Screening of ATG4B inhibitors based on the FDA-approved drugs library by FRET assay
Taking ATG4B as the target, we used the high-throughput FRET-based assay to screen the FDA-approved drugs library containing 1600 compounds (Fig. 1A). After preparing the purified recombinant target protein ATG4B and the FRET substrate FRET-GATE-16 (CFP-GATE-16-YFP) (Additional file 1: Figure S1A, B), we tested the inhibitory effect of the reported positive compound S130 [10], which has a similar IC50 to that reported illustrating the successful establishment of the assay. (Additional file 1: Figure S1C). Then twenty-four candidate compounds (marked in red) were noted through an initial screening at 10 μM (Fig. 1B). To further validate their inhibitory effect, a more rigorous and authentic approach as gel-based assay was carried out (Fig. 1C). The brake of the cleavage reaction of the substrate FRET-GATE-16 was very intuitively reflective of the compounds’ inhibitory effect on the ATG4B enzymatic activity. Some fake positive compounds with weak inhibition such as compound 438 were excluded, and further nine commercially available compounds were subjected to IC50 determination (Additional file 1: Figure S1D). The IC50 values of most of the compounds showed close to 15 μM, meanwhile, compound 669 possessed the best inhibitory activity with an IC50 of 189 nM (Additional file 1: Figure S1D), compared with the value of 80 nM [14] for FMK-9a as well as 260 nM [23]. This was also verified in the FRET- and gel-based full concentration-response plot of compound 669 (Fig. 1D, E).
Totally, compound 669, reported as Ebselen (SPI-1005, PZ-51, CAS 60940-34-3) was obtained from the FDA-approved library by high-throughput FRET assay.
Ebselen is a highly active and selective inhibitor of ATG4B
Inhibition potency for nine candidates was further estimated by in vitro cleavage assay of LC3B-glutathione-S-transferase (GST) with overexpression of ATG4B in cell lysates [5]. As shown in Fig. 2A, B, the cleavage activity of ATG4B could be still strongly suppressed by compound 669/Ebselen, meanwhile other compounds showed poor activity in this assay.
The inhibition profile of endogenous ATG4B in cells was also noted, and as shown in Additional file 1: Figure S2A, B, ATG4B was inhibited in a time- and dose-dependent manner. Given that ATG4 possesses four isoforms (A, B, C and D), while ATG4C and ATG4D have little significant substrate enzymatic activity [6], so the selectivity of Ebselen for ATG4A and ATG4B was taken into account. In the FRET-based enzymatic cleavage experiments, it was obvious from Fig. 2C that the Ebselen had a weaker inhibition for ATG4A compared to ATG4B. Whereas in the validation results in gel (Fig. 2D), 1 μM of Ebselen gave cleaved substrate (CFP-GATE-16 and YFP) appearance compared to complete inhibition of ATG4B. In consideration of ATG4B as a cysteine protease, we determined the performance of the compound for some classical cysteine proteases, Caspase 1, 2, 3, 4, 6, 8 and 9 (CASP1, 2, 3, 4, 6, 8 and 9). As shown in Fig. 2E, F, Ebselen with the concentration of 10 μM had no significant effect on most caspase activities, but a certain inhibition on CASP2 and CASP3.
Further, structure-activity relationship (SAR) studies were implemented, based on the literature [24] and commercially available structural analogues. As is shown in Fig. 2G and Additional file 1: Figure S2C, the Se atom of the “benzo[d][1,2]selenazol-3-one” backbone remained unchanged, replaced with S atom, or O atom of similar chemical properties, as well as some simple side chain group changes. Among the compounds tested (Fig. 2H, Additional file 1: Figure S2C), the Se atom exhibited a striking inhibition (IC50 < 2 μM) and S atom being the second, while the O atom or the modified S atom (SO2, O=S=O) showed no inhibitory activity. Then the activity changes by side chain group were tentatively not regular, as these data came from irregular and few side chain alterations. Although we did not get lead compounds with better activity, it also suggested that structural modification was feasible. It is noteworthy that the inhibition of Ebselen on CASP3 had been explained, because pieces of literature had reported the clear inhibition of S atom analog Ebsulfur (CAS 2527-03-9) on CASP3, and related SAR studies and patents had been published [25, 26].
In general, Ebselen showed excellent inhibition effect in gel-based assay and exhibited controllable selectivity against ATG4A and caspases.
Ebselen can covalently bind to ATG4B at Cys74
The inhibitory mechanism and pattern of binding of the compound against ATG4B were next investigated. Firstly, a molecular docking study was performed, based on the current reported binding models of Ebselen [27]. We obtained that the compound was covalently bound to the Cys74 of ATG4B (Fig. 3A, B), which is precisely the mainly active catalysis site for ATG4B enzymatic cleavage. It was known that Trp142 was a key to switching the autoinhibitory-loop of ATG4B [28], and the compound just formed two pi-pi stacking interactions with Trp142, as well as other hydrophobic interactions with Tyr143, Pro145, Gly258, Ala263, etc. Subsequently, the ATG4B C74S mutant ATG4BC74S (Cys74 to Ser74) was purified to perform in vitro thermal shift assay (TSA) for further studies (Additional file 1: Figure S3A) [29]. The thermal stability of the wild type (WT) ATG4B was strikingly improved compared with DMSO as control, while the stability of the C74S mutant significantly decreased (Fig. 3C, D). This revealed that the Ebselen may indeed covalently bind to ATG4B through the Cys74 site. Compared with the DMSO group, the stability of the C74S mutant was still a little increased with the compound (Fig. 3C, D), which also verified the existence of other interactions based on the docking results.
Mass spectrometry has always been able to provide direct evidence of protein-ligand binding [30]. Therefore, denaturing mass spectrometry was operated to prove the covalent binding of ATG4B with Ebselen. As shown in Fig. 3E, F and Additional file 1: Figure S3B, the spectrometry diagram of the protein sample without compound showed that there was a main peak of 44,522 Da (monomer molecular weight of ATG4B) and a small peak of 57,200 Da. SDS-PAGE and western blot were performed on the mass spectrometry samples, then the peak of 57,200 Da was indeed confirmed to be ATG4B (Fig. 3G). This form of ATG4B had also been observed in recent report [31], and it was speculated as an unknown aggregation form of ATG4B.
Then the compound Ebselen was added to the system, and surprisingly, the monomer ATG4B peak in the spectrum was disappeared (Fig. 3H). And the remaining two main peaks (57,200 Da and 57,475 Da) rightly had one compound molecular weight shift (Ebselen: m/z = 275 Da) (Fig. 3H–I), which showed that the Ebselen covalently binds to ATG4B with the 57,200 Da form. Non reducing SDS-PAGE is a good method to detect the aggregation form of protein, because the absence of reductant makes the disulfide bond remain. Interestingly, the monomer of ATG4B (~ 44 kDa) was also disappeared in non-reducing electrophoresis when treated with Ebselen, which was consistent with the results of mass spectrometry (Fig. 3H–J). Thus, we confirmed that Ebselen can covalently bind to ATG4B at Cys74 through in vitro TSA and mass spectrometry experiments.
Ebselen can promote ATG4B oligomerization at Cys292 and Cys361
The disappearance of the ATG4B monomer made us have to consider a novel post-translational modification of ATG4B, namely, redox and oligomers formation [32]. Subsequently, ATG4B samples were incubated with different mole ratios (1:1, 1:5, and 1: 10) of Ebselen and for different time points (0.5, 1, and 3 h), and the results were detected by non-reducing electrophoresis. As shown in Fig. 4A, with the increase of the concentration of the compound or the extension of the incubation time, the ATG4B monomers were obviously gradually reduced, then oligomerization occurred, and even aggregates accumulated on the sample wells, finally. Notably, the classical covalent ATG4B inhibitor FMK-9a in the figure did not show similar properties, although a little aggregation occurred due to the prolonged oxygen exposure in the air (Fig. 4A). The aggregates formed could be reversibly regulated by a reducing agent, because an assay to add dithiothreitol (DTT) into the reaction was operated. Whether the compounds were incubated with different concentrations of DTT in advance (Fig. 4B), or the DTT was added after incubation between protein and compounds (Fig. 4C), ATG4B aggregates could return to the level of monomer in non-reducing electrophoresis. Furthermore, we also evaluated the redox properties of analogues on ATG4B. As is shown in Additional file 1: Figure S4A, most compounds with inhibitory effect can promote the oligomerization of ATG4B. This suggested that the ability to promote oligomerization may also be an evaluation aspect in the future structure optimization.
There were many studies on the redox of ATG4B, among which the most well-known reaction sites were Cys292 and Cys361, generating intermolecular disulfide bonds to form oligomers [32]. Therefore, we expressed and purified the mutant ATG4B2CS (Cys292&Cys361 all to Ser) reportedly that could not undergo air-oxidation (Additional file 1: Figure S4B). It was intriguing that the 2CS mutant did not undergo oligomers modification when treated with Ebselen (Fig. 4D). These results clearly revealed that the oligomers of ATG4B formed by the compound depended on the classical site Cys292 and Cys361 rather than Cys74 (Fig. 4D and Additional file 1: Figure S4C). Naturally, we speculated that the covalent binding of the 2CS mutant with the Ebselen existed as the monomer and can be detected by mass spectrometry. And here in Fig. 4E, after two Cysteines were mutated to Serines (-32 Da), the molecular weight peaks of the 2CS mutant monomer were detected (44,490 Da, 44,414 Da, typically), and rightly covalent binding peaks with one compound molecular weight shift (275 Da) were also observed (44,765 Da, 44,689 Da). These above results demonstrated that Ebselen was able to covalently bind to ATG4B, especially the form of monomer. This also reconfirmed that the oligomerization of ATG4B after compound treatment was responsible for the disappearance of the monomer in the mass spectrometry.
Further, we verified the inhibition of ATG4B by oligomerization from Ebselen. As is shown in Fig. 4F, Ebselen showed higher inhibition rate against WT, compared to 2CS, which was less than 90%. Meanwhile, the WT group showed an increasing inhibition with increasing incubation time, while the 2CS showed almost no more increase. Interestingly, this result remained consistent with the increasing oligomerization of ATG4B in Fig. 4A. In other words, the gap in the inhibitory effect exhibited between WT and 2CS was a manifestation of the WT oligomerization modified by Ebselen.
In summary, ATG4B was also oligomerization modified due to oxidative regulation at Cys292 and Cys361, which might enhance the inhibitory effect of Ebselen on ATG4B.
Ebselen suppresses autophagy flux via inhibition of ATG4B
Since ATG4B is a key protein in autophagy, the experiment on the effect of Ebselen on autophagy flux was worth proposing. First of all, we determined the most classic index LC3B for autophagy flux detection in HeLa cell line [33]. When the classic mTOR inhibitor Torin 1 and lysosomal inhibitor CQ (Chloroquine) were used together, the accumulation of LC3B-II indicated that Torin 1 was an autophagy inducer as commonly known and reported (Fig. 5A, B) [33].
Returning to the role of ATG4B on autophagic flux, when it was inhibited or knocked down, its effect on autophagy had been preliminarily concluded in some studies [9, 34, 35]. In our work, compared to the control, the addition of sgATG4B lentivirus in SW620 greatly reduced the protein level of ATG4B (Fig. 5C). Obviously, the level of LC3B-II was also reduced, greatly, when combined with CQ. It meant that knockdown of ATG4B brought about inhibition of autophagy flux (Fig. 5C, D). Such conclusion was also in line with most current studies. Since, in the latest theory, it was believed that ATG4B is involved in the growth and elongation phase of phagophore by interacting with ATG9A and dynamically regulating the lipidation/delipidation of LC3B [36, 37]. Similar results were obtained that the inhibition of autophagic flux was revealed by the reduction of LC3B-II, when the inhibitor Ebselen and CQ were used together in both HCT116 and SW620 cell lines (Fig. 5E–H). More intuitive and identical results were shown in Fig. 5I, J, the number of endogenous LC3B dots as autophagosomes was decreased when compound applied. Furthermore, a gold standard long-lived protein degradation assay was operated to confirm the inhibition. As is shown in Fig. 5K, Ebselen significantly inhibited autophagic long-lived protein degradation induced by Torin 1. In all, the inhibitory properties of Ebselen with respect to autophagy were comprehensively elucidated via inhibition of ATG4B.
Ebselen suppresses the growth of colorectal cancer cells via inhibition of ATG4B
Reportedly, ATG4B was a potential therapeutic target for colorectal cancer (CRC, COAD), although not fully elucidated [9,10,11]. So further studies were urgently carried out, to demonstrate whether Ebselen could intervene or treat CRC models as a potential inhibitor of ATG4B. Firstly, the role of ATG4B in CRC was reaffirmed using bioinformatics analyses. The colorectal cancer data in the TCGA database were abstracted to investigate the relationship between COAD and ATG4B expression. As shown in Fig. 6A, the expression of ATG4B was significantly higher in CRC patients than in the normal subjects. Importantly, ATG4B expression in colon cancer cells (SW620, HCT116, and RKO) was indeed higher than that in normal colon cell (NCM460) (Additional file 1: Figure S5A). Moreover, in the occurrence and development in grades of colorectal cancer, high expression of ATG4B also appeared in grade 1, grade 2 and grade 3, which might be responsible for tumor deterioration (Fig. 6B). Meanwhile, by correlating ATG4B expression with survival curves, we concluded that high expression of ATG4B was a risk factor for CRC patients (Fig. 6C).
ATG4B knockout cells (HCT116 [10], RKO, SW620) were prepared, to confirm the effect on the growth of CRC cells when ATG4B was deficient (Fig. 6E). Firstly, colony formation assays were performed, and the ATG4B deficient cells formed significantly fewer colonies than the wild types (Fig. 6D). Cell viability was determined by CCK-8 assay, however, the effect of ATG4B deficiency or not on cell viability was not significant (data not shown). The difference became significant with the compound treatment. As shown in Fig. 6F-H, the growth of wild-type CRC cells (Particularly HCT116 and SW620) was significantly inhibited (typically, IC50 = 18.9 μM for WT HCT116), instead of the ATG4B knockout cell lines (IC50 = 28.1 μM for ATG4B KO HCT116). This also did reveal that the Ebselen was indeed working through ATG4B to suppress the growth of relevant colorectal cancer cell lines. The effects of the compound were also illustrated by cell counting and colony formation assays (Fig. 6I-J). And the results showed that Ebselen was able to significantly inhibit cell proliferation and colony formation.
Understandably, Ebselen showed nonselective cell killing at very high concentrations, so apoptosis detection was carried out. The appearance of cleaved poly ADP-ribose polymerase (PARP, substrate of the apoptosis execution protein CASP3) indicated the generation of apoptosis, when treated with the positive-induction compound Staurosporine (STS) (Additional file 1: Figure S5B-C). However, cells did not show any increase in apoptosis in response to treatment with low or high concentrations of Ebselen, in consist with the inhibition of CASP3 assayed earlier. In other words, these results above illustrated that Ebselen could indeed work to inhibit CRC cell proliferation via ATG4B, not via apoptosis. But the growth arrest effects of Ebselen on CRC cells may require further studies in the future, due to the dual inhibitory properties of the compound against both ATG4B and CASP3. However, it is noteworthy that recent studies suggested that inhibition of CASP3 could be also a potential target for cancer therapy. Because CASP3 and apoptosis could promote tumor growth, metastasis and angiogenesis in cancers, such as CRC, glioblastoma, gastric cancer, etc. [38,39,40].
In a word, we confirmed that Ebselen can suppress the growth of CRC cells via inhibition of ATG4B.
Ebselen suppresses colorectal cancer xenograft tumor growth
After the determination of the relevant in vitro cellular experiments, SW620 and HCT116 cell lines were xenografted into immune-deficient nude mice (BALB/c-nu/nu). Groups (Vehicle, 5 mg/kg, 10 mg/kg and 20 mg/kg, administration every two days) were generated randomly when tumors reached indicated sizes. The volume of tumors and body weight of mice were measured every two days (Fig. 7A). Compared with the vehicle group, the treatment group could significantly suppress the growth of tumors (Fig. 7B–D), and no obvious mice body weight change or organs damage was observed (Additional file 1: Figure S6A-B). Although the group of 5 mg/kg treatment had shown obvious anti-tumor effect, there was no difference between the medium and high concentration groups, which had better tumor suppression. Xenografts in BALB/c-nu/nu from HCT116 cell were also evaluated, showing a slightly weaker potential for tumor suppression (Additional file 1: Figure S6E–G).
After tumors were dissected, FRET assay and western blot were implemented to detect ATG4B activity and autophagy in tumors. As is shown in Fig. 7E, the activity of ATG4 (mainly ATG4B) in tumors was significantly inhibited in a dose-dependent manner. Then compared with the vehicle group, the autophagy flux in tumors of the administration groups was also significantly inhibited because of the reduction of LC3-II and the accumulation of the general autophagy cargo p62 (Fig. 7F–H, Additional file 1: Figure S6C-D). Meanwhile, Ebselen also inhibited the apoptosis of tumors in a dose-dependent manner (Fig. 7F, I). And we noticed that the background apoptosis in the vehicle group was higher, which was not coincident with the theory that the preceding apoptosis promoted the occurrence and development of tumors in some ways. So apart from the inhibition of ATG4B to generate antitumor effect, on the other hand, the inhibition of CASP3 also cannot be simply ignored. Overall, the results obtained from tumors were in keeping with the results of the previous in vitro and cellular experiments.
The tumor therapeutic potential of Ebselen via ATG4B was also further evaluated by immunohistochemical staining for Ki67, an indicator of tumor malignant proliferation (Fig. 7J, K). It showed a significant reduction of Ki67 positivity in the drug-treated group, demonstrating some reduction in tumor proliferation and malignancy. Taken together, Ebselen can suppress colorectal cancer xenograft tumor growth in vivo.