The Effect of Soy and Whey Protein Supplementation on Glucose Homeostasis in Healthy Normal Weight Asian Indians

Milk and legumes are good source of protein foods used to sustain muscle mass, but their effects on postprandial glucose homeostasis and energy metabolism may be different. This is relevant, for example, in the dietetic response to obesity or diabetes, where the intake of high-quality protein is often increased significantly. The objective of this study was to characterize the acute effect of whey and soy protein (15% vs. 30%) on glucose homeostasis, energy metabolism, and satiety. Healthy, normal body mass index (BMI) Indian adult males aged 20–35 years (n = 15) received 4 test meals (2 proteins (soy vs. whey) and 2 doses (15% vs. 30% protein: energy ratio)). Blood samples were collected serially after the meal to calculate the incremental area under the curve for plasma glucose and insulin. Energy expenditure and substrate oxidation were measured after the meal. Satiety was measured with a visual analogue scale. The insulin response, represented by the incremental area under the curve, was significantly higher for the 30% whey compared to the 30% soy protein meal (p < 0.01) but was not significantly different between the 15% protein doses. There were no differences in the plasma glucose response across protein sources or doses. The mean peak fat and carbohydrate oxidation, satiety, and energy expenditure did not differ between the protein sources and doses. In conclusion, at higher doses, whey protein has a greater insulinogenic response, compared to soy protein, and exhibits a dose-response effect. However, at lower doses, whey and soy protein elicit similar insulinogenic responses, making them equally effective protein sources in relation to glucose homoeostasis.


Introduction
Te impact of dietary protein intake on glucose homeostasis through its action on insulin secretion has gained attention especially among Asian Indians due to growing burden of type 2 diabetes (T2D) [1]. Even before changes among T2D are explored, it is imperative to understand the impact of protein intake among healthy individuals. Tis is because nonpharmacological strategies are important in the prevention and management of T2D, and data from healthy population could form the foundation for the same. Te role of postprandial hyperglycemia and its consequences have been of interest in understanding the pathophysiology of T2D. It has been established that, during the postprandial phase, a rapid increase in blood glucose levels, including hyperglycemic spikes, occurs in those with T2D [2]. Tis may be relevant as postprandial hyperglycemia has been linked to complications.
Nutrient preload, by manipulating the sequence of macronutrient ingestion during the meal, is one of the novel nutritional approaches that have proven efective in reducing postprandial hyperglycemia [3]. Te benefcial efect of noncarbohydrate nutrient preloads includes their ability to promote insulin secretion [4]. However, the mechanisms driving hyperinsulinemia are not clear. For instance, this could be linked to specifc amino acid concentrations in protein preloads [5]. In an Asian Indian context, the consumption of quality protein is poor [6]. Two common sources of protein consumed among Asian Indians are milk and legumes [7,8]. Te acute postprandial glycemic and insulinogenic efect of milk (whey) protein is diferent from legume (soy) protein [9]. Tis has been attributed to the difering proportions of branched-chain amino acid content of these proteins. However, the consumption of milk/dairy products and resultant hyperinsulinemia has been suggested to produce less than desirable long-term efects in healthy individuals, including insulin resistance [10]. Tus, as a start, there is a need to understand the efects of doses of specifc protein intakes on glucose homeostasis in healthy Asian Indians. Currently, there are no data available regarding the type (milk vs. legume) and the dose of protein intake on glucose homeostasis among Asian Indians. Terefore, the primary aim of the present study was to evaluate the acute efects of two proteins, whey protein (WP) and isolated soy protein (ISP), and two doses of each (15 and 30% of energy) on glucose homeostasis. Te secondary aim of the study was to explore the changes in energy metabolism and satiety following WP and ISP (15 and 30% of energy) among healthy Asian Indians.

Subject Recruitment.
Healthy adult males between the ages of 20-35 years were recruited from in and around St. John's Medical College and Hospital, Bengaluru, India. All participants were screened for inclusion and exclusion criteria. Participants were excluded if they were underweight (BMI <18.5 kg/m 2 ), had a history of acute weight loss, and diagnosed with T2D, hypertension, hypercholesterolemia, hyperbilirubinemia, anaemia, cancer, lactose intolerance, protein, or any food allergy, or if they were on any medication or medical condition which could afect the selected outcome measures. In total, 23 participants were screened, out of which 7 participants were excluded due to the exclusion criteria. Of the total 16 participants who were recruited and randomized, 15 participants completed all 4 experiments: 1 subject dropped out after 2 experimental days due to an acute medical condition (dengue fever) not related to the study. Te study obtained ethical approval from St John's National Academy of Health Sciences Institutional Ethics Committee on January 23, 2015. Te IEC study reference number is 156/2014 (clinical trial registration number CTRI/2018/03/012426). Te experimental protocol was explained to all the participants, and their written informed consent was obtained.

Sample Size Estimation.
Te sample size was estimated based on hyperinsulinemic response to 50% soy and whey protein [11]. Te comparison of AUC of insulin between soy and whey protein was considered as the primary outcome for estimating the sample size. To observe a minimum diference of 7.5 nmol/120 min between 30% soy and whey protein with the standard deviation of 5 nmol, 80% power and 1% level of signifcance (Bonferroni adjustment for multiple comparisons within cross overtrial) were required, and the sample size required was 12. Te current study was able to achieve a sample size of n � 15.

Method of Randomization and Concealment.
Te computer-generated randomization sequence for the order of intervention and sequence allocation was generated by an independent statistician. Te order of intervention for each participant (30% whey, 30% soy, 15% whey, and 15% soy protein) was randomly assigned (a sequence number that is 1 to 4). In total, 16 participants were randomly allocated into 4 treatment sequence using block randomization (a block size of 4). A copy of all randomization lists was maintained in a sealed envelope with an independent authority in the research institute. Investigators were blinded, and the random allocation of the study participants was carried out by an independent person.

Questionnaire and Body Composition.
Each participant underwent detailed dietary history, clinical, and anthropometric examination. Body fat and lean mass were measured by using dual X-ray absorptiometry (DXA). Whole-body and regional body compositions were estimated using the Lunar Prodigy Advanced PA+301969 (GE Medical Systems, USA) whole-body scanner, with software version 12.30. DXA scans were performed with the subject wearing light clothing and no metal objects, by the same laboratory technician. Te mass of the lean soft tissue, fat, and bone mineral for the whole body and specifc regions were obtained [12].

Experimental Details.
Participants reported to the metabolic ward in the evening prior to testing. All participants received a standard evening meal, calculated as a quarter of the daily energy requirement. Te composition of dinner was constant for all the participants and was served at the metabolic kitchen in the division of nutrition, following which all the participants slept in the metabolic ward for at least 8 hours. Te participants were woken up at 5 am, and the frst voided urine sample was collected. Tey were taken to the adjoining metabolic laboratory, intravenously cannulated and rested for 30 minutes. Basal blood samples were collected, following which their resting energy expenditure and substrate oxidation were measured by indirect calorimetry and visual analogue scales (VAS), for appetite was administered before the consumption of the test meal. Physical activity level (PAL) were calculated based on a physical activity questionnaire [13]. 2 Journal of Nutrition and Metabolism Te test meal was vanilla favored and weighed 85 g with an energy content of 400 kcal. Te test meal was a liquid meal with macronutrient content (15% soy, 30% whey protein, 50% carbohydrate, and 20% and 30% fat (Figure 1)), and gastric emptying time could play an impact on the hormonal and satiety responses [14][15][16]. Te test meal protein consisted of either isolated soy (SUPRO ® isolated soy protein) or whey protein in two doses each. Te meals were matched for all components, but the only varying components were the protein sources. Te carbohydrate source was predominantly sugar with a small amount of maltodextrin. Te fat source was high oleic sunfower oil.
Each participant randomly received 4 test meals on diferent days, separated by at least 3 days of washout. Te protein type (soy and whey) and dose (15% and 30%) of the test meals are presented in Figure 1. Te test meals were prepared by adding the preprepared meal powder to 300 ml water at room temperature and making sure it was homogenously distributed. Te test meals were consumed in 5-10 minutes, and the entire consumption was ensured by weighing the containers before and after consumption. All participants could complete their test meals. Te study measurements continued for 5 hours postprandially. Details of the study protocol are represented in Figure 1. Blood samples were collected at 15, 30, 45, 60, 90, 120, 180, 240, and 300 min after the test meal. Whole blood glucose was estimated immediately (described below). For plasma insulin measurements, blood was collected in heparinized tubes, cold-centrifuged, and plasma stored at −80°C until analysis. Te urine sample was collected at the end of the test meal for urinary nitrogen analysis. A minimum window of 3 days was maintained between the experiments. All 4 experiments were completed within 21 days.

Visual Analogue Scales.
Individual subjective indices of appetite were recorded for the duration of the experiment. Four 100 mm visual analogue scales for hunger, thoughts of food, urge to eat, and fullness of stomach were administered. Te scale was administered each time in triplicate, and a mean of 3 readings were expressed as percentage of scale [17]. Te time points during which VAS was administered have been indicated in Figure 1.  Journal of Nutrition and Metabolism 2.5.2. Biochemistry. Plasma glucose measurements were performed by the glucose oxidase method on a bedside glucose analyzer (GM9D, Analox Instruments, London, UK). Te intraassay coefcient of variation for this method (using 144.1 mg dL-1 (8 mmol L-1) standards) was <1%, while the interassay coefcient of variation has been <5%. Blood samples for insulin measurements were analyzed by electrochemiluminescence (Elecsys 2010, Roche Diagnostics, Manheim, Germany). Te intraassay CV for insulin was 3%, and the interassay CV was 1.3%. Te urinary nitrogen was analyzed by the micro-Kjeldahl method [18].

Calorimetry.
Calorimetry was performed using a ventilated hood by using an open-circuit calorimeter. Flow calibration was undertaken by burning a known quantity of 99% pure alcohol and measuring the total O 2 consumption and CO 2 production. Te measurement of minute-to-minute oxygen consumption (VO 2 ) and carbon dioxide (VCO 2 ) production was made at the baseline and at the end of every hour (last 15 minutes) for 5 hours after the experimental meal. Te respiratory quotient (RQ) was calculated as the ratio of VCO 2 to VO 2 . Te resting energy expenditure (EE) was calculated by the Weir formula [19]. Substrate oxidation was calculated from the gas exchange values using stoichiometric equations [20]. In brief, the nonprotein RQ was calculated from gas exchange corrected for protein oxidation based on the urinary nitrogen excretion at the baseline and following the protein meal by timed urine collections. Te nonprotein RQ was used to calculate fat and carbohydrate oxidation. Tese rates were examined every hour (g/min) and compared between the WP (15% vs. 30%) and ISP (15% vs. 30%).

Statistical Analyses.
Tis was a double-blind randomized trial, with randomization codes generated by an independent statistician. Baseline characteristics were reported using mean and standard deviation. Te outliers were detected using box plots and the generalized extreme studentized deviate test and were removed from statistical analysis. Assumption of normality was checked using the Kolmogrov-Smirnov test and the Q-Q plot. Non-normal data were log transformed. Te mean and the standard error of mean for plasma glucose and insulin were plotted over time.

Results
Baseline characteristics of the study participants are given in Table 1. Plasma glucose and plasma insulin responses to various test meals (mean ± SE of mean) are illustrated in Figure 2. Tere was a signifcant change in the mean plasma insulin (μU/mL) in each of the test meals over time. Te mean plasma insulin peaked at 30 minutes in all the four interventions and reached the baseline at 240 minutes, nonsignifcantly from the baseline. Te mean iAUC for plasma insulin was signifcantly diferent between the four test meals (p < 0.001), i.e., 7384, 8532, 7609, and 11057 for ISP 15%, ISP 30%, WP 15%, and WP 30%, respectively. Te post hoc test showed that the mean iAUC for plasma insulin after the ingestion of 15% WP was signifcantly lower than 30% WP (p � 0.004), 30% ISP was signifcantly lower than 30% WP (p < 0.001), and the mean iAUC for plasma insulin for 15% ISP was also signifcantly diferent from the 30% WP iAUC. However, the mean insulin of 15% ISP was not signifcantly diferent from that of 15% WP and 30% ISP. Te mean iAUC for plasma glucose (mmol/L) was 70.71, 63.53, 65.01, and 85.06 for ISP 15%, ISP 30%, WP 15%, and WP 30%, respectively. Tere was no signifcant diference in the mean iAUC of plasma glucose levels between the test meals (p � 0.11). Te peak response and iAUC for the calorimetry measure of VO 2 , VCO 2 , RQ, EE, and VAS scores are presented in Table 2. Tere was no signifcant diference in the mean iAUC values of calorimetry measures, VCO 2 (L/ min), RQ, and VAS score between the four interventions except for VO 2 (p � 0.04) and EE (p � 0.06). Te mean iAUC for VO 2 (L/min) of 30% WP was signifcantly higher than that of 15% ISP. Te mean iAUC of EE was noted to be higher in 30% WP than that in 30% ISI and 15% WP. While comparing the peak responses, the mean peak response VO 2 (L/min) was signifcantly higher for 30% WP than that for 15% WP. For EE measure, 30% WP was signifcantly higher than 15% WP meal; similarly, 30% soy was signifcantly higher than 15% soy. None of the other measures were signifcantly diferent between the test meals. Data on calorimetry measures, VO 2 , VCO 2 , RQ, and EE over the 5-hour measurement period are presented in Supplementary Figure 1. Tere was no signifcant interaction (time × group efect) noted when calorimetry measures over the 5-hour measurement period were compared between the test meals except for EE (p � 0.035). At the baseline, there was no signifcant diference in % fat and CHO oxidation across the four meals. Peak decrement response in % fat oxidation and increment in CHO oxidation over 5 hr was

Discussion
Te current study demonstrated that insulinogenic response was signifcantly higher for 30% WP than that for 30% ISP among healthy normal weight Asian Indians. At a lower dose (15%), WP and ISP elicited similar insulinogenic responses among normal weight healthy Asian Indians. Tere was no diference in plasma glucose across protein sources or doses. Te mean peak % fat, CHO oxidation, satiety, and energy expenditure did not difer between protein source and doses. Te habitual protein intake of high-quality sources is low among Asian Indians [6]. Te increased prevalence of chronic diseases, especially T2D, has promoted initiatives to explore the role of protein intake on glucose and energy metabolism [21]. Te key aspects explored as part of the present study were related to the protein consumption, i.e., the type and amount of protein. Te data from the present study demonstrated that lower doses of WP and ISP were similar (i.e., insulinogenic), but at higher doses, WP had a greater insulinogenic response. Te fact that, at the lower dose, both the protein types demonstrated similar response was promising, as this dose is translatable to clinical medicine/nutrition. For instance, the consumption of legumecontaining foods may contribute to a lower incidence of postprandial hyperglycemia preventing complications associated with it including coronary heart diseases, atherosclerosis, T2D, and carcinogenesis [22]. Te insulinreleasing capacity of WP and ISP could be attributed to their protein fraction [23]. Te mechanism by which ISP could induce hyperinsulinemia is linked to higher amino acid alanine and arginine levels [9], stimulating the secretion of glucose-dependent insulinotropic polypeptide (GIP) [9]. Te higher branched-chain amino acid (BCAA) levels in specifc leucine concentration in WP could lead to a greater glucagon-like peptide-1 (GLP-1) response, which may, in turn, be responsible for the elevated insulin release [24]. However, the pathways by which ISP and WP might operate the insulinogenic action remain unknown. Te lack of diference in insulinogenic response seen at a lower dose of protein in the present study needs further exploration. In a study on healthy individuals, the efects of casein, soy, and whey protein on various parameters including insulin response at diferent doses (10% and 25%) were studied [25]. Te study demonstrated that, at a higher dose, insulin response was greater for whey than that for soy or casein. At a lower dose, insulin concentration increased more for casein than for soy or whey with no diferences between whey and soy. Whey protein is considered a fast-absorbable protein resulting in greater aminoacidemia and a higher beta cell secretion than the ingestion of a slow absorbable protein like casein or soy protein [26]. Whether protein induced insulin response is dose dependent and there is a threshold at which the diferential response starts emerging needs further exploration. Te evidence towards the same could be seen in mouse islets models which demonstrated that insulin secretion depends on amino acid doses and glucose levels [27]. A recent systematic review and meta-analysis investigating the efect of replacing animal with plant protein on glycemic control among patients with diabetes observed that a higher plant protein intake resulted in better glycemic control [28]. However, the body of evidence comparing plant and animal protein intake on glycemic control and T2D risk has produced inconsistent results [29]. Tis is due to diferent types of commonly consumed plant proteins (e.g., soy, nut, seeds, beans, peas, and lentils) and animal proteins (e.g., meat, milk, fsh, and eggs) that have been studied, each with their own set of protein quality characteristics and nonprotein components [30]. Te current study, for the frst time, explored the 2 most consumed protein sources of animal (whey) and plant (soy) protein and their impact on glucose homeostasis among Indian population. Legume (ISP) being an afordable protein could be a good alternative source, particularly in a setup where vegetarian sources might be acceptable culturally. Te role of lean mass on glucose homeostasis afecting the cardiometabolic profle is well recognized; however, the current study did not study the impact of protein supplementation on lean mass nor strength, though the baseline DXA measurement was performed. Based on the systematic review by Lim et al., there seems to be a favorable efect on the lean mass in relation to animal protein compared to plant protein, and the beneft appears more pronounced among young individuals [31]. Meta-analyses performed in the same paper indicated that the protein source did not afect changes in strength. Similar fndings were also reported by Messina et al., indicating that resistance exercise training (RET) when supplemented with whey or soy protein resulted in signifcant increases in muscle strength but found no diference between protein groups [32]. RET is a potent stimulus than protein supplementation for increasing muscle strength [33]. It will be worth exploring the impact of protein on muscle mass and strength and their association with glucose homeostasis, especially among Asian Indians. A randomized crossover clinical trial was performed among normal weight and normoglycemic subjects to assess the efect of the diferent proteins on second meal postprandial glycemia, and the data indicated that compared with control, whey and soy protein had a signifcant reduction in postprandial glycemia. However, the study did not explore the diferent protein doses or insulin responses [34]. Gunnerud et al. explored the efcacy of premeal bolus of whey and soy protein with or without added amino acids on glycemic, insulin, incretin, and amino acid response among healthy volunteers [35]. Te premeal bolus displayed a lower glycemic response than the reference meal. However, there was no diference in the insulinemic responses between the meals. Data from the present study on the other hand demonstrated a signifcant insulin response to a higher dose (whey vs. soy protein). Tis could be due to the dose of protein used between the 2 studies (9 g vs. 15 and 30 g%). Kashima et al. studied soy protein isolate preload (20 g and 40 g) on glycemic control in young healthy subjects. Te glycemic response for the soy protein isolate (40 g) was attributed to not only exaggerated insulin response but also to the noninsulin-dependent mechanism(s), i.e., gastric emptying [36]. Te data from the present study demonstrated similar responses at higher doses and, in addition, compared soy and whey protein. A study among healthy women with similar macronutrient composition containing cod, milk, or soy protein indicated serum insulin response after the milk protein meal difered from that of the cod protein meal [37]. Te insulin/glucose ratio for the cod protein meal was lower than that for the milk and soy protein meals. Te use of food as a protein source depends on the protein fraction and can be highly variable [38]. In this context, though food as a source might be more feasible in the developing world where fnancial constraints might limit the use of protein as supplementation, more studies are required to compare the diferent food-based protein intervention and their impact on postprandial glycemia. Te insulinogenic response to diferent doses of proteins is also relevant to the exogenous insulin dose, especially among individuals with type 1 diabetes [39].
Te current study explored the thermogenic efect of WP and ISP at 2 doses (15% and 30%). Tough oxygen consumption and energy expenditure increased following the protein meal, there was no diference in either dose or the type of protein. Tere were similar trends seen for CHO and fat oxidation as well. Acheson et al. demonstrated a significant thermogenic efect after a meal containing whey (50% protein) compared to casein and soy meals [11]. Te dose used by Acheson et al. was high compared to that of the current study. Te translation of such high-dose protein consumption is not feasible or recommended. Terefore, the current study explored a dose of protein that could be easily adapted in clinical practice. Te lack of diference in the thermogenic efect in the present study could be due to the fact that it was performed in a relatively small homogenous population with the normal body weight. Te impact of body composition especially body fat, including ectopic fat and muscle mass on the energy expenditure and substrate oxidation, following protein consumption needs to be further explored. Tis is of relevance as despite the normal body weight, it is proposed that Asian Indians have a greater predisposition to develop accumulation of body fat in ectopic sites [40,41]. Te protein intake could be one of the modes by which fat could be mobilized along with exercise. Te present study only focused on healthy participants, and comparative data from obese or with type 2 diabetes could have added further value.
Te current study did not demonstrate any changes in satiety. Te questionnaire-based approach to evaluate satiety might not have uncovered subtle changes between the efects of proteins on satiety. Te mechanisms by which protein may afect satiety remain elusive. Satiety involves a complicated interaction of psychological, behavioral, and physiological mechanisms [42]. It is proposed that the satiety centre could be sensitive to serum amino acid levels, and once the levels reach a certain point, hunger would cease [43,44]. However, there is little evidence to support this.
Another possible mechanism could be the relationship between satiety and incretin hormones [45]. With habitual low protein intake among Asian Indians, it will be interesting to explore the role of incretin hormonal changes and amino acid pool to understand the impact of protein on satiety among Asian Indians.

Conclusion
In conclusion, at lower doses, high-quality soy and whey did not elicit diferent insulinogenic responses, making both equally efective protein sources for the management of glucose metabolism when used in moderation. However, at higher doses, whey protein exhibited a greater insulinogenic response than soy protein. Tis provides important insight into Asian Indians who are at greater risk of developing T2D and may be seeking to increase protein intake.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
At the time of this study, Cope M and Mukherjea R (coauthors) were part of DuPont Nutrition & Biosciences, St Louis, MO, USA.