The Neurosteroids Allopregnanolone and DHEA Modulate Resting-State Amygdala Connectivity

Abstract

The neurosteroids allopregnanolone and dehydroepiandrosterone (DHEA) are integral components of the stress response and exert positive modulatory effects on emotion in both human and animal studies. Though these antidepressant and anxiolytic effects have been well established, little research to date has examined their neural correlates, and no research has been conducted into the effects of neurosteroids on large-scale networks at rest. To investigate the neurosteroid impact on intrinsic connectivity networks, participants were administered 400 mg of pregnenolone (N = 16), 400 mg of DHEA (N = 14), or placebo (N = 15) and underwent 3T fMRI. Resting-state brain connectivity was measured using amygdala as a seed region. Compared to placebo, pregnenolone administration reduced connectivity between amygdala and dmPFC, between amygdala and precuneus, and between amygdala and hippocampus. DHEA reduced connectivity between amygdala and peri-amygdala and between amygdala and insula. Reductions in amygdala to precuneus connectivity were associated with less self-reported negative affect. These results demonstrate that neurosteroids modulate amygdala functional connectivity during resting-state, and may be a target for pharmacological intervention. Additionally, allopregnanolone and DHEA may shift the balance between salience network and default network, a finding that could provide insight into the neurocircuitry of anxiety psychopathology.

Introduction

Allopregnanolone and dehydroepiandrosterone (DHEA) are endogenously-produced neurosteroids with neuroprotective, anxiolytic, antidepressant, antioxidant, and antiglucocorticoid effects (Belelli and Lambert, 2005; Maninger, et al., 2009; Patchev, et al., 1994). Deficiencies in allopregnanolone and DHEA are related to anxiety and depressive-like behavior. For instance, endogenous levels of allopregnanolone and DHEA are decreased in animal models of depression, and these animals exhibit behavioral improvements after DHEA administration or allopregnanolone induction (Genud, et al., 2009; Pibiri, et al., 2008). DHEA and allopregnanolone administration reduce immobility in the forced swim test (Frye, et al., 2004; Urani, et al., 2001), and reduce stress and anxiety-like behavior in rodents (Frye and Rhodes, 2006; Melchior and Ritzmann, 1994). In human studies, allopregnanolone and DHEA dysregulation is associated with anxiety and depression. Individuals with major depressive disorder exhibit reduced plasma and cerebrospinal fluid (CSF) levels of allopregnanolone (Romeo, et al., 1998; Uzunova, et al., 1998) and DHEA (Barrett-Connor, et al., 1999; Morsink, et al., 2007; Wong, et al., 2011), and those with posttraumatic stress disorder (PTSD) show reduced CSF allopregnanolone (Rasmusson, et al., 2006). Furthermore, increases in serum allopregnanolone and DHEA over a course of treatment predict symptomatic improvement (Amin, et al., 2006; Olff, et al., 2007; Uzunova, et al., 1998; Yehuda, et al., 2006). Therefore, these neurosteroids show promise as targets for antidepressant and anxiolytic effects in psychiatric disorders.

Converging preclinical and neuroimaging research suggests that the amygdala, a key region in threat detection (Adolphs, et al., 1999), fear conditioning (Armony and LeDoux, 1997), and emotional salience (Whalen, et al., 2001), may be a particular locus of allopregnanolone and DHEA’s effects. Our laboratory has recently demonstrated that single-dose DHEA (Sripada, et al., 2013a) and the allopregnanolone precursor pregnenolone (Sripada, et al., 2013b) decrease amygdala activity and increase activity in medial prefrontal regions during tasks engaging cognition-emotion interactions. It had been reported that progesterone administration (which in turn increases allopregnanolone levels) modulates amygdala responsivity to emotional faces (van Wingen, et al., 2007), and increases functional connectivity between amygdala and dorsal medial prefrontal cortex (dmPFC) (van Wingen, et al., 2008). Allopregnanolone modulates GABA(A) receptor-mediated inhibitory postsynaptic currents in the central nucleus of the amygdala (Wang, et al., 2007), and microinfusions of allopregnanolone directly into the amygdala produce anxiolytic (Engin and Treit, 2007) and antidepressant (Shirayama, et al., 2011) effects. The anxiolytic and antidepressant effects of DHEA may be mediated by the amygdala as well. In animal models, DHEA administration increases BDNF concentration (Naert, et al., 2007) and 5-HT(2A) receptor expression (Cyr, et al., 2000) in the amygdala, and DHEA infusions into the amygdala modulate stress responsivity (Singh, et al., 1994). Thus, the amygdala might be a key region in the anxiolytic and emotion modulatory effects of neurosteroids.

Though mounting evidence suggests that neurosteroids modulate amygdala activity, the reported direction of amygdala modulation is somewhat inconsistent. Activation of amygdala has been reported as increased (van Wingen, et al., 2008), or decreased (Sripada, et al., 2013b; van Wingen, et al., 2007) by allopregnanolone, depending on the task employed. To fully understand the impact of neurosteroids on amygdala function, it may be necessary to examine their effect on amygdala connectivity at rest. Resting-state connectivity offers a powerful way to assess intrinsic connections between brain networks (Fox and Raichle, 2007; Fox, et al., 2005; Raichle, et al., 2001) without external demands or confounds imposed by specific tasks. The amygdala resting-state network encompasses regions associated with emotion generation, including ventral medial prefrontal cortex, insula, thalamus and striatum (Fulwiler, et al., 2012; Kim, et al., 2011; Roy, et al., 2009). Thus, resting-state analyses may provide a fuller understanding of neurosteroid modulation of the amygdala and its associated emotion production network. Previous investigations demonstrate that the amygdala is strongly positively correlated with contralateral amygdala and insula at rest, and strongly anticorrelated with precuneus (Kim, et al., 2011; Roy, et al., 2009). Due to their anxiolytic and antidepressant properties, we predicted that allopregnanolone and DHEA would modulate both positive (amygdala and insula) and negative (precuneus) connectivity within the amygdala functional connectivity network during rest (Roy, et al., 2009), and that this effect would be associated with reduced self-reported negative affect. Furthermore, given the impact of allopregnanolone and DHEA on functional connectivity between amygdala and mPFC (Alhaj, et al., 2006; Sripada, et al., 2013a; Sripada, et al., 2013b; van Wingen, et al., 2008), we hypothesized that these neurosteroids would also modulate resting-state amygdala connectivity with mPFC.

Materials and Methods

1.1 Participants

Study participants were 45 right-handed healthy male volunteers aged 18–34 years (mean ± SD = 22 ± 3.6) recruited from the community via advertisement. Exclusion criteria were a history of head injury, recent steroid use, or current or past psychiatric disorder, as assessed via the Mini-International Neuropsychiatric Interview (M.I.N.I.; Sheehan, et al., 1998). No participant was taking over-the-counter or prescription medication at the time of the experiment. Participants were given full details of the study and provided written informed consent. The study was approved by the Institutional Review Board of the University of Michigan Medical School.

1.2 Procedure

To control for any potential diurnal variation in endogenous DHEA (Stanczyk, 2006), all fMRI scans were conducted in the afternoon, and blood collection times were standardized across subjects. Baseline blood samples were drawn between 11:00 and 11:30 AM, and post-scan blood samples were drawn between 3:00 and 4:00 PM. At one hour post-drug administration, participants completed the Positive and Negative Affect Scale Expanded (PANAS-X; Watson and Clark, 1994), the Drug Effects Questionnaire (Morean, et al., 2013), and a 20-item Visual Analogue Scale. The Drug Effects Questionnaire assessed the extent to which participants (1) felt any substance effects, (2) liked the effects, (3) felt high, and (4) wanted more of the substance. The Visual Analogue Scale included the following items: Clear-headed, Anxious, Stimulated, Tired, Calm, Drowsy, Peaceful, Nervous, Hungry, Energetic, Uneasy, Relaxed, Alert, Contented, Focused, Dreamy, Restless, Nauseous, Worn Out, and Jittery. Each item consisted of a 4-inch bar with ‘Not At All’ and ‘Extremely’ indicated at the extremities.

1.3 Drug Administration

Study drugs (pregnenolone and DHEA) and matching placebo identical in appearance were obtained from Belmar Pharmacy (Lakewood, CO), which provided certificates of analysis. Participants were randomly assigned to receive a single oral dose of 400 mg pregnenolone, 400 mg DHEA, or placebo. Participants and investigators were blind to condition. Pregnenolone was administered as a precursor loading strategy to significantly increase downstream allopregnanolone levels. Although data are limited, allopregnanolone serum levels appear to reach three times baseline levels two hours after oral administration of 400 mg pregnenolone (Marx, 2007), and DHEA serum concentrations peak 60 to 480 minutes after DHEA administration (Arlt, et al., 1998). Thus, drug administration occurred two hours before neuroimaging to ensure elevated levels during the scan.

1.4 Steroid measurements

We used circulating levels of DHEA and allopregnanolone as indicators of central neurosteroid levels. In animal models, serum neurosteroid levels are closely related to levels in the hippocampus (Marx, et al., 2006a), and serum DHEA levels are closely related to levels in CSF (Kancheva, et al., 2011), though this study did not find a correlation between CSF and serum pregnenolone levels. CSF DHEA levels are also correlated with temporal cortex DHEA levels, as are CSF and temporal cortex pregnenolone levels (Naylor, et al., 2008). We collected serum samples for assay once prior to drug administration and once after the scanning session. Serum DHEA levels were determined via enzyme immunoassay (ALPCO Diagnostics, Salem, NH), and serum DHEAS levels were determined by chemiluminescent enzyme immunoassay (IMMULITE) according to the manufacturer’s directions (Siemens Healthcare Diagnostics Inc., Tarrytown, NY). Pregnenolone, pregnanolone, allopregnanolone, and androsterone levels in serum were determined by a highly sensitive and specific gas chromatography-mass spectrometry (GC/MS) method, as previously described (Marx, et al., 2006b; Marx, et al., 2006c), with modifications (the electron impact mode was utilized rather than negative ion chemical ionization). One ml of serum was extracted three times in ethyl acetate before high performance liquid chromatography (HPLC) purification using tetrahydrofuran, ethanol, and hexane in the mobile phase. All samples were injected in duplicate. Mean intra-assay coefficients of variation for pregnenolone and allopregnanolone were 0.9% and 2.9%, respectively. The limit of detection with this method was 1 pg. All neurosteroid values were natural log transformed prior to analyses.

1.5 Resting-State Paradigm

Participants underwent structural (sMRI) and functional (fMRI) scanning that included both an emotion regulation task (Sripada, et al., 2013a; Sripada, et al., 2013b) and resting state procedures. Resting-state scans always occurred prior to tasks. Participants were positioned in the MR scanner and their heads comfortably restrained to reduce head movement. Heart rate and respiration measurements were acquired for group comparisons (via Independent-Samples Kruskal-Wallis tests). A black fixation cross on a white background was displayed in the center of the screen for 8 minutes. Participants were instructed to relax and keep their eyes open and fixed on the cross.

1.6 Image Acquisition

MRI scanning occurred on a Philips 3.0 Tesla Achieva X-series MRI (Philips Medical Systems). After a T1 image (T1-overlay) was obtained, a T2*-weighted, echoplanar acquisition sequence [GRE; repetition time, 2000 ms; echo time, 25 ms; flip angle, 90°; field of view (FOV), 22 cm; 42 slice; thickness/skip, 3.0/0 mm matrix size equivalent to 64 × 64] was collected. After discarding three initial volumes to permit thermal equilibration of the MRI signal, 240 volumes were acquired over 8 minutes. After acquiring the functional volumes, a high-resolution T1 scan was obtained for anatomic normalization [26 FOV; thickness/skip, 1.0/0 mm]. Participants viewed a fixation cross through MR-compatible liquid crystal display goggles (NordicNeuroLabs http://www.nordicneurolab.com).

1.7 Preprocessing

A standard series of processing steps was performed using statistical parametric mapping (SPM8; www.fil.ion.ucl.ac.uk/spm). Scans were reconstructed, motion-corrected, slice-time corrected, realigned to the first scan in the experiment to correct for head motion, and co-registered with the high-resolution sagittal images. Normalization was performed using the voxel-based morphometry toolbox implemented in SPM8 (www.fil.ion.ucl.ac.uk/spm). Scans were normalized to standard space, segmented into tissue types, bias-corrected, and iteratively realigned using DARTEL. Smoothing was performed with an 8 × 8 × 8 mm³ kernel. Motion parameters (mean displacement, mean angle) were compared across drug conditions via Independent-Samples Kruskal-Wallis tests, and runs with any movement greater than 3 mm were excluded.

1.8 Data Analysis

Right and left amygdala seed region ROIs were constructed from cytoarchitectonically determined probabilistic maps of the human basolateral and centromedial amygdala, developed by Amunts and colleagues (2005) within the SPM Anatomy Toolbox (Eickhoff, et al., 2005). These subregions were combined to form a single, whole amygdala seed. We extracted the spatially averaged time series from right and left amygdala ROIs for each participant. Since resting-state functional connectivity measures low-frequency spontaneous BOLD oscillations (.01 – .10 Hz band) (Fox, et al., 2005), the time-course for each voxel was band-passed filtered in this range. Next, motion scrubbing and nuisance regression were performed. During motion scrubbing, individual frames with excessive head motion were censored from the time series, following Power et al. (2012). The target high motion frame, but not those flanking it, was removed in accordance with the findings of Satterthwaite et al. (2013). Nuisance regression was performed to remove the effects of nuisance variables unrelated to neuronal activity. Covariates of no interest included six motion regressors generated from the realignment step noted above. In addition, we included five principal components of the BOLD time series extracted from white matter and cerebrospinal fluid masks, which has been demonstrated to effectively remove signals derived from the cardiac and respiratory cycle (Chang and Glover, 2009). This method is comparable to the COMPCOR method and is complementary to the RETROICOR method (Behzadi, et al., 2007). In generating the white matter and cerebrospinal fluid regressors, subject-specific masks were first created using VBM-based segmentation implemented in SPM8. Masks were eroded using FSL-Erode to eliminate border regions of potentially ambiguous tissue-type. A spatially averaged time series was extracted from these masks, and the first five principal components were included in the regression. The residuals from this regression were then retained for further analysis. We did not perform global-signal regression, as it had been suggested that this may produce spurious anti-correlation with orthogonal networks that increase in proportion to the size of the those networks (Anderson, et al., 2011). Pearson product-moment correlation coefficients were calculated between average time courses in the seed regions of interest (ROIs) and all other voxels of the brain resulting in a 3-dimensional correlation coefficient image (r-image). These r-images were then transformed to z-scores using a Fisher r-to-z transformation.

1.9 Whole Brain and Region of Interest (ROI) Analysis

Z-score images from the individual functional connectivity analyses were entered into second-level random-effects analyses (one-sample and two-sample t-tests) implemented in SPM8. At the second level, for between-group comparisons, we set a whole-brain voxel-wise significance threshold at p < .05, cluster-level corrected for multiple comparisons across the entire brain (cluster volume > 1220 voxels); these cluster-level thresholds were calculated using Monte-Carlo simulations (AFNI AlphaSim, http://afni.nimh.nih.gov/afni/doc/manual/AlphaSim). To exclude a general effect of drug administration on the BOLD response (van Wingen, et al., 2007), the primary visual cortex (V1) was used as the control region (Amunts, et al., 2000). Reported voxel coordinates correspond to standardized Montreal Neurologic Institute (MNI) space. To assess for correlations with mood, PANAS-X scores were correlated with data extracted from regions of group difference (averaged across the entire cluster).

Results

2.1 Participants

Sixteen participants were administered pregnenolone, 14 were administered DHEA, and 15 were administered placebo. Groups did not differ by age (F(2,44) = .71, p = .5). Pregnenolone administration significantly increased serum pregnenolone, pregnanolone, and allopregnanolone levels, and DHEA administration increased serum DHEA, DHEAS, and androsterone levels [p < .001; details can be found in (Sripada, et al., 2013a; Sripada, et al., 2013b)]. There were no significant differences in subjective drug effects (p > .4), and participants’ guesses of which drug they received did not deviate from chance (χ² (6) = 3.80, p = .7). There were no significant differences between drug administration and placebo groups in self-reported sedation (p > 0.5).

2.2 Motion and Physiological Variables

There were no movements greater than 3 millimeters and no motion differences between DHEA and placebo groups (p > .2). In the pregnenolone administration group, mean displacement was reduced (mean = .066 mm) as compared to the placebo group (mean = .099 mm, p = .006). To correct for this discrepancy, we implemented motion scrubbing procedures. There were no group differences in heart rate or respiration (p > .7).

2.3 fMRI Findings

Under the placebo condition, the left amygdala seed showed positive connectivity with right amygdala, peri-amygdala, bilateral hippocampus and superior temporal gyrus, and anti-correlations with posterior cingulate cortex (see Table 1). The right amygdala seed showed positive connectivity with left amygdala, bilateral peri-amygdala and bilateral hippocampus, and anti-correlation with precuneus (see Table 1).

Table 1.

Contrast Map and Brain Region	Cluster Size	MNI Coordinates (x y z)	Analysis (z)
Left Amygdala positive correlations under PBO
Superior Temporal Gyrus/Left Insula	8153	−27 −7 −29	7.28
Medial Frontal Gyrus	59	−6 41 −23	4.45
Right Lingual Gyrus	13	15 −40 −2	4.29
Right Heschl’s Gyrus	16	39 −25 7	3.92
Inferior Frontal Gyrus	178	−48 11 28	3.83
Right Insula	14	30 −7 10	3.71
Left Angular Gyrus	51	−57 −70 25	3.48
Left Amygdala negative correlations under PBO
Right Middle Frontal Gyrus	52	36 47 13	4.12
Posterior Cingulate Cortex	13	6 −43 22	3.61
Right Amygdala positive correlations under PBO
Right Parahippocampal Gyrus	1935	24 −10 −23	7.66
Superior Temporal Gyrus	1766	−33 −4 −29	6.31
Right Inferior Temporal Gyrus	13	48 −49 −26	5.37
Left Inferior Temporal Gyrus	436	−54 −67 −5	4.71
Right Precentral Gyrus	1015	60 −4 46	4.09
Left Precentral Gyrus	19	−48 −7 25	3.99
Inferior Parietal Lobule	23	−48 −34 49	3.84
Supplementary Motor Area	58	12 −22 52	3.79
Right Paracentral Lobule	14	6 −40 67	3.64
Left Paracentral Lobule	35	−9 −31 73	3.61
Right Amygdala negative correlations under PBO
Cingulate Gyrus	71	−3 −19 25	4.27
Precuneus	111	12 −64 31	3.98
Left Precuneus	12	−9 −70 34	3.38

Data thresholded at p < 0.001, uncorrected, extent threshold k = 10. MNI, Montreal Neurologic Institute; PBO, placebo.

Compared to placebo, pregnenolone administration reduced connectivity between the left amygdala seed and a cluster encompassing dmPFC and precuneus (see Table 2; Figure 1). Pregnenolone also reduced connectivity between the right amygdala seed and left hippocampus (see Figure 1).

Table 2.

Contrast Map and Brain Region	Cluster Size	MNI Coordinates (x y z)	Analysis (z)
Left Amygdala ALLO > Placebo
Left Superior Frontal Gyrus	2687	−9 −28 76	3.84
Superior Medial Frontal Gyrus (dmPFC)	196	−12 41 37	2.82
Precuneus	199	−9 −40 55	2.86
Right Amygdala ALLO > Placebo
Left Middle Temporal Gyrus	1522	−45 −19 −11	3.91
Left Hippocampus	18	−36 −16 −11	2.84
Left Amygdala DHEA > Placebo
Left Supramarginal Gyrus	2038	−66 2 7	3.62
Left Insula	165	−36 −1 −8	2.81
Left Amygdala	8	−24 −7 −14	2.40
Left Amygdala DHEA > ALLO
Medial Frontal Gyrus	2049	−15 50 10	3.88
dmPFC	396	−12 53 10	3.58

Cluster-level significance set at p < .05, whole brain corrected for multiple comparisons. Within each significant cluster, Z-score and associated a priori region are noted along with MNI atlas coordinates of peaks, in bold. ALLO, allopregnanolone; dmPFC, dorsal medial prefrontal cortex; MNI, Montreal Neurologic Institute.

Figure 1.

(A) Pregnenolone administration reduced functional connectivity between left amygdala and a cluster encompassing dmPFC and precuneus (x=−11). (B) Pregnenolone administration reduced functional connectivity between right amygdala and left hippocampus (y=−16). Bar graphs depict percent signal change. Error bars depict standard error of the mean. dmPFC = dorsal medial prefrontal cortex. HPC = hippocampus.

Compared to placebo, DHEA reduced connectivity between the left amygdala seed and a cluster encompassing left amygdala, left peri-amygdala, and left insula (see Table 2; Figure 2). There were no significant differences between DHEA and placebo groups in connectivity with a right amygdala seed.

Figure 2.

DHEA reduced functional connectivity between left amygdala and (A) peri-amygdala (y=−7) and (B) left insula (y=4). Bar graphs depict percent signal change. Error bars depict standard error of the mean.

We also conducted an exploratory analysis of differences in amygdala functional connectivity between pregnenolone and DHEA administration groups. Compared to DHEA, pregnenolone reduced functional connectivity between left amygdala and dmPFC (see Table 2).

2.4 Correlations with PANAS-X

As compared to placebo, DHEA administration reduced PANAS-X negative affect score at trend level (t(27) = 1.87, p = .07) (as reported in Sripada, et al., 2013a). To probe the neural underpinnings of this effect, correlations were computed between PANAS-X negative affect score and the values extracted from regions of group difference. These included: (1) the dmPFC and precuneus-encompassing cluster that exhibited reduced connectivity with left amygdala in the pregnenolone group, (2) the left hippocampus-encompassing cluster that exhibited reduced connectivity with right amygdala in the pregnenolone group, and (3) the left insula-encompassing cluster that exhibited reduced connectivity with left amygdala in the DHEA group. Across all three groups (pregnenolone, DHEA, and placebo), there was a positive correlation between PANAS-X negative affect score and functional connectivity between right amygdala and the hippocampus-encompassing cluster (r = .294, p = .05), indicating that reduced connectivity with hippocampus was associated with less negative affect. Across groups, there was also a positive correlation between PANAS-X negative affect score and functional connectivity between left amygdala and the insula-encompassing cluster (r = .291, p = .05). Additional correlations were computed between PANAS-X negative affect score and the values extracted from anatomical insula, dmPFC, hippocampus, and precuneus. Across all three groups (pregnenolone, DHEA, and placebo), there was a positive correlation between PANAS-X negative affect score and left amygdala functional connectivity with anatomical precuneus (r = .319, p = .035).

2.5 Correlations with Neurosteroid Levels

To further probe relationships between neurosteroids and connectivity patterns, correlations were computed between delta serum neurosteroid levels (endpoint minus baseline) and the values extracted from regions of group difference. Across all three groups, functional connectivity between the right amygdala seed and left hippocampus-encompassing cluster was negatively associated with delta serum pregnenolone (r = −.513, p < .001), delta serum pregnanolone (r = −.326, p = .031), and delta serum allopregnanolone (r = −.553, p < .001). Across all three groups, functional connectivity between the left amygdala seed and left insula-encompassing cluster was negatively associated with delta serum DHEAS (r = −.436, p = .003), and delta serum androsterone (r = −.390, p = .009). Additionally, functional connectivity between left amygdala and anatomical left insula was negatively correlated with delta serum DHEAS (r = −.302, p = .047), and functional connectivity between right amygdala and anatomical precuneus was negatively correlated with delta serum pregnanolone (r = −.301, p = .047). Neither pregnenolone nor DHEA significantly modulated connectivity between amygdala and primary visual cortex (left amygdala p > .4; right amygdala p > .7), indicating that drug effects on other brain regions were not attributable to general changes in the BOLD response.

Discussion

To interrogate the neural underpinnings of anxiety- and mood-altering properties of neurosteroids, we investigated the impact of single-dose pregnenolone and DHEA on patterns of resting-state functional connectivity of the amygdala. Compared to placebo, pregnenolone administration reduced connectivity between amygdala and dmPFC, between amygdala and precuneus, and between amygdala and hippocampus. Compared to placebo, DHEA reduced amygdala connectivity with insula and with peri-amygdaloid regions. These connectivity patterns were associated with less self-reported negative affect. To our knowledge, this is the first investigation of resting-state functional connectivity after neurosteroid administration, and the first to demonstrate that neurosteroids modulate intrinsic connectivity networks in a way that is relevant to negative affect. Our results demonstrate that allopregnanolone and DHEA reduce connectivity within the resting-state amygdala network, and may also shift the balance between intrinsic connectivity networks, a function that could provide insight into the neurocircuitry of anxiety psychopathology.

We found that DHEA reduced functional connectivity within left amygdala and between left amygdala and left peri-amygdala, suggesting reduced connectivity within this emotion generation network. Converging evidence from multiple lines of investigation suggests that higher anxiety is associated with greater amygdala activity. Greater amygdala activation is associated with greater state (Bishop, et al., 2004) and trait (Etkin, et al., 2004) anxiety, and a metaanalysis of neuroimaging studies in various anxiety disorders concluded that elevated amygdala activity is present across anxiety disorders (Etkin and Wager, 2007). Furthermore, both animal and human connectivity data suggest that amygdala and extended amygdala function as a coordinated unit to process threat (Liberzon, et al., 2003; Oler, et al., 2012). It is still unclear, however, if greater interamygdala connectivity is associated with higher levels of anxiety. For example, Pantzatos and colleagues (2012) have argued that right and left amygdala are more coupled while viewing neutral faces than fearful faces. However, given the strong link between amygdala activation and negative emotional response, it is reasonable to interpret reduced interamygdala connectivity as potentially contributing to reduced emotional reactivity. This reduction of connectivity within amygdala and between amygdala and peri-amygdala may thus be relevant DHEA’s anxiolytic effects (Maninger, et al., 2009; Melchior and Ritzmann, 1994).

DHEA administration also reduced functional connectivity between left amygdala and left insula. The insula is implicated in interoception (Oppenheimer, et al., 1992), disgust (Britton, et al., 2006), emotion processing (Stein, et al., 2007), emotional recall (Phan, et al., 2002), and anticipation of aversive stimuli (Simmons, et al., 2008). Amygdala and insula are structurally interconnected (Aggleton, et al., 1980), and negative emotion induction enhances coupling between insula and amygdala in healthy volunteers (van Marle, et al., 2010). We have previously reported that amygdala exhibits greater resting-state connectivity with insula in PTSD patients as compared to healthy and combat-exposed control participants (Sripada, et al., 2012a; Sripada, et al., 2012b). This finding has been replicated by others (Rabinak, et al., 2011). Thus, reduced connectivity between amygdala and insula could represent a protective/resilience factor in the development of pathological anxiety states. In support of this hypothesis, functional connectivity between amygdala and insula was positively associated with negative affect in our sample. DHEA reduced connectivity between amygdala and insula, and also reduced PANAS-X at trend level. However, given the lack of a statistically significant decrease in PANAS-X scores, it is possible that the association between negative affect and amygdala-insula connectivity may be due to spontaneous variations in mood rather than a drug-induced effect.

During rest, the amygdala shows positive functional connectivity with ventral medial prefrontal regions, insula, thalamus, and striatum, and anti-correlations with superior frontal gyrus, bilateral middle frontal gyrus, posterior cingulate cortex, and precuneus (Fulwiler, et al., 2012; Kim, et al., 2011; Roy, et al., 2009). Notably, the observed amygdala resting-state connectivity map shows substantial overlap with the salience network, an intrinsic connectivity network responsible for detecting and orienting to salient stimuli, implicated in homeostatic regulation, interoceptive, autonomic, and reward processing (Cauda, et al., 2011; Dosenbach, et al., 2007; Seeley, et al., 2007; Sridharan, et al., 2008). Amygdala, along with dorsal anterior cingulate cortex, anterior insula/inferior frontal gyrus, and ventral striatum, are key nodes in this network. Both allopregnanolone and DHEA have demonstrated effects on this circuit in neuroimaging studies. DHEA’s reduction of amygdala to insula connectivity and interamygdala connectivity suggests this neurosteroid may modulate the intrinsic connectivity of the salience network. During task-based studies, there is preliminary evidence to suggest that DHEA modulates anterior cingulate cortex (Alhaj, et al., 2006; Sripada, et al., 2013a) and insula activity (Alhaj, et al., 2006), and that allopregnanolone influences activity in amygdala (Ossewaarde, et al., 2010a; van Wingen, et al., 2007; van Wingen, et al., 2008) and ventral striatum (Ossewaarde, et al., 2010b), all components of the salience network. We have previously suggested that stronger within-salience network connectivity is associated with greater anxiety (Sripada, et al., 2012a; Sripada, et al., 2012b). In addition, both PTSD and major depressive disorder were found to be associated with greater salience network activity in recent meta-analyses (Hamilton, et al., 2012; Patel, et al., 2012). Since elevations in neurosteroids have been associated with symptomatic improvement in anxiety disorders (Rasmusson, et al., 2006), the observed reduction of intrasalience network connectivity could suggest a potential mechanism for this anxiolytic effect.

Finally, pregnenolone reduced amygdala connectivity with dmPFC, precuneus, and hippocampus. Across groups, greater amygdala to precuneus connectivity was associated with greater PANAS-X negative affect [ten items assessing constructs of fear, hostility, guilt, and sadness (Watson and Clark, 1994)], suggesting that a reduction in this connectivity is associated with less negative affect. In monkeys, the amygdala is structurally connected with dorsal precuneus (Leichnetz, 2001), ventral precuneus (Parvizi, et al., 2006), and dmPFC/dorsal anterior cingulate cortex (Ghashghaei, et al., 2007) and typically shows negative functional connectivity with the precuneus and dmPFC at rest (Roy, et al., 2009; Zhang and Li, 2012). This is consistent with the key roles of the dmPFC and precuneus in the default network, a network that is anti-correlated with salience network at rest (Fox, et al., 2005). The default network is associated with stimulus-independent, internally-focused thought, including spontaneous cognition, autobiographical memory, prospection and mind-wandering (Menon, 2011; Qin and Northoff, 2011; Spreng, et al., 2009; Toro, et al., 2008), thus this network is active during non-structured tasks such as resting state. The current data suggest that allopregnanolone might further reduce functional connectivity between amygdala and default network, contributing to enhanced segregation between default network and salience network.

Recently, our group has proposed that decreased segregation, i.e. inappropriate connectivity between salience network and default network, contributes to anxiety psychopathology (Sripada, et al., 2012b). Among participants with PTSD, we found that greater functional connectivity between default network and salience network regions was correlated with elevated PTSD symptoms (Sripada, et al., 2012b). Similarly, enhanced cross-network connectivity between default network and thalamus (a node in salience network) has been observed in PTSD in a different study (Yin, et al., 2011). Furthermore, enhanced connectivity between default network and amygdala predicted the development of PTSD symptoms in acutely traumatized individuals (Lanius, et al., 2010), and enhanced connectivity between default network and insula was associated with higher self-reported anxiety in a separate study (Dennis, et al., 2011). Precuneus exhibits greater connectivity with amygdala in generalized anxiety (Strawn, et al., 2012) and panic disorder (Pannekoek, et al., 2013a), and with anterior cingulate cortex (a key node in salience network) in social anxiety disorder (Pannekoek, et al., 2013b). Moreover, in healthy controls, acute laboratory stress increases amygdala resting-state connectivity with precuneus and posterior cingulate cortex (Veer, et al., 2011), though the authors note that this connectivity pattern might indicate healthy rather than pathologic recovery and adaptation in the post-stressor period. Another study in healthy individuals demonstrated that viewing fearful as compared to neutral face enhanced amygdala to precuneus coupling (Pantazatos, et al., 2012). Together, these data evince a consistent pattern in which anxiety disorders or induction of anxiety is associated with reduced segregation between salience network and default network. Thus, allopregnanolone’s enhancement of segregation between default network and salience network highlights the potential contribution of this neurosteroid to the amelioration of anxiety.

Though both pregnenolone and DHEA reduced amygdala intrinsic connectivity, pregnenolone exerted a greater effect than DHEA on connectivity with medial prefrontal regions. These differences may be due to differing mechanisms of action. DHEA appears to have anxiolytic-like actions (Maninger, et al., 2009; Melchior and Ritzmann, 1994), however it demonstrates modest negative modulation of GABA(A) receptors (Imamura and Prasad, 1998). Thus, its anxiolytic-like actions in rodents may involve other mechanisms. In contrast, allopregnanolone has very pronounced effects at the GABA(A) receptor, and decreases GABA(A) receptor responses with 20-fold higher potency than benzodiazepines and barbiturates (Majewska, et al., 1986). Allopregnanolone’s greater GABAergic activity, in combination with the greater wealth of evidence linking allopregnanolone to amygdala modulation, may help to explain the somewhat stronger modulation of amygdala connectivity under allopregnanolone versus DHEA.

There are several limitations to this study. Limitations of our sample include the fact that the number of participants in each group was relatively small, thus, our results should be considered preliminary. Additionally, our sample consisted of healthy male individuals without mood or anxiety disorder diagnoses. Thus, extrapolations to women or to clinical populations should be made with caution. The between-group design of our study does not allow for a baseline comparison of functional connectivity. Thus, it is difficult to determine whether between-group differences are solely due to the drugs administered, or partially due to natural variation. Future studies should implement a crossover design to address this issue. Additionally, we did not obtain baseline measurements of subjective negative affect, and therefore group differences in PANAS-X score may also be partially attributable to natural variation. Limitations of our intervention include the fact that we measured serum levels of allopregnanolone and DHEA, and not CSF or brain levels. However, in animals, neurosteroid levels are highly correlated (Kancheva, et al., 2011; Marx, et al., 2006a). Additionally, in the allopregnanolone manipulation, we administered pregnenolone, allopregnanolone’s precursor. Though we have framed our results in terms of an allopregnanolone manipulation, it is possible that our results may be attributable to increases in pregnenolone in addition to allopregnanolone. Since allopregnanolone is not commercially available for clinical use, it was necessary to administer pregnenolone as a precursor loading strategy to increase downstream allopregnanolone levels, and our results demonstrate that oral administration of pregnenolone increases allopregnanolone levels sevenfold (see Sripada, et al., 2013b). We also did not measure levels of additional downstream steroids. Pregnenolone administration may increase levels of pregnenolone sulfate and progesterone (Marx, et al., 2009), and DHEA administration may increase estradiol and testosterone (Schmidt, et al., 2005). Thus, it is possible that increases in these downstream steroids, rather than increases in only allopregnanolone and DHEA, contributed to the reported findings. Furthermore, though the bulk of DHEA research supports an anxiolytic effect for DHEA, some studies suggest an anxiogenic role (e.g., Dmitrieva, et al., 2001; van Goozen, et al., 2000). Thus, further studies should assess the behavioral impact of DHEA’s reduction of amygdala resting-state functional connectivity.

Turning to methodological limitations, our results may have been impacted by the reduced average head displacement of .03 mm in the pregnenolone group. For instance, Van Dijk and colleagues have suggested that small head movements can inflate measures of local functional coupling and deflate measure of long-range functional coupling (Van Dijk, et al., 2012). However, we implemented scrubbing procedures as described by Power et al. (2012), which have been demonstrated to significantly reduce confounds introduced by head movement. In addition, it has been reported that DHEA and pregnenolone may potentially impact neurovascular coupling (e.g., see Liu and Dillon, 2004; Naylor, et al., 2010). PET or arterial spin labeling studies could examine this issue directly, however we compared BOLD signal across treatment groups in control regions (i.e. visual cortex) and found no significant differences. Another potential limitation is that different methodologies were used for assaying different neurosteroids: GC/MS for pregnenolone and allopregnanolone, and enzyme immunoassay for DHEA and DHEAS. DHEA values derived from these two assay methods, however, show excellent correlation (Stanczyk, 2006), and immunoassay is much less labor-intensive and less expensive than GC/MS. In contrast, there are currently no commercially available methodologies for allopregnanolone and pregnenolone that approach the sensitivity and specificity of GC/MS, and it remains the ‘gold standard’ quantification approach for these steroids. It also does not appear optimal to use GC/MS for sulfated neurosteroids such as DHEAS, given recent challenges and controversies regarding their quantification (see Ebner, et al., 2006; Higashi, et al., 2003; Liere, et al., 2004; Liu, et al., 2003; Marx, et al., 2009; Schumacher, et al., 2008). Finally, we have conceptualized decreased within-salience network connectivity as relevant to reduced anxiety; however decreased connectivity within this network may also have drawbacks. For instance, stronger salience network connectivity has been associated with better working memory (Li, et al., 2012). Future studies should investigate neurosteroid modulation of salience network in anxiety-disordered populations to further assess the benefits and disadvantages of these patterns.

Conclusion

In summary, allopregnanolone and DHEA reduced amygdala functional connectivity with other regions within amygdala/salience network, and reduced amygdala connectivity with regions of default network in our participants. These findings suggest that neurosteroids modulate amygdala functional connectivity during the resting state, and may shift the balance between salience network and default network at rest. These findings provide insight into the neurocircuitry of anxiety, and suggest that allopregnanolone and DHEA may be potential targets for pharmacological intervention for anxiety psychopathology.