j ourna l homepage: www.e lsev ie r .com/ locate /phb
Circadian and ultradian components of hunger in humannon-homeostatic meal-to-meal eating
Elizabeth C. Wuorinen a,⁎, Katarina T. Borer b
a College of Science and Mathematics, Department of Biology and Physical Education, Norwich University, Northfield, VT 05663, United Statesb School of Kinesiology, The University of Michigan, Ann Arbor, MI 48109, United States
H I G H L I G H T S
• No relationship of peak hunger ratings to previous energy deficit• Hunger in human meal-to-meal eating appears to be under circadian control.• Suggested ultradian hunger pacemaker synchronized with final gastrointestinal signals.• Hormones related to meal digestion/absorption transiently suppress central hunger arousal.
⁎ Corresponding author at: College of Science and Mathand Physical Education, Norwich University, 158 HarmoUnited States. Tel.: +1 802 485 2848; fax: +1 802 485 2
A unifying physiological explanation of the urge to initiate eating is still not available as human hunger in meal-to-meal eating may not be under homeostatic control. We hypothesized that a central circadian and a gastroin-testinal ultradian timingmechanism coordinate non-deprivationmeal-to-meal eating.We examined hunger as afunction of time of day, inter-meal (IM) energy expenditure (EE), and concentrations of proposed hunger-controlling hormones ghrelin, leptin, and insulin.Methods: In two crossover studies, 10 postmenopausal women, BMI 23–26 kg/m2 engaged in exercise (EX) andsedentary (SED) trials.Weightmaintenancemealswere provided at 6 h intervalswith an ad libitummeal at 13 hin study 1 and 21 h snack in study 2. EE during IM intervals was measured by indirect calorimetry and includedEX EE of 801 kcal in study 1, and 766–1051 kcal in study 2. Hunger was assessed with a visual analog scale andblood was collected for hormonal determination.Results:Hunger displayed a circadian variationwith acrophase at 13 and 19 h andwas unrelated to preceding EE.Hunger was suppressed by EX between 10 and 16 h and bore no relationship to either EE during preceding IMintervals or changes in leptin, insulin, and ghrelin; however leptin reflected IM energy changes and ghrelinand insulin, prandial events.Conclusions: During non-deprivation meal-to-meal eating, hunger appears to be under non-homeostatic centralcircadian control as it is unrelated to EE precedingmeals or concentrations of proposed appetite-controlling hor-mones. Gastrointestinal meal processing appears to intermittently suppress this control and entrain an ultradianhunger pattern.
A better understanding of the controls over human food intakegrows in importance as the prevalence of obesity in the US rises[16,17]. The urge to eat has been studied for years in animals andhumans, but a unifying physiological explanation of hunger and theurge to initiate eating is still not available [4,21]. Current prevailingview is that an interaction of hormones reflecting either a decline in
ematics, Department of Biologyn Drive, Northfield, VT 05663,333.n).
ghts reserved.
body fat reserves or in short-term energy availability with regulatoryhypothalamic and ventral tegmental brain areas provides the necessaryinformation for the initiation of meal taking and the regulation of ener-gy balance. The available evidence is stronger for the termination ofmeals through the intervention of gastrointestinal hormones such ascholecystokinin (CCK), GLP-1, and PYY [1,2,9,25], or to suppression offeeding and body fat accumulation by adipose tissue hormones leptinand insulin [31,42] than for the stimulation of thedesire or a drive to eat.
Physiological hunger entails seeking and ingestion of nutrients astheir desirability increases either in response to extended inter-meal in-tervals or to bodyweight and body fat losses [3] although such hunger isnot necessary for food intake which can also be stimulated by environ-mental circumstances such as social facilitation [39] and food palatability
[13,28,45]. Hunger was defined a century ago as conscious detection ofperiodic bursts of gastric contractions during times when the stomachwas empty [7]. This was followed half a century later with a definitionof hunger as a central neural motive state sensitive to activation or sup-pression by internal metabolic events as well as by non-homeostatic en-vironmental stimuli [30,37]. These definitions and theoretical constructsacknowledged that hunger involves both a central neural action and pe-ripheral energy signals in detection of nutrient need, be it in the form ofdepleted energy stores or of empty gastro-intestinal tract to direct foodseeking and ingesting behaviors. This theoretical dichotomy haspersisted to the present time.
The current view of hunger places the emphasis on a central neuralmechanism associated with nucleus accumbens and ventral pallidal cir-cuits that responds to hormonal signals reflecting the magnitude ofbody fat loss or shorter-term negative energy balance to activate foodseeking and eating [3,15,20,36]. A substantial body of information impli-cates hormone leptin released from the adipose tissue in proportion tothe size of subcutaneous fat depots [8] in signaling the status of bodyfat reserves to the brain areas responsible for the hunger drive. This en-docrine signal reflecting the metabolic need then affects the interactionbetween orexigenic and anorexigenic hypothalamic and hindbrain cir-cuits [31,42]. The withdrawal of leptin inhibition over the ventralpallidal hunger circuit when its reduced basal concentration reflects de-clines in body fatmasswas shown to contribute to increased desirabilityor incentive salience of food [3,15,20,36]. The hormonal mediation ofhunger is further supported by the temporal concordance between in-creases in hunger and concentrations of the gastric hormone ghrelinduring the interval preceding spontaneous meals and decreases inboth variables upon food ingestion [10] and by suppression of hungerafter intravenous infusion of supra-physiological concentrations of thehormone [44]. Ghrelin, in addition to the inference that it reflectsshort-term inter-meal energy deficits, has also been implicated in thecontrol of hunger after body weight loss when its concentrations werefound to increase in parallel with hunger [11]. Ghrelin, however, is notessential for the control of hunger asmice lacking the capacity for ghrel-in signaling exhibit normal food intake and weight maintenance [43]. Arecent meta-analysis has reinstated the focus on the possible involve-ment of ultradian periodicities of gastric contractions in hunger byshowing a temporal concordance between pyloric pressure waves andhuman hunger and the quantity of food consumed at a meal [35].
A useful approach toward a better understanding of factors control-ling hunger may be to hypothesize that different mechanisms operatewhen meals are taken in non-deprived state from meal seeking afterbody fat and bodymass losses. A decline in plasma leptin concentrationis clearly associated with increased hunger and increased efficiency ofweight regain after bodymass and body fat losses while leptin adminis-tration in weight reduced state suppresses both hunger and the meta-bolic adaptations [33,34]. However, this relation does not hold true innon-deprivation meal-to meal eating. We have recently shown that a400-kcal energy reduction in the morning meal increases hunger, butthat energy expenditure (EE) of similarmagnitude produced by exercise,or replacement of withheld meal calories and metabolically expendedcalories do not. Under these conditions, declines in plasma leptin and in-creases in total plasma ghrelin reflect the short-term fluctuations in en-ergy balance but are not correlated with changes in hunger [6]. Similarto our results, Seimon et al. [35] reported that intravenous nutrient infu-sions did not affect hunger or the amount of food eaten, while both ofthese variables were correlated with pyloric pressure waves. Thus, howthe interactions betweenperipheral energy depletion signals and CNS af-fect hunger is at themoment not clear. Even Cannon andWashburn rec-ognized that gastric contractions could not be the only stimulus ofhunger, since they found that stomach denervation did not abolish eat-ing and hunger.
In the present study we consider the possible contribution of centralcircadian and ultradian control to hunger in non-deprivation meal-to-meal eating. This consideration is justified by circadian and ultradian
periodicities in both human and animal meal taking under non-deprivation conditions. Discrete meals are typically taken by rodents atultradian intervals of 3 to 4 h and display a circadian segregation toonly the waking portion of the day [27]. Humans also eat during theirwakeful period in 3-h intervals if snacks are included and in about 6-h in-tervals if more substantive main meals are considered. This suggests theoperation of a central circadianmeal and hunger timingmechanism thatmay have an ultradian entrainment to the signals related to meal diges-tion. The focus of our two studies is on the circadian and ultradian influ-ences on hunger, butwe reportmeal intakes aswell despite the evidencethat the two are not strongly correlated [6,24].
Our interest in the possible role of circadian and ultradian control ofhunger was spurred by our recent observation of progressive and uni-form increase in hunger during the 6 h inter-meal intervals regardlessof systemic changes in energy availability [6]. We also observed thathunger ratings were lower in the morning and the evening comparedto mid-day meal times despite the long duration of nocturnal inter-meal interval and similar energy balance before the mid-day and eve-ning meals. This prompted us to formulate two hypotheses regardingthe control of hunger in non-deprived state, that: (1) a central neuralcircadian oscillator activates hunger during the wakeful period of theday to produce a hunger acrophase at mid-day, and (2) hormonal con-sequences of meal eating and digestion and mechanical sequelae of di-gestive food processing inhibit this central hunger drive and thusprovide cues for ultradian meal entrainment. Our hypothesis predictsthat the magnitude of hunger in non-deprivation meal-to-meal eatingis influenced by circadian time of day and ultradian interdigestive epi-sodes, and is unrelated to inter-meal EE or concentrations of ghrelin,leptin, or insulin. The alternative homeostatic hypothesis posits thatthe hunger reflects changes in pre-meal or systemic body energy avail-ability and associated reduction in negative feedback by leptin and insu-lin over hypothalamic and ventral pallidal hunger circuits and/or inincreased activation of these circuits by circulating concentrations ofghrelin [10,21,31]. We therefore tested these hypotheses in two studiesbymeasuring humanhunger ratings in non-deprivedmeal-to-meal eat-ing as a function of time of day, interdigestive periods, themagnitude ofenergy deficit experienced since the previousmeal, and plasma concen-trations of ghrelin, leptin, and insulin.
2. Methods
2.1. Subjects
The subjects were twenty healthy postmenopausal women, ten ineach of two cross-over studies (Table 1). The subjects were of similarweight and BMI that ranged from normal 23 kg/m2 tomoderately over-weight 26 kg/m2 The inclusion criteria were: fifty to sixty-five years ofage, body mass index (BMI) of 20 to 30 kg/m2, habitual exercise ofless than 20 min three times per week, and an absence of restricted di-etary intake. The exclusion criteria included the presence of endocrineand metabolic disease other than hormonally-corrected hypothyroid-ism, active dieting, smoking, the presence of musculo-skeletal disabil-ities that would preclude exercise, and the absence of the listedinclusion criteria. All subjects signed an informed consent approved bythe University of Michigan Medical School Institutional Review Board.The studies were conducted between April and January for study 1,and between March and January for study 2.
2.2. Experimental design and general study protocol (Fig. 1)
Both studies had a crossover design with exercise and sedentary tri-als assigned in a balanced order andwith at least oneweek inter-trial in-terval. Pre-participatory screening consisted of a health questionnaireand a detailed physical examination that included measurements ofweight, height, laboratory chemistries, and thyroid function. The exer-cise screen consisted of respirometric measurements (Physio-Dyne
metabolic cart, Quogue, NY) during a treadmill test consisting of 2% slopeincreases every 3 min at the walking speed of 4.5 km/h until the maxi-mal effort was achieved, using the respiratory quotient, RQ, of 1 as thecriterion. This allowed assignment of a specific relative exercise intensitybased on percent of maximal effort. The subjects were admitted to theUniversity of Michigan General Clinical Research Center at 18 h on theday before each trial. Lights out occurred at 22 h, and subjects wereawakened at 5:30 h. An intravenous catheter was inserted into an armvein at 6 h on the day of the study for sequential blood collection.
2.3. Meals and energy intake
On the day before each trial, a standardized meal consisting of 60%carbohydrates, 25% fat, and 15% protein containing 33% of weight-maintenance calories was provided at 19 h. All trial meals also hadthis macronutrient composition. Caloric intake during the trials wasassessed from measurements of food provided and any food leftuneaten. The inter-meal intervals (IMIs) were: IMI1 from the dinnerat 19 h prior to the start of the trial day to breakfast on the day of thetrial, IMI2 from the breakfast to lunch at 13 h, and IMI3, from thelunch to the dinner at 17 h. No adjustments in the quantity of food pro-vided were made for energy expended during exercise.
2.4. Exercise
No organized physical activity was available during the sedentarytrials (SED) in both studies. Exercise trials (EX) were carried out at
6 7 8 9 10 11 12 13 114
Study 1:
Study 2:
Ad Libitum m500 kcal meal
6 7 8 9 10 11 12 13 1514
exercise
Trials:SEDEx
Trials:SEDLOWHIGH
LOW
exerciseMeal Meal
HIGH
e
Time of Day
Fig. 1.Experimental design for the two studies. Dark rectangles represent times of exercise. Verthours.
moderate intensities and consisted of one bout of 120-minute-longtreadmill walking in study 1 and in study 2, there were two periods ofexercise of different intensity (starting at 8 h and at 14 h). Duringlow-intensity trial (LOW), subjects engaged in ten 15-minute bouts ofwalking each separated by a 5-minute rest period and during thehigher-intensity trial (HIGH), subjects walked for 10 exercise bouts of7.5 minutes separated by 10-minute rest periods.
2.5. General methods and procedures
2.5.1. Energy expenditure (EE)Assessment of resting non-exercisemetabolismwas donewith indi-
rect calorimetry apparatus (DeltaTrac II, SensorMedics, Yorba Linda,California) for 15 min upon subject's awakening in the morning, aftereach meal, after each exercise period as well as in study 2, at 12 h,18 h, 0 h, and 4 h using a hood covering the face and with the subjectin reclining position. Exercise EE was measured by indirect calorimetryduring the first 30-minute period of each hour of exercise with thesubjects breathing through a mouthpiece connected with tubing toa metabolic cart (Physio-Dyne, Quogue, New York). For timeswhen respirometric measurements were not performed, datawere extrapolated from the collected measurements. Conversionof respirometric data to calories was done by the method of Frayn[18].
2.5.2. Hunger ratingsHunger was assessed with a 100 mm visual analog scale (VAS) vali-
dated for variation in energy balance [22] where hunger was anchoredwith zero at one end and extreme valuation at the opposite end of thescale. The distances marked were converted to percentages of the fullscale.
2.5.3. Blood collectionBlood was removed at hourly intervals and for 15 min and 30 min
intervals at the start of meals and exercise. Blood was collected intoice-chilled EDTA-coated tubes containing 250 KIU/ml blood of aprotinin(Sigma Chemical, St Louis, MO) and 10 μl/ml blood of dipeptidylpeptidase-IV inhibitor (Linco Research, St Louis, MO). Plasma was keptfrozen at −80 °C.
2.5.4. Hormone measurementsConcentrations of total ghrelin, leptin, and insulin were measured
with radio-immunoassays (Linco Research, St Louis, MO). The intra-and inter-assay coefficients of variation (CVs) were, respectively,
165 17 18 19 2120 12322 24 2 3 4 5 6
Ad Libitum snackMeal
eal
16 17
LOW
HIGH
xercise
Time of Day
ical arrows indicate the timing of themeals in the studies. The abscissa shows time of day in
Table 2Energy intake (EI) during three main meals and energy expenditure (EE) during inter-meal intervals (IMIs).
Condition EEa during IMI1(18 to 6 to 7 h)
EIa at 6 or 7 h EEa during IMI2(6 or 7 to 13 h)
EIa at 13 h EEa during IMI3(13 to 19 h)
EIa at 19 h EIa at 21 h
SED 1b 709 ± 34.6(2.97 ± 0.14)
465 ± 12.5(1.95 ± 0.05)
344 ± 23.5(1.44 ± 0.1)
747 ± 76.4(3.13 ± 0.32)
SED 2 734 ± 20.5(3.07 ± 0.09)
450 ± 13.9(1.88 ± 0.06)
404 ± 10.9(1.69 ± 0.05)
450 ± 13.9(1.88 ± 0.05)
354 ± 38.3(1.48 ± 0.16)
450 ± 13.9(1.88 ± 0.06)
256 ± 44.2(1.07 ± 0.19)
EX 1 736 ± 19.4(3.08 ± 0.08)
468 ± 8.9(1.96 ± 0.04)
801 ± 24.6(3.35 ± 0.10)
699 ± 79.8(2.93 ± 0.33)
EX 2 L 736 ± 20.0(3.08 ± 0.08)
450 ± 13.9(1.88 ± 0.06)
986 ± 32.9(4.13 ± 0.14)
450 ± 13.9(1.88 ± 0.06)
838 ± 24.8(3.51 ± 0.10)
450 ± 13.9(1.88 ± 0.06)
319 ± 62.3(1.34 ± 0.26)
EX 2 H 748 ± 25.7(3.13 ± 0.11)
450 ± 13.9(1.88 ± 0.06)
1051 ± 37.3(4.4 ± 0.16)
450 ± 13.9(1.88 ± 0.06)
766 ± 29.8(3.21 ± 0.12)
450 ± 13.9(1.88 ± 0.06)
353 ± 73.6(1.48 ± 0.31)
L refers to lower intensity and H to higher-intensity exercise.a Total kcal expended during the IMI or eaten at a meal, (total MJ).b Refers to study number.
10.1% and 11.8% for ghrelin, 9.1% and 14.2% for leptin, and 2.2% and 20%for insulin.
2.5.5. Statistical analysesData are presented as means and standard errors of the mean,
and alpha of 0.05 or less was the criterion of significant difference.Between-treatment effects were evaluated with a two-way repeat-ed measures mixed model ANOVA (treatments versus postprandialand exercise time points) using SAS software version 9.3 (SAS Insti-tute, Cary, NC). Repeated measures ANOVA was used to comparepeak pre-meal hunger ratings and hormone concentrations at thefour meal times, at 6 to 7 h, at the two acrophases at 13 h, 19 h, aswell as at 21 h. Tukey's and Dunnett–Hsu's post-hoc analyseswere used to compare between-treatment effects. The 24-hourday was divided into inter-meal segments during which energy ex-penditure (EE) was calculated from indirect calorimetry and ex-trapolations of the data. The EE values during the IMIs werecorrelated with peak pre-meal hunger ratings using Pearson prod-uct moment correlation analysis (Excel 2007).
3. Study 1
The hypothesis being tested was that the hunger in meal-to-mealnon-deprivation eating will be positively correlated with increases inpre-meal energy expenditure (EE) and with changes in plasma ghrelinconcentrations and negatively correlated with changes with plasmaleptin concentrations.
3.1. Experimental design and procedures
An intravenous catheter was inserted into an arm vein at 6 h onthe day of the study for sequential blood collection. A 500 kcalmeal was provided at 6 h making the IMI1 11 h long. The morningmeal consisted of a bagel with peanut butter and jelly and had tobe consumed in its entirety. The mid-day meal was provided at13 h in ad libitum amounts making the IMI2 7 h long. The mid-daymeal consisted of baked chicken breast, cooked rice and corn, saladwith ranch dressing, potato roll, margarine, banana, vanilla icecream with chocolate icing, and a beverage containing dextrose.Exercise took place between 10 and 12 h and was carried out at46 ± 1.4% of VO2 max. The study ended at 17 h rendering the IMI3incomplete.
3.2. Results
3.2.1. Energy expenditureOvernight IMI1 resting EEwas 709 and 736 kcal in SED and EX trials,
respectively (Table 2). During IMI2, EE in EX trial was 801 kcal of which
exercise itself contributed 537.7 ± 35.7 kcal. In the IMI2 of SED trial, EEwas 43% lower (p b 0.0001).
3.2.2. Hunger ratingsPrior to themorningmeal, hunger was rated at between 42 and 52%
of full VAS scale and abruptly declined to between 14 and22% aftermealconsumption (Fig. 2, top). The hunger ratings increased linearly be-tween themorning andmid-daymeal in SED trial. In the EX trial, the ini-tial rise in the hunger rate was comparable to that in SED trial but wasinterrupted by a 25.3% decline in the ratings during 2 h of exercise(F1,18 = 2.22, p b 0.05). In both the SED and EX trial, a peak hunger rat-ing of about 77%was recorded before themid-daymeal. Hunger ratingsbefore the morning and mid-day meal were significantly different(F1,9 = 4.5, p b 0.05).
3.2.3. Energy intakeDuring the ad libitum mid-day meal, similar amounts of energy
were consumed, 747 kcal during the SED trial, and 699 kcal duringthe EX trial (Table 2). This represented 117% overconsumption inSED trial for the EE during IMI2, and only 15% overconsumption ifboth IMI1 and IMI2 EE as well as calories eaten during the morningmeal were taken into account. By comparison, in the EX trial, mid-day meal represented about a 13% underconsumption for the EEduring IMI2, and 24% underconsumption if both IMI1 and IMI2 EEas well as calories eaten during the morning meal were accountedfor.
3.2.4. Relationship between pre-meal EE and hungerThere was no relationship between the pre-meal EE during the IMI1
and IMI2 on one hand and peak hunger ratings on the other before thestart of the meals in either SED (open circles) or EX (solid circles) trials(r = 0.06; Fig. 3, top).
3.2.5. Changes in hormone concentrationsLeptin concentrations started to decline toward the end of ex-
ercise bout in IMI2 (Fig. 4, top) and attained significance relativeto leptin concentration in SED trial 1 h postexercise (F4,36 =39.1, p b 0.0001). Leptin concentration during EX trial remainedsignificantly lower relative to SED trial throughout the remainderof study period. No trial differences were seen in either postpran-dial insulin responses during the first half of IMI2 and IMI3 pe-riods or in postabsorptive fasting insulin levels during EX whichtook place during the second half of IMI2 (Fig. 4, center). Plasmaghrelin concentrations rose during the postabsorptive period ofIMI2 and declined after food ingestion. A trend for ghrelin concen-tration to rise during exercise did not reach significance (Fig. 4,bottom).
0
10
20
30
40
50
60
70
80
906:
00
7:00
8:00
9:00
10:0
0
11:0
0
12:0
0
13:0
0
14:0
0
15:0
0
16:0
0
17:0
0
Per
cen
t H
un
ger
Time of Day
Study 1 Sedentary
Exercise **
0
10
20
30
40
50
60
70
80
90
6:00
8:00
10:0
0
12:0
0
14:0
0
16:0
0
18:0
0
20:0
0
6:00
Per
cen
t H
un
ger
Time of Day
Study 2 Sedentary
Low
High
*
A
B
Fig. 2. Hunger ratings in study 1 (A) and study 2 (B). Arrows represent meal times; rect-angular bars represent exercise. Hunger was suppressed during the exercise bout duringthe late morning in study 1 (F1,18 = 2.22, p b 0.05). Exercise had no effect on hunger(B) during the morning exercise bout of study 2 but suppressed it significantly and tothe same extent at both exercise intensities (p b 0.05) during afternoon exercise.
0
20
40
60
80
100
120
0 200 400 600 800 1000 1200
Per
cen
t H
un
ger
Energy Expenditure (Kcals)
Study 1 Sedentary
Exercise
0
20
40
60
80
100
120
0 200 400 600 800 1000 1200
Per
cen
t H
un
ger
Energy Expenditure (Kcals)
Study 2
Sedentary
Low Intensity Exercise
High Intensity Exercise
A
B
Fig. 3. Correlation between energy expenditure and peak hunger in study 1 (A) and study2 (B). Correlation between energy expenditure during IMI 1, IMI2 and peak hunger priorto the meals during those time periods in study 1, r = 0.06 and during IMI 1, IMI2, IMI3and peak hunger in study 2, r = 0.0002.
The hypothesis being tested was that hunger in meal-to-meal non-deprivation eating will be positively correlated with the pre-meal EEof exercise independently of exercise intensity and with changes inplasma ghrelin concentration, and negatively correlated with plasmaleptin concentration.
4.1. Experimental design and procedures
An intravenous catheter was inserted into an arm vein at 6 h on theday of the study for sequential blood collection. Standardized mealssimilar to those in study 1 were provided at 7 h, 13 h, and 19 h, eachproviding 33% of the daily energy intake and supplying 25 kcal/kg ofsubject's body weight. An ad libitum snack, which consisted of bagelsand cream cheese, was offered at 21 h to permit the measurement ofany compensatory consummatory response. The durations of IMIs
were 12 h for IMI1, 7 h for IMI2, and 6 h for IMI3. The post-dinnerIMI1 was interrupted by the ad libitum snack 2 h after its start.Lower-intensity EX bouts (LOW) were carried out at 36%, and higher-intensity bouts (HIGH) at 52%, of maximal effort. The trials ended at6 h the next day.
4.2. Results
4.2.1. Energy expenditureOvernight IMI1 resting EEwas similar in SED and EX trials and ranged
between 734 and 748 kcal (Table 2). During IMI2, EE of EXwas similar inLOW (986 kcal) and HIGH (1051 kcal) trials and was 144% and 160%higher respectively than the EE in the SED trial (404 kcal), a differencethatwas significant at p b 0.0001. A similar EE difference between EX tri-als and SED trial was obtained in IMI3 with EE of EX in LOW trial(838 kcal) and HIGH trial (766 kcal) twice as high as in SED trial(354 kcal).
4.2.2. Hunger ratingsPrior to themorningmeal, hunger was rated at between 23 and 38%
of full VAS scale and abruptly declined to below 10% after morningmealconsumption (Fig. 2, bottom). During the IMI2, the hunger ratings
increased linearly between the morning and mid-day meal and in par-allel in both SED and both EX trials and reached a nearly identicalpeak rating of 80% before themid-daymeal. Second bout of EX occurredduring the postprandial phase of IMI3 and caused a similar suppressionof hunger ratings in both HIGH and LOW trials relative to the SED con-dition (F1,18 = 4.66, p b 0.02 and F1,18 = 3.63, p b 0.05, respectively).In both EX and SED trials, a peak hunger rating of about 80% was againattained before the evening meal The difference between the hungerratings before themorningmeal on one hand andmid-day and eveningmeals on the other, was significant (F2,9 = 8.12, p b 0.02 and F2,9 =4.98, p b 0.05). At any given meal, there was no between-trial differ-ence in peak hunger ratings.
4.2.3. Energy intakeAll food in the three main meals allotted and adjusted to subjects'
body mass was consumed in entirety. The ad libitum evening snack didnot produce compensatory increases for EE during EX trials (Table 2).The snack energy consumed was 256 kcal in SED trial and 319 and353 kcal, respectively, in LOW and HIGH exercise trials. If only IMI3 EEis considered, the snack in SED trial represented a 50% overconsumptionfor the preceding EE. On the other hand, the snack energy intake appro-priately matched the EE during preceding LOW (−8%) and HIGH (5%)EX trials. If EE and energy consumption during IMIs 1 through 3 are con-sidered, then total daily energy intake exceeded by 8% the EE in the SEDtrial, and resulted in a 34 to 35% energy deficit in the two EX trials.
4.2.4. Relationship between pre-meal EE and hungerThere was no relationship between the pre-meal EE during the IMI1
and IMI2 and peak hunger ratings before the start of the meals in eitherSED (open circles), LOW (solid triangles) or HIGH EX (solid squares) tri-als (r = 0.0002; Fig. 3, bottom).
4.2.5. Changes in hormone concentrationsDeclines in plasma leptin concentrations started at the end of first
EX bout in IMI2 in both LOW and HIGH trials and became more pro-nounced during the second EX bout in IMI3. The declines in plasmaleptin concentrations attained significance relative to SED trial bythe end of the second exercise bout (F2,16 = 11.48, p b 0.001)and remained depressed through the night (Fig. 4, top). Exercisesuppressed insulin secretion to a similar extent in LOW and HIGHtrials during both IMI2 and IMI3 (morning: F2,16 = 4.04, p b 0.05; af-ternoon: F2,16 = 34.45, p b 0.0001). No treatment effect was evidentin the postprandial insulin response to the third meal and eveningsnack (Fig. 4, center). Plasma ghrelin concentrations rose prior tothe meals and declined after food intake but no significant time ortrial effects were found (Fig. 4, bottom).
4.2.6. Circadian pattern of maximal hunger ratingsIn both studies, peak hunger ratings at the morning meal and the
evening snack were significantly lower than those at mid-day and eve-ning meals (F1,9 = 4.5, p b 0.05 in study 1; F2,9 = 8.12, p b 0.02 andF2,9 = 4.98, p b 0.05 in study 2; Fig. 5).
5. Discussion
This study allows four principal conclusions relevant to the under-standing of the control of human hunger during non-deprivationmeal-to-meal eating. First, at habitual meal times, hunger ratings inthe two studies (Fig. 2), and the amount of food consumed when itwas offered in ad libitum amounts (Table 2), bore no relationship topre-meal EE that was incurred by variable intensity and duration of ex-ercise exposure (Fig. 3). A lack of a relationship between pre-meal EEand the size of spontaneous meal was previously described in rats[27]. Hunger ratings and food consumption also were unrelated to var-iations in pre-meal energy balance caused by exercise or intravenous in-fusions of nutrients in our earlier study [6] and studies of others [35]. An
absence of a correlation between the energy expended prior to a mealtime and the magnitude of hunger ratings or food consumption at thesubsequent meal suggest that homeostatic signals reflecting energyavailability may not play an important role in the initiation of non-deprivation meal-to-meal eating or the amount of food consumed at ameal.
Second, in the current study, we found no support for the role ofchanges in the plasma concentrations of leptin, insulin, or ghrelin inthe control of hunger or ad libitum food consumption. Exercise EEresulted in a sustained decline in plasma leptin concentration in bothstudies (Fig. 4). Declines in leptin concentration were associated withsuppression of hunger during EX, a change in the opposite direction tothat postulated to operate in a homeostatic model of appetite control[21,31,42]. In addition, there was poor temporal concordance betweenthe timing of the rise in hunger ratings and changes in plasma concen-tration of leptin. In study 1, hunger ratings changed before leptin con-centration, and in study 2, hormone changes preceded hunger changes.
Changes in plasma insulin concentration also lent no support to theview that insulin has an influence on hunger or food consumption. Insu-lin was close to the basal fasting level in both exercise and SED trials instudy 1 at the timewhen exercise was associated with hunger suppres-sion and low plasma insulin should have facilitated hunger. In study 2,there againwas a poor temporal concordance between insulin and hun-ger changes. Exercise reduced postprandial insulin response during bothIMI 2 and 3 but suppressed hunger only during the postprandial part ofIMI3. We found no evidence that increased energy expenditure beforea meal was associated with consistent increases in plasma ghrelin al-though exercise was associated with a non-significant amplification ofpre-meal ghrelin concentration rise in study 1. In contrast to several-hour long changes in plasma leptin concentration in response to 400 to1700-kcal EE expenditure of exercise, plasma ghrelin changes were oflow magnitude and abolished after meal eating. Thus the expected in-verse relationship between declines in plasma insulin and leptin[31,42] or increases in ghrelin [10] on one hand, and increased hungeron the other, was not seen in the present studies. In effect, plasma leptinand insulin concentrations changed in the opposite of the expected di-rection when both the hunger and the hormones were affected by exer-cise in the present studies. Thus, we show in this study that neitherhunger nor food consumed during a meal in non-deprived state bears ahomeostatic relationship to calories expended before the meal or thehormones inferred to reflect pre-meal energy availability.
The third important finding of this study is a demonstration of circa-dian patterning of hunger ratings during the wakeful portion ofnycthemeral period. Low hunger ratings were found at dark–light tran-sition in themorning and light–dark transition in the evening. A hungeracrophase was apparent between 10 and 19 h, and hunger suppressionby exercise was confined to this phase of the rhythm and ineffectiveoutside this timeframe. Thus, no suppression of hunger by exercisewas seen during IMI2 in study 2 when exercise took place between 7and 9 h (bottom of Fig. 2), but was effective in doing so during thesame IMI in study 1, where exercise started at 10 h (top of Fig. 2). Fur-ther, exercise suppressed the appetite during IMI3 in study 2 where ex-ercise started at 14 h and the magnitude of the suppression was afunction of EE and not of exercise intensity. This suggests that suppres-sion of hunger by exercise EE requires the hunger to exceed a thresholdthat is attained during its mid-day 10 to19 h acrophase.
Circadian control of feeding in mammals was shown to depend onextensive interconnections between the master circadian clock in thehypothalamus, the suprachiasmatic nucleus (SCN), and the brain areasimplicated in the control of feeding [29]. Neural pathways throughwhich the photo-entrainable SCN controls behavioral and endo-crine rhythms related to energy balance include direct projectionsto subparaventricular zone (SPZ), medial preoptic area (MPOA),dorsomedial hypothalamic nucleus (DMH), and the paraventricularnucleus (PVN) of the hypothalamus [19,29]. DMH, which is inner-vated both by the SCN and the SPZ, also controls circadian pattern
Circadian Control of Human HungerSed Study 2Sed Study 1Low Ex Study 2High Ex Study 2Exercise Study 1
Fig. 5. Circadian and ultradian patterning of human hunger revealed by mean visual ana-log scale ratings of hunger in studies 1 and 2. Arrows representmeal timing for the studies;6:00— study 1; 7:00— study 2; 13:00— both studies; 19:00— study 2; 21:00— ad libitumsnack study 2. Rectangular bars represent exercise.
of feeding, sleep–wakefulness, and locomotor activity. SCN also in-fluences the circadian control of food intake, metabolism, and ener-gy expenditure through its fibers projecting to the arcuate nucleus(ARC), the ventromedial hypothalamus (VMH), and the ventralpart of the lateral hypothalamus (LH), all areas implicated in thecontrol of feeding and energy regulation. Many interneurons fromthe SCN inhibit the PVN through γ-aminobutyric acid neurotrans-mission to facilitate parasympathetic (PS) functions. Consequently,most viscera receive SCN-dependent circadian time cues via theirPS and/or sympathetic innervations that reflect metabolic and di-gestive events at peripheral sites [19].
Evidence for a circadian control of feeding and energy regulation issupported by lesioning studies. Destruction of SCN results in the lossof all bioenergetic circadian responses including the drinking patternand locomotor activity [26,38]. Destruction of VMH and ARC nucleiwithin the medial basal hypothalamus disrupts circadian alternationbetween active and inactive periods of food seeking and eating and re-sults in extended 24-hour overeating [23]. Further, when the nocturnalpart of the circadian sleep–wake cycle in humans is truncated, inappro-priate overeating during extendedwakeful periods ensues [32] contrib-uting to obesity and associated health risk factors [40]. Similarly, aseasonal change in the length of circadian exposure to light produceschanges in feeding and body fat accumulation in some mammals [12].All of these lines of evidence are consistent with the inference thatmeal tomeal eating in humans is guided by circadian changes in the in-tensity of hunger.
The fourth inference resulting from our studies is that hunger innon-deprivation meal to-meal eating displays an ultradian patternthat is associated with the duration of gastro-intestinal food transit
Fig. 4. Study 1 (left) and study 2 (right) concentrations of leptin (top), insulin (middle), and ghrstudy 1, exercise suppressed hunger (F1,18 = 2.22, p b 0.05), however therewas no significant dleptin, reflecting the energy deficit of exercise in the afternoon (Fdf4,36 = 39.1, p b 0.0001). Thsignificant difference in ghrelin although concentrations tended to be higher during the exercthere is hunger suppression. In study 2, onebout ofmorning exercise suppressed leptin in the aftconcentrations in the afternoon and evening hours (Fdf2,16 = 11.48, p b 0.001). Concentrationsing: Fdf2,16 = 4.04, p b 0.05; afternoon: Fdf2,16 = 34.45, p b 0.0001); however there was no diffetween trials on ghrelin concentrations, thus no connection when hunger was different.
and not with the magnitude of pre-meal EE. We show that the pro-gressive and linear inter-meal rise in hunger is coincident with apostprandial decline in insulin concentration (middle of Fig. 4) andwith increases in ghrelin concentration during the postabsorptiveportion in the IMI (bottom of Fig. 4). A similar concordance is seenbetween the progressive increases in hunger and plasma concentra-tions of gastric inhibitory peptide released from the proximal part ofthe gastrointestinal tract [6]. We also show that the magnitude ofpre-meal hunger bears no relationship to preceding EE during theIMI (Fig. 3). These associations suggest that inter-meal rises in thehunger ratings are influenced by changes in the postprandial con-centrations in plasma insulin and digestive hormones, and possiblyalso by gastrointestinal motility induced by inter-meal gastrointesti-nal emptying rather than by sensing and reacting to changes in ener-gy availability. A presence of an endogenous ultradian rhythm ofgastric contractions that free runs in the absence of eating [7] andbecomes attenuated with time [5] allows a conclusion that the char-acteristics and digestive consequences of food eaten to fullness athabitual meal times can entrain this ultradian IMI hunger rhythm.
This study has several limitations. One is the use of postmenopausalsubjects to the exclusion of other age and gender segments of the pop-ulation.We have chosen thismodel to avoid the possible interference ofsex hormones in the control of hunger and energy regulation. Anotherlimitation is the small sample size. A larger population group mayhave shown a significant difference in the variables measured. Otherlimitations include provision of fixed quantities of food atfixed intervalsand a single ad libitummeal to assess compensatory reactions of hungerand food intake to pre-meal EE. Our failure to vary the duration of IMIscould affect the results as it was shown that consummatory compensa-tion for preceding negative energy balance may increase with a longerIMI [41]. Finally, we did not test the circadian hypothesis by measuringhunger across the 24 h period or used experimental conditions thatwould display free-running rhythms and entrainment. We agree thatthis study is observational rather than interventional concerning the cir-cadian variation of the hunger, but feel justified in making circadian in-ference on the strength of our diurnal data. Additional studies shouldaddress these study limitations.
Collectively, and despite these limitations, our data are consistentboth with the central origin of hunger drive as first conceptualizedby Morgan [30] and Stellar [37] and more recently elaborated byBerridge [3], Figlewicz and Benoit [15], Fulton et al. [20] and Sipolset al. [36], and with the peripheral contributions to hunger fromthe gastrointestinal mechanical [7,35], and hormonal [9] signals asrevealed by others. We add to this concept the proposition for a cir-cadian organization of the intensity of this central hunger drive. Wereport that human hunger reaches an acrophase during the 10 to19 h mid-portion of the day when it can be blocked by ultradian ep-isodes of gastrointestinal filling or exercise. Our data also suggestthat hormonal and probably gastrointestinal motor signals associat-ed with emptying of the gastrointestinal tract after a meal most like-ly guide progressive inter-meal increases in hunger. They most likelyentrain an endogenous ultradian rhythm of gastric contractionsdemonstrated a century ago by Cannon and Washburn [7]. Thusour inference of a circadian central drive state and the ultradian pat-terning of human hunger in non-deprived state resembles a proba-bly widespread blueprint exemplified by the feeding mechanism ina blowfly, where the brain facilitates meal initiation when the cropis empty, the meal is terminated by a neural negative feedback
elin (bottom) graphs. Arrows representmeal times; rectangular bar represents exercise. Inifference in leptin and insulin concentrations at this time. Therewas a significant change inere was no significant difference in the concentrations of insulin in study 1. There was noise bout, which reflects the energy deficit, rather than lower as would be expected whenernoonhours reflecting the energydeficit. Similarly, twobouts of exercise decreased leptinof insulin were suppressed following the first two meals during the exercise bouts (morn-rence after the evening meal and ad libitum snack. There was no significant difference be-
from a full crop, and the next meal is initiated when this feedbackwanes upon depletion of the crop contents [14]. Our hunger growson a time scale consistent with the emptying of the gastrointestinaltract and with changes in hormone concentrations associated withdigestion and absorption of the meals and not in response to pre-meal changes in energy availability. A circadian SCN oscillator facili-tates maximal hunger during the mid-portions of the waking period.The non-homeostatic timing of meal-to-meal eating and influencesof food palatability and some other environmental factors may con-tribute to, and in part account for, the progressive rise of obesity inthe US [16,17] and therefore deserves attention and further study.
[2] Batterham RI, Cohen MA, Ellis SM, Le Rouz CW, Withers DJ, Frost GS, et al. Inhi-bition of food intake in obese subjects by peptide YY (3-36). N Engl J Med2003;349:941–8.
[3] Berridge KC. “Liking” and “wanting” food rewards: brain substrates and roles in eat-ing disorders. Physiol Behav 2009;97:537–50.
[4] Blundell JE, Goodson S, Halford JC. Regulation of appetite: role of leptin in signalingsystems for drive and satiety. Int J Obes Relat Metab Disord 2001;25(Suppl. 1):S29–34.
[5] Boldireff WN. Le travail periodique de l'appareil digestif en dehors de la digestion.Arch Des Sci Biol 1905;11:1–157.
[6] Borer KT, Wuorinen EC, Ku K, Burant C. Appetite responds to changes in meal con-tent, whereas ghrelin, leptin, and insulin track changes in energy availability. J ClinEndocrinol Metab 2009;94:2290–8.
[7] CannonWB,Washburn AL. An explanation of hunger. Am J Physiol 1912;29:441–54.[8] Considine RV, Considine EL, Williams CJ, Hyde TM, Caro JF. The hypothalamic
leptin receptor in humans: identification of incidental sequence polymorphismsand absence of the db/db mouse and fa/fa mutations. Diabetes 1996;45(7):992–4.
[9] Cummings DE, Overduin J. Gastrointestinal regulation of food intake. J Clin Invest2007;117:13–23.
[10] Cummings DE, Purnell JO, Frayo RS, Schmidova K, Wisse BE, Weigle DS. A prepran-dial rise in plasma ghrelin levels suggests a role in meal initiation in humans. Diabe-tes 2001;50:1714–9.
[11] Cummings DE, Weigle DS, Frayo RS, Breen PA, Ma MK, Dellinger EP, et al. Plasmaghrelin levels after diet-induced weight loss or gastric bypass surgery. N Engl JMed 2002;346:1623–30.
[12] Dark J, Zucker I, Wade GN. Photoperiodic regulation of body mass, food intake, andreproduction in meadow voles. Physiol Behav 1998;65:277–88.
[13] De Castro JM, Bellisle F, Dalix AM. Palatability and intake relationships in free-livinghumans: measurement and characterization in the French. Physiol Behav2000;68(3):271–7.
[14] Dethier VG, Solomon RL, Turner LH. Sensory input and central excitation and inhibi-tion in the blowfly. J Comp Physiol Psychol 1965;60:303–13.
[15] Figlewicz DP, Benoit SC. Insulin, leptin, and food reward: update 2008. Am J Physiol2009;296:R9-19.
[16] Flegal KM, Carroll MD, Kit BK, Ogden CL. Prevalence of obesity and trends in thedistribution of body mass among US adults, 1999–2008. JAMA 2012;307:491–7.
[17] Flegal KM, Carroll MD, Ogden CL. Prevalence of obesity in the United States,2009–2010. NCHS Data Brief; Jan: 1–8 2012.
[18] Frayn KN. Calculation of substrate oxidation rates in vivo from gaseous exchange. JAppl Physiol 1983;55:628–34.
[19] Froy O. Metabolism and circadian rhythms—implications for obesity. Endocr Rev2010;31:1–24.
[20] Fulton S, Woodside B, Shizgal P. Modulation of brain reward circuitry by leptin. Sci-ence 2000;287:125–8.
[21] Guyenet SJ, Schwartz MW. Regulation of food intake, energy balance, and body fatmass: implications for the pathogenesis and treatment of obesity. J Clin EndocrinolMetab 2012;97:745–55.
[22] Hill AJ, Blundell JE. Nutrients and behavior: research strategies for the investigationof taste characteristics, food preferences, hunger sensations and eating patterns inman. J Psychiatr Res 1982;17(2):203–12.
[23] Kakolewski JW, Deaux E, Christiansen J, Case B. Diurnal patterns in water and foodintake and body weight changes in rats with hypothalamic lesions. Am J Physiol1971;221(3):711–8.
[24] King NA, Burley VJ, Blundell JE. Exercise-induced suppression of appetite: effectson food intake and implications for energy balance. Eur J Clin Nutr 1994;48:715–24.
[25] Kreymann B, Williams G, Ghatei MA, Bloom SR. Glucagon-like peptide-1 7-36: aphysiological incretin in man. Lancet 1987;2:1300–4.
[27] Le Magnen J. Peripheral and systemic actions of food in the caloric regulation of in-take. Ann NY Acad Sci 1969;157(2):1126–57.
[28] Levitsky DA. The non-regulation of food intake in humans. Hope of reversing the ep-idemic of obesity. Physiol Behav 2005;86:623–32.
[29] Maywood ES, O'Neill JS, Chesham JE, Hastings MH. Minireview: the circadian clock-work of the suprachiasmatic nuclei—analysis of a cellular oscillator that drives endo-crine rhythms. Endocrinology 2007;148:5624–34.
[30] Morgan CT. Physiological psychology. New York: McGraw-Hill; 1943.[31] Morton GJ, Cummings DE, Baskin DG, Barsh GS, Schwartz MW. Central nervous sys-
tem control of food intake and body weight. Nature 2006;443(7109):289–95.[32] Nedeltcheva AV, Kilkus JM, Imperial J, Schoeller DA, Penev PD. Insufficient sleep un-
dermines dietary efforts to reduce adiposity. Ann Intern Med 2010;153:435–41.[33] Rosenbaum M, Goldsmith R, Bloomfield D, Magnano A, Weimer L, Heymsfield
S, et al. Low-dose leptin reverses skeletal muscle, autonomic, and neuroendo-crine adaptations to maintenance of reduced weight. J Clin Invest 2005;115:3579–86.
[34] Rosenbaum M, Sy M, Pavlovich K, Leibel RL, Hirsch J. Leptin reverses weightloss-induced changes in regional neural activity responses to visual food stimuli. JClin Invest 2008;118:2583–91.
[35] Seimon RV, Lange K, Little TJ, Brennan IM, Pilochiewicz AM, Feltrin KL, et al.Pooled-data analysis identifies pyloric pressures and plasma cholecystokinin con-centrations as major determinants of acute energy intake in healthy, lean men.Am J Clin Nutr 2010;92:61–8.
[36] Sipols AJ, Baskin DG, Schwartz MW. Effects of intracerebroventricular insulin infu-sion on diabetic hyperphagia and hypothalamic neuropeptide gene expression. Dia-betes 1995;44(2):147–51.
[37] Stellar E. The physiology of motivation. Psychol Rev 1954;61:5–22.[38] Stephan FK, Zucker I. Circadian rhythms in drinking behavior and locomotor ac-
tivity of rats are eliminated by hypothalamic lesions. Proc Natl Acad Sci U S A1972;69:1583–6.
[39] Strobele N, De Castro JM. Effect of ambience on food intake and food choice. Nutri-tion 2004;20(9):821–38.
[40] Van Cauter E, Spiegel K, Tasali E, Leproult R. Metabolic consequences of sleep andsleep loss. Sleep Med 2008;9(S1):S523–8.
[41] Verger P, Lanteaume MT, Louis-Sylvestre J. Human intake and choice of foods at in-tervals after exercise. Appetite 1992;18:93–9.
[42] Woods SC. The control of food intake: behavioral versus molecular perspectives. CellMetab 2009;9(6):489–98.
[43] Wortley KE, Anderson KD, Garcia K, Murray JD, Malinova L, Lin R, et al. Genetic de-letion of ghrelin does not decrease food intake but influences metabolic fuel prefer-ence. Proc Natl Acad Sci U S A 2004;101:8227–32.
[44] Wren AM, Seal LJ, Cohen MA, Brynes AE, Frost GS, Murphy KG, et al. Ghrelin en-hances appetite and increases food intake in humans. J Clin Endocrinol Metabol2001;86:5992–5.
[45] Yeomans MR, Blundell JE, Leshem M. Palatability: response to nutritional need orneed-free stimulation of appetite? Br J Nutr 2004;92(S1):S3–S14.
