Abstract

The Bcl-2 family members rigorously regulate cell endogenous apoptosis, and targeting anti-apoptotic members is a hot topic in design of anti-cancer drugs. At present, FDA and EMA have approved Bcl-2 inhibitor Venetoclax (ABT-199) for treating chronic lymphocytic leukemia (CLL). However, inhibitors of anti-apoptotic protein BCL2A1/Bfl-1 have not been vigorously developed, and no molecule with ideal activity and selectivity has been found yet. Here we review the biological function and protein structure of Bfl-1, discuss the therapeutic potential and list the currently reported inhibitory peptides and small molecules. This will provide a reference for Bfl-1 targeting drug discovery in the future.

1. Introduction

The balance between cell growth and death maintains homeostasis. Cell death is usually divided into necrosis and programmed cell death. Apoptosis and autophagy are the two basic types of programmed cell death, and both are regulated by numerous signaling pathways and related factors [1e5]. Apoptosis acts as a protective process for human body. In this process, cellular structures (including chromatin, nucleosomes, cell membrane vesicles) are morphologically altered, and apoptosis factors are released from mitochondrial membrane to promote the downstream caspase pathway (Fig.1) [6,7]. And disordered, damaged, infected, and redundant cells are actively removed in this way [8,9]. Apoptosis disorder promotes tumor growth under stimulating conditions such as oxidative stress. Therefore, targeting dysfunctional apoptosis process is a promising way for cancer therapy. The endogenous apoptosis disorder is mainly induced by the inhibition of death receptor signaling pathway, the breakdown of the balance between pro-apoptotic and anti-apoptotic proteins, and the impaired caspase activity [10e12]. At present, a variety of chemotherapy drugs have been reported to restore the natural apoptotic ability of tumor cells by inhibiting these three processes, and then play an ideal anticancer effect.

The B-cell lymphoma-2 (Bcl-2) family is a signiicant regulator of intrinsic apoptosis. It can be segmented into three classes: proapoptotic proteins (Bax, Bak, Box etc.), BH3-only proteins (Bim, Bid, Bad, Bmf, PUMA, NOXA etc.) and anti-apoptotic proteins (BclxL, Mcl-1, Bfl-1, Bcl-2 etc.) [18]. All these members are composed of at least one BH (Bcl-2 homology) domain and most of them contain a hydrophobic domain at the carboxy terminal for membrane positioning (transmembrane, TM). Pro-apoptotic proteins are composed of BH1-3 regions. Among the anti-apoptotic members, Bcl-2 and Bcl-xL contain all four BH domains, while Mcl-1 and others only possess homologyin BH1-3 region (Fig. 2) [19].

Endogenous stimulation induces the oligomerization of Bax and Bak on mitochondrial outer membrane and promotes mitochondrial outer membrane permeability (MOMP). Then cytochrome c and other apoptosis-inducing factors are released to activate the downstream caspase pathway to cause cell apoptosis [20,21]. Antiapoptotic proteins, also known as pro-survival proteins, neutralize pro-apoptotic proteins and block MOMP [22,23].

Two hypotheses for how the Bcl-2 family regulates apoptosis were proposed: direct and indirect. The direct model considers BH3-only members as the agonists or inhibitors. The agonist BH3only proteins directly activate Bax/Bak protein to promote apoptosis, while the inhibitor BH3-only proteins bind to antiapoptotic members and release Bim, tBid and Puma to activate Bax/Bak. Indirect model suggests that only BH3-only proteins bind to anti-apoptotic Bcl-2 family members, and then Bax/Bak is released to initiate apoptosis [24]. There’s evidence suggesting that both the two mechanisms may coexist in apoptosis regulation [25].

At present, anti-apoptotic proteins of Bcl-2 family have become a hot topic for tumor therapy [26e30]. By simulating the structure of BH3 peptides, it is possible to develop inhibitors targeting the hydrophobic cavity of the anti-apoptotic members. Thereby the ininite proliferation of cancer cells can be controlled through this way [31]. Nowadays inhibitor GX15-070 (Obatoclax) [32], ABT-263 [27] and ABT-737 [33,34] have entered clinical trials. In addition, the Bcl-2 inhibitor ABT-199 (Venetoclax) [26,32] jointly explored by AbbVie and Roche has been approved by FDA for curing chronic lymphocytic leukemia (CLL), small lymphocytic lymphoma (SLL), acute myelogenous leukemia (AML) and other indications. And Ascentage Pharma’s APG-2575 has obtained orphan drug qualiication for the treatment of Waldenstrom Macroglobulinemia (WM).

Bfl-1 (human), also known as BCL2A1 (murine) or GRS, is one of the Bcl-2 anti-apoptotic members. In 1991, Amos Orlofsky et al. used granulocyte-macrophage colony-stimulating factor (GM-CSF) to stimulate bone marrow-derived macrophages, and found that Bfl-1 showed an early transcriptional induction response to cytokines [35]. The expression of Bfl-1 was subsequently observed in a variety of tissues, primarily in the hematopoietic system. It is called Bcl-2 related gene expressed in fetal liver (Bfl-1) because it is identiied in embryonic liver. Bfl-1 has been observed to be associated with a variety of lymphomas and leukemias, particularly with diffuse large B-cell lymphoma (DLBCL). Overexpression of Bfl1 mRNA has also been observed in solid tumors such as breast cancer, gastric cancer, melanoma, and it is mainly relevant to advanced or metastatic disease stages [36]. In addition, Bfl-1 inhibits the autophagy of endothelial cells by binding to Beclin-1 protein, and induces the proliferation of macrophages and mast cells in allergic reactions, indicating that it makes effects in the immune system [37].

Fig. 1. Anti-apoptotic mechanism of Bfl-1 protein. Under various stimulating conditions, Bfl-1 protein competes with BH3-only protein NOXA, PUMA, tBid and other proteins to prevent Bax and Bak from gathering on the mitochondrial outer membrane, hinders the release of apoptosis factors, and thereby inhibits the downstream caspase pathway.

Fig. 2. The domain composition of Bcl-2 family members. The pro-apoptotic proteins usually contain the BH1-3 domain, the anti-apoptotic proteins contain the BH1-4 domain, and the BH3-only proteins, as the name suggests, are only composed of BH3 domain. In addition, Bfl-1 protein is similar to other anti-apoptotic proteins of Bcl-2 family, containing BH14 domain, but has no obvious transmembrane C-terminus.

The structure of Bfl-1 is similar to other anti-apoptotic proteins. It has BH1-4 hydrophobic plaques and binds to pro-apoptotic proteins such as NOXA and Bid [38]. Its amphiphilic C-terminus is different from other hydrophobic members, and also plays a role in membrane localization [39]. In recent years, some peptides and small molecules with inhibitory effect on Bfl-1 have been reported successively. Unfortunately, most of the existing Bfl-1 inhibitors lack ideal selectivity and potency.In this paper, the biological function and structure of Bfl-1 are briefly introduced, and the peptides and small molecule inhibitors of Bfl-1 protein are minutely listed. We hope that it can provide references for future research on Bfl-1 targeting therapeutic agents.

2. Biological function and oncogenic role of Bfl-1 protein
2.1. Bfl-1 in cancer

Just like the mechanism of other anti-apoptotic members, Bfl-1 inhibits the dimerization of Bax and Bak. Then the depolarization of mitochondrial membrane and release of apoptotic factors are prevented. By this means, Bfl-1 blocks the downstream caspase pathway and inhibits apoptosis. BH3 proiling indicates that Bfl-1 interacts with Bim, and PUMA [40], and molecular dynamics experiments show that it has high afinity with Bak and Bid [41,42]. It was also found that Bfl-1 weakly bound toBIK, NOXA and HRK by fluorescence polarization (FP) assay, which indicated a similar binding proile as Mcl-1 [43,44]. Moreover, Seckho Ha reported that Bfl-1 protein contributed to apoptosis escape by binding to the splicing variant of mouse ING1 homologous (mINGh) to obstruct the mINGh-induced epithelial cell death [45].

In certain circumstances, Bfl-1 promotes apoptosis. The a helixes at C-terminal membrane are cleaved from full-length Bfl-1 and then transform Bfl-1 into a pro-apoptotic protein. This process is mediated by m-calpain. The a5 helix of Bfl-1 promotes apoptosis via a Bax/Bak-dependent mechanism while the a9 helix through a membrane instability mechanism [39,46]; The 29 amino acidfragment at the C-terminal of Bfl-1 and green fluorescent protein form a fusion which converts Bfl-1 into a pro-apoptotic protein [47]. Combination of the fusion complex and Gemcitabine can make non-small cell lung cancer cells more sensitive to Gemcitabine [48].

At present, it has been reported that Bfl-1 assisted in maintaining homeostasis of mature B cell, and was regulated by Spleen tyrosine kinase (Syk) and Bruton tyrosine kinase (Btk), the vital factors of B-cell receptor (BCR) downstream survival [49]. In addition, Bfl-1 helps proteins such as Mcl-1 induce tumorigenic transformation in v-myc avian myelocytomatosis viral oncogene homolog (MYC) transgenic pre-B cells [50]. The mRNA of Bfl-1 overexpresses in various hematological malignancies, especially in some types of lymphoma and leukemia models [23,36,51], and has lineage-speciic and stage-speciic effects on the development of lymphocytes [52e54]. Bfl-1 is also named as Glasgow rearranged sequence (GRS), because it is involved in chromosome rearrangement in chronic myeloid leukemia and related to ibroblast growth factor.

Bfl-1 plays a role in lymphomas, including chronic lymphocytic leukemia [23,55e59], mantle cell lymphoma, acute lymphocytic leukemia and large B cell lymphoma, and especially in DLBCL cells [60,61]. Furthermore, Bfl-1 mRNA is elevated in solid tumor tissues such as breast cancer [62e64],lung cancer [48], hepatoma [65,66], gastric cancer, prostate cancer and melanoma [55,67]. Overexpression of Bfl-1 does not influence central nervous system (CNS) metastasis, but contributes to intracranial tumor growth, which has a great impact on melanoma spontaneous central nervous system metastasis [68]. Analyzing the disease stages with Bfl-1, we can ind that Bfl-1 is usually associated with advanced or metastatic diseases [69].

Fig. 3. Sequence of human Bfl-1 protein and distribution of hydrophobic BH1-4 regions and the a1-8 helixes.

Fig. 4. Crystal structure of Bfl-1/Bim (PBD code: 2VM6) and Mcl-1/Bim complex (PBD code: 2PQK). (A) Four amino acids in Bim BH3 peptide occupy four pockets of the Mcl-1 cavity respectively. Bim D67 forms a salt bond with Mcl-1 R263. (B) Five amino acids in Bim BH3 peptide occupy ive pockets of the Bfl-1 cavity respectively. Bim D157 forms a salt bond with Bfl-1 R88.

The quantity of Bfl-1 in vivo is regulated by several factors. Under oxidative stress and other stimulation conditions, the transcription of Bfl-1 is activated by NF-kB [70e72]. The transcription factor Pu.1 positively regulates Bfl-1 in neutrophils, causing mature neutrophils to escape apoptosis in a short time [73]. The oncogene melanocyte inducing transcription factor (MITF) transcriptionally regulates Bfl-1 to overexpress. Both of MITF and Bfl-1 are necessary for melanoma to grow [67]. Retinoid X Receptor agonists AGN194204 and 9-cis-retinoic acid make the BFL1 promoter stronger in naive T lymphocytes and promote the expression of the protein, which explains the decrease of naive T lymphocyte apoptosis after RXR agonist treatment [74,75]. On the contrary, Suppressors of Cytokine Signaling (SOCS) down-regulates Bfl-1 in leukemia T cells and promotes Fas-induced apoptosis [76].Isopropylated Rab receptor 1 (RABAC1 or PRA1) can induce apoptosis of AGS gastric cancer cells by inhibiting Bfl-1 [77]. MiR140-5p may enable MUC1 to inhibit the proliferation and metastasis of TNBC cells by regulating the Bfl-1/MAPK pathway [78]. At the post-translational level, Bfl-1 is regulated by the ubiquitin/ proteasome pathway, but its E3-ligase has not been found yet [79,80].

Bfl-1 protein exerts drug resistance function in a series of disease models, including fludarabine-resistant B cell chronic lymphocytic leukemia (B-CLL) cells [81,82], ABT-737-resistant OCI-Ly1 and SU-DHL-4 lymphoma cell line [83], and Venetoclax (ABT-199)resistant DHL cell line [84]. Based on this, several theories have been proposed to explain the mechanism. L.A. Simpson proposed that Bfl-1 was a target of Wilm tumor gene1(WT1). WT1 induced the expression of Bfl-1 by binding to the 71-bp fragment of Bfl-1 promoter and protected cell from apoptosis [85]. Anna EsteveArenys proposed that Bfl-1 was a target of MYC and was regulated by it [50,84]. BET bromodomain inhibitors JQ1 and CPI203 reduced the expression of Bfl-1 by inhibiting the MYC transcription program, which could overcome the resistance to Venetoclax (ABT-199) in DLBCL cells [86]. These statements prompt us that Bfl-1 is a potential ideal therapeutic target.

Fig. 5. Characteristics of Bfl-1 cavity. (A) Overlap of Bfl-1 and Mcl-1 helical structures. Purple represents Mcl-1 protein and yellow represents Bfl-1 protein. The three-dimensional structures of the two proteins are very similar, only showing signiicant differences in the transition from a3, a4 to a5 and a6. (B) C55 is a Bfl-1 speciic residue, which can be used to design selective inhibitors.

Fig. 6. (A) The L21 of NOXA and the W147 of Bim peptide are close to Bfl-1 C55, making the two residues substrates for covalent transformation. (B) Structures of helical stapled peptides NOXA SAHBA-3 (1) and Bim SAHBA-3 (2).

2.2. Bfl-1 in other diseases

In addition to carcinogenic effects, Bfl-1 plays a part in many other diseases.Inflammatory effect: Studies have shown that Bfl-1 is upregulated in endothelial cells after being induced by monocytes, thereby prevents apoptosis [87,88]. Under pro-inflammatory stimulation, Bfl-1 is crucial in the early stage of neutrophil differentiation and mature neutrophils. It promotes the survival of dendritic cells (DC) in rheumatoid arthritis (RA) and Langerhans cell histiocytosis (LCH) [89]. These results demonstrate its anti-inflammatory potential [90].Allergic effect: LgE receptor FcεRI-activated mast cells induce expression of Bfl-1 mRNA. The mast cells lacking Bfl-1 cannot be activated under allergic conditions, although they are capable of releasing granular mediators [91]. Knockout of BFL1 gene exempts the allergic reaction [92]. These results suggest that Bfl-1 can regulate allergic reactions [93e95].Inhibition of autophagy: It was reported that Bfl-1 bound to the autophagy Beclin-1 protein and hindered the formation of Beclin1-VPS34 complex. Bfl-1 prevents the virulent mycobacteria from being destroyed in host macrophages by inhibiting the autophagy process [37].

Motor neuron injury: Typical mutant Superoxide dismutase 1 (SOD1) in familial amyotrophic lateral sclerosis (ALS) upregulates Bfl-1 through activating the redox-sensitive transcription factor AP1. The a9 helix at C-terminal of Bfl-1 interacts with pro-caspase3 to inhibit the activation of caspase-3. Therefore, Bfl-1 makes a therapeutic effect in ALS by inhibiting the apoptosis of motor neurons [96,97].

3. The structure of Bfl-1 protein

The 14-kb BFL1 DNA locates in the Human Chromosome 15 (15q24.3) [98] and it includes three exons which can be translated into 175 amino acids. As one of the anti-apoptotic members, it contains the normal BH1-4 motifs (Fig. 3). The amphipathic C-terminus locates the protein to the mitochondria. As mentioned above, it also converts Bfl-1 into a pro-apoptotic protein under the control of m-calpain, and regulates the anti-inflammatory function in endothelial cells [99,100].The mRNA of the homolog Bfl-1S (short form of alternative splice variant) has a 56-bp exon inserted in the middle. This insertion leads to termination in early translation of the upstream stop codon, and results in a protein with 163 amino acids. It is shown that the sequence at C-terminus of Bfl-1S replaces ENGFVKKFEPKSGWMTFLEVTGKICEMLSLLKQYC in Bfl-1 with GKWHNHTPMLVESVAHKKRKMAL, which lacks the hydrophobic transmembrane region. The four residues KKRK make the C-terminus positively charged to work as the NLS (nuclear localization sequence) domain, then guide the Bfl-1S to enter the nucleus. Consequently Bfl-1S is a nuclear-targeted protein rather than a mitochondrial-targeted protein like Bfl-1. In addition, Bfl-1S retains the same BH1, 3, 4 domains and partially modiied BH2 domains as Bfl-1. This isomer is found in high levels in the lymph nodes and low levels in the spleen. Likewise, it delays the apoptotic process by hindering the cleavage of Bid and caspase family [101].

Bfl-1 (murine) has 72% amino acid homology with Bfl-1 (human) protein and is located in mouse Chromosome 9 [98,102]. There are four functional subtypes of murine BFL1 gene: A1-a, b, c, d. A1-a, A1-b and A1-d all have a DNA identity over 97% and identical open reading frames (ORFs) to encode BH1 and BH2 domain. Compared to these three subtypes, A1-c has a 1-bp nucleotide inserted at the position 438, resulting in ORF shift to generate a shorter sequence with only BH1 domain [45,103].

Human Bfl-1 and Mcl-1 proteins are often lumped together in Bcl-2 family due to their similar action patterns and binding cavities. The Bfl-1 protein folds to form eight a-helixes. The a2, a3, a5, a7, and a8 helix form a classically long and narrow hydrophobic groove. The groove mainly contains the BH1-3 domain (Fig. 3). By comparing the co-crystal structure of Bim (BH3-only peptide)-Bfl-1 (PBD code: 2VM6) and Bim-Mcl-1 (PBD code: 2PQK), we can ind that the three-dimensional structures of the two proteins, namely the helical distribution and the shape, are very similar. The two conformations only have signiicant differences in the transition from a2, a3 to a5, a6 (Fig. 5A). In the binding mode, four amino acids of the Bim BH3 peptide respectively occupy the four cavities of Mcl-1 protein, and its D67 residue forms a salt bond with Mcl-1 R263 residue (Fig. 4A). Similarly, ive amino acids of the Bim BH3 peptide (I148, L152, I155, F159, W147, Y163) occupy ive cavities of Bfl-1 respectively to produce hydrophobic effect, and the Bim Y163 residue occupies the groove formed by Bfl-1 F148 and V40 residue (known as the h5 hydrophobic domain). In addition, the D157 forms a salt bond with the Bfl-1 R88 residue (Fig. 4B). The Bfl-1 BH3 binding groove is characteristic by the negative charge from E78 and E80 in its central region [38,104].

In addition, the Bfl-1 C55 residue is a unique amino acid of the Bcl-2 family (Fig. 5B). The disulide bond between the thiol groups of C55 and NOXA makes a strong afinity, and this covalence exists only in humans, not in mice. Korshavn demonstrated that Bfl-1 C55 could form a speciic disulide bond with C175 residue at the a9 helical to prevent formation of the binding groove. The redox state of C55/C175 determined the position of the a9 residue and the capacity of the pocket, suggesting that the conformation controlled the binding groove of Bfl-1 [105]. In summary, the narrow hydrophobic groove and negatively charged center are the keys to design Bfl-1-targeting molecules. Based on the covalent action of C55 residue, there is a great potential to design Bfl-1 selective covalent inhibitor [44].

Fig. 7. Covalent peptide D-NA NOXA SAHB-15 F32A (3) and D-NA-NOXA SAHB-15 R31E (4) which alleviate membrane lysis by reducing the hydrophobicity (alanine scanning) and the positive polarity of the peptide chain (mutation of arginine residues to glutamate).

Fig. 8. Inhibitory activity of Fairlie’s electrophilic peptide 5 derived from BH3-only peptide Bim.

Fig. 9. Structures of peptide 6 and 7, derived from Bim BH3 peptide, featuring two cross-linking bridges. Peptide 6 has the highest binding eficiency among the bicyclic helical peptides which were modiied from single helical peptide to maintain the a helicity. Peptide 7 was modiied by replacing the acrylamide with propionamide, and the Glu at position 6 with Leu, leading to the high selectivity over other members of Bcl-2 family.

Fig. 10. (A) Sequence comparison between PUMA and selective peptide FS1 (8), FS2 (9), FS3 (10). Mutations are in red. (B) The R4 residue of the FS2 helix (9) reached into the BH3 hydrophobic cavity and formed a hydrogen bond with the aspartic acid N51 residue.

4. Bfl-1 protein inhibitors
4.1. Peptides

The main strategy for inhibiting anti-apoptotic proteins is mimicking BH3 peptides to compete with them in the binding groove. Antisense oligonucleotides, small molecules and short peptides targeting Bcl-2 members have been reported. Loren D. Walensky and Gregory H. Bird reported that they modiied the structures by installing disulide and lactam bonds on the peptides. The all-hydrocarbon structures (alpha-helix (SAHBs)) were inserted to maintain the helix stability, improve resistance to proteolysis, and enhance cell permeability [106,107]. Currently this strategy has been used in many proteins [108e111]. Here we review the reported peptide Bfl-1 inhibitors mimicking NOXA, PUMA and Bim in the following.

4.1.1. NOXA-mimics

In 2016, Walensky’s team analyzed the crystal structure of the Bfl-1-NOXA complex under oxidative conditions. They found that the NOXA C25 residue was very close to the Bfl-1 C55 residue (Fig. 6A). Then they used Coomassie staining and fluorescence scanning methods to conirm the selective disulide bond between the two residues. By inserting different acrylamide derivatives into the unnatural amino acids of NOXA (L21: 3.3 Å from Bfl-1 C55; PDB code: 3MQP) and Bim (W147: 3.6 Å from Bfl-1 C55; PDB code:(alanine scan) and the positive polarity (mutation of arginine residues to glutamic acid) to remove the ability of the peptide to cleave melanoma cells. In this way, D-NA NOXA SAHB-15 F32A (3, Fig. 7) and D-NA-NOXA SAHB-15 R31E (4, Fig. 7) were screened out. The effect of D-nipecotic acid warhead and the selectivity of two active peptides over Bcl-xL and Mcl-1 were veriied by streptavidin pulldown experiment and FP assay. It is worth mentioning that 3 and 4 are highly speciic, as they could not dissociate the complex between tBid and FLAG-Mcl-1 while the aforementioned cysteinereactive peptide 2 destroyed the tBid/FLAG-Mcl-1 complex through non-covalent bond.

Given that cancer development depends on multiple pathways, Walensky further explored the interdependence of Bfl-1 by genomic-level CRISPR-Cas9 screening. They found that DNA double-strand break-activated ATM serine/threonine kinase was the strongest dependent gene. By cell viability experiment, caspase-3/7 activation and mitochondrial membrane depolarization, they demonstrated the signiicant synergy of 4 and the ATM inhibitor AZD0156. It proved that the combination of the two treatments was a great way to destroy apoptotic resistance in cancer [113].warhead to form helical stapled peptide NOXA SAHBA-3 (1, Fig. 6B) and Bim SAHBA-3 (2, Fig. 6B). It proved that C55 was the key to target Bfl-1 protein through amino acid mutations. And repeated experiments on Mcl-1 and Bcl-xL showed selectivity of the two covalent peptides.

The liposomal release assays proved that stapled NOXA and Bim BH3 peptides prevented the inhibition of Bfl-1 to tBid, which would trigger the Bax-mediated liposome release. The Biotin Western experiment further veriied that 1 and 2 could irreversibly, selectively and covalently bind to Bfl-1 in the mixed protein of 293Ttreated cell lysates. Continuing to explore effects on targeted cancer cells, Walensky found that the uptake mechanism of 2 under the fluorescence microscope was consistent with micropinocytosis. And the two stapled peptides led to nucleus coagulation, membrane blistering, caspases process activation, cytochrome c release and other typical apoptotic cell features in melanoma cell line. In summary, installing cysteine reactive warheads in Th2 immune response NOXA and Bim BH3 peptide can eficiently and covalently target Bfl-1 [112].

In 2018, Walensky’s team truncated the N-terminus and C-terminus of NOXA and performed scanning of the short hydrocarbon ibers at i, i+4, and i+7 position on the speciic sequence (aa 26e40). The N-terminus of each coniguration was equipped with a covalent D-nipecotic acid warhead. They used denaturing gelelectrophoresis experiments and anti-HA western blotting to screen out the active peptide.

4.1.2. Bim-mimcs

David P. Fairlie’s team analyzed the binding mode of natural BH3-only peptide Bim and Bfl-1 (PDB code: 2VM6) in 2017. They found that the Bim L2 residue was close to the Bfl-1 C55 residue. Accordingly, the indole fragment of L2 was replaced with the electrophilic acrylamide to get stapled peptide 5 (Fig. 8). It was intended that this modiied peptide irst non-covalently bound to Bfl-1, and then the electrophilic group covalently bound to the nucleophilic cysteine in the protein. The trypsin digested the monovalent adduct band on the SDS-PAGE gel, and the MS/MS spectrum showed a 1:1 expected fragment, demonstrating that the acrylamide only bound to the C55 residue but not the C4 or C19 residue on the protein surface. As expected, it showed no signiicant inhibitory activity against other anti-apoptotic proteins. The IC50 value of 5 for Bfl-1 protein was 8.5 nM, Ki < 0.1 nM in FP assay. In living cells, flow cytometry demonstrated that cell penetration ability of Bim did not decrease after the addition of the acrylamide. The result of confocal microscopy showed that 5 passed through cells and transferred to mitochondria. Western blot also proved that the peptide bound to endogenous Bfl-1 [114].

In 2018, Fairlie’s team truncated multiple C-terminal and Nterminal residues of Bim, then added a hydrocarbon fragment between residue 9 and 13, and an electrophilic group acrylamide bond at position 2. They synthesized a series of molecules with two helical optimization groups (N-terminal hydrocarbon cross-link to enhance cell permeability and lactam chain to enhance a-helicity). Peptide 6 (Fig. 9) showed the highest afinity (IC50 = 6.6 nM) in FP assay, SDS-PAGE electrophoresis and protein MS/MS analysis and it only covalently bound to the C55 residue of Bfl-1. In addition, by replacing acrylamide with propionamide, and E6 with Leu (which can weaken the interaction with Bcl-2, Bcl-w or Bcl-xL protein), they obtained a selectively noncovalent Bfl-1/Mcl-1 dual inhibitor peptide 7 (Fig. 9). This is a novel dual inhibitor with strong afinity, cell permeability and plasma stability, as well as high selectivity for other Bcl-2 proteins. MTT, IP (anti-Bim immunoprecipitation) and LDH release assay veriied that it speciically bound to Bfl-1 and caused melanoma cell apoptosis [115].

Fig. 13. Inhibitory activities of indole compound BDM-49234 (14) and BDM-53787 (15).

Fig. 14. Inhibitory activities of difuryl-triazine compound 19SR (16) and 2,5-substituted benzoic acid compound 17 and 18.

4.1.3. PUMA-mimcs

As a member of BH3-only peptides, PUMA can also be used as a raw material for structural transformation. However, PUMA shows binding effect on various Bcl-2 family anti-apoptotic proteins with low selectivity. Using computer simulations, Justin M Jenson etal. created a library of candidate peptides including 107 various mutations on PUMA. Through six rounds of flow cytometry screening, three tightly and selectively-binding Bfl-1 peptides were obtained, namely FS1 (8, Fig. 10A), FS2 (9, Fig. 10A) and FS3 (10, Fig. 10A).By observing the eutectic structure, it was analyzed that the mutants bound to Bfl-1 in a different way with the original PUMABfl-1, speciically as follows: 9 moved 1.2 Å in the binding groove and rotated 17。 compared to the PUMA peptide. In addition, 9 destroyed the salt bridge between PUMA D15 and Bfl-1 R88 residue. However, the R4 residue (on the other side of 9) and the Bfl-1 N51 residue formed a hydrogen bond to make up the afinity loss (Fig. 10B). Analysis of the eutectic structure of 9-Bfl-1 also proved that the hydrogen bond was indispensable to the selectivity of Bfl-1 protein.

Fig. 15. Inhibitory activities of 2-imino-N-benzyliminothiazole compound CGDF-18320 (19) and CGDF-190-20 (20).

All three candidate peptides showed signiicant enhancement of mitochondrial outer membrane depolarization and cytochrome c release. “Unprimed” cell lines (requires activation of Bak and Bax) did not trigger cytochrome c release when they were treated with high concentrations of 8 or 9, and released high concentrations of cytochrome c upon 10 treatment. This result indicated that the three peptides did not act by activating Bax and Bak, but by competing with their activators to sensitize apoptosis. Further, by introducing acrylamide groups at the Q1 and W2 positions in 9, covalent inhibitors FS2_1gX and FS2_1fX that generated disulide bonds with C55 residue of Bfl-1 were obtained. Both the two peptides were effective against Bfl-1 at micromolar concentrations, but the mutant protein Bfl-1 C55S did not react with them [116].

Fig. 16. Inhibitory activities of Dihydropyrimidine-4,6(1H, 5H)-dione compound BDM44931 (21) and BDM-44898 (22) patented by WO2016079067A1.

4.2. Small molecules
4.2.1. N-aryl maleimides

In 2010, John C. Reed used high-throughput fluorescence polarization (FP) assay and time-resolved fluorescence resonance energy transfer (TR-FRET) assay to explore Bfl-1 inhibitors [117]. Fourteen compounds showed similar inhibitory activity, and among them compound 11 had the strongest inhibitory activity in submicromolar level (Fig. 11).A series of derivatives were synthesized later. The activity data proved the essential inhibitory effect of chlorine substituents and the double bonds at the maleimide ring. Substituents on the Nphenyl ring and the amine on the maleimide ring system also influenced the inhibitory activity. The 3,4-dichlorophenyl moiety of compound 11 was the best substitution of the N-aryl ring.

4.2.2. Sulfonylpyrimidines

In 2011, Reed’s team also used the binding effect of Bfl-1 protein and fluorescently labeled Bid BH3 peptide to screen out active molecules through FP assay. Excluding the effects of autofluorescence and using TR-FRET assay to conirm, they identiied three structures with desired inhibitory activity, which were chloromaleimide, sulfonylpyrimidine and quinolinyl-oxadiazoles. Among them, sulfonylpyrimidine series showed the IC50 as low as 1.5 mM against Bfl-1 and almost no activity on Bcl-w, Mcl-1 or Bcl-2. Later, Reed analyzed the structure-activity relationship of the sulfonylpyrimidine series. The results showed that it was feasible to replace the benzene ring group with heterocycle. When the sulfonyl side chain was truncated to methyl sulfone group or the length of the side chain was extended, the activity slightly reduced. Among them, the most active one was 4-fluorophenyl derivative CID2980973 (13, Fig. 12), whose IC50 was half of the HTS selected compound CID-1151859 (12, Fig. 12).Sulfonylpyrimidine 12 directly promoted SMAC to release from mitochondria, but not played the role when Bid stimulated the mitochondria. It suggested that sulfonymidine compounds may have biological activity independent of Bfl-1 protein [118].

Fig. 17. Structure of covalently crosslinked molecule 4E14 (23). It interacts with Bfl-1 protein through disulide fragment.

4.2.3. Indoles

Anne-Laure Mathieu et al. also used FP assay and TR-FRET analysis to screen out Bfl-1 inhibitor, but replaced the FITC-Bid BH3 peptide with FITC-Bim BH3 peptide [119]. Among the selected three types of molecules (indole, phenol, and benzimidazole), indole compounds BDM-49234 (14, Fig.13) and BDM-53787 (15, Fig.13) showed the best inhibitory activity (with IC50of 6.3 and 0.8 mM respectively in FPassay). Meanwhile,14 and 15 exerted their binding effect in TR-FRET and thermoshift assay (TSA), and inhibition in B-cell lymphoma cell line BP3 and B-lymphoblastoid cell line IM9. The two molecules had synergistic effect with Bcl-2/Bcl-xL inhibitor ABT-737. Changing the fluorescent polypeptide and the substrate protein, both molecules showed obvious selectivity over other anti-apoptotic members.

Since the structure provided a potential electrophilic Mannich base, which may react with the cysteine of Bfl-1, the reactivity of the two molecules with glutathione was detected by LC-MS analysis in vitro. The results showed that 14 may produce a less reversible crosslinking reaction with cysteine. Through FP assays, it was found that 14 but not 15 bound to Bfl-1 protein in timedependent manner.

Given that 14 bound to Bfl-1 for a short time and only the reactive group linked to the binding groove, molecular modeling of 15 was carried out to speculate its binding mode. In this pattern, three fragments of the ligand occupied different areas of the Bfl-1′s binding groove, which were consistent with the Bim hot spot I148, L152, and R153 residue respectively. The phenyl group of 15 occupied the hydrophobic pocket of Bim L152 and the 4-methoxyphenyl group partially occupied the hydrophobic pocket of Bim I148. The piperazine group may partially occupy the binding pocket of Bim R153, which can form a salt bridge to Bfl-1 E80 residue.The majority of the Bcl-2 family inhibitors have relatively large molecular weight or lipophilicity which may affect their drug function. However, the physicochemical properties of 14 and 15 are acceptable (MW~400 g/mol and log P~4) [119].

4.2.4. 2,5-Substituted benzoic acid

According to the structural characteristics and crystal structure analysis of the currently reported inhibitors, Kump et al. replaced the difuryl-triazine core of the Bfl-1/Mcl-1 dual inhibitor 19SR (16, Fig. 14) with 2,5-substituted aromatic benzoic acid to obtain compound 17 (Fig. 14). FP assay showed that the activity can be increased by 15 times. Then Kump synthesized a number of derivatives under the guidance of the binding information provided by HSQC-NMR and the co-crystal structure. Among them, the activity of compound 18 (Fig. 14) on Bfl-1 and Mcl-1 was further improved by nearly 20 times relative to compound 17 (Mcl-1 Ki = 0.094 ± 0.01 mM and Bfl-1 Ki = 0.1 ± 0.02 mM). Biolayer interferometry (BLI) experiment veriied the binding of compound 18 to the two protein. Meanwhile, compound 18 showed outstanding selectivity over other anti-apoptotic proteins. Pulldown experiment showed that 18 destroyed the binding effect between NOXA peptide and Mcl-1/Bfl-1, and flow cytometry showed that compound 18 dose-dependently induced the death of Eμ-Myc cell lines in which Mcl-1/Bfl-1 overexpressed.

Fig. 18. Structures of Gossypol (24) and Apogossypol (25), and inhibitory activities of Apogossypol derivative BI-97C1 (26) and ND646 mouse BI-97D6 (27).

Fig. 19. Inhibitory activity of natural product Gambogic acid (28).

On account of the crystal structure of Mcl-1 and similarity between Mcl-1 and Bfl-1′s cavity, Kump speculated that the carboxyl group of Microbial mediated compound 18 formed a hydrogen bond with the R88 residue,thereby anchored the molecule to the P2 pocket. The tertbutyl group was in contact with V122, A67, I118, F121, and L99 residue, while the benzene ring was in contact with F71, I92 and F95 residue. The phenethyl sulide group pointed to the a4 helix in the binding pocket, resulting in the hydrophobic reaction between phenyl group and methylene group on the K77 side chain. The shifts of key residues in HSQC-NMR conirmed these speculations [120].

4.2.5. 2-Imino-N-benzyl iminothiazole

Walensky reported in patent that they screened out a selective Mcl-1 inhibitor through high-throughput FP assay, and accidently obtained a class of Bfl-1 selective inhibitors in process of structural modiication. Their structures all contain 2-imino-N-phenyl iminothiazole group and the benzenetriol group, of which CGDF-18320 (19, Fig. 15) and CGDF-193-20 (20, Fig. 15) showed the strongest inhibitory activity and selectivity (IC50 was 264 nM and 365 nM respectively) [121,122].

4.2.6. Dihydropyrimidine-4,6(1H,5H)-dione

The inventor Leroux et al. used ligand screening through similarity and substructure search to get potential molecules BDM44931 (21, Fig. 16) and BDM-44898 (22, Fig. 16). 22 has been reported as a small molecule targeting Acyl homoserine lactone synthase [123]. The two compounds bound to Bfl-1 protein (IC50 was 1.3 μM and 1.6 μM respectively) with high selectivity against other anti-apoptotic proteins of Bcl-2 family. They exerted proapoptotic effect on BP3 and IM9 cell line in propidium iodide/ Annexin V double staining and flow cytometry, as well as delayed the growth of lymphoid tumor in immunocompromised mice [124].

4.2.7. Covalently crosslinked molecule

After developing covalent peptides targeting the Bfl-1 protein, the Walensky’s team focused on small molecules playing the same role [125]. They used intact mass spectrometry and FP assay to screen in a library consisting of 1600 compounds with cysteine reactivity and found a disulide-containing N-acetyltryptophan analog 4E14 (23, Fig. 17), which could selectively target Bfl-1 C55 residue (the EC50ofFITC-Bid binding to Bfl-1 changed from 23 nM to 1.3 μM, a 56-fold decrease) and hinder the interaction with BH3 to promote Bax-mediated cytochrome c release. Hydrogen/deuterium exchange mass spectrometry (HXMS) analyses and the eutectic structure showed that when Bfl-1 bound to 23, the a2 helix was in the state between bound and unbound to BH3. It was proposed that 23 irst reconstructed the conformation of Bfl-1 through non-covalent binding. After a2 being exposed to open the groove, 23 formed a disulide bond with the C55 residue to enhance the molecular interaction in the cavity. The docking study predicted that 23 would interact with the hydrophobic V48, L52, V74,and F95 residue in Bfl-1 cavity, and the indole hydrogen would form a hydrogen bond with E78 residue.

4.2.8. Pan-active inhibitor

Some pan-inhibitors against the Bcl-2 family also have good activity against Bfl-1, most of which are natural products.

4.2.8.1. Apogossypol derivatives. Reed’s team modiied the natural product Gossypol (24, Fig. 18) and removed two reactive aldehyde groups to eliminate off-target effects. And then, Apogossypol (25, Fig. 18), which inhibited the Bcl-2 family, was obtained. By performing 5,50 -amide and ketone substitution, they obtained an optically pure compound BI-97C1 (26, Fig. 18). 26 effectively obstructed the growth of human prostate cancer, lung cancer and lymphoma cell lines, and exhibited excellent single-agent anti-cancer eficacy in a prostate cancer mouse xenograft model. Its IC50 values for four anti-apoptotic proteins are shown in the igure below [126].One year later, Reed’s team reported a stronger inhibitor: BI97D6 (27, Fig. 18), which could also induce apoptosis of prostate cancer PC-3 cells and non-small cell lung cancer H23 cells. In addition, it showed anti-inflammatory effects in xenogeneic prostate cancer mouse models [127].

4.2.8.2. Gambogic acid. Gambogic acid (28, Fig. 19), a cage-like natural product extracted from Garcinia cambogia, has been reported by Reed to exert pro-apoptotic effect in a variety of cancer cells by inhibiting Bcl-2 family proteins and activating caspases. It showed signiicant inhibitory activity on six Bcl-2 anti-apoptotic proteins in FP assay (IC50 values are shown in the igure below). TRFRET experiments veriied its competition with BH3-only Bad peptide to Bcl-xL. The results further showed that 28 promoted the release of SMAC factor from mitochondria in acute lymphoblastic leukemia Jurkat T-cell line and acute promyelomonocytic leukemia HL-60 cell line, and facilitated hydrolysis of procaspase-3 protein. Knockout of BAX and BAK gene proved that 28 induced cytotoxicity depending on Bcl-2 family proteins [128]. The bioactivities of representative Bfl-1 small molecule inhibitors are described in Table 1.

5. Conclusion

BCL2A1/Bfl-1 is an undeveloped member of the Bcl-2 family. The narrow and shallow hydrophobic cavity and the diversiication of murine gene variants make it dificult to be targeted. Similarity between the binding cavity of Bfl-1 and Mcl-1 is also a big hurdle to achieve selectivity. At present, most reported short peptides covalently link to C55 residue, which is speciic in the cavity of Bfl-1. A few small molecules selected by high-throughput screening exerted inhibitory activity in nanomolar level. Highly potent and selective Bfl-1 inhibitors have not been produced yet. We believe that there is still a great possibility to develop potential Bfl-1 inhibitors and make progress in the treatment of gene-speciic disease such as diffuse large B-cell lymphoma in the future.