Sage Journals: Discover world-class research

Abstract

In most positron emission tomography (PET) molecular brain imaging studies, regions of interest have been defined anatomically and examined in isolation. However, by defining regions based on physiology and examining relationships between them, we may derive more sensitive measures of receptor abnormalities in conditions such as major depressive disorder (MDD). Using an average of 52 normalized binding potential maps, acquired using radiotracer [¹¹C]-WAY100635 and full arterial input analysis, we identified two molecular volumes of interest (VOIs) with contiguously high serotonin 1A receptor (5-HT_1A) binding sites: the olfactory sulcus (OLFS) and a band of tissue including piriform, olfactory, and entorhinal cortex (PRF). We applied these VOIs to a separate cohort of 25 healthy control males and 16 males with MDD who received [¹¹C]-WAY100635 imaging. Patients with MDD had significantly higher binding than controls in both VOIs, (p < 0.01). To identify potential homeostatic disruptions in MDD, we examined molecular connectivity, i.e. the correlation between binding of raphe nucleus (RN) 5-HT_1A autoreceptors and post-synaptic receptors in molecular VOIs. Molecular connectivity was significant in healthy controls (p < 0.01), but not in patients with MDD. This disruption in molecular connectivity allowed identification of MDD cases with high sensitivity (81%) and specificity (88%).

Keywords

Molecular connectivity olfactory sulcus piriform cortex raphe nucleus serotonin 1A

Introduction

According to conventional practice, ligand binding in the brain is quantified in anatomically defined regions determined a priori. However, these atlas-based regions may not respect the regional distribution of the binding site in question, resulting in insensitivity for detecting disease-related focal disturbances by positron emission tomography (PET), leading to equivocal results in, for example, PET studies of serotonin 5-HT_1A receptors in major depressive disorder (MDD). Therefore, in our study, we delineate volumes of interest (VOIs) using a data-driven approach based on serotonin 1A (5-HT_1A) receptor distribution, rather than by a priori anatomical landmarks. We are able to delineate such regions from a repository of >50 subjects previously imaged with the radiotracer [¹¹C]-WAY100635 and full arterial input analysis. Mean, fully quantitative binding potential maps generated from these subjects resulted in two clear regions of high 5-HT_1A density that do not overlap completely with anatomically defined regions. This method of VOI delineation presents a more accurate picture of receptor binding since receptor distribution may only comprise part of an anatomical region, or conversely, it may span several adjacent regions.

In addition, we evaluate the relationship between presynaptic 5-HT_1A receptors in the raphe nucleus (RN) and post-synaptic receptors in the aforementioned molecular VOIs. As RN autoreceptors regulate serotonergic transmission in the brain, we would expect post-synaptic receptors to upregulate and downregulate in proportion to the number of autoreceptors. Examining the relationship between 5-HT_1A density in the RN and these postsynaptic regions, therefore, could provide unique insight into receptor homeostasis and responsivity in MDD. We term the coupling of binding observed between the RN and these two sets of molecularly defined VOIs, akin to functional connectivity, molecular connectivity. Combining PET-derived VOI delineation and evaluation molecular connectivity, we evaluate PET data in an innovative way that sheds additional light on the pathophysiological mechanisms of MDD.

MDD is a debilitating psychiatric illness and one of the top causes of years lost to disability worldwide.¹ However, our understanding of its pathophysiology is limited, as evidenced by poor remission rates using conventional treatments.² One of the prevailing models of MDD pathophysiology involves the serotonergic system, and in particular the 5-HT_1A receptor. 5-HT_1A autoreceptors are found on the RN, the major source of serotonin in the central nervous system. When activated, these receptors inhibit RN neurons,³ thus decreasing the overall amount of serotonin released.⁴ Post-synaptic 5-HT_1A receptors are found throughout the brain, particularly in areas such as the hippocampus, insula, and temporal pole.⁵

PET has been used to measure 5-HT_1A receptor density in vivo using radiotracers such as [¹¹C]-WAY100635,⁶ a 5-HT_1A antagonist. In support of 5-HT_1A’s role in MDD, PET studies from our lab have found markedly higher 5-HT_1A binding potential in the RN in depressed human subjects compared to controls.^7,8 However, differences in post-synaptic regions have not been as robust.

One reason for this may be that, while the anatomic boundaries of the RN have been clearly defined by their 5-HT_1A somatodendritic distribution, the proper delineation of post-synaptic receptors has been less clear. Researchers often rely on structural MRI images and atlases generated from these images, such as the Automatic Anatomic Labeling template,⁹ the Desikan-Killiany atlas,¹⁰ and the Destrieux atlas.¹¹ While such labels have proven invaluable, they also present some limitations. The distribution of a receptor may not be clearly defined within an anatomical region. Voxelwise analyses can ameliorate this, but without specific hypotheses of voxel location to examine, statistical power may be compromised. Therefore, for some applications, it may be more beneficial to define post-synaptic VOIs based on their receptor distribution, which we refer to as a molecular VOI, as opposed to a structural VOI.

In a healthy state, a higher level of 5-HT_1A autoreceptors in the RN results in reduced serotonin output to the rest of the brain, affecting post-synaptic 5-HT_1A receptors, which upregulate in a compensatory manner.¹² However, the responsiveness of postsynaptic 5-HT_1A to this RN upregulation is unknown. This can be indirectly probed by examining molecular connectivity, the correlation between RN 5-HT_1A and 5-HT_1A in molecular VOIs. A similar approach has been used in anxiety disorders, in which it was shown that correlation of RN BP_F with post-synaptic BP_F was higher in patients treated with escitalopram than in untreated patients.¹³ This can be interpreted to reflect a restoration of homeostasis, or close coupling between the systems, in the treated anxious brain. It would be valuable to determine if such homeostatic disruptions are present in MDD. Due to the need for sensitive measurements, such an examination is optimally performed in molecular VOIs where there is a known high density of post-synaptic 5-HT_1A. If a disruption is found, this could indicate a connectivity deficit in the serotonergic system.

Our previous work has shown differences in raphe 5-HT_1A binding between patients with MDD and healthy volunteers. While these differences are significant in both sexes, they are greater in men than in women, such that binding differences cannot be reliably used to separate females with and without MDD, while binding differences in males can distinguish diagnostic groups.¹⁴ Post-mortem studies have also shown sex differences in binding.¹⁵ Therefore, to best capture differences, we focused on men in this study. We hypothesize that 5-HT_1A binding in all regions will be higher in MDD than in healthy controls. In addition, we hypothesize that molecular connectivity of the RN with molecular VOIs will be greater in healthy controls than in patients with MDD. As an exploratory measure, we also compare the slope of the linear regression between RN binding potential and molecular VOI binding potential for each group—in this way we can assess not only the goodness of fit (by the molecular connectivity/correlation) of the presynaptic/post-synaptic relationship, but how post-synaptic binding changes with presynaptic binding across participants. Finally, we explore whether differences in molecular connectivity can distinguish men with and without MDD to serve as a biomarker for the disorder, through the use of Euclidean distance from the regression line (D_R) between the RN and molecular VOIs. A better understanding of the interplay between pre and post-synaptic serotonin receptors may help guide better treatment and diagnosis.

Materials and methods

Participants

This study was approved by the Institutional Review Boards of Columbia University Medical Center and the New York State Psychiatric Institute, and abided by the standards of the Helsinki Declaration of 1975. Forty-one male subjects were recruited: 25 were healthy volunteers, while 16 met criteria for MDD. All participants were outpatients recruited from the community through outreach such as flyers and web postings. Data on these subjects have been reported previously^7,8,14,16 in the context of groupwise comparisons of RN binding potential (BP_F) and its potential role in diagnosis and prediction of treatment response. The current analysis is the first of these to use molecular VOIs (see below) and to compare pre and post-synaptic BP_F. Subjects were provided with a detailed description of the study protocol and written informed consent was obtained. Diagnosis was made by consensus of senior clinical faculty based on a structured clinical interview (SCID I) and on past medical information. Functional impairment and depression severity were assessed using the Hamilton Depression Rating Scale (HDRS-17), Beck Inventory, and Global Assessment Scale. Inclusion criteria for subjects with MDD included (1) age 18–65 years, (2) DSM IV criteria for MDD, (3) antidepressant free for at least four years (based on reported differences in 5-HT_1A binding with antidepressant use⁸), (4) no psychotropic medication for at least two weeks excluding benzodiazepines, which were discontinued three days prior to PET scanning, (5) absence of alcohol or substance abuse or dependence, (6) absence of lifetime exposure to MDMA (3,4-methylenedioxymethamphetamine, “ecstasy”), (7) absence of significant medical conditions, (8) absence of psychosis, schizophrenia, or bipolar disorder, and (9) ability to provide informed consent. Except for inclusion criterion (2), inclusion criteria for controls was the same as for MDD, with the addition of (1) absence of an axis I disorder, and (2) absence of a first degree relative with a history of a mood or psychotic disorder. Screening was verified via history and physical exams as well as routine blood and urine tests. Demographics data for participants can be found in Table 1.

Table 1.

Participant demographics^a.

	Control	MDD	P
Number of participants	25	16
Age	40.2 ± 13.9	40.1 ± 14.5	0.99
HDRS score	0.8 ± 1.2	26.3 ± 5.7	<0.01
BMI	24.9 ± 4.2^b	28.2 ± 6.8	0.10
Smoking status	1 (4%)	1 (7%)^c	0.68

Mean ± SD is reported.

Weight was unknown for one control participant.

Smoking status was unknown for one MDD participant.

MDD: major depressive disorder; BMI: body mass index; HDRS: Hamilton depression rating scale.

Radiochemistry and input function measures

[¹¹C]-WAY100635 preparation, as well as measurement of arterial input function, metabolites, and plasma free fraction (f_p) was performed as previously described.¹⁷

Image acquisition and analysis

Magnetic resonance imaging (MRI) and PET data were acquired as previously described.⁸ MRI was performed on either a 1.5 T Signa Advantage or a 3 T Signa HDx system (General Electric, Milwaukee, Wisconsin). PET was performed using an ECAT EXACT HR+ (Siemens/CTI, Knoxville, Tennessee); emission data were collected for 110 min as 20 frames of increasing duration. Image analysis was performed using MATLAB (The Mathworks, Natick, Massachusetts), with extensions to functional magnetic resonance imaging of the brain’s (FMRIB’s) linear image registration tool (FLIRT v.5.2¹⁸), brain extraction tool v1.2,¹⁹ and statistical parametric mapping (SPM5) normalization²⁰ and segmentation routines,²¹ as well as to the advanced normalization tools (ANTs).²² Frame-by-frame rigid body registration was performed to a reference frame in order to correct for subject motion, and a mean of motion-corrected PET image frames was coregistered to each subject’s MRI, using eight possible coregistration options.²³ Each result was visually inspected and the best coregistration was chosen.

Delineation of molecular VOIs

A separate [¹¹C]-WAY100635 study examined binding in 52 healthy controls, both male and female.⁸ Voxel binding maps were calculated using Basis Pursuit and averaged in a standard space. To determine which regions had highest 5-HT_1A binding, we used empirically derived BP_F thresholds in the cortex. This same thresholding technique was also used within the brainstem to generate the RN VOI. It should be noted the BP_F threshold was only used to create the atlas image. The RN and molecular VOIs were then used as an atlas for the current study; these were warped to the individual subject’s PET image using ANTs normalization through the individual’s MRI.

There was a difference in the proportion of controls vs. patients who were scanned on a 1.5 T vs. a 3 T scanner, with 21 controls and 1 patient scanned on a 1.5 T scanner. However, PET-MRI coregistration and VOI placement for each participant were manually checked by a trained technician. Therefore, magnet strength is unlikely to affect molecular VOI placement.

Derivation of outcome measures

Regional distribution volumes (V_T) for ¹¹C-WAY100635 were derived in the RN and molecular VOIs as described previously.⁸ The closest measure to in vitro binding potential for PET is BP_F, or the product of available binding density (B_avail, nM per g of tissue) and affinity (1/K_D, nM per ml of brain). We calculate BP_F as V_T – V_ND/f_P, where V_T is the distribution volume in a given region, V_ND is the distribution volume in a reference region (in this case, the white matter of the cerebellum), and f_P is the free fraction of radiotracer in the plasma. We have described kinetic considerations of this measure elsewhere.¹⁷ Briefly, time activity curves (TACs) measured from PET scans were fit to a two tissue compartment model with plasma-to-tissue rate constants constrained to that of the cerebellar white matter, using an arterial input function. Standard error (SE) was calculated using a bootstrapping technique that incorporates errors in plasma, metabolite, and brain data.²⁴

Region-wise comparison

A weighted linear mixed model was used to compare BP_F in molecular VOIs between healthy controls (HC) and patients with MDD. Log-transformation of BP_F was applied to meet model assumptions of normality and homoscedasticity. Weights were defined as $weight = (\frac{1}{SE (log (BPf))})^{2} = (\frac{BPf}{SE (BPf)})^{2}$ . Three-way interaction terms among diagnostic group (MDD vs. HC), region, and side (right, left) were considered. Correlation among left and right sides within each region and correlation among all regions from the same subject were modeled separately and selected based on Akaike Information Criteria (AIC). Possible correlation structures considered included unstructured (UN), compound symmetry (CS), direct product of UN (UN@UN) and direct product of CS (UN@CS). UN@CS was the best for both models. That is, the correlation between two sides from the same region was modeled as CS and the correlation among all four regions (left and right OLFS, left and right PRF) was modeled as UN. P-values less than 0.05, Bonferroni corrected, were considered statistically significant and analysis was performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

Molecular connectivity

To test molecular connectivity between the RN and molecular VOIs, Pearson correlation was calculated between BP_F in the RN and BP_F in each molecular VOI in all participants. Differences between coefficients in HC and MDD participants were tested using the r.test function in the R package ‘psych’; this function uses a Fisher R to Z transform to determine significant differences between two independent correlation coefficients.

If the linear regression between RN BP_F and molecular VOI BP_F in controls represents the typical relationship between regions, then deviations from this regression could be calculated as the Euclidean distance of a given participant from the control regression line r, which we denote as D_R. In this way, the insight gained by differences in molecular connectivity may also be translated into a clinically useful measure. To this end, we calculated D_R of RN BP_F versus molecular VOI BP_F. As shown in Figure 1, we calculated the vertical (y′) and horizontal (x′) distance from the regression line given the individual participant’s x and y values, where x = RN BP_F and y = BP_F of a molecular VOI. Given that y′, x′, and the distance between them, h, form a right triangle, we can calculate the angle θ

θ = {tan}^{- 1} (x' / y')

Figure 1.

Schematic on calculation of D_R (Euclidean distance of point p from regression line r). If y′ is the vertical distance from r (the regression line for healthy controls) and x′ is the horizontal distance from r, then x′, y′, and h form a right triangle. The angle θ can thus be calculated as tan⁻¹(x′/y′). Using the angle θ, we can then calculate the distance from r, D_R, as the altitude of this triangle using the equation sin θ = D_R/y′, or D_R = y′ sin θ.

Using this value of θ, we can construct a new right triangle with the distance from the point p to the regression line (D_R) as the altitude. Using the formula

\begin{matrix} sin θ = D_{R} / y' \\ \to D_{R} = y' sin θ \end{matrix}

we calculated the distance of a given subject from the control regression line. To determine whether D_R, or deviance from the typical relationship between pre and post-synaptic 5-HT_1A receptors, was related to MDD symptom severity, we measured the correlation between D_R and HDRS scores. Finally, we used the pROC package²⁵ in R²⁶ to construct receiver operating characteristic curves for varying values of D_R. To calculate sensitivity and specificity of predicting diagnosis to a given D_R threshold, pROC uses random sampling with replacement to create 2000 bootstrapped datasets.

We did not explore age differences in the above analyses, as we have not found there to be a relationship between age and binding potential.²⁷

Results

Molecular VOIs

Through our thresholding technique, we were able to identify for the first time two PET-derived regions that did not overlap with discrete anatomical regions. The first of these was a small region lying at the posterior aspect of the olfactory sulcus (OLFS). The second was a broad strip of medial temporal cortex that encompassed piriform, olfactory, and entorhinal cortex (referred to as piriform cortex or PRF). The volumes of these regions were 440 mm³ for left OLFS, 520 mm³ for right OLFS, 4072 mm³ for left PRF, and 3624 mm³ for right PRF. Images of these regions can be found in Figure 2. To determine whether using these regions provided an advantage over traditional atlas-based VOIs, we compared binding in each molecular VOI to its closest structural atlas-based VOI (left and right anterior cingulate for left and right OLFS, respectively, and left and right hippocampus for left and right PRF, respectively) using t-tests. Results indicated that both left and right PRF had significantly higher BP_F than left and right hippocampus (p < 0.01, Bonferroni corrected), but higher binding in the left and right OLFS compared to left and right anterior cingulate was nonsignificant (p = 0.08 and 0.32 for left and right respectively, Bonferroni corrected, Figure 2).

Figure 2.

Molecular VOIs labeled in sagittal (a), coronal (b), and oblique views (c) as a 3D image projection of an averaged [¹¹C]-WAY100635 image. OLFS: olfactory sulcus; PRF: piriform cortex. (d) Molecular VOIs show greater binding potential than nearby structurally defined regions. ACN: anterior cingulate; HIP: hippocampus.

BP_F comparison in OLFS and PRF

In both molecular VOIs, OLFS and PRF, bilaterally, males with MDD had higher BP_F than did healthy controls (p < 0.01, Bonferroni corrected, see Figure 3). Average BP_F in subjects with MDD was 82% greater in left OLFS, 54% greater in right OLFS, 67% greater in left PRF, and 67% greater in right PRF, than the average BP_F of healthy controls.

Figure 3.

In males, patients with major depressive disorder (MDD) show consistently greater 5-HT_1A binding in the raphe and molecular VOIs compared to controls. Error bars represent standard error. *p < 0.01, Bonferroni corrected.

Molecular connectivity

Molecular connectivity (correlation coefficients) between the RN and the four molecular VOIs in healthy male controls ranged from 0.73 in left OLFS to 0.86 in right PRF. Further, the molecular connectivity between RN BP_F and molecular region BP_F was significant in all cases (p < 0.01 for all regions, Bonferroni corrected). In contrast to these significant correlations in healthy controls, in males with MDD, molecular connectivity was lower, ranging between 0.42 in left OLFS to 0.54 in right OLFS. Further, all the correlations were nonsignificant after Bonferroni correction (see Table 2 and Figure 4). Fisher R to Z transforms showed that molecular connectivity was significantly different between male controls and patients with MDD only in right PRF, though this did not survive multiple comparisons correction (p = 0.12). In addition to the differences in correlation, the slope of the fit line of RN to molecular VOIs (which describes the relationship between pre and post-synaptic 5-HT_1A density in controls vs. MDD) was examined. In control males, slopes of the regression lines were all greater than one, ranging from 2.39 to 3.66. In males with MDD, by contrast, slopes were less than one, ranging from 0.44 to 0.69. ANCOVA revealed slopes between groups in all molecular VOIs to be significantly different (p < 0.01, Bonferroni corrected).

Table 2.

Molecular connectivity.

Region	Group	R	R²	%Difference in R²	P ^a
Left olfactory sulcus	Control	0.73*	0.54	67%	0.68
Left olfactory sulcus	MDD	0.42	0.18	67%	0.68
Right olfactory sulcus	Control	0.78*	0.61	52%	0.84
Right olfactory sulcus	MDD	0.54	0.29	52%	0.84
Left piriform cortex	Control	0.77*	0.60	70%	0.40
Left piriform cortex	MDD	0.42	0.18	70%	0.40
Right piriform cortex	Control	0.86*	0.75	71%	0.12
Right piriform cortex	MDD	0.47	0.22	71%	0.12

Bonferroni corrected, determined by Fisher R to Z transform.

P < 0.05, Bonferroni corrected.

MDD: major depressive disorder.

Figure 4.

(a–h) Molecular connectivity (correlations) between raphe nucleus BP_F and BP_F in left OLFS (a,e), right OLFS (b,f), left PRF (c,g), and right PRF (d,h) in males. Outliers (defined as outside the interval [75% quantile ± 1.5*interquartile range] in either the raphe nucleus or in molecular ROIs) are marked in red. (i) Receiver operating characteristic (ROC) curve for the use of distance from the regression line (D_R), using the linear regression between raphe binding potential (BP_F) and right piriform cortex (PRF) BP_F, as a diagnostic test in major depressive disorder.

Euclidean distance from regression line (D_R)

Patients with MDD had significantly higher D_R (the distance of a given participant from the “typical” autoreceptor to post-synaptic receptor relationship) than healthy controls (p < 0.01 for all regions; see Table 3). Further, in right PRF (the region with the largest difference in molecular connectivity between groups), D_R correlated negatively with MDD severity, though the correlation did not reach significance (r = −0.35, p = 0.18). When D_R in this region was used to predict diagnosis, it performed with 88.00 ± 12% specificity and 81.25 ± 20% sensitivity at a threshold D_R of 4.57 mL/cc in this region (see Figure 4).

Table 3.

Distance from regression line (D_R).

Region	Side	Control distance from regression line (SE)	MDD Distance from regression line (SE)	P
Olfactory Sulcus	Left	4.81 (1.25)	17.49 (3.72)	<0.01*
Olfactory Sulcus	Right	4.51 (0.95)	16.70 (3.54)	<0.01*
Piriform Cortex	Left	4.06 (1.15)	15.88 (3.55)	<0.01*
Piriform Cortex	Right	3.17 (0.78)	14.72 (3.51)	<0.01*

P < 0.05.

SE: standard error, defined as (standard deviation ÷ √(n)); MDD: major depressive disorder.

Discussion

As the RN is the central output of serotonin to the remainder of the central nervous system, we would expect pre and post-synaptic 5-HT_1A binding to be correlated in a healthy state. However, the relationship between these two entities in MDD has been unknown. Using PET-derived measures, we were able to assess molecular connectivity, i.e. the relationship between pre and post-synaptic 5-HT_1A receptors in molecularly defined regions for the first time. Importantly, we find this connectivity to be disrupted in MDD in men; RN autoreceptor binding potential (BP_F) is not significantly correlated with that in molecular VOIs, as it is in controls. As BP_F is significantly higher in both the RN and molecular VOIs in MDD, such a relationship may not have been observable through traditional analyses. The significance of this finding is that reduction of molecular connectivity in MDD suggests a disruption in serotonin transmission and homeostasis.

It is likely that both molecular VOIs are connected to the RN via the medial forebrain bundle, and that while the OLFS is innervated by projections from the median RN, PRF is innervated by projections from the dorsal RN.²⁸ That the OLFS lies directly between the subgenual ACC and ventromedial PFC, both heavily implicated in MDD,^29,30 further supports its involvement in depressive symptomology. Both molecular VOIs are largely implicated in olfactory function, a sensory modality that has not been well studied in MDD. However, removal of the olfactory bulb in rats is used as a model for MDD.^31,32 In addition, clinical studies have shown differences in olfactory sensitivity in MDD—two studies showed a decrease,^33,34 while one showed an increase.³⁵ A recent MRI study showed lower OLFS sulcal depth in MDD.³⁶ Using these molecularly defined VOIs could allow greater exploration of the role olfaction plays in MDD, and whether these effects pertain to sensation in general.

We found that 5-HT_1A binding in the RN was significantly correlated with 5-HT_1A binding in molecular VOIs in healthy controls, but not in the MDD group, which suggests dysregulation. This has implications for future MDD monitoring and treatment and further suggests that examining either pre or post-synaptic receptors in isolation (as is most often performed) may not yield a complete picture. Although a direct comparison of molecular connectivity across groups did not reach statistical significance, there are several reasons to investigate this further. (1) In an analysis with outliers (marked in red in Figure 4) removed, molecular connectivity differences were accentuated, with bilateral PRF showing significantly greater connectivity in control males than males with MDD (p = 0.02 and 0.01 for left and right PRF, respectively, Bonferroni corrected). In future analyses with larger cohorts, such outliers may not significantly affect overall connectivity findings. (2) Molecular connectivity differences in right PRF were significant before multiple comparisons correction, indicating that this region would be a good a priori choice for future studies. This region in particular may be valuable to study for several reasons. (a) It is large and therefore not as susceptible to partial volume effects³⁷ as smaller regions may be. (b) As a composite region including olfactory, piriform, and entorhinal cortex, it may have far reaching neural effects; entorhinal cortex in particular is known to have widespread afferent fibers from the cortex³⁸ and is critical in sending efferent fibers to the hippocampus via the perforant pathway.³⁹ (3) Molecular connectivity slopes were significantly different between healthy controls and patients with MDD, indicating an altered relationship between pre and post-synaptic receptors in molecular VOIs. It is important to note that there need not be a difference in correlation (molecular connectivity) for this to be true—two groups can have an equal goodness of fit but have differing linear relationships. Moreover, correlations may be more sensitive to noise and variability than regression slopes—therefore, a difference in slopes between diagnostic groups strongly suggests that there is a real difference in the relationship between pre and post-synaptic receptors in MDD.

It is plausible that molecular connectivity differences are the result of differences in structural connectivity of serotonergic neurons between the RN and molecular VOIs. Unfortunately, there is relatively little agreement in the literature with respect to diffusion tensor imaging (DTI) studies in MDD—many regions of the brain show differences in structural connectivity in MDD, but these regions are not consistently implicated. The most common of these regions are areas of frontal and temporal cortex,^40,41 which partially validate our current findings, as our molecular VOIs lie in these regions. In a large multi-site study, our group found no differences in structural connectivity between depressed patients and healthy controls, though fractional anisotropy of a strip of right medial orbitofrontal cortex correlated negatively with depression severity.⁴² This region lies directly anterior to, and may overlap with, the right OLFS. However, a previous study from our lab showed that 5-HT_1A and serotonin transporter levels do not correlate. If transporter levels are taken as a proxy of axonal density, one potential implication of this is that autoreceptor density and innervation density might not share a direct relationship.⁴³ A DTI study involving these regions in a larger sample would clarify the relationship between molecular and structural connectivity—we have performed similar DTI studies between the RN and anatomical regions.^44,45 In addition, these regions could be assessed with functional connectivity; while molecular VOIs have not been examined in this context, there have been studies examining functional connectivity of the RN in healthy controls.^46,47

A second possibility is that reduced molecular connectivity is due to a disproportionately higher BP_F in the RN with respect to molecular VOIs in MDD, which may be dependent on receptor upregulation alone, independent of structural connectivity or innervation density. This explanation is more consistent with our finding of a lack of association between 5-HT_1A and serotonin transporter. Based on preclinical studies, we hypothesize that post-synaptic changes occur as a result of presynaptic changes. In this case, in healthy controls, regression slopes indicate that a difference in 5-HT_1A autoreceptors between participants would correspond to a two to threefold difference in post-synaptic receptors. This was the opposite in MDD; in these patients, a difference in autoreceptors would only correspond to around half of that difference post-synaptically. Even though all regions examined had greater binding potential in subjects with MDD than in controls, these differences in molecular connectivity slope indicate that BP_F differences are not uniform; compensatory 5-HT_1A upregulation in MDD may not be proportional to increases in 5-HT_1A autoreceptor density.

As molecular connectivity is not correlated with depression severity, it may not be a causative mechanism of MDD. However, it may still serve as a biomarker—to that end, we explored the possibility of using molecular connectivity for diagnostic purposes through the use of Euclidean distance from the presynaptic/postsynaptic 5-HT_1A regression in controls, D_R. This measure can give an indication of how much an individual deviates from the typical presynaptic/postsynaptic receptor relationship. Males with MDD had much higher D_R values than controls in all regions, suggesting that D_R may serve as a biomarker. Our lab has previously explored the potential for a biomarker in the RN;¹⁴ in males, RN 5-HT_1A thresholds showed high specificity and sensitivity for diagnosing MDD. Our current results indicate both left and right PRF show 81% sensitivity and 88% specificity. This is a slight increase from our previous study. Therefore, a biomarker that examines both pre and post-synaptic receptors may be more efficacious, and provide more information about serotonergic function, than one that only examines autoreceptor binding. The high sensitivity and specificity of such a biomarker exceed the diagnostic accuracy of current clinical measures,⁴⁸ further supporting its use—using D_R, we may be able to study subtypes of MDD, evaluate targeted therapies, and identify people at increased risk.⁴⁹ Should molecular connectivity show a relationship with DTI, it can also open the door for less invasive and costly surrogates.

There are limitations to this study. Our sample consists of 25 healthy controls and 16 patients with MDD; our findings should be replicated in a larger sample. There is also a possibility that the smaller sample size of the MDD group relative to the control group contributes to the lack of significance in molecular connectivity, though it should be noted that the correlation coefficient would not be expected to change given the variance in this group. In addition, it would be useful to examine sex differences in molecular connectivity. Given evidence that sex hormones may play a role in 5-HT_1A expression and action,^50,51 future investigations in molecular connectivity should record and covary for differences in menstrual cycle at the time of PET scans.

An important caveat to our study of WAY100635 is that this receptor also binds to the dopamine D4 receptor, though at lower affinities than 5-HT_1A.⁵² As D4 has been found in areas such as the entorhinal cortex,^53,54 which comprises part of our PRF region, this may have an effect on PRF binding potential. The close correspondence of binding in controls between PRF and RN, which does not have significant D4 binding,⁵⁴ gives us confidence that our findings are based on the serotonin system. However, further study of the dopamine D4 receptor in MDD may provide insight that the serotonin system alone cannot elucidate.

Conclusions

We have identified, for the first time, two molecular VOIs, defined by serotonin 1A (5-HT_1A) receptor distribution, that show significantly higher binding in MDD. In studying the association between presynaptic and post-synaptic 5-HT_1A in the RN and these molecular VOIs in males, which we term molecular connectivity, we found that while RN autoreceptors are tightly correlated with post-synaptic receptors in healthy controls, they are nonsignificantly correlated in MDD. Further, on average, healthy controls had two to three times greater molecular connectivity slopes than patients with MDD. Examining PET-derived data in this way suggests that structural and compensatory mechanisms of post-synaptic receptor regulation are compromised in MDD. Individual differences in molecular connectivity are so large that they may even serve as a biomarker for the disorder. Finally, in applying molecular connectivity to other disorders, we may be able to identify patterns of disruption that we would not be able to examine through analysis of single regions.

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work was supported by the following grants: F30MH109412 (PI: Pillai), R01MH40695 (PI: Mann), P50MH062185 (PI: Mann), R01MH074813 (PI: Parsey), R01MH090276 (PI: Parsey), and K01MH091354 (PI: DeLorenzo) awarded by the National Institutes of Health, a Clinical and Translational Science Award (CTSA) from Columbia University, and grants from the American Foundation for the Prevention of Suicide (AFSP) and the National Alliance for Research on Schizophrenia and Depression (NARSAD).

Acknowledgments

We acknowledge the biostatistical support provided by the Biostatistical Consulting Core at School of Medicine, Stony Brook University, and the anatomical support by Dr. Mary F. Kritzer and Dr. Efrain Azmitia in helping us identify molecular VOIs. We also acknowledge the support in patient recruitment carried out by Dr. Greg Sullivan.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Oquendo receives royalties for the use of the Columbia Suicide Severity Rating Scale. Her family owns stock in Bristol Myers Squibb. Mann receives royalties for commercial use of the C- SSRS from the Research Foundation for Mental Hygiene and has stock options in Qualitas Health, a startup company making omega-3-fatty-acid products. Pillai, Zhang, Yang, Parsey, and DeLorenzo report no biomedical financial interests or potential conflict of interest.

Authors’ contributions

Rajapillai Pillai processed and analyzed the data and wrote the manuscript. Mengru Zhang and Jie Yang performed statistical analysis and wrote relevant sections of the methods. J. John Mann, Maria A. Oquendo, and Ramin V. Parsey acquired PET data. Christine DeLorenzo oversaw this project and provided feedback on writing and analysis. All authors read and approved of this manuscript.

References

Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015; 386: 743–800.

Hierholzer

. Remission rates for depression in STAR*D study. Am J Psychiatry 2006; 163: 1293; author reply 1293–1294.

Sprouse

Aghajanian

. Electrophysiological responses of serotoninergic dorsal raphe neurons to 5-HT1A and 5-HT1B agonists. Synapse 1987; 1: 3–9.

Blier

Pineyro

el Mansari

, et al. Role of somatodendritic 5-HT autoreceptors in modulating 5-HT neurotransmission. Ann N Y Acad Sci 1998; 861: 204–216.

Varnas

Halldin

Hall

. Autoradiographic distribution of serotonin transporters and receptor subtypes in human brain. Hum Brain Mapp 2004; 22: 246–260.

Forster

Cliffe

Bill

, et al. A pharmacological profile of the selective silent 5-HT1A receptor antagonist, WAY-100635. Eur J Pharmacol 1995; 281: 81–88.

Parsey

Oquendo

Ogden

, et al. Altered serotonin 1A binding in major depression: a [carbonyl-C-11]WAY100635 positron emission tomography study. Biol Psychiatry 2006; 59: 106–113.

Parsey

Ogden

Miller

, et al. Higher serotonin 1A binding in a second major depression cohort: modeling and reference region considerations. Biol Psychiatry 2010; 68: 170–178.

Tzourio-Mazoyer

Landeau

Papathanassiou

, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 2002; 15: 273–289.

10.

Desikan

Segonne

Fischl

, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage 2006; 31: 968–980.

11.

Destrieux

Fischl

Dale

, et al. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage 2010; 53: 1–15.

12.

Mann

. Neurobiology of suicidal behaviour. Nat Rev Neurosci 2003; 4: 819–828.

13.

Hahn

Lanzenberger

Wadsak

, et al. Escitalopram enhances the association of serotonin-1A autoreceptors to heteroreceptors in anxiety disorders. J Neurosci 2010; 30: 14482–14489.

14.

Kaufman

Sullivan

Yang

, et al. Quantification of the serotonin 1A receptor using PET: identification of a potential biomarker of major depression in males. Neuropsychopharmacology 2015; 40: 1692–1699.

15.

Boldrini

Underwood

Mann

, et al. Serotonin-1A autoreceptor binding in the dorsal raphe nucleus of depressed suicides. J Psychiatr Res 2008; 42: 433–442.

16.

Miller

Brennan

Ogden

, et al. Elevated serotonin 1A binding in remitted major depressive disorder: evidence for a trait biological abnormality. Neuropsychopharmacology 2009; 34: 2275–2284.

17.

Parsey

Slifstein

Hwang

, et al. Validation and reproducibility of measurement of 5-HT1A receptor parameters with [carbonyl-11C]WAY-100635 in humans: comparison of arterial and reference tisssue input functions. J Cereb Blood Flow Metab 2000; 20: 1111–1133.

18.

Jenkinson

Smith

. A global optimisation method for robust affine registration of brain images. Med Image Analy 2001; 5: 143–156.

19.

Smith

. Fast robust automated brain extraction. Hum Brain Mapp 2002; 17: 143–155.

20.

Ashburner

Friston

. Nonlinear spatial normalization using basis functions. Hum Brain Mapp 1999; 7: 254–266.

21.

Ashburner

Friston

. Unified segmentation. Neuroimage 2005; 26: 839–851.

22.

Avants

Tustison

Song

, et al. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage 2011; 54: 2033–2044.

23.

DeLorenzo C, Klein A, Mikhno A, et al. A new method for assessing PET-MRI coregistration. In: Medical Imaging 2009: Image Processing (Vol. 7259, p. 72592W). International Society for Optics and Photonics.

24.

Ogden

Tarpey

. Estimation in regression models with externally estimated parameters. Biostatistics 2006; 7: 115–129.

25.

Robin

Turck

Hainard

, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinform 2011; 12: 77.

26.

Team

. R: A language and environment for statistical computing, Vienna, Austria: R Foundation for Statistical Computing, 2013.

27.

Parsey

Oquendo

Simpson

, et al. Effects of sex, age, and aggressive traits in man on brain serotonin 5-HT1A receptor binding potential measured by PET using [C-11]WAY-100635. Brain Res 2002; 954: 173–182.

28.

Azmitia

Segal

. An autoradiographic analysis of the differential ascending projections of the dorsal and median raphe nuclei in the rat. J Comp Neurol 1978; 179: 641–667.

29.

Drevets

. Functional anatomical abnormalities in limbic and prefrontal cortical structures in major depression. Prog Brain Res 2000; 126: 413–431.

30.

Mayberg

Lozano

Voon

, et al. Deep brain stimulation for treatment-resistant depression. Neuron 2005; 45: 651–660.

31.

Kelly

Wrynn

Leonard

. The olfactory bulbectomized rat as a model of depression: an update. Pharmacol Ther 1997; 74: 299–316.

32.

Song

Leonard

. The olfactory bulbectomised rat as a model of depression. Neurosci Biobehav Rev 2005; 29: 627–647.

33.

Pause

Miranda

Goder

, et al. Reduced olfactory performance in patients with major depression. J Psychiatr Res 2001; 35: 271–277.

34.

Pause

Raack

Sojka

, et al. Convergent and divergent effects of odors and emotions in depression. Psychophysiology 2003; 40: 209–225.

35.

Gross-Isseroff

Luca-Haimovici

Sasson

, et al. Olfactory sensitivity in major depressive disorder and obsessive compulsive disorder. Biol Psychiatry 1994; 35: 798–802.

36.

Takahashi

Nishikawa

Yucel

, et al. Olfactory sulcus morphology in patients with current and past major depression. Psychiatry Res 2016; 255: 60–65.

37.

Erlandsson

Buvat

Pretorius

, et al. A review of partial volume correction techniques for emission tomography and their applications in neurology, cardiology and oncology. Phys Med Biol 2012; 57: R119–R159.

38.

Van Hoesen

Pandya

. Some connections of the entorhinal (area 28) and perirhinal (area 35) cortices of the rhesus monkey. I. Temporal lobe afferents. Brain Res 1975; 95: 1–24.

39.

Van Hoesen

Pandya

. Some connections of the entorhinal (area 28) and perirhinal (area 35) cortices of the rhesus monkey. III. Efferent connections. Brain Res 1975; 95: 39–59.

40.

Liao

Huang

, et al. Is depression a disconnection syndrome? Meta-analysis of diffusion tensor imaging studies in patients with MDD. J Psychiatry Neurosci 2013; 38: 49–56.

41.

Sexton

Mackay

Ebmeier

. A systematic review of diffusion tensor imaging studies in affective disorders. Biol Psychiatry 2009; 66: 814–823.

42.

Olvet

Delaparte

Yeh

, et al. A comprehensive examination of white matter tracts and connectometry in major depressive disorder. Depress Anxiety 2016; 33: 56–65.

43.

Strupp-Levitsky

Miller

Rubin-Falcone

, et al. Lack of association between the serotonin transporter and serotonin 1A receptor: an in vivo PET imaging study in healthy adults. Psychiatry Res 2016; 255: 81–86.

44.

Pillai

Malhotra

Rupert

, et al. Relations between cortical thickness, serotonin 1A receptor binding, and structural connectivity: a multimodal imaging study. Hum Brain Mapp 2018; 39: 1043–1055.

45.

Delorenzo

Delaparte

Thapa-Chhetry

, et al. Prediction of selective serotonin reuptake inhibitor response using diffusion-weighted MRI. Front Psychiatry 2013; 4: 5.

46.

Beliveau

Svarer

Frokjaer

, et al. Functional connectivity of the dorsal and median raphe nuclei at rest. Neuroimage 2015; 116: 187–195.

47.

Bar

de la Cruz

Schumann

, et al. Functional connectivity and network analysis of midbrain and brainstem nuclei. Neuroimage 2016; 134: 53–63.

48.

Williams

Jr. Pignone

Ramirez

, et al. Identifying depression in primary care: a literature synthesis of case-finding instruments. Gen Hosp Psychiatry 2002; 24: 225–237.

49.

Peterson

Weissman

. A brain-based endophenotype for major depressive disorder. Annu Rev Med 2011; 62: 461–474.

50.

Estrada-Camarena

Fernandez-Guasti

Lopez-Rubalcava

. Participation of the 5-HT1A receptor in the antidepressant-like effect of estrogens in the forced swimming test. Neuropsychopharmacology 2006; 31: 247–255.

51.

Raap

DonCarlos

Garcia

, et al. Estrogen desensitizes 5-HT(1A) receptors and reduces levels of G(z), G(i1) and G(i3) proteins in the hypothalamus. Neuropharmacology 2000; 39: 1823–1832.

52.

Chemel

Roth

Armbruster

, et al. WAY-100635 is a potent dopamine D4 receptor agonist. Psychopharmacology 2006; 188: 244–251.

53.

Lahti

Roberts

Cochrane

, et al. Direct determination of dopamine D4 receptors in normal and schizophrenic postmortem brain tissue: a [3H]NGD-94-1 study. Mol Psychiatry 1998; 3: 528–533.

54.

Primus

Thurkauf

, et al. II. Localization and characterization of dopamine D4 binding sites in rat and human brain by use of the novel, D4 receptor-selective ligand [3H]NGD 94-1. J Pharmacol Exp Ther 1997; 282: 1020–1027.

Molecular connectivity disruptions in males with major depressive disorder

Abstract

Keywords

Introduction

Materials and methods

Participants

Radiochemistry and input function measures

Image acquisition and analysis

Delineation of molecular VOIs

Derivation of outcome measures

Region-wise comparison

Molecular connectivity

Results

Molecular VOIs

BPF comparison in OLFS and PRF

Molecular connectivity

Euclidean distance from regression line (DR)

Discussion

Conclusions

Footnotes

Funding

Acknowledgments

Declaration of conflicting interests

Authors’ contributions

References

BP_F comparison in OLFS and PRF

Euclidean distance from regression line (D_R)