The beneficial effects of a high dietary intake of polyunsaturated fatty acids (PUFA) are well known. In general, evaluation of the dietary intake of PUFA is difficult (1), but the amount of PUFA supplied by the diet is directly related to the levels of these fatty acids in adipose tissues (2). Therefore, their concentration in adipose tissue can be used as a measure for the intake of PUFA (1). The ratio of polyunsaturated to saturated fatty acids in subcutaneous fat seems to reflect the degree of unsaturation of dietary fat (3); however, information about PUFA concentration in deeply located depots is difficult to obtain noninvasively.
A detailed knowledge of the spatial distribution of PUFA is also desirable since it has been demonstrated that PUFA have different metabolic effects upon the various adipose tissues. For example, they limit the hypertrophy of abdominal white adipose tissue (WAT) depots (4) and stimulate nonshivering thermogenesis in brown adipose tissue (BAT) (5). It was demonstrated that during feeding enriched with PUFA, different trends in fatty acids content and distribution were found in BAT in comparison to WAT (6).
The biochemical assay of PUFA is relatively easy to perform, but methods for imaging the fatty acids in full anatomical detail are still lacking. Nuclear magnetic resonance (NMR) is a proven tool for studying lipids in organic tissues (7–10) and may solve the problem of PUFA-mapping. Proton magnetic resonance spectroscopy has been applied to evaluate the polyunsaturation degree (PUD) in tissue extracts (7). In vivo localized spectroscopy has yielded NMR spectra of fat in selected tissues (8). A noninvasive determination of linoleic acid content in subcutaneous WAT has been performed by 13C NMR spectroscopy (9, 10). Encouragingly, the NMR data have been found to be consistent with those obtained by gas chromatography in biopsies (9). A method which allows assessing and mapping the PUD of different tissues may be used to monitor the effect of diet, metabolism, drugs, or treatments upon the PUD of target tissues. Recently, it has also been reported that PUFA accumulation occurs in apoptotic cancer cells (11), raising the possibility of in vivo detection of tumor growth and therapies by chemical shift imaging (CSI).
In the present work, mapping of PUFA in phantoms and in vivo was achieved with a protocol based on NMR CSI. The proposed method is a modification of a technique, based on NMR spectroscopy, previously applied for PUD quantitation in tissue extracts (7). Here, the spectroscopic algorithm was repeated on a pixel-by-pixel basis to produce quantitative maps of the PUD parameter. Several phantoms with a different PUD and the thoracic region of the rat were examined. In the thoracic area, different types of adipose tissue, in particular the axillary WAT and interscapular BAT (IBAT), are present. The fat content of these tissues has been analyzed in previous studies (12–16). The aim of this work is to demonstrate the feasibility of whole-body imaging of PUFA with high spatial resolution.
MATERIALS AND METHODS
Vegetable oils with different PUFA content were chosen: olive oil (from Turri, Verona), peanut oil (from Unilever S.P.A., Milano), and corn oil (from Carapelli, Firenze). Fish oil (cod-liver) was obtained from Farma P S.R.L. (Verona). Animal fat was produced by pooling different deposits of rat WAT and by extracting the triglyceride fraction according to Folch et al.'s method (17).
High-resolution (HR) NMR-spectroscopy experiments were performed with a Bruker DRX-500 (Bruker Spectrospin, Karlsruhe, Germany), equipped with an 11.7T magnet (Oxford Ltd., Oxford, UK), operating at 499.94 MHz for 1H. Proton NMR spectra were acquired at 25°C under nonsaturating conditions with the following parameters: time of repetition (TR) = 2.98 sec, 128 averages, spectral width = 5 kHz, flip angle = 90°, number of data points = 32k. Integration of fatty acid peaks was performed after baseline correction with a fifth degree polynomial. Following the lettering assignment of Zancanaro et al. (7), the peak A (0.9 ppm), related to the terminal -CH3 of the fatty acid chain, and the peak G (2.8 ppm), related to the diallylic -CH2 group of the chain, were considered. The PUD coefficient was defined by the ratio:
where G and A indicate the areas of respective peaks after baseline correction.
Chemical Shift Imaging
Imaging experiments were carried out using a Biospec System (Bruker Spectrospin), equipped with a 4.7T horizontal Oxford magnet, 33 cm bore, and SMIS (Surrey Medical Imaging Systems, UK) gradient insert. A 72-mm internal diameter birdcage coil was used.
A spin-echo, CSI sequence, consisting of two slice-selective RF pulses (90° and 180°) and two phase encoding cycles, was used. Experiments on phantoms were performed with the following parameters: field of view (FOV) = 6 × 6 cm, matrix size = 32 × 32, slice thickness = 6 mm, TR = 1 sec, time of echo (TE) = 5.25 ms; 512 complex points (zero-filled to 1024) were acquired with bandwidth (BW) = 6009.62Hz, and number of experiments (NEX) = 4. The acquisition time was 68 min.
Spectra were processed using Matlab (MathWorks, Natick, MA). In each pixel, frequency shifts caused by B0 inhomogeneities were corrected by setting the resonance of the main fat peak at 1.3 ppm. In some of the pixels, the shape of the spectral lines did not allow this correction; these pixels were not considered for the study.
In the remaining pixels (about 90%) the areas of A and G peaks were estimated after baseline subtraction: with an automated procedure, baselines were approximated by finding the straight line that maximizes the area under the spectral peak. In general, peaks which are superimposed upon a smoothly varying background may be described by a peak function plus a background function. Under the hypothesis that the background curve is a slowly varying function under the peak, it may reasonably be interpolated by a second-degree polynomial (18). In our approach, for each peak, we utilized instead a first-degree polynomial, since it allows one to apply a simple algorithm to optimize the background removal: the straight line is tangent to the curve when it leaves the maximal area above it. In this way, no fit procedure, potentially unstable, is applied; an area evaluation is far easier than a peak fitting. Moreover, the spectral width of the peak is left as a degree of freedom throughout the pixels, which seems reasonable since the spectral width can be different due to nonuniform shimming.
The method was validated on oil phantoms. The areas A and G in the spectrum of each pixel were considered as variables and plotted together. Linear regression was performed and the slope coefficient was multiplied by 1.5 to give the PUD according to Eq. .
The CSI measurements of PUD in the phantoms were compared with the HR results: statistical correlation (Pearson's test) and linear regression were finally calculated.
PUFA Mapping in Animals
Sprague-Dawley rats, 80 days old, about 200 g, fed ad libitum, were used. Immediately before MRI evaluation the rats were sacrificed by ether inhalation. The IBAT deposit was initially located with a sagittal scout image. Then an axial section was obtained at the level of the maximum thickness of the IBAT deposit. This section was chosen because it also includes the large axillary deposit of WAT, allowing for a direct comparison between BAT and WAT.
A single-slice, spin-echo sequence was acquired as morphological reference with the following parameters: FOV = 6 × 6 cm, matrix size = 128 × 128, slice thickness = 2 mm, TR = 1 sec, TE = 15.8 ms, acquisition time = 2 min. After shimming over the whole probe volume, water signal was suppressed by using a single gaussian 10 ms pulse followed by spoiled gradients. Then localized shimming was performed and a typical linewidth of 0.2 ppm was achieved on the fat peaks. A 2D CSI sequence was acquired and processed according to the above procedure. About 85% of the pixels in the adipose tissues were included in data analysis, since the remaining were not corrigible for chemical shift effect. The values of the PUD parameter were calculated for each animal using the procedure described above and a statistical comparison between IBAT and axillary WAT was performed. Pixel-by-pixel values of PUD were also calculated and used to generate a parametric map of PUD in the deposits of adipose tissue. The map was overlapped to the morphological reference image after gaussian smoothing in order to compare the relative PUD of the IBAT and the axillary WAT deposits.
In Vivo Test
The feasibility of in vivo PUFA-mapping was tested on rats anesthetized by inhalation of a mixture of O2 and air containing 1–2% halothane. The same CSI protocol as before was used with parameters: FOV = 8 × 8 cm, matrix size = 32 × 32, slice thickness = 6 mm, TR = 1 sec, TE = 5.25 ms; 512 complex points (zero-filled to 1024) were acquired with BW = 6009.62 Hz and NEX = 4. The inguinal region was chosen for its better shimming conditions, in vivo, than the thoracic one.
The tests performed on oil and fat phantoms demonstrated good agreement between the CSI and the HR measurements of PUD. In Fig. 1a,b, HR and CSI spectra of corn oil are shown: the fatty acid peaks, clearly resolved in HR spectrum, are also distinguishable in CSI spectrum. The PUD of the considered oils and the fat phantom, measured with both techniques, ranges from 0.1–1.3 (PUD is greater than 1 in long-chain fatty acids due to the high number of double bounds), reflecting different PUFA contents. A scatterplot of the area of peak G vs. peak A, relative to all the pixels examined in three of the phantoms, is shown in Fig. 2. Linear regression of data, in each phantom, has a positive intercept in the G-axis (or equivalently negative intercept in the A-axis): this feature has also been found by analyzing the animal plots (see below) and is probably due to the algorithm of baseline correction. However, the slope coefficient is less sensitive to this effect and has been used for evaluating the PUD. The correlation between HR and CSI measurements is very high (R2 = 0.998; P < 0.005) and their ratio is close to 1 (m = 0.989 ± 0.028 SEM).
Data obtained from rats are illustrated in Figs. 3–5. In Fig. 3a, a representative parametric map of PUD overlapped to the morphological image is reported. In the morphological image, a thoracic region with different types of adipose tissue is shown: the large deposit of IBAT is visible dorsally while WAT axillary deposits are also recognizable. Typical spectra from pixels of the two deposits are reported in Fig. 3c. Qualitative analysis of the maps suggests that PUD distribution is more uniform in BAT deposits than in WAT. This finding is confirmed by the scatterplot of G vs. A (Fig. 4), relative to all the pixels examined in all the animals; IBAT tissue presents a higher correlation (R2 = 0.765, n = 118) between A and G than axillary WAT (R2 = 0.418, n = 86), both correlation coefficients being statistically significant (P < 0.005). In each rat, IBAT has been found more polyunsaturated than WAT (Fig. 5), even if in some animals the standard error over PUD is very high. When considering the sample including all the rats, we obtained a PUD of 0.68 ± 0.03 (SEM) of IBAT and 0.48 ± 0.06 of WAT; moreover, a paired t-test comparing the IBAT and WAT values of the single rats provided a significant discrimination between the tissues (P < 0,005).
Experiments performed under anesthesia demonstrated the feasibility of an in vivo PUFA mapping using the CSI protocol. In the parametric map of the inguinal region (Fig. 3b) a PUFA-rich (PUD = 1.00 ± 0.08 SEM) WAT deposit is visible.
Recently, MR techniques provided relevant information about the different types of adipose tissue. These techniques can quantify small fat deposits, study their structural characteristics, or describe reactivity to drug treatments (for a recent review, see Zancanaro et al. (19)). Moreover, MRI techniques allow the calculation of some “functional” parameters of the adipose tissue, like the hydrolipidic ratio (16). However, NMR-based techniques have found few applications in studying the presence of PUFA in tissues. While some efforts have been made with spectroscopic methods (7–10), the CSI techniques have scarcely been exploited. Nevertheless, CSI allows one to selectively obtain, pixel-by-pixel, the spectra of several chemical compounds, such as lactate, phosphocreatine, ATP, and thus appears very promising for the study of PUFA. This work explores the feasibility of whole-body imaging of PUFA and the results demonstrate that, in studies on adipose tissue, this technique has several advantages with respect to other biochemical or spectroscopic methods: it has a higher spatial resolution, it can monitor many different tissues at the same time, and it can analyze deep tissues (i.e., visceral fat) that are not visualized with other noninvasive methods. The main limits of this technique are the long acquisition time, instrument cost, and the inability to distinguish individual fatty acids.
When applied to lipid phantoms, our CSI measurements of the PUD correlated well with the data obtained by HR-spectroscopy, indicating that it is possible to quantitatively map the PUD of fatty acids. The phantom tests also confirmed the validity of the baseline correction applied in our procedure. The nonzero intercept in the linear regression could indicate a systematic underestimation of area A or an overestimation of area G: this especially since the intercepts in all data plots have the same sign. They are smaller in phantoms (from 0.01–0.68, arbitrary units) than in animals (from 0.27–1.40), indicating that shimming conditions are possibly relevant to the goodness of baseline correction; however, the order of magnitude is the same, revealing that the method, validated on phantoms by HR, is still reliable in animals. The fact that shimming conditions could influence the baseline correction is probably a minor reason for spread of A and G values in phantoms (Fig. 2), while a major reason, for some pixels, is the volume-averaging effect due to the presence of air.
The high spatial resolution of the method allows assessment of PUD in different areas of adipose tissues. When applied to the rat thoracic region, our protocol gives quantitative PUD maps of the two main kinds of adipose tissue, BAT and WAT. In the IBAT, the mean PUD value was 0.68 and the PUFA composition was relatively more homogeneous than in WAT, as attested by the high correlation between areas of peaks A and G, and good reproducibility over the group of animals. In the axillary WAT, the mean PUD was 0.48 and the data indicated that the PUD values in this tissue were rather spread. This spreading could be due to a lower homogeneity in the lipid constituents of the axillary WAT relative to BAT. In effect, the axillary region is anatomically very complex and the WAT in this area is not present in its pure form, but is mixed with vascular, nervous, and lymphatic formations that can alter the homogeneity of the lipid content. In contrast, the homogeneity we have found in the IBAT is in agreement with the histological data that show a homogeneous distribution of adipocytes in this tissue (13). The difference in PUD values we have found between IBAT and axillary WAT is statistically significant and confirms a different metabolism of PUFA in these two tissues, as reported in previous studies (5, 6).
The absolute value of PUD may depend on several factors, such as age, diet, acclimatization, etc. Several authors have calculated the ratio of PUFA vs. the total fatty acid content, both in WAT (Field et al. (20) in humans; Valero-Garrido et al. (2) in rats) and in BAT (Raclot et al. (6) in rats). This parameter cannot be evaluated by NMR and differs from PUD since, in the calculation of the latter, PUFA with two or more double bounds contribute with a stronger weight: so PUD is usually higher than the PUFA / fatty acids ratio.
The higher value of PUD we found in inguinal WAT, in vivo, may reflect the different morphology of this tissue with respect to the axillary WAT (21).
In conclusion, the application of CSI-based mapping seems to open new possibilities for studying the tissue metabolism of fatty acids and could be useful in monitoring dietetic PUFA intake or pharmacological/experimental treatments involving adipose tissue. However, further evaluations are necessary to clarify the results obtained by this method; in particular, some anatomically different regions of WAT and BAT need to be mapped and different species should be studied. The possibility to perform a PUFA mapping in nonadipose tissues must also be evaluated with different acquisition protocols. The method appears potentially applicable also in clinical inquiry on humans, even if more study is needed for this purpose, particularly to evaluate the spectral discrimination at lower magnetic field.
The authors thank Prof. Marco Villa and Miss Giovanna Bissoni for proofreading.