In Vivo PET Imaging of Preynaptic Density Reveals Stress-Associated Synaptic Deficits Related to Behavioral and Molecular Alterations in Rats

Abstract

Background.

Preclinical research indicates chronic stress can induce synaptic loss in corticolimbic brain regions regulating mood and cognition. Presynaptic density can now be measured in vivo using radioligands targeting synaptic vesicle protein 2A and positron emission tomography (PET). We conducted a first in vivo PET study to investigate chronic stress-induced synaptic density changes in rats and examined correlates with behavior and protein expression.

Methods.

Male and female Long-Evans rats were exposed to chronic unpredictable stress (CUS; n=24/sex) and compared with controls (n=12/sex). Sucrose preference and novel object recognition (NOR) were used to assess stress-related behavioral phenotypes. PET with [¹⁸F]SynVesT-1 was used to measure synaptic density in a subset of rats (n=8-9/group/sex). Prefrontal cortex (PFC) and hippocampal proteins were quantified via LC-MS/MS (n=5/group/sex), followed by pathway analysis and linear regression to examine molecular profiles associated with CUS and correlated with synaptic density as measured by PET.

Results.

Synaptic density was lower in the PFC of CUS rats relative to controls (d=0.94, p=0.012) and correlated with sucrose preference (r=0.35, p=0.042). Synaptic density was also lower in hippocampus (d=0.55, p=0.017), which correlated with NOR (r=0.35 p=0.045). Differentially expressed proteins were enriched for transcriptional regulation and metabolic pathways. Proteins implicated in synaptogenesis and neurodegeneration were positively and negatively correlated, respectively, with synaptic density.

Conclusions.

We demonstrate that [¹⁸F]SynVesT-1 PET can be used for in vivo quantification of synaptic density in a rodent model of chronic stress. This method can, therefore, facilitate translational research investigating synaptic mechanisms in stress-related pathology and treatment response.

INTRODUCTION

Stress has long been recognized as a significant risk factor for developing psychiatric illnesses such as depression and anxiety disorders(1, 2). Animal model-based research has provided consistent evidence that persistently high stress levels results in dendritic arbor atrophy and synaptic loss in corticolimbic brain regions that act as key mediators of emotional regulation and cognition, such as prefrontal cortex (PFC) and hippocampus(3, 4). It was presumed that these synaptic and neuroplastic changes contribute to mood and cognitive symptoms of stress-related psychiatric disorders, a hypothesis supported by converging evidence from clinical neuroimaging, human postmortem, and preclinical studies(3-7). For example, neuroimaging research provides strong evidence of lower PFC and hippocampal volumes among individuals with mood, anxiety, and trauma disorders relative to psychiatrically healthy controls, with smaller volume being inversely correlated with length of illness and severity of depressive symptoms(8, 9). Postmortem analyses conducted in similar clinical populations consistently report lower synapse number(10) and deficits in expression of pre- and postsynaptic factors, synaptic signaling, and neuroplastic molecules within the same corticolimbic regions(10-13). Parallel research in preclinical models further implicates that chronic stress induces structural and functional synaptic deficits, which contribute to behavioral and cognitive changes. Specifically, repeated or chronic stress exposure in rodents results in anhedonia-like behavior(14-17), as well as deficits in PFC-dependent working memory(18) and hippocampal-dependent spatial memory(19, 20), with corresponding dendritic arbor atrophy, spine density, and synaptic protein expression in PFC and hippocampus(14, 16-18, 21).

While evidence collectively supports the hypothesis that synaptic loss and dysfunction within corticolimbic regions plays a central role in the etiology of stress-related psychiatric disorders, direct assessment of synaptic density in living humans has only recently become feasible. Historically, measurements of synaptic density were done postmortem; however, recent advances allow us to measure synaptic density in vivo using positron emission tomography (PET) with radioligands targeting the synaptic vesicle protein 2A (SV2A) such as [¹¹C]UCB-J, [¹⁸F]SynVesT-1, and [¹⁸F]SynVesT-2(22-24). With SV2A being homogeneously expressed in presynaptic vesicles across mammalian species and neuronal subtypes(25-28), measurement of SV2A expression can be used as a proxy for synaptic density, similar to other presynaptic proteins, such as synaptophysin, commonly used in postmortem work(23, 29, 30). By extension, PET quantification of SV2A radioligands can be used as an in vivo measure of synaptic density(22, 23, 31). Using this approach, our group published the first in vivo evidence of lower synaptic density in association with mood and cognitive symptoms in individuals with stress-related psychiatric disorders(32). Specifically, the findings of this study suggested that lower corticolimbic synaptic density contribute to symptom severity.

Preclinical investigations utilizing chronic stress paradigms, such as chronic unpredictable stress (CUS), continue to provide invaluable insight regarding physiological, cellular, and molecular processes subserving behavioral and cognitive changes associated with stress(33-35). Specifically, studies utilizing the CUS model consistently report reductions in pre- and post-synaptic protein expression (such as synapsin I and PSD95, respectively(14)), loss of dendritic spines and neurite atrophy(14, 36), and blunted neuronal proliferation(37) and synaptic plasticity(14, 38), with findings being particularly robust in medial prefrontal cortex and hippocampus(39). With the advent of SV2A PET, the same in vivo synaptic measure can be applied to these preclinical models that is being used in clinical populations. Importantly, while many studies have investigated synaptic density in fixed tissue, in vivo methods facilitate new and invaluable avenues of research, such as longitudinal evaluation of synaptic remodeling in during stress or in response to putative treatments. Further, in contrast to other in vivo synaptic imaging methods currently available (e.g., two-photon microscopy)(40-42), SV2A PET is scalable, allowing for multiple animals and brain regions (the entire brain) to be assessed in a single imaging session. For example, Corvo et al.(43) used [¹¹C]UCB-J PET to show synaptic deficits in mice following chronic corticosterone exposure. However, this study was conducted in exclusively males.

The overarching goal of the current study was to test whether PET could be used to detect stress-related presynaptic deficits in vivo and in both male and female animals. Secondarily, we aimed to test the hypothesis that anhedonia-like behavior and deficits in hippocampal-dependent memory would be associated with loss of presynaptic density as measured by PET. Exploratory analyses were conducted to examine potential sex-differences in CUS-related presynaptic decline and behaviors. We additionally performed postmortem LC-MS/MS-based proteomics to examine molecular profiles associated with chronic stress and identify proteins correlating with presynaptic density as measured by PET.

METHODS AND MATERIALS

Animals and Experimental Design

Female (n=36) and male (n=36), eight-week-old Long Evans rats (Charles River Laboratories, Wilmington, MA) were pair-housed in a climate-controlled room and maintained on a 07:00-19:00 light/dark cycle with access to water ad libitum. Rats were acclimated to the facility for one week with daily handling prior to group randomization and the start of experimental procedures as summarized in Figure 1. All experimental procedures were performed as approved by the Yale IACUC and according to NIH and institutional guidelines and the Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Figure 1. Experimental timeline.

[A] An overview of the study design, experimental groups, and subsets of rats used for behavioral tests [sucrose preference test (SPT) and novel object recognition (NOR)], positron emission tomography (PET) imaging, and postmortem protein quantification [label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS)] and proteomic analysis. [B] Individual stressors were delivered two per day during for 24 consecutive days to rats in the CUS group.

Chronic Unpredictable Stress

Rats assigned to the CUS group were exposed to a variety of mild-moderate stressors (Figure 1B) for 24 consecutive days, with two individual stressors presented daily in a pseudorandom order. Rats in the home-cage control (CON) condition were left undisturbed except for standard husbandry and weekly body weight measurements also performed in CUS rats.

Behavioral Testing

The Sucrose Preference Test (SPT)(14, 44) and the Novel Object Recognition (NOR) test of hippocampal-dependent learning and memory(45-47) were performed as previously described and as detailed in the Supplement. The Preference Index ($\text{SPI}$) was calculated as:

$$\text{SPI}=\frac{(1\%SucroseConsumption)-(WaterConsumption)}{(1\%SucroseConsumption)+(WaterConsumption)}$$

For NOR, the Discrimination Index ($\text{DI}$) was calculated as:

$$\text{DI}=\frac{(NovelInvestigationtime)-(FamiliarInvestigationtime)}{(NovelInvestigationtime)+(FamiliarInvestigationtime)}$$

Positron Emission Tomography

[¹⁸F]SynVesT-1 was synthesized on-site as previously described (48, 49). Rats from each group were selected at random and transported to the Yale PET center. Imaging procedures are detailed in the Supplement. In brief, rats were and positioned in the PET camera. [¹⁸F]SynVesT-1 (mass: 0.15±0.056 nmol; dose: 0.45±0.18 mCi) was delivered as a bolus intravenous injection, followed by a 60-minute dynamic acquisition.

Reconstructed images were co-registered to the Sigma rat magnetic resonance template(50), which used to define the a priori regions of interest (ROIs): prefrontal cortex (PFC; medial and orbitofrontal) and hippocampus (dorsal, ventral, and subiculum). The primary PET outcome measure was non-displaceable binding potential (BP_ND) as estimated by the simplified reference tissue model with fixed k₂’(SRTM2)(51) with cerebellum as the reference region (Figure S1).

Label-free Quantitative LC-MS/MS Proteomics

Rats were euthanized immediately after the PET scan. Microdissections were performed to isolate PFC and hippocampus and tissue was flash-frozen and stored at −80°C until further processing. Samples from each group were selected at random for LC-MS/MS label-free protein identification and quantification (details in the Supplement).

Statistical Analysis

Percent change of body weight from baseline was assessed by a repeated measure analysis of variance (ANOVA) with time as the repeated measure and fixed factors of group (CON, CUS) and sex (male, female). Two-(Group: CON, CUS) by-two (Sex: male, female) factorial ANOVAs were used to test for group differences in behavioral outcome measures and [¹⁸F]SynVesT-1 BP_ND in a priori ROIs (PFC, hippocampus). Post hoc tests comparing CON and CUS within each sex were performed with Bonferroni’s correction for multiple comparisons. Pearsons’s correlations were used to test for relationships between BP_ND and behavioral measures. All tests were two-tailed, with alpha at p<0.05. Statistical analyses were performed in SPSS 29.0.1.0 (IBM Corp., Armonk, NY) and visualized using Prism 10.2.2 (GraphPad Software, LLC., San Diego, CA).

For proteomic data analysis, independent sample t-test were used to determine differentially abundant proteins (DAPs) between CON and CUS samples. Given the exploratory nature of these analyses, a liberal significance threshold of p<0.05(unadjusted) and log₂(fold-change) > ±1.5 was selected. Visualization of hierarchical clustering heatmaps was conducted using Qlucore Omics Explorer v 3.7 (Qlucore AB, Lund, Sweden); Venn diagrams were generated using InteractiVenn (https://www.interactivenn.net/)(52). Functional analyses of DAPs were performed using Ingenuity Pathway Analysis software (QIAGEN Redwood City, CA). Exploratory linear regressions examining relationships between DAPs and BP_ND were conducted using Qlucore Omics Explorer with a significance threshold of p<0.01(unadjusted).

RESULTS

Behavioral Correlates of Chronic Stress Exposure

Weight Gain

A repeated measures ANOVA testing for Group and Sex differences in body weight (Figure 2) indicated a significant effect of time (F_5,64=866.031, p<0.001), with weight generally increasing across all groups. There were additionally significant main effects of sex (F_1,68=22.757, p<0.001), group (F_1,68=28.522, p<0.001), and two-way (group*sex: F_1,68=10.230, p=0.002; group*time: F_5,64=16.170, p<0.001; sex*time: F_5,64=35.033, p<0.001) and three-way (group*sex*time: F_5,64=6.371, p<0.001) interactions. Post hoc tests indicated CUS males gained less weight relative to CON-males starting the first week of CUS (p_adj.<0.001), a trajectory that continued through Day 24 (all p_adj.<0.001). In contrast, CUS-females gained less weight the first week of CUS (p_adj.<0.001) but were no different from CON-females at Day 14 through Day 24 (p_adj.=0.434-0.962).

Figure 2. Attenuated weight gain is more pronounced in CUS males.

Body weight over the course of the 24-day CUS paradigm, presented as %change from baseline (CUS Day 1). Data are graphed as the Mean±SD. Post hoc tests (CON vs. CUS): ***p_adj.<0.001, ****p_adj.<0.0001.

Blunted Sucrose Preference Following CUS

After Day 24 of CUS, we observed a significant main effect of group (F_1,69=17.588, p<0.001), such that CUS rats, overall, displayed lower $\text{SPI}$ relative to CON (Figure 3A). There was no effect of sex (F_1,69=0.314, p=0.577), or group*sex interaction (F_1,69=1.024, p=0.315).

Figure 3. Behavioral phenotypes associated with CUS.

[A] The Sucrose Preference Index, as measured during the SPT. [B] Total distance traveled, and [C] time spent in the center zone of the open arena during the Habituation trial of NOR. [D] Time spent investigation the two identical objects during the Familiarization trial of NOR, and [E] the Discrimination Index from the Test trial of NOR, calculated as a measure of object recognition memory. Individual values are plotted with Mean±SD. Results of significant post hoc tests (Bonferroni-corrected p-values) between CON and CUS groups within each sex are provided.

Novelty-induced Hyperlocomotion and Limited Exploration

When introduced to the empty arena for the first time during the Habituation trial of NOR, CUS rats traveled greater distances relative to CON (F_1,67= 9.514, p=0.003) and females traveled greater distances as compared with males (F_1,67= 7.522, p=0.008; Figure 3B). Additionally, CUS rats spent less time in the center zone as compared with CON (F_1,67= 6.388, p=0.014), and females spent less time in the center zone relative to males (F_1,67= 4.672, p=0.034; Figure 3C). The sex*group interactions were not significant for distance traveled (F_1,67= 0.942 p=0.335) or for center zone time (F_1,67= 0.091 p=0.764).

Deficits in Hippocampal-Dependent Learning and Memory

During the Familiarization trial of the NOR test, there were no differences between CON and CUS rats (F_1,66=,0.681 p=0.412) regarding the total amount of time spent investigating the two identical objects (Figure 3D); however, there was a significant main effect of sex (F_1,66=6.374, p=0.014) with females generally spending more time with the objects relative to males. Twenty-four hours later, during the Test trial, CON-males (p=0.019, p_adj.=0.077) and females (p<0.001, p_adj.=0.003), spent more time with the novel object than the familiar object (Figure S3, left). Conversely, no such difference in investigation times between the two objects was observed among CUS rats (Figure S3, right). Further, CON rats had better object recognition memory (i.e., $\text{DI}$) relative to CUS rats (main effect of group: F_1,67=19.952, p<0.001). This effect of CUS on object recognition memory was observed in male (p_adj.=0.005) and female (p_adj.=0.001) rats. There was no effect of sex (F_1,67=1.251, p=0.267), or group*sex interaction (F_1,67=0.091, p=0.764) on object recognition memory.

In Vivo SV2A PET Assessment of Synaptic Density

Regional Measures of SV2A Density Following CUS are Sex-Specific

As compared with CON, CUS rats had significantly lower PFC BP_ND values (F_1,30=7.088, p=0.012), including 13.3% and 12.7% lower BP_ND in CUS-males and CUS-females relative to the same-sex CON group, respectively (Figure 4A). We additionally observed main effects of group (F_1,30=6.384, p=0.017), sex (F_1,30=26.140, p<0.001), and a group*sex interaction (F_1,30=4.668, p=0.039) on hippocampal BP_ND (Figure 4B). Post hoc tests revealed significantly lower hippocampal BP_ND values in CUS-females (−16.7%, p=0.002), but not CUS-males (−1.8%, p=0.798), relative to respective CON groups.

Figure 4. Regional In Vivo PET Measures of SV2A Density Following CUS and Associations with Behavioral Phenotypes.

PET quantification of SV2A (i.e., [¹⁸F]SynVest-1 BP_ND) in [A] prefrontal cortex (PFC) and [B] hippocampus (HIP). Individual values are plotted with Mean±SD. The %difference between CON and CUS within each sex are provided (*p_adj<0.05). Correlations between PFC SV2A density and [C] center exploration time during the NOR habituation trial and [D] sucrose preference. [E] The correlation between HIP SV2A density and the NOR discrimination index. [F] The relationship between sucrose preference and NOR discrimination index.

Associations between SV2A Density and Behavioral Measures

For rats with [¹⁸F]SynVesT-1 PET, we examined relationships between regional SV2A density and performance on behavioral tasks. We observed significant positive correlations between PFC BP_ND values and time spent in the center zone during the Habituation trial of NOR (r=0.355, p=0.043; Figure 4C) and sucrose preference (r=0.351, p=0.042; Figure 4D). There was no relationship between PFC BP_ND and the NOR $\text{DI}$ (Figure S4A). However, there was a significant positive correlation between the $\text{DI}$ and hippocampal BP_ND (r=0.346, p=0.045; Figure 4E). Conversely, hippocampal BP_ND was not related to time spent in the center zone or sucrose preference (Figure S4B, C). Also of note, across all rats (i.e., with and without PET), sucrose preference and the NOR $\text{DI}$ were positively correlated (r=0.246, p=0.039; Figure 4F).

Exploratory Proteomics

Sex Differences in the Molecular Response to CUS

In PFC, we identified 258 DAPs between CON and CUS. Specifically in females, we observed 435 DAPs. In males, 61 DAPs between CON and CUS were identified. (Figure 5A). In hippocampus, fewer DAPs were observed overall (N=54), between female groups (N=108) and between male groups (N=78) (Figure 5B). Notably, there was minimal overlap between individual DAPs identified in for females and males in PFC (n=7) and hippocampus (n=1). Full lists of DAPs are provided in the Supplement.

Figure 5. Sex differences in the molecular response to CUS.

Number and overlap of differentially abundant proteins (DAPs; p<0.05, log₂(fold-change) > ±1.5) between CUS and CON in female, male, and across all rats for [A] prefrontal cortex and [B] hippocampus. Results of the IPA pathway enrichment analysis of DAPs in [C] prefrontal cortex and [D] hippocampus for DAPs identified in female (top) and male (bottom) rats. Bars indicate statistical significance [−log(BH p-value)], with the dotted line at 1.3 representing significance threshold. Circles indicate the proportion of DAPs in the study dataset overlapping with the Ingenuity Knowledge Base canonical pathway (Ratio).

Ingenuity Pathway Analysis (IPA) was used to explore potential functional relevance of the observed patterns of DAPs.(53) Top enriched pathways in females and males are illustrated in Figures 5C and 5D, for PFC and hippocampus, respectively. In PFC, there was high representation of pathways involved in mitochondrial function and metabolic pathways. For example, in females, mitochondrial dysfunction [−log(BH p-value)=7.9, ratio=0.08] and integration of energy metabolism [−log(BH p-value)=5.5, ratio=0.1]; and in males, creatine-phosphate biosynthesis [−log(BH p-value)=1.6, ratio=0.3]. In females, synaptogenesis [−log(BH p-value)=5.3, ratio=0.07] and neurotransmitter/neuropeptide signaling were additionally represented in the top enriched pathways. In hippocampus, transcriptional regulation pathways were enriched, including processing of capped intron-containing pre-mRNA [−log(BH p-value)=31.6, ratio=0.1] and spliceosomal cycle [−log(BH p-value)=8.2, ratio=0.2] pathways in females, and ribosomal quality control signaling [−log(BH p-value)=1.5, ratio=0.02] in males. Hedgehog [−log(BH p-value)=1.7, ratio=0.03] and Notch4 [−log(BH p-value)=1.45, ratio=0.03] signaling pathways were also enriched in male hippocampus.

Regional Associations Between SV2A PET and Neuroprotein Abundance

Regardless of CUS exposure or sex, we identified 27 proteins with relative abundance correlating with PFC SV2A density at a significance threshold of p<0.01 (Figure 6A, Table 1), of which 12 were positively correlated and 15 were negatively correlated. The relative abundance of 70 proteins correlated with hippocampal SV2A density (Figure 6B, Table 2), 26 of which were positively correlated and 44 negatively correlated.

Figure 6. Relationships between SV2A PET and structural and functional synaptic protein abundance.

Heatmaps of all significant (p<0.01) positive (r>0) and negative (r<0) correlations between SV2A density as measured by PET ([¹⁸F]SynVesT-1 BP_ND) and normalized protein abundance (Z score) assessed postmortem in [A] PFC and [B] hippocampus. Representative correlations for individual proteins [C] positively and [D] negatively correlated with PFC SV2A density and [E] positively and [F] negatively correlated with hippocampal SV2A are displayed with Pearson’s correlation (r) and significance (uncorrected p) values.

Table 1.

Correlations of p<0.01 in PFC between protein abundance and [¹⁸F]SynVesT-1 BP_ND

Accession	Gene Symbol	p-value	r-statistic
Synaptic Function/Signal Transduction
D4A5J1	Kbtbd11	0.0068	0.5414
A0A8I6GF35	Kctd10	0.0081	−0.5303
A0A8I5ZR85	Vamp1	0.0030	0.5915
Neuronal Adhesion/Extracellular Matrix
A0A0G2JW27	Nrcam	0.0021	0.6122
Q68HB6	Pcdh7	0.0038	0.5785
Kinases/Phosphotases
F1LRI7	Aak1	0.0089	0.5240
Dhehydrogenases/Oxidoreductases
O08651	Phgdh	0.0005	−0.6762
A6I0J6	Rfk	0.0068	−0.5419
A0A8I5XWT5	Sirt3	0.0091	0.5222
Lipid Signaling & Membrane Dynamics
F7F8T0	Extl2	0.0006	0.6710
A6HK92	Apoh	0.0039	−0.5768
Lipid/Small Molecule Transport
A0A8I6G7U4	Crabp1	0.0055	−0.5559
P55051	Fabp7	0.0072	−0.5384
Actin & Microtubule-Associated
Q5PQP4	Cdc42ep2	0.0084	−0.5276
F7F0R0	Tmod3	0.0084	−0.5273
Transcriptional Regulator/RNA Processing
A0A0H2UI20	Ctbp1	0.0056	0.5538
A0A8I5ZLP6	Ddx42	0.0056	−0.5548
D3ZPW6	Lage3	0.0079	−0.5320
A0A8I6ASC1	Nhsl1	0.0079	−0.5322
A0A8I5Y176	Ranbp9	0.0085	0.5271
A0A8I5ZP60	Snw1	0.0068	−0.5416
Innate Immunity & Complement Proteins
Q8R5H8	Faim	0.0024	−0.6036
Q66HG9	Mavs	0.0019	0.6176
A6IQG6	Mx2	0.0032	−0.5883
Protease Inhibitors & Protein Homeostasis
Q7TMC3	Saa4	0.0067	−0.5426
Poorly Characterized
F1LQ27	Fam98c	0.0049	0.5621
G3V9L1	Tex2	0.0023	0.6056

Table 2.

Correlations of p<0.01 in hippocampus between protein abundance and [¹⁸F]SynVesT-1 BP_ND

Accession	Gene Symbol	p-value	r-statistic
Synaptic Function/Signal Transduction
O35431	Apba2	0.0059	0.5510
P07171	Calb1	0.0055	0.5552
F7FQB8	Kpna4	0.0035	0.5835
A0A8I5ZTG9	Nisch	0.0008	0.6607
Q6B345	S100a11	0.0081	−0.5301
A0A8L2QW99	Tprg1l	0.0094	0.5201
Q7TQ11	Vtn	0.0010	−0.6503
Neuronal Adhesion/Extracellular Matrix
B4F7F3	Arvcf	0.0028	0.5951
M0RC17	Chl1	0.0068	0.5419
P07722	Mag	0.0095	−0.5194
G3V8G6	Hapln2	0.0049	−0.5631
Kinases/Phosphotases
A0A8I6AUH0	Camk1d	0.0073	0.5372
G3V6J5	Cdk17	0.0079	0.5321
Q01986	Map2k1	0.0083	0.5281
O70150	Pnck	0.0067	0.5431
Q80Z30	Ppm1e	0.0022	0.6083
P63319	Prkcg	0.0038	0.5776
Dhehydrogenases/Oxidoreductases
G3V7I5	Aldh1b1	0.0039	0.5761
G3V7J0	Aldh6a1	0.0013	0.6365
Q6AXW1	Glrx2	0.0096	0.5182
A0A8I6AMY4	Sdr39u1	0.0017	0.6216
Q76EQ0	Srr	0.0027	0.5975
P07632	Sod1	0.0095	−0.5191
Mitochondrial/Metobolic Enzymes
A0A0G2JSP8	Ckm	0.0030	−0.5907
A6J7V3	Hdhd3	0.0009	0.6536
F1LZC5	Ndufa13-ps1	0.0057	−0.5538
Q03346	Pmpcb	0.0041	0.5731
A0A0U1RS25	Upf1	0.0071	0.5394
Lipid Signaling & Membrane Dynamics
A0A8I6A1M5	Acot2	0.0054	0.5569
D4A7H9	Chmp7	0.0095	0.5188
P49620	Dgkg	0.0028	0.5963
A0A8I6A444	Dgkz	0.0005	0.6811
A0A8I5ZKL2	Gmppa	0.0057	0.5537
P04182	Oat	0.0044	0.5695
Accession	Gene Symbol	p-value	r-statistic
Lipid/Small Molecule Transport
P24090	Ahsg	0.0063	−0.5470
A0A0G2JU23	Apoa1	0.0030	−0.5923
P04638	Apoa2	0.0027	−0.5970
P02651	Apoa4	0.0023	−0.6052
A0A0H2UI39	Apoc3	0.0031	−0.5897
A0A8I5ZW66	Gc	0.0026	−0.5995
A0A8L2QBL5	Ttr	0.0058	−0.5520
Actin & Microtubule-Associated
Q99PD4	Arpc1a	0.0073	0.5368
A0A8L2Q7T7	Cd2ap	0.0038	0.5785
A0A8I6GEA5	Grip1	0.0038	0.5777
A0A8I5ZQW6	Ksr1	0.0043	0.5712
P68370	Tuba1a	0.0037	0.5800
P83888	Tubg1	0.0049	0.5628
Q5RKI0	Wdr1	0.0043	0.5707
Transcriptional Regulator/RNA Processing
A0A8I6G7K6	Arrb1	0.0079	0.5315
Q9Z1T4	Cnksr2	0.0075	0.5352
F1MAA2	Cops7a	0.0065	0.5448
Q6AYD3	Pa2g4	0.0003	−0.6950
A0A8I6A3U2	Snrpc-ps3	0.0087	−0.5256
G3V6Q3	Wdr13	0.0033	0.5867
Innate Immunity & Complement Proteins
F7F389	C9	0.0082	−0.5295
G3V615	Cfb	0.0004	−0.6918
A6JY48	Cx3cl1	0.0086	0.5262
Q66HG9	Mavs	0.0057	−0.5529
Coagulation & Fibrinolysis Factors
G3V843	F2	0.0042	−0.5724
P14480	Fgb	0.0033	−0.5854
A0A8I5ZVK5	Plg	0.0087	−0.5252
Protease Inhibitors & Protein Homeostasis
D4ABB4	Fbxl15	0.0098	0.5166
A0A8I6AJP9	Ist1	0.0061	0.5484
A0A0G2JZ73	Serpina1	0.0041	−0.5735
A0A0G2JYK0	Serpina3l	0.0031	−0.5891
P09006	Serpina3n	0.0079	−0.5316
Q5M7T5	Serpinc1	0.0096	−0.5184
A0A8I6ANY2	Ube2b	0.0062	0.5477
D4AA63	Ubqln2	0.0089	0.5234
Poorly Characterized
A0A8I6APK7	Fam114a2	0.0065	0.5447

DISCUSSION

To our knowledge, this is the first study to perform in vivo [¹⁸F]SynVesT-1 PET in the CUS model. We show lower PET-based quantification of SV2A in rats following exposure to CUS, reflecting lower synaptic density. Specifically, we found lower SV2A in PFC and hippocampus, brain regions that act as key mediators of emotional regulation and cognition and are highly implicated in the pathology of stress-related psychiatric disorders(54). This result is consistent with prior studies demonstrating stress-induced synaptic decline in the same corticolimbic regions that have primarily relied on postmortem quantification of postsynaptic elements (14, 16, 18, 21). Additionally, Corvo et al. recently reported similar results in an alternate chronic stress model and SV2A PET with the radiotracer [¹¹C]UCB-J, specifically demonstrating synaptic deficits in male mice following 10-weeks of corticosterone exposure that was mitigated by 6-weeks concomitant fluoxetine treatment(43). Together with our findings, these studies provide proof-of-concept evidence that in vivo PET is effective for detection of chronic stress-induced synaptic alterations.

In addition to synaptic alterations, we replicated a number of the behavioral responses to chronic stress previously reported in the CUS model. First, we observed attenuated weight gain in CUS rats that was much more pronounce in males. Attenuated weight gain in this context likely reflects a combination of behavioral and physiological responses to stress. Work by Kuti, et al.(55) provides evidence that chronic stress is associated with hypophagia (suggesting potential changes in appetite or motivation to feed) as well as metabolic shifts, including increased energy expenditure and a shift in fuel utilization to preferentially burn fat over carbohydrates. Unfortunately, this study like many others, was exclusively in male mice. The limited number of studies that have included females have yielded more variable effects in females and potentially a more complex relationship between stress and neuroendocrine factors contributing to bodyweight(56-58) that warrants further investigation. Second, male and female CUS rats displayed a blunted sucrose preference. Historically, lack of sucrose preference has been interpreted as an anhedonia-like phenotype in rodents induced by chronic stress that correlates with synaptic loss and dendritic arbor atrophy in PFC, with both behavioral and synaptic changes being reversible with antidepressant agents such as ketamine(3, 4, 14-17). It is therefore notable that we observed a specific positive correlation between sucrose preference and PFC SV2A density. Additionally, CUS rats traveled greater distances than CON rats during the habituation phase of the NOR, potentially indicative of novelty-induced hyperlocomotion, which, when combined with diminished exploration of the center zone, is consistent with an anxiety-like phenotype. Similar to sucrose preference, the total time a rat was willing to venture to the center of the open arena was positively correlated with PFC SV2A density. Furthermore, females, regardless of stress, displayed enhanced locomotion and center avoidance relative to males. This observation could reflect sex differences in the utilization of coping strategies, an interpretation consistent with reports of female rodents being more likely to exhibit an active phenotype relative to males; for example, darting rather than freezing in classical fear conditioning(59, 60), and being more successful in active avoidance tasks(61, 62). Lastly, novel object recognition was impaired in CUS rats, with worse NOR correlating with lower hippocampal SV2A density, in line with previous reports of stress-induced hippocampal synaptic dysfunction subserving hippocampal-dependent cognitive impairment(63-65). Importantly, there was no relationship between novel object recognition and PFC SV2A, nor relationships between hippocampus SV2A and sucrose preference, yielding confidence in the regional specificity of the behavioral tasks. Together, these findings add to the growing evidence supporting of the theory that chronic stress can result in synaptic loss, which in turn may contribute to mood and cognitive symptoms in stress-related psychiatric disorders.

When exploring the impact of CUS on neuroproteome, lack of overlap between female and male DAPs and pathway enrichment adds to the continually expanding evidence-base of sex-differences in molecular, physiological, and behavioral response to stress (see reviews(66-68)). This also further highlights inclusion of both sexes in biomedical/biobehavioral research remains critical towards a better understanding of sex-differences in disease, improve diagnosis, and in informing the development of targeted and sex-appropriate interventions (see additional reviews (69-71)). While sex-differences were profound, some general themes did emerge from the pathway analyses. In PFC, mitochondrial function and metabolic pathways were highly represented, a finding consistent with preclinical literature demonstrating the effects of stress on mitochondrial structure and function, as well as the growing body of clinical research implicating stress-induced mitochondrial and metabolic dysfunction in the pathophysiology of neuropsychiatric disorders (see review by (72)). Notably, there is substantial evidence of sexual dimorphism of mitochondria structure and function, with these fundamental biochemical differences significantly impacting energy production and other cellular processes that ultimately contribute to sex-differences in neuropathophysiology susceptibility and presentation(73-76). Thus, sex-specific therapies targeting mitochondrial dysfunction represents a promising avenue for future research.

In hippocampus, pathways involved in transcriptional regulation, such as mRNA processing and spliceosome cycle, were highly enriched. Stress, particularly through the action of stress steroids, has been shown to impact gene and protein expression in diverse ways, including through transcriptional regulation and alternative splicing(77, 78). Furthermore, successful mRNA processing and spliceosome function are critical for regulating gene expression and maintaining cellular proteostasis, while dysregulation of these processes is associated with various diseases(77-79) including stress-related neuropsychopathology(80-82). However, it is important to recognize that much of the physiological and molecular stress response is adaptive and promotes resilience(83). Thus, further investigation is needed to determine if the molecules and pathways identified here represent successful adaptations or failed resilience.

It should be remembered that SV2A is a single presynaptic protein that is being used as a proxy for “synapses” that are, in reality, dynamic structures involving multiple cells/cell types. Therefore, in addition to differential abundance analyses, we explored relationships between SV2A PET and protein abundance to provide insight into the greater neuroproteomic landscape underlying the SV2A PET measure. Irrespective of sex or stress, SV2A density correlated with both structural and functional synaptic proteins. Specifically, factors associated with promoting plasticity were positively correlated with BP_ND. For example, glutamate receptor interacting protein 1 (Grip1) and vesicle-associated membrane protein 1 (Vamp1, also known as synaptobrevin 1): Grip1 has been shown to regulate AMPA receptor trafficking and synaptic targeting and is required for long-term potentiation (LTP), learning and memory(84). Vamp1, like SV2A is an integral membrane protein of synaptic vesicles necessary for coordinated Ca2+-triggered synaptic neurotransmitter release(85). Reduced Vamp1 has also notably been observed in mouse models of Alzheimer’s Disease (AD) and in human AD postmortem brain(86). Conversely, phosphoglycerate dehydrogenase (Phdgh) was negatively correlated with SV2A density. Phdgh is an enzyme required for the biosynthesis of D-serine, a co-agonist of NMDA receptors regulating their function(87). Neuronal dysfunction and cognitive impairment have been linked to serine metabolism imbalance in aging and AD(88), and elevated phdgh levels in the brain and blood have been proposed as contributing to AD progression and an early biomarker of AD, respectively(89). Factors associated with cardiovascular disease and neurodegeneration, for example, calcium-binding protein S100a11(90) were additionally negatively correlated with SV2A density. Collectively, these findings provide evidence supporting the interpretation of SV2A density as measured by PET imaging as a proxy for synaptic density, and potentially, synaptic function.

This study has notable limitations. Firstly, we conducted only molecular synaptic measurements (albeit, both in vivo and postmortem). In the future, we plan to incorporate structural measures, such as dendritic arbor morphology and dendritic spine quantification(14, 16, 17), as well as physiological and electrochemical measures of synaptic function(14, 17). Second, this study used BP_ND as the PET outcome measure, with cerebellum as the reference region. In reality, a reference region should be devoid of, or have very minimal, specific radiotracer binding, which, for SV2A radiotracers, does not exist. Additionally, the cerebellum has historically been overlooked in the context of stress-related pathology despite evidence demonstrating cerebellar responsivity to stress and being critically involved in cognition and emotional processing(91-94). While we found no significant group or sex differences in cerebellum radiotracer uptake in this study cohort (Figure S1), permitting its use as a “pseudo-reference region”, there is still the potential of biasing the outcome measure. This limitation can be avoided by using volume of distribution (V_T) as the PET outcome measure rather than BP_ND. Calculating V_T requires derivation of the arterial input function acquired through arterial blood sampling and radiotracer metabolite quantification, a highly invasive (almost necessarily terminal) procedure in rodents(28). However, recent studies have demonstrated application of an image-derived input function and population-based metabolite correction as a non-invasive alternative(95) that we intend to adopt going forward. Third, PET imaging was performed under isoflurane anesthesia, which has been demonstrated to impact [¹⁸F]SynVest-1 dynamics, and subsequently, the outcome measure(96). Importantly, all rats that underwent imaging were similarly exposed to isoflurane. Further, all proteomics samples were from rats that underwent imaging/isoflurane anesthesia, thus minimizing bias. Finally, correlational analyses were performed across CON and CUS groups, a limitation stemming from a moderate sample size and restricted dynamic range for a number of outcome measures. It is therefore plausible that observed relationships are driven by group, rather than individual differences.

Strengths of this study include the use of both male and female rats and the integration of behavioral measures, in vivo imaging, and postmortem evaluations. Because of these key study design features, our findings provide unique translational insight regarding molecular mechanisms contributing to synaptic changes that can be measured in vivo using translational molecular imaging as well as further highlighting sex-differences in the behavioral and neurobiological stress response. The CUS model employed here plays a significant role in research aimed at understanding the pathophysiology of stress-related disorders and mechanisms of treatment response(3, 4, 97). Moreover, SV2A PET has been rapidly applied to clinical studies in stress-related psychiatric disorders, including our own studies demonstrating lower SV2A density in people with significant depressive symptoms and correlations between SV2A and mood and cognitive symptoms(32, 98). In conclusion, this study uses the same in vivo measure of synaptic density in the animal model that is currently being used in the target clinical population. This is a critical advancement towards true mechanistic translational research that can facilitate the process from preclinical discovery through to clinical application, aiding in the development of novel treatments to provide relief to those impacted by mental illness.

Supplementary Material

Supplement Methods and Figures S1-S4

Supplemental Tables S1-S2