**4. Discussion**

In order to identify new bioactive peptides from fermented milk, antioxidant peptide enriched fractions were extracted and further purified with the aim of identify the sequence of the included peptides. More in detail, the fraction 5–30% ACN was selected for the high presence of peptides able to prevent oxidative stress. In fact, when Caco-2 cells were pretreated for 24 h with enriched peptide fractions, a protective effect on the viability was apparent when oxidative stress was induced by TbOOH (Table 1). Therefore, 5–30% ACN fraction was further purified by RP-HPLC obtaining five major fractions, selected on the basis of the highest peptide relative abundance (Figure 1). These collected fractions, called **6**, **7**, **8**, **9** and **10**, were studied for their antioxidant properties in vitro and in Caco-2 cells (Figure 1). In particular, fraction **6** showed a powerful activity in protecting cells from oxidative stress and thus this fraction was selected for the following studies. The sequences of the peptides included in fraction **6** were identified by using LC-MS/MS analysis and the appropriate software. From these data, 23 peptides were chosen according to the peptide coverage for a specific sequence of the reference proteins (see Supplementary Materials). The novelty of the peptides was demonstrated by their absence in the BIOPEP database. When present, their function was not related to an antioxidant effect. Three criteria have been considered for peptide selection: firstly, the maximum protein sequences coverage, in order to map each protein sequence as much as possible from N-terminus to C-terminus; secondly, the best match score, indicated by #PSM, between the peptides given by the Proteome Discoverer Software and the reference proteins; thirdly, the length of peptides, as peptides with less than 20 residues were preferred instead of longer ones. Peptide sequences derived from alignments with other proteins, such as short isoform of polymeric immunoglobulin receptor, glycosylation-dependent cell adhesion molecule, histatherin and 3-phosphoshikimate 1-carboxyvinyltransferase were not taken into account, as they were considered less important. Subsequently, the selected peptides were synthesised, characterized and tested for their antioxidant properties. Two in vitro antioxidant tests, ABTS and DPPH, were performed and, as apparent in Table 3, peptides **Q-13-N**, **K-10-S**, **I-15-A**, **N-10-L**, **F-9-F**, **Q-14-R**, **N-15-M**, **F-12-F** and **I-14-L** were the most active in the scavenging tests (in vitro). However, when all the 23 peptides were screened in a cellular model for their action against oxidative stress induced by TbOOH testing cell viability (Table 3), only peptides **V-10-N**, **Q-14-R**, **A-17-E**, **A-12-Q**, **V-12-Q**, **N-15-M**, **V-9-E**, **E-11-F** and **K-15-L** were e ffective. Subsequently, in order to confirm the antioxidant power of the peptides, ROS production was evaluated in cells pre-treated for 24 h with the peptides and then subjected to an oxidative stimulus induced by TbOOH. As shown in Figure 2, the most active peptides against ROS production were **N-15-M**, **E-11-F**, **Q-14-R**, **E-18-H**, **H-18-Q**, **A-17-E**, **D-17-T**, **S-17-Q**, **V-9-E**, **P-9-E** and **F-12-F**. From the analysis of all the obtained results (in vitro and in cell environment), **N-15-M**, **E-11-F**, **Q-14-R** and **A-17-E** emerged as the most active peptides. For this reason, these four peptides were used for further analysis. In particular, the mechanism of action of the selected bioactive peptides, involved in the protective e ffects against oxidative stress in Caco-2 cells, was investigated. We focused on Keap1/Nrf2 pathway because it is the main regulatory system in oxidative stress conditions. In fact, when an oxidative imbalance occur, Keap1 and Nrf2 dissociate and Nrf2 translocates to the nucleus where it can bind ARE, promoting the overexpression of antioxidant enzymes such as glutathione reductase (GR), NADPH quinone oxidoreductase (NQO1), superoxide dismutase (SOD1) and thioredoxin reductase 1 (TrxR1). Therefore, to understand the involvement of Keap1/Nrf2 pathway, the translocation of Nrf2 from the cytosol to the nucleus was considered in cells treated with **N-15-M**, **E-11-F**, **Q-14-R** and **A-17-E** for 24 h. As result, **N-15-M**, **Q-14-R** and **A-17-E** increased the levels of Nrf2 present in the nucleus (Figure 3) suggesting that they activate the Keap1/Nrf2 system. Subsequently, as the amount of Nrf2 in the nucleus increased, we observed also an increase of antioxidant enzymes gene expression. To this purpose, GR, TrxR1, NQO1 and SOD1 gene expression was measured in Caco-2 cells treated with the four peptides for 24 h and again **N-15-M**, **Q-14-R** and **A-17-E** were able to increase the gene expression levels of the antioxidant enzymes (Figure 4) and the consequent protein expression (Figure 5) estimated with WB analysis. Moreover, TrxR1 and GR enzymatic activities were measured in cells treated in the same conditions and **N-15-M** and **Q-14-R** increased the activity of the tested antioxidant enzymes (Figure 6). In order to confirm our observations, molecular docking analysis between the structure of Keap1 and the four peptides was performed (Figure 7). The results showed that **N-15-M**, **Q-14-R** and **A-17-E**, but not **E-11-F**, interacted with Keap1 in the Kelch domain with specific amino acid residues, involved also in the binding between Keap1 and Nrf2 (Table 4). The Kelch repeats sequence of Keap1 (AA 327–609) is responsible for the binding to Nrf2 which participates with the DLG (AA 29–31) and ETGE (AA 79–82) motifs [21]. In particular, the ETGE motif guarantees a strong binding of Nrf2 to Keap1. Specific amino acids residues in the Kelch repeats (especially Arg380 and Arg415) facilitate the binding to the transcription factor. Our peptides are able to interact with Keap1 sequence with many of the amino acids involved in the binding to the ETGE motif. As apparent, **Q-14-R** interacts with Arg 380, Asn 382, Arg 415, Ser 508, Tyr 525 and Gln 530, while peptide **N-15-M** interacts with Arg 380, Asn 382, Arg 415, Tyr 525 and Ser 602. Furthermore, the bioactive peptides examined are able to interact with several other amino acids of the Kelch domain such as Arg 336, Asn 387, His 432, Gly 433, Ile 435, Gly 509, His 575 and Thr 576. Of note, molecular docking approach showed that **E-11-F** did not interact with Keap1.

The overall results suggested that the antioxidant e ffects highlighted in the cells treated with **N-15-M**, **Q-14-R** and **A-17-E** were due to the interaction of the bioactive peptides with the Keap1 pocket, which causes the disruption of the binding with Nrf2 and the subsequent activation of the signaling cascade. All these findings were in agreemen<sup>t</sup> with our previous results [28].

Finally, we studied the capability of these peptides to cross the intestinal barrier. Using the Transwell ® technique, we administered the four peptides to di fferentiated Caco-2 cells in the apical compartment (Figure 8). After 10 and 120 min the apical and basolateral compartments were collected and analyzed by RP-HPLC and mass spectrometry. We observed that an appreciable amount of peptides were able to reach the basolateral compartment partly with modifications as showed by the mass spectrometry analysis (Figure 8). Table 5 reports in detail each peptide and breakdown fragments. More specifically, **N-15-M** and **E-11-F** showed the best intestinal barrier crossing capacity. In fact, the

0.13% and 0.21% of these peptides reached the basolateral compartment, respectively. On the other hand, **Q-14-R** and **A-17-E** underwent a slight fragmentation by the brush border peptidases and only the 0.02% and 0.05%, respectively, reached the basolateral compartment. This observation leads to think that the peptides ingested orally can reach the blood circulation and, if not cleaved, they can exert their beneficial e ffects in many organs and tissues.
