Neurotransmitter and metabolic effects of interferon-alpha in association with decreased striatal dopamine in a Non-Human primate model of Cytokine-Induced depression

Abstract

Inflammatory stimuli administered to humans and laboratory animals affect mesolimbic and nigrostriatal dopaminergic pathways in association with impaired motivation and motor activity. Alterations in dopaminergic corticostriatal reward and motor circuits have also been observed in depressed patients with increased peripheral inflammatory markers. The effects of peripheral inflammation on dopaminergic pathways and associated neurobiologic mechanisms and consequences have been difficult to measure in patients. Postmortem tissue (n = 11) from an established, translationally-relevant non-human primate model of cytokine-induced depressive behavior involving chronic interferon-alpha (IFN-a) administration was utilized herein to explore the molecular mechanisms of peripheral cytokine effects on striatal dopamine. Dopamine (but not serotonin or norepinephrine) was decreased in the nucleus accumbens (NAcc) and putamen of IFN-a-treated animals (p < 0.05). IFN-a had no effect on number of striatal neurons or dopamine terminal density, suggesting no overt neurodegenerative changes. RNA sequencing examined in the caudate, putamen, substantia nigra, and prefrontal cortical subregions revealed that while IFN-a nominally up-regulated limited numbers of genes enriching inflammatory signaling pathways in all regions, robust, whole genome-significant effects of IFN-a were observed specifically in putamen. Genes upregulated in the putamen primarily enriched synaptic signaling, glutamate receptor signaling, and inflammatory/metabolic pathways downstream of IFN-a, including MAPK and PI3K/AKT cascades. Conversely, gene transcripts reduced by IFN-a enriched oxidative phosphorylation (OXPHOS), protein translation, and pathways regulated by dopamine receptors. Unsupervised clustering identified a gene co-expression module in the putamen that was associated with both IFN-a treatment and low dopamine levels, which enriched similar inflammatory, metabolic, and synaptic signaling pathways. IFN-a-induced reductions in dopamine further correlated with genes related to excitotoxic glutamate, kynurenine, and altered dopamine receptor signaling (r = 0.78–97, p < 0.05). These findings provide insight into the immunologic mechanisms and neurobiological consequences of peripheral inflammation effects on dopamine, which may inform novel treatment strategies targeting inflammatory, metabolic or neurotransmitter systems in depressed patients with high inflammation.

1. Introduction

Dopamine neurons innervating the ventral and dorsal striatum play an important role in reward-based learning, motivation, and voluntary movement, and may contribute to multiple depressive symptoms (Chaudhury et al., 2013; Dunlop and Nemeroff, 2007). One pathophysiologic pathway thought to influence dopamine in depression is inflammation. For example, human neuroimaging studies have consistently reported that peripheral administration of inflammatory cytokines (e.g., interferon-alpha [IFN-a]) or cytokine inducers (e.g., vaccination, endotoxin) affects dopamine-rich cortical and subcortical brain regions that contribute to inflammation-related symptoms of reduced motivation and motor slowing (Brydon et al., 2008; Capuron et al., 2007; Capuron et al., 2012b; Eisenberger et al., 2010; Harrison et al., 2015). Clinical and translational evidence has further shown that peripheral inflammatory cytokines decrease striatal dopamine availability in association with reduced motivation (Capuron et al., 2012b; Felger et al., 2013b; Yohn et al., 2016). While several potential mechanisms for the effects of peripheral inflammation on striatal dopamine availability and release have been proposed (Felger and Miller, 2012; Felger and Treadway, 2017), the precise neurotransmitter and relevant immunologic signaling pathways by which inflammation affects dopamine in relation to depressive symptoms are not well understood.

A great deal of knowledge about how inflammatory cytokines impact the brain and behavior has come from studying patients with hepatitis C virus or malignant melanoma administered peripheral therapy with IFN-a, which induces activation of other inflammatory cytokines and chemokines in blood and cerebrospinal fluid in association with depressive behaviors in both humans and non-human primates (Felger et al., 2007; Felger et al., 2013a; Felger et al., 2013b; Raison et al., 2009). Positron emission tomography in patients administered IFN-a for hepatitis C revealed increased striatal fluorodopa (18F) uptake, but decreased turnover, consistent with reduced dopamine availability and release that correlated with symptoms of reduced motivation (Capuron et al., 2012b). Our work in a translational non-human primate model of chronic IFN-a-induced depressive behavior also demonstrated decreased striatal dopamine release, as measured by in vivo microdialysis, which correlated with reduced effort-based sucrose consumption (a measure of reward sensitivity and anhedonia in monkeys)(Felger et al., 2013b). Studies in rodents have further confirmed decreased striatal dopamine release after peripheral inflammatory cytokine exposure in association with reduced effort-based responses for sucrose reward (Yohn et al., 2016), and demonstrated that stress-induced peripheral cytokines (e.g., IL-6) can access specific brain regions, including notably the nucleus accumbens (NAcc) of the ventral striatum, where peripheral immune cells traffic in animals that are susceptible to depressive and anhedonic behavior (Menard et al., 2017).

Inflammatory cytokines that access the brain are thought to affect availability of dopamine and other monoamines through increasing oxidative stress and usurping of key cofactors like tetrahydrobiopterin that are required for their synthesis (Cunnington and Channon, 2010; Felger et al., 2013a; Kitagami et al., 2003). Accordingly, our work in non-human primates also demonstrated that peripheral IFN-a-induced decreases in striatal dopamine release were reversed by administration of the immediate dopamine precursor, levodopa (L-DOPA)(Felger et al., 2015; Felger et al., 2013b). Additional mechanisms by which inflammation may impact dopamine transmission include effects on its packaging, release, and reuptake. For example, interleukin (IL)-1 and tumor necrosis factor (TNF) have been shown to decrease vesicular monoamine transporter 2 expression in rat cell lines (Kazumori et al., 2004), whereas stimulation of mitogen activated protein kinase (MAPK), a common inflammatory signal transduction pathway, increased dopamine transporter activity in a human cell line (Moron et al., 2003). Striatal dopamine may be further affected by inflammation-related increases in glutamate signaling, either directly or via excitotoxicity and downstream oxidative processes that impact its synthesis (Felger and Treadway, 2017). Better understanding the neurotransmitter and relevant signal transduction cascades by which inflammation affects dopamine may inform therapeutic strategies for patients with depression and higher inflammation.

Indeed, a significant proportion of patients with depression exhibit evidence of increased inflammation in both the periphery and central nervous system (CNS), including elevations in inflammatory cytokines like IFN-a and IL-6 and their signal transduction pathways (Bekhbat et al., 2021; Dantzer et al., 2008; Mostafavi et al., 2013; Sforzini et al., 2024; Shelton et al., 2011; Zunszain et al., 2013). We and others have also found that elevated inflammatory markers in patients with depression are related to lower functional connectivity within corticostriatal circuits that subserve reward processing and goal-directed behaviors involving dopamine-rich regions of the ventral and dorsal striatum and prefrontal cortex (PFC) in association with anhedonia and motor slowing (Felger et al., 2016; Rengasamy et al., 2021; Yin et al., 2019). Recent evidence suggests that these functional neuroimaging findings may be due to the impact of inflammation on the availability and release of striatal dopamine. Indeed, acute pharmacological challenge with L-DOPA increased functional connectivity within a ventral striatum to PFC reward circuit specifically in patients with higher inflammation (Bekhbat et al., 2022), which correlated with reduced symptoms of anhedonia. Similar functional neuroimaging biomarkers of the potential effects of inflammation on dopamine in relevant corticostriatal circuits were reversed by anti-inflammatory challenge with the cytokine antagonist infliximab (Treadway et al., 2024). As more nuanced approaches to reverse the impact of inflammation on the brain in patients with depression and other psychiatric illness are needed (Miller and Raison, 2023), translationally-relevant animal models can provide a window into potential targets for new therapeutic development.

Herein, postmortem tissue collected from our established non-human primate model of inflammation-induced depressive behavior involving chronic (4-week) IFN-a administration (Felger et al., 2007; Felger et al., 2015; Felger and Lotrich, 2013; Felger et al., 2013b) was used to examine peripheral cytokine effects on concentrations of dopamine and its metabolites compared to controls in sub-regions of the striatum (NAcc, caudate and putamen) by high-performance liquid chromatography (HPLC), with exploratory analyses for the other monoamines. Integrity of striatal neurons and dopaminergic axon terminals was also assessed by immunohistochemistry (IHC), and gene expression signatures associated with IFN-a effects on dopamine were determined by RNA-sequencing (seq), in fixed and fresh tissue respectively. To determine whether key transcriptional evidence of the neurobiological consequences of IFN-a were specific to the striatum, gene expression was also examined in midbrain dopamine neurons (consistent with substantia nigra [SNC]) and PFC regions BA25, BA24a, and BA24c (Paxinos et al., 1999)(representing subgenual, rostral, and dorsal subregions of the anterior cingulate cortex) by RNA-seq. We hypothesized that IFN-a would decrease striatal dopamine, but not other monoamines, and that that these effects would be accompanied by transcriptional signatures reflecting the immunologic, metabolic and neurotransmitter mechanisms and consequences of IFN-a effects on dopamine.

2. Methods

Animals and IFN-a treatment.

Post-mortem brain tissue was collected from 6 male and 5 female rhesus monkeys (Macaca mulatta) aged 10 to 14 years that were administered IFN-a (rHu-IFN-α-2b [Schering-Plough, Kenilworth, NJ] 20 MIU/m² s.c., n = 7, 4 male and 3 female) or saline (n = 4, 2 male and 2 female) between 7–10 am in equivalent volume as IFN-a (0.5–1.5 ml), 5 days/week for 4 weeks, consistent with high dose IFN-a therapy for malignant melanoma (Capuron et al., 2007). Animals were sacrificed shortly after the last dose of IFN-a. All animals had been treated previously with IFN-a, but never for more than 4 weeks and never more than 2 times per year. Animals were free of any immune or other treatment administration for at least 12 weeks prior to starting the study. The larger sample size in the IFN-a-administered group (n = 7 for IFN-a versus n = 4) was designed to overcome greater anticipated variability in the treated animals, as previously observed in tau-induced neuroinflammatory outcomes in rhesus macaques (Beckman et al., 2024), and to provide sufficient power (80 %) to detect significant correlations between assessed variables of moderate to large effect sizes in the 7 treated animals at alpha = 0.05. Live animal procedures were approved by the IACUC of Emory University, conducted following guidelines established by the Animal Welfare Act and the NIH for the housing and care of laboratory animals, and performed in accordance with institutional regulations. All efforts were made to minimize suffering of the animals.

Tissue processing and dissections.

Animals were deeply anesthetized with pentobarbital (100 mg/kg i.v.) and transcardially perfused with 1.5 L of cold 0.1 % saline with heparin. After perfusion, brains were removed from the skull. The left hemisphere was fixed in cold 4 % paraformaldehyde for 72–96 h. This hemisphere was sliced coronally in 1–1.5 cm blocks, sunk in increasing concentrations of sucrose (10 to 30 %) in 0.1 M phosphate buffer, and stored at − 80 °C until sectioning. Blocks were cut into 50 μm-thick coronal sections using a vibrating microtome and stored at − 20 °C in an anti-freeze solution, containing 30 % ethylene glycol and 30 % glycerol in phosphate buffer, until immunohistochemistry. The right hemisphere was sliced coronally into 1–1.5 cm blocks and micro-dissections were collected for RNA extraction and HPLC. One sample was taken from NAcc (15 ± 2 ng) for HPLC, whereas two samples were collected from the much larger basal ganglia structures, caudate (23 ± 9 ng) and putamen (22 ± 9 ng), for both RNA-seq and HPLC. One sample each was also taken from the SNC as well as BA24a, 24c and 25 (22 ± 7 ng) subregions of the anterior cingulate cortex of PFC (according to the Paxinos et al. Rhesus Monkey Brain Atlas (Paxinos et al., 1999)) as reference regions for the striatum for RNA-seq. For gene expression, samples were stored in RNAlater (Sigma-Aldrich, St. Louis, MO). For HPLC, samples were placed in a solution of 0.5 M perchloric acid and 0.01 % (w/v) ascorbic acid for each gram of brain tissue at 4 °C and homogenized with a tissue homogenizer (PowerGen 125, Fisher Scientific, USA), centrifuged at 4 °C for 30 mins at 13,000 rpm and the supernatant aliquoted. Samples were then stored at −80 °C until analysis.

HPLC.

Concentrations of monoamines and their metabolites were determined by HPLC with electrochemical detection using an internal standard by Brainsonline.org/Charles River according to established procedures and as described previously (Felger et al., 2015; Felger et al., 2013b). Samples (10 μl) were mixed with internal standard (3,4-Dihydroxybenzylamine, DHBA, 20 μl), then 20 μl was injected onto the LC system by an automated sample injector (Gilson XL233, Gilson, France). Chromatographic separation was performed on a reversed phase Hypersil BDS C18 (150 × 2.1 mm, 3 μm) analytical column (Thermo Scientific Keystone, USA) held at a temperature of 32.5 °C. Components were separated using isocratic sodium acetate buffer, pH = 4.1. Detection was performed electrochemically using a potentiostate (Antec Leyden, model Intro, The Netherlands) fitted with a glassy carbon electrode set at + 600 mV vs. Ag/AgCl (Antec Leyden, the Netherlands).

IHC and quantitative image analysis.

Tissue was stained for potential IFN-a effects on axon terminals and neurons in fixed tissue with antibodies against tyrosine hydroxylase (TH; AB152, 1:000, Millipore Sigma, Burlington, MA) and NeuN (ab177487, 1:5000, Abcam, Cambridge, MA). Images were captured using a Leica SP8 system (Leica Microsystems, Buffalo Grove, IL) located at the Emory Integrated Cellular Imaging Core. Representative images (2 from each sub-region of striatum) were taken at 20X magnification. Tyrosine hydroxylase (TH) was quantified by fluorescence intensity. NeuN + neurons were quantified using CellProfiler software (https://cellprofiler.org/)(Carpenter et al., 2006).

RNA-seq and analysis.

RNA-seq was conducted at the Yerkes Nonhuman Primate Genomics Core laboratory. Total RNA was prepared using RNeasy kits (QIAGEN, Hiden, Germany). Polyadenylated transcripts were purified on oligo-dT magnetic beads, fragmented, reverse transcribed using random hexamers, and incorporated into barcoded cDNA libraries based on the Illumina TruSeq platform, according to the manufacturer’s instructions. Libraries were validated by Agilent Bioanalyzer analysis and quantified, pooled, and clustered on Illumina TruSeq v3 flow cells and sequenced on an Illumina HiSeq 1000 at 100-nt SR, targeting approximately ~ 18–20 M reads per sample. Alignment was performed using STAR (Dobin et al., 2013); parameters were set using the annotation as a splice junction reference, and unannotated, noncanonical splice junction mappings and nonunique mappings were removed from downstream analysis. Transcripts were annotated using MacaM assembly and annotation of the Indian rhesus macaque genome (Mmul_10)(Warren et al., 2020). Transcript abundance estimates were calculated with htseq-count (Anders et al., 2015). Low expressed transcripts were filtered out, retaining approximately 12,000 annotated genes per brain region with at least 10 reads across 4 samples (the smallest group size). DESeq2 version 1.42.1 was used for normalization and differential expression analysis (Love et al., 2014). Normalized read counts were used for visualization of results, whereas regularized log expression data were generated and used for differential expression estimation between IFN-a-treated and control samples for each brain region, while adjusting for sex. A false discovery rate (FDR) < 5 % was used in conjunction with a ≥ 20 % difference (1.2-fold change)(Boyle et al., 2024; Cole et al., 2015; Cole et al., 2003; Mehta et al., 2013; Miller et al., 2008; Torres et al., 2013) to define differentially expressed genes, which were therefore predictive based on both statistical significance (p-value) and effect size (fold change).

Identified genes were examined for significantly represented pathways using WikiPathways and KEGG (Kyoto Encyclopedia Gene and Genome) databases as implemented in clusterProfiler (Yu et al., 2012) and GeneGo MetaCore (St. Joseph, MI, USA)(Ekins et al., 2007; Han et al., 2016; Mehta et al., 2013). An FDR significance threshold of q < 0.1 was used herein for pathway analysis (de Kluiver et al., 2019; Hulsegge et al., 2009; Jansen et al., 2016; Mostafavi et al., 2014; Storey and Tibshirani, 2003; Yang et al., 2008; Yang et al., 2014; Zhou et al., 2018). Nominal pathway significance threshold was p < 0.05.

Weighted Gene Co-Expression Network Analysis (WGCNA).

WGCNA (Langfelder and Horvath, 2008), an unsupervised hierarchical clustering method that has previously been applied in macaque brain transcriptomic studies (Arcego et al., 2024; Hawes et al., 2022), was used to first construct a putamen gene co-expression network and then identify clusters (i.e., “modules”) of highly inter-correlated genes that associated with phenotypic “traits” including IFN-a treatment as well as striatal dopamine and its metabolite DOPAC. Each gene co-expression module was summarized by its first principal component (called “module eigengene” [ME]), which was subsequently tested for association with the above-mentioned traits. Functional characterization of genes comprising modules of interest was made via pathway analysis as described above. To verify the quality of clustering (e.g., examine whether most IFN-a-associated genes were clustered into the same few modules rather than spread across many different modules), for each gene, its correlation with IFN-a treatment was calculated whereby a gene was deemed IFN-a-associated at nominal p-value < 0.05. The percent of all IFN-a-associated genes found within modules of interest was then calculated and reported. An identical procedure was performed to assess how well dopamine-associated genes were clustered. For details of WGCNA see Supplement.

qPCR.

Quantitative PCR was conducted on transcripts from RNA-seq that exhibited significant relationships in targeted analyses by using predesigned TaqMan Gene Expression Assay probes (Applied Biosystems)(Felger et al., 2012) for dopamine receptor 2 (DRD2; Rh01014211_m1), N-Methyl-D-aspartic acid (NMDA) receptor subunit 2a (GRIN2a; Mf03986722_m1) and NMDA receptor subunit 2b (GRIN2b; Rh01002011_m1) with TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA) and ABI 7900HT Sequence Detection System (Applied Biosystems). All samples were run in duplicate. Expression levels were expressed in arbitrary units and normalized relative to the housekeeping gene POL2RA (previously shown not to be influenced by IFN-a)(Felger et al., 2011b) to control for differences in cDNA loading.

Statistics.

HPLC and IHC data were analyzed using general linear models including a within-subject, repeated-measures variable (brain region) and between-subject variable (treatment) to determine main effects of IFN-a versus control or potential condition-by-region interactions. Post-hoc tests with Bonferroni correction were used to identify significant treatment effects within each sub-region of striatum. Significant results from general linear models were confirmed when controlling for sex of the animals as covariates. Exploratory analyses were conducted to examine potential sex-by-treatment interactions on HPLC outcomes (see Supplement). Variables that did not pass normality or equal variance tests were natural log transformed for multivariate analysis. All tests were two-tailed with α = 0.05. Statistical analyses were conducted using SPSS 24 (IBM, Armonk, NY). Dopamine-gene correlations within the IFN-a-treated group were performed in R (Barfield et al., 2012).

3. Results

IFN-a decreased tissue content of dopamine in the nucleus accumbens and putamen without impacting dopamine terminals or striatal neuron density.

Analysis of monoamines and their metabolites by HPLC revealed that IFN-a decreased tissue content of dopamine (F[1,9] = 7.86, p < 0.05). Post-hoc tests revealed decreased dopamine in the NAcc and putamen (p < 0.05), but not caudate (p = 0.82; Fig. 1A), and the effects of IFN-a in these regions remained significant when controlling for sex (p < 0.05). For 3,4-dihydroxyphenylacetic acid (DOPAC), but not homovanillic acid (HVA)(p = 0.19), a treatment by region interaction was observed (F [2,18] = 5.87, p < 0.05; Fig. 1B–C) indicating decreased DOPAC only in the putamen (p < 0.05), which also remained significant when controlling for sex (p < 0.05). Exploratory analyses for serotonin and norepinephrine and their metabolites showed a trend for IFN-a to decrease serotonin in NAcc (p = 0.06; Fig. 1D–E) but not other regions (as assessed by independent T-test due to data below detection limits in two animals for the NAcc, two for putamen, and one for caudate). No effect of IFN-a was seen for 5-hydroxyindoleacetic acid (5-HIAA), norepinephrine or its metabolite dihydroxyphenylglycol (DHPG; n = 1 below detection)(all p > 0.49; Figure S1).

Fig. 1.

Decreased tissue content of dopamine in the nucleus accumbens (NAcc) and putamen, but not caudate, of animals administered interferon-alpha (IFN-a) (red bars) compared to controls (blue bars) (A). Similarly, IFN-a significantly lowered DOPAC only in the putamen (B), whereas HVA was not significantly changed (C). Moreover, the effects of IFN-a on monoamines was specific for dopamine as only a trend was observed for decreased serotonin in the NAcc (p = 0.058) but not other striatal subregions, with no significant change in norepinephrine (D-E). Abbreviations: NAcc, nucleus accumbens; DOPAC, dihydroxyphenylacetic acid; HVA, homovanillic acid; IFN, interferon. Data are shown as Mean ± SEM. *p < 0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

As with previous findings in IFN-a-treated humans and monkeys, no effect of IFN-a on TH + terminals (p = 0.96) in the striatum was observed (Figure S2A,C). IFN-a also had no effect on the number of NeuN + neurons (p = 0.81)(Figure S2B,D). All results remained similar when controlling for sex.

IFN-a induced classic inflammatory pathways and affected genes related to intracellular metabolism and glutamate and dopamine receptor signaling.

Whole genome analysis of differentially expressed genes (at 20 % fold change and FDR-adjusted p < 0.05) identified 2392 genes significantly regulated by IFN-a in the putamen, but very few genes in other striatal and cortical regions (Fig. 2A), suggesting that IFN-a had robust transcriptomic effects specifically in the putamen.

Fig. 2.

RNA-Seq revealed that interferon-alpha (IFN-a) had profound transcriptomic effects in the putamen. Number of genes differentially expressed at the threshold of 20 % fold change and FDR-adjusted p-value < 0.05 between IFN-a-treated and control animals across the assessed regions of the striatum and three regions of cortex are shown (A), followed by a volcano plot of the 2392 differentially regulated genes in the putamen (B). Within the putamen, IFN-a increased genes related to inflammation (IL-6, MAPK signaling), interferon response (PI3K), glutamate signaling, and metabolism (Insulin, HIF-1)(C). IFN-a downregulated pathways related to oxidative phosphorylation, cytoplasmic ribosomal proteins, tight junction, and a signaling cascade regulated by dopamine receptors (D). Abbreviations: IFN-a, interferon-alpha; SNC, substantia nigra pars compacta; WP, WikiPathways; KEGG, Kyoto Encyclopedia Gene and Genome; MAPK, mitogen-activated protein kinase; NMDA, N-methyl-D-aspartate; IL, interleukin; MEK; mitogen-activated protein kinase kinase; ERK, extracellular signal-regulated kinase; PI3K, phosphatidylinositol 3-kinase; AKT, protein kinase B; OXPHOS, oxidative phosphorylation; PDGFR, Platelet-derived growth factor receptor.

Of the 1462 genes significantly up-regulated in putamen (Table S1A), top genes included SV2C, a synaptic protein that mediates dopamine release and is disrupted in Parkinson’s disease (Dunn et al., 2017), and ERC2, a key structural component of neurotransmitter release (Hida and Ohtsuka, 2010)(Fig. 2B, circled). Although serotonin concentrations were unaltered in the putamen as reported above, we also observed an upregulation of the serotonin 2C receptor gene HTR2C which is known to regulate dopaminergic circuitry (Mickey et al., 2012) (Fig. 2B). Consistent with the idea that upregulated genes (such as SV2C and ERC2) and pathways may represent compensation for low dopamine post-IFN-a, multiple pathways related to dysregulated synaptic signaling were enriched (Synaptic signaling associated with autism spectrum disorders, Disruption of postsynaptic signaling by copy number variations, Regulation of intrinsic membrane properties and excitability of striatonigral medium spiny neurons in Huntington’s disease; Figure S3). Interestingly, glutamate receptor signaling pathways were also enriched (Glutamatergic synapse, NMDA-dependent postsynaptic long-term potentiation in CA1 hippocampal neurons)(Fig. 2C), which included genes encoding NMDA, AMPA and metabotropic glutamate receptor subtypes (Figure S4). Altered synaptic transcripts were accompanied and possibly driven by Phosphatidylinositol 3-kinase (PI3K) and Mitogen-activated protein kinase (MAPK) signaling pathways, which are induced by inflammatory stimuli including IFNs and act to influence growth factors and synaptic transmission (Platanias, 2005). Indeed, IFN-a-upregulated genes enriched inflammatory pathways such as IL-6 signaling via MEK/ERK and PI3K/AKT cascades and TNF signaling, along with growth- (EGFR, BDNF) and metabolism-related pathways (Insulin signaling, Hypoxia- and receptor-mediated HIF-1 activation)(Fig. 2C and Table S1B).

The 930 genes down-regulated by IFN-a in the putamen (Table S2A) were primarily related to metabolism (Electron transport chain OXPHOS system in mitochondria) and protein synthesis and degradation (Cytosolic ribosomal proteins; Proteasome degradation)(Fig. 2D). Other down-regulated gene pathways were also related to dopamine receptor signaling (Transactivation of PDGFR in non-neuronal cells by Dopamine D2 receptor; Glutamic acid regulation of Dopamine D1A receptor signaling) (Fig. 2D and Table S2B).

Using a less stringent significance criteria (20 % fold change with nominal p < 0.05), 665 and 510 differentially expressed genes were uncovered in the caudate and SNC, as well as 315, 359, and 840 genes in BA24a, 24c, and 25, respectively (see Supplement and Fig S5). Pathway analysis of these nominally changed genes revealed that caudate (and to some extent SNC) displayed evidence of IFN-a-induced changes in neurotransmitter/synaptic signaling, whereas cortical regions either had few pathways (BA24a) or enriched mainly inflammatory pathways (BA24c and 25)(Table S3, all p < 0.05). These results reaffirmed that transcriptional effects of IFN-a relevant to dopamine were specific to the striatum and strongest in the putamen (Fig S5).

Co-expression analysis revealed synaptic and metabolic pathways in association with IFN-a treatment and striatal dopamine.

Unsupervised clustering via WGCNA revealed that putamen gene data clustered into 17 co-expression modules, each arbitrarily assigned a color name (Fig. 3A). Modules Turquoise and Blue (“MEturquoise”, “MEblue”) were of primary interest based on their ME association with both IFN-a treatment (Turquoise: r = 0.81, p = 0.002; Blue: r = −0.83, p = 0.001) and tissue content of dopamine in the putamen (Turquoise: r = −0.7, p = 0.018; Blue: r = 0.69, p = 0.018).

Fig. 3.

Unsupervised clustering via WGCNA revealed two gene modules in the putamen, Turquoise and Blue, whose eigengenes (ME) correlated with both IFN-a treatment and tissue content of dopamine (A). Module Turquoise contained 498 genes which were concurrently associated with IFN-a treatment and low dopamine (p < 0.05). These genes enriched inflammatory (IL-6), synaptic signaling, and metabolic (mTOR, insulin) pathways (B). Conversely, Module Blue contained 314 genes that were negatively associated with IFN-a treatment and higher dopamine levels (p < 0.05), which primarily enriched oxidative phosphorylation and proteasome degradation pathways (C). Abbreviations: IFN-a, interferon-alpha; DA, dopamine; WP, WikiPathways; KEGG, Kyoto Encyclopedia Gene and Genome; EGFR, epidermal growth factor receptor; mTOR, mechanistic target of rapamycin; PI3K, phosphatidylinositol 3-kinase; AKT, protein kinase B; IL, interleukin; MEK; mitogen-activated protein kinase kinase; ERK, extracellular signal-regulated kinase; NMDA, N-methyl-D-aspartate; OXPHOS, oxidative phosphorylation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Despite not considering exposure or outcome variables (e.g., associations with IFN-a treatment or dopamine levels) in its clustering algorithm to create unbiased networks based on gene co-expression patterns alone, WGCNA effectively clustered IFN-a- and dopamine-related genes into well-defined modules. Specifically, the turquoise module contained n = 1390, i.e., 91 % of all genes that were positively associated with IFN-a treatment (p < 0.05), as well as n = 556, or 91.7 % of all genes negatively associated with dopamine levels in the putamen (p < 0.05). A large overlap was observed (i.e., genes that both positively associated with IFN-a and negatively with dopamine, n = 498 genes), and significantly enriched similar pathways as those reported in Table S1B for the 1462 IFN-a-upregulated genes (as determined by DESeq2). These included synaptic signaling (Constitutive and regulated NMDA receptor trafficking, Synaptic signaling pathways associated with autism spectrum disorder) and metabolic (mTOR, Insulin signaling through PI3K/AKT) pathways (all q < 0.05; Fig. 3B and Table S4). Conversely, the blue module contained n = 1139, or 89.2 % of all genes negatively associated with IFN-a treatment (p < 0.05), and n = 377, or 92 % of all genes positively associated with dopamine levels in the putamen (p < 0.05). Of these, n = 314 genes overlapped, which primarily enriched OXPHOS and Proteasome pathways (q < 0.05; Fig. 3C and Table S5).

Targeted analysis of gene expression related to glutamate receptors, cytokine signaling, and dopamine neurotransmission and associations with dopamine concentrations.

To further examine IFN-a-induced increases in NMDA-dependent glutamate receptor signaling as observed by RNA-seq (e.g., driven by GRIN2/NR2 NMDA subunits; see Figure S4), and because excessive glutamate signaling, excitotoxicity, and associated oxidative stress may have downstream effects on dopamine by affecting key elements of its synthetic cascade (Cunnington and Channon, 2010; Felger and Treadway, 2017), we examined whether the ratio of transcripts for protective versus excitotoxic NMDA receptor subunits (GRIN2a/GRIN2b ratio) (Costa et al., 2012; Liu et al., 2007) correlated with dopamine in IFN-a-treated animals. A positive relationship was observed whereby animals with a lower protective GRIN2a/GRIN2b ratio also had lower dopamine (Costa et al., 2012) (r = 0.78, df = 5, p < 0.05)(Fig. 4A), and a statistical trend for this relationship was found also by PCR (r = 0.77, df = 5, p = 0.07). Notably, indices of glutamate excitotoxicity (GRIN2A/GRIN2B ratio) and neurotoxic kynurenine metabolism (HAAO, encoding the enzyme for quinolinic acid synthesis, an NMDA receptor agonist (Vancassel et al., 2018)) were both positively associated with genes involved in PI3K signaling (ITPKB, ITPKC, and IP6K1 correlated with the GRIN2A/GRIN2B ratio, while PDPK1 and PIK3CG correlated with HAAO expression; all r > 0.64, p < 0.05). These findings support the idea that IFN-a-induced activation of PI3K signaling may serve as a converging mechanism for neurotoxic kynurenine metabolism and glutamate excitotoxicity, with downstream effects on dopamine. Consistently, we found that mediators of inflammatory cytokine (the IFN-a-increased gene IL6ST) and PI3K (PIK3CG linked to kynurenine metabolism per above) signaling were directly related to low dopamine levels in the putamen of IFN-a-treated animals (r = −0.81, df = 5, p < 0.05 and r = −0.97, df = 5, p = 0.0003 respectively)(Fig. 4B–C).

Fig. 4.

Exploratory analyses revealed associations between neurotransmitter, excitotoxicity, and inflammation-related genes and tissue content of dopamine in the putamen. The ratio of transcripts for protective versus excitotoxic N-methyl-D-aspartate (NMDA) receptor subunits (GRIN2a/GRIN2b) was positively correlated with dopamine in animals administered interferon-alpha (IFN-a)(A). Gene expression of mediators of inflammatory cytokine and IFN-a-induced PI3K signaling (the genes IL6ST and PIK3CG) were inversely associated with dopamine (B-C). Consistent with IFN-a-induced decreases in dopamine receptor-mediated signaling, an association was found between expression of the dopamine- and cAMP-regulated neuronal phosphoprotein DARPP32 that was significantly downregulated by IFN-a and tissue dopamine concentrations (D).

In light of previous neuroimaging findings of IFN-a-induced decreases in DRD2 binding by PET imaging in the putamen and caudate of IFN-a-treated rhesus monkeys (Felger et al., 2013b), targeted analysis examined the effect of IFN-a versus control on DRD2 gene expression. DESeq2 results revealed nominal trends toward decreased DRD2 expression in the putamen (fold change: 0.88, p = 0.14) and caudate (fold change: 0.83, p = 0.1). Decreased DRD2 gene expression was confirmed by PCR where a region-by treatment interaction (F[1,9] = 5.44, p < 0.05) indicated decreased DRD2 in putamen (p < 0.05) but not caudate (p = 0.47). Next, expression of genes that are regulated by dopamine receptor signaling was examined in the putamen. Consistent with lower dopamine signaling through its receptors, DARPP32 (gene name PPP1R1B), a dopamine receptor-regulated gene, was significantly downregulated by IFN-a and contributed to the enrichment of a dopamine-related downregulated pathway (per Table S2B). Dopamine levels also positively correlated with the expression of DARPP32 (r = 0.79, df = 5, p < 0.05)(Fig. 4D).

4. Discussion

Examination of postmortem tissue from IFN-a-treated rhesus monkeys revealed that dopamine, but not serotonin or norepinephrine, was significantly decreased in the NAcc and putamen of IFN-a-treated animals. IFN-a treatment led to differential expression of thousands of whole genome-significant genes specifically in the putamen involving inflammatory and synaptic pathways, shifts in intracellular metabolism, and neurotransmitter mechanisms relevant to glutamate and dopamine receptor signaling, which correlated with decreased tissue content of dopamine. Findings support previous data from our lab and others indicating that the striatum is particularly sensitive to the effects of peripheral inflammation (Capuron et al., 2007; Dowell et al., 2016; Harrison et al., 2015; Menard et al., 2017), and that this heightened immune sensitivity may drive decreased dopamine availability (Capuron et al., 2012a; Felger et al., 2015; Felger et al., 2013b; Felger and Treadway, 2017). These data also provide mechanistic information and translational implications in support of the use of therapies that modify aspects of the inflammatory response, improve metabolic function, or that target glutamate or dopamine receptor signaling to reverse the behavioral effects of increased inflammation in patients with depression.

Consistent with previous findings in IFN-a-treated monkeys (intact dopamine transporter binding by PET imaging and reversal of IFN-a-induced decreases in striatal dopamine release by levodopa)(Felger et al., 2015; Felger et al., 2013b) and humans (increased uptake of [18] F-fluorodopa with PET)(Capuron et al., 2012a), we found no evidence of a change in TH⁺ dopaminergic terminals. Additionally, no change in the number of neuron cell bodies (NeuN⁺ nuclei) in striatum was observed, further suggesting lack of neurodegenerative processes. Consistently, case reports of patients who developed a Parkinson’s-like syndrome following chronic IFN-a indicate a functional reversibility upon treatment discontinuation or with levodopa administration (Bersano et al., 2008; Wangensteen et al., 2016), implicating IFN-a-induced dopamine deficiency, rather than progressive dopaminergic neurodegeneration. Reduced tissue content of dopamine without loss of dopamine terminals supports the hypothesis that inflammation primarily reduces tissue content of dopamine or other monoamines through limiting key aspects of their synthesis (Cunnington and Channon, 2010; Felger et al., 2013a; Kitagami et al., 2003). Specifically, inflammatory stimuli may impair uptake (Vumma et al., 2017) and synthesis of tyrosine (Felger et al., 2013a), dopamine’s primary amino acid precursor, with the latter mechanism involving inflammation-induced diversion of the essential cofactor tetrahydrobiopterin toward oxidative processes and away from monoamine synthesis (i.e., conversion of phenylalanine to tyrosine), leading to the development of interferon-induced fatigue (Felger et al., 2013a). Moreover, relationships between peripheral inflammation and increased monoamine reuptake (thus depleting monoamine availability) have been uncovered by PET imaging of the serotonin transporter in the brain (Krishnadas et al., 2016; Yang et al., 2021). Intracellularly, cytokines increase serotonin transporter activity via p38 MAPK signaling (Gajeswski-Kurdziel et al., 2024) with some evidence of MAPK pathway involvement in dopamine transporter activity (Moron et al., 2003). While a trend for decreased serotonin was only seen in NAcc (the striatal sub-region with the highest serotonin concentration, albeit ~ 10 times lower than that of dopamine), other inflammation-sensitive regions where serotonin is a more prominent neurotransmitter may be equally susceptible to the inflammatory and oxidative effects of IFN-a (Lu et al., 2012).

RNA-Seq profiling of the caudate, putamen, SNC, and prefrontal cortical subregions revealed subtle (nominally significant) IFN-a-induced increases in genes enriching inflammatory signaling pathways in all regions, including a greater number of these genes differentially regulated in BA25 (subgenual cingulate) than other cortical regions, consistent with post-mortem depression literature (Labonte et al., 2017). The most pronounced effects of IFN-a were in the putamen, as evidenced by 2392 whole genome-significant genes. In contrast with the acute effects of IFNs on murine brain gene expression (Hobson et al., 2023; Wang et al., 2008), following chronic peripheral IFN-a administration in rhesus monkeys we did not observe upregulation of canonical IFN-inducible genes and pathways such as STAT1. Instead, non-canonical IFN-a downstream pathways (Platanias, 2005) including PI3K and MAPK were strongly enriched. Interestingly, in patients receiving IFN-a therapy, early activation of p38 MAPK as well as longitudinal modulation of a MAPK-related pathway (ERK5) in peripheral immune cells were associated with subsequent development of depression and fatigue (Felger et al., 2011a; Hepgul et al., 2016). Enrichment of inflammatory (e.g., IL-6) signaling through PI3K/MAPK cascades was also evident in the putamen and driven by inflammation-specific genes like IL6ST encoding gp130, mirroring previous whole blood mRNA findings from IFN-a-treated patients (Hepgul et al., 2016). Therefore, this result extends our previous finding of elevated cerebrospinal fluid IL-6 following IFN-a therapy in patients (Raison et al., 2009) by demonstrating increased IL-6 signaling in brain tissue. Although the brain cell type origin of these inflammatory signals cannot be determined with certainty herein, in the context of IFN-a stimulation, they are likely derived from activated microglia (Hobson et al., 2023; Martins et al., 2022; O’Connor et al., 2009).

Accordingly, IFN-a induced transcriptional evidence of shifts away from oxidative phosphorylation (OXPHOS) and toward increased glycolysis (e.g., insulin, HIF-1 signaling), metabolic alterations that are associated with microglia/macrophage responses to inflammatory stimuli (Lauro and Limatola, 2020) and partly observed in peripheral immune cells of patients with depression and high inflammation in relation to symptoms of anhedonia and psychomotor slowing (Bekhbat et al., 2021; Bekhbat et al., 2020). Additional IFN-a-downregulated pathways included ribosomal proteins involved in cellular growth (Cheng et al., 2019), consistent with IFN-a’s known antitumor effects including inhibition of cell growth (Grander et al., 1997). IFN-a-induced upregulation of components of the PI3K/AKT and MAPK/ERK signaling cascades also contributed to the enrichment of growth and survival pathways (e.g., EGFR, BDNF). Together, these results are consistent with gene signatures of cellular metabolism and growth in peripheral blood immune cells associated with development of fatigue in IFN-a-treated patients (Thomas et al., 2024), which may reflect IFN-induced inflammatory activation of PI3K-AKT, mTOR, and MAPK components that are utilized by growth and differentiation factors (Platanias, 2005). However, in the brain, it may also reflect increased signaling through glutamate receptors, and/or compensatory shifts in response to low dopamine via signal transduction pathways downstream of G protein-coupled dopamine receptors, considering interactions of these neurotransmitter receptors with PI3K-related BDNF/TrkB/GSK3 cascades (Beaulieu et al., 2011; Seillier et al., 2022; Zhao et al., 2022).

Given that IFN-a exerted a multitude of effects on synaptic, metabolic, and neurotransmitter-relevant gene pathways as described above, we attempted to isolate those related to IFN-a’s impact on dopamine by examining Modules Turquoise and Blue, which comprised genes associated with both IFN-a and dopamine levels. Pathway analysis of these Modules suggested that IFN-a-induced decreases in dopamine were related to upregulation of synaptic signaling (Regulation of intrinsic membrane properties and excitability of striatonigral medium spiny neurons in Huntington’s disease; Synaptic signaling pathways associated with autism spectrum disorder) and glutamatergic (Glutamatergic synapse; NMDA-dependent postsynaptic long-term potentiation in CA1 hippocampal neurons) pathways. Consistent with IFNs’ ability to regulate synaptic transmission and ion channels (Defaye et al., 2024; Liu et al., 2016), components of synaptic signaling such as voltage-gated ion channels (CACN, SCN2A, KCNQ, KCNMA1 genes), membrane transporters (SLC24A2), and proteins promoting presynaptic vesicle release (SV2C, ERC2) may have been upregulated to compensate for low dopamine or potentially exert protection against inflammation (Liu et al., 2016; Zhou et al., 2020). Further associations between gene pathways representing cellular energetic and mitochondrial changes and both IFN-a treatment and low dopamine support the view that they may not be simply the consequences of inflammation, but rather involved in the neurotransmitter and plasticity changes that contribute to behavioral change (Chiu et al., 2017; Lacourt et al., 2018; Sforzini et al., 2024; Vichaya et al., 2016).

Primary translational findings from this study were relationships between IFN-a-induced reductions in dopamine and genes reflective of excitotoxic glutamate, cytokine, and altered dopamine receptor signaling. Increased gene expression in NMDA-mediated glutamate pathways, a reduced protective GRIN2a/GRIN2b ratio (Costa et al., 2012; Liu et al., 2007) and upregulation of PI3K components (PIK3CG) linked to neurotoxic kynurenine metabolism, in IFN-a-treated animals with low dopamine indicates that, in addition to the multitude of ways that cytokines can increase glutamate, a shift toward excitotoxic glutamate receptor and kynurenine signaling may contribute to reduced dopamine availability in the context of inflammation (Felger and Treadway, 2017). Consistent with this hypothesis, inhibition of NMDA signaling with memantine was shown to prevent inflammation-related decreases in dopamine in the putamen of non-human primates infected with simian immunodeficiency virus (Meisner et al., 2008). Therefore, therapies that block glutamate signaling may confer benefits for patients with high inflammation and depression (Kiraly et al., 2017; Yang et al., 2015). In accordance with previous findings (Felger et al., 2013b), IFN-a reduced DRD2 expression, and decreased dopamine in IFN-a-treated animals also correlated with lower expression of genes encoding signal transduction proteins downstream of dopamine receptors such as DARPP32. Dopaminergic therapies, particularly those that agonize DRD2, may be therapeutic in reversing the effects of inflammation on dopamine-relevant circuits in the brain that drive symptoms of reduced motivation and motor slowing (Bekhbat et al., 2022; Felger et al., 2016). Moreover, inflammatory pathways for IFNs and ILs and their signal transducers, as reflected by PIK3CG and IL6ST expression, associated with lower dopamine in IFN-a-treated animals. Anti-inflammatory strategies to reduce peripheral inflammation, such as cytokine antagonists, are obvious choices for reducing inflammation in the body to benefit the brain (Hodes et al., 2016; Miller et al., 2017) and may prevent striatal dopamine loss (Meulendyke et al., 2012). Finally, PI3K/ATK/mTOR and related metabolic pathways (e.g., HIF-1) may serve as modifiable targets for therapies such as rapamycin that inhibit mTOR, oxidative stress, and/or glycolytic processes while promoting OXPHOS (Bekhbat, 2024; Singh et al., 2023; Tye et al., 2022).

There are several limitations of this study. First, we did not have the power to examine both the neurobiological and immunologic effects of IFN-a and administer an interventional treatment to reverse these effects within the available sample (n = 11). Thus, our non-human primate model provided translational relevance for peripheral inflammation’s effects on the brain but was limited by the experimental design. Also, due to prioritization of HPLC for limited tissue from the NAcc, RNA-seq was only performed in the putamen and caudate. IFN-a, however, exerted similar effects on dopamine within the NAcc as putamen and thus should reflect similar inflammatory and neurotransmitter-related changes. The sample employed herein was small but consistent with our previous work and with that of other non-human primate studies employing 4 to 8 animals per group (Beckman et al., 2024; Czoty et al., 2000; Felger et al., 2007; Felger et al., 2015; Felger et al., 2013b; Sawyer et al., 2012), even for RNA-seq (Hawes et al., 2022; Raper et al., 2016; Sandler et al., 2014). This study also employed bulk RNA-seq. Cellular sources of IFN-a-induced neurotransmitter, cytokine, and metabolic changes can be probed by single cell-based approaches in future work. Despite these limitations, IFN-a exerted potent effects on the putamen that uncovered both statistically and translationally significant results.

5. Conclusions

Together these findings provide a window into the immunologic and neurobiological mechanisms and consequences of peripheral inflammatory cytokine effects on striatal dopamine, which may have translational implications for patients with high inflammation. Primary results support prior findings that the NAcc and putamen are particularly sensitive to the effects of peripheral inflammation, suggesting vulnerability of reward and motor domains. Although behavioral measurements were not available in this study, the observed dopamine reductions replicate previous findings in this model, where IFN-a decreased striatal dopamine release in association with reduced reward sensitivity and anhedonia (Felger et al., 2013b). Furthermore, the pronounced transcriptomic changes in the putamen mirror the Parkinson’s hallmark of putamen-heavy dopamine dysfunction underlying greater motor deficits (Broussolle et al., 1999; de la Fuente-Fernandez, 2013; Manza et al., 2016; Otsuka et al., 1996), and align with our previous finding of reduced in motor speed in IFN-a-treated patients (Majer et al., 2008). RNA-seq uncovered evidence that IFN-a increased glutamate and decreased dopamine receptor signaling, an effect that was verified by low tissue content of dopamine in IFN-a-treated animals correlating with shifts in glutamate and dopamine receptor-related gene expression. Further mechanistic studies in other preclinical models (e.g., midbrain organoids) and humans investigating how IFN-a directly affects dopamine signaling and linking neurochemical changes to individual behavioral outcomes will enhance the translational relevance of these findings. These data provide molecular and cellular support for our previous work and hypotheses in patients with high inflammation and depression suggesting that treatments that target inflammation, metabolism, glutamate or dopamine receptor signaling may provide therapeutic benefit for these individuals.

Data availability

Data will be made available on request.