Male Wistar ST rats, aged 13 weeks or older (SLC Co. Ltd, Japan), and male C57BL/6 mice, aged 8 - 10 weeks (SLC Co. Lit), were used. Animals were purchased 1 week before the experiments. Rats and mice were fed a standard solid chow (CE-2, Japan Clea Co. Ltd, Japan) and allowed tap water ad libitum. Animals were maintained in compliance with the “Guiding Principles for the Care and Use of Animals in the field of Physiological Sciences,” published by Physiological Society of Japan. The protocol was approved by the Animal Experiment Committee of Kobe Gakuin University and Research Laboratories for Health Science and Food technologies, Kirin Company, Limited. Animals were sacrificed after the test by using Somnopentyl.

Lactobacillus fermentum strain NBRC 3961 was purchased from the National Institute of Technology and Evaluation Biological Resource Center, Japan. Lactobacillus pentosus strain JCM 8333, was obtained from the Japan Collection of Microorganisms, Japan. Pediococcus pentosaceus strain NK-2, was supplied by Biotech Japan Co. Ltd. Each of these strains is known to be found in fermented milk and vegetables. Leuconostoc oeni strain Elios 1, which is used for adding flavor to wine, was obtained from SCETI Co. Ltd, Japan, and Lactobacillus paracasei strain KW 3110, which we have previously studied for its function in immunity, was available from our own laboratory [17] .

The Lactobacillus fermentum strain NBRC 3961, Lactobacillus pentosus strain JCM 8333, and Lactobacillus paracasei strain KW 3110 were cultured at 37˚C, 30˚C, and 30˚C, respectively for 24 h in static MRS broth (Difco, BD Bioscience, USA). Cells were harvested by centrifugation and washed with phosphate buffered saline (PBS, pH 7.4). The harvested cells were suspended and adjusted to a concentration of 10¹¹ colony forming units (CFU) with PBS. The cell suspensions were stored at 4˚C until use. To prepare LAB filtrates, the harvested cells were adjusted to 10¹⁰ CFU with PBS and then incubated at 37˚C for 5 h. After incubation, cells were removed by centrifugation and the supernatant was filtered through a filter membrane with a pore size of 0.22 μm. LAB filtrates were stored at −80˚C until use.

The GTT has been described in detail elsewhere [11] -[16] . Figure 1 and Figure 2 show the principles of the test. In summary, when perfectly round ceramic ball bearings are placed in the conical tube (Figure 1), the flat segments prevent the spheres from occluding the lumen. When blood is added to this tube, it flows through the narrow gaps caused by the ball and exits in droplets into an adjacent collecting tube. The latter is trans-illuminated by a light emitter, and a sensor opposite the emitter generates a signal whenever a drop of blood interrupts the light path. In essence, the instrument detects the time interval (d, s) between consecutive blood drops. At the start, blood flow is rapid and hence d is small. Subsequently, the flow rate gradually decreases and hence d increases. When the d exceeds 15 s (occlusion-d), the instrument displays the “Occlusion Time (OT)”, which is the time elapsed from the detection of the first drop of blood until detection of occlusion-d. Later, the flow is completely arrested. Eventually, due to thrombolysis, the flow is restored as indicated by the detection of blood drop(s). There is also an arbitrarily pre-set value of d (200 s) for lysis measurement (lysis-d). When d between the last drop before and the first drop after occlusion exceeds the value of lysis-d, the instrument displays the “Lysis Time (LT)”, which is calculated as follows: LT = [time of first drop with d > (lysis-d)] − [time of last drop with d < (lysis-d)]. Blood flows at 37˚C by gravity through the narrow gaps formed between the larger ball bearing and inner wall of the tube, where initial shear rates (15,000 - 18,000 s⁻¹) activates platelets. These activated platelets remain single, because the very short transit time and high shear stress prevent aggregation. In contrast, in the space distal to the large ball bearing (i.e. the space between the two ball bearings), the shear and turbulent flow favor the formation of large platelet aggregates. Furthermore, in this space between the ball

bearings, activated platelets generate thrombin and initiate blood coagulation. Flow then carries these fibrin- stabilized platelet aggregations into the lower gaps where they are captured, resulting in occlusion of the gaps and arrest of flow. An increase or decrease in OT indicates, respectively, inhibition or enhancement of platelet reactivity. An increase or decrease in LT indicates, respectively, inhibition or enhancement of spontaneous thrombolysis. For the present study, an automated GTT instrument was purchased from Montrose Diagnostics Ltd., UK. For each sample, six parallel measurements were made. The intra-individual variations on OT and LT are 6.07% and 33.67%, respectively [15] .

Thirty minutes after anaesthetization with somnopentyl (60 mg/kg intramuscularly), blood was withdrawn from the abdominal aorta of a rat that had been fasted overnight but allowed water ad libitum. Non-anticoagulated blood was mixed with saline kept at 37˚C (1:1), and then 2.7 ml of the diluted blood and 0.3 ml of LAB filtrate or PBS (control) were mixed by inversion in a syringe, and 3.0 ml of the mixture was applied to the GTT tube.

The laser-induced thrombosis method has been previously described in detail [18] . In brief, mice were anaesthetized with somnopentyl (65 mg/kg intramuscularly). A polyethylene tube (PE-10, Becton Dickinson and Company, USA) was placed in the left femoral artery to inject the dye, and the carotid artery (350 - 450 µm in diameter) was exposed by incision. The mouse was placed on a special microscope stage (Olympus Model BH-2, Olympus Co. Ltd, Japan), and then Evans blue dye (30 mg/kg) was injected intra-arterially. Subsequent to the dye, the centre of the exposed carotid artery was irradiated with a laser (Model Neo-50MS, 25 mW power under the objective lens, Nihon Kagaku Engineering Co. Ltd., Japan). Thrombus formation at the site of irradiation was monitored under epi-illumination and simultaneously recorded on videotape using a CCD camera (Model TMC-7, Takenaka System CO. Ltd, Japan).

LAB filtrates, cell suspensions or PBS (control) were administered by a gastric tube in a volume of 7.7 ml/kg body weight [18] . They correspond to 8 × 10¹⁰ and 8 × 10¹¹ CFU/kg body weight of LAB filtrates and cell suspensions, respectively. The same volume of LAB filtrate, cell suspension or PBS was given again 30 min after the first treatment. The mouse was then anaesthetized and thrombosis experiments started 90 min after the second oral administration. The antithrombotic or prothrombotic effect was assessed by calculation of the total thrombus size. A smaller size indicated higher antithrombotic activity [19] .

Details of this technique have been described elsewhere [20] . An image of the thrombus was computer-analyzed every 10 s. The area of thrombus was outlined, and the mass of thrombus was calculated by multiplication of the area and gray-scale analysis. Image analysis was performed using Image J software (Image Processing and Analysis Java version 1.30, National Institutes of Health, MD, USA). Thrombotic status was expressed as the total sum of mass measurements in the first 10 min after irradiation.

1) Dry matter weight

The dry matter weight of LAB filtrates was measured by the freeze drying method. The NET weight was calculated by subtracting the weight of PBS used for extraction.

2) Protein assay

Protein concentration was measured using the DC protein assay kit according to the manufacturer’s instructions (Biorad, USA).

3) Saccharides

The saccharide content was measured using the phenol-sulphuric acid method with an analytical curve in glucose [21] .

4) Organic acids

Oxalic acid, malic acid, citric acid, succinic acid, pyruvic acid, acetic acid, lactic acid, and phosphoric acid were measured using the Agilent 7100 Capillary Electrophoresis system (Agilent Technologies, USA), with electrophoresis buffer containing 15 mM 2,6-pyridin dicarbonate, 2 mM adenosine 5’-monophosphate, and 0.035 mM hexadimethrine bromide, adjusted to pH 5.55 by NaOH.

Proteins contained in LAB filtrates from Lactbacillus fermentum NBRC 3961 and Lactobacillus pentosus JCM 8333 were separated in a 4% - 12% gradient SDS-PAGE gel and stained with colloidal Coomassie. The bold protein bands corresponding approximately to molecular weights of 38 kDa and 47 kDa were excised from the gel. In-gel tryptic digestion of the isolated proteins was carried out as previously described and the digested tryptic peptides were desalted using C-18 Zip-tips [22] . MALDI mass spectrometry was carried out on a Ultraflex II MALDI-TOF/TOF-MS spectrometer (Bruker Daltonics systems, USA). Peptide solution (0.7 μl) was mixed with 0.7 μl of 10 μg/μl HCCA (α-cyano-4-hydroxycinnamic acid) in 90% aqueous acetonitrile: 0.5% trifluoroacetic acid. The mixture (1 μl) was spotted onto a MALDI target plate and allowed to crystallize. Identification of proteins from MALDI-TOF/TOF-MS spectra was achieved using MS/MS Ion Search of Mascot (Matrix Science, USA) against SWISS-PROT database [23] .

GTT measurements were analyzed by repeated ANOVA, followed by post-hoc Dunnett’s test. Thrombus size was analyzed by unpaired ANOVA, followed by post-hoc Dunnett’s test. Values were expressed as the mean ± SEM. A value of P < 0.05 was considered as the limit of significance.

The effects of LAB filtrates on platelet reactivity and endogenous thrombolysis are shown in Table 1. The undiluted filtrates of Lactobacillus paracasei KW 3110 and Pediococcus pentosaceus NK-2 significantly shortened LT but had no effect on OT, suggesting weak antithrombotic activity in vivo. Lactobacillus pentosus JCM 8333 and Leuconostoc oeni Elios 1 significantly shortened LT but had no effect on OT. These results suggested, however, greater antithrombotic activity in vivo compared with Lactobacillus paracasei KW 3110 and Pediococcus pentosaceus NK-2. Lactobacillus fermentum NBRC 3961 significantly prolonged OT and shortened LT, suggesting strongest antithrombotic activity in vivo amongst the LAB strains tested. On the basis of these

Filtrates from the culture media of lactic acid bacteria were mixed with non-anticoagulated rat blood (diluted two-fold with saline), and the effect on shear-induced platelet reactivity and endogenous thrombolytic activity was assessed by GTT in vitro. The results were expressed as the mean ± SEM. The control was PBS. ^*P < 0.05; ^**P < 0.01. n = 6 in each group.

findings, three strains Lactobacillus fermentum NBRC 3961, Lactobacillus pentosus JCM 8333 and Leuconostoc oeni Elios 1 were selected for further investigation in vivo.

The antithrombotic effects in vivo of these three selected LAB strains were measured the He-Ne laser-induced thrombosis test in mice after oral administration. The results are shown in Figure 3. The filtrate of Lactobacillus fermentum NBRC 3961 showed significantly antithrombotic effects in vivo. The filtrate of Lactobacillus pentosus JCM 8333 also showed significantly antithrombotic effects in vivo. However, the filtrate of Leuconostoc oeni Elios 1 did not show a significant antithrombotic effect in these experiments in vivo.

The results are shown in Figure 4. As the filtrate of Leuconostoc oeni Elios 1 did not show antithrombotic effect in vivo, this strain and others were given at a cell concentration 10 times higher than the original cell concentration of other LAB filtrate. All cell suspensions tested showed antithrombotic effect under these conditions in vivo.

Analysis of filtrates from Lactobacillus pentosus JCM 8333 and Lactobacillus fermentum NBRC 3961 identified 80% to 90% of the ingredients in the whole dry matter weights. Protein was the main ingredient, and comprised 60% to 70% of dry matter. Organic acids (lactic acid and acetic acid) were about 5% to 10%, and saccharides were about 10% (Figure 5). SDS electrophoresis detected about 20 different proten bands, the most prominent of which were 38 and 47 kDa. These peptides were identified as GAPDH and enolase respectively, by MALDI- TOF/TOF-MS spectrometry analysis (Figure 6). The origin of the enolase in the Lactobacillus pentosus JCM 8333 filtrate was identified L. plantarum, most likely because these two species are genetically close.