Effects of anti‐inflammatory drugs on the expression of tryptophan‐metabolism genes by human macrophages

Abstract Several lines of evidence link macrophage activation and inflammation with (monoaminergic) nervous systems in the etiology of depression. IFN treatment is associated with depressive symptoms, whereas anti‐TNFα therapies elicit positive mood. This study describes the actions of 2 monoaminergic antidepressants (escitalopram, nortriptyline) and 3 anti‐inflammatory drugs (indomethacin, prednisolone, and anti‐TNFα antibody) on the response of human monocyte‐derived macrophages (MDMs) from 6 individuals to LPS or IFN‐α. Expression profiling revealed robust changes in the MDM transcriptome (3294 genes at P < 0.001) following LPS challenge, whereas a more limited subset of genes (499) responded to IFNα. Contrary to published reports, administered at nontoxic doses, neither monoaminergic antidepressant significantly modulated the transcriptional response to either inflammatory challenge. Each anti‐inflammatory drug had a distinct impact on the expression of inflammatory cytokines and on the profile of inducible gene expression—notably on the regulation of enzymes involved in metabolism of tryptophan. Inter alia, the effect of anti‐TNFα antibody confirmed a predicted autocrine stimulatory loop in human macrophages. The transcriptional changes were predictive of tryptophan availability and kynurenine synthesis, as analyzed by targeted metabolomic studies on cellular supernatants. We suggest that inflammatory processes in the brain or periphery could impact on depression by altering the availability of tryptophan for serotonin synthesis and/or by increasing production of neurotoxic kynurenine.


INTRODUCTION
Autoimmune and inflammatory diseases are commonly associated with mood disorders and several lines of evidence indicate inflammation may give rise to or exacerbate them. 1,2 Studies of animals exposed to proinflammatory challenges, ranging from LPS administration Abbreviations: COX, cyclooxygenase; MDD, major depressive disorder; MDMs, monocyte-derived macrophages; SSRI, selective serotonin reuptake inhibitors; TCA, tricyclic antidepressant TNF blockade improves depressive symptoms in patients with rheumatoid arthritis 8 and in a subgroup of patients with MDD and elevated levels of C-reactive protein (an acute phase protein). 9 One plausible sequence of events is that peripheral proinflammatory cytokines such as TNF-, IL-1 , and IL-6, induced the expression of enzymes involved in tryptophan catabolism, for example, indoleamine dioxygenase, kynurenine hydroxylase, and kynureninase, 10 thereby reducing the availability of synaptic serotonin [11][12][13] ; which is a proximal cause of depressive symptoms. 14 Inflammation-induced changes in tryptophan metabolism can also lead to increased synthesis of kynurenine and its metabolites, many of which are known to be glutamatergic agonists and/or neurotoxic. 15 Thus, inflammation could produce the "double hit" both reducing the availability of serotonin and increasing production of kynurenine.
The same mechanisms may explain the clinical observation that patients with depressive symptoms in the context of peripheral inflammation ("inflamed depression") are less responsive to monoaminergic antidepressant drugs, so called "treatment-resistant depression." 16 Selective serotonin reuptake inhibitors (SSRIs), for example, are thought to increase the synaptic availability of serotonin (by blocking active uptake by presynaptic neurons expressing the serotonin transporter protein (SERT/SLC6A4)). 17 If synaptic serotonin is reduced by an inflammatory response, and SSRIs are not anti-inflammatory, 18 SSRIs are likely to be less effective in the presence of an inflammatory stimulus. Additionally, the neurotoxic effects of kynurenine and its metabolites, which are exacerbated by inflammation, are mediated by glutamatergic-related mechanisms that are not modulated by monoaminergic antidepressants. Treatment-resistant depressiondue to reduced serotonin availability and increased neurotoxicity by non-serotonergic mechanisms-is thus predictable in the context of peripheral inflammation. 12,13,19,20 Here we report on a series of experiments designed to test the mechanistic connections between inflammatory and monoaminergic systems in a human primary cell model, monocyte-derived macrophages (MDMs). Macrophages are both initiators and mediators of inflammation-associated pathology 21 and have long been implicated in inflammation induced depression, 20 largely based on the inflammatory induced activation of tryptophan depletion pathways. 10,19,20 Microglia, the macrophages of the brain, 22 play a crucial role in neuronal homeostasis, and the impairments in neuronal function associated with many clinical disorders, including major depression. 23 Unlike human macrophages, mouse macrophages do not induce tryptophan uptake or enzymes associated with tryptophan metabolism in response to proinflammatory signals, 24 making them a poor model for such studies. In the current study, we have used human MDMs, differentiated by CSF1, to study the impacts of 2 antidepressant drugs and 3 anti-inflammatory drugs on the transcriptional response to 2 proinflammatory challenges: LPS and IFN . In addition to acting as a primary human cell model for microglia, these cells are also used to examine tryptophan depletion mechanisms during peripheral inflammation, and how drugs may influence this effect.
LPS is a well-studied Gram-negative bacterial cell wall endotoxin that acts through activation of the TLR4 receptor, an archetypal pattern recognition receptor that signals through two well-defined com-plementary signaling pathways to induce proinflammatory cytokines and IFN target genes. 25,26 As part of the FANTOM5 consortium, we have recently generated extensive promotor-level data on the time-course of the response of MDM. 27 IFN is a type-1 IFN and was chosen as a stimulus because of the clinical data indicating that this agent can cause a depressive syndrome in patients. 6,7 We have previously compared and contrasted the transcriptional response of mouse macrophages to type-1 IFN and LPS treatment. 28 Three classes of anti-inflammatory drugs were selected to examine their effects on both inflammatory stimuli: a neutralizing antibody for the inflammatory cytokine TNF ; a nonsteroidal antiinflammatory drug (indomethacin); and a steroidal anti-inflammatory drug (prednisolone). Neutralizing anti-TNF antibodies act by binding to TNF in circulation, thus blocking its proinflammatory effects mediated by action at TNF receptors. 29 TNF acts in an autocrine manner on stimulated macrophages to induce downstream targets and amplify the initial induction of proinflammatory target genes. 27,30 Indomethacin acts by inhibiting cyclooxygenase (COX) or prostaglandin synthase enzymes (COX1/COX2; PTGS1/PTGS2). 31 The anti-inflammatory activity of COX inhibitors is thought to depend on preventing inducible prostaglandin production, which can also act in an autocrine manner in macrophages through inducible prostaglandin receptors. 25 Both COX-inhibiting and TNF -inhibiting drugs have been shown to have some antidepressant and anxiolytic effects in patients with psychologic symptoms in the context of medical inflammatory disorders. 32 Synthetic glucocorticoids, such as prednisolone, are amongst the most commonly prescribed anti-inflammatory agents, and act by inducing multiple feedback repressors of inflammation, including I B and DUSP1. 33 The responses to glucocorticoids also differ radically between humans and mice, due to the gain and loss of glucocorticoid response elements in enhancers, 33 so there is a clear need to study their effects in humans systems. The glucocorticoid ligand/receptor complex may also directly repress signaling by interfering with the activation of the inflammatory transcription factor NF-B 34 leading to a genome wide blockade of NF-B interactions with chromatin. 35 We also studied the action of two monoaminergic antidepressant drugs on MDMs, a SSRI (escitalopram) and a tricyclic antidepressant (TCA; nortriptyline). SSRIs are the most commonly prescribed drugs for treating depression and their efficacy depends at least partly on increasing synaptic availability of serotonin (5-HT). However, 5-HT also plays an important role in immune signaling, 36 and SSRIs have been shown to enhance the cytolytic function of NK cells, to enhance B cell numbers, and to inhibit 5-HT uptake and immune signaling by dendritic cells, 37 suggesting that immune mechanisms might also contribute to their therapeutic efficacy. The TCAs are less selectively serotonergic, and also block multiple classes of receptors for acetylcholine, histamine and noradrenaline. The use of TCAs in the treatment of the residual symptoms of inflammatory bowel disease, 38 suggests they may have some efficacy as anti-inflammatory agents. 39 Supplemental Fig. S1 illustrates the known mode of action of each drug examined here.
We analyzed the gene expression profiles of MDM in response to inflammatory challenge in the presence and absence of anti-inflammatory or monoaminergic antidepressant drug treatment.
Network-based methods were deployed to represent the expression changes induced by inflammatory challenges, and the modulation of response by each of the 5 drugs tested. We tested the specific prior hypotheses (i) that proinflammatory challenges will cause changes in expression of genes related to tryptophan metabolism; (ii) that monoaminergic antidepressants (nortriptyline, escitalopram) will attenuate inflammation-induced changes in tryptophan-related genes; or (iii) that anti-inflammatory drugs (prednisolone, indomethacin, anti-TNF antibody) will attenuate the regulation of tryptophan-related genes; (iv) that inflammatory activation of human macrophages results in decreased production of tryptophan and increased production of kynurenine; and (v) that anti-inflammatory drugs can attenuate changes in tryptophan metabolism and kynurenine production.

Ethics and donors
Human CD14 + mononuclear cells were isolated from fresh blood of volunteer donors under ethical approval from Lothian Research Ethics Committee (11/AL/0168).

Cell culture
Human peripheral blood monocytes were isolated from 320 mL blood samples by Ficoll gradient separation of buffy coats followed by MACS CD14 + selection (Miltenyi Biotec Ltd., Bisley, UK). They were then cultured at 5 × 10 5 cells/well in 1 mL on a 12-well plate in RPMI supplemented with penicillin/streptomycin, glutamax (Invitrogen, Loughborough, UK), and 10% fetal calf serum for 7 d in the presence of rhCSF-1 (a gift from Chiron, Emeryville, CA) at 10 4 U/mL to produce MDM. All donors were medically healthy and between 20 and 50 years old (3 male, 3 female). IFN (SRP4594; Sigma-Aldrich, Gillingham, UK) was used at 50 U/mL, within the range reported to be present in blood serum following IFN -2b/Ribavirin therapy for hepatitis C patients. 40 LPS from Salmonella enterica serotype minnesota (Re 595, L9764; Sigma-Aldrich) was used at 10 ng/mL, which is just maximal for inducible proinflammatory gene expression. 28

Optimization of inflammatory stimulation and drug concentrations
Cell viability was measured to optimize the concentrations of drugs used (other than anti-TNF ). This was performed using CellTiter-Glo  43,44 The amount of anti-TNF antibody used was calculated to be 5 g/mL. This was based on the suppliers ND 50 value of 0.01-0.04 g/mL in the presence of 0.75 ng/mL TNF and our own measurements of maximum TNF production from stimulated macrophages (<30 ng/mL). We then used the following equation to calculate the amount of antibody required: [Complex] . IFN (SRP4594, Sigma-Aldrich) was supplemented at 50 U/mL as this concentration yields a robust inflammatory response in F I G U R E 1 Study design and differential gene expression in response to inflammatory challenge. (A) Schematic representation of experimental design. Fully differentiated human MDMs (day 8) were generated from 6 individuals aged 20-30 years, 3 male and 3 female. Cells were then pretreated with either escitalopram, nortriptyline, an anti-TNF antibody, indomethacin, prednisolone, or controls (a nonspecific IgG antibody with DMSO vector), or untreated. Cells were then exposed to LPS or IFN challenge and harvested at either 7 or 24 h, or cultured for 24 h with no inflammatory stimulus. (B) Numbers of significantly differentially expressed genes (DEGs) following inflammatory stimulation (P < 0.001). (C) Venn diagrams displaying the overlap of DEGs between IFN and LPS challenges at early or late time points macrophages and is within the range reported to be present in blood serum following IFN -2b/Ribavirin therapy for hepatitis C patients. 40 Although many in vitro studies typically use 100 ng/mL of LPS, we used

Quantitation of selected neurotransmitter metabolites
To confirm the predicted metabolic effects of these treatments on the expression of genes related to tryptophan metabolism, tryptophan and kynurenine concentrations were measured by HPLC/MS in the supernatant of cells after 24 h incubation (Supplemental Materials).

Quantitation of cytokine production in sample supernatants
To compare protein production to transcript levels, cytokine production was measured in the supernatant of cells by use of TNF-(#KAC1751) and IL-6 (#KHC0061C) ELISA kits (ThermoFisher, Runcorn, UK). Precoated plates were used according to manufacturer's instructions. Briefly, plates were blocked for nonspecific binding. Supernatants, or standard controls, were then allowed to bind to the precoated antibodies before washing and addition of a secondary HRP-conjugated antibody. Absorbance was read at 450 nm by a plate reader and cytokine concentration was calculated from the standard curve.

Transcriptional regulation in MDM by LPS and IFN
LPS significantly modulated the expression of 3294 genes and IFN challenge 743 genes in human MDM, consistent with previous observations. 24,27,50 Both challenges showed the greatest effect at 7 h ( Fig. 1 and Supplemental Table S1). As previously observed in mouse macrophages, 28 IFN responsive genes were largely a subset of the LPS response: approximately 2 out of 3 of the IFN responsive genes were also differentially expressed following LPS treatment.
A sample-to-sample correlation network (Fig. 2)  Genes with a similar expression profile tended to group closely together. Regulated genes grouped into 9 major clusters of strongly coexpressed genes: 4 clusters of genes that were up-regulated and 5 clusters of down-regulated genes (Fig. 2D).

Effects of monoaminergic antidepressant drugs on whole transcriptome
At the nontoxic doses used in this study, neither antidepressant drug had a significant effect on the macrophage response to either LPS or IFN (Supplemental Fig. S3).  Table S1 for a summary and Supplemental Table S3 for details. The effect of indomethacin was more complex than simply inhibiting the responses to LPS and IFN , consistent with the known ability to prevent expression of prostaglandins, which act as feedback inhibitors of gene expression. 52 In many cases, it appeared to further elevate the expression of genes already up-or down-regulated, particularly those associated with lipid metabolism.

Regulated expression of genes involved in tryptophan metabolism
Amongst the many genes that demonstrated a significant interaction between the two inflammatory stimuli and drug treatment, we focused on pathways for tryptophan catabolism and transport. This pathway relates to the prior hypothesis that tryptophan metabolism could be a key mechanism linking peripheral inflammation to depression. 10,20,53 A schematic representation of this metabolic pathway is illustrated in Fig. 4. LPS also caused significant down-regulation of genes coding kynurenine metabolic enzymes (AFMID and CCBL7). Other tryptophan transporters in human macrophages, the heavy chain Y+L SLC3A2 (CD98) and SLC7A7 are expressed constitutively at high levels. 27 IFN challenge resulted in a similar profile of effects on the expression of genes encoding the kynurenine metabolic enzymes, but did not produce significant elevation of the tryptophan transporter genes (Fig. 4).
Anti-inflammatory drugs significantly modulated the on the inducible expression of genes encoding serotonin transporters and kynurenine metabolic enzymes. Interestingly, the direction of change in gene expression caused by anti-inflammatory drug treatment was always opposite in sign to the change caused by activating stimulus. In other words, 4 of the 5 genes that were up-regulated in response to challenge (SLC16A10, SLC7A5, IDO1, and KYNU), were less strongly induced after drug treatments; and all 3 genes that were downregulated in response to challenge (SLC7A8, AFMID, and CCBL7) were less strongly repressed (Fig. 4).
Thus, all anti-inflammatory drugs tested were found to impact tryptophan-or kynurenine-related gene expression. Indomethacin had effects mainly on tryptophan transporter genes; prednisolone on kynurenine metabolism genes; and anti-TNF on both class of genes.

Effects of macrophage activators (LPS, IFN ) and drugs on tryptophan-related metabolite concentrations
In keeping with the gene expression data, LPS treatment led to 70-80% depletion of tryptophan in the medium and accumulation of kynurenine. IFN exerted a similar effect, but the magnitude of change in these metabolites was less. Of the three anti-inflammatory drugs tested, only indomethacin significantly attenuated LPS-induced reduction in tryptophan availability (Fig. 5). Both anti-TNF and indomethacin significantly reduced kynurenine production to ∼70% and ∼50% following IFN stimulation but did not reduce kynure-nine following LPS treatment (Fig. 5). The kynurenine to tryptophan (Kyn/Trp) ratio was used as a measure of tryptophan catabolism overall (Fig. 5). Anti-TNF treatment significantly moderated IFN -induced increases in Kyn/Trp (P = 0.018); and indomethacin treatment likewise significantly moderated LPS-induced increases in the ratio of kynurenine to tryptophan (P = 0.026).

DISCUSSION
Evidence for a link between systemic inflammatory disease and depression (or depressive behavior in animals 3,5 ) is now overwhelming. 2 We investigated the effects of LPS or IFN challenge on gene expression by CSF1-cultured MDMs. The scale of the transcriptional response to LPS and IFN in these data was comparable to previous studies. LPS is itself known to induce type 1 IFNs (mainly IFN ) in MDM, which act in an autocrine manner via IFNAR1, itself induced by LPS. 27 Accordingly, the response to exogenous type 1 IFN (IFN ) was largely a subset of the LPS response, as reported in previous studies. 24,28,50,51,54 None of the anti-inflammatory agents we examined completely prevented the response to either challenge and, although prednisolone and anti-TNF both reduced expression Both stimuli greatly increased expression (∼3000 fold following LPS) of the gene encoding the enzyme indoleamine 2,3-dioxygenase (IDO1). IDO1 catalyzes the primary reaction in conversion of tryptophan to kynurenine; thereby reducing its availability as a precursor for serotonin metabolism. While this process has been F I G U R E 4 Effects of proinflammatory challenge and anti-inflammatory drugs on expression of genes related to tryptophan and kynurenine metabolism. The pathway model summarizes the metabolic role of 8 proteins involved in tryptophan transport or catabolism as illustrated. The profile of expression changes for each gene following proinflammatory challenge is shown by the line graphs during each drug treatment. Significance is depicted for every significant interaction of anti-inflammatory drug treatment with proinflammatory challenge, and color coded for each drug treatment. *P < 0.05, **P < 0.01 associated with immunosuppression through Treg activation and effector T cell suppression, 56 inflammation-induced expression of this enzyme leads to tryptophan depletion both locally and systemically in chronic inflammatory states and has also been linked to alterations in mood. 5,53,57 Similarly, the ability of host microbiota to control tryptophan metabolism has been functionally linked to influencing mood. 58 Furthermore, polymorphisms in IDO1 have been associated with susceptibility to IFN -induced depression in hepatitis patients. 59 This may arise both from the depletion of tryptophan, and the generation of neurotoxic metabolites such as kynurenine. 5,60 However, tryptophan depletion in depressed patients was reportedly independent of kynurenine pathway activation. 61

F I G U R E 5 Metabolomic analysis of effects of inflammatory challenge and anti-inflammatory drug treatment on tryptophan concentration
and tryptophan/kynurenine concentration ratio. Tryptophan (A) and kynurenine (B) concentrations in cell culture supernatants 24 h after pretreatment of the 3 anti-inflammatory drugs before each inflammatory challenge. (C) Ratio of kynurenine to tryptophan levels detected in the supernatant. Significance was calculated using a ratio paired t test. *P < 0.05 **P < 0.001 Anti-TNF reduced transcription of tryptophan catabolism enzymes, IDO1 and KYNU, following IFN stimulation but no significant effect of the drug was observed on these elements following LPS treatment. This is likely due to the transient induction of feed-forward and negative feedback loops of TNF transcription following inflammatory stimulus. Like IFN , TNF was predicted to be involved in an autocrine loop in MDM elicited by LPS, since the TNF receptor is also induced. 27 Our data (Supplemental Table S2) confirm and extend evidence of autocrine TNF signaling as a feed-forward activator of macrophage gene expression 30,62,63 and identify the subset of inducible genes dependent upon that stimulus. The lack of impact of anti-TNF on LPS-inducible IDO1 may reflect the magnitude of the response. Alternatively, activation of IDO1 by IFN-requires co-stimulation by TNF , which increases the occupancy of IFNresponse elements. 64,65 It may be that LPS provides this second signal independently of TNF . 30,62,63 A study examining the treatment of primary murine hippocampal cells with ibuprofen, a nonselective COX inhibitor like indomethacin, identified TDO2 as the most significantly affected gene. 66  To directly assess tryptophan uptake and catabolism following each treatment, we measured the concentrations of tryptophan and kynurenine in the supernatant. Indomethacin was the only inhibitor that impacted tryptophan levels during LPS challenge, likely due to a transcription repression of the tryptophan transporter. Prednisolone, while modulating expression of numerous inducible genes, did not produce any alteration in overall tryptophan catabolism (Supplemental Fig. S5), despite reducing expression of tryptophan catabolism-related genes such as IDO1, KYNU, and CCBL1.
A recent review examined the anti-inflammatory effects of antidepressant drugs. 67 We found no evidence that either of the monoaminergic antidepressant drugs tested (TCA or SSRI) exerted any effect on the response in MDMs to LPS or IFN . This is in contrast to reports that TCAs reduce proinflammatory signaling in phagocytes, 68 and SSRIs have been reported to alter macrophage differentiation and inflammatory signaling. 67,69 SSRIs have also been reported to modulate glucocorticoid actions on monocytes. 70 Although we found no direct effect of SSRIs or TCAs on macrophage responses to either LPS or IFN , we did observe significant cytotoxicity at concentrations that were <20-fold lower than circulating blood concentrations in treated patients. In support of this, nortriptyline has previously been shown to induce autophagy in macrophages. 39 Indeed, many antidepressants are cationic amphipaths, and are therefore lysosomotropic 71 likely accumulating selectively in phagocytic cells.
Selective toxicity to macrophages may therefore contribute to the results reported in other studies, which used far higher concentrations of SSRIs than used here. 69 There is already some evidence that anti-inflammatory drugs can have antidepressant efficacy. 32 Our results are compatible with the mechanistic interpretation that this may be at least partly attributable to the effects of anti-inflammatory drugs on "normalization" of an inflammation-induced bias in tryptophan and kynurenine metabolism. However, the results also indicate that this is unlikely to be the sole mechanism, and each agent may produce distinct patterns of regulation of genes that can impact indirectly on neuronal function expression. If each anti-inflammatory agent is distinct in its actions, it may be that combinations would have novel potential efficacy.