Targeting cancer addiction for SALL4 by shifting its transcriptome with a pharmacologic peptide

Significance Hepatocellular carcinoma (HCC) is leading cause of death due to late discovery and lack of effective treatment. The finding of the Sal-like 4 (SALL4)–NuRD interaction in HCC opens a new therapeutic direction targeting an epigenetic regulator. Here, we identified the SALL4–NuRD binding site through structural resolution of the crystal complex, providing valuable insight for the development of antagonists against this interaction. Our subsequent design of a therapeutic peptide has demonstrated the possibility to develop a first-in-class drug targeting the SALL4–NuRD interaction in HCC. Furthermore, we discovered that the therapeutic peptide exhibits robust antitumor properties and works by inhibiting the repressive function of SALL4. Our work could also be beneficial to a broad range of solid cancers and leukemic malignancies with elevated SALL4. Sal-like 4 (SALL4) is a nuclear factor central to the maintenance of stem cell pluripotency and is a key component in hepatocellular carcinoma, a malignancy with no effective treatment. In cancer cells, SALL4 associates with nucleosome remodeling deacetylase (NuRD) to silence tumor-suppressor genes, such as PTEN. Here, we determined the crystal structure of an amino-terminal peptide of SALL4(1–12) complexed to RBBp4, the chaperone subunit of NuRD, at 2.7 Å, and subsequent design of a potent therapeutic SALL4 peptide (FFW) capable of antagonizing the SALL4–NURD interaction using systematic truncation and amino acid substitution studies. FFW peptide disruption of the SALL4–NuRD complex resulted in unidirectional up-regulation of transcripts, turning SALL4 from a dual transcription repressor-activator mode to singular transcription activator mode. We demonstrate that FFW has a target affinity of 23 nM, and displays significant antitumor effects, inhibiting tumor growth by 85% in xenograft mouse models. Using transcriptome and survival analysis, we discovered that the peptide inhibits the transcription-repressor function of SALL4 and causes massive up-regulation of transcripts that are beneficial to patient survival. This study supports the SALL4–NuRD complex as a drug target and FFW as a viable drug candidate, showcasing an effective strategy to accurately target oncogenes previously considered undruggable.

Sal-like 4 (SALL4) is a nuclear factor central to the maintenance of stem cell pluripotency and is a key component in hepatocellular carcinoma, a malignancy with no effective treatment. In cancer cells, SALL4 associates with nucleosome remodeling deacetylase (NuRD) to silence tumor-suppressor genes, such as PTEN. Here, we determined the crystal structure of an amino-terminal peptide of SALL4 (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) complexed to RBBp4, the chaperone subunit of NuRD, at 2.7 Å, and subsequent design of a potent therapeutic SALL4 peptide (FFW) capable of antagonizing the SALL4-NURD interaction using systematic truncation and amino acid substitution studies. FFW peptide disruption of the SALL4-NuRD complex resulted in unidirectional up-regulation of transcripts, turning SALL4 from a dual transcription repressor-activator mode to singular transcription activator mode. We demonstrate that FFW has a target affinity of 23 nM, and displays significant antitumor effects, inhibiting tumor growth by 85% in xenograft mouse models. Using transcriptome and survival analysis, we discovered that the peptide inhibits the transcription-repressor function of SALL4 and causes massive upregulation of transcripts that are beneficial to patient survival. This study supports the SALL4-NuRD complex as a drug target and FFW as a viable drug candidate, showcasing an effective strategy to accurately target oncogenes previously considered undruggable.
SALL4 | RBBp4/NuRD | peptidomimetic | HCC | structural guided design S al-like 4 (SALL4) is a nuclear factor central to the maintenance of stem cell pluripotency (1)(2)(3), specifically expressed in fetal cells. SALL4 is down-regulated or absent in most adult tissues but is reactivated in solid (4-7) and hematological malignancies (8)(9)(10)(11), often associated with poor prognosis (12)(13)(14). In embryonic stem cells, SALL4 forms a core transcriptional network with Oct4, Nanog, and Sox2 to drive embryonic stem cell (ESC) self-renewal (1)(2)(3). It was also observed that SALL4 plays a dual role in transcription activity in ESCs, activating self-renewal and repressing differentiation related transcripts at the same time (2,15). Recently, we and others found that activation of SALL4 is a key driver in up to 55% of hepatocellular carcinoma (HCC) patients (14,16,17). It also serves as a biomarker for progenitor-like HCC (14,16), an aggressive subset of HCC characterized by stem cell features, invasive potential, and poor patient prognosis (17). In HCC, we reported that the phosphatase and tensin homolog deleted on chromosome 10 (PTEN)-AKT-PI3K pathway is altered by the reexpression of SALL4 (16). Furthermore, we have also reported that SALL4 acts as a repressor of PTEN by recruiting the nucleosome remodeling deacetylase (NuRD) complex (18). NuRD is a chromatin remodeling complex that silences key regulators in ESCs and adult cells (19,20); it has two independent enzymatic activities whereby nucleosomes are repositioned by the CDH3/4 ATPase subunits in order for the histone deacetylase (HDAC1/2) subunits to access their targets and repress target genes. The retinoblastoma binding protein 4 (RBBp4) is a subunit of NuRD. It is a WD40 repeat-containing protein, which consists of a seven-bladed β-propeller domain. In NuRD, RBBp4 acts as a chaperone in nucleosome assembly by bringing together histones H3 and H4 onto newly replicated DNA (21).
HCC is a fast-growing malignancy with a median survival of 11 mo. With less than 30% of patients able to receive potentially curative treatments like surgery and liver transplant, it is now the second leading cause of death worldwide (globocan.iarc.fr/) due to the lack of effective treatment options for the majority of HCC patients. The multikinase inhibitor Sorafenib is currently the first-line treatment available for patients with advanced HCC. However, this drug shows frequent adverse effects and only prolongs survival for an average of 3 mo (22). Promising new drugs that have reached clinical trials so far have met with failure. Therefore, there is an urgent need to explore alternative approaches for the treatment of this deadly disease. SALL4, Significance Hepatocellular carcinoma (HCC) is leading cause of death due to late discovery and lack of effective treatment. The finding of the Sal-like 4 (SALL4)-NuRD interaction in HCC opens a new therapeutic direction targeting an epigenetic regulator. Here, we identified the SALL4-NuRD binding site through structural resolution of the crystal complex, providing valuable insight for the development of antagonists against this interaction. Our subsequent design of a therapeutic peptide has demonstrated the possibility to develop a first-in-class drug targeting the SALL4-NuRD interaction in HCC. Furthermore, we discovered that the therapeutic peptide exhibits robust antitumor properties and works by inhibiting the repressive function of SALL4. Our work could also be beneficial to a broad range of solid cancers and leukemic malignancies with elevated SALL4. however, falls into the class of what is termed as "undruggable" targets, as a nuclear factor lacking a typical, druggable pocket for inhibitor binding. In this report, we discovered that the SALL4-NuRD interaction offers an intriguing potential therapeutic target, as SALL4 is expressed uniquely in a selected population of cancer cells with poor prognosis, and not in normal adult cells. Furthermore, we demonstrated that by targeting SALL4-NuRD, we are able to reverse the repression function of SALL4 in tumorsuppressor transcripts, switching SALL4 from a dual-mode transcription factor to single mode lacking transcriptional repression function. To that end, we resolved the RBBp4-SALL4(1-12) complex, determined key residues involved in the SALL4-RBBp4 binding, and designed a pharmacologic peptide (RRKFAKFQWI, named FFW hereafter) that blocked the interaction with high affinity, and reversed the repression function of SALL4 on target genes. The massive up-regulation of treatment-specific transcripts resulted in apoptosis, activation of tumor suppressors, and cell adhesion molecules, and subsequently showed a significant effect on inhibiting xenograft formation in mice. In addition, this report also highlights a viable and effective strategy incorporating structural analyses to accurately design a therapeutic approach against candidate genes that are otherwise considered undruggable.
We next determined the crystal structure of the RBBp4-SALL4(1-12 aa) complex to obtain structural information of the SALL4-RBBp4 protein-protein interaction site. The complex was solved at 2.7-Å resolution (SI Appendix, Table S1). RBBp4 forms a seven-sheet β-propeller (residue  with an N-terminal α-helix (Fig. 1C). All 12 residues of the SALL4 peptide are well defined in the electron density map, with 9 of the residues making favorable interactions with RBBp4. The substrate binding site of RBBp4 is highly acidic, with eight glutamic acid and two aspartic acid residues within 5 Å of the SALL4 peptide. This negatively charged interface binds the predominantly positively charged SALL4 peptide, which has five basic residues: Arg3, Arg4, Lys5, Lys8, and His11 ( Fig. 1 D and E). Arg3 and Lys5 of SALL4 form charged interactions with Glu275, Glu319, and with Glu126, Glu179, respectively, in RBBp4 ( Fig. 1F and SI Appendix, Table S2). Arg4 forms a salt bridge with Glu231, whereas His11 makes π-cation interactions with Trp42. Several unique hydrogen bonding contacts were observed between Arg3:Lys376, Arg4:Phe321, Arg4:Arg129, and Pro9:Ser73 of the RBBp4-SALL4 complex ( Fig. 1F and SI Appendix, Table S2). Additional hydrogen bonds between residues Ser2 to Gln6 and Gln10 to Ile12 stabilize the SALL4 peptide, and several hydrophobic interactions stabilize the complex. The Arg4 side chain is deeply buried into RBBp4 (buried surface area 211 Å 2 ), whereas Lys5 and Pro9 bind in shallow grooves.  and shallow acidic interaction surface, suitable for developing a peptide inhibitor to block the interaction. In this regard, we undertook an integrated approach by combining computational studies with biophysical assays to determine the important amino acids for the SALL4-RBBp4 interaction. The structural analysis of the complex indicated that Arg3, Arg4, and Lys5 of SALL4 are crucial for binding. We confirmed this structural hypothesis using computational alanine scanning, and found that alanine substitutions of Arg3, Arg4, and Lys5 greatly affected the binding free energy (19,16, and 14 kcal/mol, respectively) ( Fig. 2A). Based on the in silico analysis, we subsequently performed a biochemical alanine scan of these residues with a series of mutant peptides using a fluorescence polarization assay ( Fig. 2 B and C and SI Appendix,  Fig. S1). These findings confirmed that Arg3, Arg4, and Lys5 are essential residues involved in the SALL4-RBBp4 interaction. We next selected a representative double mutant of essential residues (MSARAQAKPQHI; MUT R3A,K5A) and compared its effect with the WT peptide on cell viability in SNU398 HCC cells, which express high levels of SALL4 RNA and protein (16). While the WT peptide exerted an inhibitory effect on cell number, the MUT peptide did not (Fig. 2D). Further qPCR analysis revealed a threefold increase in PTEN expression (P < 0.0001) in cells treated with the WT peptide compared with untreated cells, whereas MUT-treated cells again showed no significant change (P value not significant) (Fig. 2E). These findings indicate that the MUT peptide could not block the RBBp4-SALL4 interaction and failed to release the suppressive complex from the PTEN promoter, unlike its WT counterpart.
Optimization of a Candidate Therapeutic Peptide. After determining the key interactions between RBBp4 and SALL4 (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12), and by using structural data of the RBBp4-SALL4(1-12) complex as a platform, we then undertook a peptide substrate-based approach to design and optimize a potent inhibitor of the SALL4-RBBp4 interaction. We first determined the minimum length required for bioactivity through a truncation analysis of the WT peptide (SI Appendix, Table S3, peptide 1). Removing the first two Nterminal residues, Met and Ser (SI Appendix, Table S3, peptides 2 and 3), increased the peptide binding affinity to RBBp4 compared with the WT (IC 50 = 0.60 and 0.36 vs. 1.0 μM, respectively); yet C-terminal truncations resulted in a marginal loss of binding affinity (SI Appendix, Table S3) (IC 50 = 1.29, 0.80, and 1.91 μM for peptides 5, 6, and 7, respectively). Removal of a key binding residue (peptide 4) abolished the binding affinity (IC 50 > 20 μM). To further improve binding potency, peptide 3 was selected as the sequence template and subjected to a systematic single-residue mutation analysis with alanine substitutions. Substituting the nonessential residues of peptide 3 with Ala (SI Appendix, Table S3, peptides 11-16) yielded more potent peptides, whereas substituting the essential residues of peptide 3 with Ala abolished binding (SI Appendix, Table S3, peptides [8][9][10]. This suggests that nonessential residues of peptide 3 could be replaced with other amino acid residues for sequence optimization. Systematic substitution of the nonessential residues (SI Appendix, Table S3, peptide 20-45) revealed a marked reduction of IC 50 upon replacement of Gln-4, Pro-7, and His-9 of peptide 3. Gln4 sits in a small binding pocket formed by Pro43, His71, and Glu395 of RBBp4, which is able to accommodate amino acid  residues with hydrophobic side chains. Consequently, substituting Gln4 with Leu or Phe (peptides 23 and 24) led to improved binding affinities, particularly Phe, which induced a sevenfold enhancement in binding affinity over that of peptide 3 (IC 50 0.05 vs. 0.36 μM) (SI Appendix, Table S3). Additionally, a Phe substitution (peptide 34) for Pro7 further increased the potency of the peptide (IC 50 0.17 vs. 0.36 μM) (SI Appendix, Table S3). Using computer modeling, the aromatic ring of Phe7 was found to be solventexposed and not involved in RBBp4 binding. However, its backbone amide proton could be involved in H-bonding to the sidechain carboxyl moiety of Glu395, an interaction lacking in Pro7 (SI Appendix, Fig. S2). In an effort to enhance the π-π interaction to RBBp4, His9 was replaced by aromatic residues Trp (peptide 41), and the substitution improved the binding affinity by threefold compared with that of peptide 3 (IC 50 0.12 vs. 0.36 μM) (SI Appendix, Table S3). These three substitutions (Gln4Phe, Pro7Phe, and His9Trp) were incorporated into peptide 46 (RRKFAKFQWI, named FFW hereafter). Fluorescence polarization assay confirmed the high potency of FFW, with a >43-fold increase in affinity compared with the original 12-residue WT peptide (IC 50 = 0.023 vs. Fig. S3 and Table S1).
FFW Reverse-Transcription Repression by SALL4-RBBp4. To elucidate the regulatory pathways that FFW could affect, a penetratin sequence (PEN) was attached to the N terminus of FFW to facilitate penetration into cells in culture (Fig. 3A). The FFW peptide is a highly basic and hydrophobic molecule. To test the ability of FFW to penetrate cells without the penetratin sequence, we treated both peptides to SNU398 cells and MTT assays were conducted after 72 h of incubation. The results demonstrated that although FFW is hydrophobic, the FFW peptide did not show any effect on cell viability without the conjugation of the penetratin sequence (Fig. 3B). We have also conjugated FITC to the amino terminus of FFW and subjected it to fluorescence microscopy, but could not detect any uptake of the peptide by the cells (SI Appendix, Fig. S4).
In addition to PEN-FFW, the PEN sequence was also conjugated to WT and MUT peptides. To test if the PEN-FFW peptide could disrupt endogenous SALL4-NuRD interaction, we performed coimmunoprecipitation after peptide treatment using SALL4 antibody (Fig. 3C). We found that the endogenous SALL4-RBBp4/HDAC/NuRD interaction was abrogated after PEN-FFW treatment, but not the controls. We next sequenced RNA transcripts (RNA-seq) from SNU398 cells treated with PEN alone, and different PEN conjugated peptides (30 μM, 8 h). Unbiased hierarchical clustering of the whole transcriptome data revealed that the PEN-FFW-treated sample was distinctly clustered away from the rest of the samples (Fig. 3D), whereas PENand PEN-MUT-treated samples (controls) were closely clustered together with similar transcriptome profiling, and the PEN-WT-treated sample was clustered in between. Strikingly, heatmap presentation of differentially expressed genes (DEGs) demonstrated that PEN-FFW treatment induced almost exclusively gene activation, in which 99.5% of the PEN-FFW DEGs were up-regulated (575 transcripts) and only 0.5% of the DEGs (three transcripts) were down-regulated ( Fig. 3E and Dataset S1). This indicate that the PEN-FFW treatment had reversed the repressive function of SALL4, switching it from a dual-action transcription factor to a single-activator mode. This observation was further confirmed by M-A plot (log fold-change versus log mean intensity) of PEN-FFW versus controls (Fig. 3F), in which a majority of the DEGs are up-regulated. Next, gene set enrichment analysis (GSEA) was performed using the whole transcriptome between PEN-FFW or PEN-WT and the two negative controls (PEN and PEN-MUT), respectively. GSEA analysis revealed a significant concordance between genes upregulated in cells treated with PEN-FFW and genes downregulated by overexpressing oncogenes like KRAS and ALK ( Fig.  3G and Datasets S2 and S3). Genes down-regulated in HCC, for example ATP1A2 and LCN8, were also specifically up-regulated in PEN-FFW-treated samples, while further analysis showed that targets of HNF4a, a key liver transcription factor, were also enriched in PEN-FFW-treated cells. Interestingly, PEN-FFWenriched genes were associated with the repressive H3K27Me3 mark before peptide treatment, indicating these transcripts had undergone epigenetic modulation following disruption of the SALL4-NuRD interaction. (Fig. 3G and Dataset S3).

SALL4-RBBp4 Disruption Induced Apoptosis and Enhanced Cell Adhesion.
Next, we conducted chromatin immunoprecipitation coupled with next-generation sequencing (ChIP-seq) on SNU398 cells to detect the SALL4 binding regions in the genome to compare with the PEN-FFW DEGs. Of the 7,883 SALL4 binding peaks, 60% are located at promoter regions, with H3K27Ac marks enriched around the SALL4 peaks (Fig. 4A). Interestingly, 57 of the 575 PEN-FFW DEGs have strong SALL4 binding peaks within 2 kb of the transcription start sites, suggesting that these transcripts are direct targets of the SALL4-RBBp4/NuRD interaction (Fig. 4B). Furthermore, we observed that chromatin remodelers and transcripts encoding molecules that facilitate transcription activation are among the highest expressed DEGs of these 57 transcripts (Fig. 4C). These include: POLR2A, the largest subunit of POLII; pre-mRNA splicing factor SRRM2; and chromatin remodeling protein SRCAP. Interestingly, both of the H3K4-specific methytransferases, MLL2 and SETD1B, also demonstrate promoter occupancy by SALL4 (Fig. 4D) and up-regulation after SALL4-RBBp4/NuRD disruption, implying that the SALL4-RBBp4/ NuRD complex strongly regulates chromatin status.
A significant proportion of the PEN-FFW DEGs are miRNA and noncoding RNAs (23% of the transcriptome). We focused on annotated transcripts and performed pathway analysis for these transcripts. Intracellular Ca 2+ concentration is tightly regulated in cells. An unexpected finding of the pathway analysis revealed that a group of transcripts involved in calcium signaling, such as calmodulin (CALML6) and GRIN1, are up-regulated (Fig.  4E), suggesting that SALL4-RBBp4 disruption might induce a Ca 2+ influx in PEN-FFW-treated cells. Because Ca 2+ could act as a secondary messenger to activate apoptosis (23), we detected transcripts encoding molecules involved in the apoptosis pathway, such as CARD14, RAPSN, CHRNG, and BCL2L14, in the PEN-FFW DEGs. To further investigate the effect of these gene changes, three liver cancer cell lines with high expression of SALL4, SNU398, Huh7, and Hep3B (16) were treated with different peptides for 6 h (SI Appendix, Fig. S5) or 24 h (Fig. 4F). The cells were stained with propidium iodide (PI) and anti-Annexin V antibody to assess the apoptotic cell population by flow cytometry. After 24 h, PEN-FFW treatment significantly increased the apoptotic population (PI + /Annexin V + , Annexin V + ) from 16.6% in PEN control to 32.5% after PEN-FFW treatment in SNU398 (Fig. 4 F, i); from 14.3% in PEN control to 73.5% in PEN-FFW treated Huh7 cells (Fig. 4 F, ii); and from 13.5% in PEN control to 51.7% in PEN-FFW treated Hep3B cells (Fig. 4 F, iii), supporting previous data from transcriptome analysis.
Loss of cell-cell or cell-extracellular matrix (ECM) contact is often observed in cancer as transformed cells gain motility and invasiveness. In contrast, gain of cell-cell and cell-ECM contact could inhibit cancer cell growth and prevent metastasis. In our transcriptome analysis, the most distinct cellular pathway associated with PEN-FFW DEGs were transcripts encoding molecules involved in cell-cell adhesion and ECM interactions, including CDH4, Claudin-5, -18, and -20, collagen (COL4A4, COL5A1), integrin (ITGA10), laminin (LAMA5), and vitronectin (VTN). Furthermore, a group of cadherins were also upregulated (CDH4, FAT2, PCDH12, PCDHAG12) after PEN-FFW treatment. We therefore hypothesize that the disruption of SALL4-RBBp4 could tighten cell-cell contacts and limit mobility of these cells. To examine this hypothesis, we performed a cell-invasion assay in Boyden chambers in which serum-starved SNU398, Huh7, or Hep3B cells were treated with different peptides (10 μM). Cells migrating toward 10% serum and invading the membrane pores after 24 h were stained with DAPI, and quantified by counting cells per microscope field-of-view. Compared with PEN-and PEN-MUT-treated cells, cell migration ability was indeed markedly impaired in PEN-FFW-treated cells. For both SNU398 and Huh7, more than an 85% reduction of migrated cells was observed in the PEN-FFW-treated group, compared with PEN ( Fig. 4G) (P < 0.001 and P < 0.0001, respectively). For Hep3B, the migration activity is low compared with the other two cell lines, in which only 44 cells were found to have passed the membrane after PEN treatment, and this number dwindled to 14 cells after PEN-FFW treatment. Wound-healing assays were also performed to assess the migration activity of PEN-FFW-treated cells. We found that PEN-FFW-treated cells had slower migration activity compared with PEN-MUT and PEN controls in SNU398 (Fig. 4H), Huh7, and Hep3B cells (SI Appendix, Fig. S6).
Beside the pathways mentioned previously, a class of tumor suppressor genes was also restored after PEN-FFW treatment (Fig. 4E). TSC1/2, which inhibits mTORC1 (24), was enriched 2.5-fold after PEN-FFW treatment (Dataset S1). Another example, LRRC4, which functions to delay cell cycle progression (25), was increased after PEN-FFW treatment. A negative regulator of RAS, DOK3, was also found up-regulated in the PEN-FFW treated cells. Taken together, our data suggest that the disruption of the SALL4-RBBp4 interaction by the PEN-FFW peptide could lead to up-regulation of cell adhesion molecules to suppress migration, Ca 2+ influx to induce apoptosis, and a group of tumor suppressors that oppose various oncogenic processes.
Prognostic Value of PEN-FFW Up-Regulated Genes. Because PEN-FFW treatment leads to significant apoptosis of SNU398 liver cancer cells (Fig. 4F) and dramatic inhibition of xenograft tumor growth (as presented below; see also Fig. 6), we were interested in evaluating the potential prognostic value of these PEN-FFW up-regulated DEGs in patients. Thus, we first overlapped these up-regulated genes with genes that are significantly up-regulated in normal liver samples compared with liver cancer patients in three independent cohorts (Fig. 5A, schematic representative of the analysis). Across the three cohorts, 26 of these PEN-FFW up-regulated DEGs show consistent overexpression in adjacent nontumor samples compared with the tumor (Fig. 5B and Dataset S4). We next explored their diagnostic abilities in separating HCC samples from normal tissues using the receiver operator curve (ROC), and were able to identify nine of them with consistently high area under curve across all three cohorts (Dataset S5). Three examples of these genes (IGFALS, GNAO1, and ECM1) are depicted in Fig. 5C.
Concomitantly, to further evaluate the prognostic relevance of these 26 PEN-FFW up-regulated DEGs, we examined their survival differences in two independent HCC cohorts. We were able to identify eight genes with favorable survival difference for the patients with high expression compared with those with low expression (Fig. 5A and Dataset S6). By overlapping the two gene sets from ROC and survival analysis, we were able to identify four PEN-FFW up-regulated DEGs having both high diagnostic and high prognostic values, and they are IGFALS, SLC22A1, ASPG, and FTCD. The favorable survival difference from these four genes is depicted in Fig. 5D, using two independent cohorts. The prognostic value and diagnostic value of these four genes are highlighted in bold in Datasets S5 and S6.

Therapeutic Peptide FFW Leads to Tumor Regression in Xenograft
Models. Having developed the highly potent FFW peptide affecting SALL4 transcription, we next compared its efficacy to WT and mutant peptides at the cellular level. SNU398 cells were treated with PEN-MUT, PEN-WT, and PEN-FFW for 72 h with a series of concentrations, and cell viability was measured with the MTT assay. We observed that PEN-FFW conferred a 4-fold improvement in reducing cell viability compared with the PEN-WT (Fig. 6A) (EC 50 7.6 μM vs. 30 μM, respectively), and more than 13-fold compared with PEN-MUT (EC 50 > 100 μM). To further confirm the specific therapeutic effect of PEN-FFW in SALL4-expressing HCC cells, we employed SNU398, Huh7, and Hep3B, three HCC lines with high levels of SALL4 (16); SNU387, a HCC line with undetectable levels of SALL4; and THLE-2 and THLE-3, immortalized normal liver epithelial cell lines with undetectable levels of SALL4, and treated these cells with PEN-MUT, PEN-WT, or PEN-FFW peptides (SI Appendix, Fig. S7). Our data demonstrate the specificity of PEN-FFW in targeting SALL4-high HCC cells (SNU398, Huh7, and Hep3B), and not SALL4-low HCC cells (SNU387) or immortalized normal liver epithelial cell lines THLE-2 and THLE-3, highlighting a clear therapeutic window for treatment.
To further test the therapeutic effect of PEN-FFW in vivo, SNU398 cells were implanted subcutaneously into the flanks of NOD/SCID/γ-mice (NSG), and the mice were randomly grouped for peptide treatments (n = 5 per group) (Fig. 6B). For mice treated with PEN (control) or PEN-MUT, tumors progressively increased in size, showing that neither peptide was able to inhibit tumor growth (Fig. 6C). In contrast, although PEN-WT impaired tumor growth (P = 0.001), PEN-FFW induced a much stronger therapeutic effect (P = 0.0008) with a tumor growth inhibition of   85%. PEN-FFW-treated mice also displayed the smallest tumors (Fig. 6D), with significantly lower tumor weight (μ = 88 mg in PEN-FFW, μ = 564 mg in PEN-WT, and μ = 1,550 mg in PEN control) (Fig. 6E). To access toxicity, we tabulated the body weight of the mice at end point (SI Appendix, Fig. S8) and also traced the body weight change throughout the experiment, and found no significant difference, suggesting the overall well-being of the mice receiving the peptide treatments.
To benchmark against current therapy used in advanced stage HCC, we compared PEN-FFW-treated SNU398 tumor xenografts to that of Sorafenib-treated xenografts in a separate set of experiments (SI Appendix, Fig. S9) (n = 6 per treatment group). Interestingly, PEN-FFW treatment resulted in stronger antitumor activity than Sorafenib compared with vehicle-treated groups, albeit not statistically significant. Furthermore, mice treated with PEN-FFW in combination with Sorafenib showed the slowest rate of tumor growth. These data suggest a potential therapeutic effect of PEN-FFW in treating HCC patients, including those refractory to Sorafenib treatment. To further test this hypothesis, we developed a second xenograft model with a chemo-resistant HCC cell line, PLC8024, which is both SALL4 + and CD133 + , and also radio-resistant (26, 27) (n = 6 per treatment group). In this chemoresistant model, we observed greater tumor growth (+1.5-fold) in the Sorafenib-treated group compared with the vehicle control group (Fig. 6F). Although PEN-FFW treatment showed minimal tumor inhibitory effect in this model, a significant synergistic effect of Sorafenib and PEN-FFW was observed in mice treated with both agents (P = 0.02, tumor growth inhibition 57% and 73% compared with control and sorafenib treated group, respectively) (Fig. 6F). This result further suggests that PEN-FFW alone, or as part of combination therapy, could bring clinical benefits to advanced stage HCC patients to overcome resistance to Sorafenib.
Drug-Like Properties of FFW. Biologics or compounds rapidly degraded in plasma generally demonstrate little efficacy in vivo. To further evaluate the potential of further development of PEN-FFW as a drug candidate, we examined the plasma stability, cell-penetration kinetics, and toxicity of the peptide. First, we monitored the degradation of PEN-FFW (1 μM) in de-identified, healthy human plasma by LC-MS/MS. At 30-min incubation (Fig.  7A, Inset), more than 90% of the peptide remained in the plasma. In comparison, the compound Eucatropine as a control was  rapidly degraded to 40% (SI Appendix, Fig. S10, Inset). Extending the experiment duration by 24 h revealed more than 50% of the peptide remained intact in the plasma after 4 h, gradually decreasing to ∼20% on the 24th hour (Fig. 7A). This suggests that PEN-FFW can potentially be developed as an intravenous drug. We subsequently conjugated FITC to the N terminus of PEN-FFW to assess the permeability and stability of the peptide in cells. Live cell imaging was performed on the treated cells at 2-min intervals for the first hour, and at 5-min intervals for the subsequent 23 h. It was observed that FITC-PEN-FFW entered cells by 18 min, and translocated to the nucleus by 22 min. At 2.5 h, FITC-PEN-FFW had entered every cell in the microscopic field, with incremental FITC-signals at 24 h ( Fig. 7B and Movie S1). Finally, potential toxic effects of the peptide were assessed with C57BL/6 mice harboring an intact immune system (n = 4). The mice were exposed to intraperitoneal administration of PEN-FFW (30 mg/kg) or vehicle (10% DMSO) every alternate day over the course of 17 d to a cumulative dose of 270 mg/kg. Mice in both groups remained alert, responsive, and did not exhibit notable signs of toxicity, such as lethargy, loss of mobility, or weight loss (Fig. 7C). After a 7-d washout period, complete blood counts and liver-function assay tests were performed, and organs were harvested for histology. To access potential liver injury caused by the peptide treatment, serum AST (aspartate aminotransferase) and ALT (alanine aminotransferase) levels were tested. Elevation of the two enzymes was not observed in the treatment group (Fig.  7D), indicating no liver damage was detected following treatment. Concurrently, no significant change in blood counts (Fig. 7E) or tissue damage (Fig. 7F) was observed in the treatment group compared with the control group.

Discussion
HCC is a deadly disease that lacks treatment options and alternative therapeutic approaches are urgently needed. In this report, we present our strategy to target HCC. We have previously found that SALL4 is elevated in a subclass of HCC enriched with hepatic progenitor cell features (16), and patients with this subtype suffer from a poor prognosis. RBBp4 is an essential subunit of NuRD complex, which in turn plays a key role in maintaining silencing of key regulators during embryonic development. A deregulated SALL4-RBBp4/NuRD pathway results in silencing of tumor suppressors, such as PTEN in HCC cells. Therefore, targeting the SALL4-NuRD pathway in HCC is a promising therapeutic approach. To further understand the mechanism of interaction, we first determined the crystal structure of the RBBp4-SALL4(1-12) complex. The crystal structure revealed a large, acidic interaction surface between the two nuclear factors, with side chains of the SALL4(1-12) peptide intercalating into the grooves of RBBp4, providing an opportunity for an engineered peptide to competitively inhibit the interaction (Fig. 8A). An integrated approach combining computational analysis, rational truncation, and systematic substitution studies was undertaken to design the therapeutic peptide. This approach facilitated the testing of peptide affinity down to single amino acid resolution, and provided real-time data feedback to the modeling, enabling further sequence enhancement. Using this approach, we were able to develop a potent SALL4-RBBp4/NuRD inhibitor that inhibits tumor growth in xenograft mouse models, targeting the "undruggable" nuclear factor.
Remarkably, the disruption of SALL4-RBBp4 with the FFW therapeutic peptide resulted in massive up-regulation of transcripts, which is very different from previous reports in which SALL4 was knocked down (16). SALL4 can both repress and activate genes, and down-regulation of SALL4 leads to both upand down-regulation of its target genes. In contrast, within the observed time period, FFW peptide treatment caused a unidirectional transcription activity shift toward only up-regulation of target genes (Fig. 8B). We hypothesize that in cancer cells, reactivation of SALL4 causes repression of tumor suppressors and epithelial markers like cadherins, resulting in a progenitorlike, undifferentiated cancer cell type. Upon peptide disruption of the SALL4-RBBp4 axis, the repression of SALL4 function is released and results in mass activation of transcripts unfavorable for cancer cell survival but beneficial to patient survival.
In the transwell cell migration assay, it was demonstrated that PEN-FFW-treated cells have impaired cell migration compared with controls, and this phenotypic change corresponds to the upregulation of transcripts encoding cell adhesion molecules and cadherins. Similar to knocking down SALL4, the disruption of SALL4-RBBp4 resulted in apoptosis (16). Together with the upregulation of tumor-suppressor transcripts, the overall transcriptome change brought by FFW is at a disadvantage to tumor progression. Simultaneously, these transcripts are associated with a good prognosis in HCC patients, as shown by survival analysis.
The use of peptides as drugs may hold significant advantages over small molecules in targeting of protein-protein interactions, as large binding surfaces typically lack defined or deep binding pockets (28,29). Thus, we chose to target the SALL4-RBBp4 protein-protein interactions using a peptide due to the large and predominantly shallow binding interface between SALL4 and RBBp4. Moreover, a well-designed peptide drug tends to possess higher selectivity with a safer toxicity profile compared with small molecules. In certain therapeutic areas in which small molecules have limited success, peptides have been shown to be valuable substitutes. For example, the development of protease-resistant stapled peptides with cell-penetrating capabilities (30) and new peptide formulations that enhance their oral bioavailability (31) have resulted in a dozen peptides in clinical trials and tens of approved peptide drugs (29). We tested the stability of PEN-FFW  in human plasma, and the results demonstrated notable stability in which 90% of the peptide was retained after 30-min incubation under physiological conditions, gradually decreasing over time with 20% still intact at 24 h. This stability profile indicates a possibility to develop an intravenous delivery route for FFW. One advantage of using peptides as drugs is minimal toxicity. The suitability of further developing FFW is further supported by the low toxicity profile of FFW in C57BL/6 mice. PEN-FFW is a specific SALL4-RBBp4/NuRD inhibitory peptide targeting SALL4 + HCC by binding to RBBp4. This represents a first attempt to target RBBp4, a chaperone protein with WD40repeats. In the nucleus, the RBBp4/NuRD complex binds to a few other lineage-specific proteins, such as BCL11B and FOG1. Several crystal structures of RBBp4 with its binding partners have been reported previously, including BCL11A (32), FOG-1 (33), and PHF6 (34). However, in the context of HCC cells, these molecules are not present or are expressed at very low abundance. Hence, we hypothesize that RBBp4/NuRD binds to different molecules in a cell-type-specific manner, and with this, the PEN-FFW peptide is specific to SALL4-RBBp4 binding in HCC cells.
A deregulated SALL4-RBBp4/NuRD pathway results in silencing of tumor suppressors in HCC cells. The SALL4-RBBp4/ NuRD interaction represents an exciting target in HCC, and unlike HDAC inhibitors, which target cells nonselectively, this SALL4-RBBp4/NuRD interaction has been observed to be exclusive to cancer cells that express SALL4. It is worth mentioning that RBBp4 is present in other nuclear complexes, such as PRC2, MLL2, and Sin 3, functioning as a chaperone. It is possible that blocking RBBp4 could disrupt the formation of these general nuclear complexes in a cell-type-independent manner. Further investigations will need to be carried out to understand how the peptide would affect these complexes.
Through resolving the structure of the RBBp4-SALL4(1-12) complex, identification of the key interacting residues, and systematic peptide studies, a potent therapeutic peptide, FFW, was demonstrated to confer robust proapoptotic and antimigration effects in SALL4-expressing HCC cells, and induced marked improvement in therapeutic efficacy in mice bearing hepatocel-lular carcinoma tumors. Mechanistically, global genomic studies revealed that this FFW could convert SALL4 from a dualregulator to an activator-only status. Overall, the FFW peptide could be the basis for development of a first-in-class drug, as there is currently no available inhibitor targeting the SALL4-RBBp4/NuRD interaction, and in addition, provide a viable therapeutic strategy for the significant subset of patients with HCC and many other cancers whose malignancies are driven by SALL4 expression.

Methods
Crystallization screens were performed with the hanging-drop vapordiffusion method using Hampton Research screens. The RBBP4 protein was purchased from SinoBiological and concentrated to 8 mg/mL in 50 mM Tris 100 mM NaCl. The concentrated RBBP4 protein was mixed with 20 mM of SALL4 peptide and crystallization drops were set up at a 1:1 ratio. Diffraction quality crystals of SALL4(1-12) -RBBp4 complexes were obtained from a reservoir solution containing 0.2 M sodium chloride, 0.1 M Bis-Tris, pH 5.5, 25% PEG 3,350. Detailed protocol of crystallization and structure determination can be found in SI Appendix, SI Material and Method.
Detailed protocols for ITC, SPR, flourescence polarization, computational alanine scanning, cell culture and treatment with peptide, coimmunoprecipitation, migration and wound-healing assay, xenograft studies, plasma stability assay, live cell imaging, RNA-seq data analysis, ChIP, and ChIP-seq analysis can also be found in SI Appendix, SI Material and Method. All animal studies were conducted with the approval of the Institutional Animal Care and Use Committee (IACUC) of the National University of Singapore.