This study employed a systems biology approach to identify TFs acting as MRs candidates in BD based on microarray data. We also explored the transcriptional activity patterns in each phase of the disease through two-tailed GSEA. Furthermore, we analyzed the primary biological processes modulated by MR candidates to understand the functional differences associated with TFs in each mood state. Using MRA, we identified 51 MRs candidates in BD, and the DMTF1 repression was the only regulon present in the three mood states of BD compared to control. Additionally, we distinguish by two-tailed GSEA the top 5 candidates found in bipolar depression, mania, and euthymia, while the DNMT1 repression was common in the three phases of the BD. Since DMTF1 and DNMT1 were found in the mood states of BD and not in healthy controls, we infer its prominence as being indicative of disease. Based on these findings, we focused on exploring the two differentiated candidates in each transcriptomic analysis.
DNMT1 is part of the family of methyltransferases found in mammals. This enzyme is responsible for the catalysis and maintenance of DNA methylation - one of the most important epigenetic changes [34]. It is known that the brain contains the highest levels of DNA methylation, and this function is essential for neuronal maturation [35]. Animal model studies have shown that chronic stress alters the expression of epigenetic enzymes and DNA methylation in brain tissue associated with depression [36]. Additionally, hippocampal administration of DNA methylation inhibitors induces antidepressant-like effects in rodents, supporting the involvement of DNA methylation in the neurobiology of depression [37]. Also, DNMT1 deletion in the forebrain has anxiolytic and antidepressant-like effects in mice through a series of behavioral tests [38]. Beyond that, brain regions (amygdala and frontopolar cortex) of suicides showed alterations in DNMT gene transcripts expression [39].
Literature shows an association between methylation of the BDNF and cognitive performance, suggesting the BDNF gene influences brain regions that mediate cognitive tasks [40]. Similarly, another study showed a large increase in methylation in the Reelin gene promoter – directly interfering with the functioning of GABAergic neurons, brain development, and neuronal migration in the post-mortem brains of patients with schizophrenia [41]. Ludwig and Dwivedi also reported hypermethylation of several genes, including BDNF and the serotonin receptor 5HTR1A, corroborating the thesis that the imbalance of epigenetic changes affects the central nervous system and probably is involved in the etiology of neuropsychiatric disorders [42]. Furthermore, Duffy and colleagues suggest that epigenetic changes may influence disease onset in BD [43].
In addition to being reported in the CNS, DNA epigenetic modifications were found in the peripheral blood of patients with psychiatric disorders. For instance, levels of DNMT1 mRNA were decreased in BD and MDD patients during depressive episodes but not during the remission period implying that the altered expression is state dependent [44]. Further, higher levels of DNMT1 and IL-6 were found in anxious subjects compared to non-anxious subjects, which also corroborates the hypothesis that changes in DNA methylation may contribute to the biology of anxiety [45]. Moreover, our findings support previous findings confirming the involvement of methyltransferases and epigenetics in psychiatry.
Advances in the study of epigenetics have also contributed to the discovery of “epigenetic clocks''. Horvath and his research group pioneered DNA methylation for biological age estimation [46]. After Horvath, other epigenetic clocks emerged, such as the Hannum clock and the GrimAge clock. Fries and colleagues support the idea that there is an accelerated aging in the DNA of bipolar patients compared to control subjects, especially those patients with a longer duration of illness [47]. Furthermore, they have observed important epigenetic alterations, mainly methylation, in patients' hippocampus, partially related to gene expression levels, cognitive impairment, and premature aging [48]. In a recent study, accelerated epigenetic aging (employing the GrimAge clock) was associated with cognitive dysfunctions, such as short-term memory, inhibition, and problem-solving, and with age at first diagnosis in a bipolar sample [49]. Finally, the use of mood stabilizers seems to have the potential to slow down epigenetic aging [50]. Hereupon, the accelerated aging observed in BD may be a consequence of environmental factors - such as chronic stress exposure, lifestyle, and mood episodes. In this sense, detecting epigenetics markers, including DNA methylation and DNMTs, may contribute to a better understanding of the molecular basis underlying the diagnosis and prognosis of BD [48], [51].
DMTF1 (DMP1, DMTF, hDMP1, MRUL) is a tumor suppressor that activates the transcription of ARF and leads to the ARF-p53 pathway that promotes cellular senescence or apoptosis [52]. The mechanism is associated with the HDM2 and ARF bond, which blocks the interaction of HDM2 with p53 in the cytosol and nucleus. In this way, inhibition of HDM2 prevents ubiquitination and proteasome-dependent degradation of p53 leading to its stabilization. [53], [54]. a p53 complex is formed due to its accumulation in the nucleus and thus activates genes related to cell cycle and apoptosis. Although the ARF-p53 is one of the main mechanisms involved in the tumor-suppressive function, interactions of ARF with other transcription factors are also associated with cell proliferation, and protection from oncogenic signals has been identified [54].
In addition, ARF-p53 pathways function as significant stress sensors in detecting cells that have undergone different types of stress, thus leading to their elimination. When activated, p53 elicits cellular responses such as restoration of cellular homeostasis by cell cycle transitory blockage to allow DNA repair, senescence, or apoptosis [54]. Cell death from p53-dependent DNA damage includes mitochondrial membrane permeabilization and release of cytochrome c to initiate the downstream caspase cascade, triggering cellular stress and context. [55]. Interestingly, TP53 polymorphisms with susceptibility to schizophrenia or TB were recently investigated in the Chinese Han population. In this study, the allele C frequency of rs1042522 was significantly higher in BD patients than that in healthy controls [56] and this single-base-pair substitution may affect apoptosis efficiency.
Several studies have shown that abnormal neuronal apoptosis is involved in two neurophysiological processes associated with the pathogenesis of BD. For instance, a post-mortem study using electron microscopy found signs of apoptosis in frontal lobe oligodendrocytes in patients with BD. The main signs were cell reduction, smaller nuclear size, and nuclei condensation [57]. Another research identified decreased levels of Bcl-2 and BDNF together with high levels of protein expression and mRNA of BAD, BAX, caspase-9, and caspase-3, molecules known as apoptotic factors, in BA 9 of BD brains [58]. In the lymphocytes of BD patients, a decreased expression of the anti-apoptotic factor HSP70 and a reduced BAX level in the cytosolic fraction were observed, indicating that it had been transferred to the mitochondria to induce apoptosis [59]. Similarly, alterations in pro-apoptotic proteins were shown as well as an increased percentage of early apoptosis in peripheral blood mononuclear cells from BD patients compared to controls [60], [61]. Pietruczuk et al. also found higher apoptosis (and Bax expression) in the lymphocytes of BD patients compared with controls. At the same time an in vitro experiment, lithium and valproate protect the patients' lymphocytes from apoptosis [62]. Considering that the BCL-2 family is one of the downstream targets of ARF-p53 [63], it is reasonable to speculate that an unbalance between cell arrest and apoptosis is an important mechanism underlying BD. Such imbalance may contribute to the impaired cellular response to stress and make cells more susceptible when exposed to stressful situations, also contributing to increased oxidative stress, mitochondrial dysfunction, inflammation, and DNA damage commonly observed in the brain and peripheral subjects with BD.
Finally, from ontology analysis, many biological processes modulated by MR candidates were shown in BD and grouped by their function, such as RNA metabolism, cellular respiration, and ribosome biogenesis. However, the function most prominent in mania, euthymia, and bipolar depression was the inflammatory or immune system.
Inflammation is a response promoted by the immune system against pathogens, diseased cells, or foreign molecules - interpreted as harmful stimuli to our body - and tissue damage, such as cuts or wounds. Thus, inflammation is defined as a fundamental defense mechanism [64]. The main characteristic of inflammation is vasodilation, which allows the arrival of more cells and defense molecules, characterizing leukocyte infiltration and edema formation. From there, several biological mechanisms take place, where macrophages - cells derived from the monocyte in the blood and which have a high power to phagocytize and destroy foreign bodies - produce cytokines, chemokines, and other proinflammatory factors. The inflammatory or immune system was the central biological process being altered in the three mood states of BD. These findings agree with emerging evidence that the immune system and inflammatory response are strongly associated with the BD pathophysiology [49], [65]–[68]. With this regard, synthesized data from 3,528 bipolar patients showed an elevation of proinflammatory immune biomarkers (CRP, IL-6, TNF-α) when compared to healthy controls. These results also indicated that while CRP and TNF-α might be regarded as “mood episode” markers of BD, IL-6 is considered a trait marker of this illness [69]. Furthermore, combined, the index CRP/IL-6 or BDNF/TNF-α and TNFR1 levels seem promising markers to differentiate BD patients from matched controls [70].
In addition to the inflammatory response involving cytokines, another critical and essential mechanism in inflammation is the coagulation cascade activation since thrombin – the enzyme that catalyzes the conversion of fibrinogen to fibrin – has pro-inflammatory functions [71], [72]. Data from patients' blood analyzed by mass spectrometry and gene ontology showed the involvement of complement and coagulation cascade in BD physiopathology by identifying specific proteins such as the third component of complement and apolipoprotein A involved in these processes [73], [74]. A pilot study also indicated the dysregulation of the coagulation cascade in treated schizophrenia and BD patients through proteome analysis, reinforcing the importance of these biological processes in psychiatric disorders [75].
Furthermore, microglia can interact with several cells of the immune system, including T cells, stimulating the release of pro-inflammatory cytokines and activation of the inducible Nitric Oxide Synthase (iNOS) enzyme – related to vasodilation during the inflammatory response, which allows the arrival of more components of the immune system [75]. The attention to be taken in cases of chronic neuroinflammation, i.e., when microglia are persistently activated, leading to the excessive production of Reactive Oxygen Species (ROS) – which in turn causes DNA damage and, as a defense mechanism, activates NF-κB, giving sequence to the neuroinflammatory chain and promotes more neuronal damage and cell death [76]. All these data suggest the therapeutic potential of targeting the inflammatory and immune systems in BD [77].
It is essential to mention the limitations of our study, such as sample heterogeneity. Some studies do not provide complete clinical information, including mood state, duration of illness, comorbidities, or pharmacotherapy, all of which may influence the results. However, the selection of datasets was performed to minimize sample differences. Thus five sets based on the same transcriptomic analysis (microarray technique) were included. Another significant issue is regarding the blood fraction used to quantify RNA. We have four datasets that used whole blood and one that used only peripheral blood mononuclear cell fraction. This blood fraction is directly linked to the patient's immunological profile, thus leading to heterogeneous sampling when merged with other datasets.