Introduction
Noncommunicable diseases represent a major challenge for health systems, further augmented by the co-occurrence of multiple chronic conditions. While both diabetes and depression present huge socioeconomic problems affecting millions of people worldwide, the role of diabetes in increasing the risk of depression is only beginning to emerge [
1].
The link between the two diseases is unclear at present due to lack of appropriate mechanistic insight. Depression causes deterioration in carbohydrate metabolism and increases the frequency of complications; it also impairs the quality of life and reduces life expectancy [
2]. In addition, the incidence of depression is markedly increased in individuals with diabetes [
3]. Since diabetes is now widely recognised as a state of chronic systemic inflammation [
4], inflammatory actions could represent a plausible mechanism through which diabetes and depression interact [
5]. However, there is insufficient experimental evidence to support this hypothesis and effective therapies are lacking.
A clear association between depressive symptoms and reduced levels of the growth factor, brain-derived neurotrophic factor (BDNF), has long been recognised [
6]. Hippocampal biopsies of individuals with major depression revealed lower levels of BDNF and its receptor, tropomyosin receptor kinase B (TrkB) [
7]. Moreover, TrkB signalling has been implicated in the action of antidepressants, even positioning this pathway as a potential predictor for the efficacy of antidepressants [
8].
BDNF is predominantly produced in the central nervous system in the form of a precursor (proBDNF) from which mature BDNF (mBDNF) is derived via proteolytic cleavage. mBDNF induces axon growth and synaptic activity and facilitates cell survival through its receptor, TrkB. ProBDNF is itself also biologically active, although it has an opposing function whereby it decreases synaptic activity and activates apoptotic pathways via neurotrophin receptor p75 (p75
NTR) [
9].
Several in vitro and in vivo studies indicate that neuroinflammatory processes, which are widely linked with the development of depression, affect BDNF expression [
10]. Notably, lipopolysaccharide or proinflammatory cytokine administration has been demonstrated to remarkably reduce mRNA and protein levels of BDNF in the brain [
11,
12].
Renin–angiotensin–aldosterone system (RAAS) inhibitors are currently one of the primary options in the treatment of diabetes and related complications [
13]. Recent findings from a large cohort of individuals with type 1 diabetes showed that RAAS-modifying medication was associated with a reduced requirement for antidepressants [
14]. Furthermore, treatment with the angiotensin receptor 1 blocker (ARB) candesartan significantly improved interpersonal sensitivity and depression scores of diabetic individuals [
15].
A local RAAS, expressing all classical signalling pathways, also exists in the brain, where it regulates a wide variety of biological functions including blood pressure, body temperature, memory, behaviour and learning [
16]. Clinical studies have confirmed that increased RAAS activity in the brain is associated with the development of depression, Alzheimer’s disease and Parkinson’s disease [
17‐
19].
We hypothesised that blockade of RAAS activity by losartan, a widely used treatment regimen in the clinic, may exert antidepressant effects in an experimental model of diabetes. Here we investigated the importance of BDNF signalling in the inflammatory, survival and apoptotic processes of diabetes-associated depression, and suggest novel targets for clinical therapy.
Discussion
The possible link between diabetes and depression is more than 300 years old and is a re-emerging paradigm. ARBs are a widely used clinical option for treatment of individuals with various diabetes-associated complications. Furthermore, clinical observations of the past decade unexpectedly revealed that ARB-treated hypertensive individuals have a lower requirement for antidepressants [
28,
29]. In addition, ARBs improved interpersonal sensitivity and depression scores of individuals with type 2 diabetes [
15]. However, the molecular mechanisms underlying these observations remain unexplored.
To test the hypothesis that ARBs may exert an antidepressant effect, we examined the effect of losartan in a rat STZ-induced experimental diabetes model. We showed for the first time that the depressive symptoms of diabetic rats can be effectively minimised by losartan treatment and that the antidepressant effect of losartan is independent of blood glucose or body weight.
Cerebrovascular abnormalities are common in diabetes and contribute to the development of depression [
30,
31]. In addition, RAAS is a central regulator of cerebral perfusion and angiotensin II is the main effector causing local vasoconstriction. Given the known associations between cerebrovascular pathophysiology and depression (termed frequently as the ‘vascular hypothesis’ [
32]), we first investigated whether alterations in cerebral perfusion might be regulated by the actions of losartan in diabetic rats. Impaired cerebral perfusion was verified by a remarkable decline in cerebral relative [
99mTc]HMPAO uptake in diabetic rats and this decline was not improved by losartan. We found that phosphorylated eNOS levels remained unchanged while endothelin-1 levels increased in diabetic brains, in parallel with elevated hippocampal ICAM-1 immunopositivity. These findings indicated vascular activation, although the reduction in these variables brought about by losartan did not reach statistical significance. Thus, our results suggest that the beneficial effects of losartan on behaviour may be independent of its direct vasoactive actions. This is in line with earlier observations showing mitigated neurological deficit by losartan without alterations in cerebral blood flow in a murine model of type 2 diabetes [
33].
Neuroinflammation is associated with both diabetes and depression [
34,
35], as indicated by the expression of proinflammatory mediators as well as microglia and astrocyte activation [
36‐
38]. Increased [
125I]CLINME-translocator protein uptake in diabetic rats suggested microglial activation, which was confirmed by immunohistochemistry, similar to that seen in other studies after 6 weeks of STZ-induced diabetes in the hippocampus [
38]. Microglia and astrocyte activation was associated with the induction of an NF-κB-mediated inflammatory response with elevated
Il1a,
Il6 and
Tnf mRNA expression in diabetic rats, which was prevented by losartan. In line with this, losartan treatment decreased the response of astrocytes, but not of microglia, suggesting that diabetes-induced neuroinflammatory responses are modulated by losartan mainly via astrocytes. Previous studies in lipopolysaccharide-induced inflammation and hypertensive models showed that ARBs augment astrocyte and microglia activation [
39,
40].
How mood disorders such as depression are influenced by inflammatory processes is not well understood. The causal role of proinflammatory mediators, namely IL-1 and TNF, has been widely proposed [
41]. For example, IL-1-mediated actions may change neuronal network activity, the release of neurotransmitters, autonomic nervous system and hypothalamic–pituitary–adrenal axis responses and levels of growth factors among others [
42,
43]. Recent studies indicate that neuroinflammatory processes may reduce BDNF expression [
11,
12]. However, it remains unclear whether inflammation-induced changes in BDNF levels in the brain are associated with the formation of depression-related behaviours and whether this can be reversed therapeutically. Here, for the first time, we demonstrated changes in both forms of hippocampal BDNF in diabetes-associated depression. We also showed earlier that proBDNF and mBDNF levels were reduced in diabetes and were elevated by the antidepressant fluvoxamine [
44]. Our present results indicate that diabetes-associated neuroinflammation, reduced BDNF levels and impaired BDNF signalling can be effectively reversed by an ARB. While microglia sense and respond to diabetes-induced effects, leading to long-lasting changes in their phenotype, this was not affected by losartan. On the contrary, losartan decreased astrocyte activation, reduced proinflammatory factors and restored BDNF signalling, all of which may contribute to the alleviation of depressive symptoms. Literary data also strengthen our point regarding the existence of this relationship. Diniz et al infused a TrkB receptor antagonist into the ventral hippocampus and prelimbic prefrontal cortex of rats and this prevented the antidepressant effect of losartan [
45]. Furthermore, they subjected BDNF-haploinsufficient mice to FST and found that losartan decreased immobility time in wild-type rats but not in the rats with reduced BDNF levels. All these data support the direct link between losartan, BDNF and depression.
It is important to identify the exact molecular pathways through which impaired BDNF signalling may lead to the exacerbation of depressive symptoms in diabetes. Here, we found that proapoptotic p75NTR–JNK–B cell lymphoma 2-associated X protein (BAX) signalling is not activated, while the TrkB–ERK–CREB pathway is substantially impaired after 7 weeks of diabetes. However, the fact that despite losartan elevating proBDNF levels, proapoptotic p75NTR–JNK–BAX signalling was still suspended is rather interesting. It is plausible that losartan activated extra- and intracellular cleavage enzymes thereby promoting fast conversion of proBDNF to mBDNF. We confirmed this hypothesis by detecting increased furin and MMP3 levels in losartan-treated diabetic rats. Based on these results we postulate that proBDNF immediately transforms to mBDNF instead of activating its own receptor and downstream signalling.
The present study is the first to investigate the effect of ARB treatment on TrkB–ERK–CREB signalling in a model of diabetes. TrkB, p-ERK and p-CREB levels were decreased in the diabetic rat hippocampus, indicating that this pathway is repressed. Our results are in accordance with previous studies showing that TrkB and CREB activation is reduced in the hippocampus in diabetes [
46,
47]. Losartan treatment elevated the TrkB, p-ERK and p-CREB levels, indicating that this pathway is activated and promotes neuronal survival. Immunohistochemical analysis also lent support to p-CREB levels being massively increased by losartan in the diabetic hippocampus.
CREB is known to trigger the expression of several neuroprotective proteins including B cell lymphoma 2 (BCL2) and BDNF [
48]. Tanaka observed that p-CREB-positive neurons co-express BCL2 in the brain [
49] and Ramirez et al showed that neurotrophins upregulated BCL2 expression [
50]. In line with these observations, we show in the present study that losartan increased
Bcl2 expression and suggest that this may counteract the detrimental effects of diabetes-associated inflammatory changes in hippocampal neurons.
Based on these data we suggest that diabetes-induced neuroinflammation is, at least in part, responsible for decreased BDNF levels, which facilitate the development of depression-like behaviour. Our data also indicate that the upregulation of BDNF and p-CREB by ARBs may contribute to their neuroprotective effects in diabetic individuals and could be selectively targeted to alleviate some of the depressive symptoms associated with diabetes. Losartan also acts to restore the normal levels of BDNF via facilitating the conversion of proBDNF to mBDNF. In conclusion, our study suggests a novel potential of ARBs in diabetes-associated depression and may open up an additional therapeutic option for diabetic individuals.
Acknowledgements
Open access funding provided by Semmelweis University (SE). The authors would like to thank the Nikon Microscopy Center at IEM, Nikon Austria GmbH and Auro-Science Consulting Ltd. for technical support for fluorescence imaging. The authors acknowledge E. Mikics (Department of Behavioral Neurobiology, Institute of Experimental Medicine, Budapest, Hungary) for critical discussion of behaviour test data. We also thank D. Mathe (Department of Biophysics and Radiation Biology, Semmelweis University, Budapest, Hungary), D. Szollosi (Department of Biophysics and Radiation Biology, Semmelweis University, Budapest, Hungary) and CROmed Ltd. (Budapest, Hungary) for providing SPECT quantification methodologies. The authors specifically thank A. Molnar (MTA-SE Lendület Diabetes Research Group, Budapest, Hungary) and I. Kovacs (Department of Ophthalmology, Semmelweis University, Budapest, Hungary) for the statistical evaluation and E. Szkibinszkij (MTA-SE Lendület Diabetes Research Group, Budapest, Hungary) for technical assessment.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.