The HBN was prepared and kindly provided by Professor Lim Yang Mooi from University Tunku Abdul Rahman, Malaysia. A voucher specimen was deposited in the Nature Inspired UM Natural Products Library, University Malaya (voucher number UMCNA1801). Briefly, the raw edible birds nest from Aerodramus fuciphagus (white nest) from swiftlet houses was cleaned and made to powder. The raw cleaned EBN powder was then suspended in distilled deionized water with the ratio of 0.2% (w/v) for 24 h at 4°C. The mixture was then heated at 80°C in the distilled water for an hour. The extracts were allowed to cool and further centrifuged at 2,700 g to collect the supernatant for subsequent lyophilization. The dried extract was stored at −20°C until further analysis. Chemical profiling of HBN was carried out using LCMS-QTOF. Solvents system in LCMS consist of 0.1% formic acid in water (A) and acetonitrile (solvent B) were used with the following gradient: starting with 100% B and reduction to 50% B at 18 min, and finally 5% B from 18 to 30 min. The system controller was stopped at 20 min. The solvent’s flow rate was 0.8 ml/min. Samples (10 μl) were injected in to a C18 reversed-phase column (150 × 4.6 mm i.d, 3.0 µm particle size). Mass spectrometric detection was performed with a quadrupole-TOF-MS operated in the positive mode. Information dependent acquisition was conducted using a TOF-MS survey scan 100–1,100 Da (100 ms) and up to 10 dependent TOF MS scans 100–1,100 Da (100 ms) with Collision Energy (CE) of 45 V with Collision Energy Spread (CES) of ± 30 V. The identification of the peaks was conducted with Metlin database.

Healthy db/db and C57BL/6J male mice (8 weeks old) were purchased from Jackson Laboratory and Monash University (Sunway Campus, Malaysia), respectively. The mice were allowed 2 weeks for acclimatization to the housing facility. All the experimental procedures were approved by the University of Malaya Animal Care and Ethics Committee (Ethics Reference No: 2015–180709) and accredited by Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). Animal study was carried out in strict accordance with the established institutional guidelines and the NIH guidelines on the use of experimental animals. The animals were housed in individual ventilated cages with temperature (24 ± 1°C), lighting condition (12 h light/dark cycles), room pressure (10 Pa) and humidity (50–70%) with free access to standard chow (Specialty Feeds Pty Ltd., Glen Forrest, Australia) and tap water ad libitum.

Animals were divided into five groups with six mice in each group as follows:

Group 1: C57BL/6J mice-nondiabetic control mice (received distilled water) via oral gavage for 28 days

HBN and anti-diabetic agent, glibenclamide (positive control) was dissolved in distilled water. Body weight was measured weekly by using a digital weight scale. Fasting blood glucose (FBG) with blood pricked from tail vein was measured weekly by using a digital glucometer (Accu-Check, Roche Diagnostics). At the end of 28 days, the animals were sacrificed by carbon dioxide (CO₂) inhalation and blood were collected immediately by cardiac puncture. Liver, pancreas and skeletal muscle were collected for investigating the morphological changes and immunohistochemistry. Some part of the liver and epididymal fat were collected for Western blot.

Oral Glucose Tolerance Test was performed on overnight fasted animals 2 days before sacrificing the animals at the end of the HBN treatment period. Briefly, glucose load of 3 g/kg of glucose was given orally and blood glucose levels were checked by tail pricking at 0 min (prior to glucose load), 30, 60, 90, and 120 min after loading of glucose (Kunasegaran et al., 2017).

Blood samples were allowed to clot for 30 min at room temperature, centrifuged at 2,000 rpm for 10 min at 4°C to obtain serum and stored at −80°C immediately until further use. Serum insulin levels were determined using enzyme-linked immunosorbent assay (ELISA) kit (Mercodia Mice Insulin ELISA Kit’s instructions’ (i-DNA Biotechnology) according to the manufacturer’s protocol. In brief, insulin in the sample reacts with peroxidase-conjugated anti-insulin antibodies in the microtiter wells. The samples were washed to remove the unbound enzyme-labeled antibody. The bound conjugate was detected by reaction with 3, 30, 5, 50-tetramethylbenzidine, then stopped with acid which gave a colorimetric endpoint. Optical density was measured by using a micro-plate reader (Tecan, Mannedorf, Switzerland) at wavelength of 450 nm.

The volume/unit of IL6 and TNF-α protein complex were measured in the blood serum by using ELISA kits (Biosource International Inc., Camarillo, CA) following manufacturer’s guidelines. IL-6 and TNF-α were determined from a standard curve and their levels were expressed in pg/ml.

Liver, pancreas and skeletal muscle were harvested and immediately fixed in 10% buffered formalin overnight, then embedded in paraffin and manually cut into 5 μm thick sections by using a microtome (Histo-line laboratories, ARM-3600, Viabrembo, Milan, Italy). Sections were then dewaxed in two changes of xylene, hydrated in two changes of 100% of ethanol, followed by 95 and 80% of ethanol and finally rinsed with H₂O. Sections were then stained with hematoxylin and eosin (H and E). The stained sections were dehydrated with 80% ethanol followed by 95% ethanol, placed in two changes of 100% of ethanol and cleansed with two changes of xylene. Histopathological changes were viewed by using a phase contrast microscope (Nikon H600L, Nikon DS camera control Unit DS-U2, Version 4.4, Tokyo, Japan), with an attached photograph machine (Nikon H600L).

For immunochemistry, the sections were incubated with 0.01 M citrate buffer, pH 6.0 for 10 min at 100°C and then 3% H₂O₂ in phosphate buffered saline (PBS) to neutralize the endogenous peroxidase. Blocking for non-specific binding was done with normal serum (Santa Cruz Biotechnology, Santa Cruz, CA, United States). Following that, sections were incubating with primary polyclonal antibodies for insulin (sc-9168), GLUT-4 (sc-53566) and GLUT-2 (sc-9117) from Santa Cruz Biotechnology, at a dilution of 1:500 in 5% normal serum for 1 h at room temperature. The sections were then rinsed three times with PBS before incubation with biotinylated secondary antibody for 30 min at room temperature, exposed to avidin and biotinylated HRP complex in PBS for another 30 min (ImmunoCruzTM ABC Staining System). 3,30-Diaminobenzidine (DAB) were used to visualize antibody binding site that produced dark-brown precipitate. Sections were counterstained with hematoxylin-eosin (H and E) for nuclear staining.

Protein samples obtained from the liver and epididymal fat were lyzed in ice-cold 1X RIPA buffer consists of EGTA 1 mM, EDTA 1 mM, NaF 1 mM, leupeptin 1 μg/ml, aprotinin 5 μg/ml, PMSF 100 μg/ml, sodium orthovanadate 1 mM, and β-glycerolphosphate 2 mg/ml. The lysates were centrifuged, and the supernatant was used for Western blotting. Protein concentrations were quantified using standard Lowry assay protocol by (Bio-Rad Laboratories, Hercules, CA, United States). Twenty micrograms of protein samples were electrophoresed at 100 V through 7.5 or 10% SDS-polyacrylamide gels based on the size of target proteins and transferred to an Immobilon-P polyvinylidene difluoride membrane (Millipore, Billerica, MA, United States). The membranes were incubated with 3% bovine serum albumin (BSA) in 0.05% Tween 20 PBS with gentle shaking to block non-specific binding. The membranes were then incubated with primary antibodies against insulin receptor β (1:1,000, Cell Signaling), p-IRS1 (1:1,000, Cell Signaling), IRS1 (1:1,000, Cell Signaling), p-PI3K (1:1,000, Cell Signaling), p-AKT (1:1,000, Cell Signaling), AKT (1:1,000, Cell Signaling), NFκB (1:1,000, Cell Signaling), NOX4 (1:1,000, Cell Signaling), SOD-1 (1:1,000, Cell Signaling) and GAPDH (1:10,000, Abcam) overnight at 4°C. Following that, the membranes were washed three times with TBS-T and incubated with horseradish peroxidase-conjugated secondary antibodies (DakoCytomation, Carpinteria, CA, United States) at a dilution of 1:10,000 for 1 h. Then, enhanced chemiluminescence detection system (ECL reagents, Millipore Corporation, Billerica, MA) was added onto the membrane exposed on X-ray films and developed by SRX-101 (Konica, Wayne, NJ) for visualization. Quantity One software (Bio-Rad) was used for densitometry analysis. The respective protein expression levels were calculated as ratio of each target band/appropriate internal control and expressed over normal control.

All results are presented as mean ± standard error of mean (SEM) for the number of rats (n) in each group. Data were analyzed for statistical significance using Student’s t-test for unpaired observations and, for comparison of more than two groups, one-way ANOVA followed by Bonferroni’s multiple comparison test was performed using the statistical software GraphPad Prism version 4 (GraphPad Software Inc., San Diego, CA, United States); p value of less than 0.05 was considered to indicate statistically significant difference.

Figure 1 shows a LCMS- QTOF chromatogram of HBN. A total of 20 metabolites were identified. Based on the chemical MS library and comparison to the literatures, 17 metabolites (Table 1) were tentatively identified including the bio-marker compounds, sialic acid. Three types of sialic acid were determined in HBN. They are N-Acetylmuramic acid, N-Acetylneuraminic acid and N-Acetyl-2,7-anhydro-alpha-neuraminic acid. The HBN used for this study contained 1.26 µg sialic acid/mg of dried weight (the major active compound in HBN) and it was standardized to contain the same amount of sialic acid for every batch prepared. Other metabolites found were vitamin D and 3-Phenyl-5-ureido-1,2,4-triazol. As EBN contain high amount of protein, a few amino acids and peptides such as threonine, proline, leucine and valine were also detected in the sample. Secondary metabolites such as octylamine, cascarillin, roscovitine, terbutaline, lupinine and N-(octadecanoyl)-1-beta-glucosyl-sphinganine were also found in HBN sample.

The body weight of db/db mice were significantly higher than the non-diabetic mice at day 0 (46.8 ± 0.75 vs. 27.8 ± 0.65 g) and remained high until end of treatment at day 28 (39.00 ± 2.19 vs. 28.7 ± 0.61 g). There were no significant differences in the body weights for all db/db groups at day 0. The body weight of all db/db mice decreased slightly from day 14 onwards. There was no significant difference in body weight between vehicle, HBN or glibencamide treated db/db mice at day 28 (Figure 2A).

The diabetic groups demonstrated a significantly elevated fasting blood glucose (>20 mM) compared to the non-diabetic control (<10 mM) on day 0 and remained high throughout the 28 days. The db/db mice treated with HBN 150 mg/kg demonstrated a significant decrease from day 14th while HBN 75 mg/kg and glibenclamide treated groups started showing a significant decrease from day 21. There was about 20% reduction in blood glucose level at the end of treatment at day 28 for all treated groups compared to db/db mice (Figure 2B).

The diabetic groups demonstrated a significantly increased in glucose tolerance compared to the non-diabetic control. Treatment with HBN (75 and 150 mg/kg) and glibenclamide decreased significantly the glucose tolerance by about 7–10% in db/db mice (Figures 2C,D). The HBN (150 mg/kg) and glibenclamide treatment showed significant decrease in glucose tolerance starting 60 min after glucose administration and continue to decrease at 90 and 120 min while the treatment with HBN (75 mg/kg) showed significant decrease in glucose tolerance starting 90 min after glucose administration till the end point at 120 min.

Insulin expression was significantly reduced in pancreatic islets of db/db mice compared to control group. In diabetic rats treated with HBN (75 and 150 mg/kg), increased insulin expression was observed. Higher insulin expression was observed with 150 mg/kg HBN treatment. Treatment with glibenclamide also significantly increased pancreatic insulin content (Figure 3A). Serum insulin were markedly elevated in db/db mice indicating hyperinsulinemia. Supplementation of HBN (75 and 150 mg/kg) and glibenclamide (1 mg/kg) in the db/db mice significantly decreased the elevated serum insulin (Figure 3B).

Pro-inflammatory cytokines, IL-6 and TNF-α were increased by two folds in the serum of db/db mice. Treatment with HBN (150 mg/kg) and glibenclamide significantly reduced the production of pro-inflammatory cytokines of IL-6 and TNF-α. Although not significant, treatment with 75 mg/kg HBN also demonstrated a decrease in serum IL-6 and TNF-α levels in db/db mice (Figures 4A,B).

Islet of Langerhans appeared small and degranulated in the diabetic group compared to control group. The acinar cells were swollen, and small vacuoles were observed in almost all acinar cells. Interlobular ducts were lined with flattened epithelium. However, in diabetic rats treated with HBN extracts or glibenclamide, the islets were bigger and less necrotic (Supplementary Figure S1A). The liver from normal control mice showed normal histological structure, regular distinct hepatocytes with sinusoidal spaces arranged radially around the central vein. The diabetic group showed some fatty changes with necrosis and necrobiosis in the hepatocytes. However, diabetic mice treated with HBN (75 and 150 mg/kg) and glibenclamide showed decreased fatty changes with lower necrosis (Supplementary Figure S1B).

In liver (Figure 5) and adipose tissue (Figure 6), the protein expression of insulin receptors (IRβ) and the downstream proteins of insulin signaling, p-IRS1, PI3K and p-Akt was downregulated in the db/db mice compared to non-diabetic mice. Treatment of HBN (150 mg/kg) and glibenclamide in the diabetic mice upregulated the IRβ, p-IRS1, PI3K and p-Akt in both liver and adipose. Treatment with HBN (75 mg/kg) in liver and adipose tissue significantly upregulated PI3K and p-Akt. Although this dose demonstrated a slight upregulation of IRβ and p-IRS1, it was not significant.

In contrast, the inflammatory protein NFkB were upregulated by 1-fold in both the liver (Figure 5) and adipose tissue (Figure 6) of the db/db group. The treatment with HBN (75 and 150 mg/kg) and glibenclamide reversed the upregulation of NFkB. Similarly, the reactive oxygen species (ROS) marker, NOX4 protein was upregulated in the liver and adipose tissue of the diabetic group and this upregulation was reversed following treatment with HBN (75 and 150 mg/kg) and glibenclamide. In contrast, the antioxidant protein SOD-1 was downregulated in both liver (Figure 5) and adipose tissue (Figure 6) of the diabetic group. Treatment with HBN (75 and 150 mg/kg) and glibenclamide in the diabetic animals upregulated the decreased protein to control level.

GLUT-2 expression was highest in the pancreatic islets (Figure 7A) and liver (Figure 7B) of non-diabetic group and significantly lower in the pancreatic islets and liver of the diabetic group. Treatment of diabetic mice with HBN (75 and 150 mg/kg) and glibenclamide resulted in higher GLUT-2 expression in the pancreatic islets and liver compared to untreated diabetic group.

Immunostaining showed lower levels of GLUT-4 protein expression in the liver (Figure 8A) and skeletal muscle (Figure 8B) of db/db group compared to the C57BL/6J group. Treatment of diabetic mice with HBN (75 and 150 mg/kg) and glibenclamide showed significantly higher GLUT-4 expression in the liver and skeletal muscle.