More than 600 bacterial strains were isolated from the soils of Northern Taiwan using a medium that contained 1% squid pen powder (SPP) as the sole source of carbon/nitrogen. Of these, TKU042 demonstrated the strongest inhibitory activity (97%) with an IC₅₀ value of 3.9 ± 0.12 mg/mL; it was, therefore, chosen for further investigation. This potent strain was initially identified as Paenibacillus sp. based on 16S rDNA sequences. The name of the species was identified using an analytical profile index (API); however no match was found. Therefore, the TKU042 strain was simply labeled as Paenibacillus sp. Many strains of Paenibacillus have been reported as possessing potential industrial, agricultural, medical, and health food applications, such as the production of enzymes from P. curdlanolyticus B-6 [24], P. barengoltzii [25], P. barengoltzii [26], and P. ehimensis MA2012 [27], exopolysaccharides from P. mucilaginosus TKU032 [22] and P. macerans TKU029 [23], antioxidants from Paenibacillus sp. TKU036 [20], biosurfactants from P. macerans TKU029 [23], biological control agents from P. ehimensis MA2012 [27] and P. polymyxa AC-1 [28] or biofertizlizers from P. polymyxa [29,30]. Based on our recent literature review, the biosynthesis of aGIs by the genus Paenibacillus has not yet been reported.

During fermentation, the source of C/N was proposed as the significant factor in the synthesis of aGIs since it could influence the production of some aGI-related enzymes [31]. Similar phenomena were found in our previous reports, which showed SPP as the most suitable C/N source for the production of exopolysaccharides and antioxidants by isolated strains of Paenibacillus [20,21,22,23]. Three sources of C/N: 1% SPP, 1% shrimp head powder (SHP), and 0.8% nutrient broth (NB), were investigated for the production of aGIs by Paenibacillus sp. TKU042 (Figure 1). The inhibitory activities of the fermented NB, SPP, and SHP reached 100% (at day 1), 100% (at day 3), and 70% (at day 9), respectively (Figure 1A). To analyze aGI productivity, the culture supernatants were diluted appropriately to obtain aGI activity, and expressed as U/mL. As shown in Figure 1B, there were remarkable differences among the three culture supernatants. NB seemed to be the best C/N source for aGI production, showing productivity (1200 U/mL at day 4) approximately 5.3- and 10-fold higher than those of SPP (220 U/mL at day 3) and SHP (120 U/mL at day 6), respectively. Throughout fermentation, the cell growth of Paenibacillus sp. TKU 042 was also monitored by measuring the absorbance of the cell solution at 660 nm. As shown in Figure 1C, there was no relationship between cell growth and aGI productivity.

Soybean has been widely used in many studies as the sole source of C/N to synthesize aGIs [8,11,12,17,18]. To enhance aGI production, the soybean-containing medium has been supplemented with various carbohydrates. For example, Streptomyces lavendulae GC-148 produced an aGI in medium containing 2% glucose and 1.5% soybean [8], and the Bacillus species required 4% soybean peptone supplemented with 5% carbon source (glucose, galactose, lactose, or sorbitol) [11]. In this study, Paenibacillus sp. TKU 042 used NB for aGI production, diverging from the other reports.

NB was confirmed as the best source of C/N, and chosen to establish optimal parameters, including cultivation temperature, culture volume, concentration of NB, and the amount of seed culture inoculated (Figure 2).

The results showed that aGI productivity was highest on the sixth day of cultivation at 25 °C (1700 U/mL) and the fourth day of cultivation at 30 °C (1550 U/mL) (Figure 2A). Taking cultivation time into account, 30 °C was selected to study the effects of culture volume. As shown in Figure 2B, aGI productivity was highest on the sixth day of cultivation in 50 mL-medium (1800 U/mL) followed by the fourth day of cultivation in 100 mL-medium (1600 U/mL). After considering the recovered volume of culture supernatant, 100 mL-medium (at 30 °C) was selected to investigate the effects of NB concentration on aGI productivity. As seen in Figure 2C, the highest aGI productivity was found on the fourth day when either 0.4% or 0.8% NB was used. 0.8% NB was ultimately selected for further research since aGI productivity declined slightly after day 4 when 0.4% NB was used. The inoculated volume of the seed culture had no significant effect on aGI production (Figure 2D).

Overall, aGIs were effectively produced by Paenibacillus sp. TKU042 in the 0.8% NB-containing medium (100 mL of medium in a 250 mL-Erlenmeyer flask) at 30 °C and an initial pH of 6.85, using a reciprocal shaker at 150 rpm for 4 days. The culture conditions before and after the optimization study are summarized in Table 1; in short, the IC₅₀ and productivity of Paenibacillus sp. TKU 042 aGIs were increased approximately 23-fold after optimization.

To evaluate the potential of Paenibacillus sp. TKU 042 aGIs as anti-diabetic drugs, the inhibitory specificity of FNB was tested against eight kinds of commercial enzymes. As shown in Table 2, FNB showed strong inhibitory activity against bacterial (Bacillus stearothermophilus) α-glucosidase (IC₅₀ = 36.7 ± 1.7 µg/mL), yeast (Saccharomyces cerevisiae) α-glucosidase (IC₅₀ = 81 ± 4.3 µg/mL), and rat α-glucosidase (IC₅₀ = 101 ± 5.1 µg/mL), but had a weaker response against rice α-glucosidase (IC₅₀ = 508 ± 17 µg/mL). On the other hand, the anti-diabetic drug acarbose demonstrated better inhibitory activity against α-glucosidases from bacteria (0.012 ± 0.001 µg/mL), rice (2.92 ± 0.67 µg/mL), and rat (107 ± 2.5 µg/mL), but showed weaker inhibitory activity against yeast α-glucosidase (1395 ± 5 µg/mL).

α-Glucosidase from S. cerevisiae has been commonly used to evaluate aGI activity in many studies [18,32,33,34]. Compared to yeast and other sources of α-glucosidases, rats may be a more valuable resource since the animal enzyme is closer to that of humans [32]. In this study, both FNB and acarbose inhibited α-glucosidases originating from rats and yeast. Among them, FNB showed much stronger aGI activity than acarbose against α-glucosidase from yeast but approximately the same amount of inhibition against rat α-glucosidase. Therefore, FNB could be a potential candidate for α-glucosidase inhibition.

α-Amylase inhibitors have been reported to be a useful treatment for type 2 diabetes [35]. Consequently, the inhibitory activities of FNB against α-amylases were also studied. As shown in Table 2, FNB showed no inhibitory activity against porcine pancreatic and B. subtilis α-amylases. Similarly, FNB demonstrated no inhibitory activity against proteases from bromelain or papain. It is well-known that there are several proteases in the gastrointestinal tract to help ingest proteins. Since FNB showed no inhibitory activity against these tested proteases, it may have the potential to avoid disturbing the ingestion of proteins by the use of aGIs.

Several previous studies [20,22] suggested that antioxidant compounds could be produced by the genus Paenibacillus. Therefore, the antioxidant property of FNB was also tested, using the 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity assay mentioned in the experimental methods section, to provide more bioactive proof of FNB’s usefulness. The antioxidant activity was calculated as both a percentage and IC₅₀ value. α-Tocopherol, a commercial antioxidant compound, showed stronger activity than FNB, generating IC₅₀ values of 0.0247 ± 0.0012 and 2.23 ± 0.14 mg/mL, respectively. However, FNB can achieve the same amount of activity (approximately 100%) as α-tocopherol. Overall, FNB could be used as a health food due to its valuable bio-functions, including its potential for α-glucosidase inhibition and acceptable antioxidant activity.

The same concentrations (20 mg/mL) of unfermented NB (UNB) and fermented NB (FNB) solutions were analyzed by HPLC. The differences in the HPLC finger prints of NB before and after fermentation are clearly observed in Figure 3. After fermentation, some major peaks disappeared (the peaks at retention times of 6, 13.5, and 17.5 min) or reduced their area (the peak at 11 min), while some new peaks appeared at 6.5 and 13.2 min, and the peak at 17.3 min, enhanced its area. UNB was tested for α-glucosidase inhibition but no activity was observed. The differences in both the HPLC finger prints and the inhibitory activity of UNB and FNB suggest that the aGIs were produced by fermentation, since none existed in the NB beforehand. Each fraction of aGI activity in FNB was analyzed on the HPLC finger print; three fractions showed aGI activity (≥85% at the concentration of 200 µg/mL). The isolation and identification of these aGIs will eventually be conducted. The HPLC finger print of acarbose was also analyzed; based on the difference in retention time, we confirmed that the active compound in FNB aGIs was not acarbose.

The thermal stability of FNB aGIs was determined by first treating FNB at 40–100 °C for 30 min, and then testing the residual aGI activity. FNB aGIs were strongly thermostable at all temperatures, with relative activities greater than 95% (Figure 4A). They maintained a relative activity of 92% even when heated at 100 °C for 30 min. The high thermal stability of FNB aGIs suggests they may have the potential to remain active for a long time.

pH stability is also an important factor. After treating FNB with a large range of pH values at 37 °C for 30 min, the residual aGI activity was analyzed at pH 6.8. As shown in Figure 4B, aGI activity remained at 87%–96% after 30 min pre-incubation at pH ranges from 1–13. It is well-known that the stomach is very acidic; therefore, potential aGIs should be stable and able to work well in the acidic environments of the gastrointestinal tract. Bacteria-produced aGIs that show thermal and pH stability are rarely reported. In the previous reports [32,35], aGIs from ELC, a methanolic extract from a medicinal plant (Euonymus laxiflorus Champ), were reported to be heat-stable at 97 ± 3 °C (heating 30 min) with a relative activity of 90%, demonstrating the same thermal stability as FNB. However, the relative activity of this extract (50%) was lower than that of FNB (80%–93%) in the acidic environment (pH 4–6).

Three doses of FNB (100, 200, and 400 mg/kg) were used on ICR mice to evaluate their impact on plasma glucose. As shown in Figure 5A, significant reductions in plasma glucose were observed 0.5 and 1 h after sucrose loading for 100 and 400 mg/kg.

The FNB with 200 mg/kg body weight (bw) was ultimately the best, as it reduced blood sugar for 1.5 h. This result is comparable with a dose of acarbose at 25 mg/kg, but better than acarbose at 12.5 mg/kg (Figure 5B).

In the combined FNB and acarbose group, two doses of FNB (100 and 200 mg/kg) and one dose of acarbose (12.5 mg/kg) were used (Figure 5C). The combination of FNB at 100 mg and acarbose at 12.5 mg had no effect compared to acarbose (12.5 mg/kg) administration alone. On the other hand, the combination of FNB at 200 mg/kg and acarbose at 12.5 mg/kg was much better at reducing blood glucose than acarbose alone at 12.5 mg/kg, slightly higher than acarbose at 25 mg/kg, and comparable to arcabose at 50 mg/kg. The number of mice with diarrhea was also recorded. No ill effects were observed in the control, FNB at 100, 200, or 400 mg, acarbose at 12.5 mg or the combination of FNB (100, 200 mg) with acarbose at 12.5 mg. However, for the groups who received 25 and 50 mg of acarbose, diarrhea was observed in 3/8 mice (37.5%) and 5/8 mice (62.5%), respectively.

These results show that oral FNB loading can reduce plasma glucose levels in mice, and the combination of FNB (200 mg/kg) and acarbose (12.5 mg/kg) can produce approximately the same effect as acarbose alone at 25 and 50 mg. As such, the acarbose dosage can be reduced two- to four-fold, resulting in reduced side effects.

Nutrient broth was purchased from Creative Life Science Co., Taipei, Taiwan, squid pens were acquired from Shin-Ma Frozen Food Co. (Yilan, Taiwan), and shrimp head power (SHP) was obtained from Fwu-Sow Industry (Taichun, Taiwan). Rat α-glucosidase (intestinal acetone powders from rat) was purchased from Sigma Aldrich, Singapore. Acarbose, Saccharomyces cerevisiae α-glucosidase, Bacillus stearothermophilus α-glucosidase, and 2, 2-diphenyl-1-picrylhydrazyl (DPPH) were purchased from Sigma Chemical Co., St. Louis, MO, USA. Rice α-glucosidase (Type 4) and porcine pancreatic α-amylase (Type VI-B) were purchased from Sigma Aldrich, St. Louis, MO, USA. The proteases of bromelain and papain were obtained from Challenge Bioproducts Co., Ltd., Yunlin, Taiwan. When possible, reagents, solvents and other common chemicals were obtained at the highest grade.

To determine rat intestinal α-glucosidase inhibition, the techniques from the previous report were followed [36], with modifications. One hundred mg of rat-intestinal acetone powder was suspended in 4 mL of 0.1 mol/L potassium phosphate buffer (NPB), pH 7. The suspension was sonicated several times for 12 s at 4 °C and then centrifuged for 20 min at 10,000× g. The residues were twice suspended with 3 mL of 0.1 mol/L NPB at pH 7, as described above. The resulting supernatant was mixed together and dialyzed for 12 h at 4–6 °C, and then used for the assay. The α-glucosidase, sample solutions and buffer were pre-mixed at volumes of 50, 50 and 100 μL, respectively. After pre-incubation at 37 °C for 20 min, the reaction started when 50 μL of p-nitrophenyl glucopyranoside (10 mmol/L) was added to the mixture. After 20 min at 37 °C, 100 μL Na₂CO₃ solution (1 mol/L) was added to terminate the reaction. The absorbance of the final mixture was measured at 410 nm [37] and used to calculate inhibition (%) using the following equation: α - glucosidase   inhibitory   activity   ( % )   =   ( A   −   B ) / A   ×   100 where “A” is the absorbance of the control (buffer was used instead of sample with the same volume) and “B” is the absorbance of the final mixture containing the sample and α-glucosidase [38]. The IC₅₀ value was defined and determined as per the previous study [32]. The α-glucosidase and the samples were prepared in 0.1 mol/L potassium phosphate buffer (pH 7).

General α-glucosidase and α-amylase inhibition assays were carried out as described in our previous papers ([32,35], respectively). Acarbose dissolved in the same buffer (pH 7) was also analyzed as a control.

One hundred and twenty microliters of each sample (at different concentrations) was mixed with 30 µL of a methanolic solution containing 0.75 mM DPPH radicals in the well of a plate with 96 wells. The mixture was kept in the dark for 30 min before absorbance at 517 nm was measured [39]. Radical scavenging activity was calculated using the following formula: Antioxidant   activity   ( % )   =   [ ( A B   −   A A ) / A B ]   ×   100 where AB and AA stand for absorption of the blank sample (t = 0 min) and absorption of the tested sample solution (t = 30 min), respectively. α-Tocopherol dissolved in MeOH was also analyzed as a control.

About 60 soil samples were randomly collected from Northern Taiwan. For convenience, half of the soil samples were collected from the campus of Tamkang University (Tamsui, New Taipei City, Taiwan). Ten milliliters of sterile distilled water was added to 1 g of ground soil and stirred. Half a milliliter of the soil suspension was then added to medium containing 0.1% K₂HPO₄, 0.05% MgSO₄·7H₂O, and 1% squid pen powder (SPP), supplemented with 1.5% agar in a plastic Petri dish containing 10 mL of medium (pH 7.2). After 1–3 days of incubation at 37 °C, the single bacterial colonies that appeared were sub-cultured several times on the same medium.

All of the isolated strains were incubated in a 250-mL Erlenmeyer flask with 50 mL of the same medium, but without agar. Fermentation was performed at 37 °C and a 150 rpm shaking speed. After 1–4 days of fermentation, the culture was centrifuged at 500 rpm for 10 min to remove the medium residue. The solution was then centrifuged again at 4000 rpm for 20 min to separate the cells from the culture supernatant. The supernatant was then used to test yeast α-glucosidase inhibitory activity. The cells were dissolved in the same volume of distilled water as the supernatant and measured at 660 nm to detect bacterial growth. The screening of active bacteria was based on the inhibitory activity of the supernatant.

Three sources of C/N (1% SPP, 1% SHP, and 0.8% NB) were used to cultivate Paenibacillus sp. TKU042; conditions were set at 30 °C, a shaking speed of 150 rpm and a 100/250 mL medium/flask volume ratio for nine days. NB appeared to be most suitable for synthesizing aGIs and was chosen to test optimization of specific parameters, including concentration of NB (0.4%, 0.8%, 1.2%, 1.6%, 2.0% w/v), medium volume (50, 100, 150, and 200 mL), culture temperature (25, 30, and 37 °C) and the amount of seed culture (0.5, 1, 1.5, 2, 3, and 5 mL of bacterial seed solution at OD_660nm = 0.38). The culture supernatant obtained was centrifuged again at 4000 rpm for 20 min, then used to test yeast α-glucosidase inhibitory activity.

The pH stability of the sample was determined as per the methods described in our previous paper [35]. To determine thermal stability, the sample was exposed to a temperature range of 40–100 °C for 30 min before testing for inhibition against yeast α-glucosidase, using the general α-glucosidase inhibition assay [32].

Seventy-two seven week-old male ICR mice obtained from The National Laboratory Animal Center (No. 128, Sec. 2, Academia Rd., Nangang Dist., Taipei City 11529, Taiwan) were randomly divided into nine groups (8 mice per group). All had equivalent mean plasma glucose levels and body weights, including the control group (orally administered with distilled water only). Three experimental groups received oral FNB (100, 200, or 400/kg), three experimental groups received oral acarbose (12.5, 25 or 50 mg/kg), and two experimental groups received both FNB and acarbose (100 mg of FNB combined with 12.5 mg of acarbose, or 200 mg of FNB combined with 12.5 mg of acarbose). This animal study was conducted in accordance with the guidelines and approval of the Institutional Animal Care and Use Committee of the National Research Institute of Chinese Medicine, Ministry of Health and Welfare. (IACUC No.104-706-1, 29 December 2014).

The sucrose tolerance test was performed following the methods described in the previous report [40], with slight modifications. Mice were fasted for 16 h and were then given different concentrations of distilled water (control group), acarbose, or mixtures of the solutions described in detail above for 20 min. A sucrose solution was orally administered at 3 g/kg bw and blood was sampled and measured after 0.5, 1, 1.5 and 2 h.

Statistical Analysis Software (SAS-9.4, provided by SAS Institute Taiwan Ltd, Minsheng East Road, section 2, Taipei, Taiwan 149-8) was used to analyze the significant differences of the calculated IC₅₀ values, the α-glucosidase inhibition (%) and the plasma glucose levels.