A Critical Risk Factor for a Major Side Effect of Interferon-Alpha Therapy: Activated Indoleamine 2,3-Dioxygenase 1 is Related to Depressive Symptoms A Critical Risk Factor for a Major Side Effect of Interferon-Alpha Therapy: Activated Indoleamine 2,3-Dioxygenase 1 is Related to Depressive Symptoms

Hepatitis C virus (HCV) infection affects approximately 170 million people worldwide. Interferon-alpha (IFN-α) is a cytokine that is related to early viral infection and has both antiviral and antiproliferative properties. The current standard treatment for long-term chronic hepatitis C (CHC) consists of combination therapy with IFN-α and ribavirin, which has a broad spectrum antiviral effect. Despite the potential therapeutic benefits of IFN-α, its administration often causes many side effects, such as somatic and neuropsy chiatric symptoms. Depression is a serious and frequently occurring side effect of IFN-α therapy, and this is one of the major reasons for cessation of the therapy. Therefore, in order to avoid the discontinuation of INF-α therapy owing to depressive symptoms, it is important to identify the risk factor(s) leading to the onset of associated depressive symptoms. In this chapter, we introduce our novel findings on the association between IFN treatment and the onset of depression in CHC patients as well as the potential neu robiological mechanisms by which depression may arise. We also highlight a potential approach for predicting the onset risk of depression as a side effect in these patients. pathway, which finally changes in the alternation of 5-HT turnover. Our findings suggest that inflammatory pathways that lead to the activation of IDO1 may be a novel therapeu tic target in patients suffering from inflammation-associated depression, for example, HCV or cancer therapy. Our results also suggest the monitoring of TRP-KYN metabolites during immunotherapy might assist in predicting the onset risk of depression as a side effect in these patients. However, further insight into the role of each downstream KYN pathway metabo lite in the pathological process is needed to understand, and to clarify the relationship with complex neurotransmitters.


Introduction
Hepatitis C virus (HCV) infection is a global health problem. Up to 85% of HCV-infected patients may develop long-term chronic hepatitis C (CHC), a disease state associated with serious clinical sequela, including liver cirrhosis, hepatic fibrosis, and hepatocellular carcinoma [1][2][3][4]. It has been estimated that up to 20% of CHC patients will develop hepatic cirrhosis over a 20-25-year period, and these individuals are at an increased risk for developing end-stage hepatic diseases or hepatocellular carcinoma [4]. Therefore, aggressive antiviral treatments to successfully induce viral remission constitute a major strategy for reducing the morbidity and mortality associated with CHC.
Immunotherapy with interferon-alpha (IFN-α) is commonly used to treat CHC and several types of malignancies owing to its antiviral, antiproliferative, and immunoregulatory effects [5]. In clinical trials, more than 50% of CHC patients treated with combination therapy using IFN-α and ribavirin achieved a sustained viral response, defined as undetectable HCV in the blood 6 months following the end of treatment [4]. Despite the efficacy of IFN-α in CHC treatment, IFN-α therapy causes serious side effects; early signs include somatic symptoms (anorexia, pain, insomnia, fever, and fatigue). Prolonged therapy causes neuropsychiatric symptoms including depressive states, anhedonia, anxiety, and cognitive impairment. In particular, depression is a serious and frequently occurring side effect of IFN-α therapy, and this leads to discontinuation of the therapy in up to 45% of patients [6,7]. Therefore, in order to avoid the discontinuation of IFN-α therapy owing to depressive symptoms induced by the cytokine, it is important to identify the risk factor(s) leading to the associated depressive symptoms.
IDO1 is an extrahepatic enzyme that catalyzes the conversion of TRP to KYN, which can produce many neuroactive metabolites such as 3-hydroxykynurenine (3-HK), kynurenic acid (KA), and quinolinic acid (QUIN). Intriguingly, QUIN levels in CSF have been found to correlate with the severity of depressive pathology [10], and post-mortem studies have shown increased microglia QUIN levels in the frontal cortex of severely depressed patients [11].
In the current chapter, we present the findings of our latest study, which demonstrates the association between IFN treatment and changes in the TRP-KYN pathway in the blood of HCV patients. To do so, we investigated the effect of chronic Ifn gene expression on depression-like behavior and levels of brain TRP-KYN metabolites in mice. Our results suggest the possibility for the prediction of onset risk of depression as a side effect in HCV patients.

Classification of IFNs
IFNs were first introduced in 1957 as antiviral molecules. Based on their receptor types on the cell membrane surface, IFNs are classified into type I and type II. IFN type I mainly consists of IFN-α/β, while IFN type II consists of IFN-γ. IFN type I is a family of cytokines in which their amino acid sequence similarity reaches 30-80%. They are produced by a wide variety of cells, including fibroblasts, epithelial cells, and hepatocytes [12,13]. However, in most viral infections, plasmacytoid dendritic cells (pDCs) are probably the major source of these cytokines. In contrast, IFN type II (IFN-γ) is a single gene cytokine unrelated in structure to IFN-α/β, which is produced largely by macrophages, natural killer (NK) cells, and T lymphocytes [12].

Immune response for HCV infection and IFN induction
The host response against HCV infection is first triggered when a pathogen-associated molecular pattern (PAMP), presented by an infecting virus, is recognized and engaged by specific PAMP receptors expressed on the host cells. This leads to the activation of signals that ultimately induce the expression of antiviral effector genes [14,15] (Figure 1). For RNA viruses, protein, and nucleic acid products of infection or replication have been identified as viral PAMPs. These are engaged by specific toll-like receptors (TLRs) or nucleic acid-binding proteins that serve as PAMP receptors [15][16][17]. The viral RNA of HCV contains each of these PAMP signatures, and is adequate to trigger the host response when introduced into naïve cells [18,19]. In hepatocytes, which is the target cell of HCV infection, independent pathways  [14]. (1) HCV RNA binding to RIG-I or TLR3 results in the activation IRF-3. The dimer of phospho-IRF-3 translocates to the nucleus, interacts with transcription partners and binds to the cognate-DNA PRD in the promoter region of IRF-3 target genes. (2) IRF-3 activation leads to the induction of IFN-β production. (3) Secreted IFN-β from the infected cells binds to the IFN-α/β receptor, and results in activation of the JAK-STAT pathway. The ISGF3 complex translocates to the nucleus, where it binds to the ISRE on target genes to direct ISG expression. IRF-7 is one of the ISGs and it is activated after expression through viral PAMP signaling. (4) The IRF-7 dimer and heterodimer with IRF-3 binds to VRE in the promotor region of IFN-α genes resulting in the production of various IFN-α subtypes and establishing a positive-feedback loop for IFN amplification. It is the IFN-α component of the host response that is exploited by the current IFN-based therapy for HCV infection [14]. of retinoic acid-inducible gene I (RIG-I) and TLR3 signaling construct two major pathways of host defense triggered by double-stranded (ds) RNA [19][20][21]. Viral PAMP binding to RIG-I or TLR3 results in the phosphorylation and activation of interferon regulatory factor 3 (IRF-3) by TANK-binding kinase 1 (TBK-1) and I kappa B kinase ε (IKK-ε) [14,22]. The dimer of phospho-IRF-3 translocates to the cell nucleus, interacts with its transcription partners, including CREB-binding protein (CBP)/p300, and binds to the cognate-DNA positive regulatory domain (PRD) in the promoter region of IRF-3 target genes, such as IFN-β [14,23]. The engagement of PAMP receptors also leads to the synthesis of IFN-α/β, tumor necrosis factor (TNF), and a variety of other cytokines, which are largely produced by mainly pDCs that express TLRs in abundance. IFN-α/β produced by pDCs activates NK cells, thereby enhancing their cytotoxic potential and stimulating their production of IFN-γ. IFN-α/β produced by pDCs also modulates the activation of CD8 + T cells, which produce additional IFN-γ and represent the central players in the pathogen-specific adaptive immune response [12].

The antiviral effect of IFNs on HCV
IFN-α mediates a wide range of biological activities including antiproliferation, immunomodulation, and antiviral responses. IFN-α/β acts to induce the antiviral response in cells. These cells can be far from IFN-α/β production site and IFN-α/β interacts with specific cell surface receptors, type I IFN receptors (interferon-alpha receptor 1 (IFNΑR1) and IFNΑR2; Figure 1). IFNARs signal to the nucleus via Janus kinase-1 (Jak1) and tyrosine kinase 2 (Tyk2) phosphorylation of the signal transducers and activators of transcription (STATs) [24]. The classic IFN-α/β signaling pathways activate STAT1/STAT2 heterodimers and the trimeric IFN-stimulated gene factor (ISGF) complex containing IRF-9, which activate the expression of specific subsets of genes controlled by promoters containing interferon-stimulated response elements (ISRE; Figure 1) [15]. Interferon-stimulated genes (ISGs) are the genetic effectors of the host response, although the details of the signaling mechanisms by which IFN-α/β and IFN-γ induce the transcription of ISGs are still being defined [25]. IRF-7 is a transcription factor and an ISG. It is expressed in many tissue types, including complex liver tissue, in response to IFN. IRF-7 is activated after expression via viral PAMP signaling pathways that overlie with the IRF-3 activation pathway. IRF-7 phosphorylation, dimerization, and heterodimerization with IRF-3 lead to bind its cognate virus-responsive element (VRE) in the promotor region of IFN-α genes. Then, this binding results in the production of various IFN-α subtypes. The transcription effector action of IRF-7 also promotes diversification of the ISG response, establishing a positive-feedback loop that amplifies IFN production, and antiviral action [14]. This increases the plenty of RIG-I and viral PAMP signaling modules whose continued signaling acts to amplify IFN production and the host response. The medicinal administration of IFN-α promotes an antiviral reaction against HCV infection by stimulating ISG expression via the IFN-α/β receptor and the JAK-STAT pathway. In addition to stimulating ISG expression, IFN-α induces or promotes the maturation of immune effector cells, and enhances the production of other cytokines by resident hepatic cells to indirectly modulate the cell-mediated defenses and adaptive immunity to HCV [15]. Viral trigger and control of the host response may elucidate cellular tolerance for HCV RNA replication and influence the outcome of infection.

Side effect: IFN-induced depression and tryptophan metabolism
IFN-α has been shown to develop depression in many diseases, not only CHC, but also in melanoma, chronic myelogenous leukemia, and renal cell carcinoma [26]. However, CHC patients may be more susceptible to developing IFN-induced depression than patients with other disorders, possibly due to a baseline 5-HT system dysfunction. Depression in CHC patients may result from changes in platelet 5-HT function, with decreased 5-HT concentrations during CHC infection compensated for by a decrease in reuptake and metabolism [1]. Immune activation, particularly by IFN-γ, affects the catabolism of TRP, a precursor of 5-HT, by inducing expression of IDO1. IDO1 is the first and rate-limiting enzyme that converts TRP to N-formyl-L-kynurenine, which is further metabolized to QUIN (Figure 2). IFN treatment of CHC patients results in a decrease in plasma TRP and an increase in plasma KYN [8]. Another clinical study with cancer patients has shown that immunotherapy with IFN-α significantly increases the severity of depressive symptoms, which is related to a depletion of serum 5-HT and induction of the catabolism of TRP to KYN [27]. Thus, TRP catabolism switches from the 5-HT pathway to the KYN pathway, resulting in a decrease in 5-HT levels.
IDO1 is induced by several pro-inflammatory cytokines including IFNs (IFN-α/β, γ), TNF-α, and interleukin 6 (IL-6). It is also widely accepted that IFNs, especially IFN-γ, are essential factors for IDO1 induction since two ISREs and IFN-γ-activated site (GAS) element sequences are found in the 5′-flanking region of the IDO1 gene [28]. Recent preclinical studies in mice have demonstrated that pharmacological inhibition of IDO1 enzymatic activity or genetic deletion of IDO1 abrogates acute and chronic inflammation-dependent behavioral changes induced by peripheral or central administration of lipopolysaccharide (LPS) [29][30][31][32][33]. Additionally, it has been reported that peripheral administration of KYN alone can induce depression-like behavior in rats [34]. In a clinical study, patients receiving IFN-α therapy showed increases in the total Montgomery-Asberg Depression Rating Score (MADRS), an index of depressive symptoms similar to the KYN/TRP ratio; this indicates IDO1 activity and the KYN/KA ratio, which reflects a neurotoxic challenge [35]. These findings suggest that only TRP depletion itself may not be required for the induction of behavioral changes as a result of IDO1 activation; and that KYN and its neuroactive metabolites are more related to cytokine-induced depression-like behaviors than TRP depletion. However, it is still unclear whether direct activation of IDO1 and KYN metabolites plays a definitive role in the induction of depressive symptoms by IFN-α treatment.

Changes in the levels of serum TRP and its metabolites in HCV patients with IFN-α therapy
In order to further clarify the relationship between the IDO1-induced KYN pathway and the development of depressive symptoms during IFN-α therapy, we conducted a study in which we measured TRP metabolites of the KYN pathway in the serum of HCV patients undergoing IFN-α therapy.
A total of 49 patients (32 males and 17 females; mean age 54.0 ± 2.3 years) suffering from CHC were recruited. Table 1 shows the clinical characteristics of patients with HCV. In this study, most of patients were treated with recombinant (r) IFN-α 2b or pegylated (PEG)-IFN-α 2b (21 patients (42.9%) received each medicine, respectively). Five patients (10.2%) were treated with natural (n) IFN-α, and others received PEG-IFN-α 2a (2.0%) and rIFN-α 2a (2.0%), individually. All interferons have almost the same efficiency and induce about the same activation of the KYN pathway [36]. No patient had a past record of psychiatric treatment, and all were off from depressive symptoms prior to IFN-α treatment. They did not take any antidepressant medications during the study period. At an average of 104.2 ± 15.8 days after the IFN-α administration, some patients presented with apathy, social isolation tendencies, melancholy, depressed mood, and an intention to stop IFN administration. Patients who felt depressed mood were referred for psychiatric  "Depression (−)": HCV patients without depression, "Depression (+)": HCV patients with depression following IFN-α therapy [47]. HCV, hepatitis C virus; AST, aspartate aminotransferase; ALT, alanine aminotransferase. For all HCV patients, blood was collected before the onset of IFN-α therapy, as well as 2 and 4 weeks after the onset of therapy, and after the end or cessation of therapy. See Figure 3a for a detailed blood sampling schedule. * p<0.05 versus Depression (−) [47]. For all HCV patients, blood was collected before the onset of IFN-α therapy as well as 2 and 4 weeks after initiation of treatment. There was a no significant time difference for blood sampling between depression (−) and (+) patients (Table 1b and Figure 3).

(b) The time points of blood sampling
Previous studies suggested that IDO1-mediated TRP metabolism could be implicated in the development of depression, as a side effect of IFN-α therapy in HCV patients. We also found that HCV patients showed decreased TRP and increased KYN concentrations without any changes in KA, AA, and 3-HAA concentrations during IFN-α therapy (Figure 3b and  [38]. Our results showed that the level of 3-HK in the serum significantly increased in depression (+) patients are consistent with these findings. We also investigated the ratios of 3-HK/KA (reflecting neurotoxic indices) [39,40] and KYN/TRP (reflecting IDO1 activity) in depression (−) and depression (+) HCV patients during IFN-α treatment (Figure 3c and Table 2b). The ratios of KYN/TRP and 3-HK/KA in both groups increased during treatment. However, in depression (+) patients, the ratios of KYN/TRP and 3-HK/KA increased much larger in depression (−) patients during treatment (Table 2b). In these patients, the serum KYN/TRP and 3-HK/KA ratios increased more at the diagnosis of depression, but at 70.3 ± 9.1 days post therapy, they returned to the same levels as before onset of the therapy (data not shown). The severity of depression was not assessed during treatment, using neither the MADRS nor Hamilton Depressing Rating Scale. Therefore, we could not clearly show the direct association between the aggravation of depressive symptoms and changes in TRP metabolites. However, our results suggest that HCV patients with a high sensitivity for IDO1 activation by IFNs are highly susceptible to the depression-related side effects of IFN-α treatment.

The effects of chronic Ifn-γ gene expression on depression-like behavior in mice
We hypothesized that the high induction of IDO1 and the imbalance of TRP metabolites induced by IFNs in humans may be related to psychiatric side effects, such as depression.
Previous studies have shown that all three IFNs (IFN-α, -β, and -γ) induce strong IDO1 activity in human peripheral blood mononuclear cells [41,42]. In contrast, in mouse, IDO1 is induced more markedly by IFN-γ than IFN-α, which has only a weak direct IDO1-stimulatory effect. Therefore, we investigated whether IDO1 activity induced by Ifn-γ gene transfer impaired behavior in mice.  To conduct this experiment, for murine Ifn-γ gene transfer, the plasmid pCpG-Muγ was constructed by inserting a BglII/NheI murine Ifn-γ cDNA fragment into the BglII/NheI site of the pCpG-mcs vector (Figure 4a). The prepared plasmid pCpG-Muγ was dissolved in normal saline and injected into the tail veins of the mice for over 5 s on day 0. The injection volume was approximately 9% (v/w) of body weight. To eliminate the possibility of tissue damage or inflammation by the hydrodynamic injection, a control plasmid, which was the empty vector without the Ifn-γ gene (pCpG-mcs), was injected (0.05 pmol/mouse; IFN-γ transfected (−) mice). A previous study demonstrated that sustained IFN-γ concentrations were observed in mice receiving pCpG-Muγ at a dose of 0.2 pmol/mouse and more than 1000 pg/mL of IFN-γ was detected in the serum from 6 to 31 days after injection of pCpG-Muγ [43]. We also confirmed that the injected plasmid, pCpG-Muγ (IFN-γ transfected (+) mice) significantly increased IDO1 activity in the frontal cortex over a dose of 0.05 pmol/mouse compared to IFN-γ transfected (−) mice (Figure 4c). Therefore, the plasmid dose was fixed at 0.05 pmol/mouse for subsequent experiments, which corresponded to 0.10-0.12 μg of DNA/mouse.  [47]. βGlo MAR, β-globin matrix attachment region; mCMV enh, mouse cytomegalovirus enhancer; hEF1 prom, human elongation factor1 promoter; I140, synthetic 5′UTR containing an intron 140; MCS, multi cloning site; SV40 pAn, Simianvirus 40 polyadenylation; IFN-β S/MAR, interferon β gene scaffold/matrix attachment region; EM2K, CpG-free version of the bacterial EM7 promoter; Zeo, Zeocin; R6K ori, R6K origin.
In order to clarify whether the activation of IDO1 by IFN-γ-affected behaviors, three tests, open-field test (OFT), the Y-maze test, and forced swimming test (FST), were performed in mice. Mice were transfected with either a pCpG-mcs plasmid (control vector) that did not contain the Ifn-γ gene [IFN-γ-transfected (−) mice] or a pCpG-Muγ plasmid that long-lasting expressed Ifn-γ [IFN-γ transfected (+) mice]. No significant differences in locomotor activity of the OFT was observed between IFN-γ transfected (−) and (+) mice. Similarly, in the Y-maze test, no significant differences in the alternation behavior were detected between the two groups of mice. However, in the FST, immobility time was significantly longer in IFN-γtransfected (+) mice (Figure 5a). Our findings strongly suggest that IDO1 induction by IFN-γ is a critical factor in depression-like behaviors but not in short-term memory or locomotor activity in mice.

Changes in the levels of TRP and its metabolites in the serum and frontal cortex of mice following chronic Ifn-γ gene expression
In order to further elucidate the relationship between the IDO1-induced KYN pathway and the development of depression-like behavior in mice transfected with the pCpG-Muγ plasmid, we measured TRP metabolites in the serum and frontal cortex of these mice.
The serum and the frontal cortex were corrected from mice immediately following behavioral testing to determine the levels of TRP, KYN, KA, 3-HK, 3-HAA, and AA (Figure 5b and c). The concentration of serum TRP was significantly decreased in IFN-γ transfected (+) mice compared to IFN-γ-transfected (−) mice. In contrast, the levels of serum KYN and 3-HK were significantly increased in the IFN-γ-transfected (+) mice (Figure 5b). In the frontal cortex, IFN-γ transfected (+) mice had significantly higher KYN and 3-HK levels than the IFN-γ-transfected (−) mice. The TRP and KA levels in the frontal cortex tended to be lower in the IFN-γ-transfected (+) mice (Figure 5c). The activation of IDO1 by Ifn-γ gene transfer significantly modified the levels of TRP and its metabolites not only in the serum, but also in the frontal cortex of mice. These results suggest that an alternative explanation for the participation of IDO1 in IFNγ-induced depression-like behavior is the generation of neuroactive TRP metabolites. This interpretation is consistent with our clinical data and previous studies by O'Connor et al. and Wichers et al. [32,35].

The effects of Ido1 gene-deficiency on depression-like behavior, changes in TRP metabolism, 5-HT, and its turnover in the frontal cortex of mice following chronic Ifn-γ gene expression
Additionally, we evaluated the role of IDO1 in the development of depression-like behavior after Ifn-γ gene transfer using Ido1 gene knockout (KO) mice, and determined the levels of TRP metabolites in the frontal cortex.
The increase in time spent in an immobile posture in the Ifn-γ-transfected (+)/wild type mice was significantly improved in Ido1 KO mice (Figure 6a). In wild type mice, Ifn-γ gene transfer significantly increased the concentrations of KYN and 3-HK in the frontal cortex by 4.7-and 2.5-fold, respectively. In contrast, Ido1 KO mice withdrew these changes in Ifn-γ gene transfer  mice (Figure 6b). The levels of KYN and 3-HK in the frontal cortex after Ifn-γ gene transfer were considerably lower in Ido1 KO mice than in wild type mice. Even though we cannot exclude the possibility that genetic deficient in Ido1 and the resulting modifications in TRP metabolites could influence other behavioral tests, our results clearly demonstrate that Ido1 KO mice do not show depression-like behavior and do not intensify TRP metabolites after Ifn-γ gene transfer.
Other studies have emphasized that the 5-HT pathway is also relevant to depression. In a clinical study, it has been shown that levels of TRP and 5-hydroxytryptophan, a precursor of 5-HT, were significantly decreased from their baseline levels in the serum of HCV patients during IFN-α therapy [44]. Thus, we speculate that biological mechanisms underlying the IFN-α treatment induced-depressive symptoms are linked not only to the activated IDO1 and KYN pathway but also to a dysfunction of the 5-HT system. To clarify on the basis of the neurotransmitter changes in depression-like behavior after Ifn-γ gene transfer, we measured the concentrations of 5-HT and its metabolite 5-hydroxyindoleacetic acid (5-HIAA) in the frontal cortex of wild type and Ido1 KO mice (Figure 6c). We showed that Ifn-γ gene transfer produced a trend toward increased 5-HIAA levels in wild type mice but not in Ido1 KO mice. These results indicated that Ifn-γ gene transfer induced a potential increase in IDO1-induced 5-HT turnover. A raised 5-HT turnover suggests a process by which the availability of 5-HT to be released by neurons is decreased to compensate for neuronal dysfunction associated with depression-like behavior promoted by Ifn-γ gene transfer. Correspondingly, previous clinical studies have shown that brain 5-HT turnover is significantly increased in MDD patients without medication and decreased following selective serotonin reuptake inhibitors (SSRI) therapy [45,46].
Taken together, an alternative interpretation for the involvement of IDO1 in IFN-γ-induced depression-like behavior may be that depression is related to not only the generation of neuroactive TRP metabolites but also to the alteration of serotoninergic neurotransmission.

Conclusion
The levels of TRP metabolites in the serum of HCV patients changed significantly. In particular, the increase in serum 3-HK concentration in depressive HCV patients was much larger than that in HCV patients without depressive symptoms. The ratios of serum KYN/ TRP, reflecting IDO1 activity, and 3-HK/KA were increased in depressive and non-depressed HCV patients with therapy. However, the increase in serum KYN/TRP and 3-HK/KA ratios in depressive patients was much higher than that of non-depressive HCV patients. When the Ifn-γ gene was transfected into normal mice, depression-like behavior significantly increased. Additionally, Ifn-γ gene transfer to mice induced dramatic changes in TRP metabolite concentrations in the serum and the prefrontal cortex. On the other hand, genetic deletion of Ido1 abrogated the enhanced depression-like behavior after Ifn-γ gene transfer. In conclusion, our results clearly show that IDO1 is a critical molecular regulator of the depressive pathology induced as a side effect of interferon therapy. Moreover, the depressive symptoms are induced via increases in degradation of TRP and neuroactive metabolites along the KYN pathway, which finally changes in the alternation of 5-HT turnover. Our findings suggest that inflammatory pathways that lead to the activation of IDO1 may be a novel therapeutic target in patients suffering from inflammation-associated depression, for example, HCV or cancer therapy. Our results also suggest the monitoring of TRP-KYN metabolites during immunotherapy might assist in predicting the onset risk of depression as a side effect in these patients. However, further insight into the role of each downstream KYN pathway metabolite in the pathological process is needed to understand, and to clarify the relationship with complex neurotransmitters.