Improvement of glucose and lipid metabolism via mung bean protein consumption: clinical trials of GLUCODIA™ isolated mung bean protein in the USA and Canada

The aim of the present study was to confirm the effects of a commercially available mung bean protein isolate (GLUCODIA™) on glucose and lipid metabolism. The main component of GLUCODIA™ is 8S globulin, which constitutes 80 % of the total protein. The overall structure of this protein closely resembles soyabean β-conglycinin, which accounts for 20 % of total soya protein (soya protein isolate; SPI). Many physiological beneficial effects of β-conglycinin have been reported. GLUCODIA™ is expected to produce beneficial effects with fewer intakes than SPI. We conducted two independent double-blind, placebo-controlled clinical studies. In the first (preliminary dose decision trial) study, mung bean protein was shown to exert physiological beneficial effects when 3·0 g were ingested per d. In the second (main clinical trial) study, mung bean protein isolate did not lower plasma glucose levels, although the mean insulin level decreased with consumption of mung bean protein. The homeostatic model assessment of insulin resistance (HOMA-IR) values significantly decreased with mung bean protein. The mean TAG level significantly decreased with consumption of mung bean protein isolate. A significant increase in serum adiponectin levels and improvement in liver function enzymes were observed. These findings suggest that GLUCODIA™ could be useful in the prevention of insulin resistance and visceral fat accumulation, which are known to trigger the metabolic syndrome, and in the prevention of liver function decline.

is a beneficial dietary nutrient for reducing the risk factors that cause diabetes, obesity and CVD. Therefore, β-conglycinin is regarded as the main component producing the beneficial physiological effects of SPI. Tachibana et al. (12) studied normal adult rats consuming a diet containing β-conglycinin over 4 weeks, reporting decreased blood glucose and TAG levels and increased adiponectin levels. These changes were attributed to increased insulin sensitivity (i.e. improved glucose tolerance) of the liver. Several clinical trials have reported that continuous consumption of 2·5-5 g of β-conglycinin resulted in a reduction in visceral fat, serum TAG levels and lipid concentrations, and an increase in HDL-cholesterol concentration (13)(14)(15)(16)(17) . It is postulated that if highly purified β-conglycinin is efficiently and commercially isolated from soyabeans using food industrial methods, it would be possible to demonstrate the beneficial effects by continuous consumption of a small amount of protein. A number of simplified β-conglycinin isolation methods have been reported as being available in the food industry (18)(19)(20)(21)(22) . However, these methods are poor in practicality and have not yet been put to practical use.
The mung bean is a popular food crop in Asia, South America, Australia and the USA (23) . It is generally eaten as a boiled porridge, and mung bean starch is utilised in the form of noodles (bean-starch vermicelli), although the isolated protein itself is not typically consumed as a food. The main component of mung bean protein, accounting for over 80 % of the protein, is 8S globulin. 8S globulin has exhibited the highest degree of sequence identity (68 %) and structural similarity with β-conglycinin (24) . Mung bean protein isolate (MuPI) is expected to exhibit a four times stronger beneficial function on human health than SPI, in which β-conglycinin accounts for only 20 % of the total protein. 8S globulin was found to contain bioactive peptides that block angiotensin II-converting enzyme activity in vitro and was therefore hypothesised to alleviate hypertension (25) . An in vivo study in hamsters found that mung bean supplementation in their diets could decrease plasma cholesterol and TAG concentrations by up-regulating cholesterol-7α-hydroxylase (CYP7A1) mRNA and protein levels (26) . Of particular importance to the present study, GLUCODIA™, which is a commercially available MuPI produced as a by-product during starch production, was found to lower blood TAG levels in normal rats by inducing adiponectin and reducing TAG synthesis via insulin signalling (27) . In the present study, we examined the effect of GLUCODIA™ in a clinical trial setting and assessed its effects on metabolic syndrome-related factors.
Experimental methods (used in both the pre-study and the main study)

Experimental food
MuPI (GLUCODIA™) was manufactured using a method identical to that used for the production of SPI from the by-product containing the protein at the time of mung bean starch production and administered in the form of candy. GLUCODIA™, consisting of 92 % protein, 3 % minerals and 5 % water, was sterilised for food use and spray-dried.
The composition of a test candy is shown in Table 1. In the control experiment, control candy with the same taste as the test candy was produced by replacing MuPI with the milk protein casein (Table 1). Therefore, in the case of the 6-g group, 4·4 g of 8S globulin were taken per d: 8S globulin (4·4 g) = GLUCODIA™ (6·0 g) × protein content (92 %) × 8S globulin content for total protein (80 %). So, the 3-g group consumed 2·2 g as 8S globulin, and the 1·5-g group consumed 1·1 g 8S globulin, respectively.

Subjects
Test subjects were healthy men and women. Individuals undergoing treatment for hyperlipidaemia, diabetes, or liver dysfunction and those with food allergies were excluded from the study. Each study screened subjects based on health diagnostic results as shown in Table 2. Subjects were of non-child-bearing potential and included the following: women with a past history of hysterectomy, oophorectomy, or bilateral tubal ligation; post-menopausal (natural or surgically with >1 year since last menstruation) women; and 100 100 100 100 * The 3-g test and control chewable tablets were used in the pre-and main studies.
women who agreed to use a medically approved method of birth control and who had a negative urine pregnancy test result. Subject recruitment was conducted at each research institution. The following recruitment methodologies were implemented to identify subjects for the study: (A) Online recruitment: online advertising was ran almost daily throughout the study recruitment period.
(B) Database: a recruitment database was searched to identify potentially eligible subjects.
A participant withdrew from the study if it was determined by the qualified investigator that it was not in the participant's best interest to continue. This included adverse events (AE) or serious AE related to the investigational product causing clinically significant illness, the need for a prohibited concomitant medication, or a female participant who became pregnant during the course of the trial.

Test design
Each study was a double-blind, placebo-controlled trial. The outline of the test design is shown in Table 2. The numbers of subjects who participated in the pre-study and main study were set with reference to the results of clinical studies on soyabean protein (β-conglycinin) (28,29) . The subjects in the test group were instructed to consume specified pieces of candy twice daily, before breakfast and dinner. Subjects in the control group were administered the casein candy in the same manner. The subjects recorded information, such as the presence or absence of consumption of the test candy, whether or not they took medications, alcohol consumption, and their physical state, in a diary from the start of the trial to the conclusion of the diet consumption period. Furthermore, the subjects recorded their diet and dietary content for the duration of the test period.

Data collection
Data collection for each study was comparable except for body composition, which was assessed using dual-energy X-ray absorptiometry in the main study. Plasma samples were collected from each participant periodically and beginning at the start of the consumption period. Plasma samples were collected at 09.00 hours in the morning after fasting from 22.00 hours the previous night. The following serum biochemistry measurements were recorded: changes in fasting plasma glucose (FPG) and fasting insulin levels, lipid panels (total cholesterol, TAG, LDL-cholesterol, HDL-cholesterol, NEFA) and hepatic functional enzyme (aspartate aminotransferase, serum glutamic oxaloacetic transaminase, alanine aminotransferase (ALT), serum glutamic pyruvic transaminase and γ-glutamyltransferase) levels. The homeostatic model assessment of insulin resistance (HOMA-IR) was calculated as follows, based on each subject's FPG and fasting insulin levels: HOMA-IR = fasting insulin level (μU/ml) × FPG level (mg/ dl)/405 (HOMA-IR = fasting insulin level (μU/ml) × FPG level (mmol/l)/22·5). Adiponectin level was measured only in the main study. The study safety indicators included the following: changes in blood pressure and heart rate, and blood safety parameters (complete blood count and comprehensive metabolic panel results: blood urea N, creatinine, total protein, albumin, globulin, total bilirubin, Ca, Cl, Na and K levels). The frequency and severity of AE and serious AE were recorded in detail, based on information obtained from interviews during each visit.
In the data analysis of the main study, data of subjects with HOMA 1·7 or more were analysed as low insulin sensitivity subjects, as well as analysis of all subjects except those who dropped out.

Statistical analysis
Paired t tests were used to assess within-group changes from baseline to each time point in all studies. Comparisons among product groups were assessed by one-way ANOVA (pre-study), and Student's unpaired t test was used to assess differences between the groups at the end time of the test (main study). Data in Tables 3-5 are presented as means and standard deviations and a significance level (two-tailed) of <0·05 was considered to indicate a significant difference. A significance level (two-tailed) between 0·05 and 0·10 was set as a statistically significant tendency.

Adverse events
An AE was defined as any untoward medical occurrence in a clinical investigation participant who was administered an investigational product which did not necessarily have a causal relationship with the product. An AE could be any unfavourable and unintended sign (including an abnormal laboratory finding), symptom, or disease temporally associated with the use of a product, whether or not it was considered related to that product. Pre-existing conditions which worsened during the study were reported as AE. During the study, participants recorded AE in their diary. At each visit the participant was asked 'Have you experienced any difficulties or problems since I saw you last?'. Any AE were documented in the study record and were classified according to the description, duration, intensity, frequency and outcome. The qualified investigator assessed any AE and decided causality. Intensity of AE was graded on a three-point scale (mild, awareness of event but easily tolerated; moderate, discomfort enough to cause some interference with usual activity; severe, inability to carry out usual activity), and reported in detail in the study record. The causality relationship of the investigational product to the AE was assessed by the qualified investigator as either: Most probable: There was a reasonable relationship between the investigational product and the AE. The event responded to withdrawal of investigational product (dechallenge) and recurred with rechallenge when clinically feasible.
Probable: There was a reasonable relationship between the investigational product and the AE. The event responded to dechallenge.
Possible: There was a reasonable relationship between the investigational product and the AE. Dechallenge information is lacking or unclear.
Unlikely: There was a temporal relationship to investigational product administration but there was no reasonable causal relationship between the investigational product and the AE.
Not related: No temporal relationship to administration of the investigational product or there was a reasonable causal relationship between non-investigational product, concurrent disease or circumstance and the AE.

Ethical considerations
Each study was conducted according to the guidelines set out in the Declaration of Helsinki, and written informed consent was obtained from all subjects/patients. Informed consent was obtained from each participant at the screening visit prior to any study-related activities being performed. Institutional review board approval of the protocol of pre-study was granted

Pre-study
Subjects. The characteristics of the subjects are shown in Table 3. During the evaluation of efficacy, data from three subjects markedly reduced the accuracy of that evaluation; therefore, the three subjects were excluded from the analysis. The reasons for exclusion were: (1) non-compliance with failure to consume the study product (n 2); and (2) lost to follow-up (n 1). Finally, data from twenty-nine subjects were analysed.
Evaluation of safety. There were no statistically significant differences in blood pressure, heart rate and the blood safety FPG, fasting plasma glucose; HOMA-IR, homeostatic model assessment of insulin resistance. * Significant difference between baseline and week 4 data (P < 0·05). † Significant difference between test and control data (P < 0·05). ‡ Significant tendency between test and control data (P < 0·1). § To convert glucose from mg/dl to mmol/l, multiply by 0·0555. ǁ To convert TAG from mg/dl to mmol/l, multiply by 0·0113.
parameters within each group and between the test and control groups at any time point. The frequency and intensity of AE were established from interviews occurring during each visit. One subject complained of psychological AE (irritability, impatience and fatigue), and two subjects complained of gastrointestinal AE (diarrhoea, abdominal cramping, bloating and nausea). By the qualified investigator, intensities of all observed AE were classified as 'mild' on a three-point scale, and the causality relationships of the investigational product to these AE were assessed as 'not related'.
Fasting plasma glucose, insulin and homeostatic model assessment of insulin resistance results. Mean FPG, insulin and HOMA-IR are shown in Table 3. The FPG levels in the 6-g and 3-g groups decreased during the study period. There were significant differences between the net changes in FPG level in the 6-g and control groups, and between the 3-g and control groups, respectively (P < 0·05, for both). The insulin levels in the test groups decreased dose-dependently. However, the net changes did not differ significantly between the test and control groups. Similarly, the HOMA-IR values in the test groups showed a dose-dependent decrease (Table 3). There was a significant tendency in the net change in HOMA-IR values between the 6-g and the control groups (P < 0·1).

Main study
Subjects. Six subjects were removed from the per-protocol analysis (Fig. 1) for the following reasons: three subjects did not complete the study (continuous intakes of the test meal were refused); one subject was lost to follow-up; one subject withdrew consent from the study; and one subject was withdrawn due to poor compliance with the study protocol.
Evaluation of safety. There were no significant differences in the haematological and clinical chemistry parameters between the two groups. Differences in the following variables were significant within the test group (corpuscular Hb, 29·6 (SD 6·6) to 29·3 (SD 8·0) pg, P < 0·05; corpuscular Hb concentration, 340 (SD 33) to 338 (SD 42) g/l, P < 0·05; urea, 5·2 (SD 1·1) to 5·6 (SD 1·4) mmol/l, P < 0·05). In the control group, differences in the following variables were significant: corpuscular volume (88·1 (SD 14·1) to 87·8 (SD 12·2) fl; P < 0·05), corpuscular Hb (30·3 (SD 5·2) to 29·9 (SD 5·1) pg; P < 0·01) and Na (141·3 (SD 9·4) to 140·0 (SD 6·1) mmol/l; P < 0·05). However, these changes were within the normal clinical ranges and were not of clinical significance. A total of eight AE were reported, of which five occurred in the test group and three in the control group. The AE were experienced by five subjects: five AE were reported by two subjects in the test group, and three AE were reported by three subjects in the control group. The five AE in the test group were: dyspepsia (n 2), infrequent bowel movements (n 1) (both gastrointestinal disorders), headache (n 1) and thirst (n 1); the three AE in the control group were: nausea (n 1), upper abdominal pain (n 1) (both gastrointestinal disorders) and headache (n 1). By the qualified investigator, intensities of all observed AE were classified as 'mild' on a three-point scale, and the causality relationships of the investigational product to these AE were assessed as 'unlikely' or 'not related', respectively.
Fasting plasma glucose, insulin and homeostatic model assessment of insulin resistance (all subjects). The mean FPG and insulin test results are shown in Table 4. The FPG levels in the test and control groups did not change during the meal consumption period. Fasting insulin levels in the test group showed a significant decrease from 14·9 (SD 6·0) to 10·8 (SD 4·7) µU/ml (P < 0·01). In contrast, fasting insulin levels in the control group showed a slight increase. The HOMA-IR values in the test group showed a significant decrease tendency, from 3·8 (SD 1·8) to 2·9 (SD 1·9) (P < 0·01), while these values increased in the control group (Table 4).
Lipid panel (all subjects). TAG levels in the test group showed a significant decrease tendency from 153 (SD 70) to 142 (SD 70) mg/dl (1·73 (SD 0·79) to 1·60 (SD 0·79) mmol/l) (P < 0·1). Conversely, TAG levels in the control group slightly increased ( Table 4). The total cholesterol and LDL-cholesterol levels in the test and control groups did not change during the meal consumption period. HDL-cholesterol levels in the two groups showed a significant decrease (test group, P < 0·05; control group, P < 0·01, Table 4). NEFA levels decreased in the test group and increased in the control group.
Hepatic functional enzymes (all subjects). Aspartate aminotransferase and γ-glutamyltransferase levels in the test and control groups did not change during the meal consumption period. ALT levels in the test group significantly decreased from 35·3 (SD 15·1) to 30·6 (SD 13·6) IU/l (P < 0·05). Conversely, ALT levels increased slightly in the control group (Table 4).
was a significant difference between the net changes in the test and control groups (P < 0·01).
Background information on participants with low insulin sensitivity. Subjects with HOMA-IR ≥ 1·7, as mild or above insulin resistance subjects, were selected (30) . The background information on these subjects is shown in Table 5.
Lipid panel (participants with low insulin sensitivity). TAG and NEFA levels in the test group decreased considerably. The net changes in TAG levels in the test and control groups were −15 (SD 37) and 10 (SD 33) mg/dl (−0·17 (SD 0·42) and 0·11 (SD 0·37) mmol/l), respectively, and a significant difference was observed between the two groups (P < 0·05). The net changes in NEFA levels in the test and control groups were −85 (SD 379) and 45 (SD 286) mEq/l, respectively. The difference between the groups was not statistically significant (Fig. 3).
Body composition (participants with low insulin sensitivity).
The BMI of the test group decreased slightly while the BMI in the control group increased (data not shown). Body composition measurements by dual-energy X-ray absorptiometry revealed a decrease in body fat mass and an increase in lean body mass in the test group (body fat mass, 33·8 (SD 31·4) to 33·6 (SD 35·6) kg; lean body mass, 48·7 (SD 40·3) to 49·0 (SD 42·2) kg). Conversely, in the control group, body fat mass increased and lean body mass decreased (body fat mass, 34·2 (SD 47·3) to 34·7 (SD 48·3) kg; lean body mass, 57·1 (SD 57·2) to 56·9 (SD 52·1) kg). The differences in body fat mass and lean body mass within each group and between the test and control groups were not statistically significant.

Discussion
The present two studies on MuPI (GLUCODIA ™ ) are promising phase II clinical studies that provide a clear scientific rationale for the investigation of specific parameters, such as glucose and liver enzyme chemistry, in future trials. The safety profile of MuPI was evaluated during these two studies and the dose was both safe and well tolerated. Safety parameters remained within clinically acceptable ranges, and all subjects were able to complete the study except for dropouts due to personal circumstances. There were no significant differences in safety parameters between the groups. In the pre-study, FPG levels in the MuPI 6-g and MuPI 3-g groups decreased after the consumption of the study product (P < 0·1, for both) while an increasing trend was observed in Fig. 2. Box plots of net changes in the parameters of glucose metabolism among participants with low insulin sensitivity: (a) fasting plasma glucose (FPG); (b) insulin; (c) homeostasis model assessment of insulin resistance (HOMA-IR). Whiskers represent the minimum and maximum values, and the boxes represent the lower and upper quartiles. ** Significant difference between the test (GLUCODA™) and control groups (P < 0·01). To convert glucose from mg/dl to mmol/l, multiply by 0·0555. the control group. There were significant differences in the net changes in FPG levels between the MuPI 6-g and 3-g groups and the control group (P < 0·05, for both). The trend for fasting insulin was similar to that for FPG. The fasting insulin levels in the MuPI 6-g and 3-g groups showed a decreasing trend, while an increasing trend was observed in the control group. HOMA-IR values for both the MuPI 6-g and the 3-g groups decreased, although these changes were also not statistically significant. The majority of the lipid parameters were within the normal range at baseline. However, TAG levels for the MuPI 3-g group were high at baseline. They decreased progressively and, by week 4, these were under 100 mg/dl (1·13 mmol/l). BMI showed a similar trend to FPG and insulin. Based on these results, the intake of 3 g of MuPI was considered the minimum dose required to produce physiological effects.
Results from the main study showed that mung bean protein was effective at significantly reducing insulin concentrations after an 8-week supplementation. A significant decrease in HOMA-IR and increase in adiponectin levels in the test group relative to the control group were observed. Additionally, though it was not significant, a decreasing trend in NEFA in the test group relative to the control group was observed. Insulin resistance has emerged as a major pathophysiological factor in the development and progression of diabetes since it is evident in susceptible individuals at the early stages of diabetes, and particularly type 2 diabetes. Insulin resistance is usually defined as a value greater than the 75th percentile value for nondiabetic subjects according to the WHO (31) . Furthermore, assessment of insulin resistance using the HOMA-IR is a key index for the primary prevention of diabetes and is recommended for screening of high-risk groups. Different HOMA-IR cut-off values have been reported (30) . This study examined subjects with a HOMA-IR value over 1·7 in order to evaluate the improvement of consuming GLUCODIA™ on glucose tolerance in subjects with the most mild insulin resistance. The beneficial effects of GLUCODIA™ on glucose tolerance were greater in subjects with low insulin sensitivity. In this study, there were no significant differences between the groups in terms of body fat, fat body mass and lean body mass by body dual-energy X-ray absorptiometry scans (32) . However, decreasing trends in percentage of body fat and fat body mass and an increasing trend in lean body mass were observed among subjects in the MuPI group. Furthermore, plasma adiponectin concentrations significantly increased in the MuPI group. The negative correlation between visceral adiposity and adiponectin levels may be partly explained by the increased secretion of this adiponectin from accumulated visceral fat (33) . Therefore, the decreasing trends in percentage of body fat and fat body mass suggest a reduction in visceral fat. A decrease in body weight and body fat, resulting from a decrease in visceral fat, may be more predictably expected with ingestion of MuPI over a longer period. Blood lipid parameters and hepatic functional enzymes (aspartate aminotransferase, ALT and γ-glutamyltransferase) were found to be within the normal range, except for TAG and ALT. TAG levels at baseline for the MuPI group were slightly above the normal range, although at week 8 TAG levels had decreased, and this trend was particularly noticeable among subjects with low insulin sensitivity (HOMA ≥ 1·7). Thus, there were significant differences in the net changes in TAG levels between the MuPI group and the control group. High ALT has been reported to be an independent risk factor for non-alcoholic fatty liver disease (NAFLD) (34) . NAFLD is known to be associated with diabetes and is an important feature of the metabolic syndrome and insulin resistance (35) . In their study on mice, Watanabe et al. (36) reported that MuPI may play a preventative role in NAFLD by reducing hepatic lipid accumulation. MuPI has been reported to reduce plasma TAG levels by the active suppression of fatty acid synthase by inhibition of sterol regulatory element-binding protein 1 (SREBP-1) expression in rat liver (27) . The results obtained in these studies are in good agreement with the results reported for β-conglycinin (12,37) , which closely resembles the 8S globulin of MuPI with respect to structural properties (24) . Therefore, 8S globulin and β-conglycinin may have peptides of consensus sequence that exert these beneficial effects. However, Watanabe et al. (36) reported that MuPI may have different gene expression systems from SPI, including β-conglycinin, in de novo lipogenesis (DNL)-related gene-expression control. It is known that DNL, fatty acid uptake and fatty acid oxidation are involved in the process of hepatic lipid accumulation (38) . Moreover, Watanabe et al. (36) reported that the suppressive effect of MuPI on DNL-related gene expression was not induced by a mixture of amino acids matching the composition of MuPI, and this effect is not related to the amino acid composition of MuPI per se. This report postulated the existence of a unique MuPI peptide sequence, distinct from the sequence in soya protein contributing to improvement of the lipid metabolism in the liver.
MuPI (GLUCODIA ™ ) is safer than soya protein in terms of being a non-GM organism and there are no reported cases of allergies as in soyabeans. In the present study, it was shown that GLUCODIA ™ has beneficial effects on the metabolic syndrome and it is a food material that can be expected to have these effects at a smaller amount of intake compared with the intake recommended by the Food and Drug Administration for soya protein.