The primary materials used in this study were soybeans and tempe, sourced from a small-scale enterprise (SME) in Bogor, Indonesia. The tempe produced by this SME has obtained Indonesia National Standard (SNI), HACCP, BPOM (Indonesia Food and Drug Surveillance Agency), and Halal certifications, indicating compliance with national and international food safety standards. The tempe production process at this SME adheres to good manufacturing practices (GMP) and meets export standards. The process involves cleaning soybean seeds, soaking and boiling, dehulling, inoculation with Rhizopus spp., plastic packaging, and fermentation at room temperature (28–30 °C) for 48 h.

Tempe flour and soybean flour were processed separately according to the method described by Mahdi et al. (21). Fresh tempe was cut into 2 × 2 cm² cubes and then comminuted using a chopper. Soybean flour was produced through the following steps: cleaning, soaking, boiling, dehulling, and comminuting using a chopper. Both the comminuted tempe and soybeans were dried using a fluidized bed dryer (FBD) at 50 °C for 6 h. Drying with a FBD at temperatures below 60 °C prevents the oxidation and degradation of antioxidant components like polyphenols (Elgamal et al.). Additionally, FBD ensures the sample’s surface dries evenly, a notable advantage over traditional tray or vacuum drying methods. Subsequently, the dried materials were ground into a fine powder using a grinder and sieved through a 100-mesh sieve.

The antioxidant activity of the samples was determined using the 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging assay. This method assesses the ability of antioxidant compounds in the samples to stabilize the DPPH radical. The procedure was adapted from Abdurrasyid et al. (22). Initially, 1 g of the powdered beverage sample was mixed with 10 mL of a methanol–water solution (80:20, v/v). The mixture was homogenized using a vortex mixer for 1 min and then centrifuged at 4 °C and 3,000 rpm for 45 min. Subsequently, 0.2 mL of the supernatant was added to 3.8 mL of 0.1 mM DPPH solution. The mixture was incubated in the dark for 30 min. The absorbance of the incubated samples was measured at 517 nm (A1) using a spectrophotometer. In parallel, a standard curve was constructed using serial dilutions of ascorbic acid (25–200 μg/mL) as a reference standard. This curve was used to convert the absorbance values into antioxidant capacity expressed as mg ascorbic acid equivalents (AAE)/100 g. Additionally, a blank sample (A0) was measured without the sample in the test solution. The DPPH radical scavenging activity was calculated using the following formula:

The isoflavone content of tempe flour and soybean flour was determined using a high-performance liquid chromatography (HPLC) method as described by Astawan et al. (23). Two grams of each sample were mixed with 30 mL of a 1:4 hydrochloric acid: acetonitrile solution to hydrolyze isoflavone glycosides into isoflavone aglycones. The mixture then extracted using a sonicator for 30 min. Thenincubated in a water bath for 1 h. After the hydrolysis process was complete, the sample was cooled and centrifuged. The supernatant was collected and filtered using a 0.45 μm syringe filter. A 20 μL aliquot of the filtrate was injected into an HPLC system equipped with a C-18 column (15 cm × 4.6 mm i.d., 5 μm particle size). The separation was performed using a reverse-phase mode, elution isocratic with a mobile phase consisting of a methanol:1 mM ammonium acetate mixture (6:4, v/v) at a flow rate of 1 mL/min. The UV detector was set at a wavelength of 265 nm, column temperature: 30 °C. Subsequently, the isoflavones were quantified using external standards of daidzein and genistein and a calibration curve. The concentrations of daidzein and genistein in the samples were then calculated based on the sample peak areas and the calibration equations.

Cholesterol binding activity was determined using the Lieberman–Burchard (LB) method, as described by Muharni et al. (24) and Imtihani et al. (25). The LB method is a colorimetric assay that reacts cholesterol with the LB reagent. Samples were extracted by maceration with 70% ethanol for 3 × 24 h at room temperature, followed by sonication for 30 min. Preliminary analysis indicated that the optimal wavelength for measurement was 415 nm with an operating time of 45 min. A standard curve was constructed using serial concentrations of cholesterol (100, 150, 200, 250, and 300 ppm) to quantify the cholesterol content. Five milliliters of each sample extract and simvastatin (positive control) were mixed with 300 ppm cholesterol solution (A) in test tubes, vortexed for 30 s, and incubated at 37 °C for 60 min. After incubation, the tubes were centrifuged at 4,000 rpm for 5 min. The supernatant was transferred to new tubes, 1 mL of acetic anhydride was added, and the mixture was vortexed for 30 s and kept in the dark for 30 min. Subsequently, 0.1 mL of concentrated sulfuric acid was added, vortexed, and incubated in the dark for 45 min. Finally, the absorbance of the solution was measured at 415 nm. The cholesterol concentration in the test samples and positive control (B) was determined using the standard curve equation. A blank sample (C) was also measured, which consisted of the sample solution without sulfuric acid, to serve as a color baseline. The calculation was performed using the following equation:

Pancreatic lipase inhibitory activity was determined according to the methods described by Pradono et al. (26) and Aji et al. (27). The assay was performed by mixing 0.1 mL of lipase (1 mg/mL) with 0.2 mL of 70% ethanol extract of the sample (the same for each concentration) and 0.7 mL of Tris–HCl buffer (pH 8.0) in a test tube. The mixture was homogenized and incubated at 37 °C for 15 min. After incubation, 0.1 mL of p-nitrophenyl palmitate (p-NPP) was added, and the mixture was incubated for an additional 30 min at the same temperature. Finally, the absorbance was measured at 410 nm using a spectrophotometer. Orlistat was used as a positive control (100% inhibition) and was prepared following the same procedure. The percentage of inhibition was calculated as follows:

Note: A0, Absorbance of blank (without inhibitor); A, Absorbance of sample.

The antioxidant activity data was analyzed using SPSS 22. An independent t-test was performed at the 5% significance level. For multivariate data from GC–MS analysis, visualization was performed using SIMCA 18 (Umetrics, Umea, Sweden). Principal component analysis (PCA) was initially applied to the multivariate data to visualize sample discrimination and assess the initial model quality. If the PCA was satisfactory, orthogonal projection to latent structures (OPLS) analysis was performed. OPLS analysis generated a correlation model between metabolites (X) and bioactivity (Y). The output data from OPLS analysis included score plots, S-plots, Y-related coefficient plots, and VIP plots. The generated model was validated using R²X, R²Y, Q², CV-ANOVA, and permutation tests.

Soybeans and tempe contain a variety of antioxidant compounds. Consuming foods rich in antioxidants offers numerous health benefits. Many chronic diseases and metabolic syndromes are caused by oxidative stress in the body (29). The isoflavones in soybeans and tempe can prevent this from happening. Soybean isoflavones can act as anti-atherosclerotic agents by inhibiting oxidative stress, increasing nitric oxide (NO) availability, reducing LDL size, and preventing LDL peroxidation (30). Table 1 shows that the antioxidant activity of tempe flour (31.56 ± 2.27%) was significantly higher (p < 0.05) than that of soybean flour (27.99 ± 1.57%). This result is consistent with the isoflavone content of the samples. The content of daidzein and genistein in tempe flour was also significantly higher (p < 0.05) compared to soybean flour.

Based on Barus et al.’s study (31), the antioxidant activity of tempe can reach 53%–84%. The values might decrease in powdered samples due to the drying process, which can reduce bioactive antioxidant components. However, this finding is consistent with the study by Kuligowski et al. (32) who observed that the isoflavones daidzein and genistein increased by 5–8 times during tempe fermentation. This demonstrates that the content of both isoflavones is indeed higher in tempe than in soybeans. The fermentation process in tempe increases the aglycone forms of daidzein and genistein due to the activity of microorganisms that hydrolyze the glycosidic bonds, namely β-glucosidase (23). The increase in aglycone isoflavones enhances their bioavailability and antioxidant activity, as both compounds are flavonoids that can donate electrons to free radicals. In addition to isoflavones, which are derivatives of flavonoids, soybeans also contain other antioxidant compounds such as phenolic compounds.

The results of the Lieberman–Burchard analysis of tempe flour and soybean flour (Figure 1) showed significantly different binding percentages. Tempe flour had the highest binding value (27.66%), followed by soybean flour (14.98%), and the lowest was the positive control, statin (6.09%). These results indicate that tempe flour and soybean flour have a greater potential to control blood cholesterol levels compared to statins, which work through a different mechanism. The positive control, statin, showed the lowest percentage decrease in cholesterol, presumably due to the lack of direct binding with free cholesterol. This limited cholesterol-lowering effect of statins can be attributed to their mechanism of action, which primarily involves inhibiting HMG-CoA reductase, the rate-limiting enzyme in cholesterol biosynthesis, rather than directly binding to cholesterol molecules (33). The findings are largely comparable to the study by Zendrato et al. (34), where Porang glucomannan extract exhibited a cholesterol-binding capacity of 13.42–45.56% (at concentrations of 150–750 μg/mL). In contrast, research by Musa et al. (35) on Saurauia vulcani Korth. (Actinidiaceae), a tropical medicinal plant, reported a higher cholesterol-binding capacity ranging from 20 to 80%.

The cholesterol-binding activity of tempe flour and soybean flour was analyzed using the photometric Lieberman–Burchard (LB) method. The LB method can be used for both qualitative and quantitative analysis. The principle of this analysis involves measuring the amount of free cholesterol in a solution that is not bound to metabolites in the sample. Free cholesterol undergoes derivatization with acetic anhydride and H₂SO₄, resulting in a strongly chromogenic blue–green cholesterol derivative, known as an aromatic sulfonic acid. The unbound cholesterol was quantified spectrophotometrically using a method adapted from Adu et al. (36). A higher absorbance value correlates with a higher concentration of free cholesterol, as evidenced by a more intense color of the solution. Conversely, if a sample effectively binds cholesterol, the resulting solution will exhibit a lighter color and a lower absorbance value. The samples used in the cholesterol binding activity test were ethanol extracts. Ethanol was chosen as the extraction solvent because it is considered a safe solvent and can effectively extract bioactive compounds, especially antioxidant derivatives such as phenolics, flavonoids, and saponins from the material. Ethanol can degrade non-polar walls through semi-polar interactions, allowing intracellular phenolic-flavonoid compounds to be extracted (37–39).

A positive correlation was also found between the cholesterol-binding activity and the antioxidant activity of the samples; the higher the antioxidant activity, the greater binding cholesterol activity. The cholesterol binding mechanism in the LB reaction involves the interaction of the cholesterol hydroxyl (-OH) group with antioxidant compounds, such as phenols and flavonoids. This interaction leads to the formation of stable hydrogen bonds, effectively preventing cholesterol from reacting with sulfuric acid. The crucial role of the hydroxyl group in this process is underscored by its involvement in the formation of a pharmacophore, as described by Musa et al. (35). The OH group in phenols is also thought to have the potential to bind with cholesterol, forming hydroxysterols. Hydroxysterols are another form of cholesterol that has accepted an oxygen atom. The remaining free cholesterol reacts with acetic anhydride and H₂SO₄, forming a colored compound.

This model of cholesterol binding in the LB analysis represents one of the body’s regulatory mechanisms for maintaining stable cholesterol levels (homeostasis). Cholesterol homeostasis can be controlled by four mechanisms: (1) cholesterol absorption from food through the small intestine, (2) cholesterol synthesis by various body tissues, especially the liver, and intestines, for release into the blood plasma, (3) cholesterol excretion through the excretory system regulated by the liver, and (4) conversion of cholesterol to various hormone compounds required by humans (40). Cholesterol exists in two forms: free cholesterol and cholesterol ester. Only monomeric cholesterol can be absorbed by the small intestine/basolateral membrane (41). Thus, bound cholesterol is presumably unable to be absorbed or is inhibited from absorption and can be excreted through feces, similar to the mechanism of dietary fiber binding bile acids or cholesterol. The percentage of cholesterol binding in this LB analysis is expected to reflect the difference in the cholesterol binding capacity of the samples. Tempe flour and soybean flour are known to have the potential to control blood cholesterol levels.

The results showed that 58 metabolites were detected (Table 2). Of these, 52 compounds were successfully annotated using an in-house library, and only six compounds remained unidentified. Of the 52 annotated compounds, there were six bioactive compounds in soybean flour and tempe flour, namely 3-hydroxy-3-methylglutaric acid, 4-aminobutyric acid, meso-erythritol, 3-phenyllactic acid, daidzein, and genistein. In addition, there were 20 amino acids and their derivatives, 8 organic acids and their derivatives, 10 carbohydrates and their derivatives, 1 mineral, and 7 other compounds.

The GC–MS metabolomic data was processed using SIMCA 18 and showed significant differences in the characteristics of soybean flour and tempe flour based on their metabolites. PCA analysis showed that both samples could be well clustered in different quadrants (Figure 3). The metabolite components in tempe flour clustered in the positive quadrant relative to the x-axis, while the metabolite components in soybean flour were separated and located in the negative quadrant relative to the x-axis. Both flours had distinct compounds.

Multivariate data analysis using PCA can illustrate class differences in a dataset, such as mapping differences in chemical profiles (53). The output of PCA is a score plot. The R²X value obtained from this PCA model was 0.87 with a Q² value of 0.85. Both values are greater than 0.4, indicating that the resulting model is a good representation of the data (54). Metabolomics analysis serves as a valuable approach for comprehensively characterizing the dynamic metabolic shifts that occur during fermentation, enabling a deeper understanding of the biochemical alterations within the food matrix. Fermentation in food alters the food structure, which can increase or decrease the levels of certain compounds due to degradation processes (55). In the field of foodomics, metabolomics analysis is useful for comprehensively capturing these changes at one time. The complexity of fermentation products can be unraveled by identifying thousands of metabolites using analytical instruments. The metabolomics method used was an untargeted profiling analysis or an exploratory method that does not limit the discovery of specific compounds. By employing an untargeted approach, this study maximizes the potential for identifying a diverse range of metabolites, including novel compounds with potential biological significance (56).

Tempe fermentation is the result of synergistic activity between bacteria and fungi. Lactic acid bacterial fermentation is the first stage of the tempe-making process, which occurs during the soaking of boiled soybeans for 12 h at room temperature. Lactic acid bacteria utilize available sugars within the soybean substrate for growth and metabolic activity, resulting in the production of organic acids. This microbial activity leads to a significant reduction in sugar content, as evidenced by the diminished levels observed in tempe flour (Figure 4).

Tempe fermentation, facilitated by the action of Rhizopus spp. fungi, results in the accumulation of amino acids due to the degradation of proteins. The proteolytic activity of Rhizopus spp. leads to an increase in amino acids such as lysine, leucine, valine, and tyrosine in tempe, contributing to its characteristic bitter taste (28). Fermentation significantly alters the metabolic composition of soybeans, influencing both nutritional and sensory characteristics (28).

Metabolomic analysis revealed the presence of several derivative amino acids, including ornithine and pyroglutamic acid. Ornithine, a non-protein amino acid derived from arginine, has been shown to exhibit stress-relieving and sleep-promoting effects in clinical studies (57), while pyroglutamic acid is a derivative of glutamic acid. Pyroglutamic acid, another non-protein amino acid, is commonly found in fermented products produced by lactic acid bacteria (58) and exhibits anti-inflammatory and antidepressant effects (59). Both ornithine and pyroglutamic acid were found in higher quantities in tempe flour compared to soybean flour.

The OPLS score plot with Pareto scaling in Figure 5A illustrates the grouping based on cholesterol binding activity. Tempe and soybean flours are consistently separated into distinct quadrants. The OPLS model exhibited excellent performance with R²X = 0.98, R²Y = 0.99, and Q² = 0.92, indicating a strong model fit. Figure 5B reveals that the cholesterol binding points align with the tempe flour samples in the positive quadrant of the x-axis. This suggests that the high cholesterol binding activity is likely attributed to the group of metabolites located in the same quadrant. Tempe flour samples clustered closely together in the same quadrant as the cholesterol binding points, indicating that tempe contains metabolites that significantly contribute to its high cholesterol binding activity. Metabolites close to the cholesterol points included amino acids and organic acids. These metabolite components are hypothesized to have a positive correlation with cholesterol-binding bioactivity.

Orthogonal Projections to Latent Structures (OPLS) analysis was employed to examine the correlation between cholesterol binding activity and anti-lipase activity with the metabolite profiles of the two samples. OPLS can be used to assess the relationship between two data matrices, such as chemical profiles and bioactivity data (60). Consequently, this analysis can identify metabolites that contribute to bioactivity. The output of OPLS includes score plots, S-plots, Y-related coefficient plots, and VIP plots. The resulting OPLS model was validated using permutation tests.

Figure 6A presents the OPLS score plot, demonstrating the grouping of samples based on their lipase inhibition activity. Similar to the cholesterol-binding activity, tempe and soybean flour samples cluster distinctly in separate quadrants. The model exhibits excellent performance with R²X = 0.98, R²Y = 0.99, and Q² = 0.93. Tempe samples consistently cluster in the positive quadrant, while soybean flour samples group in the negative quadrant relative to the x-axis.

The OPLS loading plot (Figure 6B) illustrates the relationship between individual metabolites and lipase inhibition. The lipase inhibition point is in the positive quadrant of the x-axis, closely aligned with one tempe sample but within the same quadrant as three other tempe samples. Similarly to the cholesterol-binding activity, all soybean flour samples are positioned in the opposite quadrant. Metabolites closest to the lipase inhibition point, such as amino acids, exhibit a positive correlation with lipase inhibition activity. This suggests that a higher abundance of these metabolites in tempe samples is associated with increased lipase inhibition.

To specifically identify the metabolites contributing to the OPLS model and their correlation with cholesterol binding and anti-lipase activities, Y-related coefficient, VIP, and S-plot analyses were conducted. Metabolites with VIP scores greater than 1 are considered to have a strong influence on the OPLS model and the separation of samples based on the relationship between X (metabolites) and Y (bioactivity) (61). Positive correlation coefficients (>0) indicate a strong positive correlation, while negative coefficients (<0) suggest a negative or reducing correlation (62). S-plots visualize the intensity of the correlation between metabolites and bioactivity based on their quadrant location. Metabolites in the upper right quadrant have a strong positive correlation, while those in the lower left quadrant have a weak negative correlation. Metabolites in the center have a negligible correlation. The combination of VIP, Y-related coefficient, and S-plot provides a comprehensive assessment of the importance of metabolites in the OPLS model and their relationship with bioactivity.

Supplementary Figures S1A–C, S2A–C present the scores for Y-related coefficient, S-plot, and VIP for cholesterol binding and lipase inhibition activities, respectively. Both bioactivities showed a strong positive correlation with amino acids and organic acids, as indicated by the Y-related coefficient and S-plot. Of the analyzed metabolites, 50 exhibited a positive correlation coefficient (>0), while 8 had a negative correlation coefficient (<0) in the Y-related coefficient plot.

The highest correlation score for both bioactivities was observed for phosphate metabolites, while the lowest score was found for lactic acid (−2.28). These results align with the S-plot, which shows phosphate metabolites in the upper right quadrant and lactic acid in the lower left quadrant. A significant advantage of the S-plot is its ability to visually identify metabolites with negligible correlations to bioactivity, as these metabolites are scattered in the central region of the plot. The study revealed that several organic acids, such as malonic acid and phenyl lactic acid, as well as certain carbohydrate derivatives, exhibited weak or negligible correlations with bioactivity. Metabolites with negligible correlations had correlation values close to zero.

To identify the specific metabolites contributing to model formation and sample separation, VIP scores were calculated for both models. The results showed that both models had 12 metabolites with VIP scores greater than 1.0, as presented in Table 3, and Y-related coefficient values ranging from 0.4 to 1.29. Lactic acid was the only metabolite that exhibited a negative correlation with both bioactivities but was a dominant contributor to the OPLS modeling. Of the remaining 11 biomarker metabolites, one was a mineral (phosphate), two were carbohydrates or their derivatives (trehalose and inositol), one was a lipid component (glycerol), four were amino acids (alanine, glycine, tyrosine, valine), and three were organic acids (4-aminobutyric acid, malic acid, pyroglutamic acid).

Phosphates, trehalose, inositol, and glycerol, metabolites formed during fermentation, contributed significantly to the clustering of tempe and soybean flour samples in PCA and OPLS analyses. The presence of these metabolites in higher concentrations in tempe flour can be attributed to the action of fungal enzymes. For instance, Rhizopus spp. possess phytase, which hydrolyzes phytate into phosphate and inositol (63), while yeasts metabolize glucose to produce trehalose and glycerol (8, 64, 65). Regarding the bioactivity of cholesterol binding and lipase inhibition, there have been no reports that the four metabolites identified can directly bind to cholesterol or inhibit lipase activity. Phosphate, for example, is an essential mineral for human health, particularly for bone health. However, a study by Trauttvetter et al. (66) found that the minerals calcium and phosphorus can form amorphous molecules that bind to bile acids in the small intestine, thus reducing blood cholesterol levels. Similarly, a meta-analysis by Tabrizi et al. (67) reported that inositol supplementation in patients with metabolic diseases can help improve blood lipid profiles by reducing cholesterol and triglyceride levels. Likewise, trehalose, a disaccharide component, can improve blood lipid profiles. In vivo studies have shown that trehalose can suppress fat absorption in the small intestine and increase its excretion via feces (68).

Amino acids and organic acids were also major contributors to the increased bioactivity and the OPLS modeling. While the specific mechanism of alanine’s influence on lipid metabolism remains unclear, glycine has been shown to reduce the risk of heart disease by lowering homocysteine levels (69). Branched-chain amino acids like valine have anti-atherogenic effects, and when combined with leucine, can reduce triglyceride and total cholesterol levels while increasing HDL-cholesterol (70). Furthermore, a synergistic effect between tyrosine and tryptophan has been observed, where specific ratios of these amino acids can effectively normalize triglyceride levels and reduce LDL-cholesterol in animal models of high-fat diet-induced obesity (71).

Gamma amino butyric acid (GABA) or 4-aminobutyric acid is similar compound which have four-carbon non-protein amino acid (72), a neurotransmitter, has been reported to have beneficial effects in preventing dyslipidemia. Malic acid and pyroglutamic acid, with VIP values of 1.02–1.19, indicated their significant roles in cholesterol binding and lipase inhibition. Studies on plums, rich in malic acid, have shown anti-lipase activity (73). Given that lipase activity is optimal at a pH of 7.5–8.5 (74), the presence of organic acids in fermented products may inhibit lipase activity by creating an acidic environment (75).

The OPLS models were validated using R²Y and Q²Y values, with values closer to 1 indicating a better fit. The obtained R²Y and Q²Y values for both cholesterol binding and lipase inhibition models were high, suggesting good model performance (Supplementary Figure S3). Further validation was performed using permutation tests and CV-ANOVA (Supplementary Tables S1, S2), which confirmed the robustness of the models. While the study provided valuable insights, the relatively small sample size (n = 8) limited the generalizability of the findings. Future studies with larger sample sizes are recommended to enhance the robustness of the models and provide more conclusive evidence.

GC–MS analysis identified six dominant bioactive compounds: 3-hydroxy-3-methylglutaric acid, 4-aminobutyric acid (GABA—Gamma aminobutyric acid), meso-erythritol, daidzein, and genistein in tempe flour, while 3-phenyllactic acid was predominantly found in soybean flour (Figure 7). These bioactive compounds are known to exert various effects on cholesterol metabolism. Based on the literature, only meso-erythritol and 3-phenyllactic acid have not been directly linked to cholesterol metabolism.

Meso-erythritol, a sugar alcohol, is classified as a food additive by the European Commission Regulation No 231/2012 and is produced through carbohydrate fermentation. With a very low glycemic index (GI: 0), it is considered safe for consumption by individuals with diabetes and obesity (76). While it has negligible caloric content and minimal side effects, excessive consumption may lead to laxative effects. However, studies have shown that consumption of up to 50 g of erythritol does not induce laxative symptoms (77).

3-Phenyllactic acid is a compound produced by lactic acid bacteria and exhibits antifungal properties (78). To date, there is limited literature on its potential hypocholesterolemic or hypolipidemic effects. Therefore, this bioactive compound may have a weak or even negative contribution to cholesterol reduction and lipolysis inhibition based on the OPLS analysis

Another bioactive component, 3-hydroxy-3-methylglutaric acid (HMG) or meglutol, is an organic acid complex with hypolipidemic effects and is commonly found in seeds (79). Recent research by Iman et al. (80) detected increased levels of this compound in fermented edamame, suggesting a role in regulating blood cholesterol. Subsequent research by the same group (81) demonstrated that meglutol can lower LDL-cholesterol in plasma and is found in the highest concentrations in pea tempe. The mechanism of action involves inhibiting endogenous cholesterol synthesis by acting as an inhibitor of HMG-CoA reductase, the enzyme that converts mevalonate to cholesterol (82). This mechanism is similar to that of statins, which are widely used to treat hypercholesterolemia.

While the mechanism of gamma aminobutyric acid (GABA) in lowering blood cholesterol is less well understood, numerous in vivo and clinical studies have shown that GABA supplementation can reduce total cholesterol, LDL-cholesterol, and triglycerides while increasing HDL cholesterol (83–85). GABA has also been shown to prevent atherosclerosis by inhibiting the formation of foam cells in macrophages, thus reducing the risk of blood clots (86). Additionally, GABA can decrease adipogenesis and lipogenesis while increasing energy expenditure. In obese rats, GABA supplementation prevented weight gain despite a high-fat diet (87).

GABA is produced through the decarboxylation of glutamate. It can be synthesized chemically or, more environmentally friendly, through fermentation. During fermentation, lactic acid bacteria produce glutamate decarboxylase (GAD) enzyme, which catalyzes the synthesis of GABA from L-glutamate using pyridoxal-5-phosphate (PLP) as a cofactor (88, 89). Therefore, the fermentation process in tempe production enhances GABA concentration.

Similar to GABA, the soybean isoflavones daidzein and genistein have been reported to have hypocholesterolemic effects, although the mechanisms are not fully understood. Genistein is more potent than daidzein in lowering plasma cholesterol (90, 91). Genistein has the potential to inhibit cholesterol synthesis by targeting HMG-CoA reductase. Computational studies by Hermanto et al. (92) suggest that genistein has a high affinity for HMG-CoA reductase and low toxicity, making it a promising candidate for cholesterol reduction. Isoflavones also exhibit anti-obesity effects, with genistein preventing fat accumulation and daidzein reducing inflammation in obese individuals (93). Isoflavone consumption can prevent adipogenesis through specific mechanisms in the body (94, 95).