Measurement of Human Brown Adipose Tissue Volume and Activity Using Anatomical MRI and Functional MRI

Abstract

The existence of brown adipose tissue (BAT) in humans has previously been assessed in vivo via sequential 18F-FDG PET/CT imaging. We developed a MRI protocol to detect BAT mass based on BAT’s property of having higher water-to-fat ratio than white adipose tissue (WAT). We showed that the signal contrast obtained between water-saturation and without water-saturation was higher in BAT than in WAT in fast spin echo images and in T2-weighted images. The water-to-fat ratio was also higher in BAT via contrasting the water and fat images of the Dixon method. The MRI measured volume and location of BAT was similar to PET/CT results in the same subjects. In addition, we also demonstrated that cold challenges (14 °C) led to significant fMRI BOLD signal increases in BAT.

Brown adipose tissue (BAT) is a unique organ whose role is to consume fat calories by generating heat in response to cold exposure (1–4). BAT energy expenditure may also be used to counterbalance the accumulation of stored energy in white adipose tissue (WAT), making induction of BAT thermogenesis a promising target for treating obesity and metabolic disease. Unfortunately, these goals cannot be met until BAT volume and activity in humans can be quantified, which has been a challenge since the only currently available method for imaging BAT uses a combination¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT), which gives an adequate estimate of activity, but is not sensitive to BAT in the thermoneutral state. Using rodent models, we recently showed that a combination of MRI and functional magnetic resonance imaging (fMRI) can be used to measure BAT volume and activity (5). However, the principal rodent interscapular BAT depot lies in a defined anatomical location and is composed homogeneously of brown adipocytes. By contrast, human BAT is a heterogeneous mixture of white and brown adipocytes in multiple sites, making its imaging particularly challenging. In the current work, we demonstrate how MRI/fMRI can be used in humans to quantify BAT volume under thermoneutral conditions and also monitor the dynamic changes in BAT blood flow in response to cold stimulation.

The appeal of MRI is that it has the potential to distinguish BAT and WAT based on two independent physiological factors: BAT has higher water-to-fat ratio than WAT and BAT has a high density of mitochondria and blood vessels (6–9). Efforts to evaluate animal BAT volume and distribution using in vivo techniques have employed MR spectroscopy (MRS) (6,10,11), MR imaging (MRI), and a combination of the two, such as chemical shift imaging (CSI) (7). These MR studies were based on the fact that BAT differs from WAT by the fraction of saturated fatty acids, diunsaturated fatty acids (6), and water content (6,7,11). However, a reliable MR protocol to measure BAT volume and distribution in living adult humans is still lacking. Here we used sequences available in most clinical MR scanners to access BAT location and function. This approach is preferable to¹⁸F-FDG PET/CT, which is limited by several factors, a critical one being the requirement to activate the tissue by exposing the human subject to some kind of challenge such as a cold stress. Thus, PET/CT permits detection of an unknown and variable fraction of the total BAT volume and reveals only the tissue avidly involved in glucose utilization, but none of the other potential substrates, such as fatty acids (12,13). On the other hand, MRI is capable of assessing BAT volume using water-fat separation and monitoring BAT activity through the blood oxygenation level dependent (BOLD) mechanism continuously throughout the period of cold exposure.

MATERIALS AND METHODS

Five healthy volunteers were recruited for MRI/fMRI studies. Four out of the five subjects had a previous¹⁸F-FDG PET/CT study demonstrating detectable cold-activated BAT in the cervical, supraclavicular, and upper thoracic depots (14) (Supplementary table 1). All MRI studies were carried out using a 1.5 Tesla Avanto “TIM” system (Siemens). The test subject wore a temperature-controlled water circulation cooling vest (Polar Products Inc, Akron, Ohio). The circulating water was set to room temperature (22–24 °C) during BAT anatomical scans and during baseline period of the fMRI scans. The circulating water was switched to 13–16 °C to provide mildly cold stimulation during fMRI scans. In euthyroid volunteers, the temperature used in the vest does not induce shivering, which was confirmed by regular questioning the subjects during the scan.

MRI anatomical scans for locating BAT

All anatomical scans were acquired at spatial resolution of 1.48mm×1.48mm×2.5mm. BAT mass was assessed via two methods: water-saturation efficiency, and water-to-fat contrast ratio. As the high water-to-fat ratio in BAT makes a water suppression-based “water-saturation” (WS) routine more effective in BAT than in WAT (15). Thus, BAT mass was segmented from WAT mass by contrasting signal intensity with and without water-suppression (Equation 1). WS contrast was obtained from a pair of multi-slice fast spin echo (FSE) images (TR 3000ms, effective TE 98ms) with and without water-saturation, and also from a pair of T2 weighted (T2W) images (TR 1990ms, TE 8.4ms) with and without water-saturation. Only the fat areas with significant WS contrast (>10% WS contrast) from both FSE and T2W were regarded as BAT to exclude false signal contrast such as from blood vessels.

$$\mathrm{W}\mathrm{S}\text{contrast}=[\mathrm{S}\mathrm{I}(\mathrm{w}/\mathrm{o}\mathrm{W}\mathrm{S})-\mathrm{S}\mathrm{I}(\mathrm{w}\mathrm{W}\mathrm{S})]/\mathrm{S}\mathrm{I}(\mathrm{w}/\mathrm{o}\mathrm{W}\mathrm{S})\times 100(\%)$$ eq. 1

In addition to WS contrast, BAT was distinguished from WAT using Dixon’s method (TR 6.59ms, TEs 2.38ms and 4.76 ms) (16). Dixon method separates fat from water by acquiring two images in the same scan, one at an echo time when fat and water signals are in-phase, and another when fat and water signals are out-of-phase. Water-only and fat-only images are then generated mathematically from these two sets of images. Since water content is higher in BAT than in WAT, BAT mass thus has higher water-to-fat contrast (Equation 2):

$$\text{water-fat contrast}=\mathrm{S}{\mathrm{I}}_{\text{water}}/\mathrm{S}{\mathrm{I}}_{\text{FAT}}$$ eq. 2

We tested the water-to-fat contrast method in three subject to identify BAT mass at thermoneutral state.

FSE and T2W were post-registered to the Dixon images using FLIRT, a rigid registration routine in FsFast (FreeSurfer Functional Analysis Stream). The Dixon fat image was used to exclude non-fat signals for all anatomical scans. Subcutaneous fat was manually segmented and was excluded from BAT analysis. Both water saturation efficiency and water-fat contrast were calculated according to eq. 1 and eq. 2 on a pixel-by-pixels basis.

fMRI scans for probing BAT activity to cold challenge

BAT responding to cold challenge was assessed using 3D-FLASH sequence repetitively (TR 20ms, TE 1.85ms). Two baseline points were acquired before switching to cold challenge (13–16 °C), and followed by repetitive 3D-FLASH scans over the next 30 minutes. fMRI data was motion-corrected and registered to Dixon images, masked by Dixon fat images to exclude signal changes from non-fat area. Map of percent signal changes to cold challenge was masked by statistical significance of a student t-test (P<0.01).

RESULTS

The water-saturation efficiency approach used two independent MRI sequences: conventional T2 weighted (T2W) spin echo sequence and fast spin echo (FSE) sequence. A color map shows the fat area where the water-suppression analysis led to significant signal reduction (WS contrast > 10%) and thus the location of BAT (Fig. 1A). Similarly, the water-to-fat contrast method via Dixon’s method also detected BAT mass in the same areas (Fig. 1B). The BAT mass located by MRI was matched with previous cold-activated 18F-FDG PET/CT results in four of the subjects (one subject did not have PET/CT study (Fig. 1C). Both MRI and cold-activated¹⁸F-FDG PET/CT studies yielded similar volume of BAT in the cervical, supraclavicular, and upper thoracic depots (repeated measures ANOVA P = 0.11)(see volume estimation inTable 1). The degree to which the different methods demonstrate the cervical and supraclavicular BAT distribution can be appreciated through a 3D-reconstruction (Fig. 2 and Supplementary Videos 1–3). 3D PET/CT images were generated using PET/CT Viewer freeware (17). MRI methods also identified BAT in the deeper cervical depots, such as the carotid sheath, that have been seen using anatomical dissection(18) but were uncertain in the cold-activated PET/CT study.

Figure 1.

A. Detecting BAT mass (color overlay) by water-suppression efficiency using traditional spin echo imagines (SE) and fast spin echo (FSE) images. Color map was WS contrast, masked by a mask generated from the overlapping masks of FSE and SE methods. B. Detecting BAT mass (red-yellow overlay) via water-fat contrast using Dixon’s method. Blue tone: water image, green tone: fat image, red: water-to-fat contrast (indicator for BAT). Insert showed the non-subcutaneous fat mask generated by Dixon’s method. C.BAT is located via cold-activated 18F-FDG PET imaging. MRI and PET methods detected BAT mass at similar locations (eg. cross hairs). Note all images were from the same subject but with slight anatomical differences between PET and MRI due to arm-up (PET) versus arm-down (MRI) posture. Images were from subject #1 in Table 1.

Table 1.

Estimated BAT volume (ml)

Subject ID		MRI		PET 18F-FDG
		water saturation contrast	water-fat ratio
		FSE & SE	Dixon
#1	Left	18.1	19.5	20.2
	Right	21.4	24.8	21.4
#2	Left	22.6	18.2	22.4
	Right	21.0	14.7	20.1
#3	Left	13.8	n.a.	27.8
	Right	14.8	n.a.	28.9
#4	Left	23.3	n.a.	28.9
	Right	29.6	n.a.	28.6
#5	Left	21.5	24.3	n.a.
	Right	27.4	31.0	n.a.

^*Percent of BAT volume measured by FSE & SE that was identified via¹⁸F-FDG-PET/CT

Note, only area survived WS contrast criteria (>10%) in FSE and SE method will be counted toward BAT volume.

Figure 2.

3D reconstruction of BAT mass using MRI (with FSE-SE and Dixon) and PET-CT (subject #2 in Table 1).

Having demonstrated the potential of MRI to measure BAT volume in the thermoneutral state, we moved to functional MRI (fMRI) to measure BAT activity in response to cold stimulation. fMRI detects metabolic activity by measuring hemodynamic changes associated with oxygenation demand. Although fMRI is traditionally used for measuring neuronal activity, it is equally powerful in detecting metabolic activity in other organs, including BAT(5) as BAT requires glucose and oxygen to be delivered rapidly through the blood when BAT is stimulated (13,19,20). 3 shows an example of fMRI detection of BAT activity upon exposure to cold stimulation through a water-circulating vest (13–16 °C)(14) for 60 minutes while in the MRI scanner (n = 3 studied). The BOLD signal increased in BAT by 10.7 ± 1.8 % (P < 0.01) upon cold exposure. The degree of BOLD signal changes was much larger than in the typical brain fMRI study. The high BOLD signal changes in BAT was likely due to larger hemodynamic response to fulfill a higher metabolic demand in BAT and also decreased T1 relaxation time resulting from increased tissue temperature due to BAT thermogenesis.

DISCUSSION

Understanding human BAT physiology requires an accurate measurement of its volume in addition to its activity in response to stimulation, and we show here how MRI/fMRI provides several potential advantages over the current standard, 18F-FDG PET/CT. The variable, heterogeneous composition of BAT means that in the thermoneutral resting state PET/CT does not have the resolution to distinguish BAT from WAT. In contrast, the MR sequences used here rely on water content, which can detect BAT even without stimulation.

In our BAT volume estimation, Dixon’s fat image was as a mask for gross fat tissue. Since the estimation of BAT volume was further estimated via WS and fat-water contrast, the fat mask from Dixon’s method serves as a rough guideline and has minimal impact on the final BAT volume estimation. Even in the case the fat tissue was underestimated by the Dixon’s mask, the primary BAT region in our study appeared to be in consistent with the PET-CT measurement.

Cold-activated glucose uptake, as measured by accumulation of 18F-FDG, does distinguish BAT from WAT, and since it correlates with blood flow (20), is a useful measure of BAT activity. However, the ultimate fate of glucose in BAT and how well it predicts the amount of thermogenesis or BAT volume has not been established and may have substantial interindividual variation. One cannot automatically assume a proportionate correlation between¹⁸F-FDG PET/CT activity and BAT volume measured by MRI even as the two methods localized BAT to similar anatomical sites in our healthy volunteers. Future research should evaluate with greater precision how glucose uptake correlates with BAT water content and hemodynamic parameters evaluated by BOLD, which could be determined through the application of PET/MRI. With the ability to measure human BAT volume and activity using nonionizing fMRI, it will now be possible to study BAT physiology serially and in much more diverse populations.

Figure 3.

fMRI detection of BAT activity upon cold challenge (13–16 °C). Degree of BOLD signal changes (red-yellow map) is superimposed on anatomical images (in grey scale). Upon cold stimulation, significant BOLD signal increases were found in the regions identified as having BAT for subjects #1 – #3 (such as areas indicated by green circles). Note that subject #1 is the same subject as the one in figure 1.