The Nrf2 Pathway in Depressive Disorders: A Systematic Review of Animal and Human Studies

There is increasing interest in the involvement of antioxidative systems in protecting from depression. Among these, Nrf2 occupies a central place. We aimed to review the role of Nrf2 in depression. For this reason, we conducted a PubMed search using as search strategy (psychiatr*[ti] OR schizo*[ti] OR psychot*[ti] OR psychos*[ti] OR depress*[ti] OR MDD[ti] OR BD[ti] OR bipolar[ti] OR Anxiety[ti] OR antidepress*[ti] OR panic[ti] OR obsess*[ti] OR compulsio*[ti] OR “mood disord*”[ti] OR phobi*[ti] OR agoraphob*[ti] OR anorex*[ti] OR anorect*[ti] OR bulimi*[ti] OR “eating disorder*”[ti] OR neurodevelopm*[ti] OR retardation[ti] OR autism[ti] OR autistic[ti] OR ASM[ti] OR adhd[ti] OR “attention-deficit”[ti]) AND nrf2, which on the 9th of March produced 208 results of which 89 were eligible for our purposes. Eligible articles were studies reporting data of Nrf2 manipulations or content by any treatment in human patients or animals with any animal model of depression. Most studies were on mice only (N = 58), 20 on rats only, and three on both rats and mice. There were two studies on cell lines (in vitro) and one each on nematodes and fish. Only four studies were conducted in humans, one of which was post mortem. Most studies were conducted on male animals; however, human studies were carried out on both men and women. The results indicate that Nrf2 is lower in depression and that antidepressant methods (drugs or other methods) increase it. Antioxidant systems and plasticity-promoting molecules, such as those in the Nrf2–HO-1, BDNF–TrkB, and cyclic AMP–CREB pathways, could protect from depression, while glycogen synthase kinase-3β and nuclear factor κB oppose these actions, thus increasing depressive-like behaviours. Since Nrf2 is also endowed with tumorigenic and atherogenic potential, the balance between benefits and harms must be taken into account in designing novel drugs aiming at increasing the intracellular content of Nrf2.


Introduction
Recently, much research has been devoted to the study of inflammation and its role in the context of a wide variety of pathological conditions. Inflammation and oxidative stress were found to enhance each other, thus establishing a pathological state [1], which may be found in various psychiatric disorders. The variety of external stimuli to which organisms are subjected triggers adaptive responses, which are designed to restore homoeostasis through a fine balance between oxidation and antioxidant activity [2].
An imbalance between the generation of reactive oxygen species (ROS) and antioxidant defences results in increased oxidative stress [3], with a consequent increase in neuroinflammation, mitochondrial dysfunction, and cell degeneration processes such as apoptosis and ferroptosis, which proved to be crucial in many psychiatric disorders. This is not surprising if we consider that the brain consumes massive doses of oxygen and contains high concentrations of oxidative lipids, thus being extremely vulnerable to oxidative stress-induced damage [2].
In this framework, it is important to draw attention to nuclear factor erythroid-2 (Nrf2), the main endogenous negative regulator of oxidation [4]. Its activation determines the expression of numerous antioxidants and cytoprotective genes capable of modulating oxidative stress. Nrf2 also presides over the regulation of genes involved in the oxidative stress-related pathological processes mentioned above. Consequently, dysregulation in Nrf2 expression with reduced cortical levels may contribute to the aetiopathogenesis of numerous pathological conditions, including psychiatric disorders and neurodegenerative diseases.
Nrf2 is a protein transcription factor composed of 605 amino acids, encoded by the NFE2L2 gene and belonging to the Cap'n'collar (CNC) family of transcription factors [5]. It contains seven highly conserved functional domains called Nrf2-ECH homology 1 (Neh1-Neh7). Neh1 and Neh3 interact with specific DNA sequences called antioxidant response elements (ARE) [6], thus promoting the transcription of enzymes with antioxidant activity. Neh2 interacts with Kelch-like ECH-associated protein 1 (Keap1), the main negative regulator of Nrf2. Neh4 and Neh5 interact with the cyclic adenosine monophosphate (cAMP)-response element-binding protein (CREB), which also promotes transcriptional activation [7]. In contrast, Neh6 can link to β-transducin repeat-containing protein (β-TrCP) and is involved in Keap1-independent degradation of Nrf2. Finally, Neh7 inhibits the Nrf2-ARE signalling pathway by binding to retinoic X receptor alpha (RXRα) [8]. Therefore, the activity of Nrf2 is subjected to both Keap1-dependent and Keap1-independent regulation.
Keap 1 represents the main Nrf2 suppressor; it forms a homodimer capable of binding ETGE and DLG motifs (stronger and weaker binding sites of Keap-1, respectively; the former is located in the loop region of the antiparallel β-sheet, while the latter is N-terminal to the α-helix [9]) included in the Neh2 domain of Nrf2. Under physiological conditions, the Keap1-Nrf2 complex combines with the E3-ubiquitin ligase Cullin 3 (Cul3) complex, leading to ubiquitination and proteasomal degradation of Nrf2 [10]. Negative regulation is also mediated by the phosphorylation of Nrf2 by glycogen synthase kinase-3 beta (GSK-3β) and mitogen-activating protein kinase (MAPK) [11].
When oxidative stress increases and ROS accumulation occurs, there is a dissociation of the Keap1-Nrf2 complex induced by conformational changes of a Keap1 domain, inhibiting ubiquitination and subsequent degradation of Nrf2. Dissociated from Keap1, Nrf2 is free to translocate into the nucleus and bind specific genomic sequences in order to promote antioxidant enzyme transcription. At the same time, there is a positive Keap1-independent regulation mediated by other kinases. The kinases involved in the phosphorylation and subsequent activation of Nrf2 include protein kinase C (PKC), casein kinase II (CK2), protein kinase R (PKR), c-Jun N-terminal kinase (JNK), and extracellular regulated kinases (ERKs) [12][13][14][15]. Finally, brain-derived neurotrophic factor (BDNF) can also promote the activation and subsequent migration of Nrf2 into the nucleus [16].
In addition to its antioxidant activity, Nrf2 is also directly involved in oxidative stress-related pathological processes by regulating their activation. In particular, there is a direct cross-talk between Nrf2 and p62, an autophagy key protein. Nrf2 can promote the expression of genes involved in autophagy, while p62 can compete with Nrf2 in binding to Keap1 in a positive feedback that is associated with cytoprotection [6,17]. In addition, Nrf2 is also involved in a special form of autophagy called mitophagy, the alteration of which allegedly plays an important role in psychiatric disorders. While it preserves mitochondrial integrity, mitophagy entails the elimination of damaged or redundant mitochondria through autophagy [18].
Furthermore, recent studies of Nrf2 found that its inducers promote the suppression of the pro-inflammatory phenotype of microglia through regulating BDNF, the reduction of which is found in many psychiatric disorders characterised by neuroinflammation [16,[19][20][21]. The existence of reciprocal regulation has emerged between BDNF and Nrf2; the latter, in connecting with the exon I promoter of bdnf, can activate BDNF; on the other side, BDNF can increase the nuclear translocation of Nrf2, thus promoting its antioxidant activity [16].
Finally, several studies also cast light on the involvement of Nrf2 in ferroptosis, an iron-dependent form of programmed cell death characterised by the accumulation of lipid peroxides (lipids damaged by oxidation). Notably, Nrf2 does not only regulate numerous genes involved in iron metabolism and homoeostasis but also promotes the basal expression of the lipid hydroperoxidase, glutathione peroxidase-4 (GPX4), which converts toxic lipid peroxides to nontoxic lipid alcohols [22]. Impaired response to oxidative stress has been shown in animal models for stress disorders, such as post-traumatic stress disorder (PTSD) [23][24][25][26], but only a few studies have focused on Nrf2 [27][28][29]. PTSD and depression are long considered to represent stress disorders and share common neurobiological patterns [30]. Recently, the Nrf2-depression connection has received attention [2]. We decided to search the literature for studies investigating the ties between depression paradigms in the animal and/or major depressive disorder in humans and Nrf2 as a proxy of a mechanism that counters oxidative stress. Establishing such a relationship would allow us to concentrate on the production of drugs that would promote the search for drugs interfering with intracellular oxidative processes.

Methods
To systematically review the ties between depression and Nrf2, we first conducted a PubMed search involving all possible mental and psychiatric disorders and then focused on depression. The inclusion of other than depressive disorders/states was to make sure that depression during the course of other mental disorders was not dealt with. We employed the following search strategy: ( Following search performance, we characterised the nature of all ensuing records and labelled them accordingly. This resulted in their being either included or excluded. The inclusion/exclusion labelling with the reasons for exclusion is shown in the Online Supplemental material and in Figure 1, where the PRISMA flow diagram is displayed. In carrying out our review, we followed the 2020 Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement [31]. The 2020 PRISMA Checklist may be found in the Online Supplement. We assessed the Risk of bias (RoB) of the included studies with the Cochrane RoB 2.0 tool [32]. We performed an evaluation of the RoB for each included study. The results are shown in the Online Supplement.
Eligibility was based on being an original study on any animal or tissue, including humans, on investigating depression or depression models and providing data on Nrf2 levels. All other studies were excluded. Excluded were case reports, opinion articles, such as editorials, letters to the editor, comments of other work, reviews, and meta-analyses (however, we hand-searched their reference lists to identify other possibly eligible studies), and studies not providing data. Eligibility for each paper was established with the consen-sus of all authors obtained through Delphi rounds, in which all authors participated, either in-person or online. The same applied to the compilation of the RoB. Eligibility was based on being an original study on any animal or tissue, including humans, on investigating depression or depression models and providing data on Nrf2 levels. All other studies were excluded. Excluded were case reports, opinion articles, such as editorials, letters to the editor, comments of other work, reviews, and meta-analyses (however, we hand-searched their reference lists to identify other possibly eligible studies), and studies not providing data. Eligibility for each paper was established with the consensus of all authors obtained through Delphi rounds, in which all authors participated, either in-person or online. The same applied to the compilation of the RoB.

Results
Our search, eventually conducted on the 10th of March 2023, yielded 208 results on PubMed, of which 89 studies were eligible, as summarised in Table 1 (human studies, of which two used cell lines in vitro) and Table 2 (animal studies). In particular, there were 78 articles labelled Depression and 11 Depression and Anxiety; these amounted to 89 articles. The remaining 116 studies were excluded. Depression-free articles focused on autism (N = 17), schizophrenia (N = 10), anxiety (N = 9), attention-deficit/hyperactivity disorder (N = 3), insomnia (N = 2), and bipolar disorder alone (N = 1). Other articles that did not meet the inclusion criteria were Opinions (N = 1), Case reports (N = 1), and Reviews (N = 23), while many articles were off-target (N = 49) in that they did not focus any psychiatric disorder and were unfocused in their designs or they were unrelated to the subject of our inquiry. Furthermore, there were two duplicates and a retracted paper, but another paper from the same group had not been retracted and dealt with the same issue as the retracted one. Publication dates spanned from 23-March-2006 to 1-March-2023 for the searched papers and 23-April-2013 to 1-March-2023 for the eligible ones. The included and excluded studies with their reasons for exclusion are shown in the Supplement. The selection process and reasons for exclusion are depicted in the PRISMA flowchart ( Figure 1).
Of the 89 studies included in this review, most were conducted on mice only (N = 58; 65.17%), 20 were conducted on rats only (22.47%), and three on both rats and mice (3.37%), four on humans (4.49%), two on human cell lines in vitro (2.25%), one on fish and worms each (1.12%). Of the 58 studies carried out on mice only, 50 used only male animals, seven used only female, and one both used animals of both sexes; 31 used C57BL/6 strains; one reported unspecified C57 mice (which were presumably C57BL/6, based on other articles by the same group of authors), 10 used unspecified Swiss strains, five used CD-1, eight used Balb/c, five ICR, four Kunming, one BXD Recombinant Inbred, and one Murphy Roths Large lymphoproliferative Mouse (MRL/lpr); in eight of these studies, investigators used more than one strain. All 20 studies conducted on rats used only male animals, 14 Sprague Dawley and six Wistar, of which two were Hannover and two albino Wistar, while all three studies that employed both mice and rats were conducted using male-only animals, all three used Sprague Dawley rats (1 also Groningen, a strain characterised by high aggression levels [33]) and C57BL/6 mice. The four human studies included patients and matched controls of both sexes, while the only post mortem study did not report the sex or the age of included patients. The study that used fish employed the Japanese rice fish, medaka (Oryzias latipes), both male and female, and the one that used nematodes used Caenorhabditis elegans, while of the two conducted on cell lines, one used the macrophage RAW26.7 line (primary CD14+ monocytes from human donors of both sexes transformed in macrophages through 1-week Colony Stimulation Factor-1 stimulation) and the other used human neuroblastoma SH-SY5Y cells. Of the 85 studies that specified the sex of the animals they used in their experiments, 79 used males (92.94%) and only 12 used females (14.12%). Limiting the sex of animals to the 81 rodent studies, which constituted the bulk of eligible studies included in this review, it results that 73 studies employed male-only animals (90.12%), while only seven studies (9.59%) used female-only animals. This shows a strong bias toward the use of male animals in depression studies of rodent models that cannot be easily translated to humans, given that the majority of people with depression are women [34,35] or female adolescents [36].
Of the eligible studies, most were conducted in China (N = 49, 55.06%; only three were located in Beijing, China's capital, while six studies were conducted in the Guangdong Province, i.e., three in Shenzhen, two in Guangzhou, and one in Zhanjiang), eight in Brazil (8.99%) and seven in Spain, of which one was a multinational study shared with other four countries (6.97%), five in Japan (5.62%), four in Egypt (4.49%), three in the US (3.37%), two in South Korea and France (2.25% each), two in Germany and Poland, with one multinational shared (1.35% each), one each in India, Iran, Italy, Nigeria, Pakistan, Serbia, and Turkey (1.12% each), and one in both Romania and Sweden, sharing the same multinational study as the other three (0.22% each). Of the 89 included studies, 15 did not use any specific drug to identify its effects on Nrf2 but rather focused on the effects of specific animal depression models on the entire antioxidant system. These 15 studies were conducted in China (N = 8), Spain (N = 3), Serbia, the USA, Brazil, and South Korea (N = 1 each). Plant extracts or animal tissue extracts were tested in 29 Chinese studies, three Japanese studies, one Pakistani, one Nigerian, and one South Korean study, for a total of 35 studies, representing 39.33% of all included studies, with China accounting for 82.86% of these studies and plant extract using studies for 59.18% of all Chinese studies.                          [105] ♂Sprague Dawley rats; BW 180-200 g LPS-induced DLB Islamabad, Pakistan Rats divided into 2 groups of 10, one treated with fluoxetine (5 mg/kg) + LPS, CAR20 or CAR50 + LPS → single dose of each × 5 days, ip + LPS → or single dose or after CAR20/50 (3rd, 4th day 1 mg/kg), and another, divided into 3 subgroupings (10 animals/group): ATRA + LPS, ATRA + LPS + CAR20/50, ATRA + LPS+ fluoxetine. ATRA ip injection 30 min before LPS. At Day 2, → behav tests (FST, LDB, E + M, sucrose SPIT) → sacrificed → Hippoc and cortical tissue collected and processed to RT-PCR and ELISA for Nrf2/HO-1. Immunohistochemical analysis for OS-related molecules LPS ↓ Nrf2/HO-1 expression in cortex and Hippoc compared to Ctrl (p < 0.05). CAR20/50 + LPS ↑ Nrf2 and HO-1 compared to the LPS group (p < 0.01) LPS ↑ ROS and DLB. CAR20/50 ↓ DLB and↑ Nrf2/HO-1; no differences between CAR20/50 + ATRA + LPS and ATRA + LPS   [108] ♂adult Swiss mice BW 25-30 g CUMS-induced depression Fortaleza, Ceará, Brazil Mice were exposed to a variety of stressful events (restraint, tilted cage, intermittent circle between lights on and off, constant light, water deprivation, wet cage, electric shock) × 28 days; from the 14th day they received DMF 50 and 100 mg/kg or fluoxetine 10 mg/kg or plac. On the 29th day → behav tests (OFT, FST, SPT, NOR). Mice divided into two groups, Ctrl and CUMS, which were subdivided into plac (carboxymethyl cellulose 0.5%), DMF50, DMF100 or fluoxetine 10 mg/kg (FLU) groups. 14 days after the beginning of the procedure, treatments were started in all groups, 30 min before daily stress application in CUMS groups.

Discussion
In this review, we collected data on animal and human studies on depression and the role of Nrf2 as a probe of the antioxidant system function and found an antidepressant effect to be the consequence of the activation of the Nrf2-HO-1 pathway, with no studies pointing to the opposite direction. In this context, the role of neuroinflammation in being associated with depressive-like states clearly emerges, as well as the role of antioxidant systems in countering neuroinflammation and its markers, with Nrf2 standing at its crossroad. We may thus suppose that depression involves multiple systems and the antioxidant system, and Nrf2, in particular, counteract many actions of the failing multi-system state associated with depressive behaviour. Many of the substances that help overcome the alterations caused in the organism by depression are of natural origin and constitute long-used remedies for many other conditions. The antioxidant research line in depression that focuses on the actions of Nrf2 is relatively recent, dating back to 2013 [43], although the substance was isolated in 1994 [126], almost three decades ago.
The rate of publications on this issue has considerably increased across the years, with papers focusing on psychiatric disorders increasing almost exponentially from 26 in 2020 to 37 in 2021 and to 58 in 2022, witnessing increased interest and awareness of its importance in the pathophysiology of psychiatric disorders. Studies on depression reflect a similar trend. From eight studies in 2019, which increased to 16 in 2020, 18 in 2021, and 23 in 2022 to the beginning of 2023.
Most of the studies were conducted on rodents, mainly mice and, most importantly, male animals. This limits the extension of results to female animals, but it should be noted that findings in females were not dissimilar from those in males.
Anti-inflammatory or anti-neuroinflammatory effects of employed antidepressants and putative antidepressant substances, such as the many plant extracts used in the studies we considered here, were shown to be related to various improvements in neurochemical indexes and behavioural tests [43,48,59,71,76,81,85,91,95,108,111,123,125].
Studies were consistent in measuring cyclic AMP responsive element binding protein (CREB) activity, in that increased phospho-CREB or CREB levels were associated with an antidepressant response [38,46,81,83,103,109,112]; remarkably, studies on peripheral tissues in man [38] and hippocampal content in mice [81] pointed to the same direction. Additionally, studies of BDNF showed that increased BDNF activity was associated with antidepressant effects and that animals displaying depressive features have low BDNF contents in the periphery and the brain [44,52,57,57,58,66,67,69,81,83,85,88,93,111,112,114,121]. It has recently been shown that most, if not all, antidepressant drugs bring about their antidepressant effects by binding the tyrosine kinase receptor 2 (TrkB), a molecule tied to the action of BDNF [127,128]. The findings of the studies reviewed here quite match this concept; increased TrkB/BDNF signalling was related to antidepressant effects, and low levels were associated with depressive behaviour [11,50,66,69,84,93,103,111,120]. Increased activity also of protein kinase B (Akt-1), which is an antioxidant in the wingless-GSK pathway and carries on some of the biological [128][129][130] and clinical [131] actions of lithium, which also inhibits GSK-3β [132][133][134], was found to be associated with antidepressant effects in the studies included in this review [42,45,47,74,75,93,94,110,119]. On the opposite side are found nuclear factor kappa B (NF-κB) and glycogen synthase kinase (GSK) pathways; the higher the NF-κB [38,39,67,69,73,75,80,85,91,98,100,103,113,115] and GSK3β levels [45,79,109], the worse the depressive indexes. Therefore, it appears that on one side, there are antioxidant systems and growth factors promoting plasticity, such as Nrf2-HO-1, BDNF, Akt, TrkB, and CREB, which are protective from depression and, on the other side, GSK-3β and NF-κB, which promote depressive behaviour ( Figure 2). Therefore, we may suppose that new antidepressant drug discovery could involve promoting drugs that increase Nrf2-HO-1, BDNF, Akt, TrkB, and CREB activity, while downplaying GSK-3β and NF-κB. However, it is not so simple to identify the characteristics of drugs that act only where they will wishfully carry on the desired effect because drugs acting within cells must cross plasma membranes selectively, reach their intracellular target in those cells needing their actions, and not everywhere. Depression is a multi-system and multi-organ condition but not an all-system derangement, so the presence of an extraordinarily penetrating drug in some cells could be related to undesirable or adverse effects. Furthermore, not all actions of Nrf2 are good for the body. Nrf2 may promote tumours [135,136] and atherogenesis [137,138], so any drug acting on Nrf2 must avoid these two and maybe other unknown potentially harmful actions. nalling was related to antidepressant effects, and low levels were associated with depressive behaviour [11,50,66,69,84,93,103,111,120]. Increased activity also of protein kinase B (Akt-1), which is an antioxidant in the wingless-GSK pathway and carries on some of the biological [128][129][130] and clinical [131] actions of lithium, which also inhibits GSK-3β [132][133][134], was found to be associated with antidepressant effects in the studies included in this review [42,45,47,74,75,93,94,110,119]. On the opposite side are found nuclear factor kappa B (NF-κB) and glycogen synthase kinase (GSK) pathways; the higher the NF-κB [38,39,67,69,73,75,80,85,91,98,100,103,113,115] and GSK3β levels [45,79,109], the worse the depressive indexes. Therefore, it appears that on one side, there are antioxidant systems and growth factors promoting plasticity, such as Nrf2-HO-1, BDNF, Akt, TrkB, and CREB, which are protective from depression and, on the other side, GSK-3β and NF-κB, which promote depressive behaviour ( Figure 2). Therefore, we may suppose that new antidepressant drug discovery could involve promoting drugs that increase Nrf2-HO-1, BDNF, Akt, TrkB, and CREB activity, while downplaying GSK-3β and NF-κB. However, it is not so simple to identify the characteristics of drugs that act only where they will wishfully carry on the desired effect because drugs acting within cells must cross plasma membranes selectively, reach their intracellular target in those cells needing their actions, and not everywhere. Depression is a multi-system and multi-organ condition but not an all-system derangement, so the presence of an extraordinarily penetrating drug in some cells could be related to undesirable or adverse effects. Furthermore, not all actions of Nrf2 are good for the body. Nrf2 may promote tumours [135,136] and atherogenesis [137,138], so any drug acting on Nrf2 must avoid these two and maybe other unknown potentially harmful actions.

Limitations
This review has several limitations in that the animal studies focused mostly on rodents (mice and rats), and few focused on other animals (fish and nematodes) or cell lines; therefore, the results cannot extend to other animal species. Human studies were only four (one post mortem), and their methodologies were too different from animal protocols; hence the results from the latter are not translatable to the former. Furthermore, a sex bias exists in animal studies (mostly males) and not in the few human studies. Depression is overrepresented in women in humans, so it might be that testing it in male animals is not a good idea, but female rodents pose significant problems in conducting scientific experiments using the paradigms of the studies included in this review. The animal models of depression are not easy to translate to humans and into clinical practice.

Conclusions
Nrf2 is at the crossroad of many cellular actions. It directs antioxidant pathways and receives many intracellular signals to which it attempts to respond in a balanced way. Its function appears to be impaired in depression, and it is also possible that its manipulation could prove to be beneficial in human depression; however, there is much to discover. There is hope to discover drugs that cross the blood-brain barrier, or use some phytopharmaca or their derivatives, such as gastrodin, or use drugs stimulating anti-inflammatory cytokines that could counteract neuroinflammation. From the studies included in this review, depression appears to be strictly tied to inflammatory mechanisms. Provided that new drugs acting on the Nrf2 antioxidant pathway can avoid tumorigenesis and atherogenesis, Nrf2 can be a useful target for novel drug development.