Background
There is accumulating evidence that depression may develop in response to activation of the innate immune system [
1‐
3]. Exposure of volunteers to a low dose of endotoxin induces depressed mood that correlated with cytokine expression, independent of sickness behaviors [
4]. Recently, a low dose of endotoxin given to volunteers was for the first time shown to induce anhedonia, one of the primary features (diagnostic = DSM IV) for depression [
5]. An increase in negative affect follows typhoid vaccine injections and similar to endotoxin exposure, these mood changes correlate with the induction of cytokine secretion [
6]. Studies such as these provide a correlation between mood changes and inflammation, but a direct cause-effect link between activation of the innate immune system and mood changes came with human cytokine immunotherapy. Cancer immunotherapy and cytokine treatment for hepatitis C viral infection induces symptoms of depression in a significant percentage of patients [
7,
8]. These symptoms develop on a background of neurovegetative symptoms that are very similar to inflammation-induced sickness behavior [
3]. Together with the Reichenberg study [
4] showing a dissociation between depression and overt sickness, there is now strong evidence that depression does not fully overlap with sickness and that depression may be caused by cytokines in the brain.
In a rodent preclinical model, activation of the immune system reliably induces depression-like behavior assessed by several criteria including decreased preference for a sweetened (saccharin) solution over water, as an index of anhedonia, decreased sexual behavior [
9], decreased preference for a sweetened (sucrose) solution over water, increased time of immobility during the forced swim test (FST) [
10] and increased time of immobility during the tail suspension test (TST) [
11]. LPS induces transient sickness with the changes in preference for a sweetened solution or immobility in the FST and TST still being evident after the disappearance of sickness; i.e. after locomotor activity, social exploration of a novel juvenile, body weight or food intake have normalized. These depression-like behaviors are reversed by anti-depressants and importantly by minocycline which attenuates LPS-induced expression of brain cytokines [
9,
11‐
14]. In all of these studies, depression-like behaviors continued after the acute immune response that was induced by LPS and the minocycline study clearly indicated that the cytokine response was requisite for the development of depression-like behaviors. These types of studies support the human data that inflammation is causative in the development or maintenance of depressive disorders.
Until recently, IGF-I has not been evaluated for anti-depression actions on a background of inflammation. We showed that i.c.v. IGF-I did not affect the acute sickness response that was induced by i.c.v. LPS. In contrast, IGF-I tempered cytokine expression and depression-like behavior [
11]. In that study, IGF-I also increased the central expression of BDNF, a neurotrophin with well-characterized anti-depressant activity. For that work, gene expression was quantified in cDNA prepared from the entire perfused brain of mice [
11,
13]. Whether, the LPS or IGF-I responses were global or localized with a specific brain region was not examined. However, following a single LPS injection, pro-inflammatory cytokines, IL-1β, TNFα and IL-6 are all similarly elevated in the hippocampus and frontal cortex of mice [
15]. Following repeated LPS injections, IL-1β is elevated in the frontal cortex, hippocampus and striatum [
16]. These studies suggest that LPS induces a global inflammatory response within the brain and justified our previous use of total brain mRNA as the source for cDNA to quantify an immune response following LPS. However, it is clear from studies with humans that the frontal cortex plays a unique role in depression [
17‐
20]. Similarly with rodents, electrical stimulation of the frontal cortex elicits hedonic vocalizations [
21], whereas lesions reduce play behavior [
22]. The antidepressant effect of fluoxetine on immobility in the TST was shown to correlate with BDNF expression in the frontal cortex, but not in the hippocampus [
23] implicating a unique role for the frontal cortex in depression-like behavior of mice. Thus in the current study, expression of genes associated with inflammation were quantified in the frontal cortex of mice.
IGF-I is well recognized as a neuroprotective hormone and paracrine growth factor displaying activities in a variety of neuropathologies [
24,
25]. IGF-I is cleaved within the brain to release an N-terminal tripeptide [
26]. GPE, like IGF-I, is neuroprotective both
in vivo and
in vitro [
27]. Lower amounts of GPE provide protection when given i.c.v. compared to i.p. suggesting a central site of action; indeed GPE does not appear to act outside the nervous system. Thus GPE may represent a centrally active IGF-derivative, separating it from the global role of IGF-I. Importantly, purification of IGF-I from brain yields IGF-I lacking the first 3 amino acids [
26,
28,
29] suggesting that GPE is naturally produced in the brain. Indeed, GPE is found in the normal brain [
30] and a protease that releases GPE is found in brain [
31,
32]. Protease inhibitors, that prevent the release of GPE from IGF-I, block IGF-I inhibition of GNRH secretion [
33,
34]. This later finding strongly suggests that GPE mediates at least part of the central action of IGF-I. These finding suggest that GPE is a natural product of central IGF-I and has naturalistic neuroprotective actions. A central function for GPE, outside of cell survival such as behavior modification, has not been reported. The current study shows that like IGF-I its' natural cleaved product, GPE, has behavior modifying activity. In the present series of experiments, we show that both IGF-I and GPE administered centrally are able to prevent and cure depression-like behavior induced by peripheral administration of LPS.
Discussion
Three new findings are evident from the current studies, 1) although they both share an inflammatory component, sickness and depression-like behavior are independently modulated by the IGF system, 2) when added exogenously, two naturally occurring products found in the naïve brain, mature IGF-I and GPE, act centrally to temper LPS-induced changes in depression-like behavior, and 3) GPE can temper the central innate immune response to LPS providing a possible mechanism by which to regulate depression-like behavior. A direct behavior modifying role of GPE has never been reported. Previous studies with the GPE peptide had shown biological activity associated with neuroprotection using preclinical models for Alzheimer's, Parkinson's disease, stroke, multiple sclerosis and Huntington's [
27], but a central behavioral effect that is independent of neuroprotection had yet to be reported. Neither IGF-I or GPE affected behavior of naïve (non-LPS) mice, but had anti-depressant activity when tested against the inflammation-inducer LPS in the TST and FST. It's not known whether the effect of GPE can be extended to other models of depression as GPE was not able to reverse LPS-induced decrease in sucrose preference, used as an index of anhedonia.
Other studies have shown that IGF-I has anti-depressant activity using various rodents models. Voluntary exercise increases blood and brain IGF-I levels. This increase is in part responsible for exercise-induced anti-depressant activity [
39]. An anti-IGF-I antibody blocked the decreased time of immobility during the FST of exercised mice [
40], indicating that exercise-induced increases in IGF-I has a naturalistic and endogenous role in controlling behavior. Similarly, rough-and-tumble play of rats increases IGF-I levels in the frontal and parietal cortex and also increases hedonic 50-kHz ultrasonic vocalizations [
41]. The play-induced increase in hedonic vocalizations was dependent on IGF-1R activation and i.c.v. administration of IGF-I increased hedonic vocalizations and decreased approach latency to self-administered play. IGF-I also decreased time of immobility in the FST, decreased consumption latencies in the novelty-induced hypophagia test using naïve mice and increased sucrose consumption of mice subjected to chronic unpredictable stress [
40,
42‐
44]. Activating endogenous IGF-I, by freeing it from inhibitory IGF-binding proteins with NBI-31772, decreased immobility in the TST. The action of IGF-I and NBI-31772 was blocked by JB1, an IGF-1R antagonist [
44] supporting a mechanism that requires activation of IGF-1R. A single i.c.v. injection of IGF-I to rats decreased immobility during the FST [
42] while JB1 blocked the anti-depressant activity of IGF-I [
43]. Data such as these, especially with naïve animals, show that the behavior modifying role of endogenous and exogenous IGF-I is independent of IGF-I's well documented neuroprotective actions and that the anti-depressant activity of IGF-I can be quantified with several tests that have proven to be reliable indices of depression-like behaviors. Blocking the anti-depressant activity of exogenous IGF-I with JB1, which prevents IGF-1R receptor activation, shows that IGF-1R activation is requisite for IGF-I action, but does not show whether all the effects are mediated by the IGF-1R. With this in mind we determined whether the naturalistic product of IGF-I cleavage within the brain that acts independent of the IGF-1R [
45] can exert additional anti-depressant activity.
In the current study, GPE had very similar actions compared to IGF-I. GPE, given before or after LPS, elicited an anti-depressant effect with mice using either the TST or FST. Like IGF-I, GPE did not alter the sickness response to i.p. LPS. The lack of an effect of these peptides on sickness is not unexpected. LPS given i.p. elicits a strong peripheral immune response and also sends inflammatory-initiating signals to the brain via the vagus nerves and other immune-to-brain communication pathways [
46]. The acute peripheral immune response, subsequent delivery of cytokines to the brain via the humoral and neural communication pathways is unlikely to be altered by i.c.v. administration of either IGF-I or GPE. Thus in this model, the sickness response appears to be largely under the control of the peripheral inflammatory input to the brain. These primary signals subsequently initiate an activation of the innate immune system of the brain which is tempered by GPE (Figure
5). Decreased neuroinflammation in response to IGF-I [
11] and GPE (current study) correlate with changes in depression-like behavior when assessed using the TST and FST. The anti-depressant activities of IGF-I and GPE in the current experiments assessed with the TST and FST does not appear to be a result of an overall attenuation of psychomotor retardation as locomotor activity was unaffected by either IGF-I or GPE. GPE was unable to alter mouse performance in a test for anhedonia. Sucrose preference was not affected by GPE. Attempts to assess the activity of IGF-I on sucrose preference were complicated by the ability of an acute IGF-I challenge to enhance sucrose consumption in naïve mice (data not shown), reflecting its insulin-like activity and ability to directly alter glucose metabolism. Differential effects with the TST and FST versus sucrose preference suggest that GPE plays a specific role in alleviating only some symptoms of inflammation-induced depression, helplessness/despair, without affecting another, anhedonia. An hedonic response may require mature IGF-I or chronic exposure. In a study by Duman, chronic subcutaneous infusion of IGF-I (~1-1.5 mg/day for 14 d, vs. a 1 mg bolus in the current work) did not increase sucrose consumption in naïve mice (an effect we have found with bolus IGF-I injection i.c.v.), but this constant chronic infusion of IGF-I reversed the lowered sucrose consumption by mice subjected to chronic unpredictable stress [
40]. Our new finding however provides a novel model to help define central pathways controlling specific behavioral symptoms.
We previously found that IGF-I induced the expression of several BDNF transcripts [
11]; another possible mediator of its anti-depressant activity. One of the more interesting aspects of BDNF biology is the differential expression and regulation of specific transcripts. All transcripts produce mature BDNF, but in different cells types. Transcripts initiating from exons I, II and III are expressed predominantly in neurons and transcripts initiating from exons IV, V and VI are expressed by both neurons and astrocytes [
47]. Normal expression of all transcripts may be necessary for wellness, as a knockout of even a single transcripts causes depression-like behavior of mice [
48]. Duloxetine, an SNRI class anti-depressant, increases the expression of only 4 of the 9 transcripts [
49] indicating specificity of treatment. Thus we again examined multiple BDNF transcripts. However, GPE did not induce BDNF expression and thus BDNF may not play a significant role in the anti-depressant role of GPE. Only a single dose of GPE was tested in the present study; a dose with proven neuroprotective action [
35,
36]. The present results call for a more extensive investigation of the dose-response effect of GPE on inflammation-induced depression and moreover on the ability of endogenously generated GPE to mediate part of the anti-depressant activity of IGF-I.
Clearly from studies mentioned above, the anti-depressant activity of mature IGF-I is dependent of IGF-1R signaling, which also mediates IGF-I's neuroprotective action. The receptor that mediates GPE action on behavior has not been characterized. GPE was shown to displace glutamate from its receptor, but GPE's neuroprotective activity is not clearly related to glutamate/NMDA receptor (NMDA-R) binding as it competes with glutamate binding with a very low affinity (30 μM) even though neuroprotective activity is evident at 0.1 nM [
33,
34,
45,
50,
51]. This discrepancy suggests that GPE activity is dependent on an as yet uncharacterized mechanism [
52]. Peptidase inhibitors that block processing of many proteins including the release of GPE from intact mature IGF-I with the generation of des-(1-3)-IGF-I, block IGF-I inhibition of GNRH secretion [
33,
34] suggesting that GPE mediates at least part of the central action of IGF-I. Mature intact IGF-I has antidepressant activity, but it is still not known if des-(1-3)-IGF-I, acting via the IGF-1R, or the generation of GPE from exogenous IGF-I play a significant role in the antidepressant activity of the intact mature peptide. However, since blockage of IGF-1R activity abrogates anti-depressant activity of the mature peptide [
44] and GPE does not activate the IGF-1R [
45], clearly IGF-I itself has anti-depressant activity through the IGF-1R independent of GPE. Thus it appears that IGF-I has the potential to generate two distinct signaling events that result in anti-depressant activity; an IGF-1R mediated action and an IGF-1R independent GPE-mediated action.
One common link between IGF-I and GPE that may be related to their anti-depressant activity is their common ability to attenuate the activity of quinolinic acid [
53,
54]. Induction of tryptophan dioxygenase activity by inflammation plays a major role in depression as shunting of tryptophan metabolism toward the generation of metabolites such as quinolinic acid has been linked to inflammation-associated depression [
55]. Using mice as a preclinical model, i.p. LPS induces IDO1 activity in the periphery and brain [
56] with an increase in central and plasma kynurenine/tryptophan ratios [
13]. Increased IDO1 expression is associated with the generation of kynurenine metabolites (kynurenic acid, 3-hydroxykynurenine and quinolinic acid) that are active neuromodulators. Increased levels of these modulators are associated with inflammation-dependent depression [
57]. It is not clear as to which of these metabolites have a causative role in depression or if periphery or brain-generated tryptophan metabolites are responsible for depression-like behaviors but induction of peripheral or central IDO1 expression correlates with depression-like behavior of mice. IDO1 expression was quantified using two qPCR assays. IDO1 protein is translated from multiple transcripts with the full length transcript being less widely distributed in the body than other transcripts [
58]. Although not reported for brain samples, in the placenta the major transcript does not include much of exon 1 [
59]. An assay designed to probe for expression of IDO1's full-size transcript (probe spanning exons 1 and 2) resulted in weak amplification in controls (Ct = 38.7). Expression of all major IDO1 variants was quantified using a probe that spans exons 3 and 4. IDO1 ex. 3-4 amplification is much stronger in control brain samples, Ct = 27.4, than that of IDO1 ex. 1-2. IDO2 and TDO2 expression were quantified with probes that span exons 3 and 4 and 4 to 5, respectively. These assays quantify all major transcripts. Amplification of IDO2 in control mice was slightly stronger than that for IDO1 ex. 3-4 with a Ct value of 24.4, whereas amplification of TDO2 was weaker, Ct = 30.5. Although Ct values do not imply differences in mRNA levels within samples, because of probable differences in assay-to-assay qPCR efficiencies, the extremely low Ct value for IDO1 ex. 1-2 compared to all other dioxygenases indicate that in control animals it may be a minor source for protein translation. However, expression of IDO1 ex. 1-2 may play an important role in neurophysiology because of its strong induction with neuroinflammation; in this case following administration of LPS. In our current model, IDO1, IDO2 AND TDO2 mRNA expression, which are rate-limiting enzymes for the conversion of intracellular tryptophan to kynurenine, were all increased by LPS. Expressions were unaffected by GPE; similarly IGF-I did not attenuate LPS-induced dioxygenase expression (unpublished data). These data seem to suggest that IGF-I and GPE do not act by altering the consequences associated with elevated dioxygenase expression. However, it remains to be determined if IGF-I or GPE alter the effects of elevated tryptophan dioxygenase expression downstream of the enzyme; i.e. by directly attenuating the activity of quinolinic acid or other downstream products such as 3'-hydroxykynurenine or kynurenic acid.
Cytokine and neurotrophin mRNA expression was quantified in the prefrontal cortex as this region has been strongly implicated in the regulation of positive affect. In rodents, electrical stimulation of the frontal cortex elicits hedonic vocalizations by rodents, [
21], whereas lesions reduce rough-and-tumble play behavior [
22] both associated with a positive emotional state. Although many studies show a strong correlation between changes in the frontal cortex and hippocampus, the antidepressant effect of fluoxetine was shown to correlate with BDNF expression in the frontal cortex, but not in the hippocampus [
23]. With humans, several positive affective stimuli increase neuronal activity in the frontal cortex [
17‐
20]. Here we show that LPS induces expression of central cytokine and GPE reduces their expression in a brain region intimately associated with depression-like behaviors. The opposing effects of LPS and GPE on cytokine expression parallel changes in depression-like behavior.
The neurotrophic hypothesis for depression asserts that neuronal plasticity is dysfunctional because of insufficient neurotrophic support and that classic anti-depressants ultimately reactivate plasticity in the brain [
60‐
62]. IGF-I is a classic neurotrophin; inducing the proliferation, survival, differentiation and maturation of cells of the central nervous system [
63] and directly supporting neuronal plasticity [
64]. GPE is also a neurotrophin; although characterized primarily by its ability to support neuronal survival and stimulation of glia proliferation [
65]. IGF-I is expressed by neurons within the brain [
66‐
68] and expression was reduced by the administration of LPS (Figure
7). This reduction would translate into low central levels of both IGF-I and GPE within the frontal cortex. This reduced protein expression would result in lower neurotrophic support along with the lower expression of BDNF and VEGF (Figure
7). By administration of IGF-I or GPE neurotrophic support is exogenously provided and depression-like behavior is attenuated.
In conclusion, data in this manuscript show that exogenous central administration of either IGF-I or GPE results in anti-depressant-like activity, assessed using the TST and FST. The behavioral changes did not occur with naïve mice but these neurotrophins attenuated the behavioral changes induced by i.p. LPS. These data extend the anti-depressant activity of the IGF system by showing that a proteolytic product derived from IGF-I, GPE, has anti-depressant activity paralleling that of the mature peptide. Both IGF-I and GPE offset the inflammatory response elicited by LPS which may in part be responsible for anti-depressant activity.
Competing interests
RD has received honoraria from Bristol-Myers Squibb and Lundbeck Laboratories and works as a consultant for Lundbeck Laboratories. RD and KWK have received honoraria from Astra-Zeneca.
Authors' contributions
RHM designed the experiments, performed surgery, treated animals, performed behavior tests, analyzed data and wrote much of the manuscript. SP performed surgery, treated animals, performed behavior tests, analyzed data and did all the PCR analyses. ML, KWK and RD helped in the design and interpretation of the experiments and editing the manuscript. All authors have read and approved the final version of this manuscript.