly, no significant differences in whole-brain FC were observed for either the right or left CA1. Finally, there were no significant differences between groups (CHR-P placebo/diazepam vs. HC) or conditions (CHR-P diazepam vs. placebo) in FC strength using the anterior hippocampus as a seed on an ROI or whole-brain level. Since there was a difference in IQ between the groups, supplementary analyses correlated within each group IQ scores with hippocampal-ROI FC parameter estimates for models in which there was a significant difference between HC and CHR-P placebo groups. These analyses showed no significant correlations (Supplementary Table S2).
Voxel-wise whole-brain functional connectivity results for the CA1. Significant clusters showing differences (Z > 2.3, P FWE < 0.05) in functional connectivity between HC and CHR-P placebo (a) and CHR-P diazepam (b) for the CA1. Areas showing functional hyperconnectivity (CHR-P placebo/diazepam > HC) are displayed in red colourbar, whilst areas displaying functional hypoconnectivity are displayed in blue. N.B., no significant differences were found for the anterior hippocampus, nor for any of the regions (CA1 or anterior hippocampus) when contrasting CHR-P diazepam versus placebo. CHR-P, ‘clinical high-risk for psychosis’; HC, ‘healthy controls’.
Discussion
The main finding of the current study was that a single, non-sedating dose of the GABA-enhancing drug diazepam partially normalised CA1 dysconnectivity to a cortico-limbic-striatal circuit in individuals at CHR-P. More specifically, CHR-P individuals in the placebo condition (compared to HC) showed lower CA1-vmPFC and CA1-NAc FC. Diazepam significantly increased CA1-vmPFC FC compared to placebo, and the lower CA1-vmPFC and CA1-NAc FC observed in the placebo condition was normalised to HC levels in the diazepam condition. We observed more complex results for CA1-amygdala FC, as CHR-P individuals in the placebo condition showed lower and higher FC compared to HC, which were still present in the diazepam condition. Previously, we demonstrated that diazepam normalised increased hippocampal and subfield regional cerebral blood flow in the same CHR-P individuals, and here we extend this work by showing that diazepam can also partially normalise CA1 dysconnectivity to a downstream circuit. Taken together, these results indicate that GABA-enhancing compounds can rescue brain function in a psychosis-relevant circuit in CHR-P individuals, and therefore, show promise as a novel treatment strategy for clinical intervention in this group.
Our finding of lower CA1-vmPFC and CA1-NAc FC contralaterally (but normal FC ipsilaterally) in CHR-P individuals in the placebo condition (vs. HC) is consistent with prior rs-fMRI reports of subtle dysconnectivity in sub-clinical psychosis populations (i.e., most studies report lower FC but some studies report no differences) (Aberizk et al., Reference Aberizk, Sefik, Addington, Anticevic, Bearden, Cadenhead, Cannon, Cornblatt, Keshavan, Mathalon, Perkins, Stone, Tsuang, Woods and Walker2023; Allen et al., Reference Allen, Hird, Orlov, Modinos, Bossong, Antoniades, Sampson, Azis, Howes, Stone, Perez, Broome, Grace and McGuire2021; Edmiston et al., Reference Edmiston, Song, Chang, Yin, Zhou, Zhou, Jiang, Wei, Xu, Tang and Wang2020; Kozhuharova et al., Reference Kozhuharova, Saviola, Diaconescu and Allen2021; Liu et al., Reference Liu, Li, Liu, Yan, Wang, Sun, Fan, Song, Xu, Chen, Chen, Wang, Guo, Wan, Lv, Yang, Li, Lu, Yan and Liu2020; Tian et al., Reference Tian, Meng, Yan, Zhao, Liu, Yan, Han, Yuan, Wang, Yue, Zhang, Li, Zhu, He and Zhang2011; Waltmann et al., Reference Waltmann, O’Daly, Egerton, McMullen, Kumari, Barker, Williams and Modinos2019; Wang et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021). In contrast, studies in first-episode and chronic schizophrenia samples more consistently report lower FC between these regions, both contralaterally and ipsilaterally (Blessing et al., Reference Blessing, Murty, Zeng, Wang, Davachi and Goff2020; Cen et al., Reference Cen, Xu, Yang, Mei, Chen, Zhuo, Xiang, Song, Wang, Guo, Wang, Jiang, Xu, Li and Liu2020; Duan et al., Reference Duan, Gan, Yang, Cheng, Gao, Shi, Zhu, Liang and Zhao2015; Edmiston et al., Reference Edmiston, Song, Chang, Yin, Zhou, Zhou, Jiang, Wei, Xu, Tang and Wang2020; Fan et al., Reference Fan, Tan, Yang, Tan, Zhao, Chen, Li, Song, Wang, Jin, Zhou, Milham, Zou and Zuo2013; Gangadin et al., Reference Gangadin, Cahn, Scheewe, Hulshoff Pol and Bossong2021; Knöchel et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Kraguljac et al., Reference Kraguljac, White, Hadley, Hadley, Ver Hoef, Davis and Lahti2016; Liu et al., Reference Liu, Li, Liu, Yan, Wang, Sun, Fan, Song, Xu, Chen, Chen, Wang, Guo, Wan, Lv, Yang, Li, Lu, Yan and Liu2020; Nelson et al., Reference Nelson, Kraguljac, Maximo, Briend, Armstrong, Ver Hoef, Johnson and Lahti2022; Qiu et al., Reference Qiu, Lu, Zhou, Yan, Du, Zhang, Xie and Zhang2021; Samudra et al., Reference Samudra, Ivleva, Hubbard, Rypma, Sweeney, Clementz, Keshavan, Pearlson and Tamminga2015; Sarpal et al., Reference Sarpal, Robinson, Lencz, Argyelan, Ikuta, Karlsgodt, Gallego, Kane, Szeszko and Malhotra2015; Song et al., Reference Song, Yang, Chang, Wei, Yin, Zhu, Zhou, Zhou, Jiang, Wu, Kong, Xu, Wang and Tang2022; Wang et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021; Xue et al., Reference Xue, Chen, Wei, Chen, Han, Wang, Zhang, Song and Cheng2023; Zhou et al., Reference Zhou, Shu, Liu, Song, Hao, Liu, Yu, Liu and Jiang2008). This may suggest that in psychosis vulnerability stages, as hippocampal hyperactivity begins to drive glutamatergic input to the cortico-limbic-striatal circuit (Floresco et al., Reference Floresco, Todd and Grace2001; Ghoshal & Conn, Reference Ghoshal and Conn2015; Grace, Reference Grace2016), there is preserved or perhaps even elevated temporal coherence (i.e., increased FC) between the hippocampus and these other regions. As CHR-P symptoms persist, hippocampal hyperactivity and dysrhythmia may lead to excitotoxic loss of efferent projecting hippocampal glutamatergic neurons (Schobel et al., Reference Schobel, Chaudhury, Khan, Paniagua, Styner, Asllani, Inbar, Corcoran, Lieberman, Moore and Small2013) and consequentially uncoupling with downstream circuitry, which may further deteriorate following the onset of psychosis. Given that there are fewer connections contralaterally (e.g., right CA1 to left amygdala) than ipsilaterally (e.g., right CA1 to right amygdala) (Roesler, Parent, LaLumiere, & McIntyre, Reference Roesler, Parent, LaLumiere and McIntyre2021), it is likely that reduced FC would first be observed contralaterally, whilst ipsilateral connections may be preserved. In support of this model, experiments in MAM-treated rats demonstrated that NAc hyperactivity, due to hippocampal dysfunction, drives a striatal-midbrain circuit loop (Lodge & Grace, Reference Lodge and Grace2007) which increases phasic dopamine efflux in the NAc itself (Grace, Floresco, Goto, & Lodge, Reference Grace, Floresco, Goto and Lodge2007). Importantly, this increase in phasic dopamine can potentiate the hippocampal drive on the NAc (Goto & Grace, Reference Goto and Grace2005), which may result in reduced hippocampal-NAc FC. This inverse relationship of hippocampal hyperactivity and reduced hippocampal-striatal FC has been observed previously in individuals at CHR-P, as higher hippocampal glutamate levels (indicative of hyperactivity) was associated with reduced hippocampal-striatal FC (Allen et al., Reference Allen, Hird, Orlov, Modinos, Bossong, Antoniades, Sampson, Azis, Howes, Stone, Perez, Broome, Grace and McGuire2021). In accordance with this, reduced CA1-NAc FC was the most robust finding in our sample of individuals at CHR-P (i.e., it was observed bilaterally in the CA1), in whom we have previously demonstrated hippocampal hyperactivity (Livingston et al., Reference Livingston, Kiemes, Devenyi, Knight, Lukow, Jelen, Reilly, Dima, Nettis, Casetta, Agyekum, Zelaya, Spencer, De Micheli, Fusar-Poli, Grace, Williams, McGuire, Egerton and Modinos2024). Beyond illness chronicity, the more pronounced reductions observed in hippocampal FC in individuals with psychotic disorders compared to those at CHR-P might be related to antipsychotic treatment. For instance, we observed reductions in right, but not left, CA1-vmPFC FC in our sample of antipsychotic-naïve individuals at CHR-P, compared to the more robust observations in antipsychotic-treated individuals with schizophrenia (Fan et al., Reference Fan, Tan, Yang, Tan, Zhao, Chen, Li, Song, Wang, Jin, Zhou, Milham, Zou and Zuo2013; Knöchel et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Liu et al., Reference Liu, Li, Liu, Yan, Wang, Sun, Fan, Song, Xu, Chen, Chen, Wang, Guo, Wan, Lv, Yang, Li, Lu, Yan and Liu2020; Samudra et al., Reference Samudra, Ivleva, Hubbard, Rypma, Sweeney, Clementz, Keshavan, Pearlson and Tamminga2015; Song et al., Reference Song, Yang, Chang, Wei, Yin, Zhu, Zhou, Zhou, Jiang, Wu, Kong, Xu, Wang and Tang2022; Wang et al., Reference Wang, Yin, Sun, Jiang, Chao, Dai and Tang2021; Zhou et al., Reference Zhou, Shu, Liu, Song, Hao, Liu, Yu, Liu and Jiang2008). Whilst cognitive symptoms which are present in the prodrome may worsen following the onset of psychosis (Bora & Murray, Reference Bora and Murray2014; Catalan, de Pablo, Aymerich, et al., Reference Catalan, Salazar de Pablo, Aymerich, Damiani, Sordi, Radua, Oliver, McGuire, Giuliano, Stone and Fusar-Poli2021; Dong, Mao, Ding, et al., Reference Dong, Mao, Ding, Wang, Bo, Li, Wang and Wang2023), chronic antipsychotic treatment may also play a role in further cognitive impairment related to hippocampal-PFC FC uncoupling (Haddad et al., Reference Haddad, Salameh, Sacre, Clément and Calvet2023; McCutcheon, Keefe, & McGuire, Reference McCutcheon, Keefe and McGuire2023).
We found both higher and lower CA1-amygdala FC in individuals at CHR-P in the placebo condition compared to HC. Prior rs-fMRI studies have found lower (Tian et al., Reference Tian, Meng, Yan, Zhao, Liu, Yan, Han, Yuan, Wang, Yue, Zhang, Li, Zhu, He and Zhang2011; Xue et al., Reference Xue, Chen, Wei, Chen, Han, Wang, Zhang, Song and Cheng2023) and normal (Walther et al., Reference Walther, Lefebvre, Conring, Gangl, Nadesalingam, Alexaki, Wüthrich, Rüter, Viher, Federspiel, Wiest and Stegmayer2022) hippocampal-amygdala FC in individuals with psychotic disorders. However, hippocampal-amygdala FC was increased in people with schizophrenia with paranoia versus no paranoia (Walther et al., Reference Walther, Lefebvre, Conring, Gangl, Nadesalingam, Alexaki, Wüthrich, Rüter, Viher, Federspiel, Wiest and Stegmayer2022), and higher hippocampal-amygdala-PFC FC was associated with higher fear/anxiety in individuals with early psychosis (Feola, Beermann, Manzanarez Felix, et al., Reference Feola, Beermann, Manzanarez Felix, Coleman, Bouix, Holt, Lewandowski, Öngür, Breier, Shenton, Heckers, Brady, Blackford and Ward2024). Whilst amygdala dysfunction is associated with negative symptoms of schizophrenia (Ghoshal & Conn, Reference Ghoshal and Conn2015), it is also implicated in clinically distinct comorbid anxiety/mood disorders, which are more common in those at CHR-P (Fusar-Poli et al., Reference Fusar-Poli, Nelson, Valmaggia, Yung and McGuire2014; Achim et al., Reference Achim, Maziade, Raymond, Olivier, Mérette and Roy2011). This increased affective component might explain the higher hippocampal-amygdala FC observed in our sample of individuals at CHR-P compared to HC. Furthermore, the findings in our study appeared to be hemisphere-dependent (i.e., the right CA1 showed increased FC to the right amygdala and decreased FC to the left amygdala). This was also observed at the whole-brain level, whereby the right CA1 showed hyperconnectivity with a right medial temporal network, including the amygdala, but hypoconnectivity with a left hippocampal network and frontal regions of the left default mode network. Increased hippocampal FC with the medial temporal lobe has been observed previously in the psychosis spectrum (Avery, Rogers, McHugo, et al., Reference Avery, Rogers, McHugo, Armstrong, Blackford, Vandekar, Woodward and Heckers2022; Knöchel et al., Reference Knöchel, Stäblein, Storchak, Reinke, Jurcoane, Prvulovic, Linden, van de Ven, Ghinea, Wenzler, Alves, Matura, Kröger and Oertel-Knöchel2014; Li, Liu, Deng, et al., Reference Li, Liu, Deng, Li, Jiang, Li and Li2023), and therefore, according to the model outlined above, increased FC to the amygdala may be observed given the close proximity, number of bidirectional connections (van Staalduinen & Zeineh, Reference van Staalduinen and Zeineh2022), and proposed GABAergic alterations in this region in psychosis (Du & Grace, Reference Du and Grace2016). Furthermore, this pattern of intra-hemispheric hyperconnectivity and inter-hemispheric hypoconnectivity has been found previously in individuals with psychotic disorders, indicating increased local network segregation and decreased remote network integration (Hadley et al., Reference Hadley, Kraguljac, White, Ver Hoef, Tabora and Lahti2016).
The main effect of diazepam versus placebo in CHR-P individuals on CA1 FC to the cortico-limbic-striatal network was a bilateral increase in CA1-vmPFC FC. Furthermore, all decreases in CA1-vmPFC and CA1-NAc FC in CHR-P individuals in the placebo condition compared to HC were not present in the diazepam condition. The general direction of the drug effect (that is, increasing FC) is in line with our predictions and with prior pharmacological rs-fMRI studies using acute doses of GABA-enhancing drugs in healthy individuals (Blanco-Hinojo et al., Reference Blanco-Hinojo, Pujol, Macià, Martínez-Vilavella, Martín-Santos, Pérez-Sola and Deus2021; Feng, Yu, Wang, et al., Reference Feng, Yu, Wang, Wang, Park, Wilson, Deng, Hu, Yan and Kong2019; Flodin et al., Reference Flodin, Gospic, Petrovic and Fransson2012; Frölich et al., Reference Frölich, White, Kraguljac and Lahti2020; Licata et al., Reference Licata, Nickerson, Lowen, Trksak, Maclean and Lukas2013). GABA-enhancing drugs, such as diazepam, are positive allosteric modulators of the GABAA receptors via the benzodiazepine site (Engin, Benham, & Rudolph, Reference Engin, Benham and Rudolph2018). Most commonly, benzodiazepine binding leads to increased hyperpolarisation of post-synaptic glutamatergic pyramidal cells (Engin et al., Reference Engin, Benham and Rudolph2018), reducing their activity (Venkat, Chopp, & Chen, Reference Venkat, Chopp and Chen2016). The mechanism by which inhibition of neural activity in one brain region can result in increased FC to another has been recently elucidated by a chemogenetic fMRI study in mice. Rocchi and colleagues (Rocchi, Canella, Noei, et al., Reference Rocchi, Canella, Noei, Elorette, Fujimoto, Stoll, Fujimoto, Bienkowska, London, Fleysher, Russ and Rudebeck2022) demonstrated that either acute or chronic inhibition of the PFC led to increases in FC with direct thalamo-cortical output regions. The spiking activity was reduced but became more rhythmic and phase-locked to low-frequency oscillatory rhythms, leading to an increase in FC with connecting regions. Therefore, through this mechanism, it is likely that downregulation of hippocampal hyperactivity under diazepam (which we have demonstrated previously in this sample) led to increases in FC with connecting output regions.
Interestingly, the effect of diazepam on CA1-vmPFC FC showed the least inter-individual differences between people at CHR-P, whilst the effects in the amygdala and NAc were more varied. This may be due to the fact that the vmPFC, similar to the hippocampus, contains a high number of benzodiazepine receptors (Nørgaard, Beliveau, Ganz, et al., Reference Nørgaard, Beliveau, Ganz, Svarer, Pinborg, Keller, Jensen, Greve and Knudsen2021). Consequently, similar local effects on neural activity in the hippocampus and vmPFC might have also contributed to a more robust increase in temporal coherence between them. Increases in hippocampal-PFC FC under benzodiazepine versus placebo have previously been reported (Licata et al., Reference Licata, Nickerson, Lowen, Trksak, Maclean and Lukas2013), along with increases in FC to somatosensory and occipital regions,(Liang et al., Reference Liang, Zhang, Xu, Jia, Zang and Li2015; Wein, Riebel, Seidel, et al., Reference Wein, Riebel, Seidel, Brunner, Wagner, Nothdurfter, Rupprecht and Schwarzbach2024) which also have a high number of benzodiazepine binding sites (Nørgaard et al., Reference Nørgaard, Beliveau, Ganz, Svarer, Pinborg, Keller, Jensen, Greve and Knudsen2021). Furthermore, as noted earlier, the largest alterations in hippocampal FC observed in individuals at CHR-P in the placebo condition compared to HC were with the amygdala. This suggests that hippocampal-amygdala FC was the most perturbed out of the cortico-limbic-striatal regions. Given the proposed role of the amygdala in the initiation of hippocampal hyperactivity (Berretta et al., Reference Berretta, Lange, Bhattacharyya, Sebro, Garces and Benes2004) and PV+ interneuron loss (Zhu & Grace, Reference Zhu and Grace2023), and the high number of connections between these regions (van Staalduinen & Zeineh, Reference van Staalduinen and Zeineh2022), a single dose of diazepam may not have been sufficient to regulate altered hippocampal-amygdala FC in individuals at CHR-P. In support of this, benzodiazepines have been shown to either increase (Licata et al., Reference Licata, Nickerson, Lowen, Trksak, Maclean and Lukas2013) or decrease (Flodin et al., Reference Flodin, Gospic, Petrovic and Fransson2012) hippocampal-amygdala FC in healthy individuals. This suggests the pharmacological effects of GABA-enhancing compounds on this circuitry are inherently complex, without the presence of potential alterations to the GABAergic system in individuals at CHR-P.
Finally, we found no differences in FC strength between groups or drug conditions for the anterior hippocampus to the cortico-limbic-striatal circuit. This was unexpected, based on preclinical evidence (Grace, Reference Grace2010) and current theories about the pathophysiology of psychosis (Heckers & Konradi, Reference Heckers and Konradi2015; Knight et al., Reference Knight, McCutcheon, Dwir, Grace, O’Daly, McGuire and Modinos2022; Lieberman et al., Reference Lieberman, Girgis, Brucato, Moore, Provenzano, Kegeles, Javitt, Kantrowitz, Wall, Corcoran, Schobel and Small2018). However, the anterior hippocampus contains subfields beyond the CA1 and subiculum, such as the CA2/3, which largely only have intra-hippocampal projections (Shinohara & Kohara, Reference Shinohara and Kohara2023). Therefore, inclusion of this signal may increase noise, making it difficult to detect subtle FC alterations between the anterior hippocampus and the cortico-limbic-striatal circuit within individuals at CHR-P. In line with this, whilst preclinical evidence focuses on the anterior hippocampus, it specifically identifies the anterior CA1 as the site of dysfunction (Gergues, Han, Choi, et al., Reference Gergues, Han, Choi, Brown, Clausing, Turner, Vainchtein, Molofsky and Kheirbek2020).
This study had several strengths. We used a gold standard randomised, double-blind, placebo-controlled, crossover study design in a sample of antipsychotic-naïve individuals at CHR-P. The hippocampus and CA1 subfield were segmented with a high degree of accuracy using novel computational methods (Pipitone et al., Reference Pipitone, Park, Winterburn, Lett, Lerch, Pruessner, Lepage, Voineskos and Chakravarty2014), allowing the generation of study-specific hippocampal and subfield masks. We acquired rs-fMRI data using an advanced multi-echo sequence, allowing robust data cleaning and removal of non-physiological noise with advanced methodological techniques such as TEDANA (Kundu et al., Reference Kundu, Voon, Balchandani, Lombardo, Poser and Bandettini2017). This led to high-quality data, as within-group/condition resting-state FC networks for the CA1 to the rest of the brain replicated those found previously (Ezama et al., Reference Ezama, Hernández-Cabrera, Seoane, Pereda and Janssen2021). We were able to contextualise baseline differences and direction of drug effects in the CHR-P group by comparing them with data from a HC group. Finally, we used advanced statistical mixed-effects modelling, which is optimal for examining both inter-group differences without assuming uniform variance and also for investigating within-subject effects (Beckmann, Jenkinson, & Smith, Reference Beckmann, Jenkinson and Smith2003). This study also had some limitations. Our sample size of CHR-P individuals was reduced from 24 down to 18 after quality control, but retrospective power analysis demonstrated that the diazepam versus placebo analyses (mean Cohen’s d = 0.83) had an achieved power of 91%. Additionally, this study was not powered to investigate the relationship between FC alterations and symptoms, which would require a much larger CHR-P sample. Due to limitations with the resolution of rs-fMRI, we were not able to investigate differences in FC from specifically the anterior CA1 (our study-specific mask only had ~30 voxels per hemisphere), which is of particular relevance for psychosis (Grace, Reference Grace2010).
In conclusion, this study provides evidence that a single dose of a non-specific GABA-enhancing drug, such as diazepam, can normalise CA1 FC alterations with the vmPFC and NAc in individuals at CHR-P. Conversely, CA1-amygdala FC was greatly perturbed in people at CHR-P under placebo compared to HC and was largely unaffected by diazepam challenge. Given this mechanistic evidence, future research is warranted with extended treatment durations to link these neurobiological changes to symptoms and clinical outcomes, including psychosis prevention.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725101268.
Funding statement
This research was funded by the Wellcome Trust and The Royal Society (202397/Z/16/Z to Gemma Modinos) and the National Institute for Health and Care Research (NIHR) Maudsley Biomedical Research Centre (BRC). The views expressed are those of the authors and not necessarily those of the Wellcome Trust, NIHR, or the Department of Health and Social Care. For the purpose of open access, the author has applied a CC-BY public copyright licence to any Author Accepted Manuscript version arising from this submission. Nicholas R. Livingston was funded by an MRC DTP PhD studentship at the time of data collection and analysis. Paulina B. Lukow was in receipt of a PhD studentship funded by the NIHR Maudsley BRC at the time of data collection. Owen O’Daly is funded by the Maudsley BRC. Luke A. Jelen was supported by an MRC Clinical Research Training Fellowship (MR/T028084/1) at the time of data collection. Thomas J. Reilly is supported by an MRC Clinical Research Training Fellowship (MR/W015943/1). PFP is supported by the European Union funding within the MUR PNRR Extended Partnership initiative on Neuroscience and Neuropharmacology (Project no. PE00000006 CUP H93C22000660006 “MNESYS, A multiscale integrated approach to the study of the nervous system in health and disease”). Anthony A. Grace received funding from USPHS NIMH MH57440. M. Mallar Chakravarty receives salary support from the Fonds de Recherche Québec – Santé and from a James McGill Professorship. M. Mallar Chakravarty also receives research support from Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council – Canada, McGill University’s Health Brains for Health Lives (a Canada Research Excellence Fund Iniative), and TRIDENT (a New Frontiers in Research Fund program).
Competing interests
Anthony A. Grace has received consulting fees from Alkermes, Lundbeck, Takeda, Roche, Lyra, Concert, Newron, and SynAgile, and research funding from Newron and Merck. Steve C. R. Williams has recently received research funding from Boehringer Ingelheim and GE Healthcare to perform investigator-led research. Alice Egerton has received consultancy fees from Leal Therapeutics. Gemma Modinos has received consulting fees from Boehringer Ingelheim. The remaining authors have no disclosures to declare.
Ethical standard
The authors assert that 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.
Open access
