The present investigation has shown that a non-sweet carbohydrate in the human mouth produces a similar central neural response to that obtained with glucose, suggesting there may be a class of so far unidentified oral receptors that respond to the caloric property of carbohydrate independently of those for sweetness. Furthermore, there has been speculation that the brain responses to glucose in the mouth may mediate emotional and behavioural responses associated with rewarding stimuli and the present results demonstrate, for the first time, evidence of such a link with the improvements in exercise performance that were obtained with both glucose and maltodextrin in the mouth. For convenience the discussion will concentrate first on the exercise studies (Studies 1A and 2A) and then the brain imaging data (Studies 1B and 2B) rather than following the sequence in which they are presented in the Results section.
The effect of oral carbohydrate on exercise performance
A 1 h time trial is a demanding form of exercise during which power output typically declines steadily before a final sprint to the finish. Success in this event depends on minimising this loss of power and several studies have reported that oral carbohydrate feeding enhances performance compared to water or an artificially sweetened placebo (Anantaraman et al. 1995; Below et al. 1995; Jeukendrup et al. 1997). However, there is no clear metabolic explanation for the ergogenic action of the carbohydrate and several groups have speculated about a ‘non-metabolic’ (McConell et al. 2000) or ‘central’ (Jeukendrup et al. 1997) action. Carter et al. (2004a) first produced evidence of such a central effect and proposed a mechanism involving the activation of higher brain centres by carbohydrate-sensitive receptors in the oral cavity. The two exercise studies (Studies 1A and 2A) have shown that repeated exposure of the oral cavity to a carbohydrate solution, containing either glucose or maltodextrin, improves performance of a simulated 1 h cycle time trial. The 2% reduction in time to complete the set work load and corresponding increase in mean power output in Study 1A, as a result of swilling the glucose solution, and the 3.1% increase in Study 2A, with the maltodextrin solution, were virtually the same as the 2.9% improvement reported by Carter et al. (2004a) using non-sweet maltodextrin. Carter and colleagues used a plain water placebo and although the maltodextrin solution was non-sweet, some of their subjects reported a slight difference in texture of the two solutions. In the present studies care was taken to disguise the solutions with a strong artificial sweetener and while, post exercise in Study 1A, some subjects rated the glucose solution to be more viscous, they did not have any reason to associate this with the presence of glucose in the solution nor were they aware of the hypothesis behind the study. We conclude, therefore, that the results reported here confirm the suggestion of Carter et al. (2004a) that carbohydrate in the oral cavity improves performance during simulated 1 h cycle time trials.
A very similar pattern of changing power output, or pacing strategy, was observed in the glucose and maltodextrin trials when compared to placebo (Fig. 3), with power declining during most of the exercise before increasing in the final phase as subjects completed a ‘sprint finish’. The trend in both the glucose and maltodextrin trials was for power to be better maintained compared with placebo, particularly in the latter stages of exercise, allowing subjects to complete the set workload in a significantly shorter time. Despite the higher power generated there were no differences in RPE in either glucose solution or maltodextrin trials compared to placebo, exactly as found by Carter et al. (2004a). It has been noted that subjects tend to alter their power output during self-paced exercise so that their RPE remains relatively constant (Cole et al. 1996). The fact that the subjects in the glucose and maltodextrin trials were working at a higher power yet reporting the same RPE as in the placebo trials, suggests that oral exposure to carbohydrate evokes a central response that enables subjects to increase their power output by reducing the perception of a given workload.
A recent investigation failed to find a benefit from rinsing the mouth with a carbohydrate solution during a running time trial (Whitham & McKinney, 2007). As acknowledged by the authors, the lack of performance improvement in those studies may have been due to the study design as the participants had to make a conscious decision to alter the pace of the motorised treadmill and, as such, running speed was far more consistent throughout the performance test compared with the variable power outputs typically observed during a cycling time trial. An exercise performance test that requires conscious alteration of power output may therefore lack the sensitivity to observe the proposed unconscious central effect of a carbohydrate mouth rinse.
In the present investigation the glucose, maltodextrin and placebo solutions were artificially sweetened, demonstrating that the observed enhancement in exercise performance was independent of sweetness. This is consistent with previous studies which have reported an improvement in exercise performance with a carbohydrate solution when both the carbohydrate and placebo beverages were matched for sweetness, flavour and colour (Below et al. 1995; Jeukendrup et al. 1997). Furthermore, it has been shown that sweetness per se is not an important factor for carbohydrate supplementation to improve exercise performance in a hyperthermic environment (Carter et al. 2005).
The current studies and those of Carter et al. (2004a) that have reported a positive effect of a carbohydrate mouth rinse on exercise performance used a design in which subjects began exercise after a prolonged fast (> 6 h fast). For that reason, in both the exercise and fMRI studies we purposefully observed the brain responses to oral glucose, maltodextrin and saccharin with participants in a similar fasted state. Different pre-exercise nutritional practices could explain the discrepancy in some of the reports concerning carbohydrate ingestion during high-intensity exercise. The studies showing a performance enhancement have all entailed subjects commencing exercise following an overnight fast (Neufer et al. 1987; Below et al. 1995; Millard-Stafford et al. 1997) or in a post-absorptive state (> 4 h) (Anantaraman et al. 1995). Conversely, a common feature of investigations that fail to report an ergogenic action from carbohydrate feeding during high-intensity exercise is that subjects receive a meal designed to ‘top-up’ endogenous carbohydrate stores, ∼2 h prior to exercise (Desbrow et al. 2004; Burke et al. 2005). Similarly, the investigation of Whitham & McKinney (2007), which reported no improvement with a carbohydrate mouth rinse, had subjects commence exercise following a shorter period of fasting (4 h) than the present studies. Pre-exercise feeding may influence the brain responses to an oral carbohydrate stimulus during subsequent exercise as it is likely that the activation of brain regions associated with feeding and reward, such as the orbitofrontal cortex and striatum, are modulated by homeostatic regulation and the current physiological state of the body (Small et al. 2001). An interesting series of future studies would be to compare both the effect on exercise performance and brain responses of an oral carbohydrate stimulus in the fed and fasted states.
The effect of oral carbohydrate on brain responses
Studies 1B and 2B both used glucose as one of the tastants and while the brain responses were broadly similar, there were some differences. In Study 2B, glucose activated the orbitofrontal cortex and the adjoining rostral part of the anterior cingulate cortex, which was not seen in Study 1B. The difference may have been due to the proximity of the orbitofrontal cortex to the air-filled sinuses leading to signal dropout and artefacts (Wilson et al. 2002) and consequently negative findings from the orbitofrontal cortex should be treated with caution (Kringelbach, 2005). The use of a specific set of imaging parameters to minimize distortion artefacts in the orbitofrontal cortex, as used in other investigations (de Araujo et al. 2003a; de Araujo & Rolls, 2004; Frank et al. 2008), might have improved the consistency of data obtained in this area of the brain.
Previous functional neuroimaging studies of the anterior insula/frontal operculum, believed to be the human primary taste cortex, have shown it to be sensitive to various oral stimuli, including glucose, salt (O’Doherty et al. 2001) and umami (de Araujo et al. 2003a). The present study revealed activation in this region in response to both glucose and saccharin solutions (Fig. 5A). Comparable activation in response to glucose and saccharin was also found in the dorsolateral prefrontal cortex (Fig. 5B) which is suggested to have a role in the preparation and selection of cognitive responses (Rowe et al. 2000) and has previously been shown to be sensitive to a range of different taste stimuli (Kringelbach et al. 2004). Activation within this region is believed to reflect an engagement in cognitive and attentional processing induced by taste input. However, despite the inability of the subjects to distinguish between the glucose and saccharin solutions on a range of subjective measures (sweetness, pleasantness and viscosity), glucose activated a number of brain regions that were unresponsive to saccharin. These included the anterior cingulate cortex and the right caudate, that forms part of the striatum. These brain regions, in particular the dopaminergic pathways within the striatum, are believed to mediate the emotional and behavioural response to rewarding food stimuli (Berridge & Robinson, 1998; Kelley et al. 2002; Rolls, 2007). These observations are very similar to those of Frank et al. (2008) who found that, compared to a similar sweet-tasting sucralose solution, only sucrose activated the cingulate cortex, parts of the striatum and ventral tegmental area. The current study therefore supports the idea that it is not sweetness that is required for the activation of particular reward-related regions of the brain but rather some other property of natural sugars, possibly the caloric content.
The importance of the presumed caloric-content rather than the sweetness of natural carbohydrates is underlined by the similar cortical responses produced by oral glucose and maltodextrin (Figs 7 and 8), despite the obvious differences in perceived sweetness between the two solutions. Regions of common activation were found in the primary taste cortex (Fig. 7A) and a medial region of orbitofrontal cortex (Fig. 7B), which is the putative secondary taste cortex (Rolls, 2007). Previous functional neuroimaging studies have revealed activation in the orbitofrontal cortex from a variety of taste stimuli (O’Doherty et al. 2001; Small et al. 2001; de Araujo et al. 2003a; de Araujo & Rolls, 2004) and this is where the current hedonic value of an oral stimulus is thought to be represented since activation of this region can be suppressed by satiety (Small et al. 2001). The conjunction overlay mask also revealed clusters of common activation in response to glucose and maltodextrin in the dorsolateral prefrontal cortex (Fig. 8A) and a small area of the right caudate (Fig. 8B). Thus, despite not being rated as ‘pleasant’ as the glucose tastant, the complex carbohydrate solution still produced activation within a region of the ventral striatum, a crucial interface for many well-established motivational circuits in the brain (Kelley et al. 2002).
The orbitofrontal cortex is an important area of convergence for somatosensory inputs produced by the texture of food in the mouth (Rolls, 2007). Single-neuron recordings in the primate orbitofrontal cortex have revealed a population that responds to the oral texture of carboxymethylcellulose, a tasteless thickening agent used in the food industry (Verhagen et al. 2003). These oral texture responses have since been extended to the human brain with de Araujo & Rolls (2004) reporting activation in a lateral region of orbitofrontal cortex from both an oral fat (vegetable oil) stimulus and carboxymethylcellulose solution with a similar viscosity. Consequently, the central activation observed in Study 2B, particularly from the maltodextrin solution, might have been due to activation of oral somatosensory receptors (Simon et al. 2006) rather than specific taste receptors. There are, however, two reasons for thinking this is not the case. The first is the observation of de Araujo & Rolls (2004) that a rostral part of anterior cingulate cortex, where it borders the medial orbitofrontal cortex and the ventral striatum, was activated by oral fat independently of its viscosity. This led the authors to propose that activation in these brain regions may indicate the energy content of a food. The fact that maltodextrin in the present study produced a very similar response in this part of the brain would suggest that the complex carbohydrate solution was providing more than a simple somatosensory stimulus. A second reason for doubting that activation from the maltodextrin solution was a consequence of its viscosity is that although subjects were able to detect a difference, this solution would not normally be described as ‘viscous’. de Araujo & Rolls (2004) comment that the 18% sucrose solution they used had a viscosity of 2 centipoise (cP) compared with 50 cP for a carboxymethylcellulose solution which provided activation comparable to that of their sucrose solution in the primary taste cortex. The maltodextrin solution we used (18%) had a similar low viscosity of ∼2 cP and consequently is unlikely to have evoked a somatosensory response that was any greater than that provided by the control solution with a viscosity of 1–2 cP.
These observations raise the question of how the maltodextrin solution is sensed in the oral cavity. From the manufacturer's specifications (see Methods) and our measurements of free glucose, the maltodextrin solution would have contained ∼1.6% mono- and disaccharides, about 50 times less than the concentration of the comparison glucose solution. We are not aware of any dose–response studies regarding central activation by oral glucose, but it seems unlikely that the sweet mono- and disaccharides in the maltodextrin solution at this low concentration would generate the same activation as the pure glucose solution, especially as the maltodextrin solution was not perceived as ‘sweet’ by the subjects.
The mammalian sweet taste receptor combines two G-protein-coupled receptors, T1R2 and T1R3, which respond to both natural sugars and artificial sweeteners (Nelson et al. 2001). These taste receptor cells found primarily on the tongue, are innervated by afferent fibres that transmit information to taste regions in the cortex via the thalamus (Simon et al, 2006). Recent work using transgenic mice that lack the T1R3 protein suggests that natural caloric sugars activate taste afferents differently from non-caloric artificial sweeteners (Damak et al. 2003; Zhao et al. 2003). T1R3-knock-out (KO) mice showed no behavioural attraction to artificial sweeteners yet there was only a modest reduction in preference to caloric sugars (Damak et al. 2003) and T1R3-KO mice still had a detectable gustatory nerve response to natural sugar. More recently, Delay et al. (2006) reported that the detection threshold for sucrose was indistinguishable between T1R3-KO and wild-type mice. These results indicate that there are T1R3-independent taste receptors for natural carbohydrates in mice. The fact that in our experiments maltodextrin activated very similar brain areas compared to glucose and was not perceived as sweet suggests that there may also be human T1R3-independent taste receptors which have subtly different projections in the brain compared to the taste receptor cells that co-express T1R2 and T1R3 and convey sweetness.
The concentration of the glucose and maltodextrin solutions used in the fMRI experiments were nearly 3 times greater than those used in the exercise studies. This raises the obvious question of whether the lower concentrations used during the time trials would also activate the brain regions we have identified. The higher concentrations used in the fMRI studies were chosen to replicate and allow comparisons with earlier work on glucose tasting (e.g. O’Doherty et al. 2001). The intention was to determine whether it was possible that glucose and maltodextrin activate similar areas of the brain. This is clearly the case and lends strong support to the hypothesis that the oral carbohydrate in the exercise studies was acting via central neural pathways. However, definitive proof would require fMRI studies with lower carbohydrate concentrations. The current studies here, therefore demonstrate that the presence of either glucose or maltodextrin in the oral cavity can improve performance of a self-paced exercise task of ∼1 h duration while there is, at least, the potential for both carbohydrates to activate brain regions believed to mediate emotional and behavioural responses to a rewarding sensory stimulus (Kringelbach, 2004). For example, the dopaminergic system of the ventral striatum has been implicated in arousal, motivation and the control of motor behaviour (Berridge & Robinson, 1998). Prolonged exercise, such as a cycle time trial, generates a great deal of afferent information arising from muscles, joints, lungs, skin and core temperature receptors which may, over time, be perceived as unpleasant and consciously, or unconsciously, lead to an inhibition of motor output manifesting as ‘central’ fatigue. Individuals tend to regulate their physical activity to keep their levels of discomfort within acceptable limits and this has become known as the ‘Central Governor Model’ (Noakes, 2000; St Clair Gibson et al. 2001; Lambert et al. 2005). It is not clear which brain pathways are involved in this inhibitory activity but one possibility is a decrease in activity of dopaminergic pathways affecting either reward or the motor functions of the basal ganglia. Conversely, increased activity of these pathways might counteract the effects of fatigue and we suggest that this is the mode of action of oral carbohydrate. Central stimulants such as caffeine and amphetamines are well known to enhance performance during prolonged exercise (Gerald, 1978; Chandler & Blair, 1980) and administration of methamphetamine stimulates activity of reward circuitry in the human brain, including the medial orbitofrontal cortex and ventral striatum (Völlm et al. 2004), in a similar way to the oral carbohydrates used in the present investigation.
Probably the major limitation of the work presented here is that in the fMRI studies the subjects were all tested in the rested condition and it is possible that some of the consequences of exercise, such as hyperthermia, may alter the brain responses to oral carbohydrate. Our present speculations about mechanisms would therefore be strengthened if the response to a carbohydrate solution in reward-related brain regions was observed when individuals perform an intense exercise task. The possible interaction between this affective response and activation of brain regions that modify central motor drive during exercise could also be determined. Unfortunately it would be difficult to replicate the intense nature of a cycle time trial inside an fMRI scanner, as dynamic whole-body exercise is likely to cause substantial head movement that would create distortion artefacts in the BOLD signal. A further limitation of the present work is that the fMRI studies only observed the central neural responses to tasting saccharin while in the exercise studies the masking sweet taste was provided by a mixture of saccharin and aspartame. The only way in which this might invalidate our results would be if aspartame activated similar brain areas to the caloric carbohydrates. This seems unlikely and it is notable that the only other non-nutritive sweetener, sucralose, that has been tested in this way showed patterns of activation that were distinct from those of a natural carbohydrate (Frank et al. 2008).
In summary, we have shown that both sweet and non-sweet carbohydrate in the human mouth activate a variety of brain areas, some of which may be involved in reward and the regulation of motor activity. We suggest that activation of these regions of the brain may provide a mechanism to explain the improvement in exercise performance that is observed when carbohydrate is present in the mouth. The findings also support the existence of oral receptors sensitive to the caloric value of carbohydrate and which are independent of sweetness.