Twenty-three Lactiplantibacillus plantarum strains, including L. plantarum CCFM571, CCFM595, CCFM8610, CCFM438, CCFM726, CCFM408, CCFM8661, CCFM634, CCFM175, CCFM242, CCFM361, CCFM259, CCFM382, Lp45, FJSWX14-5, DYNDL58M4, PS3-9, FFJND7-L5, HY9-10, M2-05-R02, FJSZJ4-L5, 4L-4 and VNMWLT1M12, were obtained from Research Center of Food Biotechnology in Jiangnan University (Wuxi, China). The cultivation of all L. plantarum strains was in MRS broth (Hopebio, Qingdao, China) at 37°C for 18h.

The B[α]P-binding ability of the 23 L. plantarum strains was estimated as previously described (21). The cultured biomass was centrifuged at 5,000 × g for 15 min and washed twice with ultrapure water to acquire cell pellets. The cell pellets were re-suspended in the ultrapure water containing 10µg/mL B[α]P (Sigma-Aldrich, St Louis, MO, USA). The suspension was centrifuged after incubation for 2h at 150 rpm and 37°C, and the residual B[α]P concentration in the supernatant was analyzed. B[α]P levels were measured using a high performance liquid chromatography (HPLC) equipped with a Waters Atlantis C18 reverse-phase column (4.6×250 mm×5 µm, 30°C; Waters Corporation, Milford, MA, USA). The mobile phase was acetonitrile:water (88:12). The injection volume was 20μL, with a flow rate of 1mL/min, and the fluorescence detection wavelength was 406 nm.

The B[α]P-binding abilities of the Lactiplantibacillus strains are expressed as the B[α]P removal rate, which was calculated as follows:

Removal rate (%)=[(C_i–C_r)/C_i]×100%, C_i and C_r are the initial and residual B[α]P level, respectively.

Male adult BALB/c mice (8-week-old) were purchased from Slack limited company (Shanghai, China). Mice were kept in cages at a constant temperature (22°C ± 1°C) and humidity (55% ± 10%) under a 12-h/12-h light/dark cycle and had free access to food and water. All procedures and protocols of mice experiments were performed according to the guidelines of the Animal Care and Use Committee and the Ethics Committee of Jiangnan University (JN.No20190915b0481210).

The mice were divided into four groups and allowed to acclimatize to their environment for 1 week. The experimental schedule is shown in Table 1 . Group 1 (control group) was administered skim milk and corn oil without B[α]P. Group 2 (model group) was administered B[α]P dissolved in corn oil at a dose of 50 mg/kg b.w. Groups 3 and 4 (CCFM8661 and CCFM382 groups, respectively) were administered 2×10⁹ CFU of L. plantarum strains CCFM8661 and CCFM382, respectively, and 50 mg/kg b.w. of B[α]P. All treatments were administered via oral gavage for 5 weeks. Mice were fasted for 12h before sacrifice. Blood samples were collected and centrifuged at 3,000 × g for 15 min to obtain the serum, which was used for biochemical analysis. Colon and brain tissues were immediately washed with 0.9% saline and separated into two parts: one part was stored at -80°C for subsequent measurements, and the other part was fixed with 4% formalin for histopathological analysis.

The test was performed as previously described with minor modifications (22, 23). The size of apparatus used for the open-field test was 50cm×50cm square with four white walls. The edge region is a 15cm area near the walls and the rest of the field was the central area. Each mouse can move freely within the apparatus for 15min. A camera was used to record their movements, and the data were analyzed using EthoVision (Noldus, Wageningen, Netherlands). The apparatus was cleaned with 75% ethanol after each test to eliminate any possible odor cues. The total distance traveled and the times spent in the center and the edge regions were calculated to measure anxiety-like behavior.

The mouse feces were mixed with acetonitrile (1:2) by vortex oscillation. After centrifugation at 10000 rpm for 10 min, the 3-OH B[α]P levels in supernatant was analyzed by HPLC-fluorescence detection (24). The mobile phase was methanol:water (97:3, pH4.5). The injection volume was 20 μL, with a flow rate of 0.5 mL/min, and the fluorescence excitation and emission wavelength were 365 and 450 nm, respectively.

The MDA level and SOD activity were measured using ELISA kit according to the operating instructions of the manufacturer (Jiancheng Bioengineering, Nanjing, China).

Colon and brain tissue (0.1g) samples were lysed in TRIzol reagent (Ambion, USA) for RNA extraction. cDNA was synthesized using the RevertAid (Thermo Fisher Scientific, Waltham, MA, USA). Gene expression levels were determined using validated primers for Gapdh, Zo-1, Occludin, Claudin-1, CYP1A1, Bax, Bcl-2, and p53 ( Table 2 ) and iTaq Univeral (Bio-Rad, Hercules, CA, USA) on an RT-qPCR system (BioRad-CFX384) (25). The PCR program comprised initial denaturation at 95°C for 2 min, followed by 40 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s, and finally, 72°C for 5 min. Relative quantification of these target gene expression levels was performed after normalization to Gapdh gene expression levels using the 2^−ΔΔCt method.

Colon tissue was fixed in 10% formalin saline for 24 h and then embedded in paraffin. The paraffin was sliced into 5-μm-thick sections. After sectioning, the tissue samples were stained with H&E (26).

Total DNA in feces was extracted using the FastDNA Spin Kit (MP Biomedicals, Santa Ana, CA, USA). The V3-V4 region of the 16S rRNA gene was amplified using 341F/806R primers. The library was built and sequenced on an Illumina MiSeq PE300 platform.

Fecal samples (50mg) were dispersed in 50μL of saturated NaCl solution, acidified with 5% (v/v) H₂SO₄, and SCFAs were extracted with 1mL of diethyl ether. SCFA levels were tested by GC-MS (27).

Statistical analyses were performed using Prism version 7 (GraphPad, San Diego, CA, USA). Significant differences were evaluated using a one-way analysis of variance. Microbiota-related analyses, including alpha diversity and biodiversity richness, were assessed with the QIIME (version 1.17) and R (version 3.5.0) software.

The B[a]P-binding abilities of the 23 L. plantarum strains are presented in Figure 1 . The B[a]P-binding abilities were significantly different among the different L. plantarum strains. L. plantarum CCFM8661 had the highest B[a]P removal capacity, with a removal rate of 60.9%; thus, this strain was selected as the target strain for subsequent animal experiments. CCFM382 had the lowest removal rate of only 1.6%; thus, it was selected as the negative reference strain. The protection of these two L. plantarum strains against B[α]P-induced toxicity in the gut and the brain were compared.

The levels of 3-OH B[α]P in the feces of mice significantly increased after B[a]P treatment, which was up to 0.27 μg/g ( Figure 2 ; P<0.05). The effects were significantly reversed by oral administration of L. plantarum CCFM8661 (P<0.05), but not L. plantarum CCFM382 (P>0.05). The 3-OH B[α]P level in the CCFM8661 and CCFM382 groups were 0.14μg/g, and 0.21μg/g, respectively.

The open-field test was used to evaluate the spatial cognitive ability of the experimental animals. The distance moved was used to represent autonomous activity ability, and the time spent in the central area was used to reflect their spatial cognition ability in a new environment. Mice with poor cognitive ability would quickly leave the central area and move along the periphery, thereby spending less exploration time in the central area. As shown in Figure 3 , B[α]P-treated mice traveled a shorter total distance and spent less time in the center than mice in the control group (P<0.05). However, mice in the CCFM8661 and CCFM382 group traveled a longer distance and spent more time in the central zone than those in the model group (P<0.05). Importantly, L. plantarum CCFM8661 had a more significant increase on these two parameters than CCFM382. These results suggested that anxiety-like behavior caused by B[α]P can be better reversed by CCFM8661 supplementation.

Malondialdehyde (MDA) levels were dramatically higher in the model group than those in the control group ( Figure 4B , P<0.05). Of the two L. plantarum intervention groups, only L. plantarum CCFM8661 sharply decreased the MDA levels (P<0.05). Superoxide dismutase activity was not dramatically different between the groups ( Figure 4A ).

The expression levels of Bcl-2 and p53 in the brain increased dramatically in the model group, but L. plantarum CCFM8661 dramatically reduced the expression levels of these genes ( Figures 4D , E ; P<0.05). Although the decrease expression of p53 in CCFM382 group was also observed, the difference was not significant (P>0.05). In addition, the Bax expression level in the brain tissue of the control, model, and two L. plantarum-intervention groups were no significant differences ( Figure 4C ).

The mRNA expression levels of Zo-1 and occludin in the colon were dramatically lower in B[α]P-exposed mice than in control mice ( Figure 5A ; P<0.05). Oral administration of CCFM8661 and CCFM382 significantly increased occludin expression (P<0.05). The mRNA expression levels of claudin-1 were increased in the model group (P<0.05) but were not significantly affected by L. plantarum intervention (P>0.05).

Levels of the B[a]P-metabolizing enzyme CYP1A1, a member of the P450 enzyme family, were increased about six-fold after B[a]P treatment ( Figure 5A ; P<0.05), and L. plantarum CCFM8661, rather than L. plantarum CCFM382, significantly reduced its expression (P<0.05), thus alleviating the damage to the colon caused by B[a]P.

Colonic histopathology was normal in mice in the control group ( Figure 5B ). However, B[α]P treatment led to serious injury to the colon, including crypt destruction, inflammatory cell infiltration, and severe ulceration. Colonic ulcers were reversed after L. plantarum CCFM382 administration, but there was still moderate inflammatory cell infiltration. Treatment with L. plantarum CCFM8661 significantly alleviated the colonic histopathological lesions to almost normal levels. Therefore, oral administration of L. plantarum CCFM8661 had a better alleviative effects on colonic damage induced by B[α]P.

To determine whether the protective effects of L. plantarum CCFM8661 on B[α]P-induced colonic damage involved changes in the gut microbiota, the α diversity and composition of gut microbiota were measured. The number of observed species and the Shannon index were used to represent the gut microbiota richness and diversity, respectively. The Faith_pd index represents phylogenetic diversity, which is a qualitative measure of community richness. As shown in Figure 6A , the observed species and Faith_pd indices decreased significantly in the model group, indicating that the richness and diversity of the gut microbiota were dramatically reduced (P<0.05). Oral supplementation of L. plantarum CCFM8661 significantly increased these two indexes, but L. plantarum CCFM382 supplementation only elevated the observed species index (P<0.05). The results showed that L. plantarum CCFM8661 had a stronger regulating ability than CCFM382 on gut microbiota diversity.

The α diversity results confirmed that B[α]P and L. plantarum CCFM8661 could, indeed, change the composition of the intestinal microbiota. Thus, changes in the gut microbiota composition were further explored ( Figure 6 ). At the phylum level, Bacteroidetes, Firmicutes, Actinobacteria, and Proteobacteria accounted for more than 95% of the gut microbiota ( Figure 6B ). In the control group, these four predominant bacterial phyla accounted for 67.1%, 30.1%, 0.7%, and 1.8% of the gut microbiota, respectively. The abundance of Bacteroidetes, Firmicutes, and Actinobacteria in the model group was 52.2%, 38.9%, and 5.8%, respectively, whereas the abundance of Proteobacteria (2.3%) remained almost unchanged. After supplement with L. plantarum CCFM8661, the abundance of Bacteroidetes (62.7%), Firmicutes (30.8%) and Proteobacteria (1.1%) gradually recovered to the levels of the control group, while the abundance of Actinobacteria (5.0%) was similar to their abundance in the model group, without significant changes. The abundances of these four bacterial phyla in CCFM382 group accounted for 57.6%, 36.8%, 3.9% and 1.1% respectively. The results showed that CCFM382 also affected the gut microbiota composition, but the effects were not as strong as that of CCFM8661. At the genus level, Muribaculaceae, Bacteroides, Alistipes, and Lachnospiraceae NK4A136 accounted for 23.6%, 20.6%, 5.6%, and 4.4%, respectively, in the control group ( Figure 6B ). The abundance of Muribaculaceae and Bacteroides in the model group decreased to 13.8% and 8.9%, respectively, and the abundance of Alistipes and Lachnospiraceae NK4A136 increased to 10.8% and 7.9%, respectively, compared with those in the control group. In the CCFM8661 group, the abundance of Bacteroides increased to 14.6% and the abundance of Lachnospiraceae NK4A136 decreased to 2.5%. Moreover, after B[α]P treatment, the abundance of Ruminococcus, Lachnospiraceae, and Bacteroides decreased significantly and the abundance of Turicibacter increased ( Figure 6C , P<0.05). L. plantarum CCFM8661 reversed the abundance of these four genera to their levels observed in the control group (P<0.05), while L. plantarum CCFM382 only reversed the abundances of Turicibacter and Bacteroides.

Compared to the levels in the control group, the levels of SCFAs, including acetic acid, butyric acid, isobutyric acid and propionic acid, were dramatically reduced in mice treated with B[α]P ( Figure 6D ). L. plantarum CCFM8661 sharply increased the levels of these SCFAs (P<0.05) to levels almost the same as those in the control group. However, the administration of CCFM382 significantly increased only the levels of isobutyric acid among the SCFAs in B[α]P-treated mice (P<0.05). The results showed that L. plantarum CCFM8661 had a better effect than CCFM382 on increase of SCFA levels.

The parameters that were significantly affected by B[a]P and L. plantarum CCFM8661 were selected to assess their correlation using Pearson’s correlation coefficients ( Figure 7 ). The Zo-1 expression levels were positive correlated with the abundance of Ruminococcus (r=0.91), Lachnospiraceae (r=0.99), and Bacteroides (r=0.98), and negatively correlated with the Turicibacter abundance (r=-0.95), indicating that the gut microbiota was closely related to gut barrier function. The 3-OH B[a]P levels in feces had a strong positive correlation with Turicibacter abundance (r=0.97) but a negative correlation with the abundance of Ruminococcus (r=-0.97), Lachnospiraceae (r=-0.98), and Bacteroides (r=-0.99) and Zo-1 expression levels (r=-0.96). These results confirmed that the gut microbiota and the gut barrier affected 3-OH B[a]P excretion in the feces, and vice versa. Moreover, Bcl-2 expression levels showed a strong positive correlation with the abundance of Turicibacter (r=0.95) but a negative correlation with the abundance of Lachnospiraceae (r=-0.94) and Bacteroides (r=-0.91). p53 levels were positively correlated with Turicibacter abundance (r=1.00) and negatively correlated with the abundance Ruminococcus (r=-0.98), Lachnospiraceae (r=-0.93) and Bacteroides (r=-0.93). MDA levels were positively correlated with Turicibacter abundance (r=0.97) and negatively correlated with the Ruminococcus abundance (r=-0.94), indicating that the gut microbiota was related to changes in tumor- and oxidative stress- related parameters in the brain.

Heatmap analysis was used to identify the similarities and differences among the four groups, as similar groups would cluster together in this analysis. As shown in Figure 8 , the CCFM8661 treatment group was clustered with the control group, and the CCFM382 treatment group clustered with the model group, indicating that CCFM8661 had a strong protection against B[a]P-induced damage, which may almost recover to the levels of the control group, while CCFM8661 only had a little protective effects.