Resting-state neural signatures of depressive symptoms in acute HIV

Abstract

Depressive symptoms are often elevated in acute and chronic HIV. Previous neuroimaging research identifies abnormalities in emotion-related brain regions in depression without HIV, including the anterior cingulate cortex (ACC) and amygdala. However, no studies have examined the neural signatures of depressive symptoms in acute HIV infection (AHI). Seed-based voxelwise resting-state functional connectivity (rsFC) for affective seed regions of interest (pregenual ACC, subgenual ACC [sgACC], bilateral amygdala) was computed for 74 Thai males with AHI and 30 Thai HIV-uninfected controls. Group analyses compared rsFC of ACC and amygdala seed regions between AHI and uninfected control groups. Within the AHI group, voxelwise regression analyses investigated the relationship between depressive symptoms and rsFC for these affective seed regions. Group analyses revealed alterations in rsFC of the amygdala in AHI versus uninfected controls. Depressive symptoms associated with decreased rsFC between ACC regions and posterior cingulate/precuneus, medial temporal and lateral parietal regions in AHI. Symptoms of depression also correlated to increased rsFC between ACC regions and lateral prefrontal cortex, sgACC, and cerebellum in AHI. Similar to the ACC, depressive symptoms associated with decreased rsFC between amygdala and precuneus. Of blood biomarkers, only HIV RNA inversely correlated with rsFC between posterior sgACC and left uncus. We found that depressive symptoms in AHI associate with altered rsFC of ACC and amygdala regions previously implicated in depression. Longitudinal research in this cohort will be necessary to determine whether these early alterations in rsFC of affective network regions are related to persistent depressive symptoms after combination antiretroviral therapy.

Introduction

Research in acute HIV infection (AHI)—the time period within days to weeks after transmission—indicates that elevated concentrations of the virus (i.e., plasma HIV RNA) can be detected in blood and semen (Pilcher et al. 2004a, b; Stekler et al. 2008) as well as in the central nervous system (CNS) within approximately a week after transmission (Valcour et al. 2012). In addition to these viral and neurobiological changes, psychiatric symptoms are often elevated in individuals with AHI and early HIV infection (Kelly et al. 1998; Atkinson et al. 2009; Moore et al. 2011; Weber et al. 2013; Gold et al. 2014; Hellmuth et al. 2017). For instance, clinically significant depressive symptoms have been reported to occur at rates between 48-55% of people with AHI and early HIV infection (Atkinson et al. 2009; Moore et al. 2011; Gold et al. 2014; Hellmuth et al. 2017), which is higher than the general population prevalence rates of 5-16% (Kessler et al. 2003; Steel et al. 2014).

Elevated rates of depressive symptoms in HIV infection may be due to premorbid factors (e.g., gender, socioeconomic status, or co-morbid conditions), a response to the HIV diagnosis, or could be caused or worsened by other viral or immunological factors. Longitudinal studies in chronic HIV infection have further shown that depression, one of the most common psychiatric disorders in this population (Bing et al. 2001; Rabkin 2008; Arseniou et al. 2013), predicts worse health outcomes, such as greater declines in CD4 cell count, more rapid disease progression, and greater risk for mortality (Patterson et al. 1996; Ickovics et al. 2001; Leserman 2008).

A review of these findings indicates that the relationship between chronic depression and/or depressive symptoms and poorer health outcomes in chronic HIV infection appears in studies conducted both before and after the widespread availability of combination antiretroviral therapies (cART) (Leserman 2008; Arseniou et al. 2013). In addition, depressive symptoms at clinical and subclinical levels in individuals with chronic HIV have been consistently associated with poorer medication adherence (Gonzalez et al. 2011; Uthman et al. 2014), which may explain some of the negative health outcomes reported in longitudinal studies.

A recent study from members of our group reported an association between clinically significant depressive symptoms, higher plasma HIV RNA, lower CD4 cell count, and higher plasma neopterin in individuals with AHI before ART initiation (Hellmuth et al. 2017). Given the prevalence and putative impact of depression in HIV infection, an improved understanding of the early biological and neurological correlates of symptoms of depression in AHI prior to long-term ART is crucial. To our knowledge, however, no studies have yet investigated whether depressive symptoms in AHI are also associated with alterations in neurobiological functioning.

Resting-state functional connectivity (rsFC) is a neuroimaging technique that is sensitive to dynamic fluctuations in cognition and mood. rsFC measures correlations in spontaneous low frequency fluctuations in the blood oxygenation level-dependent (BOLD) response while participants are at rest (i.e., not engaged in a task; Fox and Raichle 2007 for review). Numerous rsFC studies describe large-scale functional network abnormalities across cortical and subcortical brain systems in psychiatric conditions, including depression (Greicius 2008; Menon 2011; Kaiser et al. 2015; Williams 2016).

One neural circuit that has been frequently implicated in depression is the affective network, which includes the anterior cingulate cortex (ACC) and amygdala (Mayberg et al. 1999; Mayberg 2003; Drevets et al. 2008; Sheline et al. 2010; Davey et al. 2012; Kaiser et al. 2015; Williams 2016). Previous neuroimaging research in chronic and early HIV infection documents abnormal structure and function of the same bilateral amygdala and ACC regions within the affective network (Ances et al. 2012; Thomas et al. 2013; Behrman-Lay et al. 2016; Spies et al. 2016; Clark et al. 2017; Thames et al. 2018). However, research directly investigating the relationship between symptoms of depression and rsFC of these limbic brain regions has been more limited in HIV-infected populations, with only one known study. Specifically, McIntosh and colleagues (2018) found that reduced rsFC between the posterior sgACC and right and left amygdala predicted greater depressive symptoms in participants with chronic HIV infection who were on long-term ART. Together with previous research in non-HIV populations with depression, these findings suggest that depressive symptoms in HIV may produce similar disruptions in rsFC of ACC and amygdala regions found in uninfected cohorts. However, studies to date have not evaluated the relationship between symptoms of depression and rsFC of these affective regions in individuals during the earliest stages of disease.

The purpose of the current study was to identify the resting-state neural signatures of depressive symptoms in 74 Thai males with AHI. We also compared rsFC of emotion-related brain regions in participants with AHI to 30 Thai HIV-uninfected controls (CO), with all participants undergoing resting-state functional magnetic resonance imaging on the same scanner. To examine rsFC across the whole brain, we performed seed-based voxelwise analyses for ACC and amygdala regions. We hypothesized that symptoms of depression in AHI would be associated with similar patterns of rsFC of ACC and amygdala regions reported in depression in non-HIV samples. Given that depression in AHI and chronic HIV infection has been associated with abnormalities in biological markers of HIV infection (e.g., CD4 cell count), we also investigated whether rsFC signatures of depressive symptoms would be associated with HIV-related biomarkers in AHI.

Methods

Participants

Participants seeking HIV testing who met criteria for AHI were enrolled at the Anonymous clinic of the Thai Red Cross AIDS Research center in Bangkok, Thailand, as previously described (De Souza et al. 2015). All participants in the present study underwent imaging, laboratory assessments, and clinical follow-up as outlined in a broader protocol designed to investigate the earliest biological dynamics of acute infection and response to treatment (SEARCH 010/RV 254, ClinicalTrials.gov identifier NCT00796146; De Souza et al. 2015). Note, imaging was optional for participants in this cohort. Only baseline data were included in the present study.

Participant inclusion criteria were the following: confirmed AHI in Fiebig stage I-V (Fiebig et al. 2003), age ≥18 years, and imaging acquired on 3.0 Tesla MRI scanner. Individuals were excluded for the following reasons: presence of pre-HIV psychiatric disorders (e.g., schizophrenia) or past or current substance use disorder defined by clinical exam using DSM-5 criteria, confounding neurological conditions (e.g., head injury with loss of consciousness), history of opportunistic CNS infections (e.g., toxoplasmosis), or contraindications for MRI (e.g., pregnancy, claustrophobia, metals). None of the participants in this study had co-infection with hepatitis or syphilis (positive serology for serum VDRL) or were taking medications for psychiatric conditions at diagnosis of AHI. Of the 94 AHI participants, full imaging data were available for 74, after excluding participants due to positive serum syphilis serology (n = 10) and excessive motion during the resting-state scan (n = 10; as described below). MRI for the AHI group was performed before combination ART initiation in the majority of participants (n = 61), and a mean (SD) of 1.1(0.4) days after combination ART initiation in the remaining participants (n = 13). The breakdown by Fiebig stage was as follows: Fiebig I (HIV RNA+, p24 antigen−, HIV IgM−, n = 7), stage II (HIV RNA+, p24 antigen+, HIV IgM−, n = 15), stage III (HIV IgM+, Western Blot −, n = 43), stage IV (HIV IgM+, Western Blot indeterminate, n = 5), stage V (Western Blot+ without p31 band, n = 4).

A comparison group was included, comprised of demographically similar, HIV uninfected controls (n = 30) enrolled from concurrent HIV protocols (RV304/SEARCH 013: NCT01397669), after excluding participants due to excessive motion (n = 4; as described below). Note, some of the participants in the present study (49 AHI, 23 controls) overlap with those in a recently published study examining HIV effects (Samboju et al. 2018). All control participants were scanned on the same MR system with identical scanning sequences and parameters. MRI scans for both control and AHI participants were collected in the morning between 7am and 9am for most participants (n = 91), while remaining participants were scanned in the afternoon (n = 10) or evening (n = 3). Prior to the scan, all participants completed an MRI screening form to determine eligibility for scanning (e.g., metal implants, claustrophobia, pregnancy). See Table 1 for demographic and clinical information for all AHI and control participants.

Table 1.

Participant demographic, depressive symptom, and clinical characteristics

	AHI (n = 74)*	Uninfected controls (n = 30)*	P-values
Demographic Information
Agea	28.9 (8.9)	30.2 (5.4)	p = .461
Sex (% Male)a	100%	66.7%b	p < .0005
Education levela			p = .045
No certificate/primary school certificate	2.7%	0%
Less than high school certificate/vocational certificate	5.4%	23.3%
High school certificate or higher vocational or diploma	35.1%	33.3%
Bachelor degree or higher	56.8%	43.3%
Motion Summary Measure
Average RMS	28.5 (14.3)	27.6 (9.9)	p = .754
Depressive Symptom Measure
HADS-Dc	5.2 (4.2)	3.2 (1.9)	p = .005
HIV-Related Clinical Variables
Time since infection (days)	19.0 (6.6)	—	—
Fiebig stage I, n (%)	7 (9.5%)	—	—
Fiebig stage II, n (%)	15 (20.3%)	—	—
Fiebig stage III, n (%)	43 (58.1%)	—	—
Fiebig stage IV, n (%)	5 (6.8%)	—	—
Fiebig stage V, n (%)	4 (5.7%)	—	—
CD4	394.8 (209.4)	—	—
CD8	680.9 (507.0)	—	—
CD4:CD8 ratio	0.8 (0.5)	—	—
HIV RNA (log 10 copies)	6.0 (1.1)	—	—

Notes. AHI = Acute HIV; RMS = relative root mean squared displacement; HADS-D = Hospital Anxiety and Depression Scale-depression subscale score.

^aSex and education level were significantly different between groups (chi-square tests). There were no significant group differences in age (independent samples t-test).

^bFor control group, 26.7% female (8), 66.7% male (20), and 6.7% transgender (2)

^cBased on data for n = 64 AHI subjects and n = 20 for uninfected control participants*; total HADS-D scores ranged from 0 to 18. There were significant group differences in depressive symptoms (independent samples t-test).

This study was approved by institutional review boards from all participating sites and all participants provided written and informed consent.

Depression Symptom Measure

Within the AHI group, we administered the English or validated Thai version of the Hospital Anxiety and Depression Scale (HADS), a 14-item questionnaire used to assess current depression and anxiety symptoms (i.e., over the past week) (Nilchaikovit 1996; Nüesch et al. 2009). Language proficiency in English or Thai was evaluated during the consent process and was used to determine the HADS version to use. Only depression subscale scores (HADS-D) were used in the present study. HADS-D scores range from 0 to 21 and were calculated for AHI participants with complete HADS data (n = 64; complete HADS data missing for n = 10 AHI participants; HADS-D, Cronbach’s α = .86). The HADS was administered prior to other testing in the AHI group, including lumbar puncture, to minimize the impact of procedures on these assessments. Note, HADS data were not collected for a subset of the control participants (n = 10) and individual item scores were unavailable, so reliability could not be calculated for the uninfected control group.

HIV-Related Clinical Variables

Quantification of plasma HIV RNA followed standard measures as in previous work (De Souza et al. 2015). Corresponding blood HIV RNA, CD4 cell counts, and CD8 cell counts were measured the day of enrollment into the study and within a mean (SD) 1.6 (±1.7) days of MRI acquisition. Baseline cerebrospinal fluid (CSF) HIV RNA was only available for 12 participants and thus was not considered for analysis. Estimated duration of infection was calculated using a median of participants’ self-reported exposure date(s), as previously described (Valcour et al. 2012).

MRI Data Acquisition

All structural and functional MRI data were acquired using the same Philips Ingenia 3T MRI scanner equipped with a 15-channel volume head coil for signal excitation and reception. High-resolution T1-weighted structural images were acquired using a standard T1-weighted imaging turbo field echo (T1W 3D TFE) scan (TR/TE/flip angle (FA): 8.1ms/3.7ms/8°, matrix:256x256x165, field of view (FOV): 256mm x 256mm, voxel size: 1mm x 1mm x 1mm, number of slices: 165, no gap). Resting-state functional images were acquired while participants eyes were closed using T2*-weighted Echo Planar Imaging (EPI) sequence (TR/TE/FA: 1600ms/22ms/70°, matrix size: 128x128, FOV: 224mm x 224mm slice thickness: 1.75mm x 1.75mm x 3.5mm, number of slices: 9694 axial slices). The resting-state scan time was about 7 minutes (262 volumes).

Preprocessing and Motion Analysis for Resting-State fMRI Data

The resting-state functional data were processed in AFNI, FSL, and ANTs (Cox 1996; FMRIB Software Library; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/; http://stnava.github.io/ANTs/). First, we preprocessed data by applying the following steps: (1) images were deobliqued (3dWarp), (2) first three volumes were omitted (3dcalc), (3) motion corrected by rigid body alignment to the first EPI acquisition (3dvolreg), (4) the 3D+time series were despiked to remove time series outliers (3dDespike), and (5) temporally filtered (band-pass: 0.009 Hz < f < 0.08 Hz; 3dTproject) and spatially smoothed with a 3D 4-mm full-width half-maximum (FWHM) Gaussian kernel (3dmerge). Next, the skull-stripped anatomical scan for each participant was rigidly coregistered first with the T1, then with the EPI, and diffeomorphically aligned to Montreal Neurological Institute (MNI)-152 template space using a symmetric normalization algorithm in ANTs (Avants and Gee 2004). Normalized T1 anatomical images (i.e., aligned to the EPI and in original space) were also segmented into gray matter, white matter, and CSF using FAST in FSL (Zhang et al. 2001). White matter and CSF segments were used as masks to extract a representative time series from each tissue type. A whole-brain mask was used to extract the average time series for the global signal.

Final preprocessing steps were performed in a GLM (3dDeconvolve) to account for motion and other typical nuisance variables (as in Ciric et al. 2017): (1-12) six motion parameters (three translations, three rotations) obtained from the rigid body alignment of EPI volumes and their six derivatives, (13-14) the white matter time series and its derivative, (15-16) the ventricular (CSF) time series and its derivative, and (17-18) a whole-brain or global signal time series and its derivative. To further control for individual motion within the GLM, volumes were censored for excessive motion, as described in the following paragraph. These final preprocessed resting-state data were used in the functional connectivity analysis detailed below.

Although global signal regression can be beneficial in preprocessing to remove physiological noise from resting-state data when cardiac and respiratory measures are not collected (Birn et al. 2006), this step has also been shown to introduce anti-correlations (Murphy et al. 2009). Therefore, we also performed the same functional connectivity analysis using final preprocessed resting-state files without global signal regression and report the results from both methods.

Excessive motion was assessed using the following criteria (as in Samboju et al. 2018): maximum framewise motion displacement > 3 mm, and/or total scan time < 3 min after censoring all time points with framewise motion displacement > .2 mm and extreme timeseries displacement. Fourteen participants (10 AHI; 4 control) in the present study were excluded based on these criteria. We selected these motion thresholds based on recommendations from previous research (Power et al. 2012; Yan et al. 2013). Average relative root-mean-squared (RMS) displacement was used as a summary measure of individual participant motion (as in Ciric et al. 2017).

Functional Connectivity Analysis

Building on prior research identifying ACC and amygdala dysfunction in depression, we performed seed-based voxelwise functional connectivity analyses for the following affective seed regions of interest (ROIs): pregenual ACC (pgACC: 1, 40, 16; Andrews-Hanna et al. 2007), anterior subgenual ACC (anterior sgACC: 0, 38, −9; Berman et al. 2011), posterior subgenual ACC (posterior sgACC: 1, 25, −9; Mayberg et al. 1999), and right and left amygdala (as in Motzkin et al. 2015). ACC seed ROIs were generated using 3dcalc in AFNI, to create a 6-mm spherical seed centered on the coordinates for each ROI in MNI space. Left and right amygdala ROIs were created using the atlas-defined anatomy from the Talairach daemon in AFNI and aligned to MNI space (as in Motzkin et al. 2015). The transformation matrix from the registration procedure described above was used to transform each seed from MNI space to original space (3dfractionize), with the accuracy of seed locations for each participant verified by one of the authors (L.N).

For each participant, the mean resting-state BOLD time series from each seed ROI was included in a GLM (3dDeconvolve) to compute the correlation between each seed ROI’s time series and all other voxels in the brain. To create the voxelwise correlation maps for each seed ROI, we performed the following steps: (1) used the output from the GLM to convert R² values to correlation coefficients (r), and (2) converted the correlation coefficients to z-scores via Fisher's r-to-z transform (as in Philippi et al., 2015). The transformation matrix from the registration procedure described above was also used to align the correlation maps for each participant to MNI-152 space. The resulting whole brain z-score maps were then entered into the second level statistical analyses.

Statistical Analysis

To address the main aim of the study, we performed separate voxelwise regression analyses (3dttest ++ in AFNI) within the AHI group to examine the relationship between symptoms of depression and whole-brain rsFC for each ACC and amygdala ROI. Given that few resting-state studies have examined rsFC of emotion-related regions in AHI, we also investigated group differences in rsFC of ACC and amygdala regions between AHI and HIV uninfected controls, we performed unpaired two-sample t tests on the whole-brain z-score connectivity maps derived from the seed-based voxelwise rsFC analyses for each seed ROI. To assess the effect of Fiebig stages on rsFC of ACC and amygdala regions in AHI, we conducted supplemental analyses to compare rsFC for early Fiebig stages I and II (n = 22) versus later Fiebig stages III-V (n = 52) using unpaired two-sample t tests as described for the group analysis comparing AHI and controls. Lastly, we also performed all of these analyses using the final preprocessed resting-state files without global signal regression.

We applied a cluster-level family-wise error (FWE) correction approach with a whole-brain mask (3dClustSim in AFNI version updated May 2018) and cluster-extent thresholding (Forman et al. 1995; Carp 2012) to correct for multiple comparisons. To address the non-Gaussian nature of fMRI data, we used the autocorrelation function (-acf) to calculate the FWHM for each participant (3dFWHMx in AFNI; Eklund et al. 2016). The cluster-extent threshold corresponded to the probability of finding a random noise cluster at a predefined uncorrected voxelwise threshold of p < 0.001. Using this voxelwise cluster correction and Bonferroni correction for the number of seed ROIs (n = 5), a FWE cluster-corrected size of ≥122 voxels was significant at p_FWE < .01 (.05/5 = .01) in the analyses reported below. In all figures, findings are overlaid on the normalized mean anatomical image in MNI template space.

Associations between rsFC and HIV-related variables.

We also examined the correlations between significant rsFC of ACC and amygdala regions associated with affective symptoms and HIV-related biological and infection relevant variables. Specifically, we calculated the average z-scores for voxels within each significant cluster from the main regression analyses. Next, we performed Pearson's correlations between each average z-score and each HIV-related biological measure (CD4, CD8, CD4:CD8, plasma HIV RNA). Lastly, we conducted multiple linear regression analyses to determine whether the rsFC results associated with depressive symptoms in AHI remained significant after controlling for HIV-related biological measures and infection relevant variables (duration of HIV infection and Fiebig stage). All correlation and regression analyses described in this paragraph were completed in SPSS (version 25; SPSS/IBM, Chicago, IL).

Results

Participant characteristics

There were no statistically significant differences between AHI and uninfected control participants in age (AHI: mean = 28.9±8.9; uninfected control: mean = 30.2±5.4; t(102) = .74, p = .461). However, there were significant group differences in education level (χ(3) = 8.04, p = .045), with fewer control participants reporting high school or higher levels of education (Table 1). There were also significant group differences in gender (χ(2) = 27.29, p < .001), with more females in the control group (n = 10) than the AHI group (n = 0; Table 1). There were significant group differences in depressive symptom scores (t(82) = −2.92, p = .005), with higher depressive symptoms in the AHI group as compared with the subsample of control participants with HADS-D scores (Table 1). The AHI group had a mean (SD) CD4 count of 395 (±209) cells/uL, 6.0 (±1.1) log₁₀ copies HIV RNA and estimated duration of infection of 19.0 (±6.6) days. Most AHI participants were classified into early Fiebig stages I (9.5%), II (20.3%), and III (58.1%) with 67 (90.5%) participants showing signs of acute retroviral syndrome defined by a standardized checklist and completed by a physician.

There were no significant differences in depressive symptoms for Fiebig stages I/II versus III-V (t(62) = −. 14, p = .88). The AHI group with brief exposure to ART had no differences in demographic variables (ps = .43-.90), depressive symptoms (t(62) = 1.07, p = .29), or HIV-related biological variables compared to those without ART initiation (ps = .34-.96).

Within the AHI group, 25% of participants met criteria for clinically relevant depressive symptoms (HADS-D score ≥ 8). There was a significant difference in CD4:CD8 ratio (t(62) = 2.15, p = .04) between the AHI group with clinically relevant depressive symptoms versus without (HADS-D score < 8). For the remaining HIV-related variables, there were no differences between individuals with versus without clinically relevant symptoms of depression (CD4, p = .37; CD8, p = .26; log₁₀ copies of plasma HIV RNA, p = .13; duration of HIV infection, p = .42; Fiebig stage, p = .52).

Evaluation of excessive motion for resting-state data

There were no significant differences in average RMS motion between the AHI and HIV uninfected groups (t(102) = −0.31, p = .75) and there were no significant correlations between RMS and depressive symptoms (r = −.08, p = .51) or HIV-related measures (CD4, p = .71; CD8, p = .94; CD4:CD8, p = .23; log₁₀ copies HIV RNA, p = .60; duration of HIV infection, p = .16; Fiebig stage, p = .32. Therefore, RMS was not included as a covariate in either group or regressions analyses reported below.

rsFC of ACC and amygdala regions in AHI versus controls

There were no significant differences in rsFC of ACC for AHI versus uninfected controls (each p_FWE > .01). By contrast, there were significant group differences in rsFC of amygdala regions. Specifically, rsFC between left amygdala and right parahippocampal gyrus was significantly greater in the AHI group compared to controls (p_FWE < .01; Figure 1; Table 2). There was also significantly decreased rsFC (i.e., reduced negative correlations) in the AHI group between right amygdala and right cerebellum and left cerebellar tonsil compared to controls (p_FWE < .01; Figure 1; Table 2). These results were largely the same without global signal regression, except there was now a large bilateral cerebellar cluster instead of two smaller right and left cerebellar clusters as identified in the analysis with global signal (Table S1).

Figure 1. Differences in rsFC of the amygdala in AHI versus controls.

A. Greater rsFC between the left amygdala and right parahippocampal gyrus in AHI versus control groups; Decreased rsFC (i.e., decreased negative correlations) between the right amygdala and the B. right cerebellum, and C. left cerebellar tonsil in AHI versus control groups. The amygdala seed ROIs (red) and all results were displayed on the group average structural MRI in MNI template space. Bar graphs display the average z-scores for each significant cluster for each group. All results survived whole-brain cluster correction, including Bonferroni-correction for number of ACC seed regions (p_FWE < 0.01, p < 0.001 uncorrected). Color bar depicts the t-statistic from the unpaired two-sample t tests.

Table 2.

Group differences in rsFC for ACC and amygdala seed ROIs

Seed ROI	Cluster location	MNI coordinates (x, y, z)	Cluster size (voxels)	t-value	Average connectivity AHI a, b (n = 74)	Average connectivity CO a, b (n = 30)
pgACC	n/a	--	--	--	--	--
Anterior sgACC	n/a	--	--	--	--	--
Posterior sgACC	n/a	--	--	--	--	--
L amygdala	right parahippocampal gyrus	15, −12, −22	244	4.76	7.18	−1.06
R amygdala	right cerebellum	21, −36, −28	713	4.83	−0.52	−4.88
	left cerebellar tonsil	−17, −40, −34	187	4.37	−2.40	−5.04

Notes. AHI = Acute HIV, CO = HIV-negative control group; pgACC = pregenual anterior cingulate cortex, sgACC = subgenual anterior cingulate cortex. (p_FWE = .01, Bonferroni-corrected for number of seeds, p = .05/5 = .01; uncorrected p = .001).

^aLPI coordinate order

^bAverage connectivity of seed ROI to cluster location (t-values), without any covariates. There were no significant group differences in rsFC of ACC seed ROIs (each p_FWE > .01).

When controlling for education level and gender in the model, all rsFC findings remained significant, including for left amygdala and right parahippocampal gyrus (F(1,100) = 22.40, p < .001, partial η² = .18), and for right amygdala and right cerebellum (F(1,100) = 14.80, p < .001, partial η² = .13) and left cerebellar tonsil (F(1,100) = 14.24, p < .001, partial η² = .12). These findings remained significant in the same analyses without global signal regression (ps < .001). We also performed sensitivity analyses with unpaired two-sample t tests on the whole-brain z-score connectivity maps of ACC and amygdala ROIs in only male participants. The findings indicated that for AHI versus uninfected male control participants (n = 20), the results were comparable with those from the full sample for rsFC of ACC and amygdala ROIs. Specifically, there were still no significant findings for ACC regions. For the amygdala, the results were similar to those with the full sample, though the cluster sizes were smaller and no longer significant with the male only sample comprised of fewer participants (p_FWE > .01). We found the same results for the sensitivity analyses without global signal regression.

rsFC of ACC and amygdala regions related to Fiebig stage in AHI

We also performed group analyses comparing early Fiebig stages (I and II) to late Fiebig stages (III-V) within the AHI group. There were no significant differences in rsFC for any ACC or amygdala regions (p_FWE > .01). These results remained non-significant without global signal regression.

rsFC of ACC and amygdala regions related to depressive symptoms in AHI

Within the AHI group, depressive symptoms were related to unique rsFC signatures for ACC and amygdala regions. For both anterior and posterior sgACC regions, greater depressive symptoms were associated with reduced rsFC with posterior cingulate cortex/precuneus bilaterally (p_FWE < .01; Figure 2; Table 3). For the posterior sgACC, elevated depressive symptoms were related to increased rsFC with lateral prefrontal cortex, posterior sgACC, and cerebellum (p_FWE < .01; Figure 2; Table 3). Greater depressive symptoms were also associated with decreased rsFC between posterior sgACC and left uncus and posterior sgACC and left inferior parietal lobule (p_FWE < .01; Figure 2; Table 3). For the left amygdala, elevated depressive symptoms were related to decreased rsFC with right precuneus and increased rsFC with left fusiform gyrus extending into the cerebellum (p_FWE < .01; Figure 3; Table 3). There were no significant relationships between depressive symptoms and rsFC for the pgACC or right amygdala (p_FWE > .01). All of these rsFC findings related to depressive symptoms were largely the same without global signal regression, with the exception of the posterior sgACC and left inferior parietal lobule result (Table S2).

Figure 2. Depressive symptoms are related to distinct patterns of rsFC of sgACC regions in AHI.

A. Depressive symptoms associated with decreased rsFC between anterior subgenual ACC and right precuneus; B. Depressive symptoms associated with increased/decreased rsFC between posterior subgenual ACC and lateral and medial prefrontal, medial and lateral parietal, and cerebellar regions. The sgACC seed ROIs and all results are displayed on the group average structural MRI in MNI template space. All results survived whole-brain cluster correction, including Bonferroni-correction for number of ACC and amygdala seed regions (p_FWE < 0.01, p < 0.001 uncorrected). Color bar indicates uncorrected t values across all findings.

Table 3.

Regression results using rsFC for ACC and amygdala seed ROIs and depressive symptoms in AHI

Seed ROI	Cluster location	MNI coordinates (x, y, z)a	Cluster size (voxels)	t-value	Average connectivityb AHI (n = 64)
Depressive Symptoms
pgACC	n/a	--	--	--	--
Anterior sgACC	right precuneus	7, −54, 38	442	−5.04	6.47
Posterior sgACC	left superior frontal gyrus	−23, 64, −12	682	5.80	6.44
	left precuneus	−1, −60, 52	660	−5.95	1.09
	right superior frontal gyrus	25, 60, −14	331	6.60	7.32
	right middle frontal gyrus	45, 46, −14	302	6.47	8.94
	left anterior cingulate	−9, 22, −8	192	6.12	17.61
	left uncus	−39, −14, −44	178	−5.33	−2.71
	right culmen	45, −42, −34	159	5.49	6.84
	left inferior parietal lobule	−33, −42, 56	153	−5.05	−1.65
L amygdala	left fusiform gyrus	−41, −64, −20	280	5.37	−0.49
	right precuneus	21, −50, 44	167	−5.00	−6.04
R amygdala	n/a	--	--	--	--

Notes. AHI = Acute HIV, pgACC = pregenual anterior cingulate cortex, sgACC = subgenual anterior cingulate cortex, R = right, L = left. (p_FWE = .01; uncorrected p = .001).

^aLPI coordinate order

^bAverage connectivity of seed ROI to cluster location (t-values), without any covariates. There were no significant relationships between depressive symptoms and rsFC for the pgACC or R amygdala (p_FWE > .01).

Figure 3. Depressive symptoms are related to increased and decreased rsFC of amygdala regions in AHI.

A. Depressive symptoms associated with increased rsFC between left amygdala and left fusiform gyrus extending to cerebellum in AHI; B. Depressive symptoms associated with decreased rsFC between left amygdala and right precuneus in AHI. The amygdala seed ROIs and all results are displayed on the group average structural MRI in MNI template space. All results survived whole-brain cluster correction, including Bonferroni-correction for number of ACC and amygdala seed regions (p_FWE < 0.01, p < 0.001 uncorrected). Color bar indicates uncorrected t values across all findings.

Associations between rsFC and HIV-related variables

Only log₁₀ copies of plasma HIV RNA significantly negatively correlated with rsFC between posterior sgACC and left uncus (r = −.26, p = .038). There were no other significant correlations between the rsFC findings and CD4 cell count, CD8 cell count, CD4:CD8 ratio (all ps > .05). In our follow-up multiple linear regression analyses, all significant rsFC findings for ACC and amygdala ROIs related to depressive symptoms in AHI reported above remained significant after controlling for HIV-related biological variables, duration of HIV infection, or Fiebig stage (all ps < .00001).

Discussion

The present study examined the resting-state neural signatures of depressive symptoms in AHI. In the group analysis, we found that rsFC of the amygdala was altered in AHI when compared to that of uninfected controls. Analyses within the AHI group provide novel evidence that rsFC of affective regions are associated with depressive symptoms in the acute stage of infection. Our results revealed statistically significant relationships between depressive symptoms and decreased rsFC between ACC regions and PCC/precuneus, medial temporal and lateral parietal regions in AHI.

Symptoms of depression were also associated with increased rsFC between ACC regions and lateral prefrontal cortex, sgACC, and cerebellum in AHI. Depressive symptoms were related to both decreased and increased rsFC of amygdala with precuneus and ventral temporal cortex in AHI. There were few correlations between HIV-related biological factors and rsFC. We discuss each of these major findings in turn.

We observed increased rsFC between left amygdala and parahippocampal gyrus and decreased rsFC between the right amygdala and bilateral cerebellum in AHI as compared with uninfected controls. The present findings extend previous structural and functional neuroimaging research in chronic HIV populations reporting abnormalities in the amygdala (Ances et al. 2012; Clark et al. 2015, 2017; Behrman-Lay et al. 2016; Spies et al. 2016). While few resting-state studies to date have focused specifically on connectivity of the amygdala in HIV infection (Janssen et al. 2017; McIntosh et al. 2018), research in individuals with chronic HIV suggests that rsFC within networks including the amygdala may be altered.

Consistent with our finding of reduced rsFC of the amygdala in AHI, studies have found diminished rsFC of the salience network, which includes the amygdala, in participants with chronic HIV infection (Thomas et al. 2013; Chaganti et al. 2017). Other research in individuals with chronic HIV revealed changes in rsFC between the salience network and other cortical networks, such as the default mode network (DMN). For instance, Thomas and colleagues (2013) found trend-level decreases in rsFC between the salience network and the DMN in participants with chronic HIV as compared with uninfected participants. The results of the present study are comparable those findings because we also identified differences in rsFC between the amygdala and the parahippocampal gyrus within the DMN in AHI versus uninfected control participants.

However, our findings are different in that we showed increases (as opposed to decreases) in rsFC between the left amygdala and right parahippocampal gyrus of the DMN. These discrepancies could be explained by differences between these studies in terms of the rsFC method applied, age of participants (younger versus older adults), or disease stage (chronic versus acute). Another possible explanation is that the AHI participants in our study were experiencing elevated levels of stress related to their recent HIV diagnosis. Consistent with this possibility, previous studies in non-HIV populations have shown that exposure to acute stress is associated with increased rsFC between the amygdala and hippocampal regions within the DMN during recovery from the stress manipulation, including up to 2 hours later (Vaisvaser et al. 2013; Quaedflieg et al. 2015). Although we did not have a measure of stress for participants in the present study, we did measure psychological distress using the distress thermometer in the AHI group (Roth et al. 2000). There was no correlation between psychological distress and amygdala-hippocampal connectivity (r = .11, p = .403). Although psychological distress may not contribute to elevated connectivity in our study, it is possible that established self-report and neuroendocrine markers of stress may play a role. Thus, future work could determine whether stress in participants with AHI is related to elevated rsFC between the amygdala and DMN.

Similar to our results in AHI, previous rsFC studies in uninfected samples of adults and adolescents with major depression have found reduced rsFC between ACC regions and PCC/precuneus (Zhu et al. 2012; Connolly et al. 2013; van Tol et al. 2014). However, the findings of the current study are inconsistent with other research reporting increased rsFC between the same regions of the DMN in depression (Sheline et al. 2010; e.g., Berman et al. 2011; Liston et al. 2014). One possibility is that these inconsistencies are related to whether global signal regression was applied in preprocessing, as negative correlations or anti-correlations in rsFC tend to be more prevalent when global signal regression is included (Murphy et al. 2009). At the same time, this does not fully explain our finding of reduced rsFC, as the results were essentially the same when we performed our analyses without global signal regression. Moreover, two of the three studies reporting reduced rsFC did not include global signal regression (Zhu et al. 2012; Connolly et al. 2013).

Another potential explanation for this discrepancy is related to differences in sample characteristics such as age, severity of depressive symptoms, or gender of participants between studies with decreased versus increased rsFC. In line with this hypothesis, the three studies that reported decreased rsFC between ACC and PCC/precuneus were similar to our study in that they were in younger participants (Zhu et al. 2012; Connolly et al. 2013), in first-episode and treatment naive depressed participants (Zhu et al. 2012), and in a cohort comprised of mostly males, as was our cohort (van Tol et al. 2014). Studies that found increased rsFC between ACC and PCC/precuneus were mostly in older participants, with a longer duration of major depression, and/or with more depressive episodes, and in a cohort including more females (Sheline et al. 2010; Berman et al. 2011; Liston et al. 2014) than in the present study. It is important to emphasize that participants in the AHI group in our study were not diagnosed with major depression, but instead had varying levels of depression severity with 25% meeting clinically significant cutoffs for depressive symptoms.

It is also possible that reduced rsFC between ACC and PCC identified in AHI is related to HIV infection. In line with this proposal, several studies have reported diminished rsFC of the DMN, which includes the ACC and PCC, in early HIV and chronic HIV infection (Thomas et al. 2013; Ortega et al. 2015; Zhuang et al. 2017), and decreased beta oscillations in the DMN in participants with chronic HIV infection (Becker et al. 2013). However, further longitudinal research in this AHI cohort is warranted as our results appeared to be specific to depressive symptoms as opposed to HIV-related biological factors at this acute stage of infection.

Our results revealed diminished rsFC between ACC regions and the inferior parietal lobule and precuneus in relation to depressive symptoms in AHI. The ACC seed regions in the present study overlap with anterior portions of the DMN and the inferior parietal lobe and precuneus clusters overlap with parts of the frontoparietal network (Vincent et al. 2008; Yeo et al. 2011), suggesting that rsFC between these networks in AHI was reduced in relation to depressive symptoms. Decreased rsFC between DMN and frontoparietal networks has been reported both in major depression (Kaiser et al. 2015; Mulders et al. 2015) and in chronic HIV infection (Thomas et al. 2013, 2015).

Consistent with several previous studies in depression (Greicius et al. 2007; Zhu et al. 2012; Kaiser et al. 2015; Mulders et al. 2015), we found that greater depressive symptoms were associated with elevated rsFC within the posterior sgACC in AHI. These results extend research in individuals with chronic HIV showing relationships between depression symptomatology and altered theta activity within the rostral ACC (Kremer et al. 2015) and altered rsFC of sgACC (McIntosh et al. 2018). However, neuroimaging research in individuals with chronic HIV suggests that the effects of depressive symptoms on sgACC functioning may be distinct from HIV-related factors. For example, some neuroimaging studies in participants with chronic HIV report reduced rsFC within sgACC compared with uninfected controls (McIntosh et al. 2018), lower resting metabolism of rostral ACC (Andersen et al. 2010), and reduced volume within ACC regions (Küper et al. 2011; Kallianpur et al. 2012). Altogether, these studies suggest that sgACC functioning may be abnormally increased in relation to symptoms of depression but decreased in chronic HIV. One important area for future research will be to understand how HIV infection and depressive symptoms interact to influence ACC functioning over the course of the illness.

Our findings of greater depressive symptoms associated with increased rsFC between posterior sgACC and right middle frontal gyrus parallel prior studies in individuals with depression (Davey et al. 2012; Kaiser et al. 2015). The dorsolateral prefrontal cortex, located within the middle frontal gyrus, has also been used as a target for brain stimulation techniques in the treatment of major depressive disorder in uninfected cohorts (Fox et al. 2012, 2013; Liston et al. 2014). Altogether, these results could have implications for the use of transcranial magnetic stimulation to treat depression in HIV.

Moreover, we identified a relationship between depressive symptoms and increased rsFC between the posterior sgACC and the right cerebellum. These findings are broadly aligned with a growing body of research linking cerebellar dysfunction with depression in uninfected populations (Dutta et al. 2014; Kaiser et al. 2015; Córdova-Palomera et al. 2016). Our results are also consistent with recent resting-state fMRI studies showing aberrant rsFC between the cerebellum and other cortical regions in individuals with depression (Zeng et al. 2012; Liu et al. 2012; Ma et al. 2013; Guo et al. 2013; Kaiser et al. 2015; Gao et al. 2016; Córdova-Palomera et al. 2016). For instance, using a machine learning algorithm with rsFC of cerebellar subregions, Ma and colleagues (2013) were able to correctly classify 90.6% participants into depressed and healthy control groups. Some studies have also reported abnormal structure and function of the cerebellum in people living with chronic HIV infection (Tagliati et al. 1998; Klunder et al. 2008; Ernst et al. 2009; Elsheikh et al. 2010; Caldwell et al. 2014; Ann et al. 2016; Wang et al. 2018). One study found reduced cerebellar volume in participants with chronic HIV as compared with uninfected controls (Klunder et al. 2008). In the same study, significant correlations were found between the level of cerebellar atrophy and depressive symptoms (Klunder et al. 2008). Combined with these structural findings, our study suggests that depressive symptoms may be associated with aberrant rsFC of the cerebellum in HIV.

Besides the aforementioned group differences in amygdala connectivity, altered rsFC of the amygdala with precuneus and ventral temporal regions was also associated with depressive symptoms in AHI. Similar to our findings, abnormalities in rsFC of amygdala have been documented in adults, adolescents, and children with major depression without HIV (Mulders et al. 2015 for review). Further, disrupted amygdala connectivity has been associated with the severity of depressive symptoms in major depressive disorder (Dannlowski et al. 2009; Xie et al. 2012; Ramasubbu et al. 2014).

In terms of HIV, to our knowledge this is the first study to report a link between amygdala connectivity and depressive symptoms during the acute stage of infection. Only two previous studies have investigated the relationship between amygdala function and symptoms of depression among individuals with HIV, and both focused on chronic infection (Clark et al. 2017; McIntosh et al. 2018). McIntosh and colleagues (2018) found that reduced rsFC between sgACC and bilateral amygdala regions predicted greater depressive symptomatology in individuals with chronic HIV but not uninfected controls. In another study using task-based fMRI in participants with chronic HIV, diminished amygdala reactivity to negative emotional faces was associated with greater neuropsychiatric symptoms, including depression (Clark et al. 2017 . In addition to depressive symptoms examined in HIV-infected participants, several previous studies document the negative effects of childhood trauma exposure on the integrity of the amygdala in people living with HIV (Spies et al. 2016; Clark et al. 2017; Thames et al. 2018. Together, this research suggests that the presence of neuropsychiatric symptoms may further exacerbate amygdala dysfunction in HIV.

In the present study, only plasma HIV RNA was related to reduced sgACC-left uncus connectivity. Results from a structural neuroimaging study revealed significant correlations between HIV DNA and cortical thinning in anterior temporal and medial temporal lobe regions in participants with chronic HIV (Kallianpur et al. 2012). The absence of a strong link between HIV disease variables and connectivity in the current study suggests the presence of intermediating variables and/or interactions among predictor variables. For example, it is possible that elevated CNS inflammation may have contributed to the altered rsFC in AHI in our study (Valcour et al. 2012). Increased levels of proinflammatory cytokines (i.e., interleukin-6) and depressive symptoms have been associated with resting-state activity in the orbitofrontal cortex in chronic HIV (McIntosh et al. 2018). Once again, longitudinal research in this cohort is needed to examine interdependencies between depressive symptoms, viral related factors, and long-term inflammation and rsFC of ACC and amygdala regions.

Our results could have important implications for early interventions designed to reduce depressive symptoms or to normalize rsFC of affective brain regions in the early stages of infection. For example, brain stimulation techniques have also been used to treat depression by modulating rsFC of affective brain regions in HIV-uninfected individuals (e.g., Fox et al. 2012; Liston et al. 2014). One study has applied brain stimulation to treat depressive symptoms in chronic HIV (Knotkova et al. 2012), suggesting that this approach may also be feasible for individuals with AHI.

There are some limitations to the current study that should be acknowledged. First, given that we had incomplete depressive symptom information for the uninfected control participants, we did not directly examine interactions between HIV infection and symptoms of depression. However, the AHI group had significantly higher depressive symptom scores than the subsample of uninfected control participants with HADS-D scores. Moreover, we did find that 25% of participants with AHI had clinically significant depressive symptoms based on established clinical cutoffs for depression, suggesting that those participants had abnormally elevated depressive symptoms. Nevertheless, additional research with complete depressive symptom information in both HIV and uninfected control participants will be required to address this question. Second, the sample of AHI participants was mostly male, limiting the generalizability of the findings. The possibility that relationships between depressive symptoms and rsFC of ACC and amygdala may differ between men, women, and transgender individuals with HIV warrants further study. Third, it is possible that differences in the amount of alcohol consumption, exercise, or hours of sleep before the scan could have contributed to individual variability in rsFC. However, given that the majority of participants (n = 91) were scanned in the morning, it is less likely that participants would be under the influence of alcohol or other drugs. Fourth, we only examined the neural signatures of depressive symptoms in the present study as opposed to other psychological symptoms that may be elevated following the recent HIV diagnosis. For example, future research should determine whether acute posttraumatic symptoms in AHI contribute to differences in rsFC between AHI versus uninfected control participants.

Conclusion

In summary, we demonstrated novel findings in AHI relating depressive symptoms to rsFC of ACC and amygdala regions previously implicated in depression. Only plasma HIV RNA was associated with reduced rsFC of sgACC, suggesting that biomarkers of HIV may not contribute to all changes in rsFC associated with depressive symptoms during the acute stage of infection. Longitudinal research after initiation of ART in AHI will be required to determine whether early changes in rsFC of ACC and amygdala regions predict chronic depressive symptoms and biological markers of infection with viral suppression after long-term ART.