GLP-1 release and vagal afferent activation mediate the beneficial metabolic and chronotherapeutic effects of D-allulose

Overeating and arrhythmic feeding promote obesity and diabetes. Glucagon-like peptide-1 receptor (GLP-1R) agonists are effective anti-obesity drugs but their use is limited by side effects. Here we show that oral administration of the non-calorie sweetener, rare sugar d-allulose (d-psicose), induces GLP-1 release, activates vagal afferent signaling, reduces food intake and promotes glucose tolerance in healthy and obese-diabetic animal models. Subchronic d-allulose administered at the light period (LP) onset ameliorates LP-specific hyperphagia, visceral obesity, and glucose intolerance. These effects are blunted by vagotomy or pharmacological GLP-1R blockade, and by genetic inactivation of GLP-1R signaling in whole body or selectively in vagal afferents. Our results identify d-allulose as prominent GLP-1 releaser that acts via vagal afferents to restrict feeding and hyperglycemia. Furthermore, when administered in a time-specific manner, chronic d-allulose corrects arrhythmic overeating, obesity and diabetes, suggesting that chronotherapeutic modulation of vagal afferent GLP-1R signaling may aid in treating metabolic disorders.

T he obesity pandemic is a serious worldwide health problem: 39% of adults (1.9 billion) are overweight/obesity 1 , and these numbers are increasing 2 . Obesity is a major risk factor for type 2 diabetes, cardiovascular, and cerebral diseases. Overeating and abnormal feeding rhythm contribute to development of obesity 3,4 . However, few effective and safe medicines and/or food ingredients (supplements) are available to treat obesity and feeding disorders. One exception is the class of glucagon-like peptide-1 receptor (GLP-1R) agonists, which can be used to treat not only type 2 diabetes but obesity, hyperphagia, and hypertension [5][6][7] . Recent randomized double-blind trials showed that treatment with GLP-1R agonists reduces cardiovascular events, stroke, and nephropathy 8,9 . However, GLP-1R agonists elicit nausea and vomiting 6,10 , and increase mean heart rate via directly stimulating sinoatrial node and enhancing sympathetic nervous system tone [10][11][12] . The mechanisms underlying these adverse effects of GLP-1R agonists remain to be elucidated. However, their long life-span, penetration through the blood-brain barrier (BBB) 13 , and high therapeutic doses may allow them to directly act on central nervous system GLP-1Rs 12 , thereby causing CNS-dependent side effects. Strategies to enhance the tolerability of GLP-1-based therapy might include identification of agents that stimulate release of endogenous GLP-1 from intestinal L-cells, potentially accompanied by fewer adverse effects.
D-Allulose (D-ribo-2-hexylose, previously named D-psicose), a C-3 epimer of D-fructose, is a rare sugar existing in small amounts in nature. D-Allulose contributes no calories but substantial sweetness equivalent to glucose and 70% of sucrose 14,15 , yielding a zero-calorie sweetener. In humans, large proportions of orally administered D-allulose are absorbed and subsequently excreted via the urine without being metabolized 14 . D-Allulose has been shown to ameliorate diabetes and obesity in animals [16][17][18] , however, underlying mechanisms are largely unclear. Of particular importance is whether D-allulose alters feeding, a behavior upstream of glucose and energy homeostasis.
The previous reports of multiple beneficial effects of orally administered D-allulose prompted us to infer that it might act through stimulating GLP-1 release and/or vagal afferent signaling, both of which transmit gut signals to remote organs including the brain. Notably, GLP-1 directly interacts with vagal afferent neurons 19 , and surgical or chemical denervation of vagal afferents and specific Glp1r knockdown in vagal afferent neurons attenuate anorexigenic, insulinotropic and glycemic effects of exogenous GLP-1 and dipeptidyl peptide-4 (DPP-4; an enzyme that degrades GLP-1) inhibitors [20][21][22][23][24][25] . Thus, although it is evident that vagal afferents play a key role in metabolic effects of GLP-1, it is not clear whether D-allulose exerts its metabolic actions through GLP-1 and vagal afferent signaling.
We examined the effects of D-allulose on feeding behavior including arrhythmic hyperphagia that is characteristically associated with obesity and diabetes 3,4 and on glucose metabolism by glucose tolerance test (GTT), insulin tolerance test (ITT), and pyruvate tolerance test (PTT), and explored underlying mechanisms involving GLP-1 release and vagal afferents using Glp1r knockout mice (Glp1r KO), surgical vagotomy and Glp1r knockdown specifically in vagal afferents using viral-mediated shRNA. We found that oral administration of D-allulose induces GLP-1 release, activates vagal afferents, reduces food intake, and promotes glucose tolerance via enhanced insulin secretion and action. Subchronic administration of D-allulose ameliorates arrhythmic hyperphagia, obesity, and glucose intolerance. These effects are blunted by vagotomy and by pharmacological or genetic inactivation of the GLP-1R in vagal afferents. Our results demonstrate that D-allulose stimulates GLP-1 release, and corrects arrhythmic overeating, obesity and diabetes via vagal afferent pathways.
To further explore the involvement of GLP-1R signaling in the anorexigenic effect of D-allulose, we used Glp1r knockout (Glp1r KO) mice 26 . Oral D-allulose at 1 and 3 g kg −1 failed to alter food intake at any time point from 0.5 to 6 h in Glp1r KO mice (Fig. 2f, g), while it inhibited food intake at all time points in control wildtype (WT) C57BL/6J mice (Fig. 2c, d). I.p. administration of oxytocin, which inhibits food intake via activation of vagal afferents 27 , exerted its anorexigenic effect in Glp1r KO and WT mice to similar extents (Fig. 2e, h). These results indicate that GLP-1R signaling is required for the anorexigenic effect of Dallulose.
The results that D-allulose moderately increased insulin release only at 15 min prompted us to explore possible additional mechanisms for the glycemic action of D-allulose. First, the effect of D-allulose on insulin action was investigated with an ITT. Oral D-allulose (1 g kg −1 ), injected 1 h prior to i.p. insulin (1 IU kg −1 ), enhanced the blood glucose lowering effect of insulin at 15-240 min, with statistically significant difference (vs. saline) at 120 min and later (Fig. 3h). The AUC of blood glucose levels during 0-240 min was significantly reduced in the presence of D-allulose (Fig. 3j). The potentiation of insulin action by D-allulose was blocked by treatment with 200 nmol kg −1 Ex(9-39) (Fig. 3i, j) and attenuated in Glp1r KO mice (Fig. 3k). The effect of D-allulose on glucose production was next investigated with a PTT, which indirectly assesses hepatic gluconeogenesis. Treatment with Dallulose (1 g kg −1 ) from −60 min robustly suppressed the rise in blood glucose level and its AUC during 0-120 min (Fig. 3l, n). The action of D-allulose to attenuate the blood glucose rises in PTT was counteracted by treatment with 200 nmol kg −1 Ex(9-39) (Fig. 3m, n) and absent in Glp1r KO mice (Fig. 3o). These results indicate that oral administration of D-allulose enhances insulin action and suppresses glucose production via the GLP-1R.  Fig. 3a), nor influenced glucose tolerance, insulin action, or glucose production (Supplementary Fig. 3b-d). In ex vivo experiments, administration of D-allulose had no effect on glucose (11.2 mM)-induced insulin secretion from isolated pancreatic islets under static incubation ( Supplementary Fig. 4), indicating lack of direct interaction of D-allulose with islets. These results demonstrate that D-allulose improves glucose tolerance via potentiation of GLP-1R signaling.
D-Allulose reduces feeding via GLP-1R in HFD-fed mice. We next examined the effects of D-allulose on food intake in obese and diabetic mice. In high-fat diet (HFD)-fed obese mice, p.o. administration of D-allulose at 1 g kg −1 (Fig. 4a) and 3 g kg −1 (Fig. 4b) at the dark period (DP) onset (19:30) significantly suppressed cumulative HFD intake at 0.5-6 h. Notably, cumulative food intake for 24 h tended to be suppressed without rebound for the following 24-48 h period (Fig. 4a, b), which was accompanied by reduction in body weight gain at 24 h and even 48 h (Fig. 4a, b). These acute anorexigenic and weight-reducing effects of D-allulose were not observed in Glp1r KO mice-fed HFD for 5 weeks or longer (Fig. 4c, d). HFD-fed obese mice responded to p.o. D-allulose with a rise in portal GLP-1 concentration to a level approximately twofold higher compared to standard chow-fed mice (Fig. 4e vs. Fig. 1h). Accordingly, a threefold higher dose (600 nmol kg −1 ) of Ex(9-39) was used in HFD-fed mice and it counteracted the anorexigenic effects of D-allulose (Fig. 4f).
Furthermore, in db/db mice, a genetic obesity model due to a leptin receptor mutation, p.o. administration of D-allulose (1 g kg −1 ) suppressed food intake at 0.5-6 h in a similar manner to chow-fed and HFD-fed obese C57BL/6J mice (Fig. 4g vs. Figs. 1a and 4a). These data indicate that the actions of D-allulose to suppress food intake via GLP-1R signaling are preserved in multiple preclinical models of diabetes and obesity.
Allu improves glucose tolerance via GLP-1R in HFD-fed mice. In HFD-fed mice with elevated basal glucose levels (~200 mg dl −1 ), p.o. administration of D-allulose (1 g kg −1 ) at −60 min significantly lowered basal blood glucose level at 0 min, and suppressed rises in blood glucose level at 15-120 min and its AUC for 0-120 min period in ipGTT (Fig. 5a, b). D-Allulose at −60 min tended to decrease plasma insulin at 0 min (Fig. 5c) and increase plasma insulin at 15 min after i.p. glucose injection (Fig. 5c, d), suggesting that D-allulose might partially restore the insulin secretory response to glucose.
In HFD-fed mice fasted moderately for 4 h, administration of saline at −60 min did not alter blood glucose at 0 min (Fig. 5e), and subsequent i.p. injection of insulin (1.5 IU kg −1 ) lowered blood glucose level at a rate slower than that in normal lean mice, without reaching significant difference at 30 min of ITT (Fig. 5e vs. Fig. 3h), suggesting insulin resistance. In contrast, administration of D-allulose at −60 min significantly attenuated basal glucose levels at 0 min (Fig. 5e). Moreover, in the presence of Dallulose, i.p. insulin (1.5 IU kg −1 ) significantly decreased blood  . h-k Allu potentiated insulin action to lower blood glucose (h) and its AUC (j) in ITT, and these effects were abolished in the presence of Ex(9-39) in C57BL/6J mice fasted 4 h (i, j) and in Glp1r KO mice (k). n = 5-11. l-o Allu suppressed blood glucose elevation (l) and its AUC (n) in PTT, and these effects were completely blocked in the presence of Ex  in C57BL/6J mice fasted overnight (m, n) and in Glp1r KO mice fasted overnight (o). n = 5-6. Different letters indicate p < 0.05 by two-way ANOVA followed by Tukey's test (e), and *p < 0.05, **p < 0.01 by two-way ANOVA followed by Bonferroni's test vs. saline group (a-c, h, l). In d,  (Fig. 5e). These results suggest that D-allulose significantly restored insulin action in HFD-fed mice. Moreover, single p.o. administration of D-allulose did not change blood glucose levels at 0 min in fasted (16 h) HFD-fed mice ( Fig. 5f). Subsequently, the increases in blood glucose and AUC by challenge with pyruvate were significantly suppressed in the presence of D-allulose compared to saline (Fig. 5f, g). These data indicate that D-allulose significantly suppressed glucose production in HFD-fed mice with insulin resistance. Thus, D-allulose augmented insulin action, suppressed glucose production and enhanced insulin secretion in obese mice, complementary actions improving glucose tolerance. In hyperglycemic Glp1r KO mice fed HFD for 5 weeks or longer, p.o. administration of D-allulose neither decreased the elevated basal blood glucose level, nor improved glucose tolerance in ipGTT (Fig. 5h, i). Insulin levels trended higher at 15 min of ipGTT after D-allulose in WT mice (Fig. 5d), but not in HFD-fed Glp1r KO mice (Fig. 5j). Furthermore, treatment with Ex(9-39) counteracted the effects of D-allulose to lower basal blood glucose (0 min) and to improve glucose tolerance under glucose challenge (0-120 min) in HFD-fed mice (Fig. 5k, l). These data indicate that D-allulose ameliorates basal hyperglycemia and impaired glucose tolerance (IGT) in HFD-fed mice via GLP-1R signaling.
The effect of D-allulose on glycemia was also examined in type 2 diabetic db/db mice with genetic inactivation of the leptin receptor, which exhibits obesity and insulin resistance. In db/db mice, p.o. administration of D-allulose (1 g kg −1 ) at −60 min significantly lowered elevated basal blood glucose level at 0 min and rises in blood glucose levels at 30 and 60 min during ipGTT (Fig. 5m). Blood glucose levels over the period till 60 min were reduced by D-allulose treatment.
Chronotherapeutic effects of Allu on hyperphagic obesity. Impaired diurnal feeding rhythm is implicated in development of obesity, diabetes, and metabolic syndrome in humans and rodents 3,4,28 . In rodents, the arrhythmic feeding is featured by hyperphagia during light period (LP), and its correction ameliorates obesity 3,4 . In the present study using obese mice fed HFD for 5 weeks or longer, food intake in LP (7:30-19:30) was approximately doubled, whereas that in dark period (DP, 19:30-7:30) unaltered, exhibiting LP-specific hyperphagia accompanied by daily hyperphagia (Figs. 6a-c and 7a-c). Hence, we examined whether subchronic administration of D-allulose daily at LP onset ameliorates LP hyperphagia, obesity, and IGT in HFD-fed mice.
Subchronic daily p.o. administration of D-allulose (1 and 3 g kg −1 ) at the LP onset 7:30 attenuated LP hyperphagia without altering DP food intake, thereby improving diurnal feeding rhythm in HFD-fed mice (Figs. 6a-c and 7a-c). With 3 g kg −1 Dallulose, this amelioration of LP hyperphagia was accompanied by suppression of daily hyperphagia (Fig. 7c). These changes in food intake were paralleled by amelioration of elevated body weight   Supplementary Fig. 5a-c), and liver triacylglycerol (TG) content (Figs. 6f and 7g) and hepatic steatosis (Fig. 7f) without altering weights of several peripheral organs ( Supplementary Fig. 5d-i) in HFD-fed mice. Administration of 1 g kg −1 D-allulose for 9-10 days ameliorated arrhythmic feeding and visceral obesity ( Fig. 6 and Supplementary Fig. 6a-e), and tended to increase energy expenditure, fat oxidation and uncoupling protein-1 (UCP-1) expression without altering locomotor activity ( Supplementary  Fig. 6f-j). These data suggest that D-allulose may moderately enhance energy expenditure, which might additionally contribute to attenuation of visceral obesity, in agreement with previous reports that D-allulose increases energy expenditure 29 . D-Allulose treatment at 3 g kg −1 day −1 for 10 days significantly reduced the elevated basal blood glucose levels and tended to attenuate hyperinsulinemia in HFD-fed mice fasted 4 h (Fig. 7h, i). In ipGTT performed 1 day after termination of D-allulose treatment for 10 days, rises in blood glucose at 60 and 120 min were markedly suppressed (Fig. 7j), and both basal (0 min) and elevated plasma insulin levels (15 and 30 min) tended to be lower (Fig. 7k). The reductions in both blood glucose and plasma insulin levels suggested that subchronic D-allulose treatment improved glucose tolerance at least partly by ameliorating insulin resistance in HFD-fed mice, consistent with previous findings 17 .
Feeding and glycemic effects of Allu require vagal afferents. The effects of D-allulose shown here may involve the brain, however, D-allulose does not pass through BBB 30 , raising a question whether D-allulose exerts its effects at least partly through vagal afferent nerves 31,32 . The effects of 1 and 3 g kg −1 Dallulose to suppress food intake for 0.5-3 h were abolished in mice with subdiaphragmatic vagotomy (Fig. 9a vs. Fig. 9b). When the hepatic branch of vagal afferents, which sense gastrointestinal hormones 25 , was selectively denervated, the anorexigenic effect of 1 g kg −1 D-allulose for 1-3 h was blocked, whereas that at 0.5 h was unaltered (Fig. 9c vs. Fig. 9d). Subdiaphragmatic vagotomy also attenuated the ability of D-allulose to enhance glucose tolerance (Fig. 9e vs. Fig. 3a, c, e). These results indicated that the effects of D-allulose on food intake and glucose metabolism are largely mediated by vagal afferent nerves. D-Allulose activates vagal afferent neural pathways. We next examined whether p.o. D-allulose activates vagal afferent neurons and nucleus tractus solitarius (NTS) to which vagal afferents project and, if so, whether these effects are mediated via GLP-1R signaling. Immunohistochemical staining for cellular/neuronal activation markers, phosphorylation of extracellular signalregulated kinase 1 and 2 (pERK1/2) and c-Fos 27,33 , were examined. P.o. administration of D-allulose (1 g kg −1 ) induced expression of pERK1/2 in vagal afferent nodose ganglion neurons and in NTS of WT C57BL/6J mice, but not in Glp1r KO mice (Fig. 10a-h). Similarly, p.o. D-allulose (1 g kg −1 ) increased c-Fos expression in NTS (Supplementary Fig. 7). These results indicated Light period (7:30-19:30) Dark period (19:30-7:30) that p.o. D-allulose GLP-1R dependently activates vagal afferent neurons and NTS, the areas implicated in regulation of feeding 27,34 .
The anorexigenic effect of D-allulose was attenuated by vagotomy of hepatic branch (Fig. 9c, d), which suggested that GLP-1 might act through the hepatic vagal branch. We examined the effect of D-allulose on food intake in rats whose Glp1r was knocked down selectively in the ventral trunk of vagal afferents including hepatic branch, which was achieved by microinjection of AAV9-Glp1r-shRNA vector into the left NG. Three weeks after the microinjection, successful infection was confirmed by visualizing ZsGreen1 expression in whole left NG (Fig. 10k), but not in right intact NG. GLP-1R mRNA expression in the left NG significantly decreased in Glp1r knockdown (KD) rats to 41% of the level in control rats injected with AAV9-Scrambled (Scr)-shRNA vector (Fig. 10l). Neither GLP-1R mRNA expression in the right NG and hypothalamus nor anorexigenic CCK-A receptor mRNA expression in the left and right NGs, was altered in Glp1r KD rats ( Fig. 10l and Supplementary Fig. 8). These results indicated that the left NG-specific KD of GLP-1R was achieved by the Glp1r-shRNA construct used. In these Glp1r KD rats, p.o. D-allulose at 1 g kg −1 failed to significantly suppress food intake at 0.5, 1, 2, and 3 h after injection (Fig. 10m), while it significantly reduced food intake at all time points in control AAV9-Scr-shRNA-injected rats. These results collectively demonstrate that GLP-1R signaling via vagal afferents is necessary for the anorexigenic effect of oral D-allulose.

Discussion
In the present study, we found that single p.o. administration of D-allulose stimulated GLP-1 release, inhibited food intake, and promoted glucose tolerance in chow-fed and HFD-fed obese/ hyperglycemic mice. D-Allulose promoted glucose tolerance by enhancing insulin release and action and inhibiting glucose production. In HFD-fed obese mice, subchronic p.o. administration of D-allulose once daily at LP onset (7:30) attenuated LP hyperphagia to restore diurnal feeding rhythm, and thereby suppressed daily hyperphagia, obesity, adiposity, and IGT. Conversely, subchronic D-allulose at DP onset (19:30) was ineffective in attenuating LP hyperphagia and exerting the metabolic effects, showing chronotherapeutic action of D-allulose. These effects of D-allulose to attenuate feeding and promote glucose tolerance were diminished by a GLP-1R antagonist, and attenuated in Glp1r KO mice, and by vagotomy. Furthermore, the Glp1r KD specifically in the ventral trunk of vagal afferents eliminated the anorexigenic effect of D-allulose. These results identified oral D-allulose as a GLP-1 releaser that attenuates feeding and promotes glucose tolerance in lean and hyperglycemic/obese mice via a vagal afferent-mediated pathway. Furthermore, when time-specifically administered, chronic D-allulose corrected arrhythmic overeating, obesity and diabetes, suggesting a potential for chronotherapy (Fig. 10n).
The D-allulose-induced suppression of food intake at 0.5 h was not counteracted by i.p. pretreatment with Ex(9-39) (Fig. 2a, b), whereas it was blunted in Glp1r KO mice (Fig. 2f, g), vagal afferent-specific Glp1r KD rats (Fig. 10m), and vagotomized mice (Fig. 9b). These results suggest that Ex  at the dose used is unable to antagonize GLP-1R at an early period of 0.5 h possibly due to the following reasons. Ex(9-39) may not be fully absorbed at this early time point. Alternatively, i.p. administration may result in a substantial rise of Ex  in the general circulation but not in the microcirculation in the specific tissue where GLP-1 is sensed, including the intestine and portal.
Surprisingly, D-allulose stimulated release of GLP-1, but not PYY. We speculate that D-allulose may activate a subset of L-cells in the proximal or middle small intestine where the majority of Lcells express GLP-1, while most PYY-expressing L-cells appear in the distal ileum and large intestine 35,36 . Alternatively, D-allulose might evoke cellular signaling selectively linked to release of GLP-1, but not PYY, from L-cells expressing one or both peptides. The mechanism underlying selective release of GLP-1 by D-allulose remains to be elucidated.
The absence of glucose-induced GLP-1 secretion could be due to that the dose of glucose used (1 or 3 g kg −1 ) is lower than that (higher than 5 g kg −1 ) previously reported to increase plasma GLP-1 [37][38][39][40] . Although several studies documented GLP-1 secretion with glucose at 2-4 g kg −1 41,42 , others failed to observe the increment of GLP-1 with oral glucose at 2 or 3 g kg −1 40,43 . Alternatively, the time point for collecting the blood sample (60 min after oral load) might be too late to detect increase in plasma GLP-1, which was observed 5-10 min after oral glucose load in other reports 38,40,44,45 . Therefore, glucose might transiently activate proximal L cells, and is subsequently absorbed immediately in the proximal small intestine. Because D-allulose is slowly absorbed in the intestine compared to glucose 30 , it would stay in the intestinal lumen longer to stimulate L-cells.
The LP hyperphagia causes obesity in rodents 3,4 . In our study, D-allulose, administered at LP onset but not DP onset, Fig. 7 Chronotherapeutic effects of D-allulose (3 g kg −1 day −1 ) on hyperphagic obesity in HFD-fed mice. a-k HFD-fed obese C57BL/6J mice (HFD-fed) were subchronically treated for 10 days with p.o. Allu (3 g kg −1 day −1 ) or water once daily at LP 7:30. C57BL/6J lean mice fed standard chow (Chow-fed) were the lean control. Allu, compared to water, markedly attenuated increases in LP and daily HFD intake (a-c) and body weight gain (d). Subchronic Allu treatment ameliorated increased visceral WAT weight (e), hepatic steatosis (f), and elevated triacylglycerol content in liver (g) on Day 11. Scale bar, 200 µm. Allu treatment significantly ameliorated elevated basal blood glucose (h) with trend to attenuate hyperinsulinemia on Day 11 (i). In ipGTT at Day 11, rises in blood glucose at 60 and 120 min were markedly suppressed (j), and basal (0 min) and elevated plasma insulin levels at 15 and 30 min tended to decrease (k). l-s In HFD-fed Glp1r KO mice, the same subchronic Allu treatment did not alter LP HFD intake (l, n), slightly elevated DP HFD intake at Day 4 and later (m, n), and failed to significantly change body weight gain (o), visceral WAT weight on Day 11 (p), and blood glucose level (q) and its AUC (r) in ipGTT on Day 11 except an increase of blood glucose level at 30 min in ipGTT (q). Before sacrificing these Glp1r KO mice, portal vein was sampled 60 min after p.o. Allu (3 g kg −1 ) or saline and active GLP-1 was determined (s). n = 5-6. Different letters p < 0.05 and **p < 0.01 by two-way ANOVA followed by Tukey's test (a, d, h) or Bonferroni's test (j, q). #p < 0.01 by two-way ANOVA followed by Bonferroni's test vs. Day 0 (h). In m, effect of Allu treatment was significant with p < 0.05 by two-way ANOVA. *p < 0.05, and **p < 0.01 by one-way ANOVA followed by Tukey's test (c, e, g, i). *p < 0.05, and **p < 0.01 by unpaired t-test vs. saline group (n, s). Error bars are SEM ameliorated LP hyperphagia and visceral obesity in HFD-fed obese mice. Detailed mechanisms for the time-dependent effects of D-allulose remain unknown. However, since D-allulose inhibits food intake significantly for 0.5-6 h but not 24 h (Figs. 1a and 4a, b) and elevates plasma GLP-1 level for 0.5-3 h (Fig. 1d), D-allulose administered at early LP may be able to effectively suppress LP hyperphagia. Alternatively, the feeding center machinery whose impairment causes LP hyperphagia may have sensitivity to and be corrected by D-allulose administered in early LP. Notably, inversely to the correction of LP hyperphagia by D-allulose-GLP-1, Glp1r KO mice displayed deterioration of LP hyperphagia (Fig. 7n). These results collectively suggest that the physiological GLP-1 released postprandially serves to produce feeding rhythm, and that pharmacologic D-allulose and GLP-1, when administered in the time period with impaired GLP-1 rise, can ameliorate feeding arrhythmia.
The present results support that the release of GLP-1 by oral Dallulose activates vagal afferents, which in turn suppresses food intake and improves glucose metabolism (Fig. 10n). This is in consistent with previous reports that neural pathways contribute to the effect of GLP-1 and DPP-4 inhibitors on food intake and glucose metabolism 22,23,25,46 . This vagal afferent-mediated link of endogenous GLP-1 to food intake and glucose metabolism is consistent with several previous reports. Vagal afferents-specific Glp1r knockdown resulted in elevation of food intake for 1 h after overnight fasting and reduction of meal-evoked insulin secretion 23 . A study using β cell-specific Glp1r knockout showed that oral glucose-or DPP-4 inhibitor-induced elevation of endogenous GLP-1 improves glucose metabolism via extra-islet GLP-1Rs 47 . Intraduodenal infusion of metformin lowered hepatic glucose production via GLP-1R signaling and neuronal pathway including vagal afferents 21 .
We have demonstrated that oral D-allulose ameliorates obesity and diabetes via GLP-1 release relayed to vagal afferent signaling.
Currently, two incretin-based medicines, DPP-4 inhibitors and GLP-1R agonists, are clinically available. DPP-4 inhibitors expand the life-span of physiological GLP-1 and GIP, and thereby lower blood glucose. However, they are incapable of reducing body weight, possibly due to elevation of GIP that promotes adiposity [48][49][50] . GLP-1R agonists ameliorate both diabetes and obesity, decrease cardiovascular events, stroke and nephropathy in clinical trials 8,9 and ameliorate dementia and stroke in preclinical studies 51 . However, GLP-1R agonists also elicit adverse effects including a rise in heart rate, nausea and vomiting 6,10,11 , possibly by passing through the BBB and directly acting on neurons in the brain 13,52 . In contrast, the present study showed that D-allulose, a substance incapable of penetrating BBB 30 , ameliorated hyperphagia and obesity without inducing aversive behavior. This property may be due to the action of D-allulose to stimulate release of intestinal GLP-1, which interacts with restricted targets including the hepatic and/or intestinal branches of vagal afferents that serve to inform specific brain regions and peripheral organs to ameliorate hyperphagia, obesity and hyperglycemia 21,23 . This notion fits with a previous report that endogenous GLP-1 is sensed largely by the vagal afferents innervating the portal/liver areas 25 and gastrointestinal tract 21 . Together with the fact that no appreciable side effects have been reported for Dallulose 18 , D-allulose may provide a selective and safe treatment and/or prevention for obesity and diabetes.
We found that administration of D-allulose significantly lowered basal glucose level in hyperglycemic (>200 mg dl −1 ) HFDfed mice and db/db mice, while it did not alter basal glucose level (around 100 mg dl −1 ) in normal glycemic chow-fed mice. The results indicate the ability of D-allulose to ameliorate hyperglycemia without risk of hypoglycemic events, suggesting a therapeutic advantage. This specific effect could be produced by the acute D-allulose action to ameliorate insulin resistance in HFDfed hyperglycemic mice (Fig. 5e). Moreover, this glucose- dependent property of D-allulose may be due to its principal action to stimulate release of GLP-1 that works glucose dependently. Since D-allulose has sweet taste and an anorexigenic effect, consumption of D-allulose can decrease calorie intake while maintaining the pleasure of eating. The safety of D-allulose as dietary supplements has been approved by the US Food and Drug Administration (FDA, GRAS Notice No. GRN 498). Therefore, oral D-allulose may provide an efficacious, readily-applicable, and safe therapy to ameliorate obesity, diabetes, metabolic syndrome, and possibly associated cardiovascular/cerebral disorders.

Methods
Animal. Male C57BL/6J mice (H2b) and male Wistar rats were obtained from Japan SLC (Shizuoka, Japan) and male db/db mice were from CLEA Japan (Tokyo, Japan). The Glp1r −/− C57BL/6J (Glp1r KO) mice generated and provided by Dr. D. J. Drucker at University of Toronto and Dr. Y. Yamada at Akita University 26 . The animals were housed in individual cages for at least 1 week under controlled temperature (23 ± 1°C), humidity (55 ± 5%), and lighting (light on at 7:30 and off at 19:30). Standard chow (CE-2, CLEA japan) and water were available ad libitum. Mice and rats (8-to 16-weeks-old) were sufficiently habituated to handling before experiments.
Measurement of food and water intake. We performed feeding experiments with two protocols. In first protocol, the male mice were deprived of food from 18:00 (16 h fasting) with free access to water 1 day before the experiment. On next day at 9:50, D-allulose (0.3-3 g kg −1 , 10 ml kg −1 ), D-glucose (1 g kg −1 ) or saline (10 ml kg −1 ) was peroral (p.o.) or i.p. administrated, and at 10:00 standard chow (chow; CE-2) or HFD (HFD-32) was given. In second protocol, feeding experiments were started at 19:30 in male mice which had been deprived of food for 3 h (16:30-19:30) with free access to water. Then cumulative food and water intake for the following Subdiaphragmatic and hepatic-selective vagotomy. Bilateral subdiaphragmatic vagotomy was performed as previously reported 27 . A midline incision was made to provide wide exposure of the upper abdominal organ in male C57BL/6J mice anesthetized with tribromoethanol (200 mg kg −1 , i.p.). The bilateral subdiaphragmatic trunks of vagal nerves along the esophagus were exposed and cut. In the sham operation group, these vagal trunks were exposed but not cut. The vagotomized and sham-operated mice were maintained on a nutritionally complete liquid diet for human baby (Chilmil, Morinaga, Tokyo, Japan). One to two weeks after the operation, feeding experiments and GTTs were performed. After experiments, to confirm whether the surgical vagotomy was successful, an increase in stomach weight, a reported phenotype of subdiaphragmatic vagotomy 55 , was checked. Selective hepatic vagotomy was performed as described 56 . The ventral subdiaphragmatic vagal trunk was exposed as described above under anesthesia (200 mg kg −1 i.p. tribromoethanol). Since vagal hepatic branch forms a neurovascular bundle, this branch was selectively ligated by silk sutures and cut using microscissors. In the sham operation group, the hepatic branch was exposed but not cut. Hepatic vagotomized and sham-operated mice were recovered under chow (CE-2) diet for 1 week before experiments. It was reported that the hepatic vagotomy is judged successful if fasting-induced fat utilization in the epididymal fat tissue is decreased 57 . Hence, at the end of various measurements, mice were fasted overnight (16 h) and sacrificed and the epididymal fat weight was measured. We found an increased epididymal fat mass after hepatic vagotomy (sham 0.127 ± 0.010 g (n = 11) vs. hepatic vagotomy 0.190 ± 0.015 g (n = 12), p < 0.01 by unpaired t-test), confirming that the surgical hepatic vagotomy was successful. 2+ ] i in single nodose ganglion neurons. Single neurons of nodose ganglia (NGs) were prepared from male C57BL/6J mice (6-weeks-old) as described 33,58 . NGs were excised from the mice anesthetized with α-chloralose and urethane (0.1 and 1 g kg −1 , i.p.). The NGs were treated 20 min at 37°C with 0.1-0.5 mg ml −1 collagenase Ia (Sigma), 0.4-0.6 mg ml −1 dispase II (Roche, Basel, Swiss), 15 µg ml −1 DNase II type IV (Sigma), and 0.75 mg ml −1 bovine serum albumin (Sigma) in HKRB supplemented with 5.6 glucose. Separated single neurons were cultured in Eagle's MEM with 10% fetal bovine serum supplemented with 5.6 mM glucose for 12-24 h.

Measurements of [Ca
Measurements of [Ca 2+ ] i in the primary cultured single NG neurons were carried out as described previously 33,58 . Following incubating with 2 µM fura-2 AM (DOJINDO, Kumamoto, Japan) for 30 min at 37°C, the cells were mounted in chamber and superfused at 1.3 ml min −1 at 30°C with HKRB containing 5.6 mM glucose. Fluorescence ratio images at 510 nm following excitation at 340 and 380 nm were produced by an Aquacosmos ver. 2.6 (Hamamatsu Photonics, Shizuoka, Japan). When [Ca 2+ ] i changed within 5 min after addition of agents and their amplitudes were at least twice larger than the fluctuations of the baseline, they were considered responses. Only the neurons that responded to high KCl (55 mM) were analyzed.
Construction of shRNAs and viral vector production. A target sequence for rat Glp1r was chosen to design short hairpin RNAs (shRNAs) (Accession No. NM_012728.1: 5′-GTATCTCTACGAGGACGAG-3′. 880 to 898). A scrambled oligonucleotide sequence was used for specificity control (5′-TTCTCCGAACGTGTCACGT-3′). To construct shRNAs, forward oligonucleotides were designed to contain the sense and antisense sequences connected with a hairpin loop (5′-CTGTGAAGCCACAGATGGG-3′) followed by a poly(dT) termination signal. The annealed forward and reverse oligonucleotides were ligated into the pAAV-U6-ZsGreen1 vector (Takara Bio, Japan). Using this vector, the shRNA expression was driven by the mouse U6 promoter, whereas green fluorescence protein, ZsGreen1 expression was controlled by the cytomegalovirus (CMV) promoter. Briefly, AAV9-Glp1r-shRNA and AAV9-Scr-shRNA viruses were produced following triple-transfection of HEK293 cells with pAAV-shRNA, an adenoviral helper plasmid pAdeno, and a chimeric helper plasmid encoding AAV2 rep/AAV9 cap genes (pAAV2rep/AAV9cap) (a gift from Dr. James M Wilson). The rAAV particles were purified by a dual ion-exchange procedure using high-performance membrane absorption as previously described 59 . At 72 h after transfection, cell pellets were resuspended in 30 ml of Tris-buffered saline (TBS: 100 mM Tris-HCl [pH 8.0], 150 mM NaCl). Recombinant AAV was harvested by five cycles of freeze-thawing of the resuspended pellet. The crude viral lysate was initially concentrated by a brief two-tier CsCl gradient centrifugation for 3 h and then the vector fraction was dialyzed in MHA buffer (3.3 mM morpholinoethanesulfonic acid [MES], 3.3 mM HEPES [pH 7.5], 3.3 mM sodium acetate). Chromatography was performed with an AKTAexplorer 10 S fast protein liquid chromatography (FPLC) system (GE Healthcare Life Sciences). First, a dialyzed vector-containing fraction was loaded onto a cation-exchange membrane (Acrodisc unit with Mustang S membrane; Pall, East Hills, NY). The flow-through sample was further loaded onto an anion-exchange membrane (Acrodisc unit with Mustang Q membrane; Pall). After washing, bound virus on the Mustang Q membrane was eluted with a 0-2 M linear NaCl gradient in MHA buffer. Recombinant rAAV particle number was determined by quantitative real-time PCR. Primers sequences were 5′-ACCGTGTACAAGGCCAAGTC-3′ and 5′-GTCAGCTTGTGCTGGAT-GAA-3′.
AAV-mediated shRNA interference of nodose ganglion in rats. The left NG of male Wistar rats was surgically exposed under anesthesia (Tribromoethanol 200 mg kg −1 i.p.). A glass capillary (30-50 µm tip) was used to administer 2 µL AAV solution (Control vs. Glp1r knockdown (KD), 1.0 × 10 12 vector genomes ml −1 with 0.05% Fast Green) into the left NG with a BJ-110 microinjector (BEX, Tokyo, Japan). At the operation, the success of microinjection was determined by Fast Green Dye filling in the NG. We allowed animals to recover for at least 2 weeks to ensure expression of the viral constructs.
Statistical analysis. We did not perform power calculations, but the sample size and animal number for each group were chosen based on study feasibility and prior knowledge of statistical power from previously published experiments. Animals were randomly allocated to individual experimental groups and were performed without blinding. All data were expressed as means ± SEM. All groups from our data showed normal variance, and statistical analysis was performed by two-tailed unpaired t-test or by one-way or two-way ANOVA. When ANOVA indicated a significant difference among groups, these groups were compared by Dunnett's, Tukey's or Bonferroni's post hoc test. All statistical analyses were performed using Prism 5 (GraphPad Software, CA). p < 0.05 was considered significant.
Data availability. The data that support the findings of this study are available from the corresponding authors upon reasonable request.