The heart is controlled by the sympathetic and parasympathetic limbs of the autonomic nervous system with inhibitory signaling mechanisms recruited in both limbs. The aim of this study was to determine the role of inhibitory heterotrimeric G proteins in the central nervous mechanisms underlying autonomic control of the heart and its potential role in arrhythmogenesis. Mice with conditional deletion of the inhibitory heterotrimeric G protein GαO in the presympathetic area of the rostral ventral lateral medulla (RVLM) were generated to determine the role of GαO‐mediated signalling in autonomic control and electrophysiological properties of the heart. GαO deletion within the RVLM was not associated with changes in heart rate (HR) or the arterial blood pressure at rest (home cage, normal behavior). However, exposure to stressful conditions (novel environment, hypoxia, or hypercapnia) in these mice was associated with abnormal HR responses and an increased baroreflex gain when assessed under urethane anesthesia. This was associated with shortening of the ventricular effective refractory period. This phenotype was reversed by systemic beta‐adrenoceptor blockade, suggesting that GαO depletion in the RVLM increases central sympathetic drive. The data obtained support the hypothesis that GαO‐mediated signaling within the presympathetic circuits of the RVLM contributes to the autonomic control of the heart. GαO deficiency in the RVLM has a significant impact on cardiovascular responses to stress, cardiovascular reflexes and electrical properties of the heart.
The heart is controlled by the autonomic nervous system via its two functional limbs: the sympathetic and parasympathetic (Spyer 1994). Sympathovagal imbalance can trigger ventricular arrhythmias in patients with inherited “channelopathies” such as the long QT syndrome, Brugada syndrome, catecholaminergic polymorphic ventricular tachycardia (Verrier and Antzelevitch 2004), and also in patients with ischemic heart disease (Nolan et al. 1998). Sympathetic denervation has been used as an adjunct to pharmacotherapy with beta‐adrenoceptor antagonists in the management of patients with resistant ventricular arrhythmias (Vaseghi et al. 2014). However, the efficacy of current therapeutic strategies remains limited, perhaps reflecting the fact that only the downstream peripheral mechanisms are targeted.
Heterotrimeric G proteins mediate signaling via G‐protein coupled receptors (GPCRs). GPCRs are involved in autonomic control at the level of the heart via beta adrenergic and muscarinic receptors signaling via stimulatory GαS and inhibitory Gαi/o proteins. There appears to be specificity of Gα protein subtypes in vivo. Zuberi et al. (2008) have demonstrated the importance of the inhibitory G proteins Gαi2 and GαO in the autonomic control of the heart using global knockout (KO) mouse models. We have recently shown that Gαi2 is important in mediating parasympathetic influences at the level of the cardiac conduction tissue, probably by coupling to M2 muscarinic receptors (Sebastian et al. 2013). What remains unclear is the exact role played by the inhibitory G proteins in the central nervous mechanisms of autonomic regulation of the heart.
GαO protein is a key inhibitory G protein subtype in the central nervous system (CNS) (Sternweis and Robishaw 1984) including the brainstem (Parker et al. 2012). Mice with global deletion of GαO are tachycardic and display loss of diurnal rhythm in heart rate (HR) and selective loss of the low‐frequency component of HR variability with preserved total power (Zuberi et al. 2008). However, carbachol still had a negative chronotropic effect in this model, suggestive of increased sympathetic activity (Zuberi et al. 2008). The rostral ventrolateral medulla (RVLM) contains principal presympathetic circuits which generate central sympathetic drive (Head and McCarty 2002; Nalivaiko et al. 2005). Increased activity of the RVLM neurons has been implicated in the development and progression of cardiovascular diseases associated with enhanced central sympathetic drive including hypertension (Marina et al. 2015) and heart failure (Marina et al. 2011). We hypothesized that the inhibitory influences mediated via GαO proteins in the RVLM are important in controlling (restricting) sympathetic tone. This would broadly impact on the blood pressure homeostasis and cardiac function, particularly HR and ventricular excitability. To test this hypothesis, we generated mice with conditional deletion of GαO in the RVLM and studied their cardiovascular phenotype, cardiovascular reflexes, and ventricular excitability in vivo.
Mice were maintained in an animal core facility under the UK Home Office guidelines relating to animal welfare. All procedures were approved by the local animal care and use committee and performed in accord with the UK Home Office regulations. All mice were kept in a temperature‐controlled environment (21–24°C) with 12/12 h light/dark cycle. Animals were allowed ad libitum access to standard rodent chow and drinking water. Mice were studied between 8 and 12 weeks of age.
GαO flx/flx mice on a Sv129 background with coding exons 5–6 of the GαO allele on chromosome 8 flanked by loxP sites were generated using homologous recombination in mouse ES cells (Ustyugova et al. 2012). Gene targeting strategies, genotyping primers, and confirmation of selective deletion of GαO at the protein level from various tissues including the brain using western blot approaches have been previously described (Ustyugova et al. 2012). Mice with conditional deletion of GαO in the RVLM were then generated following microinjections of adenoviral vectors (AVV) coexpressing Cre and GFP under the control of a CMV promoter (Cre/GFP‐AVV) into the RVLM of GαO flx/flx mice. Control mice were produced using GFP‐only‐expressing AVV under the control of the same CMV promoter (GFP‐AVV). These vectors have been described previously (LaVallie et al. 2006; Lopez et al. 2008) and were obtained from Viraquest (USA). In preliminary studies, quantitative RT‐PCR using TaqMan gene expression assays was performed at sites of injections to confirm knockdown of GαO expression and determine the optimum dilution of viral titers to be used. Immunohistochemistry was performed at the end of all the experiments to determine colocalization of GFP expression with tyrosine hydroxylase (TH) immunoreactivity in the RVLM.
Experiments in conscious freely moving mice
Conscious freely behaving GαO flx/flx mice were first studied in their normal housing conditions and in a plethysmography chamber. Figure 1 illustrates the experimental protocol used.
Biotelemetry recordings of the systemic arterial blood pressure
Blood pressure (BP) telemetry probe implantation and recording techniques have been described in detail previously (Aziz et al. 2014). Briefly, the left carotid artery was cannulated with a gel‐filled catheter of the BP telemetry probe (model PA‐C10; Data Sciences International, St Paul, MN) which allows beat‐to‐beat biotelemetry recording of the arterial BP in conscious freely moving mouse. Heart rate (HR) was derived from the BP recording trace.
Recording of the 24‐h BP and HR profile
Animals were allowed to recover for at least 10 days after the transmitter implantation. Experiments were performed in a temperature‐controlled environment (21–24°C) with 12 h/12 h light/dark cycle. The animals were allowed to acclimatize to the experimental conditions during the recovery period from surgery. For serial BP and HR measurements, 15‐second‐long samples were taken every 30 min over 24 h using a “scheduled sampling” protocol. From the recorded dataset, it was possible to determine the diurnal changes in BP and HR during the day and night and also calculate the mean values.
Whole body plethysmography
Mice with telemetry probes implanted were also exposed to hypoxic and hypercapnic conditions. Respiratory rate (fR, breaths min−1), tidal volume (VT, arbitrary unit g−1), and minute ventilation (VE) ((fR × VT); arbitrary unit min−1 g−1) were measured using whole body plethysmography as described in detail previously (Onodera et al. 1997; Rong et al. 2003; Trapp et al. 2011) and BP and derived HR simultaneously recorded using radiotelemetry. Briefly, the mouse was placed in a plexiglass recording chamber (~200 mL) that is in turn placed on a telemetry recording mat. The chamber was flushed continuously with a mixture of 79% nitrogen and 21% oxygen (unless otherwise required by protocol) at a rate of ~1 L min−1. Concentrations of O2 and CO2 in the chamber were monitored online using a fast‐response O2/CO2 monitor (Morgan Scientific, Haverhill, MA).
The animals were allowed at least 20 min to acclimatize to the chamber environment at normoxia/normocapnia (21% O2, 79% N2, and <0.3% CO2) before measurements of baseline ventilation were taken. Hypoxia was induced by lowering the O2 concentration in the inspired air down to 10% for 5 min. After 5 min, the O2 concentration was then brought back up to 21% for a further 5 min. In a separate experiment, normoxic hypercapnia was induced by titrating CO2 into the respiratory mixture up to a level of 3% or 6% (lowering N2 accordingly) for 5 min at each CO2 level. The measurements were taken during the last 2 min before exposure to the stimulus and during the 2‐min period at the end of each stimulus, when breathing has stabilized. Simultaneous continuous recordings of changes in BP, HR, fR and VT during hypoxic and hypercapnic challenges were obtained.
In vivo experiments under general anesthesia
A separate cohort of GαO flx/flx mice were studied in the anesthetized state after Cre/GFP AVV or control GFP AVV injections into the RVLM. Mice were anesthetized with urethane (1.3 g kg−1 ip). The depth of anesthesia was monitored using the stability of BP, HR, and lack of flexor responses to a paw pinch and supplemental anesthesia was given as required. Body temperature was maintained at 37.0 ± 0.2°C using a servo‐controlled heating pad. Tracheostomy was performed to facilitate ventilation in spontaneously breathing mice. Needle electrodes were placed in a lead II configuration to record surface ECG (sampled at 2 kHz, amplified × 50–100, and filtered to a bandwidth between 5 and 100 Hz with 50 Hz notch filtering).
Determining baroreflex sensitivity
The right and left jugular veins were cannulated with polyethylene tubing (PE‐10) for administration of phenylephrine (0.5–2.5 mg kg−1 iv in 2–10 μL saline) or sodium nitroprusside (SNP, 0.1–1 mg kg−1 iv in 1–10 μL saline), respectively. The left carotid artery was cannulated with saline‐filled polyethylene tubing (mechanically stretched PE‐10 connected to PE‐50) connected to a pressure transducer to measure arterial BP (sampled at 2 kHz). Changes in the systemic arterial BP were induced by alternately administrating phenylephrine and SNP and varying the volumes of injectate. The baroreflex sensitivity curve was then derived as described by Head and McCarty (2002). The peak changes in arterial BP and HR were determined following the administration of each drug and the changes from basal values were then calculated for each pair of data points and scaled back to absolute BP and HR by the average of all basal values recorded. About 10–15 data points were obtained across the range of BP‐HR for each animal and the data points were then fitted to the following logistic equation using a least squares iterative routine on MatLab (R2010b, MathWorks, Natick, MA). where P1 = lower HR plateau, P2 = HR range, P3 = a curvature coefficient which is independent of range, and P4 = BP50, that is, the BP at half the HR range. The average baroreflex gain (BRG) or slope of the curve between the two inflection points is given by BRG = −P2 × P3/4.56 and the upper plateau = P1 + P2.
A separate group of mice underwent programmed electrical assessment of ventricular excitability. An octapolar 1.1F cardiac electrophysiology catheter was inserted into the right ventricular apex through the right internal jugular vein to pace the heart as previously described (Zuberi et al. 2010; Machhada et al. 2015). Cardiac pacing using extrastimulation protocols was performed with an S88‐Grass stimulator to record the ventricular effective refractory period at a basic drive train of 750 beats per minute (VERP750).
Data are reported as means ± standard error of the mean (SEM) unless stated otherwise. For determination of statistical significance between groups, two‐tailed Student's t test was used for parametric data with a normal distribution and the Mann–Whitney U Test was used as a nonparametric test. Fisher's exact test was used for comparing categorical data. For data with repeated sampling between multiple groups, mixed‐level modeling using both fixed and random effects was used to compare the mean effect of each group. Analysis of variance with Bonferroni multiple comparison test was used to compare the difference of effect between groups. In all the instances, P < 0.05 was considered significant.
Generation of mice with conditional deletion of GαO deletion in the RVLM and controls
Using 1:100 dilutions of Cre/GFP AVV (6 × 1010 PFU) and GFP AVV (3 × 1010 PFU), GαO expression at the sites of injections was reduced by 79% in mice post‐Cre/GFP‐AVV injections (relative expression 0.21, 95% CI 0.18–0.25) compared to 21% post‐GFP‐AVV injections (relative expression 0.79, 95% CI 0.75–0.82), when normalized to the expression in mice preinjections (n = 3 mice in triplicate for both groups). Figure 2 shows the representative brainstem distribution of GFP‐expressing cells relative to TH immunoreactivity.
Diurnal profile of HR and BP changes in conditions of GαO deletion in the RVLM
We first examined the 24‐h HR and BP profile of GαO flx/flx mice pre‐ and post‐Cre/GFP AVV (n = 6) and control GFP AVV (n = 6) injections under normal housing conditions. There was a clear circadian variation in HR and systolic BP in GαO flx/flx mice which was unchanged following injections of Cre/GFP AVV and control GFP AVV (Fig. 3).
Hemodynamic and respiratory responses to hypoxia and hypercapnia in conditions of GαO deletion in the RVLM
GαO flx/flx mice were then studied in the plethysmography chamber. GαO flx/flx mice responded to hypoxia (10% O2) with an increase in HR, respiratory rate, tidal volume, and minute ventilation, but there was no significant change in the systolic BP. There was posthypoxic depression of respiratory activity with reduced respiratory rate below baseline and normalization of HR and tidal volume. In conditions of GαO deletion in the RVLM (post‐Cre/GFP AVV injections in GαO flx/flx mice), the animals were more tachycardic at baseline (579 ± 10 vs. 504 ± 26 bpm, P = 0.023, mixed‐level modeling), during the hypoxic challenge (721 ± 7 vs. 656 ± 10 bpm, P < 0.001) and during reoxygenation (633 ± 14 vs. 560 ± 26 bpm, P = 0.049) compared to the preinjection state. There was no difference in the change in HR during hypoxic challenge or recovery comparing pre‐ and post‐Cre/GFP AVV injections. GαO flx/flx mice were also more hypertensive at baseline after the injections of Cre/GFP (152 ± 7 vs. 128 ± 5 mmHg, P = 0.025). There was a significant decrease in systolic BP on exposure to hypoxia (ΔSBP −22 mmHg, 95% CI −42 to −2 mmHg, P = 0.034, Bonferroni) which was maintained during recovery post‐Cre/GFP AVV injection. As a result, systolic BP during (132 ± 3 vs. 129 ± 5 mmHg, P = 0.671) and after hypoxia (124 ± 6 vs. 126 ± 6 mmHg, P = 0.796) were similar in pre‐ and post‐Cre/GFP AVV injections. Hypoxic ventilatory response was unaffected in conditions of GαO deletion in the RVLM (Figs. 4 and 6A, Table 1).
GαO flx/flx mice responded to 6% CO2 with a significant decrease in the HR (ΔHR‐88 bpm, 95% CI −160 to −16 bpm, P = 0.018), accompanied by an increase in the arterial BP (ΔSBP 23 mmHg, 95% CI 5–41 mmHg, P = 0.013). Hypercapnia also evoked stereotypic increases in minute ventilation. GαO flx/flx mice displayed higher HR and systolic BP at baseline after acclimatization to the chamber environment post‐Cre/GFP AVV injections (654 ± 16 vs. 587 ± 23 bpm, P = 0.012; 159 ± 4 vs. 136 ± 6 mmHg, P = 0.012). HR depression during hypercapnia was attenuated post‐Cre/GFP AVV injections (ΔHR −21 bpm, 95% CI −93 to 51 bpm, P = 0.919), resulting in a consistently elevated HR throughout (3% CO2: 616 ± 23 vs. 539 ± 28 bpm, P = 0.049; 6% CO2: 633 ± 19 vs. 499 ± 27 bpm, P = 0.003). The elevation of systolic BP on exposure to 6% CO2 was also attenuated post‐Cre/GFP AVV injections (ΔSBP 11 95% CI −7 to 29 mmHg, P = 0.248) resulting in a higher systolic BP compared to preinjection (170 ± 10 vs. 159 ± 4 mmHg, P = 0.669). In summary, there is a significant difference in HR and systolic BP of mice at 6% CO2 compared to baseline preinjection of Cre/GFP AVV which is lost with GαO deletion postinjection (Figs. 5 and 6B, Table 2).
Hemodynamic profile and baroreflex sensitivity in conditions of GαO deletion in the RVLM
GαO flx/flx mice were next studied under urethane anesthesia following injections of Cre/GFP AVV (n = 5) and GFP AVV (n = 5). In these conditions, GαO flx/flx mice had significantly higher HR post‐Cre/GFP AVV injections compared to GFP AVV injections (629 ± 3 vs. 598 ± 5 bpm, P = 0.002, t test) with no difference in systolic BP (111 ± 2 vs. 115 ± 4 mmHg, P = 0.074). This was associated with an increase in baroreflex gain (0.65 ± 0.07 vs. 0.35 ± 0.01 bpm mmHg−1, P = 0.012) (Fig. 7; Table 3).
Ventricular excitability in conditions of GαO deletion in the RVLM
Cre/GFP AVV (n = 9)‐ and GFP AVV (n = 7)‐injected mice then underwent programmed electrical stimulation. The difference in resting HR was abolished following administration of atenolol with no effect on the HR with additional administration of atropine. Cre/GFP AVV‐injected mice had significantly reduced VERP750 compared to controls (VERP750 43 ± 3 vs. 55 ± 4 ms, P = 0.033, Mann–Whitney U) and this difference in VERP750 was abolished by beta‐adrenoceptor blockade with atenolol (Fig. 8).
GαO is the major inhibitory heterotrimeric G protein expressed in the brain and mediating inhibitory signaling by several neurotransmitters. In this study, we investigated whether GαO‐mediated signaling is important in the central nervous mechanisms controlling cardiovascular activities focusing on the RVLM presympathetic circuits. Our results show that mice with deletion of GαO in the presympathetic circuits of the brainstem (RVLM) display normal HR and arterial BP profile under resting conditions. However, when stressed in a novel environment or studied under general (urethane) anesthesia, these mice show an elevated HR and abnormal hemodynamic responses during systemic hypoxia and hypercapnia as well as increased baroreflex sensitivity. Pharmacological autonomic blockade using atenolol and atropine suggested that this phenotype is largely due to an increase in sympathetic tone. Finally, this increase in sympathetic tone to the heart was found to be associated with alterations in ventricular electrophysiology as evidenced from reduced VERP.
Deletion of GαO proteins in the RVLM does not affect diurnal HR or BP under resting conditions
There were no significant differences in HR and BP profile after GαO deletion in the presympathetic area of the brainstem in conscious freely moving mice housed in standard conditions. This observation is in agreement with the findings obtained in rats where permanent experimental perturbations within the C1 are likely to be compensated as previously reported (Schreihofer et al. 2000; Madden and Sved 2003). These data suggest that impairment of RVLM function in controlling HR and BP can be fully compensated for in the awake and unstressed experimental animals.
Altered hemodynamic responses to hypoxia and hypercapnia in mice with GαO deletion in the RVLM
An increasing body of evidence is accumulating on the importance of the presympathetic CNS circuits, and more specifically the RVLM, in mediating body responses to a variety of physiological and pathological stressors (Guyenet et al. 2013). Our data add to this body of evidence of the key role played by the RVLM mechanisms under the conditions of environmental stress including hypoxia and hypercapnia. In addition, it would appear that inhibitory signaling in the brainstem mediated by GαO proteins is an important mechanism restraining sympathetic activity. For example, the RVLM is activated by increased dorsomedial hypothalamic activity in response to psychological stressors (Nalivaiko and Blessing 2001; Madden and Morrison 2004). This increased sympathetic activity is reversibly inhibited by selective serotonin‐1A (5‐HT1A) receptor agonist 8‐OH‐DPAT (Nalivaiko et al. 2005; Horiuchi et al. 2008; Ngampramuan et al. 2008). 5‐HT1A receptors are one of the many GPCRs in the brainstem that signal via inhibitory Gαi/o. In addition, GABA and enkephalins can tonically inhibit the RVLM potentially through GPCR pathways (Bowman et al. 2013). GABA, 5‐HT, and adenosine are inhibitory neurotransmitters that have also all been shown to be released in the brainstem during hypoxia (Richter et al. 1999). Loss of inhibitory signaling mediated by GαO could potentially explain the augmented HR response during the hypoxic challenge. The corelease of GABA with glutamate with similar temporal kinetics (Richter et al. 1999) suggests a role for GABA as a likely mediator of the restraining influence to the excitatory effects of glutamate during hypoxia. Indeed, there is evidence that GABAergic inhibitory mechanisms are important in modulating RVLM activity during hypoxia in rat neonatal brainstem preparations (Boychuk et al. 2012).
Increased sympathetic tone and its effect on ventricular excitability
Mice studied under urethane anesthesia have previously been shown to have reduced arterial pressure and increased HR compared to the conscious state (Desai et al. 1997; Paton and Butcher 1998). Similar findings were obtained here with mice being hypotensive and generally more tachycardic. In general, anesthesia results in a reduction of vagal tone, but urethane seems to lead to relative sparing compared to other anesthetics (Machhada et al. 2015). However, the RVLM does contain vagal preganglionic motoneurons (Izzo et al. 1993). Sequential challenges with atenolol and atropine revealed low vagal tone and demonstrated that sympathetic activity indeed dominates HR control in mice under urethane anesthesia. This suggests that increased sympathetic outflow (rather than vagal withdrawal) is responsible for the increased HR and enhanced baroreflex sensitivity seen in mice with deletion of GαO in the RVLM.
It is well known that the autonomic nervous system modulates cardiac excitability. Increase in sympathetic tone to the heart has been shown to reduce action potential duration with a corresponding increase in conduction velocity, resulting in a change in the restitution properties of action potential duration and dispersion of ventricular repolarization and increased susceptibility to ventricular arrhythmia (Ng et al. 2009; Finlay et al. 2016). Our data add to this body of evidence and demonstrates how alterations in central nervous mechanisms controlling cardiorespiratory activity can alter ventricular excitability.
G protein specificity in vivo
Parker et al. (2012) has previously shown that Gαi2 and GαO are the most ubiquitous inhibitory G proteins in the RVLM. In keeping with this, our data suggest an important functional role for GαO. We cannot exclude a role for Gαi2 as this was not studied directly with a conditional KO of Gαi2 in the brainstem. Our own previous studies showed an effect on cardiac function in global Gαi2 KO mice, though we have ascribed this largely to direct effects on cardiac tissue (Zuberi et al. 2008, 2010). Other investigators have shown a potential role for Gαi2 in the rat in mediating hypertensive responses (Wainford et al. 2015; Carmichael et al. 2016).
The A1 and A5 cell groups and the vagal preganglionic neurons lie in close proximity to the RVLM in the ventral medullary brainstem region (Strack et al. 1989). Although great care was taken to target viral delivery to the RVLM area, given the nonspecific nature of the CMV promoter, it is likely that some of the neurons in the neighboring regions would have undergone Cre‐mediated deletion of GαO. However, given that the subsequent histological analysis demonstrated expression of GFP to be distributed among TH‐immunoreactive neurons in the RVLM region, the observed phenotype is likely to be mainly due to the loss of GαO‐mediated signaling within the RVLM presympathetic circuits.
The relative contributions of the sympathetic and parasympathetic limbs of the autonomic nervous system to the phenotype observed were assessed by pharmacological blockade. These are indirect measures of autonomic function. A more direct approach would involve direct recordings of efferent nerve activity such as of the renal sympathetic nerve (Ma et al. 2002) or other experimental paradigms such as in the working heart brainstem preparations (Paton 1996).
The precise mechanism by which loss of GαO at the RVLM results in alterations in autonomic tone remains unknown. We suggest that GαO is important in mediating inhibitory influences on the RVLM with loss of this resulting in disinhibition and increased sympathetic outflow. GαO‐coupled receptors can cause pre‐ or postsynaptic inhibition either by inhibiting calcium channels or activating G‐protein‐gated K+ channels, respectively (Dolphin 1998; Luscher and Slesinger 2010). The specific neurotransmitter(s) and exact mechanism is a topic for further work in the follow‐up studies.
In summary, the data obtained in this study suggest that GαO‐mediated signaling within the presympathetic circuits of the RVLM contributes to the autonomic control of the heart. GαO deficiency in the RVLM is associated with abnormal cardiovascular responses to stress, altered cardiovascular reflexes, and ventricular excitability.
Conflicts of Interest
The authors have no conflicts of interest to declare.
AVG and AT are joint last authors.
This research was supported by the Medical Research Council (MRC Clinical Research Training Fellowship to RA), British Heart Foundation (Ref: RG/14/4/30736), Wellcome Trust (Wellcome Trust Senior Research Fellowship to AVG; Ref: 095064), and by the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences (Project Z01‐ES‐101643 to LB). This work was facilitated by the National Institute for Health Research Barts Cardiovascular Biomedical Research Unit.
- Manuscript Received: May 12, 2016.
- Manuscript Revised: June 9, 2016.
- Manuscript Accepted: June 20, 2016.
- © 2016 The Authors. Physiological Reports published by Wiley Periodicals, Inc. on behalf of the American Physiological Society and The Physiological Society.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.