Differences of Lipid Proton Compositions and Fatty Acids Between Alcoholic Fatty Liver and High-Fat Diet Fatty Liver Animal Models: 9.4 T Magnetic Resonance Spectroscopy Study

The objective of this study was to determine differences in chemical compositions according to fat deposition in the liver using an alcoholic fatty liver animal model and a high-fat diet-induced fatty liver animal model. A mouse model of chronic and binge ethanol feeding (NIAAA model), an alcoholic fatty liver model, was used to induce fatty liver according to the study protocol. The NIAAA control group had 15 mice. The NIAAA experimental group was administered with Lieber DeCarli diet powder. The high-fat diet control group was fed a general diet ad libitum. The high-fat diet fatty liver group was induced with a high-fat (60%) diet. Data were acquired by 9.4 T magnetic resonance spectroscopy for each fatty liver animal model. Fatty acids were calculated by quantifying each lipid proton through T2 correction. Difference in lipid proton (LP) for each model was identified with a multivariate statistical method. Mean differences in fatty acids among the four models were compared. The difference in LP composition was insignificant between the high-fat diet control and the experimental group. Orthogonal partial least squares discriminant analysis of the high-fat experimental group and the NIAAA experimental group showed no significant difference in the composition of each LP. However, there was a difference in the value of the composition deposited in the liver between NIAAA control and experimental groups. Comparison of each fatty acid between NIAAA control and experimental groups revealed that poly-unsaturated bond was significantly (p = 0.002) higher in the experimental group than in the control group. There were differences in total lipid and polyunsaturated bonds between NIAAA experimental and control groups. Methylene protons were deposited at lower concentrations whereas diallylic protons were deposited at higher concentrations in the NIAAA experimental group than in the control group.


Introduction
Fatty liver has various causes such as genetic, environmental, and metabolic factors. Nonalcoholic fatty liver disease is a common disease that affects more than 20-30% of the world's population [1,2]. However, continuous control of fatty liver is essential because its pathological progression can lead to the development of steatosis, fibrosis, and cancer of the liver [3,4]. When fat accumulates in the liver, it affects the oxidation of mitochondria in liver cells, generating free radicals and causing an inflammatory state. Generated liver steatosis can develop into liver fibrosis, liver cirrhosis, and hepatocellular carcinoma [5][6][7]. Once the pathological progression from fatty liver to liver cirrhosis occurs, liver steatosis becomes irreversible. Therefore, active treatment of fatty liver is essential to prevent pathological progression in the future. Many experiments have been conducted to prevent the pathological progression of fatty liver. [8,9]. Many studies have been conducted by inducing fatty liver in animals and performing interventional therapy to prevent pathological progression of fatty liver [10][11][12][13]. However, in some cases, the treatment effect was not well demonstrated when a wrong fatty liver model was selected [14]. In other words, when treating a fatty liver, the mechanism of the fatty liver should be accurately identified first and an experimental approach should be made [15].
Fatty liver is mainly divided into non-alcoholic fatty liver and alcoholic fatty liver. Non-alcoholic fatty liver is a disease in which fat is deposited in the liver due to administration of a high-fat diet. Thus, fatty liver is induced with a weight gain [15]. On the other hand, methionine and choline-deficient (MCD) fatty liver will stop breaking down fat metabolism in the liver parenchyma with weight loss due to methionine choline deficiency [16,17]. These two fatty livers only show different induction mechanisms without showing significant difference in compositions of fat deposition in the hepatic parenchyma. However, our study has confirmed that there are differences in chemical compositions of lipid protons through magnetic resonance spectroscopy in these two animal models [18]. In the case of the MCD model, macro-fat deposition was confirmed through histological evaluation. Above all, fats had higher chemical compositions in the MCD mode than the high-fat diet model in which unsaturated fatty acids were deposited. In addition, in the case of a high-fat diet-induced fatty liver, intracellular fat deposition was higher than that in MCD. It has been confirmed that fat deposit in the liver also shows a different pattern [18]. As such, in animal models of fatty liver that we know so far, early researchers studying fatty liver have recognized the same chemical composition of fat deposits in the liver. In this study, the NIAAA (mouse model of chronic and binge ethanol feeding) animal model, an alcoholic fatty liver animal model, was induced. Differences between control and experimental groups were investigated. In addition, we want to observe the difference between fats in these two animal models from the viewpoint of magnetic resonance spectroscopy by comparative analysis with a high-fat diet-induced fatty liver model, the most representative animal model of fatty liver. were prepared by diluting saline in Lieber Decarile diet powder (#710027, Dyets, Bethlehem, PA) to prepare a feed. Then 20 ml each was administered to mice to prepare the model. Mice were housed in an aseptically isolated space at room temperature (23 ± 3 °C) with a constant relative humidity of 55 ± 10%. The experimental group was administered a diet prepared by adding water and pure ethanol to Lieber Decarile diet powder (#710260, Dyets, Bethlehem, PA). During the first 1-5 days, the amount of ethanol was gradually increased from 0 to 1%, 2%, 3%, 4%, and 5% to allow adaptation to a liquid diet and alcohol. After that, during days 6 to 15, an alcoholic diet was prepared with Lieber-Decarli diet powder + Saline + 5% ethanol and administered orally. For the remaining 16 days, alcoholic fatty liver was induced by administering a high dose of ethanol with Lieber-Decarli diet powder + Saline + 31.5% ethanol. Two mice per cage (C57/BL6N, six weeks old) were bred in an environment and a total of 30 (experiment group: 15 mice; control group: 15 mice) mice were used for the alcoholic fatty liver study. In addition, high-fat diet fatty liver animal modeling was carried out using the same mouse species. The high-fat diet experimental group had 24 ICR mice under the same conditions as NIAAA modeling. For the induction of fatty liver, a high-fat diet (D12492, Research Diets, New Brunswick, NJ, 60% fat, 20% protein, and 20% Carbohydrates) was administered at 11 g/day/mouse. Water was provided ad libitum. Modeling was conducted over a total of six weeks. In addition, the high-fat diet control group had 15 ICR mice. Their breeding environment and feeding period were the same as those of the high-fat diet-induced fatty liver experimental group. These mice were allowed free access to general food and water. Animal modeling for the four groups was performed. Liver magnetic resonance spectroscopy (MRS) data were acquired within 24 h after completing all fatty liver modeling.

Acquire MRS Data and T2 Corrected Lipid Proton Quantification Data
Fat deposition rate in the liver was calculated as the ratio of the water (H 2 O) M0 value with T2 relaxation corrected and the sum of M0 values of each lipid proton for which the T2 relaxation time was corrected. Therefore, we performed a curve fitting for T2 relaxation of water by fabricating a virtual fatty liver phantom containing 10% of soybean oil. For phantom data acquisition, parameters were set as shown in Table 1 with the same device as the magnetic resonance imaging device for in vivo MRS data acquisition. Water MRS data to which multi-TE was applied were subjected to T2 relaxation graph fitting using MATLAB software (Mathworks, Natick, MA, US). After that, the T2 relaxation time and M0 value of water were calculated using the following formula: Signal Intensity (SI) of M TE = SI of M0 × e (−TE/T2) .
To obtain MRS of the liver of the mouse model of the four groups prepared for this experiment, 2% isoflurane was inhaled to minimize animal movement. Data were acquired under anesthesia. In addition, data were obtained while minimizing the effect of respiration by performing respiration gating. After that, the experimental animal was placed in the iso-center of the MRI equipment. Magnetic resonance images were acquired to set the volume of interest (VOI) for MRS acquisition. The entire liver was imaged in axial, coronal, sagittal directions using Rapid Acquisition with Refocused Echo (RPARE) pulse sequence. Parameters used for image acquisition are also summarized in Table 1. To obtain MRS of the liver parenchyma, the VOI was set as the part with a homogeneous signal intensity in the liver parenchyma while avoiding large vessels (Fig. 1). The STEAM pulse sequence was used for MRS acquisition. Parameters for data acquisition are summarized in Table 1. Obtained MRS data were quantified for each lipid proton using LCModel software. For data analysis using LCModel (version 6.3-1H, Stephen W. Provencher), the 'sptype' was designated as 'liver-11'. Data of 'liver-11' were analyzed by targeting the following seven LP quantified values and choline: methyl protons (-CH 3 , 0.90 ppm), methylene protons ((−CH 2 −)n; 1.30 ppm), ß-methylene resonance to the carboxy group (−CH 2 −CH 2 −CO; 1.60 ppm), allylic protons (−CH 2 −C=C−CH 2 −; 2.03 ppm), α-methylene resonance to the carboxyl group (-CH 2 −CH 2 -CO-; 2.25 ppm), diallylic proton s(=C−CH 2 −C= ; 2.78 ppm), and methane protons (−CH=CH−; 5.3 ppm). The most significant advantage of magnetic resonance spectroscopy is that it can quantify fat content in the liver with an in vivo test. To quantify fat content in the liver, the signal value at the moment origin (Mo) with T2 relaxation time of each lipid proton corrected is obtained. Therefore, we obtained the chemical composition ratio of each LP using T2 relaxation time data of seven lipid protons obtained from the experiment of Song et al. [19].

Differences of Lipid Proton Compositions and Fatty Acids Between…
Each LP composition ratio value = T2 corrected LP concentration value/T2 corrected 7 peak LP integrated concentration value.
The composition value of LP expresses difference in the composition of lipids deposited in the liver. In other studies, including our previous study, calculations of unsaturated/saturated fatty acid series through composition value ratio have been performed [18]. In addition, fat percentage deposition in obtained data was calculated with the following formula. These data were used for correlation analysis between fat percentage of each modeled mouse and T2 corrected LP quantification value.

Statistical Analysis
In each LP composition value, the sum of the total is 1. That is, the difference in the composition value of each LP means the chemical difference in liver fat. As a result, we investigated their differences using a multivariate statistical method. This analysis confirmed the difference between LPs with different chemical properties in each (1) Fat percentate (%) = T2 corrected 7 peak lipid protons integrated concentration value (T2 corrected 7 peak lipid protons integrated vale + T2 corrected water vlae) × 100. animal model. Spearman correlation analysis between fat percentage and each fatty acid was performed for LPs with a high contribution to fat deposition. The value of each fatty acid was calculated by the method suggested in previous studies [20,21]. In addition, the average comparison of fatty acids was identified with Mann-White's U test (SPSS version 20.0; SPSS Inc., Chicago, IL, USA).

In Vivo MRS Data of Each Animal Model and Pre-/Post-results of Lipid Percentage
In the animal model used in this study, each group had more than 15 animals. However, data showing an error of more than 10% of each lipid proton quantification were excluded due to motion artifact (respiration effect) occurring during MRS data acquisition process. Therefore, data in this study were obtained from 7 cases of the high-fat diet control group and 13 cases of the high-fat diet experimental group. In addition, 9 cases in the experimental group and 8 cases in the control group of NIAAA (alcoholic fatty liver) were analyzed. Body weight was measured just before MRS acquisition after modeling each animal in this study. T2 corrected M0 acquired by MRS and fat percentage are summarized in Table 2. The value of LP M0 was calculated by normalizing the T2 relaxation time of water (1.2866 ms) and finally quantified as a value obtained by correcting the T2 relaxation time of each LP as mentioned in the measured material method. Fat percentage induced by the high-fat diet showed a fatty liver deposition rate of over 35% in all animal models, 36.21% in the animal with the lowest deposition rate, and 66.84% in the model with the highest fat deposition (48.14 ± 12.10%). On the other hand, in the animal model fed the regular diet, the highest fat deposition rate was 45.05% and the lowest fat deposition rate was 9.93%, about 23.55% lower in fat deposition rate than the high-fat diet animal model. The NIAAA control group and the experimental group showed the same level of fat deposition (37.43% and 34.41%, respectively), showing moderate levels of fat deposition. In the case of the NIAAA control group, the model had the highest invasion rate of 52.88% and the lowest level of 16.86%. In the case of the NIAAA experimental group, the highest deposition rate was 57.10%, while the lowest deposition rate was 17.98%.

Each LP Composition Ratio Value Difference Identification, Fatty Acid Means, and Correlation Analysis
Each LP composition ratio value expresses the ratio of a specific LP to the sum of all LPs and directly affects the quantified value of Total lipid (TL), Total saturated fatty acid (TSFA), Total unsaturated fatty acid (TUSFA), Total unsaturated bond (TUSB), and Polyunsaturated bon (PUSB). The difference in T2 corrected LP was confirmed through Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) analysis as shown in Fig. 2. The difference was not clear between the high-fat diet control and the experimental group. LPs with a VIP score of more than 1 for these two groups were diallylic protons, α-methylene resonance to the carboxyl group, methane proton, and β-methylene resonance to the carboxy group expressed at 2.8, 2.3, 5.3, and 1.6 ppm, respectively. No significant group difference between the high-fat experimental group and the NIAAA experimental group was observed in the OPLS-DA analysis. LPs showing more than 1 of the VIP score value were β-methylene resonance to the carboxyl group and α-methylene resonance to the carboxyl group expressed at 1.6 and 2.3 ppm, respectively. On the other hand, a significant difference between the NIAAA control group and the Experimental group was observed in the OPLS-DA analysis. For LPs with a VIP score of 1 or higher, diallylic protons, methylene protons, and α-methylene resonance had a carboxyl group. In the case of the NIAAA control group and the experimental group, it was observed that there was a difference in the value of the composition deposited in the liver (Fig. 2c). LPs with a VIP score of 1 or higher were 1.5, 1.22, and 1.05 for diallylic protons (2.8 ppm), methylene protons (1.3 ppm), and α-methylene  Table 3 Correlation coefficient values of between fat percentage and TL were 1.000 (p = 0.000), and TUSB and PUSB were − 0.867 (p = 0.002) and 0.667 (p = 0.050), respectively in the NIAAA control group significantly. In the NIAAA experimental group, except for TL, fatty acids failed to show a significant correlation with fat percentage. In the high-fat diet control group, fatty acids that showed significant correlations with fat percentage were TL and PUSB, with correlation coefficients of 1.000 (p = 0.000) and 0.821 (p = 0.023), respectively. In the high-fat experimental group, only TL showed with fat percentage a significant correlation coefficient of 0.989 (p = 0.000). Results of comparing means through  the Mann-Whitney test are shown in Fig. 3. As a result of the average comparison of each fatty acid between NIAAA control and experimental groups, PUSB was significantly (p = 0.002) higher in the experimental group. TL (p = 0.004), TUSB (p = 0.006), and PUSB (p = 0.004) showed significant differences between highfat diet control and experimental groups. In the comparison between the NIAAA experimental group and the HF experimental group, the HF experimental group showed significantly (p = 0.004) higher value of TL. TUSFA (p = 0.035) and TUSB (p = 0.035) showed higher values in the NIAAA experimental group.

Discussion
Fatty liver is a disease caused by the accumulation of fatty acids. It is an early stage of liver disease that can lead to cirrhosis and liver cancer through liver fibrosis in the future. Many studies have been conducted to improve the therapeutic effect on liver disease by managing fatty liver and blocking the progression of the disease through appropriate intervention [10,12,22,23]. Liver disease is primarily classified into alcoholic and non-alcoholic fatty liver. Research results recently published have shown differences in characteristics of fatty acids for each mechanism of occurrence [18]. Results of this study provide basic information to determine the difference in fatty acid and LP composition between alcoholic fatty liver and high-fat diet fatty liver models to select an appropriate animal model for fatty liver animal experiments in the future.
In this study, the degree of fat generation was evaluated using a total of four animal models. Characteristics of LP were first identified by acquiring MRS from the liver parenchyma of each model. In the case of the HF group, the average fat deposition rate was 48.15% and the deposition rate of intrahepatic fat was 1.913 times higher than that of the general diet group. Through OPLS-DA analysis, there was no difference in each LP between the two groups. That is, there was no difference in chemical composition of fatty acids between the group that took the high-fat and the group took the standard diet. It was observed that the correlation between fat percentage and TL was significant in the high-fat experimental group due to fat deposition. This meant that as fat was deposited, the concentration of methylene protons increased, although concentrations of other LPs did not increase significantly.
A similar correlation was also observed in the high-fat diet control group. It was confirmed that TL was a high correlation with PUSB (correlation coefficient: 0.821). PUSB is a diallylic proton expressed at 2.8 ppm. There were no differences in results the OPLS-DA analysis between the two groups, consistent with results showing that components having the highest VIP score were diallylic protons. When comparing the mean between fatty acids through the Mann-Whitney test, TL, TUSFA, and TUSB showed statistically significant differences. In the case of TL, since it was calculated through the ratio of water and methylene protons, methylene protons in both groups increased simultaneously, which could be the main LP component of fat accumulation. In the HF diet fatty liver experimental group, concentrations of all fatty acids including TL were increased. As shown in Table 3, TUSB and PUSB were significantly increased. Since all fatty acids were normalized to methyl protons at 0.9 ppm, it meant that higher concentrations of diallylic protons (2.8 ppm) and methane protons (5.3 ppm) were formed. This was consistent with the result showing that the concentration of LP with the highest VIP score in the OPLS-DA analysis between the two groups was 2.8 ppm. However, in the correlation analysis with TL, no significant increase in the concentration was observed, suggesting that the increase was not significant.
Meanwhile, we conducted an alcoholic fatty liver model using the NIAAA method. This was because the NIAAA method was introduced to perform modeling in a shorter period than the previous chronic model. It was similar to the pathogenesis of human alcoholic hepatitis [24]. We induced alcoholic fatty liver according to the protocol presented in the experimental study of Bertola et al. [24]. We performed modeling of experimental and control groups using the Lieber-DeCarli liquid diet high in monounsaturated fat and low in carbohydrates. First, it was confirmed that there was a difference in the composition of LP between the NIAAA control group and the NIAAA experimental group through OPLS-DA analysis. In VIP scoring as a significant indicator, diallylic protons (2.8 ppm), methylene protons (1.3 ppm), and α-methylene resonance to the carboxyl group (2.3 ppm) were 1.3, 1.22, and 1.05, respectively. In the case of the NIAAA control group, fatty acids that had statistically significant correlations with fat percentage were TUSB and PUSB, including TL. PUSB was directly related to diallylic protons expressed at 2.8 ppm, with a correlation coefficient of 0.667 (p = 0.050), which meant that PUSB also increased together with the progress of intrahepatic fat deposition compared to the NIAAA Experimental group. This result supports the OPLS-DA result for the composition ratio of each LP. Diallylic protons are closely related to PUSB. This result could also be observed through the S-plot of the OPLS-DA analysis shown in Fig. 3, which meant that there was a difference in the deposition of diallylic protons (2.8 ppm) and methylene protons (1.3 ppm) in both models. That is, the significant result found in this experiment meant that fat with a difference in fatty acids was accumulated in the liver in the NIAAA experimental group and the control group. TL is a fatty acid mainly deposited according to fat deposition, even in previous studies [18]. The same results were also shown in this study's two fatty liver models because the increase in methylene protons was directly related to TL. However, in the case of the control model of NIAAA, the TL deposition occurred with higher specific gravity than the experimental group. In contrast, the deposition of PUSB was lower, confirming a difference in the chemical composition of fat between the two models.
There are several experimental limitations in this study. First, the concentration of each T2-corrected LP was the T2 relaxation time used by other researchers. However, referenced results were obtained in the same environment as the same device we tested. Thus, it was judged that error in T2 relaxation time did not occur significantly. Second, when calculating fat percentage, the water signal measures the T2 relaxation time. Correct value can be calculated with T2 corrected water signal reflecting this. Since the virtual fatty liver phantom conducted in this experiment was very different from an in vivo environment, the fat percentage presented in results of this study could not represent fat percentage accurately. However, even if the accuracy of fat percentage is not evaluated, it does not affect results of evaluating differences in fatty acid changes or fat's chemical compositions according to fat deposition.

Conclusion
In conclusion, this study demonstrated no difference in fat composition between fatty liver animals that received a high-fat diet and the control group that received a regular diet. On the other hand, there were differences in TL and PUSB between NIAAA experimental and control groups. There were also differences in total lipid and polyunsaturated bonds between NIAAA experimental and control groups. Methylene protons were deposited at lower concentrations whereas diallylic protons were deposited at higher concentrations in the NIAAA experimental group than in the control group.