Appetite-regulating hormones modulate odor perception and odor-evoked activity in hypothalamus and olfactory cortices

Abstract

Odors guide food seeking, and food intake modulates olfactory function. This interaction is mediated by appetite-regulating hormones like ghrelin, insulin, and leptin, which alter activity in the rodent olfactory bulb, but their effects on downstream olfactory cortices have not yet been established in humans. The olfactory tract connects the olfactory bulb to the cortex through 3 main striae, terminating in the piriform cortex (PirC), amygdala (AMY), olfactory tubercule (OT), and anterior olfactory nucleus (AON). Here, we test the hypothesis that appetite-regulating hormones modulate olfactory processing in the endpoints of the olfactory tract and the hypothalamus. We collected odor-evoked functional magnetic resonance imaging (fMRI) responses and plasma levels of ghrelin, insulin, and leptin from human subjects (n = 25) after a standardized meal. We found that a hormonal composite measure, capturing variance relating positively to insulin and negatively to ghrelin, correlated inversely with odor intensity ratings and fMRI responses to odorized vs. clean air in the hypothalamus, OT, and AON. No significant correlations were found with activity in PirC or AMY, the endpoints of the lateral stria. Exploratory whole-brain analyses revealed significant correlations near the diagonal band of Broca and parahippocampal gyrus. These results demonstrate that high (low) blood plasma concentrations of insulin (ghrelin) decrease perceived odor intensity and odor-evoked activity in the cortical targets of the medial and intermediate striae of the olfactory tract, as well as the hypothalamus. These findings expand our understanding of the cortical mechanisms by which metabolic hormones in humans modulate olfactory processing after a meal.

functional magnetic resonance imaging, ghrelin, human, insulin, olfaction, olfactory cortex

Introduction

Food intake and olfaction are closely intertwined (Yeomans 2006). Odors guide food search and eating can modulate olfactory function (Rolls 2005, 2015; Shanahan et al. 2021). For instance, exposure to food odors increases appetite (Zoon et al. 2016), induces salivation (Morquecho-Campos et al. 2019), and promotes food-seeking behaviors throughout the animal kingdom (Fine and Riera 2019; Zjacic and Scholz 2022). Conversely, food intake reduces the sensitivity to perceived odors in both animals and humans (Aime et al. 2012; Ramaekers et al. 2016; Shanahan and Kahnt 2022). However, the neural mechanisms underlying this interaction are not fully understood.

Food intake is governed by appetite-regulating hormones, most notably ghrelin, insulin, and leptin. Blood plasma levels of ghrelin rise before a meal and fall to a trough afterwards, whereas insulin levels surge immediately after a meal (Cummings et al. 2001). Although leptin levels are also related to meal timing and rise throughout the day, they are only weakly increased by food intake (Schoeller et al. 1997). These metabolic hormones have been shown to modulate olfaction across species. For instance, in rodents, administration of ghrelin increases sniffing frequency and enhances olfactory sensitivity (Tong et al. 2011), whereas insulin decreases olfactory sensitivity (Aime et al. 2012). In addition, leptin-deficient mice are faster to find buried food, and this effect is reversed by leptin administration (Getchell et al. 2006). Similar effects have been obtained in humans. Peripheral ghrelin levels correlate with perceived odor intensity (Ginieis et al. 2022), and systemic ghrelin infusions increase sniffing (Tong et al. 2011) and enhance food odor conditioning (Han et al. 2018). Conversely, increases in blood plasma insulin are related to lower olfactory sensitivity (Thanarajah et al. 2019), and intranasal insulin administration reduces olfactory sensitivity (Brunner et al. 2013). In contrast, peripheral leptin levels have not been shown to correlate with olfactory sensitivity or odor ratings (Ginieis et al. 2022).

These effects are likely mediated by actions on metabolic hormone receptors present throughout the olfactory system and the hypothalamus. Specifically, ghrelin receptors are present in the rodent olfactory bulb (OB), piriform cortex (PirC), amygdala (AMY), and hypothalamus (Zigman et al. 2006; Tong et al. 2011; Alvarez-Crespo et al. 2012). Similarly, in rodents, insulin receptors are present in the OB, PirC, anterior olfactory nucleus (AON), olfactory tubercule (OT), and hypothalamus (Hill et al. 1986; Werther et al. 1987; Marks et al. 1990; Aime et al. 2012; Scherer et al. 2021), and leptin receptors are found in the OB and the hypothalamus (Shioda et al. 1998; Prud’homme et al. 2009). By acting on these receptors, metabolic hormones can modulate neural activity. For instance, ghrelin increases the responsiveness of olfactory sensory neurons (Loch et al. 2015), and insulin has excitatory effects on mitral cells (Fadool et al. 2011; Kuczewski et al. 2014), which are inhibited by leptin (Sun et al. 2019). Importantly, given that these receptors are present in the output cell layers of the OB (i.e., mitral and tufted cells) (Palouzier-Paulignan et al. 2012), these hormones may affect neural processing in the olfactory cortex and throughout the rest of the brain.

The olfactory bulb (OB) sends projections via the 3 striae of the olfactory tract to the PirC and AMY (lateral stria), OT (medial stria), and AON (intermediate stria) (Nieuwenhuys et al. 2008; Echevarria-Cooper et al. 2022). These areas are considered primary olfactory cortex as they receive direct input from the OB and are also directly connected to the hypothalamus (Palouzier-Paulignan et al. 2012) that plays a key role in endocrine signaling and the regulation of food intake (Saper and Lowell 2014; Stuber and Wise 2016). While the effects of metabolic hormones on activity in the OB are relatively well established in animals, their effects on neural activity in the endpoints of the olfactory tract are less well characterized, especially in humans.

To address this question, we analyzed data from a recent study (Bhutani et al. 2019) in which we collected blood plasma levels of ghrelin, insulin, and leptin from healthy human subjects undergoing olfactory task-based fMRI after a meal. It is important to note that because OB responses cannot readily be measured with fMRI, we are unable to discern whether modulatory effects in olfactory cortical areas are driven by local actions of hormones in these areas or whether they reflect effects that are passed down from the OB. Notwithstanding these interpretational challenges, we found that higher plasma insulin and lower ghrelin levels were related to lower odor intensity ratings and decreased fMRI responses to odors in AON, OT, and hypothalamus. Moreover, exploratory analyses revealed correlations between metabolic hormones and odor-evoked fMRI activity in the parahippocampal gyrus (PHG) and the diagonal band of Broca (DBB).

Methods

Subjects

Our previous study (Bhutani et al. 2019) screened and obtained written consent from 41 healthy, right-handed, nonsmoking human subjects between the ages of 18 and 40 yr old with body mass index (BMI) values between 18.5 and 24.9 kg/m². Subjects had no history of neurological disorders. Out of the 41 potential subjects screened, 29 continued with the experiment. Three dropped out due to discomfort inside the scanner, and 1 was not included in the analysis due to a large number of missed responses. This resulted in a final sample of n = 25 subjects (10 male) that were included in the current analysis. All experimental procedures were approved by the Institutional Review Board of Northwestern University.

Study overview

Each participant completed 2 fMRI sessions, 28 d apart, which differed only by whether the subject was sleep deprived (4 h in bed) or not (8 h in bed) the night before the fMRI session. Subjects were asked to abstain from drugs, alcohol, and caffeine over the duration of the study to prevent disturbances in hormone levels or sleep. Isocaloric meals were provided for the 24 h before each fMRI session. The fMRI sessions took place at around 7 PM in the evening and prior to entering the fMRI scanner, a standardized isocaloric dinner was provided. Blood plasma was collected at baseline before the dinner and at 4 time points until the end of the session (Fig. 1A). The fMRI session lasted for 1 h and subjects were asked to rate the pleasantness and intensity of each odor (food, nonfood, or clean air). For the purpose of the current analysis, we collapsed the data collected across the 2 sessions.

Fig. 1.

Experimental design. A) Timeline of experimental day. B) Example trial of the olfactory task. C) Four out of 6 food odors (caramel, cinnamon roll, gingerbread, pot roast, garlic, and potato chip) were chosen for each participant and presented in the fMRI task along with 2 nonfood odors (fir and celery seed), and clean air.

Controlled food intake

Subjects were provided with an isocaloric diet for 24 h before each session. All meals provided to the participants for home consumption were based on their estimated caloric needs using their sex, age, and BMI (range: 1,400 to 2,600 kcal/d). A standardized, isocaloric dinner was also provided at 6 PM on the day of the fMRI session and consisted of an entrée, fruit/snack, and a small nonalcoholic and noncaffeinated drink. All subjects self-reported refraining from consuming any additional food or beverages prior to the fMRI session.

Olfactory fMRI task

Subjects entered the scanner at around 6:45 PM. Immediately before the start of scanning, subjects were instructed to rate the pleasantness, edibility, and quality of the 4 individually selected food odors (chosen based on pleasantness ratings from: caramel, cinnamon roll, gingerbread, pot roast, garlic, and potato chip) and 2 nonfood odors (fir and celery seed, Fig. 1C). Each fMRI session was comprised of 4 fMRI runs and each run contained 63 pseudo-randomized trials of olfactory stimulation. On each trial (Fig. 1B), after a 2 s countdown, a white crosshair in the center of the screen turned blue, indicating that participants should sniff for 2.5 s. Then, a rating scale appeared for 5 s, and subjects were asked to rate either the pleasantness or intensity of the odor. There was then a 1 to 8 s inter-trial interval. In total, each odor and clean air were presented 9 times per run, for a total of 63 trials per run.

Blood plasma collection and assay

In total, 5 blood plasma samples were collected per fMRI session. Samples were collected with an MRI-safe catheter placed in subjects’ left arms every 30 min beginning at baseline (i.e., prior to the isocaloric dinner) and at 4 time points until the end of the fMRI session. Prior to drawing each sample, 1.5-2 mL of blood was discarded to remove diluted blood from the dead space in the catheter. Samples were then put on ice, centrifuged, aspirated, divided into aliquots, and stored at –80 ^oC until assay. Enzyme-linked immunosorbent assay (ELISA) kits were used to determine total plasma ghrelin (human ghrelin, Millipore) and plasma leptin (human leptin, Millipore), and cobas e411 analyzer (Roche) was used for plasma insulin.

fMRI data acquisition

Functional MRI data were acquired using a Siemens 3T PRISMA with a 64-channel head–neck coil. First, we acquired a high-resolution (1 mm isotropic) T1-weighted structural scan. Then, for each run, 382 Echo-Planar Imaging (EPI) volumes were acquired with a parallel imaging sequence. The parameters were as follows: 2 s repetition time; 22 m echo time; 104 × 96 matrix size; 208 × 192 mm field-of-view; 2 × 2 mm² in-plane resolution; 90^o flip angle; 2 multiband acceleration factor; 2 mm slice thickness; 58 slices; no gap. The acquisition angle was tilted by –30^o from AC–PC to minimize signal dropout in olfactory areas (Deichmann et al. 2003; Weiskopf et al. 2006). Lastly, 10 whole-brain EPI volumes were captured with the same parameters as above, except these had 96 slices and a repetition time of 3.22 s.

fMRI preprocessing

fMRI data were preprocessed using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Functional volumes were aligned for each participant to the first acquired volume to correct for head motion. The whole-brain EPI volumes were realigned and averaged. Then, the mean whole-brain EPI was co-registered to the T1 image, and the mean functional volume was co-registered to the mean whole-brain EPI. The T1 image was then normalized to MNI space, and the normalization parameters were applied to the motion-corrected and co-registered functional images. Finally, functional volumes were spatially smoothed with an 8 × 8 × 8 mm FWHM Gaussian kernel.

Regions of interest

Regions of interest (ROI) were selected from the bilateral endpoints of the olfactory striae (PirC, AMY, OT, and AON) and the hypothalamus. The olfactory (Echevarria-Cooper et al. 2022) and hypothalamus ROIs (Neudorfer et al. 2020) were taken from publicly available atlases, and binary masks were created for each ROI by thresholding the atlases at 0.2.

Analysis of olfactory fMRI data

Normalized and smoothed functional images from both sessions were analyzed using a GLM approach. The onsets of: (i) food odors, (ii) nonfood odors, and (iii) clean air were included as separate regressors in the GLM. Additionally, the following nuisance regressors were included: smoothed and normalized respiratory trace down-sampled to scan resolution; 6 realignment parameters (3 translations and 3 rotations) calculated for each volume; the derivative, square, and square of the derivative of each realignment parameter; the absolute value of the signal difference between even and odd slices; and the variance across slices in each functional volume. Contrast maps were computed for all odors > clean air, food odors > clean air, and nonfood odors > clean air. Contrast estimates from the odor > clean air contrast was extracted from each subject in each ROI, and 1-sample t-tests (P < 0.05, 1-tailed) were conducted on the average contrast estimates in each ROI.

Analysis of hormone data

To determine if levels of metabolic hormones changed after the meal, for each hormone, we computed a 1-way repeated-measures ANOVA with time as a within-subject factor. Significant main effects of time (P < 0.05, 2-tailed) were followed up by post hoc paired t-tests (P < 0.05, 2-tailed) comparing each time point to baseline.

To reduce the dimensionality of the hormonal data, we conducted a principal component analysis (PCA) to extract independent factors explaining maximal variability in the 3 appetite-regulating hormones ghrelin, insulin, and leptin. This analysis was performed on average hormone levels (i.e., averaged across the 3 blood plasma samples drawn during the fMRI scan in each of the 2 fMRI sessions). Based on the explained variance and factor loadings of the resulting principal components, we chose to proceed with the first and second components for further correlation analysis.

Specifically, we computed Pearson’s correlation coefficients (r) between the hormonal data, and the neural data (P < 0.05, 2-tailed), or odor ratings (P < 0.05, 1-tailed). Differences between correlations were tested for significance with dependent sample z-tests (P < 0.05, one-tailed). The neural data consisted of contrast estimates extracted from each ROI (averaged across voxels) for contrasts odor > clean air, and the odor ratings consisted of intensity ratings that subjects made in response to food and nonfood odors during the fMRI session.

We further conducted an exploratory voxel-wise analysis to identify additional areas where metabolic hormones modulated odor-evoked fMRI responses. For this, we correlated the first principal component (PC1) with estimates from the all odors > clean air contrast in each voxel within a whole-brain gray matter mask (excluding cerebellum). Significant clusters were identified using a threshold of P < 0.0001 (1-tailed) and a cluster extent threshold of k > 10 voxels.

Results

Postprandial levels of metabolic hormones

Upon arrival at the lab, baseline blood plasma samples were collected, and subjects were given a standardized isocaloric dinner. Blood samples were collected immediately before, during, and immediately after fMRI scanning, with a gap of 30 min between samples (Fig. 1A).

Levels of ghrelin, insulin, and leptin changed significantly after the meal. A set of 1-way ANOVAs with repeated measures revealed a significant main effect of time on the levels of insulin (F = 29.71, P = 5.04 × 10^–11), ghrelin (F = 48.23, P = 7.68 × 10^–9), and leptin (F = 18.32, P = 1.27 × 10^–5). In addition, post hoc paired t-tests showed that every timepoint following the meal was changed significantly from baseline for insulin (all P < 6.78 × 10^–7; Fig. 2A and 2D), ghrelin (all P < 1.13 × 10^–4; Fig. 2B and 2E), and leptin (all P < 0.05; Fig. 2C and 2F).

Fig. 2.

Postprandial levels of metabolic hormones. A to C) Changes in metabolic hormone levels after the experimental dinner. At all-time points following the dinner, levels of insulin (A), ghrelin (B), and leptin (C) were significantly changed from baseline (before the meal, timepoint 0). Thick line depicts group mean and error bars show SEM, thin lines depict individual subjects’ data, and ^* denotes significant (P < 0.05) difference compared to baseline. D and E) Percent change for insulin (D), ghrelin (E), and leptin (F) from baseline.

For the purpose of this study, we averaged hormone measurements collected during the fMRI session (at 60, 90, and 120 min). Although 1-way ANOVAs with repeated measures on these 3 time points revealed a significant main effect of time for ghrelin (F = 18.41, P = 1.2 × 10^–5) and leptin (F = 20.17, P = 6.6 × 10^–5), not insulin (F = 2.81, P = 0.09), the time points for each hormone were highly correlated (insulin: r > 0.80, ghrelin: r > 0.97, leptin: r > 0.99), suggesting that averaging across time would not meaningfully affect correlations with perceptual or neural responses.

As these 3 appetite-regulating hormones were highly correlated, we utilized principal component analysis (PCA) to derive a composite measure for the hormones (Fig. 3A and 3B). The first hormonal component (PC1) explained 47.28% of the variance in the data and the second hormonal component (PC2) explained 38.35%. PC1 was mainly composed of insulin and ghrelin, with a positive contribution of insulin (0.71) and a similar, but negative contribution from ghrelin (–0.7). Leptin had a near-zero loading on PC1 (0.01) but loaded positively on PC2 (0.89).

Fig. 3.

Principal component analysis. A) Eigen values of the principal component analysis and B) individual hormone loadings for each factor. C to E) Scatter plots showing the relationship between the first hormonal component (PC1) and insulin (C), ghrelin (D), and leptin (E).

We correlated each individual hormonal measure with PC1, confirming that insulin was positively correlated (r = 0.84, P = 1.1 × 10^–7, Fig. 3C) and ghrelin was negatively correlated with PC1 (r = –0.84, P = 1.6 × 10^–7, Fig. 3D). In contrast, there was no significant correlation between leptin and PC1 (r = 0.01, P = 0.96, Fig. 3E). This suggests that any correlations between PC1 and behavioral or neural measures (see below) are primarily driven by insulin and ghrelin levels.

First hormonal component (PC1) is related to perceived odor intensity

We first tested whether differences in appetite-regulating hormones are related to differences in olfactory perception. For this, we correlated the first and second hormonal components with average odor intensity ratings (food and nonfood odors) that subjects made during the fMRI scan. We found a significant negative correlation between odor intensity ratings and PC1 (r = –0.42, P = 0.02) and no significant correlation with PC2 (r = –0.22, P = 0.15, Fig. 4A and 4B). We also tested this correlation separately for intensity ratings made for food and nonfood odors. This revealed a significant negative correlation between PC1 and intensity ratings for food odors (r = –0.44, P = 0.01) but not for nonfood odors (r = –0.32, P = 0.06). However, z-tests (for dependent samples) showed that the difference between the correlations involving food and nonfood odors was not significant (P = 0.24). Taken together, these results suggest that metabolic hormones alter the perceived intensity of odors, such that insulin reduces while ghrelin enhances olfactory sensitivity.

Fig. 4.

First hormonal component (PC1) is related to perceived odor intensity. A and B) Scatter plots showing relationships between the first (A) and second hormonal components (B) and average odor intensity ratings collected during fMRI scanning.

Hormone-neural correlates in the endpoints of the olfactory tract and hypothalamus

We next tested whether appetite-regulating hormones affect odor-evoked activity in primary olfactory brain regions (Echevarria-Cooper et al. 2022) and the hypothalamus (Fig. 5A). As a sanity check, we first asked whether fMRI activity in these ROIs responded to odors. We found significantly higher fMRI responses to odors than clean air in the PirC (t = 9.22, P = 1.2 × 10^–9), AMY (t = 8.98, P = 1.9 × 10^–9), OT (t = 4.34, P =1.1 × 10^–4), and the hypothalamus (t = 3.14, P = 0.002). However, no significant response to odors compared to clean air was found in the AON (t = 1.13, P = 0.14, Fig. 5B).

Fig. 5.

Olfactory cortical areas respond to odorized compared to clean air. A) Hypothalamus and olfactory regions of interest. B) Piriform cortex (PirC), amygdala (AMY), olfactory tubercle (OT), and hypothalamus show significantly stronger fMRI responses to odor compared to clean air. Dots depict individual subjects, error bars show the mean and SEM, and ^*Denotes significance at P < 0.01.

To determine whether olfactory processing in these areas was modulated by appetite-regulating hormones, we correlated the first hormonal component with fMRI responses to odor > clean air in the ROIs. We found significant negative correlations between PC1 and odor-evoked fMRI responses in the AON (r = –0.6, P = 0.002), OT (r = –0.42, P = 0.03), and hypothalamus (r = –0.52, P = 0.01). Correlations with odor-evoked responses in PirC (r = –0.17, P = 0.43) and AMY (r = –0.29, P = 0.16) were nominally negative but not significant (Fig. 6). No correlations were observed with PC2 (all P > 0.09). Additionally, we used a z-test (for dependent samples) which revealed no significant differences between correlations in any ROI, except between PirC and AON (z-score = 1.989, P = 0.023, 1-tailed). In post hoc analyses, we also correlated the individual hormonal measures with the same odor-evoked fMRI responses (Supplementary Table 1). Although correlative in nature, these findings suggest that appetite-regulating hormones may modulate neural processing of odors in the AON, OT, and hypothalamus.

Fig. 6.

Hormone-neural correlates in the endpoints of the olfactory tract and hypothalamus. A to E) Region of interest correlation analysis. fMRI responses to all odors > clean air in the medial and intermediate olfactory striae endpoints (OT and AON) and hypothalamus significantly correlate with PC1. No significant correlations are observed in the endpoints of the lateral stria (PirC and AMY).

In an additional post hoc analysis, we investigated whether these correlations were driven by food and/or nonfood odors. Specifically, in areas that showed significant negative correlations between PC1 and odor-evoked fMRI signals, we also computed correlations with fMRI responses to food odors > clean air and nonfood odors > clean air, separately. We found significant correlations for both food and nonfood odors in the AON (PC1–AON_food: r = –0.48, P = 0.02; PC1–AON_non: r = –0.64, P = 6.3 × 10^–4), OT (PC1–OT_food: r = –0.29, P = 0.16; PC1–OT_non: r = –0.49, P = 0.01; and the hypothalamus (PC1–Hypo_food: r = –0.46, P = 0.02; PC1–Hypo_non: r = –0.52, P = 0.01). Importantly, nominal differences in the size of the correlations with food and nonfood odors were not significant (all P > 0.07), suggesting that appetite-regulating hormones modulated the processing of food and nonfood odors to a comparable degree.

Whole-brain voxel-wise correlation analysis

Finally, we conducted an exploratory analysis to identify areas outside of our a priori ROIs where appetite-regulating hormones may modulate olfactory processing. For this, we performed a whole-brain, voxel-wise correlation analysis, and found maximum correlations (P < 0.0001, uncorrected, k > 10) between fMRI responses (odor > clean air) and PC1 in the diagonal band of Broca (DBB, MNI: [–6, 8, –18], t = –4.64) and parahippocampal gyrus (PHG, MNI: [–34, –40, –18], t = –4.88, Fig. 7). This suggests that in addition to the endpoints of the olfactory tract, appetite-regulating hormones may modulate neural processing of olfactory stimuli in the diagonal band of Broca and parahippocampal gyrus.

Fig. 7.

Whole-brain voxel-wise correlation analysis. A to C) Exploratory whole brain analysis reveals maximum correlations between fMRI responses to all odors > clean air and PC1 in the diagonal band of Broca (MNI: [–6, 8, –18]) and parahippocampal gyrus (MNI: [–34, –40, –18]). Axial (A), coronal (B), and sagittal (C) views are shown. For illustrative purposes, t-map is displayed at P < 0.001, uncorrected.

Discussion

The present study shows that appetite-regulating hormones modulate olfactory perception and odor-evoked fMRI responses in the olfactory cortex and the hypothalamus. Of note, because levels of ghrelin and insulin were highly correlated, we used a composite measure to capture their shared variance. Consequently, our results are driven by either ghrelin, insulin, or a combination of both. Behaviorally, high blood plasma insulin levels (or low ghrelin levels) were associated with lower sensitivity to odors. In the brain, high blood plasma insulin levels (or low ghrelin levels) were related to general suppression of odor-evoked activity throughout the primary olfactory cortex. Specifically, hormone levels were associated with the lower odor-evoked activity to both food and nonfood odors in AON, OT, PHG and DBB, and the hypothalamus.

As expected based on past work (Cummings et al. 2001), levels of ghrelin, insulin, and leptin were significantly changed following the experimental dinner. We also found that subjects with higher blood plasma insulin levels rated odors as less intense. This finding is in line with previous reports that elevated blood plasma insulin levels are associated with decreased olfactory sensitivity (Thanarajah et al. 2019), independent of BMI (Poessel et al. 2020). Our findings also mirror work in rats that show intracerebroventricular injections of insulin decrease olfactory detection (Aime et al. 2012). However, more research is needed to determine the causal effects of insulin on human olfaction, as results from studies using intranasal insulin have produced conflicting results so far (Hallschmid 2021). Similarly, previous studies have shown that ghrelin levels correlate positively with odor intensity ratings in humans (Ginieis et al. 2022), and administration of ghrelin increases olfactory sensitivity in animals, and exploratory sniffing in both animals and humans (Tong et al. 2011). Together with these previous findings, our data suggest that after a meal, increases in insulin and decreases in ghrelin may suppress olfactory sensitivity and thereby the ability and/or motivation to search and find food.

Importantly, we further investigated the modulatory effects of these metabolic hormones on odor-evoked activity in olfactory cortices. We found that high insulin levels (and low ghrelin levels) were associated with lower fMRI responses to odors > clean air in the AON, OT, and hypothalamus. However, because differences in correlations between areas were not significant (except between PirC and AON), these effects should not be considered specific to these areas, but rather as part of a more general effect of hormone levels on olfactory processing. Interestingly, we did not find any correlations between leptin and odor ratings or odor-evoked fMRI activity, despite mechanistic work in rodents showing that exogenously administered leptin decreases olfactory discrimination and inhibits odor-evoked activity in olfactory sensory neurons (Savigner et al. 2009; Sun et al. 2019). This suggests that meal-related increases in leptin in humans may have different effects than exogenously administered leptin in animals.

Much of the current literature considers OT as part of the ventral striatum, which is involved in reward-related processing (Murata 2020), suggesting that it could serve as an important link between olfaction and reward. Reduced OT responses associated with postprandial insulin may decrease the reward value of food-related odors, disincentivizing food-search behaviors. AON has been hypothesized to act as a “sensory gate” (Brunert et al. 2023) for olfactory function. Specifically, mechanistic work in rodents has shown that inhibition of AON eliminates odor-evoked responses in the OB (Rothermel and Wachowiak 2014). Thus, it is possible that insulin postprandial reduces olfactory sensitivity and thereby the ability to find food. Lastly, the hypothalamus is a primary target for metabolic hormones, is dense in insulin receptors, and has long been considered the primary regulator for feeding-related behaviors (Saper and Lowell 2014; Stuber and Wise 2016; Timper and Bruning 2017). The negative correlation found in this area suggests a suppression of hypothalamic response due to an increase in insulin binding. Conversely, in a hungry state, it is possible that ghrelin could increase hypothalamic activity, thereby promoting feeding and olfactory sensitivity.

In our study, fMRI responses to both food and nonfood odors were modulated in the endpoints of the medial (OT) and intermediate stria (AON). This suggests that the neuromodulatory effects of insulin and/or ghrelin are not specific to the processing of food-related odors but olfactory processing in general. Surprisingly, we did not find any significant correlations with activity in the PirC or AMY, the endpoints of the lateral stria. This could be due to a ceiling effect as fMRI responses to all odors > clean air was strongest in these regions, and potential modulatory effects of metabolic hormones may not have been strong enough to measurably alter these responses. Interestingly, we found a strong correlation between hormone levels and activity in AON, but at the group level, AON was not significantly activated by odors. The lack of a significant effect is possibly due to the high variability in AON responses, which our data suggest are related to metabolic hormones.

As we are unable to directly image the OB with fMRI, it is unclear whether modulatory effects observed in cortical areas are generated locally or driven by effects in OB. However, the anatomical localization of our results may hint at an answer. Specifically, whereas mitral cells project to all olfactory areas examined here, AON and OT predominantly receive projections from OB via tufted cells (Igarashi et al. 2012; Mori and Sakano 2021; Bhattarai et al. 2022). The significant difference in the modulatory effects in PirC and AON may suggest that our effects are driven by insulin and/or ghrelin acting on tufted cells in the OB. Alternatively, our effects could have been driven by actions on local receptors in AON and OT. It is possible that these areas then modulate OB activity through centrifugal input from OT and AON to the OB (Wesson and Wilson 2011; Rothermel and Wachowiak 2014).

Our exploratory whole-brain analysis revealed that activity in PHG and DBB correlates with circulating plasma levels of insulin and ghrelin. PHG is anatomically connected to the primary olfactory cortex (PirC, AMY, OT, AON) (Powell et al. 2004) and is involved in spatial memory encoding and retrieval (Watanabe et al., 2018). Decreased PHG activity postprandial could result in weaker encoding or retrieval of food-related spatial memories, and thereby suppress food search. The DBB is a neuromodulatory area in the basal forebrain, nestled adjacent to the olfactory tract (Mark et al. 1994; Nieuwenhuys et al. 2008). The horizontal limb of DBB (hlDBB) and OB have bidirectional innervation: neurons in hlDBB are modulated by OB stimulation, and OB projection cells are modulated by input from the hlDBB (Linster and Hasselmo 2000; D’Souza and Vijayaraghavan 2014). The functional role of DBB in olfactory processing remains to be elucidated, but it is possible that it mediates the influence of higher olfactory areas on OB activity (Rothermel and Wachowiak 2014).

It is important to consider the limitations of our work. First, we re-analyzed data from a previous study, which was not originally designed to answer this specific question. Second, the sample size was relatively small. However, the large amount of data per subject (2 h of fMRI scanning and 6 blood plasma samples) may compensate for the small n by enhancing the reliability with which individual data points were measured. Third, our measures of peripheral blood hormones may not directly correlate with levels of these hormones in the brain. In dogs, peripheral and cerebrospinal fluid (CSF) insulin levels are nonlinearly related (Woods et al. 1977), but whether this translates to human physiology is unknown. Lastly, we only measured 3 metabolic hormones, and olfactory processing could be influenced by other appetite-regulating neuropeptides, neurotransmitters, and hormones.

Taken together, our findings demonstrate a modulatory effect of appetite-regulating hormones like insulin and ghrelin on olfactory sensitivity and neural processing of odors. Specifically, olfactory perception and odor-evoked activity in the AON, OT, hypothalamus, PHG, and DBB are modulated by metabolic hormones postprandial. These findings may be useful to guide future research on metabolic hormones and olfactory processing in the treatment of diabetes and obesity.

Acknowledgments

The authors thank all research participants and colleagues who contributed to the original study.

Funding

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (R21 DK118503 to T.K.) and the Intramural Research Program at the National Institute on Drug Abuse (ZIA DA000642 to T.K). The opinions expressed in this work are the authors’ own and do not reflect the view of the NIH/DHHS.

Conflict of interest statement All authors declare no conflicts of interest.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.