1 Manuscript Title Page Title: The Amnesic Effects of Propofol on the Hippocampus Version: Accepted Authors: David Lindsay1,2,3*, Ram M Adapa1,3, David K Menon1,2,3,4 & Emmanuel A Stamatakis1,2 Author Ins7tu7onal Affilia7ons: 1) University of Cambridge Division of Anaesthesia, Cambridge, UK. 2) University of Cambridge Department of Clinical Neurosciences, Cambridge, UK. 3) Cambridge University Hospitals NHS FoundaPon Trust, Cambridge, UK. 4) University of Cambridge Department of Medicine, Cambridge, UK. *Corresponding Author: David Lindsay *Corresponding Author Address: University of Cambridge Division of Anaesthesia, Cambridge, United Kingdom. *Corresponding Author Email: dl372@cam.ac.uk 2 Abstract Background: One of the primary acPons of general anaesthePc agents, apart from inducing a state of unconsciousness, is reversible impairment of memory formaPon during the period of administraPon. Failure to induce and maintain amnesia may result in recall of accidental intraoperaPve awareness and contribute to adverse psychological health outcomes. The precise mechanism(s) of acPon by which general anaesthePc drugs achieve their amnesic effects is not fully understood. To this end we focused on the hippocampus, a region criPcal for the formaPon of new episodic explicit long-term memories of the type normally inhibited by general anaesthePcs. Methods: 25 healthy adult volunteers underwent funcPonal magnePc resonance neuroimaging whilst sedated with a plasma target controlled infusion of the anaesthePc agent propofol. The funcPonal connecPvity (synchronised neuronal acPvity with other brain regions) of the hippocampus and microanatomical hippocampal subregions was assessed at baseline, under sedaPon and during recovery. Serial plasma propofol concentraPons and responses to an auditory sPmuli semanPc decision task were measured. Post-scanning memory tesPng was conducted and memory performance was related to the fMRI data. Results: FuncPonal connecPvity changes associated with an amnesic but sub-hypnoPc depth of propofol sedaPon were predominantly characterised by a reduced connecPvity signature of the hippocampus stratum radiatum, stratum lacunosum, stratum moleculare, CA1 stratum pyramidalis and CA4/Dentate Gyrus subfields with the precuneus. Conclusions: This study provides evidence for differenPal acPons of propofol on hippocampal subdivisions and limbic circuits related to amnesic efficacy and suggests a more significant role of the precuneus in long-term memory consolidaPon than previously thought. Key Words AAGA, Amnesia, fMRI, General Anaesthesia, Hippocampus, Memory, Propofol, SedaPon 3 Introduc7on Since the advent of modern anaesthesia, the primary funcPon of general anaesthePc agents has been to induce a rapid and fully reversible state of unconsciousness. However, one of the other major effects of general anaesthePcs is anterograde amnesia, specifically inhibiPon of episodic explicit long-term memory formaPon, which may occur at sub- hypnoPc dosages.1-2 Despite extensive clinical use, the mechanisms by which general anaesthePc drugs disrupt memory formaPon remain poorly understood. Accidental awareness under general anaesthesia (AAGA) with recall remains an uncommon but potenPally life changing complicaPon of general anaesthesia.3 For AAGA with explicit recall to occur, both failure of hypnosis and failure to prevent memory formaPon are required. Just as the risk of AAGA varies between anaesthePc agents (albeit dwarfed by a range of confounding factors), the amnesic efficacy of anaesthePc drugs relaPve to their hypnoPc potency also varies by agent. Previous experiments by Veselis and colleagues found both propofol and midazolam to be profoundly amnesic at sub-hypnoPc doses, whilst thiopentone and dexmedetomidine both had only mild amnesic effects and fentanyl alone effecPvely negligible memory effects.1-2,4 This is consistent with theories suggesPng thiopentone and dexmedetomidine primarily cause amnesia by impairing percepPon and memory encoding through sedaPon/unconsciousness, whereas propofol and midazolam primarily act on early memory consolidaPon (the subsequent stage of the memory formaPon and storage sequence) to exert their amnesic effects, potenPally independently of the level of sedaPon.2,5-6 A range of clinical case studies, animal experiments and neuroimaging evidence have highlighted the central role of the hippocampus in early memory consolidaPon.7-10 The hippocampi are bilateral, convoluted archicorPcal structures located in the medial temporal lobes of the brain.11 Localized lesions in this area acquired in adulthood are associated with catastrophic impairment of new episodic memory formaPon capacity, despite apparently normal short-term memory and recall of events from long before the lesion event.9 The hippocampi are histologically divided into several subregions running longitudinally.11-13 Different subdivisions have been associated with different memory funcPons in animal models at mulPple scales, with a range of long term spaPal, paeern and social memory types localised to specific mesoscopic subregions and also at a cellular level, encoding engrams and spaPal posiPoning informaPon.9,14 Two major neural circuits have been idenPfied within the hippocampus running from the entorhinal cortex to the subiculum: the (monosynapPc) direct pathway and the trisynapPc pathway, as well as a third pathway via the smaller CA2 region (Fig 1).9 The direct pathway is thought to be sufficient for some types of memory formaPon without input from the trisynapPc pathway, which may facilitate rapid learning from single episodic experiences.15 FuncPonal magnePc resonance imaging (fMRI) is a neuroimaging technique that uPlises the difference in paramagnePc/diamagnePc suscepPbility between 4 deoxyhaemoglobin and oxyhaemoglobin respecPvely to infer changes in regional cerebral blood flow, and in turn, underlying changes in local neuronal acPvity.16-17 Pryor and colleagues have previously idenPfied differenPal effects of propofol on anterior and posterior divisions of the hippocampus using blood oxygen level dependent (BOLD) funcPonal magnePc resonance imaging.18 Their findings include a decrease in posterior hippocampus acPvaPon proporPonate to decreased memory performance.18 We hypothesise that propofol inhibits early memory consolidaPon by differenPally acPng on specific pathways within the hippocampus as a mechanism of propofol-induced amnesia. To characterise this, an experiment was performed using a novel microanatomical hippocampal subregion segmentaPon in a seed-to-voxel resPng-state funcPonal connecPvity (FC) analysis. This allowed us to invesPgate the effects of propofol sedaPon on hippocampal subregion funcPonal connecPvity, in associaPon with quanPfied drug-induced amnesic effects and measured plasma propofol levels. The results of this may offer insights into both the mechanisms of general anaesthePc acPon and why those mechanisms somePmes fail. Fig 1 – SchemaPc of simplified hippocampal formaPon internal circuits (major inputs and connecPons only, more minor pathways not shown). The Direct Pathway is highlighted in red and the TrisynapPc Pathway in purple [EC = Entorhinal Cortex, DG = Dentate Gyrus, CA = Cornu Ammonis]. EC Hypothalamus DG CA3 CA2 CA1 Subiculum Mossy Fibers Schaffer Collaterals Layer II Layer II Layer III Direct Pathway Trisynaptic Pathway 5 Materials and Methods The experimental protocol used for primary data acquisiPon has been previously described and is summarised below.19-20 Twenty-five healthy, right-handed, naPve English speaker, adult volunteers were recruited and assessed before wrieen informed consent for study parPcipaPon was obtained (age range 19 – 53 years, mean 34 years old; sex raPo 11:14 Male:Female). Ethical approval for this study was prospecPvely confirmed by the Cambridgeshire 2 Research Ethics Commieee. All volunteers were required to follow standard fasPng guidelines and safety procedures. Two senior anaesthePsts and dedicated monitoring equipment provided conPnuous monitoring throughout. The performance, safety and magnePc resonance (MR) image quality profiles of the target-controlled infusion (TCI) pharmacokinePc pumps (Carefusion – Alaris Products, Basingstoke, UK) were evaluated inside the MR scanner suite and confirmed to be within accepted limits in an earlier study.21 As the study was originally powered for a different primary analysis, sample size was not specifically pre-determined for the work reported here.20 Seda7on Volunteers were sedated with propofol (Fresenius Propoven 1%, Fresenius Kabi, Cheshire, UK) using a Marsh-model propofol TCI infusion to a sedaPon target, CpT = 1.2 μg ml-1, with a period of 10 minutes included to allow for plasma-effect site equilibraPon.22 Plasma propofol concentraPons were measured on blood samples drawn arer the equilibraPon period for the sedaPon phase and arer recovery phase imaging (iniPated 20mins arer the propofol infusion was terminated) using high performance liquid chromatography.19 PharmacokinePc model accuracy was assessed using mean predicPon error (EquaPon 1) and median absolute performance error (EquaPon 2) based on measured (mCp) and model predicted (Cp) plasma propofol concentraPons.22-23 Equation 1 𝑃𝐸$% = 𝑚𝐶𝑝 – 𝐶𝑝 𝐶𝑝 × 100 Equation 2 𝐴𝑃𝐸$% = |𝑃𝐸$%| 6 Memory & Behavioural Tes7ng At the pre-sedaPon baseline and sedaPon stages, volunteers were presented with auditory word sPmuli using the CogniPon and Brain Sciences Unit Audio SPmulaPon Tool and required to select whether the word referred to a ‘living’ or ‘non-living’ object using a bueon-box immediately before the acquisiPon of resPng-state sequences.19 ‘Time-Out’s (no response within 3 seconds), error rate and response Pme were recorded in real Pme. Memory tesPng was performed when the volunteers were fully recovered and a minimum of 30 minutes arer cessaPon of the propofol infusion, using a mixture of old (previously heard) and new word sPmuli. The Explicit Recall SensiPvity Index (ERSI), d¢ (the difference between the z-transformed ‘hit rate’, H (the proporPon of Pmes a subject correctly idenPfied a sPmulus as previously heard) and the ‘false alarm rate’, F (the proporPon of Pmes a subject incorrectly idenPfied a sPmulus as old when it was actually new)) was calculated (EquaPon 3).18 Equation 3 d' = z(H) – z(F) Memory and behavioural data were related to plasma propofol target concentraPon using Cohen’s d and Wilcoxon signed-rank tests, as well as to measured plasma propofol concentraPon using Spearman’s rank correlaPon coefficient, linear regression and Cohen’s f2 staPsPcal tests. Image Acquisi7on Structural and resPng-state BOLD funcPonal magnePc resonance images were acquired using a Siemens Trio Tim 3 Tesla MRI System (Wolfson Brain Imaging Centre, Cambridge). Each funcPonal run (baseline, sedaPon & recovery phases) included 150 volumes of 32 interleaved 3mm slices with an interslice-gap of 0.75mm and 3mm x 3mm x 3mm resoluPon, acquisiPon Pme of 2s, echo Pme of 30ms, flip angle of 78O, and bandwidth of 2442 Hz Px-1. T1-weighted structural images were acquired using a magnePsaPon prepared rapid gradient echo (MPRAGE) sequence with 1mm x 1mm x 1mm resoluPon, repePPon Pme of 2250ms, inversion Pme of 900ms, echo Pme of 2.99ms and flip angle of 9O. Image Processing & Analysis Arer removing the first 5 volumes of each funcPonal run, MR Images were pre- processed using SPM12 (StaPsPcal Parametric Mapping v12, update revision 7771, Wellcome Centre for Human Neuroimaging, UCL, UK) running on MATLAB R2021b (9.11.0.1769968) (MathWorks, NaPck, MA, USA). The slice-Pming corrected, realigned and 7 spaPally normalised funcPonal images were wrieen to a super-sampled final resoluPon of 2mm x 2mm x 2mm. FuncPonal connecPvity analysis was performed using regressor Pme series extracted from unsmoothed images with the CONN FuncPonal ConnecPvity Toolbox (version CONN22.a) and its default FSL Harvard-Oxford Atlas and Automated Anatomical Labelling atlases.24-25 The McGill ComputaPonal Brain Anatomy Laboratory (CoBrALab) 3T in vivo High ResoluPon Atlas of the Hippocampus and Subfields was used to generate hippocampus subfield regions-of-interest (ROIs).26-28 Due to MRI atlas resoluPon limits, several histological subfields are grouped together into joint ROIs: CA2 & CA3 stratum pyramidalis subfields (CA2/CA3), CA4 stratum pyramidalis & Dentate Gyrus subfields (CA4/DG), stratum radiatum/stratum lacunosum/stratum moleculare layers (Stratum). Other mapped ROIs include the: CA1 stratum pyramidalis subfield, subiculum, alveus, fimbria, fornix, mammillary bodies and the amygdala. Denoising (aCompCor), moPon regression, BOLD Pme series band- pass frequency filtering (0.008Hz – 0.09Hz) and linear detrending pipelines were applied as implemented in CONN. Within-subject seed-to-voxel connecPvity analysis was performed using weighted General Linear Models (GLMs) for each seed-target area by compuPng the correlaPons between each ROI’s Pme-series and brain-wide target voxels. Between-subject group analysis was performed using GLMs esPmated for each voxel with relevant first-level connecPvity measures for each subject. Voxel cluster-level inferences were performed using Gaussian Random Field Theory two-sided parametric staPsPcs.29 Unless stated otherwise, results are reported with staPsPcal thresholding set at p < 0.001 uncorrected cluster-forming voxel-level threshold and p-FDR < 0.05 cluster-size threshold, corrected for mulPple comparisons using False Discovery Rate correcPon.17,30 8 Results The mean (standard deviaPon) measured plasma propofol concentraPon (mCp) across all parPcipants was 0.70 μg ml-1 (0.32 μg ml-1) during sedaPon (CpT = 1.2 μg ml-1) and 0.27 μg ml-1 (0.077 μg ml-1) during the recovery phase [Wilcoxon signed-rank test p = 4.2x10- 4] (Fig 2a). The mean predicPon error (a measure of bias) for the pharmacokinePc model was -42%. The median absolute performance error (a measure of precision) was 52%. In all cases memory performance of the sedaPon period sPmuli was decreased compared to pre-sedaPon baseline, with a mean (standard deviaPon) decrease in Explicit Recall SensiPvity Index from 1.00 (0.488) to 0.264 (0.413) [p = 1.9x10-5] (Fig 2a). Mean (standard deviaPon) Response Time increased from 1110ms (160ms) pre-sedaPon to 1230ms (206ms) under sedaPon [p = 0.011]. During the propofol infusion several volunteers exhibited markedly decreased responsiveness to auditory sPmuli, with a Time-Out rate of 100% in one individual, however most remained fully responsive with a modal Time Out rate of 0% and median Time Out rate of 2.5% [p = 0.098]. The difference in error rate was also not staPsPcally significant [p = 0.18] Propofol infusion (CpT) had a proporPonately larger effect on explicit memory performance (Cohen’s d = 1.60) than any proxy-marker of responsiveness or cogniPve performance, being 2.5 Pmes greater than the effect on Response Time (d = 0.65), 3.0 Pmes greater than the effect on Time Out rate (d = 0.53) and 6.8 Pmes greater than the effect on Error Rate (d = 0.23) for the semanPc decision task at a group level. Changes in memory performance and responsiveness also correlated with measured plasma propofol concentraPon (mCp) in all domains, although this was not staPsPcally significant in the case of Response Time and Error Rate (Fig 2b-e). A previous analysis demonstrated no detectable priming or familiarity implicit memory component in memory tesPng responses.30 ResPng-state fMRI (rs-fMRI) responses have been previously shown to correlate with task performance.32-33 No significant FC paeerns were idenPfied as aeributable to confounding factors when performing a regression analysis for age, sex, height or weight in addiPon to moPon regression in funcPonal images. As one of the most criPcal structures required for new episodic long-term memory formaPon, the resPng-state funcPonal connecPvity of the whole hippocampus was assessed at baseline (pre-sedaPon) using a ler and right averaged ROI Pme series, generated using the Harvard-Oxford atlas. The hippocampus as a whole displayed a posiPve acPvity correlaPon (i.e. connecPvity) with the precuneus, right precentral gyrus, brainstem, ler frontal orbital cortex and right frontal pole (Table 1, Fig 3a). Assessing the effect of propofol sedaPon on overall hippocampus funcPonal connecPvity, a paired t-test analysis between sedaPon and baseline condiPons demonstrated a significant decrease in connecPvity with two clusters in the precuneus (Table 1, Fig 3b). 9 Fig 2 – a) Violin-Box Plots of Explicit Recall SensiPvity Index (ERSI), Response Time (RT), Time Out Rate (TO), Error Rate (ER) & measured plasma propofol concentraPon (mCp) by experimental phase [*p<0.05, **p<0.01, ***p<0.001]. Change from awake baseline in b) Explicit Recall SensiPvity Index, c) Response Time, d) Time Out Rate & e) Error Rate against measured plasma propofol concentraPon during pseudosteady-state TCI sedaPon with linear regression (blue) and 95% confidence intervals (grey) [r = Spearman’s Rank CorrelaPon Coefficient, f2 = Cohen’s f2, LR-p = Linear Regression p-value] Plasma Propofol Concentration (μg ml-1) Δ E xp lic it R ec al l S en si tiv ity In de x (u ni tle ss ) ρ = -0.412 f2 = 0.263 LR-p = 0.025 b Plasma Propofol Concentration (μg ml-1) Δ R es po ns e Ti m e (m s) c ρ = 0.261 f2 = 0.066 LR-p = 0.240 d Plasma Propofol Concentration (μg ml-1) Δ T im e O ut R at e (u ni tle ss ) ρ = 0.428 f2 = 0.349 LR-p = 9.40 x10-3 e Plasma Propofol Concentration (μg ml-1) Δ E rr or R at e (u ni tle ss ) ρ = 0.371 f2 = 0.079 LR-p = 0.190 0.4 0.8 1.2 1.6 -0.1 0 0.1 0.2 0 -0.5 0.5 1.0 0 -0.5 -1.5 -1.0 -2.0 -300 0 300 600 a Sedation Phase Awake Sedation ERSI RT TO ER mCp Awake Sedation Awake Sedation Awake Sedation Sedation Recovery (ms) (μg ml-1) 0 -1 1 2 3 2000 1500 1000 1 0.75 0.5 0.25 0 0 0.1 0.2 0.3 0.4 0 0.5 1 1.5 2 ✱✱✱ ✱✱✱✱ 0.4 0.8 1.2 1.6 0.4 0.8 1.2 1.60.4 0.8 1.2 1.6 10 The effect of stopping propofol sedaPon was also assessed for potenPal reciprocal changes. A paired t-test between recovery and sedaPon condiPons demonstrated an increase in FC with the fronto-medial cortex and right superior frontal gyrus. AnPcipaPng a return of funcPonal connecPvity with the precuneus, relaxing the voxel threshold to p- uncorrected < 0.01 did indeed display reciprocal increased connecPvity with the posterior cingulate gyrus and precuneus, albeit more superiorly than the focus of decreased FC observed during sedaPon earlier (Table 1, Fig 3c). It is important to note, however, that this threshold does not meet generally accepted significance levels. A paired t-test between recovery and baseline condiPons did not show any significant FC changes from baseline, suggesPng an effecPve return to normal rather than sub-normal or supra-normal FC. Fig 3 – Whole Hippocampus ROI, resPng-state funcPonal connecPvity overlay on standard structural image (Colour scale key: T-staPsPc; Slice levels: Montreal Neurological InsPtute z- axis coordinate): a) Baseline, b) Difference of SedaPon > Baseline, c) Difference of Recovery > SedaPon [Voxel threshold p-uncorrected < 0.01 in c] a b c Anterior Posterior Left Right 11 To invesPgate the role of individual hippocampal subregions, the hippocampus was divided into separate subfield ROIs using the CoBrALab 3T in vivo High ResoluPon Atlas of the Hippocampus and Subfields.26-28 Three subfields were idenPfied which corroborated the paeern of decreased FC with a cluster centred on the precuneus during sedaPon, as observed in the whole hippocampus ROI: CA1, the subiculum and the combined ‘Stratum’ subregions (Table 2). The CA4/Dentate Gyrus subfield also displayed decreased FC with the precuneus but this did not reach robust staPsPcal significance. Fig 4 – Change from baseline in funcPonal connecPvity of combined ‘Stratum’ ROI under sedaPon weighted by change in Explicit Recall SensiPvity Index. Orthogonal views of staPsPcally significant clusters with 3D reconstrucPon including whole hippocampus (green). (Colour scale key: T-staPsPc; Coordinates: Montreal Neurological InsPtute) 12 The subiculum also demonstrated a rebound increase in funcPonal connecPvity with the precuneus and posterior cingulate gyrus when contrasPng recovery and sedaPon phases, however this was again only at a relaxed voxel threshold (p-uncorrected < 0.005), whilst the CA1 and Stratum subfields did not show any reciprocal FC change during Recovery. In order to separate the BOLD FC changes specific to the reduced memory funcPon observed from other more general sedaPon-related and non-specific neurovascular effects, the volunteers’ Explicit Recall SensiPvity Index (d¢) scores, which can be considered a proxy measure of accurate explicit memory formaPon, retenPon and recall, were then also incorporated into the general linear model. Two further volunteers for whom a change in d¢ score and/or Response Time could not be calculated (Time-Out rate >75%) were excluded from further analysis. When weighPng for the d¢ scores, the only FC change during sedaPon observed for the hippocampus as a whole was an increase in connecPvity with a cluster centred in the thalamus. Some hippocampal subfields conPnued to display a decrease in FC with the precuneus though. The Stratum combined subfield displayed decreased FC with a cluster in the superior aspect of the precuneus and increased FC with the thalamus (Table 2, Fig 4). The CA4/Dentate Gyrus and CA1 subfields displayed similar FC changes but were staPsPcally sub-threshold. The subiculum did not show any staPsPcally significant FC changes. To further separate amnesia-specific FC changes from general sedaPon and non- specific changes, the change in Response Time, which may be considered a proxy measure of alertness and cogniPve processing speed, and measured plasma propofol concentraPons (mCp) were also included in the general linear model in turn. There was no staPsPcally significant change in FC aeributable to alertness (Response Time) per se. mCp-related changes included an increase in FC between the whole hippocampus and the medial prefrontal cortex, right inferior temporal gyrus and cerebellum during sedaPon and a decrease in FC with the right postcentral gyrus on recovery from sedaPon, which was also reflected in the main hippocampal subfields individually. 13 Cluster Rank MNI Coordinates Cluster Size (voxels) Size p- FDR Correlation (+/-) / FC Change (­/¯) Main Location x y z Whole Hippocampus Baseline FC 1 +22 -08 -22 47069 <1 x10-6 Positive Precuneus 2 -04 -06 +54 1185 <1 x10-6 Positive Right Precentral Gyrus 3 -04 -16 -46 345 <1 x10-6 Negative Brainstem 4 -30 +30 -12 171 7.7 x10-5 Positive Left Frontal Orbital Cortex 5 +26 +16 +16 143 2.5 x10-4 Negative Right Frontal Pole 6 -54 -36 +38 137 2.8 x10-4 Negative Left Posterior Supramarginal Gyrus 7 +08 -20 +26 132 3.1 x10-4 Negative Posterior Cingulate Gyrus 8 +30 +32 -14 124 4.2 x10-4 Positive Right Frontal Pole 9 -20 +16 +22 112 7.2 x10-4 Negative White Matter (Left Frontal Lobe) 10 -22 +36 +02 91 2.2 x10-3 Negative White Matter (Left Frontal Lobe) 11 -28 -68 -62 82 3.4 x10-3 Positive Left Cerebellum 12 00 -02 +18 73 5.5 x10-3 Negative Corpus Callosum 13 +42 +32 +16 57 1.5 x10-2 Positive Right Inferior Frontal Gyrus 14 +30 +36 +04 54 1.7 x10-2 Negative White Matter (Right Frontal Lobe) Whole Hippocampus Difference of Sedation > Baseline 1 -04 -58 +10 241 2.9 x10-5 ¯ Precuneus 2 +10 -52 +08 76 4.1 x10-2 ¯ Precuneus Whole Hippocampus Difference of Recovery > Sedation 1 +02 +54 +30 465 4.9 x10-4 ­ Left Frontal Pole 2 +06 -44 +32 454 4.9 x10-4 ­ Posterior Cingulate Gyrus/Precuneus 3 -02 +36 -14 359 1.9 x10-3 ­ Frontal Medial Cortex 4 +46 +08 -02 346 1.9 x10-3 ¯ Right Insular Cortex Table 1 – Whole Hippocampus ROI, resPng state funcPonal connecPvity clusters with correlaPon at A) Baseline and with change in FC comparing B) Difference of SedaPon > Baseline & C) Difference of Recovery > SedaPon [Voxel threshold p-uncorrected < 0.01 in C (above generally accepted significance thresholds)] 14 Cluster Rank MNI Coordinates Cluster Size (voxels) Size p- FDR FC Change Main Location x y z Stratum Subfield Difference of Sedation > Baseline 1 -06 -50 +10 142 2.5 x10-3 ¯ Precuneus CA1 Subfield Difference of Sedation > Baseline 1 -06 -50 +10 118 9.4 x10-3 ¯ Precuneus Subiculum Subfield Difference of Sedation > Baseline 1 -06 -58 +14 122 5.6 x10-3 ¯ Precuneus Whole Hippocampus Difference of Sedation > Baseline factoring ERSI (d¢) 1 -10 -16 +18 118 2.6 x10-3 ­ Left Thalamus Stratum Subfield Difference of Sedation > Baseline factoring ERSI (d¢) 1 +02 -68 +46 202 9.6 x10-5 ¯ Precuneus 2 -10 -16 +20 103 5.4 x10-3 ­ Left Thalamus CA4/DG Subfield Difference of Sedation > Baseline factoring ERSI (d¢) 1 +12 -64 +38 311 5.5 x10-4 ¯ Precuneus CA1 Subfield Difference of Sedation > Baseline factoring ERSI (d¢) 1 -18 -58 +24 628 7.6 x10-5 ¯ Precuneus 2 -40 -20 +26 300 1.2 x10-2 ­ Left Postcentral Gyrus 3 -10 -16 +20 258 1.8 x10-2 ­ Left Thalamus Table 2 – CA1, Subiculum & ‘Stratum’ Hippocampus ROIs, resPng state funcPonal connecPvity clusters, difference of SedaPon > Baseline; Whole Hippocampus, ‘Stratum’, CA4/DG & CA1 Subfield Hippocampus ROIs, resPng state funcPonal connecPvity clusters, difference of SedaPon > Baseline with Explicit Recall SensiPvity Index d¢ beta-coefficients [Voxel threshold p-uncorrected < 0.005 in CA4/DG and < 0.01 in CA1] 15 Discussion There are two main findings from this analysis. Firstly, that propofol sedaPon-induced amnesia is predominantly associated with reduced funcPonal connecPvity of the stratum radiatum/lacunosum/moleculare, CA1 field stratum pyramidalis & CA4 stratum pyramidalis/Dentate Gyrus subfields. Secondly, that propofol sedaPon-induced amnesia is associated with a reduced funcPonal connecPvity between the hippocampus and the precuneus. Within the hippocampus, two major neuronal pathways have long been recognised. These are known as the direct pathway and the trisynapPc pathway.9 They both originate in the entorhinal cortex and converge on a common ouxlow tract to the subiculum from the CA1 pyramidal subfield. Whilst the direct pathway neurons synapse directly in the CA1 stratum pyramidalis field, the trisynapPc pathway first synapses in the dentate gyrus and incorporates the Mossy fibers connecPng the dentate gyrus to the CA3 region and the Schaffer collaterals connecPng the CA3 and CA1 regions (Fig 1).11 The exact physiological role of each of these pathways and how inter-dependent they are remain to be elucidated. Whilst propofol may potenPally be acPng on one or both of the direct or the trisynapPc pathways, given the relaPve lack of FC changes observed in the CA2/CA3 regions, the changes in funcPonal connecPvity that we demonstrate only provide evidence for the former. This acPon may either be on the main circuit neurons themselves or through effects on the interneurons, which are thought to make up around 10% of the cells in the CA1 region.11 The observed paeern of prominent funcPonal connecPvity changes in the CA1 and CA4/Dentate Gyrus regions is not unexpected as this would be consistent with the known distribuPon of Gamma-Aminobutyric Acid Type A (GABAA) receptors (and N-Methyl-D- Aspartate (NMDA) receptors), which are expressed at significantly greater densiPes within the CA1 and dentate gyrus subregions relaPve to other hippocampal areas.11 GABAergic interneurons are thought to be both a specific target of Propofol and criPcal for hippocampal memory funcPon.34-35 The CA1 region is a common nexus of both the direct and trisynapPc pathways and it is known from previous clinical invesPgaPons in humans with lesions localised to the CA1 field (which is thought to be parPcularly suscepPble to hypoxic, metabolic and excitotoxic insults), that damage to this area results in severe deficits in episodic explicit long-term memory.7 This would also provide further evidence for a specific focal acPon of propofol, rather than non-specific acPons in which case longer, polysynapPc pathways would logically be more vulnerable to disrupPon. The precuneus is increasingly considered to be a major centre involved in memory.36- 37 It is closely associated with the posterior cingulate cortex and the rest of the Default Mode Network (DMN), which is thought to have a role in autobiographical memory more generally.8,33 16 Reduced funcPonal connecPvity between the CA1 field stratum pyramidalis, stratum radiatum, stratum lacunosum, stratum moleculare, CA4 & dentate gyrus subregions and the precuneus appears to be a characterisPc feature of propofol-induced amnesia at sub- anaesthePc plasma concentraPons. Given that previous studies have associated propofol sedaPon with impairment of early memory consolidaPon specifically, these results support a role for the precuneus in normal memory consolidaPon funcPon.4-6 The absence of significant FC change between hippocampal subregions and auditory associaPon areas in this data may also be interpreted as a lack of evidence for disrupPon of iniPal engram encoding, further supporPng an acPon of propofol on the consolidaPon phase specifically as the basis on impaired recall. The precuneus could potenPally serve a computaPonal or integraPve associaPon role in memory consolidaPon or any downstream events directly. AlternaPvely, it may have a simpler role as a passive relay or distributor hub for long range connecPons between the hippocampi and memory associated regions such as the pre-frontal cortex, or with longer- term memory storage sites in the neocortex and source areas for the brain states being encoded and consolidated in memory. By implicaPon therefore, propofol may either exert its amnesic acPon by disrupPng neuronal signalling between these areas or by acPng at one or more of these sites directly. This would also be consistent with previous research suggesPng EEG changes observed to correlate with the amnesic effects of propofol such as in the P2-N2 event related potenPal and θ-synchrony may reflect underlying memory-related hippocampal-corPcal feedback loops.5 As suggested by previous studies, this amnesic acPon may be independent of its sedaPve acPon.1-2,4-5,18 The hippocampus has been previously shown to have a large degree of structural and funcPonal connecPvity with thalamic nuclei associated with the extended hippocampal memory system, especially the anterior nuclear group and the mediodorsal nucleus.38 If indeed specific to memory changes, the observed increase in FC between the hippocampus and this region during sedaPon could reflect changes in upstream or downstream memory regulaPon or feedback due to subcircuit isolaPon. The connecPvity changes observed when weighPng for mCp may represent more general resPng-state network changes. Taken together this could potenPally indicate a division of the DMN, with fracPonal de-coupling of the memory-specific components due to propofol. All experimental results should be interpreted with cauPon and these results are from a single dataset so should be reproduced independently with other subjects and other sedaPve agents before firm conclusions can be drawn. As with almost all fMRI data, this dataset featured some artefacts such as warping related to single direcPon phase-encoding and aliasing artefact (wrapping) in some cases. Even arer correcPng for these, this analysis is pushing the limits of 3T BOLD fMRI. Unlike the neocortex, the hippocampus is mostly only trilaminar. It is also one of the most highly convoluted areas of the brain. With an isotropic resoluPon of 3mm in the raw fMRI images and a 0.75mm gap between slices in the z-axis, some parPal volume effects and over-smoothing are inevitable for very fine anatomical structures such as hippocampal subfields. In parPcular, the CA1 field stratum pyramidalis, 17 stratum radiatum, stratum lacunosum, stratum moleculare, CA4 & dentate gyrus subregions are all anatomically co-located in 3D Euclidean space (together with other histologically separate layers such as the stratum oriens) and typically have common or closely related vascular supplies.9,11-13,39 The differenPal underlying GABA receptor density in each ROI may also exert an influence. Furthermore, whether these findings are truly specific effects on memory circuits or predominantly a consequence of the high baseline resPng-state FC between the hippocampi and the precuneus (which also has a high resPng regional cerebral blood flow) remains unclear.40 Given the increase in FC observed with the medial prefrontal cortex during sedaPon when measured plasma propofol levels were included in the GLM, effects on the DMN more generally cannot be excluded. Whilst explicit recall tesPng was only performed arer volunteers had fully recovered clinically (a minimum of 30 minutes arer the propofol was stopped), as this was tested during the same session persisPng subclinical drug effects of residual circulaPng propofol are possible, although previous research suggests this is unlikely to be significant.5 The mean mCp during infusion was notably 42% lower than predicted plasma propofol concentraPon (Cp) modelled at the Pme by the pharmacokinePc pump. The pharmacokinePc model appears to have performed less well in terms of both accuracy and precision than generally reported.22 This may be related to the target being lower than the dose range typically targeted for surgical planes of anaesthesia, as used in most pharmacokinePc studies, being at the lower extreme of the model envelope and amplifying the effects of fixed errors. Whilst the change in ERSI was staPsPcally significant when compared against both CpT and mCp, the change in Response Time was significant between CpT targets but only weakly correlated with mCp and vice versa for Time Out rate, an observaPon that also reflects the pharmacokinePc model performance in this study. Some of the results observed, although physiologically plausible, fell short of staPsPcal significance with the analyPcal methods used, however it should be noted that as the hippocampal subfield ROIs are different from those ROIs originally anPcipated during experimental design, this analysis may be underpowered. A task/event-related fMRI design alternaPve to resPng-state fMRI would have allowed separaPon of remembered and forgoeen words rather than an overall memory performance index, however the focus on drug-induced changes in FC in this case and their relaPonship to the extended process of explicit memory formaPon makes resPng-state an advantageous approach. Overall, this study provides a novel insight into the relaPve subregional acPons of propofol on the human hippocampus in vivo at clinically relevant plasma concentraPons sufficient to induce anterograde amnesia but below those required to completely inhibit response to auditory sPmuli and decision-making performance. It also provides evidence for a more prominent role of the precuneus in explicit memory consolidaPon than previous theories suggest. 18 Declara7ons of Interest The authors declare that they have no conflict of interest. Funding This work was supported by The Wellcome Trust (grant no. 083660/Z/07/Z) [R.A.]; Raymond & Beverly Sackler FoundaPon Studentship; Cambridge Commonwealth Trust [R.A.]; The BriPsh Oxygen Professorship, The Royal College of AnaesthePsts [D.K.M.]; The Evelyn Trust [D.K.M.]; The Canadian InsPtute for Advanced Research (CIFAR grant RCZB/072 RG93193) [D.K.M. & E.A.S.]; Cambridge NaPonal InsPtute for Health and Care Research Biomedical Research Centre [D.K.M.] and The Stephen Erskine Fellowship, Queens’ College, University of Cambridge [E.A.S.] 19 References 1. Veselis RA, Reinsel RA, Feshchenko VA, Wroński M. The comparaPve amnesic effects of Midazolam, Propofol, Thiopental, and Fentanyl at equisedaPve concentraPons. Anesthesiology 1997; 87:749-764. 2. Veselis RA, Reinsel RA, Feshchenko VA. Drug-induced amnesia is a separate phenomenon from sedaPon. Anesthesiology 2001; 95:896-907. 3. Tashbigou SR, Vogels MF, Absalom AR. Accidental awareness during general anaesthesia – a narraPve review. Anaesthesia 2017; 73:112-122. DOI:10.1111/anae.14124 4. Veselis RA, Reinsel RA, Feshchenko VA, Johnson R (Jr). InformaPon loss over Pme defines the memory defect of Propofol: A comparaPve response with Thiopental and Dexmedetomidine. Anesthesiology 2004; 101:831-41. 5. Pryor KO, Reinsel RA, Mehta M, Li Y, Wixted JT, Veselis RA. Visual P2-N2 Complex and Arousal at the Time of Encoding Predict the Time Domain CharacterisPcs of Amnesia for MulPple Intravenous AnesthePc Drugs in Humans. Anesthesiology 2010; 113(2):313-326. 6. Veselis RA, Pryor KO, Reinsel RA, Li Y, Mehta M, Johnson R (Jr). Propofol and Midazolam Inhibit Conscious Memory Processes Very Soon arer Encoding: An Event- related PotenPal Study of Familiarity and RecollecPon in Volunteers. Anesthesiology 2009; 110:295-312. 7. Bartsch T, Butler C. Transient amnesic syndromes. Nat Rev Neurol 2013; 9:86-97. DOI: 10.1038/nrneurol.2012.264 8. Gropper MA, Miller RD, Cohen NH et al., eds. Miller’s Anesthesia, 9th ed (InternaPonal EdiPon). Philadelphia, PA: Elsevier, 2020. 9. Kandel ER, Koester JD, Mack SH, Siegelbaum SA, eds. Principles of Neural Science, 6th ed. New York, NY: McGraw-Hill, 2021. 10. Merlo SA, Belluscio MA, Pedreira ME, Merlo E. Memory persistence: from fundamental mechanisms to translaPonal opportuniPes. Transl Psychiatry 2024; 14:98. DOI: 10.1038/s41398-024-02808-z 11. Standring S, Anand N, Catani M et al., eds. Gray’s Anatomy, 42nd ed. London: Churchill Livingstone, 2021. 12. Destrieux C, Bourry D, Velut S. Anatomie chirurgicale de l’hippocampe. Neurochir 2013; 59:149-158. DOI: 10.1016/j.neuchi.2013.08.003. 13. Tatu L, Vuillier F. Structure and VascularizaPon of the Human Hippocampus. In: Szabo K, Hennerici MG, eds. FronMers of Neurology and Neuroscience, Vol 34: The Hippocampus in Clinical Neuroscience. Basel: Karger, 2014; 18-25. 14. Gardner RJ, Hermansen E, Pachitariu M et al. Toroidal topology of populaPon acPvity in grid cells. Nature 2022; 602:123-128. DOI: 10.1038/s41586-021-04268-7. 20 15. Nakashiba T, Young JZ, McHugh TJ, Buhl DL, Tonegawa S. Transgenic InhibiPon of SynapPc Transmission Reveals Role of CA3 Output in Hippocampal Learning. Science 2008; 319:1260-1264. DOI: 10.1126/science.1151120. 16. LogothePs NK. What we can do and what we cannot do with fMRI. Nature 2008; 453:869-878. DOI: 10.1038/nature06976 17. Stamatakis EA, Orfanidou E, Papanicolaou AC. FuncPonal MagnePc Resonance Imaging. In: Papanicolaou AC, ed. Handbook of FuncMonal Brain Imaging in Neuropsychology and CogniMve Neurosciences. Oxford: Oxford University Press, 2017; 43-59. 18. Pryor KO, Root JC, Mehta M et al. Effect of Propofol on the medial temporal lobe emoPonal memory system: a funcPonal magnePc resonance imaging study in human subjects. Br J Anaesth 2015; i104-i113. DOI: 10.1093/bja/aev038 19. Adapa RM, Davis MH, Stamatakis EA, Absalom AR, Menon DK. Neural Correlates of Successful SemanPc Processing During Propofol SedaPon. Hum Brain Mapp 2014; 35:2935-2949. DOI: 10.1002/hbm.22375 20. Stamatakis EA, Adapa RM, Absalom AR, Menon DK. Changes in ResPng Neural ConnecPvity during Propofol SedaPon. PLoS One 2010; 5(12):e14224. DOI: 10.1371/journal.pone.0014224 21. Adapa RM, Axell RG, Mangat JS, Carpenter TA, Absalom AR. Safety and performance of TCI pumps in a magnePc resonance imaging environment. Anaesthesia 2012; 67:33-39. DOI: 10.1111/j.1365-2044.2011.06917.x 22. Marsh B, White M, Morton N, Kenny GNC. PharmacokinePc model driven infusion of propofol in children. Br J Anaesth 1991; 67:41-48. 23. Varvel JR, Donoho DL, Shafer SL. Measuring the PredicPve Performance of Computer- Controlled Infusion Pumps. J Pharmacokinet Biopharm 1992; 20(1):63-94. DOI: 10.1007/BF01143186 24. Nieto-Castanon A, Whixield-Gabrieli S. CONN funcPonal connecPvity toolbox: RRID SCR_009550, release 22, 2022. DOI: 10.56441/hilbertpress.2246.5840. 25. Whixield-Gabrieli S, Nieto-Castanon A. Conn: A FuncPonal connecPvity toolbox for correlated and anPcorrelated brain networks. Brain Connect 2012; 2(3):125-141. DOI: 10.1089/brain.2012.0073 26. Chakravarty MM, Steadman P, van Eede MC et al. Performing label-fusion-based segmentaPon using mulPple automaPcally generated templates. Hum Brain Mapp 2013; 34:2635-2654. DOI: 10.1002/hbm.22092 27. Piptone J, Park MTM, Winterburn J et al. MulP-atlas segmentaPon of the whole hippocampus and subfields using mulPple automaPcally generated templates. Neuroimage 2014; 101:494-512. DOI: 10.1016/j.neuroimage.2014.04.054 28. Winterburn JL, Pruessner JC, Chavez S et al. A novel in vivo atlas of human hippocampal subfields using high-resoluPon 3T magnePc resonance imaging. Neuroimage 2015; 71:254-265. DOI: 10.1016/j.neuroimage.2013.02.003 21 29. Worsley KJ, Marree S, Neelin P, Vandal AC, Friston KJ, Evans AC. A unified staPsPcal approach for determining significant signals in images of cerebral acPvaPon. Hum Brain Mapp 1996; 4(1):58-73. 30. Chumbley J, Worsley K, Flandin G, Friston K. Topological FDR for neuroimaging. Neuroimage 2010; 49(4):3057-3064. 31. Adapa RM. Exploring neural correlates of higher cogniMon with sedaMon. Thesis, University of Cambridge, 2011. 32. Cole MW, Bassee DS, Power JD, Braver, Petersen SE. Intrinsic and Task-Evoked Network Architectures of the Human Brain. Neuron 2014; 83:238-251. DOI: 10.1016/j.neuron.2014.05.014 33. Vatansever D, Manktelow AE, Sahakian BJ, Menon DK, Stamatakis EA. CogniPve Flexibility: A Default Network and Basal Ganglia ConnecPvity PerspecPve. Brain Connect 2016; 6(3):201-207. DOI: 10.1089/brain.2015.0388 34. Craig MM, Misic B, Pappas I, Adapa RM, Menon DK, Stamatakis EA. Propofol sedaPon-induced alteraPons in brain connecPvity reflect parvalbumin interneuron distribuPon in human cerebral cortex. Br J Anaesth 2021; 126(4):835-844. DOI 10.1016/j.bja.2020.11.035 35. Sakimoto Y, Oo PMT, Goshima M, Kanehisa I, Tsukada Y, Mitsushima D. Significance of GABAA Receptor for CogniPve FuncPon and Hippocampal Pathology. Int J Mol Sci 2021; 22:12456. DOI: 10.3390/ijms222212456 36. Flanagin VL, Klinkowski S, Brodt S et al. The precuneus as a central node in declaraPve memory retrieval. Cereb Cortex 2023; 33:5981-5990. DOI: 10.1093/cercor/bhac476. 37. Lyu D, SPeger JR, Xin C et al. Causal evidence for the processing of bodily self in the anterior precuneus. Neuron 2023; 111:2502-2512. DOI: 10.1016/j.neuron.2023.05.013. 38. Sweeney-Reed CM, Buentjen L, Voges J et al. The role of the anterior nuclei of the thalamus in human memory processing. Neurosci Biobehav Rev 2021; 126:146-158. DOI: 10.1016/j.neubiorev.2021.02.046 39. Marinković S, Milisavljević M, Puškaš L. Microvascular Anatomy of the Hippocampal FormaPon. Surg Neurol 1992; 37(5):339-349. DOI: 10.1016/0090-3019(92)90001-4. 40. Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GJ. A default mode of brain funcPon. Proc Natl Acad Sci USA 2001; 98(2):676-682. DOI: 10.1073/pnas.98.2.676.