Introduction
Marked decreases or increases in appetite or weight are common symptoms included in the diagnostic criteria for major depressive disorder (MDD) (American Psychiatric Association, 2013). Research suggests that almost one half of adults with MDD present with hypophagic behaviors, whereas approximately one quarter present with hyperphagic behaviors attributable to the current MDD episode (Husain et al., Reference Husain, Rush, Sackeim, Wisniewski, McClintock, Craven and Hauger2005). These phenotypes differ not only behaviorally, but also in terms of brain activity and metabolism (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020). Further elucidating the neurobiological substrates of these differing MDD presentations may help develop more effective treatments.
Few studies have focused on neurobiological differences between appetite/weight phenotypes in MDD. One recent study used functional magnetic resonance imaging (fMRI) to evaluate differences in unmedicated participants with MDD who had an increase v. a decrease in appetite (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016). Participants viewed food and non-food stimuli during fMRI and rated how pleasant it would be to eat foods from a second set of visual stimuli. When presented with food v. non-food stimuli, participants with an increase in appetite (n = 16) showed significantly stronger activation in the orbitofrontal cortex (OFC), ventral striatum [including the nucleus accumbens (NAcc)], ventral pallidum, and putamen, relative to individuals with a decrease in appetite (n = 16). Additionally, participants in the decreased appetite group had lower activation in the caudal anterior and dorsal mid-insula, compared to participants in the increased appetite group. Although the authors did not examine resting-state functional connectivity (RSFC) group differences, food pleasantness ratings were positively correlated with functional connectivity between the right dorsal mid-insula and two other regions showing group differences in activation during the food stimuli task (the left medial ventral striatum and the left ventral medial prefrontal cortex); additionally, pleasantness ratings were higher in the hyperphagic MDD group than the controls and hypophagic MDD group. Moreover, another study investigated differences in fMRI responsivity to food v. non-food stimuli in participants with remitted MDD who had experienced MDD-attributed appetite changes (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019). For high-calorie food (v. non-food) stimuli, the increased appetite group had greater activation in the right putamen and right anterior insula compared to the decreased appetite group. For high- (v. low-) calorie food stimuli, the increased appetite group showed greater activation than healthy controls in the right and left caudate and the left pallidum, and participants in the decreased appetite group were characterized by greater activation than healthy controls in the left hypothalamus and the right caudate (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019). Collectively, these findings indicate that different appetite phenotypes in MDD are associated with abnormalities within regions critically implicated in reward processing (e.g. NAcc, OFC, putamen, caudate, ventral pallidum) and interoceptive perception (e.g. anterior insula, hypothalamus).
Differences in brain activation to food stimuli have also been investigated in conjunction with metabolic markers (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020). Specifically, in the hyperphagic MDD group, individuals with greater insulin resistance exhibited higher activation to food cues in the insular cortex. In combination with other findings, such as increased levels of immune markers in the increased appetite group, Simmons et al. (Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020) suggested that hyperphagic MDD is associated with metabolic and immune dysfunction, independent of body mass index (BMI).
Existing studies have examined differences in how MDD appetite phenotypes respond to food cues (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019; Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016, Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020). Because brain activity and connectivity are less constrained by specific task-based demands during rest, analyses of resting-state fMRI for hyperphagic and hypophagic MDD can provide important, complementary insights into the neural underpinnings of these presentations, revealing potential differences that exist independent of exposure to food cues. Besides using RSFC to investigate connectivity between regions, the fractional amplitude of low-frequency fluctuations (fALFF) analyses may be helpful to assess spontaneous activity in specific brain regions (Zou et al., Reference Zou, Zhu, Yang, Zuo, Long, Cao and Zang2008), since disorder-related abnormalities may be due to local neuronal activity rather than connectivity between different locations (Egorova, Veldsman, Cumming, & Brodtmann, Reference Egorova, Veldsman, Cumming and Brodtmann2017). These methods have been used to elucidate the neural correlates of MDD (Mulders, van Eijndhoven, Schene, Beckmann, & Tendolkar, Reference Mulders, van Eijndhoven, Schene, Beckmann and Tendolkar2015; Zhou et al., Reference Zhou, Hu, Lu, Zhang, Chen, Gong and Huang2017) and conditions associated with abnormal food intake (Parsons, Steward, Clohesy, Almgren, & Duehlmeyer, Reference Parsons, Steward, Clohesy, Almgren and Duehlmeyer2021). However, no investigations have compared resting-state activity or connectivity between different appetite phenotypes in MDD, despite the potential insights that may be gleaned from such investigations. In particular, the aforementioned resting-state fMRI measures may prove sensitive to detecting the appetitive, metabolic, and reward-related differences that differentiate MDD with hyperphagia from MDD with hypophagia or euphagia, given prior evidence suggesting that resting-state measures can index individual's metabolic state and perceived food pleasantness (Al-Zubaidi, Heldmann, Mertins, Jauch-Chara, & Munte, Reference Al-Zubaidi, Heldmann, Mertins, Jauch-Chara and Munte2018; Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016).
In the present study, we seek to establish a more comprehensive view of the neural underpinnings of MDD appetite and/or weight phenotypes by examining resting-state activity and connectivity. This is important in order to understand whether there are group differences in spontaneous activity and connectivity that exist independent of the response to external food cues. To achieve this, we examined differences in fALFF and RSFC in hyperphagic, hypophagic and euphagic (no change in appetite/weight attributable to the MDD episode) MDD in comparison to a healthy comparison group. To focus on the regions specified in our a priori hypotheses, we first restricted fALFF and RSFC analyses to a mask containing the brain regions reported to show differences across MDD eating phenotypes in prior research, i.e., reward and interoceptive regions. For RSFC analyses, regions exhibiting fALFF group differences were used as seeds, as was the NAcc given its role in reward processing and involvement in food-related disorders and depression (Demos, Heatherton, & Kelley, Reference Demos, Heatherton and Kelley2012; Domingo-Rodriguez et al., Reference Domingo-Rodriguez, Ruiz de Azua, Dominguez, Senabre, Serra, Kummer and Maldonado2020; Lawrence, Hinton, Parkinson, & Lawrence, Reference Lawrence, Hinton, Parkinson and Lawrence2012). Second, we explored fALFF and RSFC (using the same regions as seeds) in the whole brain. Given that RSFC to reward regions (e.g. NAcc) was greater in conditions associated with increased food intake (Parsons et al., Reference Parsons, Steward, Clohesy, Almgren and Duehlmeyer2021) and reduced in conditions associated with decreased food intake (Haynos et al., Reference Haynos, Hall, Lavender, Peterson, Crow, Klimes-Dougan and Camchong2019), we expected the connectivity between the NAcc and other reward and interoceptive regions to vary according to the MDD appetite/weight phenotype, with hyperphagic MDD showing higher connectivity than hypophagic MDD, and the euphagic and healthy comparison groups being in an intermediate position between the other two groups. We expected fALFF in reward and interoceptive regions to vary with respect to appetite/weight phenotype but did not have specific hypotheses regarding directionality given the limited existing literature.
Methods
Participants
Participants (n = 223) were drawn from the multi-site clinical trial Establishing Moderators and Biosignatures of Antidepressant Response for Clinical Care for Depression (EMBARC) (Trivedi et al., Reference Trivedi, McGrath, Fava, Parsey, Kurian, Phillips and Weissman2016). Data were collected at Columbia University, Massachusetts General Hospital, the University of Texas Southwestern Medical Center, and the University of Michigan. For the MDD group, exclusion criteria included lifetime psychotic depressive disorder, bipolar disorder, and any psychotic disorder, as well as current primary anxiety disorder, substance dependence (previous six months), and substance abuse (previous two months). Further, to increase homogeneity, individuals were excluded from the MDD group if they did not score 14 or above on the Quick Inventory of Depressive Symptomatology (QIDS) (Rush et al., Reference Rush, Trivedi, Ibrahim, Carmody, Arnow, Klein and Keller2003). For the healthy comparison group, exclusion criteria included any lifetime Axis I disorder. For all groups, further exclusion criteria included a history of neurological disorders, head injury, major medical illnesses, or current pregnancy. For the present analyses, we additionally excluded individuals reporting current diabetes or whose resting-state fMRI data did not pass quality control (Fig. 1). No participants were on anti-depressant medication in the present analyses, based on data collected at the screening and baseline visits.
Flow chart of participants' inclusion.
For the present analyses, participants with MDD were divided into groups categorized as hyperphagic MDD (i.e. individuals with MDD who experienced an increase in appetite and/or weight during the current MDD episode; n = 77), hypophagic MDD (i.e. individuals with MDD who experienced a decrease in appetite and/or weight during the current MDD episode; n = 66), and euphagic MDD (i.e. individuals with MDD who experienced no significant or detectable changes in appetite and/or weight during the current MDD episode; n = 42); in addition, healthy comparison participants without MDD (n = 38) were included. MDD participants were allocated to groups based on their responses to items assessing appetite and/or weight changes from the MDD section of the mood disorders module of the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-IV) (Kübler, Reference Kübler, Gellman and Turner2013). Specifically, individuals with scores of 3 (i.e. threshold or true) on the appetite and/or weight change item were assigned to either the hyperphagic or hypophagic groups, depending on whether the additional items indicated weight gain / increased appetite or weight loss / decreased appetite, respectively. Individuals with scores of 1 on the appetite and/or weight change item were assigned to the euphagic MDD group. Individuals with a score of 2 (i.e. subthreshold) on the appetite and/or weight change item (n = 23) were excluded from the analyses, given that the additional items indicating directionality of the change were not consistently completed for those participants and therefore sub-threshold hyperphagia/hypophagia groups would have been extremely small.
Procedures
During screening, all participants signed the consent and completed the SCID-IV, demographics, as well as the Mood and Anxiety Symptoms Questionnaire (MASQ) (Watson et al., Reference Watson, Weber, Assenheimer, Clark, Strauss and McCormick1995), the Snaith-Hamilton Pleasure Scale (SHAPS) (Snaith et al., Reference Snaith, Hamilton, Morley, Humayan, Hargreaves and Trigwell1995), and the QIDS. Following screening (range = 1–70 days; mean = 13.1 days), participants presented to the fMRI scan used in the present analyses.
Imaging acquisition and preprocessing of fMRI data
See online Supplemental Information for details on data acquisition, preprocessing, and region of interest (ROI) definition.
Statistical analyses
Demographic and clinical measures
To more fully characterize the groups, we compared them on various demographic and clinical measures. See online Supplemental Material for more information on analyses of demographic variables.
Clinical symptoms were assessed using the MASQ, SHAPS, and QIDS. As in Simmons et al. (Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016), SHAPS items assessing loss of pleasure to food and drink were excluded when calculating the score since previous analyses of food pleasantness showed differences between hyperphagic v. hypophagic MDD (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016). Similarly, QIDS items assessing appetite and weight change were excluded from the calculation of the overall QIDS score. However, for participants with MDD, we used responses to the QIDS appetite and weight change items to investigate whether group differences in fALFF and RSFC analyses were driven by MDD-related changes in appetite and/or weight. Specifically, for MDD participants who correctly followed instructions for completing the QIDS appetite and weight change items, we multiplied the item assessing decreased appetite by −1 and added it to the item assessing increased appetite, resulting in a composite appetite change score that ranged from −3 to 3; the same procedure was used to create a composite weight change score.
Functional activity and connectivity
The Functional Connectivity Toolbox v19.c (CONN) (Whitfield-Gabrieli & Nieto-Castanon, Reference Whitfield-Gabrieli and Nieto-Castanon2012) was used for all fMRI analyses. All analyses included Sex and Site as regressors, since data were collected at different locations and groups significantly differed with respect to sex frequencies. To address multiple comparisons, we used threshold-free cluster enhancement (TFCE) (Smith & Nichols, Reference Smith and Nichols2009) with 1000 permutations (statistical threshold set at p < 0.05, FWE-corrected). Finally, Cohen's effect sizes (d) were calculated for significant findings.
Fractional amplitude of low-frequency fluctuations (fALFF). We first performed mask-restricted analyses that examined group differences in fALFF in regions of interest (the NAcc, lateral OFC, hypothalamus, putamen, caudate, ventral pallidum, and anterior insula). Specifically, we investigated significant clusters within the mask including these regions. Secondly, we performed exploratory analyses examining fALFF group differences across the whole brain. In both analyses, the timeseries of each voxel was transformed to the frequency domain using a Discrete Cosine Transform, and power spectra were calculated. Voxelwise fALFF analysis was conducted by calculating the average square root of power with the 0.008–0.09 Hz frequency band for each voxel and dividing by the total power spectra (Hallquist, Hwang, & Luna, Reference Hallquist, Hwang and Luna2013). fALFF maps were standardized into subject-level Z-score maps.
Seed-based connectivity (SBC) analyses. We first performed mask-restricted analyses that examined group differences in SBC from three seed regions to voxels within regions of interest. Seed regions included the NAcc, as well as the clusters in the lateral OFC and anterior insula that showed significant fALFF differences between groups. Secondly, we performed exploratory whole-brain SBC analyses using the same three regions as seeds to evaluate group differences in connectivity involving the NAcc, lateral OFC, and anterior insula with the rest of the brain.
Sensitivity analyses
Given the evidence that hyperphagic MDD is associated with metabolic dysfunction (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020), the analyses described above were also performed with adjustment for BMI (n = 201 participants with BMI available) as an additional covariate. Additionally, to explore moderation by sex, we examined whether group differences in fALFF and RSFC differed for males and females. Finally, we used Pearson's correlations and multiple regression to examine the relationship between beta values from significant fALFF and RSFC findings and the QIDS appetite and weight change scores (n = 147 participants with change scores available), as well as BMI (see online Supplemental Materials for more information).
Preregistration
Hypotheses and data analysis plans were pre-registered on Open Science Framework (see https://osf.io/fxqdp). See online Supplemental Material for details.
Results
Analyses of demographic and clinical measures
Sample characteristics can be found in Table 1. See online Supplemental Material for further description of demographics and clinical measures.
Demographic and clinical characteristics of participants included in the imaging analyses (N = 223)
Note: p values are comparisons between groups via
a one-way ANOVA (and Tukey's Test for post hoc comparisons when applicable),
b Fisher's Exact tests, or
c Kruskal-Wallis tests.
d Quick Inventory of Depressive Symptomatology (QIDS) overall scores did not include the items measuring appetite/weight change. QIDS appetite change scores were calculated by multiplying the item assessing decreased appetite by −1 and adding it to the item assessing increased appetite, resulting in a composite appetite change score that ranged from −3 to 3; the same procedure was used to create a composite weight change score.
e Using the Structured Clinical Interview for DSM-IV Axis I disorders (Kübler, Reference Kübler, Gellman and Turner2013).
f There was no significant difference between MDD groups for the other clinical measures unless stated otherwise.
fMRI analyses
Unthresholded maps for fMRI analyses can be viewed at https://neurovault.org/collections/11836/.
Fractional amplitude of low-frequency fluctuations (fALFF)
Mask-restricted analyses. Relative to the healthy comparison group, the hyperphagic MDD group was characterized by significantly lower fALFF in the right anterior insula (TFCE = 146.73, p-FWE = 0.018, d (cluster level) = −0.942; Fig. 2), but higher fALFF in the right lateral OFC (TFCE = 125.40, p-FWE = 0.036, d (cluster level) = 0.918; Fig. 2). See Table 2 for further details. No other group differences emerged.
Resting-state fractional amplitude of low-frequency fluctuations (fALFF) in the right anterior insula and right lateral orbitofrontal cortex (OFC). Compared to the healthy comparison group, the hyperphagic major depressive disorder group showed increased fALFF in the right lateral OFC and decreased fALFF in the right anterior insula. Brain slices are presented in accordance with radiological convention (i.e. right hemisphere presented in left side of the image). Models included adjustment for sex and site. (Abbreviations: Eu = euphagic; fALFF = fractional amplitude of low-frequency fluctuations; HC = healthy comparisons; Hyper = hyperphagic; Hypo = hypophagic; MDD = major depressive disorder; OFC = orbitofrontal cortex; R = right).
Clusters exhibiting group differences in mask-restricted or whole-brain analyses of fractional amplitude of low-frequency fluctuations. Models included adjustment for sex and site
FWE, Family-wise error; HC, Healthy comparison group; Hyper, Hyperphagic major depressive disorder group; MR, Mask-restricted; OFC, Orbitofrontal cortex; R, Right; TFCE, Threshold-free cluster enhancement; unc, uncorrected; WB, Whole brain.
a Region with the most voxels within the relevant cluster.
b Only clusters with more than 20 voxels are reported.
c Montreal Neurological Institute (MNI).
Whole-brain analysis. Whole-brain analyses confirmed lower fALFF for hyperphagic MDD relative to the healthy comparison group in the right anterior insula (TFCE = 411.98, p-FWE = 0.037, d (cluster level) = −0.958; Fig. 2). See Table 2 for further details. No other significant differences were found.
Seed-based connectivity
Mask-restricted analyses. The connectivity between the right lateral OFC and the left anterior insula was significantly lower in hyperphagic MDD compared to the healthy comparison group (TFCE = 173.71, p-FWE = 0.045, d (cluster level) = −0.730; Fig. 3). See Table 3 for further details. No significant differences emerged when the NAcc or anterior insula were used as seeds.
Resting-state functional connectivity analyses with the right lateral orbitofrontal cortex (OFC) as seed. Compared to the healthy comparison group, the hyperphagic major depressive disorder group showed decreased resting-state functional connectivity to the left anterior insula and increased connectivity to the left precentral gyrus. Brain slices are presented in accordance with radiological convention (i.e. right hemisphere presented in left side of the image). Models included adjustment for sex and site. (Abbreviations: Eu = euphagic; HC = healthy comparisons; Hyper = hyperphagic; Hypo = hypophagic; L = left; MDD = major depressive disorder; OFC = orbitofrontal cortex; PreCG = precentral gyrus; R = right).
Clusters exhibiting group differences in mask-restricted or whole-brain analyses of seed-based connectivity (with the right lateral OFC as the seed). Models included adjustment for sex and site
FWE, Family-wise error; HC, Healthy comparison group; Hyper, Hyperphagic major depressive disorder group; L, Left; MR, Mask-restricted; OFC, Orbitofrontal cortex; TFCE, Threshold-free cluster enhancement; unc, uncorrected; WB, Whole brain.
a Region with the most voxels within the relevant cluster.
b Only clusters with more than 20 voxels are reported.
c Montreal Neurological Institute (MNI).
Whole-brain analysis. We found significantly higher connectivity between the right lateral OFC and the left precentral gyrus in hyperphagic MDD compared to the healthy comparison group (TFCE = 593.07, p-FWE = 0.047, d (cluster level) = 0.862; Fig. 3). See Table 3 for further details. No significant differences were found for connectivity of the NAcc, or the anterior insula, to the whole brain.
Sensitivity analysis
When BMI was included as a covariate in the analyses above, no significant differences were found among groups for fALFF or RSFC. Also, in the sex moderation analyses, the effect of group did not differ significantly between males and females for fALFF or RSFC (all p values > 0.05).
Finally, beta values were significantly correlated (all p values <0.05), in the expected direction, with both appetite and weight changes scores, with two exceptions (online Supplementary Fig. S2). Beta values for the anterior insula finding from the fALFF mask-restricted analyses were not significantly correlated with weight change scores (p = 0.15), and beta values for the R lateral OFC – L anterior insula finding from SBC mask-restricted analyses were not significantly correlated with appetite change scores (p = 0.077). Correlations between beta values and BMI were in the same direction as the appetite- and weight- change items, but generally smaller and non-significant (online Supplementary Fig. S2). In multiple regressions for appetite (or weight) change scores as a function of the beta values for the various significant fALFF and RSFC findings (see online Supplemental Materials), connectivity between right lateral OFC and left precentral gyrus was the most reliable predictor of appetite changes scores (p < 0.05) and weight change scores (p < 0.01).
Discussion
Investigating different appetite/weight change phenotypes in MDD may reveal important neurobiological differences between groups and help guide more targeted treatments. Specifically, individuals with MDD may be more likely to benefit from interventions that normalize the distinct patterns of resting-state activity and connectivity found to be aberrant for that individual's particular MDD phenotype. We investigated resting-state fALFF and seed-based RSFC in individuals with MDD presenting with different appetite/weight phenotypes. In comparison with the healthy comparison group, individuals with hyperphagic MDD showed: (1) lower fALFF in the right anterior insula and higher fALFF in the right lateral OFC; (2) lower connectivity between the right lateral OFC and the left anterior insula; and (3) higher connectivity between the right lateral OFC and the left precentral gyrus. Further, examination of correlations between the beta values for these significant findings and responses to the appetite and weight change items from the Quick Inventory of Depressive Symptomatology suggests that these differences in fALFF and connectivity are generally correlated with both appetite and weight changes; exceptions were fALFF in the right anterior insula, which was significantly correlated with appetite changes but not weight changes, and connectivity between the right lateral OFC and left anterior insula, which was significantly correlated with weight changes but not appetite changes.
The anterior insula is one of two regions containing the primary gustatory cortex; it is also implicated in interoceptive awareness (Craig, Reference Craig2009), and may play a role in communicating interoceptive changes (Barrett & Simmons, Reference Barrett and Simmons2015). Prior research in current (Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016) and remitted (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019) MDD suggests higher activity in response to the presentation of visual food stimuli in the right anterior insula for individuals with increased (v. decreased) appetite in the context of MDD. Our fALFF finding for the anterior insula extends these prior data, showing that independent of food presentation, individuals with hyperphagic MDD present with decreased spontaneous activity in the anterior insula and that these alterations are correlated with appetite changes. Notably, the anterior insula has been implicated in the pathophysiology of MDD (Sliz & Hayley, Reference Sliz and Hayley2012), and some meta-analyses of resting-state activity suggest that the amplitude of low-frequency fluctuations (ALFF) is elevated in the anterior insula in individuals with MDD (Li et al., Reference Li, Chen, Wu, Zhu, Gu, Zhao and Gong2017; Zhou et al., Reference Zhou, Hu, Lu, Zhang, Chen, Gong and Huang2017). Our findings suggest that the nature of alterations in spontaneous activity in the anterior insula may partly depend on the particular appetite phenotype of MDD.
In addition to lower fALFF in the right anterior insula, hyperphagic MDD also had higher fALFF in the right lateral OFC, compared to the healthy comparison group. The OFC is involved in valuation and decision-making. This includes the representation of stimuli (e.g. food) as rewards, with activation in the OFC encoding, for example, food pleasantness (Simmons et al., Reference Simmons, Rapuano, Ingeholm, Avery, Kallman, Hall and Martin2014) and reward value (Rudebeck & Murray, Reference Rudebeck and Murray2014), as well as tastiness (Londerée & Wagner, Reference Londerée and Wagner2020). Research has also evidenced differences in reward processing between the medial and lateral parts of the OFC (Rolls, Cheng, & Feng, Reference Rolls, Cheng and Feng2020). An early study in primates suggested that neurons in the lateral portion of OFC were activated in response to a food-reward extinction procedure, i.e., when food reward was no longer delivered as expected (Rolls, Reference Rolls2021; Thorpe, Rolls, & Maddison, Reference Thorpe, Rolls and Maddison1983). Additionally, evidence suggests that the lateral OFC may represent unpleasant stimuli, such as unpleasant odors (Rolls, Kringelbach, & de Araujo, Reference Rolls, Kringelbach and de Araujo2003). Relevant to the present study, Simmons et al. (Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016) found increased activation in response to food (v. non-food) stimuli in the right OFC for the increased appetite MDD group in comparison to healthy controls. Also notably, some meta-analyses have found increased ALFF in more medial parts of the OFC for individuals with MDD (Li et al., Reference Li, Chen, Wu, Zhu, Gu, Zhao and Gong2017). Our results for the lateral OFC extend these prior findings, by suggesting that hyperphagic MDD is associated with increased spontaneous activity in the lateral OFC independent of food-related stimuli.
Besides differences in fALFF, the hyperphagic MDD group exhibited differences in RSFC to the right lateral OFC seed region, including lower connectivity with the left anterior insula and greater connectivity with the left precentral gyrus. Until recently, few studies have examined seed-based RSFC to the lateral OFC in MDD (Mulders et al., Reference Mulders, van Eijndhoven, Schene, Beckmann and Tendolkar2015), but recent research suggests that MDD is generally associated with hyperconnectivity between the lateral OFC and diverse brain regions (Cheng et al., Reference Cheng, Rolls, Qiu, Liu, Tang, Huang and Feng2016). Our finding of hypoconnectivity between the lateral OFC and anterior insula suggests that the generally heterogeneous findings on seed-based RSFC for the insula (Mulders et al., Reference Mulders, van Eijndhoven, Schene, Beckmann and Tendolkar2015; Yin et al., Reference Yin, Chang, Wei, Jiang, Zhou, Cui and Tang2018) may be clarified by accounting for different appetite/weight phenotypes of MDD.
It is important to note that when BMI was added as a covariate, the aforementioned differences between hyperphagic MDD and healthy comparison were no longer significant. Given that hyperphagic MDD is associated with current, as well as past (Lamers et al., Reference Lamers, Rhebergen, Merikangas, de Jonge, Beekman and Penninx2012), weight gain in the context of MDD, it is not surprising that the hyperphagic MDD group had a higher BMI than the other MDD groups and healthy comparison participants, especially since many participants with MDD in the present sample have had a substantial history of depression. In fact, since hyperphagic MDD may be one cause of higher BMI in the hyperphagic MDD group, it would be expected that findings for hyperphagic MDD would overlap to some extent with those for BMI. Further, given that hyperphagic MDD may be one cause of higher BMI, adjustment for BMI or matching on BMI status could introduce bias into investigations of group differences, in addition to or instead of reducing bias (Diemer, Hudson, & Javaras, Reference Diemer, Hudson and Javaras2021). The partial overlap of findings for BMI and for MDD-related appetite/weight change is supported by the pattern of correlations for beta values from significant group differences: correlations with BMI were in the same direction, but generally smaller and less significant for BMI than for MDD-related appetite and weight changes.
Indeed, several of the regions (e.g. the insula and OFC) exhibiting differences for the hyperphagic MDD group, which had a mean BMI of 31.4, have been implicated in resting-state studies of obesity (Parsons et al., Reference Parsons, Steward, Clohesy, Almgren and Duehlmeyer2021). Further, our functional connectivity findings are generally in the same direction as those for obesity. For instance, in a recent review of resting-state studies in obesity, Parsons et al. (Reference Parsons, Steward, Clohesy, Almgren and Duehlmeyer2021) reported that obesity is associated with reduced insular and increased OFC functional connectivity, consistent with our respective findings of decreased connectivity of the OFC with the anterior insula and increased connectivity of the OFC with the precentral gyrus, in hyperphagic MDD relative to controls. In contrast, findings regarding resting-state activity in obesity, though few (Parsons et al., Reference Parsons, Steward, Clohesy, Almgren and Duehlmeyer2021), are in the opposite direction as our fALFF findings (i.e. lower anterior insula and higher OFC) for hyperphagic MDD. In our sample, the pattern of results for BMI (see online Supplemental Materials) differed from the findings reported here (i.e. for hyperphagic MDD v. healthy comparison), making it unlikely that findings for hyperphagic MDD v. healthy comparison are attributable solely to BMI differences between those groups. However, more research is needed to disentangle the effects of higher BMI and hyperphagic MDD on resting-state activity and connectivity.
It is also important to note that, despite our relatively large sample size, our analyses did not find significant differences in fALFF between the healthy comparison group and other MDD groups, or between MDD groups, in contrast to studies comparing fALFF between MDD (all subtypes combined) and healthy controls (Li et al., Reference Li, Chen, Wu, Zhu, Gu, Zhao and Gong2017; Zhong, Pu, & Yao, Reference Zhong, Pu and Yao2016; Zhou et al., Reference Zhou, Hu, Lu, Zhang, Chen, Gong and Huang2017) and to studies comparing task-based fMRI activity between MDD appetite phenotypes (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019; Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016), respectively. The absence of differences in fALFF between MDD groups may indicate that measures of spontaneous activity are less sensitive to the appetitive and metabolic differences that distinguish these groups, in contrast to activity relative to food stimuli (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019; Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016). It is also important to note that our analyses did not find any significant differences between groups in functional connectivity to the accumbens seed region, in contrast to studies comparing striatal connectivity between MDD (all subtypes combined) and healthy controls (Furman, Hamilton, & Gotlib, Reference Furman, Hamilton and Gotlib2011; Mulders et al., Reference Mulders, van Eijndhoven, Schene, Beckmann and Tendolkar2015).
Although our study had a relatively large sample and used threshold-free cluster enhancement, a method of analysis that enhances sensitivity (Smith & Nichols, Reference Smith and Nichols2009), and analyses were pre-registered, it also has some limitations. First, the healthy comparison group was relatively small. Second, with respect to the distribution of BMI, the healthy comparison group was not well matched to the MDD groups, especially the hyperphagic MDD group, reducing the precision of adjustment for BMI. Third, our hyperphagic and hypophagic MDD groups were defined based on changes in appetite or weight, which may have introduced heterogeneity into these groups if MDD-related changes in appetite and weight involve partially differing neurobiological alterations. Fourth, groups included two individuals with bulimia nervosa, and they may have included individuals with binge-eating disorder or an eating disorder not otherwise specified, which were not assessed and could have confounded our results. Fifth, the interval between assessment of MDD symptoms and actual scanning was as high as 70 days for some participants. Although research suggests that eating phenotypes in MDD are relatively stable across episodes and multi-year time frames (Lamers et al., Reference Lamers, Rhebergen, Merikangas, de Jonge, Beekman and Penninx2012; Nierenberg, Pava, Clancy, Rosenbaum, & Fava, Reference Nierenberg, Pava, Clancy, Rosenbaum and Fava1996), this does not preclude some temporal variability in appetite and weight symptoms between the screening and baseline imaging visits, and future studies should aim for smaller intervals between screening and scanning. Sixth, participants were not standardized in terms of fasting/eating, and no information was available as to the timing or content of participant's food consumption before the scanning session, which would be important to control in future studies since satiety has been shown to modulate resting-state activity and connectivity (Al-Zubaidi et al., Reference Al-Zubaidi, Heldmann, Mertins, Jauch-Chara and Munte2018). Seventh, due to signal dropout, we were unable to examine fALFF or connectivity for more medial portions of the OFC, although these regions have been implicated in both eating behavior (Rolls, Reference Rolls2021) and MDD (Li et al., Reference Li, Chen, Wu, Zhu, Gu, Zhao and Gong2017). Eighth, although the exact reliability of the fMRI measures used in our analyses is not known, most of these measures likely do not have high test-retest reliability (Noble, Scheinost, & Constable, Reference Noble, Scheinost and Constable2019; Zuo & Xing, Reference Zuo and Xing2014). Low reliability of fMRI measures may have reduced our power to detect differences between groups. For example, larger samples may be needed to detect group differences in NAcc SBC if the reliability for this measure is low. Ninth, sensitivity analyses involving QIDS appetite and weight changes scores should be treated as exploratory since these analyses are not corrected for multiple comparisons. Finally, to our knowledge, the reliability and validity of the appetite and weight change scores are not known, and these scores may not be correlated with more objective measures of appetite and weight change in the context of MDD.
Conclusion
Task-based fMRI studies have shown that appetite phenotypes in MDD are associated with altered activation in response to food stimuli in regions involved in processing interoceptive and reward-related information (Cerit et al., Reference Cerit, Christensen, Moondra, Klibanski, Goldstein and Holsen2019; Simmons et al., Reference Simmons, Burrows, Avery, Kerr, Bodurka, Savage and Drevets2016, Reference Simmons, Burrows, Avery, Kerr, Taylor, Bodurka and Drevets2020). However, little is known about differences in spontaneous activity and connectivity for these phenotypes. Our study adds to prior research by suggesting that, independent of the presentation of food stimuli, hyperphagic MDD is associated with alterations in spontaneous activity and connectivity in the anterior insula and lateral OFC which are generally correlated with changes in both appetite and weight. Although no differences were found for the other appetite/weight change groups, our results suggest that alterations in processing interoceptive and reward-related information may contribute to hyperphagic MDD.
Article information
This work was previously presented as a poster at the Society of Biological Psychiatry 2021 Virtual Meeting, and the abstract was published in the proceedings for the conference (Piccolo, Belleau, Holsen, Pizzagalli, & Javaras, Reference Piccolo, Belleau, Holsen, Pizzagalli and Javaras2021).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291722001398.
Acknowledgements
The authors would like to acknowledge Dr Lisa Nickerson and Dr Thomas Carmody for providing helpful advice and information throughout the development of this study.
Financial support
The EMBARC study was supported by the National Institute of Mental Health under award numbers U01MH092221 (Trivedi, M.H.) and U01MH092250 (McGrath, P.J., Parsey, R.V., Weissman, M.M.). DAP was partially supported by R37 MH068376. KNJ's work on the present study was supported by K23 DK120517. MP's work on the present study was supported by a DocMobility Fellowship from the Swiss National Science Foundation (P1FRP1 188057). LMH's work on the present study was supported by R01 DK104772. ELB was supported by K23 MH122668 and the Klingenstein Third Generation Foundation. This work was supported by the EMBARC National Coordinating Center at UT Southwestern Medical Center, Madhukar H. Trivedi, M.D., Coordinating PI, and the Data Center at Columbia and Stony Brook Universities. The funder has no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflicts of interest
In the last three years, the authors report the following financial disclosures, for activities unrelated to the current research: Dr Weissman: In the last 3 years has received research funding from NIMH, Brain and Behavior Foundation, Templeton Foundation and has received book royalties from Perseus Press, Oxford Press, and APA Publishing and receive royalties on the social adjustment scale from Multihealth Systems for work that she is not involved in. Dr Trivedi: Dr Trivedi reports the following lifetime disclosures: research support from the Agency for Healthcare Research and Quality, Cyberonics Inc., National Alliance for Research in Schizophrenia and Depression, National Institute of Mental Health, National Institute on Drug Abuse, National Institute of Diabetes and Digestive and Kidney Diseases, Johnson & Johnson, and consulting and speaker fees from Abbott Laboratories Inc., Akzo (Organon Pharmaceuticals Inc.), Allergan Sales LLC, Alkermes, AstraZeneca, Axon Advisors, Brintellix, Bristol-Myers Squibb Company, Cephalon Inc., Cerecor, Eli Lilly & Company, Evotec, Fabre Kramer Pharmaceuticals Inc., Forest Pharmaceuticals, GlaxoSmithKline, Health Research Associates, Johnson & Johnson, Lundbeck, MedAvante Medscape, Medtronic, Merck, Mitsubishi Tanabe Pharma Development America Inc., MSI Methylation Sciences Inc., Nestle Health Science-PamLab Inc., Naurex, Neuronetics, One Carbon Therapeutics Ltd., Otsuka Pharmaceuticals, Pamlab, Parke-Davis Pharmaceuticals Inc., Pfizer Inc., PgxHealth, Phoenix Marketing Solutions, Rexahn Pharmaceuticals, Ridge Diagnostics, Roche Products Ltd., Sepracor, SHIRE Development, Sierra, SK Life and Science, Sunovion, Takeda, Tal Medical/Puretech Venture, Targacept, Transcept, VantagePoint, Vivus, and Wyeth-Ayerst Laboratories. Dr Pizzagalli: funding from NIMH, Brain and Behavior Research Foundation, the Dana Foundation, and Millennium Pharmaceuticals; consulting fees from Albright Stonebridge Group, Boehreinger Ingelheim, Compass Pathways, Concert Pharmaceuticals, Engrail Therapeutics, Neumora Therapeutics (former BlackThorn Therapeutics), Neurocrine Biosciences, Otsuka Pharmaceuticals, and Takeda Pharmaceuticals; honoraria from the Psychonomic Society (for editorial work) and Alkermes; stock options from Compass Pathways, Engrail Therapeutics, Neumora Therapeutics (former BlackThorn Therapeutics), and Neuroscience Software. All other authors report no biomedical financial interests or potential conflicts of interest.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
