Circulating PACAP levels are associated with altered imaging measures of entorhinal cortex neurite density in posttraumatic stress disorder

ABSTRACT

Introduction: Pituitary adenylate cyclase-activating polypeptide (PACAP) regulates plasticity in brain systems underlying arousal and memory and is associated with posttraumatic stress disorder (PTSD). Research in animal models suggests that PACAP modulates entorhinal cortex (EC) input to the hippocampus, contributing to impaired contextual fear conditioning. In PTSD, PACAP is associated with higher activity of the amygdala to threat stimuli and lower functional connectivity of the amygdala and hippocampus. However, PACAP-affiliated structural alterations of these regions have not been investigated in PTSD. Here, we examined whether peripheral PACAP levels were associated with neuronal morphology of the amygdala and hippocampus (primary analyses), and EC (secondary) using Neurite Orientation Dispersion and Density Imaging.

Methods: Sixty-four (44 female) adults (19 to 54 years old) with DSM-5 Criterion A trauma exposure completed the Clinician-Administered PTSD Scale (CAPS-5), a blood draw, and magnetic resonance imaging. PACAP38 radioimmunoassay was performed and T1-weighted and multi-shell diffusion-weighted images were acquired. Neurite Density Index (NDI) and Orientation Dispersion Index (ODI) were quantified in the amygdala, hippocampus, and EC. CAPS-5 total score and anxious arousal score were used to test for clinical associations with brain structure.

Results: Higher PACAP levels were associated with greater EC NDI (β = 0.0099, q = 0.032) and lower EC ODI (β = −0.0073, q = 0.047), and not hippocampal or amygdala measures. Neither EC NDI nor ODI was associated with clinical measures.

Conclusions: Circulating PACAP levels were associated with altered neuronal density of the EC but not the hippocampus or amygdala. These findings strengthen evidence that PACAP may impact arousal-associated memory circuits in PTSD.

HIGHLIGHTS

• PACAP was associated with altered entorhinal cortex neurite density in PTSD.

• PACAP was not associated with altered neurite density in amygdala or hippocampus.

• PACAP may impact arousal-associated memory circuits.

Introducción: El polipéptido activador de la adenilato ciclasa de la pituitaria (PACAP, por sus siglas en inglés) regula la plasticidad en los sistemas cerebrales subyacentes al estado de alerta y la memoria; y está asociado con el trastorno de estrés postraumático (TEPT). La investigación en modelos animales sugiere que el PACAP modula la entrada de la corteza entorrinal (CE) al hipocampo, contribuyendo al deterioro del condicionamiento contextual del miedo. En el TEPT, el PACAP está asociado con una mayor actividad de la amígdala ante estímulos amenazantes y una menor conectividad funcional de la amígdala y el hipocampo. Sin embargo, las alteraciones estructurales afiliadas al PACAP de estas regiones no han sido investigadas en el TEPT. Es por ello que, examinamos si los niveles periféricos del PACAP estaban asociados con la morfología neuronal de la amígdala y el hipocampo (análisis primarios) y la CE (secundario) utilizando imágenes de Densidad y Dispersión de Orientación de las Neuritas.

Métodos: Sesenta y cuatro (44 mujeres) adultos (de 19 a 54 años) con exposición a trauma según el Criterio A del DSM-5 completaron la Escala de trastorno de estrés postraumático administrada por el cínico según el DSM-5 (CAPS-5), se les extrajo una muestra de sangre y se les realizó una resonancia magnética. Se realizó el radioinmunoensayo del PACAP38 y se adquirieron imágenes potenciadas en T1 y en difusión multicapa. El índice de densidad de neuritas (NDI, por sus siglas en inglés) y el índice de dispersión de orientación (ODI, por sus siglas en inglés) fueron cuantificados en la amígdala, el hipocampo y la CE. Se utilizó la puntuación total del CAPS-5 y la puntuación del estado de alerta ansioso para evaluar las asociaciones clínicas con la estructura cerebral.

Resultados: Los niveles más altos del PACAP se asociaron con un mayor NDI de la CE (β = 0,0099, q = 0,032) y un ODI de la CE más bajo (β = −0,0073, q = 0,047), y no con medidas del hipocampo o la amígdala. Ni el NDI de la CE, ni el ODI de la CE se asociaron con las medidas clínicas.

Conclusiones: Los niveles circulantes del PACAP se asociaron con una densidad neuronal alterada de la CE, pero no del hipocampo o la amígdala. Estos hallazgos refuerzan la evidencia de que el PACAP puede afectar los circuitos de memoria asociados al estado de alerta en el TEPT.

KEYWORDS:

• Posttraumatic stress disorder
• diffusion imaging
• pituitary adenylate cyclase-activating polypeptide
• entorhinal cortex
• amygdala
• hippocampus

PALABRAS CLAVE:

• Trastorno de estrés postraumático
• imágenes de difusión
• lóbulo temporal medial
• polipéptido activador de la adenilato ciclasa de la pituitaria
• corteza entorrinal
• amígdala
• hipocampo

1. Introduction

Posttraumatic stress disorder (PTSD) is characterised by intrusion, avoidance, negative emotions, and hyperarousal symptoms after exposure to severe trauma (Ressler et al., 2022). Pituitary adenylate cyclase-activating polypeptide (PACAP) is a well-established neuromodulator of stress and arousal (Hammack & May, 2015), is involved in the secretion and production of corticotropin-releasing hormone (Agarwal et al., 2005; Boucher et al., 2021; Stroth & Eiden, 2010), stress-related sleep disturbance in mice (Foilb et al., 2024), and has been associated with hyperarousal symptoms of PTSD (Ressler et al., 2011). Stress-induced morphological alterations of the medial temporal lobe (MTL), including changes in amygdala and hippocampal volume, are implicated in PTSD (Shin et al., 2006). Notably, circulating PACAP levels and PAC1 receptors (PAC1Rs, encoded by ADCYAP1R1) have been associated with modulation of MTL functioning in animal models (Johnson et al., 2020), healthy adults (Porta-Casteràs et al., 2022), and people with PTSD (Stevens et al., 2014). Research in animal models has shown that PACAP-regulated pathways influence neuronal differentiation and are involved in neurotrophic processes including synaptogenesis and axon outgrowth in stress-sensitive MTL regions (Ogata et al., 2015; Rivnyak et al., 2018). In people with PTSD, functional imaging findings suggest that an ADCYAP1R1-associated risk polymorphism increases amgydala and hippocampal vulnerability to the harmful effects of stress (Stevens et al., 2014). However, a relationship between PACAP and MTL morphometry has not been reported in PTSD. To bridge this research gap, the recent emergence of neurite orientation dispersion and density imaging (NODDI) presents a timely opportunity. NODDI provides metrics of the orientations of axons and dendrites (together known as ‘neurites’), which represent potentially functionally-relevant tissue features not discerned by traditional volumetric analyses. Consequently, this study investigated the association between PACAP and NODDI indices within MTL regions in a sample of adults with PTSD.

Multiple lines of research indicate that PACAP modulates MTL circuitry, including the amygdala, hippocampus, and entorhinal cortex (EC). PAC1Rs are densely expressed in these regions in animals (Condro et al., 2016; Hashimoto et al., 2011; Jaworski & Proctor, 2000; Johnson et al., 2020; Piggins et al., 1996). Brief application of PACAP in rat hippocampal slices enhances synaptic activity and strength within the hippocampus and modulates fear conditioning (Johnson et al., 2020; Kondo et al., 1997; Schmidt et al., 2015). Similarly, PAC1R knockout mice exhibit deficits in long-term potentiation following high-frequency stimulation of EC input to the hippocampus (Matsuyama et al., 2003; Otto et al., 2001). Within the amygdala, PACAP infusion prior to fear conditioning increases synaptic plasticity and neuronal activation in a manner that disrupts fear acquisition in the short-term, and yet facilitates retention of fear cues in the longer term (Meloni et al., 2016). Altogether, this literature highlights the role of PACAP in modulating synaptic signalling and dendritic and axonal growth within the MTL in a manner relevant to mechanisms of fear learning.

Within the PTSD literature, PACAP systems have been strongly implicated in arousal-related symptoms and underlying functioning of the amygdala and hippocampus. In a seminal report, Ressler et al. (2011) found that circulating PACAP, ADCYAP1R1 SNP polymorphisms, and ADCYAP1R1 DNA methylation were associated with PTSD diagnosis, and this has since been replicated in independent samples (Almli et al., 2013; Uddin et al., 2013; Wang et al., 2013, but see Chang et al., 2012). Circulating PACAP was also correlated with significantly greater severity of hyperarousal symptoms in women with PTSD (Ressler et al., 2011). In a functional imaging study of traumatised women, ADCYAP1R1 polymorphism predicted greater amygdala reactivity to fearful stimuli and lower functional connectivity between the amygdala and hippocampus (Stevens et al., 2014). Similarly, our group has recently shown that greater circulating PACAP was related to greater amygdala connectivity with posterior default mode network regions (Clancy et al., 2023). Together, this supports investigation of PACAP with structural indices that may influence amygdala and hippocampus function in PTSD.

Although no study has yet examined PACAP in relation to the EC in PTSD, multiple lines of research support doing so. First, the EC functions as a gateway for amygdala modulation of hippocampal activity (Roesler & McGaugh, 2022). As an intermediate node, the EC may be an additional factor impacting the previously reported associations of PACAP with amygdala and hippocampus functional connectivity. This would align with evidence that PACAP-affiliated glutamatergic neurons of the EC project to the hippocampus (Johnson et al., 2020). Second, the EC supports threat processing functions that are known to be modulated by PACAP and affected in PTSD (Feng et al., 2021; Kang & Han, 2021), including contextual fear conditioning (Feng et al., 2021; Sparta et al., 2014). Finally, the EC has especially dense expression of PACAP and its receptors, relative to other brain regions (Condro et al., 2016; Hashimoto et al., 1996; Palkovits et al., 1995). To our knowledge, this is the first study examining whether PACAP relates to EC structural features in people with PTSD.

Diffusion-weighted MRI, specifically NODDI (Zhang et al., 2012), has been used to detect differences in gray matter microstructure. NODDI provides researchers with surrogate measures of neuronal integrity hypothesized to relate to the number of neurites and complexity of dendrites (Neurite Density Index: NDI) as well as the dispersion of axons and neurons (Orientation Dispersion Index: ODI) which have been shown to be related to functional dynamics (Nazeri et al., 2015). NODDI metrics have recently been proposed as candidates for next-generation biomarkers in psychiatry (Kamiya et al., 2020); however, to our knowledge, these methods have never been applied to study structural brain correlates of the stress response in individuals with PTSD. Using NODDI could enhance understanding of cellular characteristics, differences, or changes in PTSD that might influence regional brain function.

To address this gap in the literature, this study examined whether peripheral PACAP was associated with regional NODDI measures of gray matter microstructure in the MTL of adults with DSM-5 Criterion A trauma exposure and a range of PTSD symptoms. Our primary analyses investigated the relationship of PACAP with NDI and ODI in the amygdala and hippocampus. This was motivated by prior PTSD research linking PACAP to the functioning of these two specific regions. Additionally, we conducted analyses examining PACAP in relation to NDI and ODI of the EC. These analyses were considered secondary due to the absence of pre-existing PTSD literature on PACAP and EC phenotypes, although supported by neuroscience literature on EC PACAP receptor density and EC function. To determine the clinical relevance of significant findings, we examined whether PACAP-associated NDI and ODI metrics were related to total PTSD symptom severity and anxious arousal symptoms. Examination of anxious arousal symptoms was based on previous evidence that higher peripheral PACAP levels were associated with hyperarousal symptoms and that individuals with a polymorphism in the gene encoding PAC1Rs may show greater physiological fear responses including hyperarousal and startle discrimination (Ressler et al., 2011). Factor analytic investigations of the latent structure of PTSD symptoms have shown that self-reported hypervigilance and startle symptoms cluster together as a part of an anxious arousal construct, separately from symptoms of dysphoric arousal (e.g. irritability, trouble sleeping) (Armour et al., 2012).

2. Materials and methods

2.1. Participants

One-hundred and three (103) trauma-exposed community-based adults were recruited from the greater Boston metropolitan area. Of these individuals, 84 were scanned with the Human Connectome Project (HCP) Adult Lifespan protocol and 19 with the HCP Young Adult protocol. There is evidence that NODDI metrics across gray and white matter are susceptible to differences in acquisition methods. For instance, NDI is sensitive to the choice of outer shell b-value, while ODI is sensitive to the number of gradient directions (Parvathaneni et al., 2018), and that variability of NODDI measures in shell-schemes tracks with low b-values (Lucignani et al., 2021). To avoid the possibility that NODDI measures were impacted by varying protocols in this clinically-assessed sample, we used only the participants scanned with the majority-use HCP Lifespan protocol. Twenty (20) of the 84 participants scanned with the HCP Adult Lifespan protocol did not have processed PACAP data. The final sample consisted of 64 trauma-exposed community-based adults (19–54 years old; 44 female). The distribution of age by sex is shown in Supplementary Figure 1. Participants were included if they met DSM-5 Criterion A trauma exposure for PTSD and at least two of the four PTSD symptom clusters (B-E). Exclusion criteria were left-handedness, confounding medical condition including untreated seizure disorder or neurological disorder, inability to tolerate blood draws, history of head trauma with loss of consciousness >5 min, current treatment with an antipsychotic (unless prescribed for PTSD-related sleep and nightmares), current (past month) alcohol use disorder or moderate/severe substance use disorder, current psychotic disorder, anorexia, obsessive-compulsive disorder, manic or mixed mood episode, lifetime history of schizophrenia or schizoaffective disorder, MR contraindications (e.g. metal implants and claustrophobia), positive pregnancy test on day of MRI scan (female participants), and history of receiving hormonal replacement therapy or undergoing gender confirmation surgery. All study procedures were approved by the Mass General Brigham Human Research Committee (Protocol 2019P000626). Characteristics of the sample are summarised in Table 1.

Table 1. Sample characteristics (N = 64).

Characteristic	Mean ± SD or N (%)
Age (years)	29.73 ± 8.63
Female Sex	44 (69%)
Education
Masters or Doctorate	9 (14%)
College Degree	25 (39%)
Some College	27 (42%)
High School Degree	2 (3%)
<12th grade	1 (2%)
Race/Ethnicity
Asian	4 (6%)
Black	3 (5%)
Hispanic/Latino	5 (8%)
Multiracial	10 (16%)
White	42 (66%)
PACAP (pM)	896.90 ± 686.80
LEC-5 Total	14.28 ± 8.50
Motion	0.25 ± 0.06
CAPS-5 Total	28.52 ± 10.75
CAPS-5 Anxious Arousal	2.88 ± 2.08

Note: PACAP: pituitary adenylate cyclase-activating polypeptide; LEC-5 Total: Life Events Checklist for DSM-5 total; motion: average relative motion during scan; CAPS-5: Clinician Administered PTSD Scale for DSM-5.

2.2. Clinical interviews and questionnaires

2.2.1. Life Events Checklist for DSM-5 (LEC-5)

The LEC-5 was given as a survey of 17 potentially traumatic events and used to determine DSM-5 Criterion A trauma exposure and an index trauma for the interview.

2.2.2. Clinician-Administered PTSD Scale for DSM-5 (CAPS-5)

The CAPS-5 (Weathers et al., 2018) was administered by doctoral-level psychologists and used to determine DSM-5 PTSD diagnosis and symptom severity. The CAPS-5 is currently the gold standard to determine whether an individual meets criteria for the diagnosis of PTSD. The CAPS-5 consists of 30 items to assess the onset, duration, and impact of PTSD symptoms We derived CAPS-5 total score and anxious arousal subscore. The latter equalled the sum of responses to Item 17 (E3) Hypervigilance and Item 18 (E4) Exaggerated Startle Response (Armour et al., 2012).

2.3. PACAP immunoassays

Blood samples were collected at the beginning of each study visit. Participants were instructed not to eat the morning of their visit prior to their blood draw which was scheduled between 8 and 10 AM (see additional details in the Supplement). Human plasma samples were prepared as described previously (Ressler et al., 2011; Ross et al., 2020), and all human PACAP38-specific measurements were performed at the University of Vermont, Larner College of Medicine, using double antibody sandwich ELISA immunoassays (Cat. No. HUFI02692, AssayGenie, Dublin, Ireland). Samples were centrifuged at 3500 rpm for 15 min. Plasma was extracted and stored at −80°C until analysis. Optimal sample volume was determined in dilution tests, and all values represent the mean from assay duplicates; intra-assay variation was approximately 9%. The assay midpoint was 1.1 fmol and the detection limit from the linear range of the standard curve was 0.2 fmol. Outliers with exceedingly high (n = 2) concentrations of PACAP levels were winsorized to the next highest reliable/non-outlier value (Clancy et al., 2023; Ross et al., 2020).

2.4. Imaging

MRI data were acquired on a 3.0 Tesla Siemens Prisma Scanner at the McLean Imaging Center. Anatomical whole-brain images were obtained. A T1-weighted MP-RAGE sequence was acquired; TR = 2500 ms, TE = 7.27 ms, flip angle = 8 deg, FOV = 256 × 256, voxel size = 0.8 mm isotropic. A high-angular multi-shell whole brain diffusion-weighted imaging sequence was also acquired; TR = 3230, TE = 89.5 ms; flip angle = 78deg, refocusing flip angle = 160deg, FOV = 210 × 210; matrix = 168 × 144, slice thickness = 1.5 mm, 92 slices, voxel size = 1.5 mm isotropic, b = 1500 and 3000 s/mm², ∼47 directions/shell, multiband factor = 4. An identical sequence acquired with opposite-phase encoding was also acquired and used to correct for magnetisation-induced susceptibility distortion (see below).

2.5. Image processing

T1-weighted images were processed using Freesurfer 6 (Dale et al., 1999). Freesurfer-processed T1-weighted images were skull-stripped using AFNIs 3dSkullStrip programme and corrected for signal intensity changes (Cox, 1996). Opposite phase encoding B0’s were merged for FSL’s topup programme (Andersson et al., 2003; Smith et al., 2004). Topup results were fed into FSL’s eddy programme to correct for motion, eddy current artifacts, and magnetisation induced susceptibility distortion (Andersson & Sotiropoulos, 2016). We used the –repol option to replace outliers and further improve the quality of the data (Andersson et al., 2016). Rotated vectors were used for subsequent analysis. Neurite Density Index (NDI) and Orientation Dispersion Index (ODI) were calculated using the Microstucture Diffusion Toolbox (MDT) (Harms et al., 2017). Skull-stripped T1 data were aligned to the NDI image using antsRegistrationSynQuick.sh from Advanced Normalization Tools (ANTS) (Avants et al., 2009). Diffeomorphic transformations from the skull-stripped T1 image to the NDI image were applied to each region of interest using ANTS. Voxel data for each hemisphere, measure, and region type were extracted using AFNI’s 3dmaskdump (Cox, 1996). Separate left and right hemisphere data for each region and NODDI metric were entered as dependent variables in linear mixed effects models. For post-hoc analyses with clinical measures, hemisphere data were averaged for a given measure and region type. Volumes of each hemisphere and region type were calculated prior to diffeomorphic transformation to native diffusion space. Alignment of each region of interest to the NDI image was checked with tools distributed with scikit-learn (Abraham et al., 2014). We analysed histograms of raw voxel values for all metrics bilaterally. We observed a non-normal distribution of voxel intensities in NDI metrics. To correct this bias, we filtered out voxels with NDI values of over 0.98. All individuals had at least 70% retained voxels after voxel filtering (Supplementary Table 1). Participant motion was quantified using FSLs' quality control eddy_quad programme for average motion relative to the previous sub-brick and was included as a nuisance regressor in statistical models (Bastiani et al., 2019).

2.6. Statistical analyses

Linear mixed effects models associated our NODDI measures of primary interest (NDI and ODI from the amygdala and hippocampus) and secondary interest (NDI and ODI from the EC) with our predictor of interest, circulating PACAP levels. These models included fixed effects of PACAP, hemisphere, age, sex, and participant motion, and random intercepts to account for correlations between left and right hemisphere measurements from the same subject. Interactions of PACAP with hemisphere and PACAP with sex were added separately to each model. Due to limited power to detect interaction effects, separate PACAP coefficient estimates for left and right hemispheres and males and females based on interaction models are reported in supplementary materials alongside estimates from PACAP by hemisphere interaction and estimates from PACAP by sex interaction regardless of the significance of statistical tests for interaction. Confidence intervals for linear combinations of parameters from interaction models were calculated at the individual 95% confidence level using the ‘estimable’ function from the gmodels package in R (Warnes et al., 2022). Hommel multiple testing corrections were applied separately to hypothesis testing results for models quantifying associations of primary interest (4 models; amygdala and hippocampal NDI and ODI) and quantifying associations of secondary interest (2 models; EC NDI and ODI); these corrections were carried through to tests for interaction by hemisphere and sex. Additional models were fit for regional NODDI metrics that were significantly associated with PACAP after multiple comparison correction to determine if findings remained after accounting for regional (left and right) volumes. Normal quantile-quantile plots for the fitted random intercepts and residuals for each linear mixed effects model were used for assessing model fit. Plots were generated using ‘residplot’ function from the predictmeans package in R (Luo et al., 2023).

For regional NODDI metrics that were significantly associated with PACAP after multiple comparison correction, post hoc linear regression examined their associations with clinically-relevant measures including CAPS-5 total score, anxious arousal subscore, and total score on the LEC-5. All post-hoc models controlled for age, sex, and motion. Post-hoc analyses incorporated averaged left and right EC NDI and averaged left and right ODI values given lack of statistically significant interactions of circulating PACAP with hemisphere. No multiple testing correction was applied to results of post hoc analyses. Statistical significance tests were two-sided and conducted at the alpha = 0.05 significance level after correction. No strategy for accomodating missing data was required because data were complete for the final sample. Data analysis was conducted in R-Studio (http://r-project.org/.).

3. Results

Table 2 shows the results of all linear mixed effects analyses including the coefficient estimates for all covariates. Supplementary Table 2 shows coefficient estimates by hemisphere and estimates of left and right hemisphere differences from models including interactions of PACAP with hemisphere for primary and secondary analyses. Supplementary Table 3 shows coefficient estimates by sex and estimates of sex differences from models including interactions of PACAP with sex for primary and secondary analyses. Supplementary Figure 2 shows the relationship (pairwise Pearson correlation coefficients) among all study variables and their distributions. Supplementary Figure 3 shows normal quantile-quantile plots for fitted random intercepts and residuals from linear mixed effects models including PACAP, hemisphere, age, sex, and motion for all primary and secondary NODDI measurements.

Table 2. Results of linear mixed effects models for our primary analyses (models predicting hippocampal and amygdala NDI and ODI) and secondary analyses (models predicting entorhinal cortex NDI and ODI)

	NDI × 1000		ODI × 1000
	β (95% C.I.)	q	β (95% C.I.)	q
PACAP
Amygdala	0.0041 (−0.0024, 0.011)	0.59	0.0027 (−0.0071, 0.012)	0.59
Hippocampus	0.0068 (0.00061, 0.013)	0.13^┼	0.0026 (−0.0054, 0.010)	0.59
Entorhinal Cortex	0.0099 (0.0019, 0.018)	0.032*	−0.0073 (−0.015, −0.00011)	0.047*
Hemisphere
Amygdala	−6.32 (−9.99, −2.65)	0.004*	4.36 (−1.39, 10.096)	0.27
Hippocampus	3.58 (0.19, 6.97)	0.12^┼	−0.55 (−4.37, 3.27)	0.77
Entorhinal Cortex	4.83 (−1.14, 10.80)	0.11	6.82 (1.74, 11.90)	0.019*
Age
Amygdala	0.82 (0.32, 1.31)	0.005*	0.35 (−0.38, 1.087)	0.59
Hippocampus	0.73 (0.27, 1.20)	0.008*	0.16 (−0.43, 0.76)	0.59
Entorhinal Cortex	0.86 (0.26, 1.46)	0.012*	0.37 (−0.17, 0.92)	0.17
Sex
Amygdala	−5.87 (−15.95, 4.21)	0.51	1.082 (−13.99, 16.15)	0.89
Hippocampus	−12.87 (−22.39, −3.35)	0.036*	−5.92 (−18.19, 6.35)	0.68
Entorhinal Cortex	−10.43 (−22.82, 1.95)	0.19	4.83 (−6.37, 16.02)	0.39
Motion
Amygdala	−6.069 (−69.022, 56.88)	0.86	55.54 (−38.61, 149.69)	0.86
Hippocampus	−5.41 (−64.88, 54.055)	0.86	−28.16 (−104.80, 48.49)	0.86
Entorhinal Cortex	33.97 (−43.42, 111.36)	0.52	−22.47 (−92.41¸ 47.47)	0.52

Note: Linear mixed effects results for PACAP, hemisphere, age, sex, and motion predicting NDI and ODI for each region. q-values (corrected p-values) are reported. We applied Hommel corrections for multiple comparisons for our primary (amygdala and hippocampus) and secondary analyses (entorhinal cortex) separately. Unadjusted beta values were multiplied by 1000 for ease of reporting. Unadjusted 95% confidence intervals are reported. * Indicates significance at the corrected level, ^┼ indicates significance at the uncorrected level.

3.1. Primary analysis: relationship of circulating PACAP levels with hippocampus and amygdala NODDI metrics

For analyses examining circulating PACAP in relation to hippocampal NDI, there was a positive effect of PACAP at the uncorrected level (β = 0.0068, p = .032), which did not survive correction for multiple comparisons (q = 0.13). Hippocampal NDI was positively correlated with age (β = 0.73, q = 0.008) and was greater in males (β = −12.87, q = 0.036). There was an effect of hemisphere (β = 3.58, p = .039) that did not survive correction for multiple comparisons (q = 0.12). Hippocampal NDI was not associated with participant motion. For the linear mixed effects model associating PACAP with hippocampal ODI, the effect of PACAP was not significant, nor were the effects of hemisphere, age, sex, and motion covariates.

For the linear mixed effects model associating PACAP with amygdala NDI, the PACAP association was not significant, nor were the contributions of sex and participant motion. However, we observed a significant positive effect of age (β = 0.82, q = 0.005) and hemisphere (β = −6.32, q = 0.004) on amygdala NDI. For the linear mixed effects model associating PACAP with amygdala ODI, the effects of PACAP, hemisphere, age, sex, and participant motion were all non-significant.

We found no statistically significant interactions of PACAP with either hemisphere or sex for all primary NODDI measures. Coefficient and difference estimates from interaction models by hemisphere and sex and results of statistical tests for interaction are provided in Supplementary Tables 2 and 3.

3.2. Secondary analysis: relationship of PACAP with EC NODDI metrics

For the linear mixed effects model associating PACAP with EC NDI, there was a significant positive effect of PACAP at the uncorrected and corrected levels (β = 0.0099, q = 0.032), controlling for a significant positive effect of age (β = 0.86, q = 0.012) and non-significant effects of sex and participant motion. For the linear mixed effects model associating PACAP with EC ODI, there was a significant negative effect of PACAP at the uncorrected and corrected levels (β = −0.0073, q = 0.047), controlling for non-significant effects of age, sex, and participant motion. We also observed higher EC ODI in the right hemisphere (β = 6.82, q = 0.019). Additional analyses revealed that the association between PACAP and EC NDI (β = 0.0098, p = .019) as well as the relationship between PACAP and EC ODI (β = −0.0073, p = .049) persisted after inclusion of hemispheric EC volume as a control covariate in the linear mixed effects model.

There were no statistically significant interactions of PACAP with hemisphere or sex for either EC NDI or EC ODI. EC coefficient and difference estimates from interaction models by hemisphere and sex and results of statistical tests for interaction are included in Supplementary Tables 2 and 3.

3.3. Post hoc analyses of NODDI Metrics with clinical symptoms

Neither average EC NDI (β = 4.18, p = .95) nor average ODI (β = −114.43, p = .14) was associated with CAPS-5 total score, covarying for age, sex, and motion. In addition, neither average EC NDI (β = 12.066, p = .39) nor average ODI (β = −13.608, p = .39) was associated with the anxious arousal subscore, covarying for age, sex, and motion. Neither average EC NDI (β = 33.45, p = .20) nor average EC ODI (β = −33.14, p = .26) was associated with total LEC-5 score, covarying for age, sex, and motion.

4. Discussion

Here we examined the relationship between PACAP levels circulating in blood and NODDI measures of dendritic and cellular complexity of the MTL in people with PTSD. Our findings suggest that circulating PACAP is particularly relevant to the microstructure of the EC. We found that higher PACAP levels were associated with higher NDI and lower ODI of the EC, after accounting for the effects of hemisphere, age, sex, and motion. While we had stronger a priori predictions regarding the associations between PACAP levels and the microstructure of the amygdala and hippocampus, a positive association between PACAP levels and hippocampal NDI was not significant after correcting for multiple comparisons. Finally, exploration of clinical correlations revealed that EC NODDI measures did not track with total PTSD symptom severity, anxious arousal symptoms, or cumulative exposure to traumatic life events. Altogether, these findings suggest that higher levels of peripheral PACAP are associated with EC microstructural properties in PTSD.

Our strongest finding indicates that higher PACAP levels are associated with greater NDI of the EC, which may reflect lower dendritic complexity (Radhakrishnan et al., 2020). Studies of aging have shown that greater hippocampal NDI is found in older adults compared with younger adults, and is associated with poorer episodic memory performance, thus capturing morphological changes related to impaired MTL gray matter function (Radhakrishnan et al., 2020; Radhakrishnan, Bennett, et al., 2022). Our findings of positive associations between age and NDI of the amygdala, hippocampus, and EC are consistent with this literature on aging, lending confidence that they may reflect similar underlying microstructural properties. Notably, PACAP enhances the excitability of the hippocampus through upstream regions such as the EC (Johnson et al., 2020) and via known projections (Witter et al., 2017; Witter & Amaral, 1991). Using magnetic resonance spectroscopy, we previously found lower neuronal integrity along with higher glutamate levels in the hippocampus of individuals with PTSD, which is potentially compatible with excitotoxic mechanisms (Rosso et al., 2017). Therefore, we speculate that a reduction in dendritic complexity of the EC in association with PACAP may have downstream consequences on stress-related excitotoxic-neuronal loss in the hippocampus. Further research is needed to fully test this hypothesis in relation to PACAP.

PACAP was also associated with lower EC ODI, which indicates less occupancy of the extraneurite space (Kamiya et al., 2020). Animal studies pairing immunohistochemistry and high-resolution diffusion-weighted MRI suggest that reduced ODI in gray matter is associated with selective depletion of microglia (Yi et al., 2019). Changes in glial cell occupancy may drive lower ODI values found in gray matter, as glia take up a large proportion of extra-neurite space. Hence, we speculate that higher peripheral PACAP may be associated with lower microglial density (lower ODI) in the EC in PTSD. Altogether, these associations between PACAP and altered NODDI measures have implications for studies that have found differences in EC volume (Ben-Zion et al., 2020; Lopez et al., 2017; Mo et al., 2022) and function (Mueller-Pfeiffer et al., 2013) in individuals with PTSD.

The finding that circulating PACAP levels correlate with altered EC NODDI measures is important in the context of prior research showing disrupted connectivity between the amygdala and hippocampus and greater activity of the amygdala in response to threat stimuli in trauma-exposed individuals with the ADCYAP1R1 genotype (Stevens et al., 2014). Because the EC acts as a gateway for amygdala influence on hippocampus-based functions (Roesler & McGaugh, 2022), we suggest that PACAP may induce morphological changes in the EC that hinder effective communication between the amygdala and hippocampus in individuals with PTSD. This suggestion gains credence from the known relationship between changes in NODDI measures and dynamics in resting state connectivity (Nazeri et al., 2015). We note that we did not identify associations of EC NDI and EC ODI with overall PTSD symptoms, anxious arousal symptoms, or trauma load. However, we suggest that PACAP-related microstructural alterations of the EC in PTSD may have functional relevance for other aspects of this disorder, including arousal measures not captured by standard clinical measures. Based on prior evidence that PACAP relates to hyperarousal in PTSD, digital phenotyping metrics of arousal would be worth examination in future studies. Recent work in mice shows that acute administration of PACAP can regulate sleep architecture captured using subcutaneous detectors (Foilb et al., 2024), although future work is needed to precisely model the types of long-term changes in PACAP function and sleep that might be seen in people with PTSD. Nonetheless, further research is needed to test the hypothesis that structural changes of the EC may serve as an intermediary factor influencing the relationship between PACAP and lower functional connectivity of the amygdala and hippocampus in PTSD.

Unexpectedly, this investigation did not reveal a relationship between PACAP and NDI or ODI of the hippocampus and amygdala, after adjusting for multiple comparisons. This outcome is surprising given prior research in animals demonstrating that PACAP signalling in the central nucleus of the amygdala increases anxiety-like responses (Missig et al., 2014), as well as human studies showing higher amygdala reactivity to threat stimuli and lower amygdala-hippocampus functional connectivity in women with a PACAP-receptor genetic polymorphism (Stevens et al., 2014). One potential reason for our failure to detect robust differences within the amygdala and hippocampus could be the specificity of action of PACAP within subnuclei of these regions (e.g. Meloni et al., 2019), which we did not investigate in this study.

Along these lines, in the current investigation, we looked at the hippocampus as a unitary structure without regard to its nuanced subregional anatomy and function (Amaral & Lavenex, 2007; Yassa & Stark, 2011). Our prior research has shown that high-resolution diffusion MRI can outperform whole-brain standard resolution in the detection of tensor-based microstructural differences between subregions of the hippocampus, including the dentate gyrus and CA1 (Granger et al., 2022). Thus, differences in microstructure of hippocampal subregions conducted with standard resolutions (i.e. 1.5 mm isotropic) related to various pathologies may be less meaningful or could be improved with high-resolution diffusion imaging, as others have also shown (Merenstein et al., 2023). Cognizant of this body of work, we did not pursue subregional analysis. An important avenue for future research would be to determine whether higher circulating PACAP is associated with microstructural measurements of hippocampal subfields and whether using high-resolution diffusion-weighted MRI enhances these associations. Similarly, we opted for a unitary approach to examine EC structure, despite its domain-sensitive subregions specialising in distinct episodic memory processes (Hunsaker et al., 2013; Reagh & Yassa, 2014; Schultz et al., 2012). Parsing the medial and lateral EC with high-resolution diffusion imaging represents an important future direction which, given their relevance to posterior-medial and anterior-temporal cortical networks, may provide further insights into the neurobiological implications of greater circulating PACAP in arousal-mediated memory circuits (Hisey et al., 2023; Ritchey et al., 2015). Despite the potential for increased granularity of NODDI measures in these regions to yield deeper information, to our knowledge, we are the first to apply this method to studies of MTL gray matter in PTSD.

Our results should be interpreted in the context of several limitations. The interpretation of our results is particularly challenging due to the lack of published literature applying NODDI methods to the MTL in individuals with PTSD. While it is possible that these imaging measures reflect lower dendritic complexity and glial cell density, caution must be exercised in interpreting NODDI measures at the cellular level (Radhakrishnan, Shabestari, et al., 2022). Histological validation of the relationship between PACAP and EC NODDI metrics in animal models would provide additional biological insights into the underlying mechanisms. In addition, this study had a small sample size, primarily composed of females. Though our analysis tested for interactions between PACAP levels and sex, our study was not adequately powered to detect more moderate sex differences, and absence of a statistically significant interaction should not be equated with absence of a scientifically meaningful difference. Because of prior literature suggesting that the relationship between PACAP and hyperarousal is selective to females (Ressler et al., 2011) and evidence of differential development of PACAP receptors in the female hippocampus (Shneider et al., 2010), future investigations with larger sample sizes should investigate the interplay between sex, PACAP, and subfield structure. Additionally, future studies should investigate the link between circulating PACAP levels and brain PACAP levels. It is worth noting that while we had predefined research questions about associations of PACAP with MTL regions due to prior studies of PACAP within PTSD, PACAP and its receptors are more widely expressed in other cortical and subcortical regions as well (Condro et al., 2016; Hashimoto et al., 1996; Palkovits et al., 1995). Therefore, future work should consider additional PACAP-affiliated regions pending further description of PACAP receptor distribution and relevance to PTSD pathophysiology. Additional analyses such as examining voxel-wise associations between circulating PACAP and NODDI measures within PACAP-affiliated regions or the incorporation of high-resolution diffusion imaging in human studies may yield additional insight into PACAP-related microstructural alterations in PTSD.

Despite these limitations, our investigation boasts several strengths. We implemented opposite-phase encoding to correct for magnetisation-induced susceptibility distortion, a crucial step given that the MTL is particularly susceptible to signal distortion (Olman et al., 2009). In addition, we included motion and EC volumes as additional nuisance regressors and found that the relationship between PACAP and EC NDI and ODI persisted. Finally, we observed a non-normal distribution of NDI values across all voxels of the EC on a subject-by-subject basis. Consequently, we filtered out problematic voxels that likely reflected cerebrospinal fluid or pia/dura mater. Notably, the persistence of relationships between PACAP and EC and hippocampal NDI after filtering reinforces our confidence in these findings, as initially observed strong associations remained intact despite the removal of voxels with abnormally high NDI values.

5. Conclusions

In conclusion, our findings highlight the association between circulating PACAP levels and gray matter microstructure of the EC in individuals with PTSD. Higher PACAP levels were related to higher NDI and lower ODI, suggesting less EC dendritic complexity and cellular density, respectively. These findings offer insights into cellular integrity of the MTL and motivate further investigation into the role that EC microstructure may play in PTSD-related alterations of the amygdala-hippocampus network. Future research endeavours aimed at validating these interpretations should include histological validation as a function of region and tissue type. In addition, future studies should examine subregions of the hippocampus and amygdala to further elucidate the role of PACAP within the MTL in PTSD.

Disclosures

Within the past three years, WAC has served as a consultant for Psy Therapeutics and has sponsored research agreements with Cerevel and Delix. NPD has served on scientific advisory boards for BioVie Pharma, Circular Genomics and Sentio Solutions for unrelated work. KJR has performed scientific consultation for Acer, Bionomics, and Jazz Pharma; serves on Scientific Advisory Boards for Sage, Boehringer Ingelheim, Senseye, Brain and Behavior Research Foundation and the Brain Research Foundation, and he has received sponsored research support from Alto Neuroscience. SLR is paid as secretary of Society of Biological Psychiatry, and for Board service to Mindpath Health/Community Psychiatry and National Association of Behavioral Healthcare; served as volunteer member of the Board for The National Network of Depression Centers; received royalties from Oxford University Press, American Psychiatric Publishing Inc, and Springer Publishing; received research funding from NIMH. EAO is employed by Crisis Text Line.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

Data are available through the NIMH National Data Archive (NDA; https://nda.nih.gov/edit_collection.html?id=3166) upon reasonable request to the senior author, IMR.

Additional information

Funding

This work was supported by NIH award P50MH115874 (to WAC and KJR, PDs; IMR, SLR, Project 4 PIs). In addition, the investigators were partially supported by R01MH120400 (IMR), R01MH97988 (SEH and VM).