People who do not get enough sleep often start to favor sweet and fatty foods, which contributes to weight gain. While the exact mechanisms are still unknown, lack of sleep seems to change food preferences by influencing the levels of molecules that regulate food intake. In particular, it could have an effect on the endocannabinoid system, a complex network of molecules in the nervous system that controls biological processes such as appetite.

The sense of smell is also tightly linked to how and what organisms choose to eat. Recent experiments indicate that in rodents, endocannabinoids enhance food intake by influencing the activity of the brain areas that process odors. However, it is still unclear whether the brain regions that process odors play a similar role in humans.

To investigate, Bhutani et al. examined the impact of a four-hour night’s sleep on 25 healthy human volunteers. Blood analyses showed that after a short night, individuals had increased amounts of 2-oleoylglycerol, a molecule that is part of the endocannabinoid system. When sleep-deprived people were given the choice to eat whatever they wanted, those with greater levels of 2-oleoylglycerol preferred food higher in energy. Bhutani et al. also imaged the volunteers’ brains to examine whether these changes were connected to modifications in the way the brain processed smells. This revealed that, in people who did not sleep enough, an odor-processing region called the piriform cortex was encoding smells more strongly.

The piriform cortex is connected to another region, the insula, which integrates information about the state of the body to control food intake. Lack of sleep altered this connection, and this was associated with a preference for high-energy food. In addition, further analysis showed that changes in the amounts of 2-oleoylglycerol were linked to modifications in the connection between the two brain areas. Taken together, these results suggest that sleep deprivation influences the endocannabinoid system, which in turn alters the connection between piriform and insular cortex, leading to a shift toward foods which are high in calories.

In the United States alone, one in three people sleep less than six hours a night. Learning more about how sleep deprivation affects brain pathways and food choice may help scientists to develop new drugs or behavioral therapies for conditions like obesity.

Sleep deprivation profoundly impacts food choices. When individuals are sleep-deprived, their dietary behavior shifts toward increased consumption of foods high in sugar and fat, leading to weight gain (Markwald et al., 2013; Nedeltcheva et al., 2009). These effects on ingestive behavior are likely related to sleep-dependent changes in appetite-regulating compounds, including ghrelin (Rihm et al., 2019; Spiegel et al., 2004b), leptin (Spiegel et al., 2004a), and endocannabinoids (Hanlon et al., 2016). Indeed, the endocannabinoid system (ECS) exerts strong effects on food intake (Bellocchio et al., 2010; Di Marzo et al., 2001), and levels of the endocannabinoid 2-arachidonoylglycerol (2-AG) and its structural analog 2-oleoylglycerol (2-OG) are enhanced in sleep-deprived individuals (Hanlon et al., 2016). While previous studies have tested the effects of sleep deprivation on the human brain (Greer et al., 2013; Krause et al., 2017; Muto et al., 2016; Rihm et al., 2019), the neural pathways through which sleep-dependent alterations in the ECS influence food intake have not been investigated in humans.

One likely target for sleep-dependent neuromodulation of food intake is the olfactory system. Odors serve as powerful signals for the initiation and termination of feeding behavior (Saper et al., 2002; Shepherd, 2006), and animal studies have shown that olfactory processing is modulated in a state-dependent manner (Julliard et al., 2007; McIntyre et al., 2017; Murakami et al., 2005). In rodents (Aimé et al., 2007; Aimé et al., 2014) and humans (Hanci and Altun, 2016; Stafford and Welbeck, 2011), olfaction is altered by hunger and satiety, and satiety reduces neural activity in olfactory brain regions in parallel with a suppression of feeding behavior (Boesveldt, 2017; Gervais and Pager, 1979; O'Doherty et al., 2000; Prud'homme et al., 2009; Soria-Gómez et al., 2014). Moreover, recent work across different species suggests a link between the ECS, olfactory processing, and food intake, such that endocannabinoids may directly modulate neural activity in olfactory circuits (Breunig et al., 2010; Soria-Gómez et al., 2014). However, whether odor-evoked responses in the human olfactory system are similarly modulated by the ECS, and whether this accounts for the effects of sleep deprivation on food intake, is not known.

We hypothesized that sleep deprivation is associated with a cascade of metabolic and olfactory changes, ultimately steering food choices toward energy-dense options (Simon et al., 2015). We predicted that after a night of restricted sleep, relative levels of circulating ECS compounds will be increased (Hanlon et al., 2016), leading to changes in how olfactory brain regions in the medial temporal and basal frontal lobes respond to food odors (Soria-Gómez et al., 2014). We expected that such sleep-dependent changes in olfactory processing would manifest in odor-evoked activity patterns in piriform cortex (Howard and Gottfried, 2014; Howard et al., 2009), and that effects on food intake would involve interactions with areas downstream of piriform cortex, such as the insula. Olfactory, gustatory, homeostatic, and visceral signals are integrated in the insula (Craig, 2002; de Araujo et al., 2003; Johnson et al., 2000; Livneh et al., 2017; Small et al., 2008), optimally positioning this region to regulate ingestive behavior in a state-dependent manner (Dagher, 2012; de Araujo et al., 2006).

Clinical and epidemiological studies have linked reduced sleep to elevated food intake and weight gain (Kant and Graubard, 2014; Markwald et al., 2013; Patel and Hu, 2008). This relationship has been confirmed in studies using experimentally induced sleep deprivation, demonstrating that sleep restriction increases the desire for foods high in sugar and fat content (Cain et al., 2015; Greer et al., 2013; Hogenkamp et al., 2013; Simon et al., 2015), and leads to excessive consumption of such food options (Brondel et al., 2010; Nedeltcheva et al., 2009). Several different factors have been proposed to account for this relationship (Patel and Hu, 2008), including changes in hunger induced by appetite-regulating hormones such as ghrelin and leptin (Spiegel et al., 2004a; Spiegel et al., 2004b), and the ECS (Hanlon et al., 2016). In addition, previous imaging studies have reported sleep-dependent activity changes in response to food cues (Benedict et al., 2012; Greer et al., 2013; Rihm et al., 2019; St-Onge et al., 2014), but whether and how these neural changes are related to actual food choices and the ECS has remained unclear.

In the current study, we tested the hypothesis that central olfactory mechanisms, in conjunction with the ECS, play a role in mediating the effects of sleep deprivation on dietary choices. This hypothesis was based on previous findings that experimentally induced sleep deprivation elevates relative levels of ECS compounds (Hanlon et al., 2016), and animal work suggesting that ECS activity drives changes in food intake through modulatory effects on olfactory circuits (Breunig et al., 2010; Soria-Gómez et al., 2014; Wang et al., 2012). In line with this idea, we found that sleep deprivation increased consumption of energy-dense foods, proportional to relative increases in 2-OG, and that it enhanced pattern-based encoding of odor information in the piriform cortex. Finally, we found that the effects of the ECS on food intake were mediated by changes in connectivity between the piriform cortex and the insula.

Previous animal studies have established an important role for the ECS in regulating feeding behavior (Bellocchio et al., 2010; Di Marzo et al., 2001; Rodríguez de Fonseca et al., 2001). More recently, it has been shown that levels of endocannabinoids in humans are elevated after sleep restriction (Hanlon et al., 2016), suggesting that this may drive altered dietary choices. However, unlike previous studies (Hanlon et al., 2016), we did not find a significant increase in 2-AG that paralleled increases in 2-OG. This may be due to the shorter duration of sleep restriction used in our study, but may also indicate different roles of the two compounds. Whereas 2-AG stimulates food intake and lipogenesis by activating CB1 receptors (DiPatrizio and Simansky, 2008; Osei-Hyiaman et al., 2005), the exact role of 2-OG and its relation to 2-AG is not fully understood (Murataeva et al., 2016). Our study shows that sleep-dependent increases in 2-OG are associated with changes in the consumption of energy-dense foods, providing novel evidence for a link between sleep, ECS, and dietary behavior.

We found that sleep restriction induced qualitative changes in food intake, biasing choices toward energy-dense options, without altering total calorie intake. Although some studies have shown increases in calorie intake with sleep deprivation (Al Khatib et al., 2017; Broussard et al., 2016; Markwald et al., 2013; Patel and Hu, 2008), our findings are in line with several other studies (Cain et al., 2015; Nedeltcheva et al., 2009; Simon et al., 2015) and suggest that sleep deprivation induces nuanced changes in food-based decision making, rather than simply increasing hunger or the motivation to eat.

Our results further elaborate on the effects of sleep deprivation on the human brain, suggesting that neural processing of odors is enhanced in primary olfactory brain areas after sleep restriction. Although decoding accuracy can be influenced by factors other than the strength of neural encoding, such as reduced variability (noise), our results indicate that encoding of food vs. non-food odors was more robust in the piriform cortex in a sleep-deprived state. In theory, such enhanced encoding of olfactory information could facilitate odor-evoked approach and consummatory responses (Aimé et al., 2007; Soria-Gómez et al., 2014). However, we did not observe a direct correlation between encoding in piriform cortex and food intake, and changes in odor information were not directly related to changes in circulating levels of 2-OG. It is possible that nonlinear effects or interactions among different ECS compounds (Ho et al., 2008; Murataeva et al., 2016) may have obscured a direct linear relationship between 2-OG and encoding in piriform cortex. In any case, our findings indicate that sleep-dependent increases in odor information may not directly mediate the relationship between the ECS and food intake. Instead, they suggest that interactions between olfactory cortex and downstream areas may translate altered chemosensory encoding into changes in food intake.

We found that changes in piriform-insula connectivity were correlated with the effects of sleep deprivation on food choices, suggesting a relationship between sleep-dependent food intake and neural processing in extended olfactory pathways. A formal mediation analysis showed that relative increases in 2-OG were related to reductions in odor-evoked piriform-insula connectivity, which in turn was correlated with increased choices of energy-dense food options. Although these findings need to be confirmed in future studies, they suggest a broader role for neural processing of chemosensory signals along piriform-insula pathways in the regulation of food intake. There are several possible ways by which reduced piriform-insula connectivity could promote choices of energy-dense foods in the presence of enhanced odor information in piriform cortex. Previous studies show that sensory and visceral-homeostatic signals are integrated in the insula, and that this integration is critical for guiding food intake (de Araujo et al., 2003; Livneh et al., 2017; Small, 2012). Reduced piriform-insula connectivity could reflect a diminished integration of chemosensory and homeostatic-visceral information, and a failure to adequately integrate elevated olfactory signals with homeostatic information may drive excess intake of energy-dense food. Alternatively, reduced piriform-insula connectivity could indicate an aberrant assignment of value to energy-dense foods (Balleine and Dickinson, 2000; Gottfried et al., 2003). Finally, to the degree that sleep deprivation decreases activity in other cortical areas (Greer et al., 2013; Muto et al., 2016), it is possible that reduced piriform-insula connectivity is related to diminished top-down control over elevated olfactory signals in piriform cortex, which may promote impulsive behavior in response to energy-dense food (Cedernaes et al., 2014; Krause et al., 2017).

In the current study, we compared brain responses to food odors with high palatability and non-food items with low palatability. As expected, the food odors were rated as higher in pleasantness than non-food odors. In principle, it is therefore possible that our observed effects for food vs. non-food odor encoding in the brain were fully explained by this corresponding pleasantness difference. However, our results remained significant when including pleasantness as a covariate in the statistical models, indicating that in this case pleasantness does not account for our results.

Taken together, our findings show that sleep-dependent changes in food choices are associated with changes in an olfactory pathway that is related to the ECS. This pathway is likely not restricted to sleep-dependent changes in food intake but may also account for dietary decisions more generally. In this regard, our current findings may help to guide the identification of novel targets for treatments of obesity.

All data generated or analysed during this study are included in the manuscript, supporting files and source data files.

1. 1. Aimé P
   2. Duchamp-Viret P
   3. Chaput MA
   4. Savigner A
   5. Mahfouz M
   6. Julliard AK

   (2007) Fasting increases and satiation decreases olfactory detection for a neutral odor in rats

   Behavioural Brain Research 179:258–264.

   https://doi.org/10.1016/j.bbr.2007.02.012

   • PubMed
   • Google Scholar
2. 1. Aimé P
   2. Palouzier-Paulignan B
   3. Salem R
   4. Al Koborssy D
   5. Garcia S
   6. Duchamp C
   7. Romestaing C
   8. Julliard AK

   (2014) Modulation of olfactory sensitivity and glucose-sensing by the feeding state in obese zucker rats

   Frontiers in Behavioral Neuroscience 8:326.

   https://doi.org/10.3389/fnbeh.2014.00326

   • PubMed
   • Google Scholar
3. 1. Al Khatib HK
   2. Harding SV
   3. Darzi J
   4. Pot GK

   (2017) The effects of partial sleep deprivation on energy balance: a systematic review and meta-analysis

   European Journal of Clinical Nutrition 71:614–624.

   https://doi.org/10.1038/ejcn.2016.201

   • PubMed
   • Google Scholar
4. 1. Ancoli-Israel S
   2. Cole R
   3. Alessi C
   4. Chambers M
   5. Moorcroft W
   6. Pollak CP

   (2003) The role of actigraphy in the study of sleep and circadian rhythms

   Sleep 26:342–392.

   https://doi.org/10.1093/sleep/26.3.342

   • PubMed
   • Google Scholar
5. 1. Balleine BW
   2. Dickinson A

   (2000) The effect of lesions of the insular cortex on instrumental conditioning: evidence for a role in incentive memory

   The Journal of Neuroscience 20:8954–8964.

   https://doi.org/10.1523/JNEUROSCI.20-23-08954.2000

   • PubMed
   • Google Scholar
6. 1. Barnes DC
   2. Hofacer RD
   3. Zaman AR
   4. Rennaker RL
   5. Wilson DA

   (2008) Olfactory perceptual stability and discrimination

   Nature Neuroscience 11:1378–1380.

   https://doi.org/10.1038/nn.2217

   • PubMed
   • Google Scholar
7. 1. Baron RM
   2. Kenny DA

   (1986) The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations

   Journal of Personality and Social Psychology 51:1173–1182.

   https://doi.org/10.1037/0022-3514.51.6.1173

   • PubMed
   • Google Scholar
8. 1. Bellocchio L
   2. Lafenêtre P
   3. Cannich A
   4. Cota D
   5. Puente N
   6. Grandes P
   7. Chaouloff F
   8. Piazza PV
   9. Marsicano G

   (2010) Bimodal control of stimulated food intake by the endocannabinoid system

   Nature Neuroscience 13:281–283.

   https://doi.org/10.1038/nn.2494

   • PubMed
   • Google Scholar
9. 1. Benedict C
   2. Brooks SJ
   3. O'Daly OG
   4. Almèn MS
   5. Morell A
   6. Åberg K
   7. Gingnell M
   8. Schultes B
   9. Hallschmid M
   10. Broman JE
   11. Larsson EM
   12. Schiöth HB

   (2012) Acute sleep deprivation enhances the brain's response to hedonic food stimuli: an fMRI study

   The Journal of Clinical Endocrinology & Metabolism 97:E443–E447.

   https://doi.org/10.1210/jc.2011-2759

   • PubMed
   • Google Scholar
10. 1. Birn RM
    2. Smith MA
    3. Jones TB
    4. Bandettini PA

    (2008) The respiration response function: the temporal dynamics of fMRI signal fluctuations related to changes in respiration

    NeuroImage 40:644–654.

    https://doi.org/10.1016/j.neuroimage.2007.11.059

    • PubMed
    • Google Scholar
11. Book

    1. Boesveldt S

    (2017) Olfaction and Eating Behavior

    In: Buettner A, editors. Springer Handbook of Odor. Springer. pp. 109–110.

    https://doi.org/10.1007/978-3-319-26932-0_44

    • Google Scholar
12. 1. Breunig E
    2. Manzini I
    3. Piscitelli F
    4. Gutermann B
    5. Di Marzo V
    6. Schild D
    7. Czesnik D

    (2010) The endocannabinoid 2-arachidonoyl-glycerol controls odor sensitivity in larvae of Xenopus laevis

    Journal of Neuroscience 30:8965–8973.

    https://doi.org/10.1523/JNEUROSCI.4030-09.2010

    • PubMed
    • Google Scholar
13. 1. Brondel L
    2. Romer MA
    3. Nougues PM
    4. Touyarou P
    5. Davenne D

    (2010) Acute partial sleep deprivation increases food intake in healthy men

    The American Journal of Clinical Nutrition 91:1550–1559.

    https://doi.org/10.3945/ajcn.2009.28523

    • PubMed
    • Google Scholar
14. 1. Broussard JL
    2. Kilkus JM
    3. Delebecque F
    4. Abraham V
    5. Day A
    6. Whitmore HR
    7. Tasali E

    (2016) Elevated ghrelin predicts food intake during experimental sleep restriction

    Obesity 24:132–138.

    https://doi.org/10.1002/oby.21321

    • PubMed
    • Google Scholar
15. 1. Burgess HJ
    2. Eastman CI

    (2004) Early versus late bedtimes phase shift the human dim light melatonin rhythm despite a fixed morning lights on time

    Neuroscience Letters 356:115–118.

    https://doi.org/10.1016/j.neulet.2003.11.032

    • PubMed
    • Google Scholar
16. 1. Buysse DJ
    2. Reynolds CF
    3. Monk TH
    4. Berman SR
    5. Kupfer DJ

    (1989) The Pittsburgh sleep quality index: a new instrument for psychiatric practice and research

    Psychiatry Research 28:193–213.

    https://doi.org/10.1016/0165-1781(89)90047-4

    • PubMed
    • Google Scholar
17. 1. Cain SW
    2. Filtness AJ
    3. Phillips CL
    4. Anderson C

    (2015) Enhanced preference for high-fat foods following a simulated night shift

    Scandinavian Journal of Work, Environment & Health 41:288–293.

    https://doi.org/10.5271/sjweh.3486

    • PubMed
    • Google Scholar
18. 1. Cedernaes J
    2. Brandell J
    3. Ros O
    4. Broman JE
    5. Hogenkamp PS
    6. Schiöth HB
    7. Benedict C

    (2014) Increased impulsivity in response to food cues after sleep loss in healthy young men

    Obesity 22:1786–1791.

    https://doi.org/10.1002/oby.20786

    • PubMed
    • Google Scholar
19. 1. Chang CC
    2. Lin CJ

    (2011) LIBSVM: a library for support vector machines

    ACM Transactions on Intelligent Systems and Technology 2:27:1–27:27.

    https://doi.org/10.1145/1961189.1961199

    • Google Scholar
20. 1. Craig AD

    (2002) How do you feel? interoception: the sense of the physiological condition of the body

    Nature Reviews Neuroscience 3:655–666.

    https://doi.org/10.1038/nrn894

    • PubMed
    • Google Scholar
21. 1. Dagher A

    (2012) Functional brain imaging of appetite

    Trends in Endocrinology & Metabolism 23:250–260.

    https://doi.org/10.1016/j.tem.2012.02.009

    • PubMed
    • Google Scholar
22. 1. de Araujo IE
    2. Rolls ET
    3. Kringelbach ML
    4. McGlone F
    5. Phillips N

    (2003) Taste-olfactory convergence, and the representation of the pleasantness of flavour, in the human brain

    European Journal of Neuroscience 18:2059–2068.

    https://doi.org/10.1046/j.1460-9568.2003.02915.x

    • PubMed
    • Google Scholar
23. 1. de Araujo IE
    2. Gutierrez R
    3. Oliveira-Maia AJ
    4. Pereira A
    5. Nicolelis MA
    6. Simon SA

    (2006) Neural ensemble coding of satiety states

    Neuron 51:483–494.

    https://doi.org/10.1016/j.neuron.2006.07.009

    • PubMed
    • Google Scholar
24. 1. Deichmann R
    2. Gottfried JA
    3. Hutton C
    4. Turner R

    (2003) Optimized EPI for fMRI studies of the orbitofrontal cortex

    NeuroImage 19:430–441.

    https://doi.org/10.1016/S1053-8119(03)00073-9

    • PubMed
    • Google Scholar
25. 1. Di Marzo V
    2. Goparaju SK
    3. Wang L
    4. Liu J
    5. Bátkai S
    6. Járai Z
    7. Fezza F
    8. Miura GI
    9. Palmiter RD
    10. Sugiura T
    11. Kunos G

    (2001) Leptin-regulated endocannabinoids are involved in maintaining food intake

    Nature 410:822–825.

    https://doi.org/10.1038/35071088

    • PubMed
    • Google Scholar
26. 1. DiPatrizio NV
    2. Simansky KJ

    (2008) Activating parabrachial cannabinoid CB1 receptors selectively stimulates feeding of palatable foods in rats

    Journal of Neuroscience 28:9702–9709.

    https://doi.org/10.1523/JNEUROSCI.1171-08.2008

    • PubMed
    • Google Scholar
27. 1. Gervais R
    2. Pager J

    (1979) Combined modulating effects of the general arousal and the specific hunger arousal on the olfactory bulb responses in the rat

    Electroencephalography and Clinical Neurophysiology 46:87–94.

    https://doi.org/10.1016/0013-4694(79)90053-1

    • PubMed
    • Google Scholar
28. 1. Gottfried JA
    2. O'Doherty J
    3. Dolan RJ

    (2003) Encoding predictive reward value in human amygdala and orbitofrontal cortex

    Science 301:1104–1107.

    https://doi.org/10.1126/science.1087919

    • PubMed
    • Google Scholar
29. 1. Greer SM
    2. Goldstein AN
    3. Walker MP

    (2013) The impact of sleep deprivation on food desire in the human brain

    Nature Communications 4:2259.

    https://doi.org/10.1038/ncomms3259

    • PubMed
    • Google Scholar
30. 1. Hanci D
    2. Altun H

    (2016) Hunger state affects both olfactory abilities and gustatory sensitivity

    European Archives of Oto-Rhino-Laryngology 273:1637–1641.

    https://doi.org/10.1007/s00405-015-3589-6

    • PubMed
    • Google Scholar
31. 1. Hanlon EC
    2. Tasali E
    3. Leproult R
    4. Stuhr KL
    5. Doncheck E
    6. de Wit H
    7. Hillard CJ
    8. Van Cauter E

    (2016) Sleep restriction enhances the daily rhythm of circulating levels of endocannabinoid 2-Arachidonoylglycerol

    Sleep 39:653–664.

    https://doi.org/10.5665/sleep.5546

    • PubMed
    • Google Scholar
32. 1. Haynes JD
    2. Sakai K
    3. Rees G
    4. Gilbert S
    5. Frith C
    6. Passingham RE

    (2007) Reading hidden intentions in the human brain

    Current Biology 17:323–328.

    https://doi.org/10.1016/j.cub.2006.11.072

    • PubMed
    • Google Scholar
33. 1. Hebart MN
    2. Görgen K
    3. Haynes JD

    (2014) The decoding toolbox (TDT): a versatile software package for multivariate analyses of functional imaging data

    Frontiers in Neuroinformatics 8:88.

    https://doi.org/10.3389/fninf.2014.00088

    • PubMed
    • Google Scholar
34. 1. Hirshkowitz M
    2. Whiton K
    3. Albert SM
    4. Alessi C
    5. Bruni O
    6. DonCarlos L
    7. Hazen N
    8. Herman J
    9. Katz ES
    10. Kheirandish-Gozal L
    11. Neubauer DN
    12. O'Donnell AE
    13. Ohayon M
    14. Peever J
    15. Rawding R
    16. Sachdeva RC
    17. Setters B
    18. Vitiello MV
    19. Ware JC
    20. Adams Hillard PJ

    (2015) National sleep foundation's sleep time duration recommendations: methodology and results summary

    Sleep Health 1:40–43.

    https://doi.org/10.1016/j.sleh.2014.12.010

    • PubMed
    • Google Scholar
35. 1. Ho WS
    2. Barrett DA
    3. Randall MD

    (2008) 'Entourage' effects of N-palmitoylethanolamide and N-oleoylethanolamide on vasorelaxation to anandamide occur through TRPV1 receptors

    British Journal of Pharmacology 155:837–846.

    https://doi.org/10.1038/bjp.2008.324

    • PubMed
    • Google Scholar
36. 1. Hogenkamp PS
    2. Nilsson E
    3. Nilsson VC
    4. Chapman CD
    5. Vogel H
    6. Lundberg LS
    7. Zarei S
    8. Cedernaes J
    9. Rångtell FH
    10. Broman JE
    11. Dickson SL
    12. Brunstrom JM
    13. Benedict C
    14. Schiöth HB

    (2013) Acute sleep deprivation increases portion size and affects food choice in young men

    Psychoneuroendocrinology 38:1668–1674.

    https://doi.org/10.1016/j.psyneuen.2013.01.012

    • PubMed
    • Google Scholar
37. 1. Howard JD
    2. Plailly J
    3. Grueschow M
    4. Haynes JD
    5. Gottfried JA

    (2009) Odor quality coding and categorization in human posterior piriform cortex

    Nature Neuroscience 12:932–938.

    https://doi.org/10.1038/nn.2324

    • PubMed
    • Google Scholar
38. 1. Howard JD
    2. Gottfried JA

    (2014) Configural and elemental coding of natural odor mixture components in the human brain

    Neuron 84:857–869.

    https://doi.org/10.1016/j.neuron.2014.10.012

    • PubMed
    • Google Scholar
39. 1. Howard JD
    2. Kahnt T

    (2017) Identity-Specific reward representations in orbitofrontal cortex are modulated by selective devaluation

    The Journal of Neuroscience 37:2627–2638.

    https://doi.org/10.1523/JNEUROSCI.3473-16.2017

    • PubMed
    • Google Scholar
40. 1. Howard JD
    2. Kahnt T

    (2018) Identity prediction errors in the human midbrain update reward-identity expectations in the orbitofrontal cortex

    Nature Communications 9:1611.

    https://doi.org/10.1038/s41467-018-04055-5

    • PubMed
    • Google Scholar
41. 1. Illig KR
    2. Haberly LB

    (2003) Odor-evoked activity is spatially distributed in piriform cortex

    The Journal of Comparative Neurology 457:361–373.

    https://doi.org/10.1002/cne.10557

    • PubMed
    • Google Scholar
42. 1. Johnson DM
    2. Illig KR
    3. Behan M
    4. Haberly LB

    (2000) New features of connectivity in piriform cortex visualized by intracellular injection of pyramidal cells suggest that "primary" olfactory cortex functions like "association" cortex in other sensory systems

    The Journal of Neuroscience 20:6974–6982.

    https://doi.org/10.1523/JNEUROSCI.20-18-06974.2000

    • PubMed
    • Google Scholar
43. 1. Julliard AK
    2. Chaput MA
    3. Apelbaum A
    4. Aimé P
    5. Mahfouz M
    6. Duchamp-Viret P

    (2007) Changes in rat olfactory detection performance induced by orexin and leptin mimicking fasting and satiation

    Behavioural Brain Research 183:123–129.

    https://doi.org/10.1016/j.bbr.2007.05.033

    • PubMed
    • Google Scholar
44. 1. Kahnt T
    2. Grueschow M
    3. Speck O
    4. Haynes JD

    (2011) Perceptual learning and decision-making in human medial frontal cortex

    Neuron 70:549–559.

    https://doi.org/10.1016/j.neuron.2011.02.054

    • PubMed
    • Google Scholar
45. 1. Kant AK
    2. Graubard BI

    (2014) Association of self-reported sleep duration with eating behaviors of american adults: nhanes 2005-2010

    The American Journal of Clinical Nutrition 100:938–947.

    https://doi.org/10.3945/ajcn.114.085191

    • PubMed
    • Google Scholar
46. 1. Krause AJ
    2. Simon EB
    3. Mander BA
    4. Greer SM
    5. Saletin JM
    6. Goldstein-Piekarski AN
    7. Walker MP

    (2017) The sleep-deprived human brain

    Nature Reviews Neuroscience 18:404–418.

    https://doi.org/10.1038/nrn.2017.55

    • PubMed
    • Google Scholar
47. 1. Kriegeskorte N
    2. Goebel R
    3. Bandettini P

    (2006) Information-based functional brain mapping

    PNAS 103:3863–3868.

    https://doi.org/10.1073/pnas.0600244103

    • PubMed
    • Google Scholar
48. 1. Livneh Y
    2. Ramesh RN
    3. Burgess CR
    4. Levandowski KM
    5. Madara JC
    6. Fenselau H
    7. Goldey GJ
    8. Diaz VE
    9. Jikomes N
    10. Resch JM
    11. Lowell BB
    12. Andermann ML

    (2017) Homeostatic circuits selectively gate food cue responses in insular cortex

    Nature 546:611–616.

    https://doi.org/10.1038/nature22375

    • PubMed
    • Google Scholar
49. 1. Maier SU
    2. Makwana AB
    3. Hare TA

    (2015) Acute stress impairs Self-Control in Goal-Directed choice by altering multiple functional connections within the brain's Decision Circuits

    Neuron 87:621–631.

    https://doi.org/10.1016/j.neuron.2015.07.005

    • PubMed
    • Google Scholar
50. 1. Marino M
    2. Li Y
    3. Rueschman MN
    4. Winkelman JW
    5. Ellenbogen JM
    6. Solet JM
    7. Dulin H
    8. Berkman LF
    9. Buxton OM

    (2013) Measuring sleep: accuracy, sensitivity, and specificity of wrist actigraphy compared to polysomnography

    Sleep 36:1747–1755.

    https://doi.org/10.5665/sleep.3142

    • PubMed
    • Google Scholar
51. 1. Markwald RR
    2. Melanson EL
    3. Smith MR
    4. Higgins J
    5. Perreault L
    6. Eckel RH
    7. Wright KP

    (2013) Impact of insufficient sleep on total daily energy expenditure, food intake, and weight gain

    PNAS 110:5695–5700.

    https://doi.org/10.1073/pnas.1216951110

    • PubMed
    • Google Scholar
52. 1. McIntyre JC
    2. Thiebaud N
    3. McGann JP
    4. Komiyama T
    5. Rothermel M

    (2017) Neuromodulation in chemosensory pathways

    Chemical Senses 42:375–379.

    https://doi.org/10.1093/chemse/bjx014

    • PubMed
    • Google Scholar
53. 1. McLaren DG
    2. Ries ML
    3. Xu G
    4. Johnson SC

    (2012) A generalized form of context-dependent psychophysiological interactions (gPPI): a comparison to standard approaches

    NeuroImage 61:1277–1286.

    https://doi.org/10.1016/j.neuroimage.2012.03.068

    • PubMed
    • Google Scholar
54. 1. Murakami M
    2. Kashiwadani H
    3. Kirino Y
    4. Mori K

    (2005) State-dependent sensory gating in olfactory cortex

    Neuron 46:285–296.

    https://doi.org/10.1016/j.neuron.2005.02.025

    • PubMed
    • Google Scholar
55. 1. Murataeva N
    2. Dhopeshwarkar A
    3. Yin D
    4. Mitjavila J
    5. Bradshaw H
    6. Straiker A
    7. Mackie K

    (2016) Where's my entourage? The curious case of 2-oleoylglycerol, 2-linolenoylglycerol, and 2-palmitoylglycerol

    Pharmacological Research 110:173–180.

    https://doi.org/10.1016/j.phrs.2016.04.015

    • PubMed
    • Google Scholar
56. 1. Muto V
    2. Jaspar M
    3. Meyer C
    4. Kussé C
    5. Chellappa SL
    6. Degueldre C
    7. Balteau E
    8. Shaffii-Le Bourdiec A
    9. Luxen A
    10. Middleton B
    11. Archer SN
    12. Phillips C
    13. Collette F
    14. Vandewalle G
    15. Dijk DJ
    16. Maquet P

    (2016) Local modulation of human brain responses by circadian rhythmicity and sleep debt

    Science 353:687–690.

    https://doi.org/10.1126/science.aad2993

    • PubMed
    • Google Scholar
57. 1. Nedeltcheva AV
    2. Kilkus JM
    3. Imperial J
    4. Kasza K
    5. Schoeller DA
    6. Penev PD

    (2009) Sleep curtailment is accompanied by increased intake of calories from snacks

    The American Journal of Clinical Nutrition 89:126–133.

    https://doi.org/10.3945/ajcn.2008.26574

    • PubMed
    • Google Scholar
58. 1. O'Doherty J
    2. Rolls ET
    3. Francis S
    4. Bowtell R
    5. McGlone F
    6. Kobal G
    7. Renner B
    8. Ahne G

    (2000) Sensory-specific satiety-related olfactory activation of the human orbitofrontal cortex

    NeuroReport 11:893–897.

    https://doi.org/10.1097/00001756-200003200-00046

    • PubMed
    • Google Scholar
59. 1. Osei-Hyiaman D
    2. DePetrillo M
    3. Pacher P
    4. Liu J
    5. Radaeva S
    6. Bátkai S
    7. Harvey-White J
    8. Mackie K
    9. Offertáler L
    10. Wang L
    11. Kunos G

    (2005) Endocannabinoid activation at hepatic CB1 receptors stimulates fatty acid synthesis and contributes to diet-induced obesity

    Journal of Clinical Investigation 115:1298–1305.

    https://doi.org/10.1172/JCI200523057

    • PubMed
    • Google Scholar
60. 1. Patel S
    2. Carrier EJ
    3. Ho WS
    4. Rademacher DJ
    5. Cunningham S
    6. Reddy DS
    7. Falck JR
    8. Cravatt BF
    9. Hillard CJ

    (2005) The postmortal accumulation of brain N-arachidonylethanolamine (anandamide) is dependent upon fatty acid amide hydrolase activity

    Journal of Lipid Research 46:342–349.

    https://doi.org/10.1194/jlr.M400377-JLR200

    • PubMed
    • Google Scholar
61. 1. Patel SR
    2. Hu FB

    (2008) Short sleep duration and weight gain: a systematic review

    Obesity 16:643–653.

    https://doi.org/10.1038/oby.2007.118

    • PubMed
    • Google Scholar
62. 1. Prud'homme MJ
    2. Lacroix MC
    3. Badonnel K
    4. Gougis S
    5. Baly C
    6. Salesse R
    7. Caillol M

    (2009) Nutritional status modulates behavioural and olfactory bulb fos responses to isoamyl acetate or food odour in rats: roles of orexins and leptin

    Neuroscience 162:1287–1298.

    https://doi.org/10.1016/j.neuroscience.2009.05.043

    • PubMed
    • Google Scholar
63. 1. Radloff LS

    (1977) The CES-D scale. A Self-Report depression scale for research in the general population

    Applied Psychological Measurement 1:385–401.

    https://doi.org/10.1177/014662167700100306

    • Google Scholar
64. 1. Rihm JS
    2. Menz MM
    3. Schultz H
    4. Bruder L
    5. Schilbach L
    6. Schmid SM
    7. Peters J

    (2019) Sleep deprivation selectively upregulates an Amygdala-Hypothalamic circuit involved in food reward

    The Journal of Neuroscience 39:888–899.

    https://doi.org/10.1523/JNEUROSCI.0250-18.2018

    • PubMed
    • Google Scholar
65. 1. Rodríguez de Fonseca F
    2. Navarro M
    3. Gómez R
    4. Escuredo L
    5. Nava F
    6. Fu J
    7. Murillo-Rodríguez E
    8. Giuffrida A
    9. LoVerme J
    10. Gaetani S
    11. Kathuria S
    12. Gall C
    13. Piomelli D

    (2001) An anorexic lipid mediator regulated by feeding

    Nature 414:209–212.

    https://doi.org/10.1038/35102582

    • PubMed
    • Google Scholar
66. 1. Saper CB
    2. Chou TC
    3. Elmquist JK

    (2002) The need to feed: homeostatic and hedonic control of eating

    Neuron 36:199–211.

    https://doi.org/10.1016/s0896-6273(02)00969-8

    • PubMed
    • Google Scholar
67. 1. Shepherd GM

    (2006) Smell images and the flavour system in the human brain

    Nature 444:316–321.

    https://doi.org/10.1038/nature05405

    • PubMed
    • Google Scholar
68. 1. Simon SL
    2. Field J
    3. Miller LE
    4. DiFrancesco M
    5. Beebe DW

    (2015) Sweet/dessert foods are more appealing to adolescents after sleep restriction

    PLOS ONE 10:e0115434.

    https://doi.org/10.1371/journal.pone.0115434

    • PubMed
    • Google Scholar
69. 1. Small DM
    2. Veldhuizen MG
    3. Felsted J
    4. Mak YE
    5. McGlone F

    (2008) Separable substrates for anticipatory and consummatory food chemosensation

    Neuron 57:786–797.

    https://doi.org/10.1016/j.neuron.2008.01.021

    • PubMed
    • Google Scholar
70. 1. Small DM

    (2012) Flavor is in the brain

    Physiology & Behavior 107:540–552.

    https://doi.org/10.1016/j.physbeh.2012.04.011

    • PubMed
    • Google Scholar
71. 1. Soria-Gómez E
    2. Bellocchio L
    3. Reguero L
    4. Lepousez G
    5. Martin C
    6. Bendahmane M
    7. Ruehle S
    8. Remmers F
    9. Desprez T
    10. Matias I
    11. Wiesner T
    12. Cannich A
    13. Nissant A
    14. Wadleigh A
    15. Pape HC
    16. Chiarlone AP
    17. Quarta C
    18. Verrier D
    19. Vincent P
    20. Massa F
    21. Lutz B
    22. Guzmán M
    23. Gurden H
    24. Ferreira G
    25. Lledo PM
    26. Grandes P
    27. Marsicano G

    (2014) The endocannabinoid system controls food intake via olfactory processes

    Nature Neuroscience 17:407–415.

    https://doi.org/10.1038/nn.3647

    • PubMed
    • Google Scholar
72. 1. Spiegel K
    2. Leproult R
    3. L'hermite-Balériaux M
    4. Copinschi G
    5. Penev PD
    6. Van Cauter E

    (2004a) Leptin levels are dependent on sleep duration: relationships with sympathovagal balance, carbohydrate regulation, cortisol, and thyrotropin

    The Journal of Clinical Endocrinology & Metabolism 89:5762–5771.

    https://doi.org/10.1210/jc.2004-1003

    • PubMed
    • Google Scholar
73. 1. Spiegel K
    2. Tasali E
    3. Penev P
    4. Van Cauter E

    (2004b) Brief communication: sleep curtailment in healthy young men is associated with decreased leptin levels, elevated ghrelin levels, and increased hunger and appetite

    Annals of Internal Medicine 141:846–850.

    https://doi.org/10.7326/0003-4819-141-11-200412070-00008

    • PubMed
    • Google Scholar
74. 1. St-Onge M-P
    2. Wolfe S
    3. Sy M
    4. Shechter A
    5. Hirsch J

    (2014) Sleep restriction increases the neuronal response to unhealthy food in normal-weight individuals

    International Journal of Obesity 38:411–416.

    https://doi.org/10.1038/ijo.2013.114

    • Google Scholar
75. 1. Stafford LD
    2. Welbeck K

    (2011) High hunger state increases olfactory sensitivity to neutral but not food odors

    Chemical Senses 36:189–198.

    https://doi.org/10.1093/chemse/bjq114

    • PubMed
    • Google Scholar
76. 1. Stettler DD
    2. Axel R

    (2009) Representations of odor in the piriform cortex

    Neuron 63:854–864.

    https://doi.org/10.1016/j.neuron.2009.09.005

    • PubMed
    • Google Scholar
77. 1. Suarez JA
    2. Howard JD
    3. Schoenbaum G
    4. Kahnt T

    (2019) Sensory prediction errors in the human midbrain signal identity violations independent of perceptual distance

    eLife 8:e43962.

    https://doi.org/10.7554/eLife.43962

    • PubMed
    • Google Scholar
78. 1. Wang ZJ
    2. Sun L
    3. Heinbockel T

    (2012) Cannabinoid receptor-mediated regulation of neuronal activity and signaling in glomeruli of the main olfactory bulb

    Journal of Neuroscience 32:8475–8479.

    https://doi.org/10.1523/JNEUROSCI.5333-11.2012

    • PubMed
    • Google Scholar
79. 1. Weiskopf N
    2. Hutton C
    3. Josephs O
    4. Deichmann R

    (2006) Optimal EPI parameters for reduction of susceptibility-induced BOLD sensitivity losses: a whole-brain analysis at 3 T and 1.5 T

    NeuroImage 33:493–504.

    https://doi.org/10.1016/j.neuroimage.2006.07.029

    • PubMed
    • Google Scholar
80. 1. Zelano C
    2. Mohanty A
    3. Gottfried JA

    (2011) Olfactory predictive codes and stimulus templates in piriform cortex

    Neuron 72:178–187.

    https://doi.org/10.1016/j.neuron.2011.08.010

    • PubMed
    • Google Scholar

Thank you for submitting your article "Olfactory contributions to sleep-dependent food intake in humans" for consideration by eLife. Your article has been reviewed by three peer reviewers, including, including X as the Reviewing Editor and Reviewer #1 as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Christian Büchel as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Noam Sobel (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This study examines how a manipulation of sleep duration impacts both food choice and circulating endocannabinoid system (ECS) compounds, as well as how this association is mediated by altered connectivity of neural olfactory regions. Consistent with previous work they found that sleep restriction increases circulating levels of an ECS compound (2-OG) and increases food intake (measured as energy density in food selected from a buffet). Using an MVPA approach, the authors identified regions that were selective in their representations of food vs. non-food smells, with the right piriform cortex exhibiting a group difference in representational distances of food vs. non-food odors. Using this as a seed region, the authors found that individual differences in odor vs. clean air task connectivity between the piriform cortex and insula mediated the individual differences observed in ECS compounds and food choice.

All three reviewers felt that this is potentially a very important study that addresses a very important question, with immediate interest for both scientists and the general public alike. Moreover, it is clearly the product of "high level" rigorous science. More specifically, the authors tackle their question from several directions at once. This is on one side a clear advantage, yet on the other, when you do a lot of different work, there are lots of potential complications that need to be addressed before the manuscript can be accepted.

Essential revisions:

1) Imaging analysis.

Reviewer 1 pointed out that the critical finding that sets this work apart from prior studies is the observation that task-related piriform-insula connectivity differences statistically mediate a relationship between 2-OG and food choice. Yet a lot relies on the veracity of the neuroimaging analysis. Both estimates of task effects (e.g., GLM) and connectivity measures in the BOLD response are highly sensitive to motion and physiological artifacts that can sometimes lead to spurious observations. This brings up several concerns.

First, reviewer 1 is concerned that the food > non-food difference in piriform representations may be due to differences in head motion across groups. It seems possible that sleep manipulations can lead to differences in head motion in the scanner, that in turn would alter the reliability of the representational distance estimates that the SVN decoder picks up on (leading to a potential spurious group difference). The authors should both report whether there were differences in framewise displacement measures across groups and include a displacement measure as a nuisance variable in the estimation of the group differences.

Second, reviewer 1 pointed out that while there is no statistical difference in sniff amplitude (Figure 1—figure supplement 3E), there does appear to be a mean difference such that the sleep restricted group is taking lower amplitude sniffs. Even if this isn't significant at the behavioral level, it might lead to differences in the expression of respiratory artifacts in the BOLD response. In addition, there may be differences in respiration variability, both during the odor presentation and in-between, which is also known to contribute to artifacts in the BOLD signal (see Birn et al., (2008). Therefore, the authors should include models of respiration artifacts, particularly in the connectivity estimates, in order to rule out possible spurious associations due to group differences in respiration. This concern was echoed by reviewer 3, who pointed out that, they were not convinced that there were not, for instance, motoric differences in the sniff that contribute to the effects. Although Figure 1—figure supplement 3E and F show a non-significant differences, when one cuts out C and D and overlay them up against the light, it sure appears that the odors for NDS are equal to the clean air in DS. In other words, it is not clear where the amplitude and latency measures come from, but they don't seem to capture the differences you can see if you overlap the curves in C and D. One other concern in this regard is that if the piriform-insula connectivity measure is for odors (collapsed) v. clean air, then this sniff response comparison should be for odors (collapsed) v. clean air. (One note on food odors v. non-food odors: it's confusing to me that celery seed is considered a non-food odor as this is a not-uncommon ingredient in food dishes.)

Finally, all three reviewers raised concerns about the use of different contrasts used in the analyses. Reviewer 1 points out that the search for group differences in representational distances in the piriform uses the food > non-food contrast, but then the connectivity results between piriform and insula are using the odor > clean air contrast. It is likely that this is because the food > non-food contrast did not produce a significant result. This should still be reported and clarified in the text.

2) Pleasantness confounds.

Both reviewers 1 and 2 had concerns regarding the role of subjective pleasantness preferences had on the task. Reviewer 2 pointed out that a major analysis in this manuscript is the searchlight-based multi-voxel pattern analysis contrasting the response to "food" and "non-food" odors. As clearly indicated in Figure 1—figure supplement 3, "food" odors were much (as in p = 2.5x10-6) more pleasant than "non-food" odors. Thus, why is this to be considered a contrast of "food vs. non-food" and not a contrast of "pleasant vs. less-pleasant". The impact of sleep deprivation on hedonics is also interesting, but not the aim of this manuscript. This potential confound is so blatant, that the reader must be missing something. Thus, what is missing? Why is the sharp difference in pleasantness between the edible and non-edible odorants not a concern? The authors can do one of two things: Either better explain how this relates directly to food vs. non-food odors, or run analyses to address this. For example, either regress out pleasantness differences, or select subsets of data devoid of this potential confound.

Related to this, reviewer 2 pointed out that the selection of non-food odorants was slightly odd. After all, celery seed is not that far from a food. Why didn't the authors just use perfume? It would have addressed their pleasantness difference, and it's clearly not edible. Odorant selection, however, is behind us. Thus, what the authors should at least add is the actual edibility ratings of the odorants. This should be provided in a supplementary table with the associated statistics of the differences in edibility.

3) Validity of inferences.

Reviewers 2 and 3 had concerns about the certainty of the conclusions being made. Reviewer 2 pointed out that the authors conclude that "sleep deprivation induces changes in food intake through the modulation of an olfactory pathway that is related to the endocannabinoid system". This very strong statement implies causation, that it is not 100% certain the authors have in hand. There is a relationship, but the authors to make such a strong causal claim, wouldn't they need to somehow find a way to independently manipulate the olfactory pathway, and show that it is indeed responsible for the effect? Would sleep deprivation fail to impact eating behaviour of individuals with anosmia? This is a major concern, but it is of course trivial to address: The authors should either slightly tone down the claims on causation (the manuscript is strong enough as it is) or make a better substantiated claim on causation.

Reviewer 3 pointed out that the conclusions of the paper emphasize a finding that olfaction contributes to sleep-dependent food intake in humans (see Title). This is an overstatement of the results. First, unlike other papers, the present study did not find a difference in food intake (caloric intake) based on sleep deprivation. The authors did find a difference in food preference (food decisions). Second, the measure of olfaction in the result is not olfaction per se but rather piriform-insula connectivity when sleep deprived v. not sleep deprived. This should be made more clear and up front throughout the manuscript.

4) 2-OG vs. food choice association.

All three reviewers had a concern about the association between 2-OG and food choice. An inspection of Figure 2B suggests that the association between ECS system and food choice may be driven by 5 participants. Reviewers 2 and 3 pointed out that if these outliers are removed, the association appears to become negative. The authors seem to be aware of this since they used a robust regression analysis; however, robust regression approaches simply account for differences in variance of each observation in the overall estimate. Reviewer 1 recommend a non-parametric statistic, such as a bootstrap or permutation test. (Note: This is listed as a Major concern because this is one of the critical observations for the authors' primary conclusions).

5) Compliance.

Reviewer 3 raised concerns about subject compliance. There is a lot of trust in participants. Trust that they did not consume additional food and drinks, trust they did not take caffeine, and trust in both the participant and actigraph that the sleep restriction instructions were complied with. Given the importance of these, it is surprising that participants were not kept in-lab for the sleep deprivation manipulation. The authors should address this.

6) Sniffing vs. inhalation.

Reviewer 2 raised a concern about the nature of the inhalation measure identified as sniffing. This reviewer points out that authors don't directly measure sniffing; they measure respiration with respiratory belts. However, the path from thoracic and abdominal movements to airflow patterns in the nose is complex, and more critically, it is variable. In fact, even in direction; an expanding abdomen can reflect inhalation and it can reflect exhalation. Moreover, belts don't discriminate between nasal and oral inhalation. What if a participant didn't like a particular odorant, and shifted to oral inhalation every time it was presented? How would the authors know this?

Moreover, as the authors know, sniffing is odorant dependent. Thus, the pleasant food odors may have been greeted with slightly more vigorous sniffs, yet the less pleasant non-food odors with slightly less vigorous sniffs. If this difference was very small, it would not be picked up by the belts. Such a small difference, however, may also be very consistent, and thus meaningful. All this is critical, because the authors may attribute an activation pattern to a difference in odor, where in fact it may reflect a difference in sniffing. Alas, these are indeed often very difficult to untangle, but this still needs to be addressed. This is especially true in light of the simplicity of precisely measuring sniffing at the nose. If the olfactometer uses a mask, then a simple pressure tube off the mask accurately converts sniffing. If no mask is used, then a nasal cannula provides an even better measure of sniffing. As sniffing is very easy to measure, it is not clear why the authors did not do so.

As to the current manuscript, the authors should use more careful terminology. In Figure 1—figure supplement 3, the authors should refer to respiratory amplitude, not sniff amplitude. Reviewer 2 points out that this is a problem with the entire field, but should be reasonably addressed here. In future, reviewer 2 suggests the authors precisely measure sniffing in their studies. It's cheap, easy, and informative.

Summary:

This study examines how a manipulation of sleep duration impacts both food choice and circulating endocannabinoid system (ECS) compounds, as well as how this association is mediated by altered connectivity of neural olfactory regions. […] More specifically, the authors tackle their question from several directions at once. This is on one side a clear advantage, yet on the other, when you do a lot of "things", there are lots of potential complications that need to be addressed before the manuscript can be accepted.

We thank the reviewers for their positive and thoughtful comments, and the reviewing editor for preparing the summary and the consolidated comments. Below we outline how we have addressed the essential revisions in the revised manuscript.

Essential revisions:

1) Imaging analysis.

Reviewer 1 pointed out that the critical finding that sets this work apart from prior studies is the observation that task-related piriform-insula connectivity differences statistically mediate a relationship between 2-OG and food choice. Yet a lot relies on the veracity of the neuroimaging analysis. Both estimates of task effects (e.g., GLM) and connectivity measures in the BOLD response are highly sensitive to motion and physiological artifacts that can sometimes lead to spurious observations. This brings up several concerns.

First, reviewer 1 is concerned that the food > non-food difference in piriform representations may be due to differences in head motion across groups. It seems possible that sleep manipulations can lead to differences in head motion in the scanner, that in turn would alter the reliability of the representational distance estimates that the SVN decoder picks up on (leading to a potential spurious group difference). The authors should both report whether there were differences in framewise displacement measures across groups and include a displacement measure as a nuisance variable in the estimation of the group differences.

We thank the reviewers for these suggestions. We were also particularly concerned about potential differences in head motion between the two sleep conditions. We therefore included 24 (instead of the standard 6) motion regressors along with nuisance regressors that capture within-volume motion (absolute difference between odd and even slices, variance across slices) into all GLMs.

We also directly compared head motion between the two sessions (i.e., NDS and DS). Specifically, we computed the average (across scans) of the absolute volume-by-volume displacements for each of the 6 realignment parameters for each session. We found no significant sleep-dependent differences in any of these parameters (x, P=0.476; y, P=0.962; z, P=0.922; pitch, P=0.879; roll, P=0.9182; yaw, P=0.894). In addition, we computed composite scores for translation and rotation parameters, which showed no sleep-dependent differences (translation, T₂₄=0.25, P=0.803; rotation, T₂₄=−0.14, P=0.891).

Finally, as suggested, we added these two composite scores as covariates in the second level models. Adding these covariates did not change the results for the decoding (x=20 y=8 z=−12, T₂₃=6.12, P_FWE-SVC=0.0001) or the connectivity analysis (right insula, x=40, y=6, z=0, T₂₁=5.42, P_FWE-SVC=0.021).

We have added the results of these control analyses to the manuscript.

Materials and methods section:

“To quantify and compare head motion between sessions we computed the average (across scans) of the absolute volume-by-volume displacements for each of the 6 realignment parameters for each session. We then computed composite scores for translation and rotation parameters and compared them between sessions. There were no sleep-dependent differences in these composite head motion parameters (translation, T₂₄=0.25, P=0.803; rotation, T₂₄=−0.14, P=0.891).”

Legend Figure 3:

“This result did not change when including covariates for head motion (translation and rotation) into the group-level model (x=20 y=8 z=−12, T₂₃=6.12, P_FWE-SVC=0.0001).”

Legend Figure 5:

“This result did not change when including covariates for head motion (translation and rotation) into the group-level model (x=40, y=6, z=0, T₂₁=5.42, P_FWE-SVC=0.021).”

Second, reviewer 1 pointed out that while there is no statistical difference in sniff amplitude (Figure 1—figure supplement 3E), there does appear to be a mean difference such that the sleep restricted group is taking lower amplitude sniffs. […] One other concern in this regard is that if the piriform-insula connectivity measure is for odors (collapsed) v. clean air, then this sniff response comparison should be for odors (collapsed) v. clean air. (One note on food odors v. non-food odors: it's confusing to me that celery seed is considered a non-food odor as this is a not-uncommon ingredient in food dishes.)

The reviewers point out a number of important issues. Similar to head motion, we controlled for potential respiratory confounds by including respiratory traces as covariates in all GLMs used for the decoding and connectivity analyses. We also tested whether the respiratory measures differed between sessions, the results of which are reported in the legend accompanying Figure 1—figure supplement 3.

In addition to these controls, we have computed new GLMs which include respiratory nuisance regressors that model the “respiration response function” as defined in Birn et al., (2008). Controlling for breathing using these more advanced nuisance regressors did not change the results of the decoding (x=20, y=8, z=−12, T₂₄=4.83, P_FWE-SVC=0.001) or connectivity analyses (x=40, y=6, z=0, T₂₃=5.59, P_FWE-SVC=0.011). Taken together, these new control analyses suggest that our findings are unlikely to be driven by respiratory artifacts.

Regarding sleep-dependent differences in odor-evoked respiratory amplitude and latency, we do not fully understand the reviewer’s concern when they write “it is not clear where the amplitude and latency measures come from but they don't seem to capture the differences you can see if you overlap the curves in C and D”. We fully agree that the peak of the respiratory trace for food in NDS (light pink trace in Figure 1—figure supplement 3C) appears to roughly line up with the peak of the respiratory trace for clean air in DS (dark gray trace in Figure 1—figure supplement 3D). We would like to point out that these respiratory traces (averaged across trials and subjects) are presented in panels C and D only for visualization, and the across-subjects mean values of respiratory peak amplitude and latency directly corresponding to these traces are plotted in panels E and F of the same figure using bar plots with error bars. We suspect that the confusion may have originated from the fact that the respiratory traces themselves in Figure 1—figure supplement 3C and D did not contain error bars, which are necessary to interpret the mean differences. We now include error bars on these traces to better illustrate that differences between respiratory responses to different odors are within what can be expected by chance.

Finally, as requested, we also computed statistical tests for comparing odors (collapsed) vs. clean air trials which show that there are no sleep-dependent effects in respiratory peak amplitude or latency (sleep-by-odor ANOVA on amplitude; main effect sleep F_1,24=1.45, P=0.240; main effect odor F_1,24=22.41, P<0.0001; sleep-by-odor interaction, F_1,24=1.24, P=0.277; sleep-by-odor ANOVA on latency; main effect sleep F_1,24=0.14, P=0.715; main effect odor F_1,24=3.50, P=0.074; sleep-by-odor interaction, F_1,24=0.53, P=0.474).

We have added the results of these additional analyses to the manuscript.

Legend Figure 3:

“Finally, controlling for respiratory response functions (Birn et al., 2008) did not change the result (x=20, y=8, z=−12, T₂₄=4.83, P_FWE-SVC=0.001).”

Legend Figure 5:

“In addition, controlling for respiratory response functions (Birn et al., 2008) did not change the result (x=40, y=6, z=0, T₂₃=5.59, P_FWE-SVC=0.011).”

Legend Figure 1—figure supplement 3:

“There were also sleep-depended effects on amplitude and latency when comparing odor (collapsed across food and non-food odors) vs. clean air trials (sleep-by-odor ANOVA on amplitude; main effect sleep F_1,24=1.45, P=0.240; main effect odor F_1,24=22.41, P<0.0001; sleep-by-odor interaction, F_1,24=1.24, P=0.277; sleep-by-odor ANOVA on latency; main effect sleep F_1,24=0.14, P=0.715; main effect odor F_1,24=3.50, P=0.074; sleep-by-odor interaction, F_1,24=0.53, P=0. 474).”

We address the question about the choice of celery seed as a non-food odor below.

Finally, all three reviewers raised concerns about the use of different contrasts used in the analyses. Reviewer 1 points out that the search for group differences in representational distances in the piriform uses the food > non-food contrast, but then the connectivity results between piriform and insula are using the odor > clean air contrast. It is likely that this is because the food > non-food contrast did not produce a significant result. This should still be reported and clarified in the text.

We used the food vs. non-food contrast in our decoding analysis to test for differences in encoding of odor information. For the connectivity analysis, we tested odor-evoked connectivity, thus contrasting connectivity during odor vs. clean air trials. Reviewers are correct that the contrast between food and non-food odor for sleep-dependent changes in connectivity did not reveal significant effects. We now report this in the manuscript.

Results section:

“In contrast, sleep-dependent changes in piriform connectivity for food vs. non-food odors did not show a significant relationship with changes in food choices.”

2) Pleasantness confounds.

Both reviewers 1 and 2 had concerns regarding the role of subjective pleasantness preferences had on the task. Reviewer 2 pointed out that a major analysis in this manuscript is the searchlight-based multi-voxel pattern analysis contrasting the response to "food" and "non-food" odors. As clearly indicated in Figure 1—figure supplement 3, "food" odors were much (as in p = 2.5x10-6) more pleasant than "non-food" odors. Thus, why is this to be considered a contrast of "food vs. non-food" and not a contrast of "pleasant vs. less-pleasant". The impact of sleep deprivation on hedonics is also interesting, but not the aim of this manuscript. This potential confound is so blatant, that the reader must be missing something. Thus, what is missing? Why is the sharp difference in pleasantness between the edible and non-edible odorants not a concern? The authors can do one of two things: Either better explain how this relates directly to food vs. non-food odors, or run analyses to address this. For example, either regress out pleasantness differences, or select subsets of data devoid of this potential confound.

The objective of this experiment was to study sleep-dependent brain responses to food odors corresponding to highly palatable energy-dense food items (caramel, cinnamon bun, potato chips, pot roast, garlic bread, etc.), relative to odors of objects (in principle edible) with low palatability and low energy density (celery seed and fir needle). We expected that higher palatability would be reflected in higher pleasantness ratings, and thus view the pleasantness difference as validation of our experimental design with respect to stimulus selection.

However, we understand the reviewers’ concerns and conducted additional analyses to test whether our effects are merely driven by pleasantness. In these analyses, we controlled for pleasantness in the single-subject GLMs used to estimate activity patterns evoked by food and non-food odors by adding the ratings as a nuisance regressor. Interestingly, despite the fact that food and non-food odors differed in pleasantness, we still found a significant (but much weaker) effect in our piriform ROI, such that odor encoding was enhanced in the sleep-deprived session (x=20, y=8, z=−10, T₂₄=3.12, P_FWE-SVC=0.045).

In addition, it is worth emphasizing that pleasantness ratings for food and non-food odors did not differ between the two sleep sessions, and that controlling for individual differences in the pleasantness of food vs. non-food odors in the group-level analysis did not change our decoding results (x=20, y=8, z=−12, T₂₃=6.34, P_FWE-SVC=0.0001).

We now include these results into the manuscript and mention the potential pleasantness confound in the Discussion section.

Legend Figure 3:

“In addition, results were still significant when including covariates for odor pleasantness in the first- (x=20, y=8, z=−10, T₂₄=3.12, P_FWE-SVC=0.045) and group-level models (x=20, y=8, z=−12, T₂₃=6.34, P_FWE-SVC=0.0001).”

Discussion section:

“In the current study, we compared brain responses to food odors with high palatability and non-food items with low palatability. As expected, the food odors were rated as higher in pleasantness than non-food odors. In principle, it is therefore possible that our observed effects for food vs. non-food odor encoding in the brain were fully explained by this corresponding pleasantness difference. However, our results remained significant when including pleasantness as a covariate in the statistical models, indicating that in this case pleasantness does not account for our results.”

Related to this, reviewer 2 pointed out that the selection of non-food odorants was slightly odd. After all, celery seed is not that far from a food. Why didn't the authors just use perfume? It would have addressed their pleasantness difference, and it's clearly not edible. Odorant selection, however, is behind us. Thus, what the authors should at least add is the actual edibility ratings of the odorants. This should be provided in a supplementary table with the associated statistics of the differences in edibility.

Edibility ratings did indeed differ significantly between food and non-food odors (T₂₄=12.69, P=3.87x10^-12). We have now included a new supplementary table (Supplementary file 2) summarizing the odor ratings (collected during the screening session) for all odors used in the study.

Materials and methods section:

“Odor ratings collected during screening are summarized in Supplementary file 2. Most importantly, edibility ratings for food and non-food odors collected during the screening session differed significantly between food and non-food odors (T₂₄=12.69, P=3.87x10^-12).”

3) Validity of inferences.

Reviewers 2 and 3 had concerns about the certainty of the conclusions being made. Reviewer 2 pointed out that the authors conclude that "sleep deprivation induces changes in food intake through the modulation of an olfactory pathway that is related to the endocannabinoid system". This very strong statement implies causation, that it is not 100% certain the authors have in hand. There is a relationship, but the authors to make such a strong causal claim, wouldn't they need to somehow find a way to independently manipulate the olfactory pathway, and show that it is indeed responsible for the effect? Would sleep deprivation fail to impact eating behaviour of individuals with anosmia? This is a major concern, but it is of course trivial to address: The authors should either slightly tone down the claims on causation (the manuscript is strong enough as it is) or make a better substantiated claim on causation.

We have toned down our conclusions regarding causation throughout the manuscript. In particular, we have changed the sentence referenced above to:

Discussion section:

“Taken together, our findings show that sleep-dependent changes in food choices are associated with changes in an olfactory pathway that is related to the ECS.”

Reviewer 3 pointed out that the conclusions of the paper emphasize a finding that olfaction contributes to sleep-dependent food intake in humans (see Title). This is an overstatement of the results. First, unlike other papers, the present study did not find a difference in food intake (caloric intake) based on sleep deprivation. The authors did find a difference in food preference (food decisions). Second, the measure of olfaction in the result is not olfaction per se but rather piriform-insula connectivity when sleep deprived v. not sleep deprived. This should be made more clear and up front throughout the manuscript.

The reviewer is correct that the relationship between energy-dense food choices and total calorie intake is not evident in our study. We further agree that energy-dense food choice was correlated with connectivity of the olfactory cortex, not olfaction per se.

We have changed wording throughout the manuscript and also changed the Title to:

“Olfactory connectivity mediates sleep-dependent food choices in humans”

4) 2-OG vs. food choice association.

All three reviewers had a concern about the association between 2-OG and food choice. An inspection of Figure 2B suggests that the association between ECS system and food choice may be driven by 5 participants. Reviewers 2 and 3 pointed out that if these outliers are removed, the association appears to become negative. The authors seem to be aware of this since they used a robust regression analysis; however, robust regression approaches simply account for differences in variance of each observation in the overall estimate. Reviewer 1 recommend a non-parametric statistic, such as a bootstrap or permutation test. (Note: This is listed as a Major concern because this is one of the critical observations for the authors' primary conclusions).

Following this suggestion, we computed a permutation test to verify the result from the robust regression analysis. This permutation test (100,000 permutations) confirmed that the correlation between 2-OG and food choice is statistically significant (p=0.018).

We have added this result to the manuscript.

Results section:

“Interestingly, sleep-dependent increases in 2-OG correlated significantly with increases in the energy density of food consumed at the post-scanning buffet (robust regression, β=0.47, P=0.027; permutation test, P=0.018; Figure 2B).”

5) Compliance.

Reviewer 3 raised concerns about subject compliance. There is a lot of trust in participants. Trust that they did not consume additional food and drinks, trust they did not take caffeine, and trust in both the participant and actigraph that the sleep restriction instructions were complied with. Given the importance of these, it is surprising that participants were not kept in-lab for the sleep deprivation manipulation. The authors should address this.

We agree that the issue of compliance is an inherent concern in sleep and dietary manipulation studies conducted in the field. We opted for an in-home sleep manipulation in order to render the sleep deprivation as ecologically valid as possible without the typical distractors and stressors of being in an unfamiliar hospital laboratory environment. However, we took several measures to monitor compliance and to control for potential confounds. Most importantly, we used wrist actigraphy to monitor subjects’ sleep-wake schedule, as reported in Figure 1—figure supplement 1. We also collected a time stamped self-reported sleep diary and ratings of sleep quality during both sessions (Figure 1D and E) to verify that our sleep protocol was effective.

Importantly, non-compliance with the sleep instructions would have presumably increased between subject variability, thereby decreasing the likelihood of observing significant effects. Thus, assuming that some subjects may have “cheated”, the resulting observations are a conservative estimate of the true effect of sleep deprivation on the brain activity and behavior reported here.

With regards to food intake, two of the four meals provided to the participants prior to scanning were consumed at the laboratory and imaging center, respectively. We also provided subjects with repeated instructions to not consume any additional foods or snacks in the 24 hours preceding the scanning session. Regarding the remaining two take-out meals, subjects were required to report whether they consumed all of the provided food, and they were also asked to report any additional foods or drinks consumed.

We now discuss these issues in the manuscript.

Materials and methods section:

“Our study used an in-home setting to render the sleep-deprivation protocol as ecologically valid as possible without the additional distractions and stressors of being in an unfamiliar hospital laboratory environment. However, because in-home settings come with potential limitations related to non-compliance, we took several measures to reinforce and monitor compliance with the sleep stabilization and sleep manipulation schedule.”

6) Sniffing vs. inhalation.

Reviewer 2 raised a concern about the nature of the inhalation measure identified as sniffing. […] In future, reviewer 2 suggests the authors precisely measure sniffing in their studies. It's cheap, easy, and informative.

We appreciate the methodological advice from reviewer 2. We now refer to “respiration” not “sniff” in Figure 1—figure supplement 3 and throughout the manuscript. We also now clearly state in the methods that respiratory effort bands only provide an indirect measure of sniffing.

Materials and methods section:

“Respiration, as an indirect measure of nasal sniffing, was measured using a MR-compatible breathing belt (BIOPAC Systems Inc, Goleta, CA) affixed around the participant’s torso, and recorded at 1 kHz using PowerLab equipment (ADInstruments, Dunedin, New Zealand).”

Author details

1. Surabhi Bhutani

   1. Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States
   2. School of Exercise and Nutritional Sciences, College of Health and Human Services, San Diego State University, San Diego, United States

   Contribution

   Conceptualization, Formal analysis, Investigation, Writing—original draft

   Competing interests

   No competing interests declared

2. James D Howard

   Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States

   Contribution

   Formal analysis, Investigation, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9309-3773

3. Rachel Reynolds

   Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States

   Contribution

   Investigation, Writing—review and editing

   Competing interests

   No competing interests declared

4. Phyllis C Zee

   Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States

   Contribution

   Conceptualization, Writing—review and editing

   Competing interests

   No competing interests declared

5. Jay Gottfried

   1. Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States
   2. Department of Neurology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, United States
   3. Department of Psychology, School of Arts and Sciences, University of Pennsylvania, Philadelphia, United States

   Contribution

   Conceptualization, Writing—review and editing

   Competing interests

   No competing interests declared

6. Thorsten Kahnt

   1. Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, United States
   2. Department of Psychiatry and Behavioral Sciences, Feinberg School of Medicine, Northwestern University, Chicago, United States
   3. Department of Psychology, Weinberg College of Arts and Sciences, Northwestern University, Evanston, United States

   Contribution

   Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing—original draft

   For correspondence

   thorsten.kahnt@northwestern.edu

   Competing interests

   Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0002-3575-2670

• 11,442

  Page views

• 775

  Downloads

• 17

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.