Background
The
Plasmodium parasite is estimated to infect hundreds of millions worldwide carrying an annual death toll of approximately 0.5 million [
1]. Cerebral malaria (CM) constitutes a large portion of the mortality burden. The pathogenesis of CM is complex involving a vascular component with parasite-induced venule blockage, platelet activation, blood–brain barrier (BBB) disruption and an inflammatory component with activation of astrocytes, microglia, complement system, and lymphocytes. Ultimately; the consequences of this dysfunction can be seizures, coma and death [
2‐
4]. Oxidative stress is present throughout the body from the endogenous immune response and the presence of reactive free haem from lysis of infected erythrocytes (iRBC) [
5‐
7]. Interestingly, although oxidative stress is present systemically, it has been suggested that neither cerebral mitochondrial performance [
8,
9] nor antioxidant buffering [
10‐
13] seems decisively compromised in fatal murine and human CM.
Glucagon-like peptide-1 (GLP-1), recently emerging as a front-line treatment for type II diabetes mellitus (DM), has been shown to have potent neuroprotective properties. GLP-1 receptor agonist treatment improves outcomes in experimental models of stroke, Parkinson’s disease and Alzheimer’s disease [
14‐
16]. Moreover, the GLP-1 analogue, liraglutide, is neuroprotective against murine traumatic brain injury (TBI) and this effect is closely associated with the cytoprotective cAMP response element binding protein (CREB) pathway [
17]. In addition to the potential cytoprotective effects of CREB-driven protective proteins, GLP-1 has also been suggested to promote endothelial cell nitric oxide production, up-regulate haem oxygenase-1 [
18,
19], and reduce neuroinflammation in vivo [
20‐
23], all previously reported to protect against experimental CM (ECM) [
7,
20,
24]. The broad spectrum of neuroprotective effects reported after GLP-1 receptor agonist treatment and its safety profile in humans led us to investigate its potential as a therapeutic treatment against murine CM. In this investigation, this therapeutic potential is assessed with the GLP-1 receptor agonist, liraglutide: a long-acting GLP-1 analogue designed to extend the half-life of GLP-1 receptor activation [
25] that can cross the BBB [
26].
Finally, the prevalence of obesity and type II DM is rising in malaria endemic regions of low and middle-income countries [
27] and therefore the safety of liraglutide treatment on parasite growth and metabolism was evaluated.
Methods
Liraglutide treatment of Plasmodium berghei-induced cerebral malaria
All animal experiments were approved by the Danish Animal Experiments Inspectorate according to the license 2012-15-2934-00449. Survival study experiments were designed from a power analysis that suggested 14 mice per group would be powered to 80 % at 95 % confidence for detecting a 50 % reduction in mortality. Animals were kept under standard conditions with food/water access ad libitum, and all studies were conducted to minimize suffering and in accordance with a pre-defined humane endpoint of body temperature <32 °C [
28].
Stock
Plasmodium berghei strain ANKA parasites characterized previously for disease progression were thawed and injected into a pilot mouse. After 4 days of in vivo growth, 10
4 iRBCs were injected intra peritoneal (i.p.) into female C57Bl/6j mice for modeling ECM or female Balb/c mice for modeling non-cerebral malaria (NCM) [(Taconic, Denmark), age 6–8 weeks]. To evaluate the potential neuroprotective effects of liraglutide in ECM, liraglutide was administered at 200 μg/kg twice-daily. Initial pilot experiments in ECM were performed with 50, 200, 400 μg/kg and obtained similar results. In previous work, three increasing doses of liraglutide (100, 200, and 400 μg/kg) were investigated as a protective agent in a model of murine TBI [
17] where 200 μg/kg was the most effective against brain damage. Moreover, this dose is matched to the dosing in the clinic for anti-diabetic effect in mice [
29]. Thus, high-powered experiments were performed at a dose of 200 μg/kg. Maintaining a translational focus in the study, liraglutide was administered from day 4 post-infection to mimic a critical care scenario: day 4 coincides with detection of parasitaemia, and the beginning of damage to the olfactory bulb [
30].
Thus, on day 4 post-infection mice were randomly assigned to different treatment arms: twice-daily vehicle s.c. (PBS; n = 15), twice-daily liraglutide s.c. (Novo Nordisk, DK; n = 15), or once-daily 5000 IU recombinant human Erythropoietin i.p. (Epo, Eprex
®, Janssen-Cilag, Switzerland; n = 8). Epo treatment from day 4 completely protects from ECM development [
8,
31] and was therefore included as a positive control group. This ECM study was repeated to confirm results with smaller sample sizes [n = 7 (vehicle), 8 (liraglutide)]. NCM mice were included into evaluate the interaction between liraglutide and parasite growth and to control for the cerebral component of the
P. berghei infection. These two treatment arms (vehicle and liraglutide) were included in an experiment with ECM and NCM mice (n = 10 for all groups and treatments).
Mice were monitored twice daily on day 4 and three times daily from day 5. 2 μL of blood from the tail vein were sampled from day 5 for parasitaemia enumeration. Parasites were enumerated with flow cytometry using acridine orange labeling and FACScanto flow cytometer (BD Biosciences, CA, US) as previously described [
32].
ECM progression was monitored based on presentation of ruffled fur, ataxia, hemiplegia, seizures, coma and core body temperature reduction [
28,
33]. This study was conducted with a pre-defined humane endpoint: animals were deemed ‘terminal’ when they registered a body temperature below 32 °C. Body temperature was measured three-times daily, non-invasively with an infrared thermometer (Testo 845, Testo, Germany) and further confirmed with rectal probe (DM852, Ellab, DK) if the measurement was below 32 °C as described in [
34]. Mice were killed by cervical dislocation and brain was observed after removal of skull for presence of haemorrhage to further verify the diagnosis of ECM. Finally, a group of healthy C57Bl/6j mice were included and treated from day 4 until day 8 with vehicle or liraglutide (n = 8, 8).
Reactive oxygen/nitrogen species visualization
In order to visualize ROS/RNS species generation in CM infection in real-time, the highly sensitive luminol derivative, L-012 was used [
35].
Plasmodium berghei-infected mice were injected with 75 mg/kg L-012 i.p. from day 5 post infection. Mice were anesthetized with isoflurane and chemiluminescence was imaged (dorsal orientation) with a CCD camera 15 min after injection with a 60 s exposure. Thereafter a supine acquisition was taken (60 s exposure). In vivo images were acquired once daily thereafter. Animals that were deemed terminally ill were injected with the luminescent probe and cervically dislocated 15 min post-injection. Organs were rapidly exposed for light emission and an ex vivo image was acquired (60 s). Epo prevented ECM in all mice and thus were terminated at the end of the experiment (day 9) along with uninfected mice for imaging of the organs. In a repeat experiment,
P. berghei-infected mice were administered vehicle from day 4 and ex vivo tissue images were acquired in mice deemed terminal. Images were quantified with Living Image software (Perkin-Elmer, USA). All quantifications are reported as the difference between regions of interest and background signal.
Cerebral CREB, cAMP, and nitrate/nitrite levels
Brains were removed, split into olfactory bulb, cerebellum, and two cerebral hemispheres, snap frozen in liquid nitrogen and maintained at −80 °C. One hemisphere was designated for immunoblotting and nitrate/nitrite assay and was homogenized in PhosphoSafe extraction reagent (catalogue number 71296, Millipore, DK) with protease inhibitor cocktail (mini complete, Roche, DK), protein content quantified (BioRad DC protein assay kit, DK), and stored at −22 °C.
One hemisphere was processed and assayed for cAMP determination as per manufacturers recommendations (Enzo Life Sciences, DK) in hydrochloric acid.
Immunoblotting was optimized and performed with standard western blot principles. All homogenates were thawed and mixed in sample buffer (100 mM DTT, Laemmli buffer) and heated at 95 °C for 5 min. Based on optimization experiments, 25 µg protein per animal from one cerebral hemisphere was loaded into pre-cast polyacrylamide gels (12 % bis–tris; NuPAGE, Life Technologies, DK) and gels run in MES buffer at 150 V for 60 min. Gels were transferred to PVDF membranes (BioRad) at 30 V for 60 min. Membranes were then washed in TBS and placed in blocking solution (5 % skim milk powder—(CREB) or bovine serum albumin- (phospho-CREB) TBS-Tween20 (TBST) for 1 h at room temperature. Primary antibodies were applied CREB (1:1000, Cell Signaling, 9197), pCREB (Serine 133) (1:1000, Millipore, 06-519), and GAPDH (1:10 000 Millipore, MAB 374) in appropriate blocking solution overnight at 4 °C. Membranes were washed three times in TBST and secondary antibodies were applied in appropriate blocking solution: HRP-conjugated anti-rabbit/anti-mouse (DAKO, DK) at 1:2000 and 1:3000 respectively for 1 h at room temperature. Membranes were washed three times in TBST and once in TBS. Membranes were incubated in SuperSignal Femto substrate (34095, Fisher, DK) and exposed with CCD camera (Bio-Rad Chemidoc XRS imager). Images were quantified with ImageJ software [
36]. All protein signals are reported relative to the housekeeping protein, GAPDH.
Nitrate/nitrite assay was performed as per manufacturer’s recommendations (Cayman Chemical, DK) and normalized to protein content.
Plasmodium falciparum growth and metabolism
Stock
P. falciparum parasites of CSA-selected Palo Alto (PA) strain were chosen to investigate the effects of liraglutide on parasite growth based on previous growth criteria as described in detail in [
37]. Parasites maintained at the Centre for Medical Parasitology were thawed and cultured at 37 °C in Albumax-enriched, HEPES buffered RPMI culture media: RPMI (Gibco, DK), albumax (Gibco), hypoxanthine (Sigma-Aldrich, DK), and gentamicin (Gibco) [
37]. Stock cultures were diluted to 0.5 % and cultured in sterile 96-well microplates in vehicle (PBS) or increasing doses of liraglutide (0.1–100 μg/mL). Artesunate (10 ng/mL, Sigma-Aldrich) was included as a positive control for growth inhibition. Media and treatments were removed and replenished after 24 h and parasitaemia quantified by flow cytometry [
32]. In a repeat experiment,
P. falciparum parasites were cultured in sterile 24-well plates with the same treatment groups. Media and drugs were exchanged and replenished after 24 h. After 48 h, media was removed, spun at 4 °C and the supernatant frozen at −80 °C for metabolite quantification.
Media from 24-well cultures were thawed and fed into a blood-gas and metabolite analyzer (ABL 725 Analyzer, Radiometer, DK) as described previously [
37]. Media were assayed for concentration of glucose, lactate, H
+ (pH), Na
+, K
+, and Ca
2+.
Data analysis
Data sets were tested for normality (Shapiro–Wilk) and equal variance (Barlett’s) before statistical analyses were performed. Non-normal data sets were log-transformed, and tested for normality. All data sets were either normal or normalized after log-transformation and thereafter analysis was performed with parametric tests. Survival data were analysed with log-rank test of Kaplan–Meier curves. One-way ANOVA with Holm-Sidak correction was performed on ex vivo L-012 chemiluminescence data, immunoblotting, cAMP and nitrate/nitrite assays, and cultured-parasite growth and metabolism data. Two-way ANOVA repeated measures was performed on in vivo parasitaemia data and in vivo L-012 chemiluminescence data. Normal data is presented as mean + SEM. and log transformed data as geometric mean + 95 % confidence intervals unless otherwise stated. Significant differences are reported with p values below 0.05. Statistical analyses were performed with Graphpad Prism 6 software.
Discussion
Despite the premise for a therapeutic effect of GLP-1 receptor agonism against ECM cerebrovascular pathology these results indicate that liraglutide does not protect against ECM. After pilot testing of various doses and rigourous experimentation with three high-powered studies (Fig.
1a), these data suggest that the finding presented here is robust for this analogue. With the scope of neuropathological conditions that GLP-1 treatment improves [
14‐
17], this finding is insightful for understanding both the pathogenesis of ECM and the neuroprotective mechanism of GLP-1 receptor agonism.
Plasmodium berghei ANKA infection initiates an oxidative stress component from rupturing of RBCs releasing free haem, hemozoin and immunogens and activated phagocyte respiratory bursting. Moreover, leukocytes isolated from murine and clinical CM generate higher ROS levels than those from healthy and non-CM controls [
5,
6]. ROS/RNS were visualized in vivo using a sensitive probe, L-012, a luminol derivative with a higher sensitivity and signal to noise ratio than commonly used luminol and lucigenin [
38,
39].
As suspected, these data show that mice with
P. berghei infection are burdened with an increased ROS/RNS load throughout the body (Fig.
2). However, no emission signal was detected from the damaged brains, despite strong signals from damaged peripheral tissues (lungs, spleen, adipose). The absence of cerebral ROS/RNS signal through the development of fulminant, terminal ECM is intriguing. This finding is robust through two independent infections and a total of 27 mice in terminal-stage ECM (vehicle: n = 19, liraglutide: n = 8). Past experiments with this imaging molecule have shown penetration in the brain [
35], including work in TBI where the damaged brain shows high levels of ROS/RNS [
17].
The role of ROS/RNS in the pathogenesis of ECM is indeed unclear [
8,
9,
11,
12,
40,
41]. Antioxidant treatments can be protective against ECM [
41] and lipid peroxidation increased and cerebral thiols decreased in late-stage ECM [
40,
42,
43]. Moreover, neutrophils accumulate in the ECM brain over time [
44]. Finally, micro-haemorrhage sites in the cerebellum show signs of possible stress in the vicinity of the haemorrhage [
45]. Conversely, antioxidant therapy was unsuccessful in treatment of human CM and hydrogen sulfide donors were ineffective against ECM [
10,
37]. Indeed, many antioxidant treatments can also have effects on immune function [
12,
46], which may partly explain their efficacy in ECM. Furthermore, although expression of some genes involved with oxidative stress is reduced in ECM [
11], these seem to be compensated by increased anti-oxidant gene promotion [
11,
47]: molecular markers of ROS damage (ROS-modified proteins) are largely unchanged [
11,
12], mitochondrial function is preserved in ECM [
8] and genetic removal of the primary endogenous reactive species generator NADPH oxidase does not improve ECM outcome [
12]. Furthermore, poly(ADP-ribose) polymerase-1 knock-out mice are not protected from ECM, indicating that cytopathic hypoxia is not driving ECM [
9]. These data are in support of these findings.
In a model of TBI, liraglutide reduced ROS/RNS and may have preserved mitochondrial function [
17]. It is possible that an important component of the neuroprotective effect of GLP-1 involves the downstream buffering of cellular homeostasis and/or protection of mitochondrial function. These effects may be redundant in ECM and thus, be ineffective against fatality.
ROS/RNS were elevated in terminal mice in the lungs, spleen and adipose tissue- the preferred tissues where
P. berghei parasites accumulated in previous studies [
48,
49]. Interestingly, L-012 photon emission decreased in terminal mice and may reflect vascular collapse often associated with ECM vasospasms and increased RBC/iRBC/platelet/leukocyte aggregation in vessels throughout the body [
50]. Therapeutic Epo treatment reduced reactive species in the lungs, spleen and adipose tissue. It is unclear whether this is a direct effect or may be associated with the reduced parasite load in Epo treated mice, similar to previous reports [
31].
The CREB pathway is a well-described transcriptional pathway associated with the production of a wide range of protective proteins in the brain [
51]. This appears to be the first assessment of CREB dynamics in the brains of ECM and NCM mice. It is clear from Fig.
3a, that the cerebral CREB system is activated specifically in ECM but not in healthy or NCM mice. Since CREB is involved in up-regulation of proteins with anti-oxidant properties and preservation of the mitochondria (e.g. haem oxygenase-1, peroxisome proliferator-activated receptor gamma coactivator 1-alpha, neuroglobin [
52‐
54]), activation of this system in ECM may play a role in preventing ROS/RNS damage. Indeed, previous work by Linares et al. [
11] shows that haem oxygenase-1 is markedly increased in mice with ECM and considered an important counterweight to potential oxidative damage. The activation of the CREB system observed here might be, in part, responsible for this mechanism.
In previous work, liraglutide treatment was protective after TBI and was associated with substantial pathology-dependent CREB activation and increases in CREB-regulated protective proteins [
17]. Interestingly, CREB was not activated by liraglutide in ECM. This may be related to absence of ROS/RNS stress in the brains of ECM mice whereas ROS/RNS were markedly increased in TBI. Furthermore, it is intriguing that cAMP levels were increased in healthy mice by liraglutide but this did not translate to CREB activation. A similar phenomenon was observed previously, where liraglutide treatment did not activate CREB until pathology was introduced [
17]. It seems that certain pathological events may be required for liraglutide to induce the CREB pathway. This work in ECM suggests that marked oxidative stress may be important for a therapeutic effect of GLP-1 treatment in the brain. Differences in the therapeutic effect of GLP-1 may also be related to changes in GLP-1 receptor expression during different pathological states. GLP-1 receptors are distributed in distinct brain regions in the healthy rodent brain [
55]. Nevertheless, data is lacking on the receptor distribution under different pathological states. Finally, although GLP-1 treatment is known to increase the nitric oxide synthase enzyme, no increases in nitrous products downstream of nitric oxide production were detected in the brain. This may also contribute to the lack of effect of liraglutide in ECM, where nitric oxide is a known protective agent [
20].
With an estimated 317 million people living with type II DM, a large portion of the burden resides in malaria endemic regions of the world [
27]. It is therefore of interest to investigate the safety of type II DM treatments for potential interactions between a drug and parasite growth and development of malaria. These data show that
P. falciparum growth in vitro and metabolism (Fig.
4) is unaffected by increasing doses of liraglutide (0.1–10 μg/mL). At a 100 μg/mL dosage, liraglutide treatment exerts inhibitory effects on parasite growth though is not likely to be achieved clinically due to gastrointestinal side effects associated with much lower doses [
56,
57]. With three large in vivo experiments and in vitro experiments involving the human pathogen, it does seem likely that liraglutide interacts adversely with
P. falciparum in humans. This investigation therefore suggests that liraglutide is likely safe for administration under co-morbidity circumstances with type II DM. Active post-marketing surveillance should still be applied.
Authors’ contributions
BD designed the study, conducted all the experiments, analyzed the data, and drafted the manuscript. CH designed the study and analyzed data. TS conducted the parasite growth experiments with BD. FFJ designed the study. JALK designed the study and analyzed data. All authors read and approved the final manuscript.