Chromenopyrazole−Peptide Conjugates as Small‐Molecule Based Inhibitors Disrupting the Protein−RNA Interaction of LIN28‐let‐7

Targeting the protein−RNA interaction of LIN28 and let‐7 is a promising strategy for the development of novel anticancer therapeutics. However, a limited number of small‐molecule inhibitors disrupting the LIN28‐let‐7 interaction with potent efficacy are available. Herein, we developed a novel LIN28‐inhibiting strategy by targeting selective hotspot amino acids at the LIN28‐let‐7 binding interface with small‐molecule‐based bifunctional conjugates. Starting from reported small‐molecule LIN28 inhibitors, we identified a feasible linker‐attachment position after performing a structure‐activity relationship exploration based on the LIN28‐targeting chromenopyrazoles. In parallel, a virtual alanine scan identified hotspot residues at the protein−RNA binding interface, based on which we designed a set of peptides to enhance the interaction with the identified hotspot residues. Conjugation of the tailor‐designed peptides with linker‐attached chromenopyrazoles yielded a series of bifunctional small‐molecule‐peptide conjugates, represented by compound 83 (PH‐223), as a new LIN28‐targeting chemical modality. Our result demonstrated an unexplored rational design approach using bifunctional conjugates to target protein−RNA interactions.


Introduction
Cellular events are regulated by an extensive network of interactions between RNAs and associated RNA-binding proteins (RBPs). [1] An increasing number of compounds has been identified as small-molecule binders for structured RNAs to modulate the cellular phenomena. [2] In comparison, small molecules targeting RBPs are understudied and most reported RBP inhibitors suffer from poor in vivo efficacies. [3] RBP-targeting presents an alternative approach to RNA-targeting for the treatment of diseases associated with the dysregulation of RBPÀ RNA interactions and presents an additional layer of selectivity in perturbing proteinÀ RNA interactions (PRIs). Therefore, the development of new chemical modalities targeting RBPs is highly sought-after.
The interaction between the miRNA-binding protein Lin28 and the let-7 miRNA family, initially discovered in 1984, is one of the most thoroughly investigated and characterized RBPÀ RNA interactions. [4] The let-7 miRNAs are responsible for the stabilization of differentiated cell states by regulating CDK/cytokine proteins and LIN28 mRNA, being key regulators of the immune system. [5] Oncogenes like RAS, [6] MYC, [7] and HMGA2 [5c] are all downregulated by let-7 and the dysregulation of let-7 has been associated with cancers. LIN28 is expressed in two different isoforms in mammals, LIN28A and LIN28B. While LIN28A is exclusively found in the cytosol and interferes with the pre-let-7 element processed by DICER, the LIN28B isoform is mainly localized in the nucleus and prevents the processing of pri-let-7 ( Figure 1A). [8] LIN28 overexpression has been found in many types of tumors and is associated with poor prognosis. [9] Therefore, inhibition of upregulated LIN28 to restore let-7 levels is emerging as an appealing anticancer strategy. Concomitantly, small-molecule inhibitors of varied chemical scaffolds disrupting the LIN28-let-7 interaction have been identified ( Figure 1B). [10] However, the inhibition mechanism for many reported LIN28 inhibitors was only partially investigated or is not clear. Furthermore, even the most potent inhibitors reported to date showed limited micromolar potency and poor cellular activities.
In this context, we started with the structural optimizations based on small-molecule LIN28 inhibitors and identified an amenable position for linker attachment on the chromenopyr-azole scaffold. [10a] We then rationally designed and synthesized a set of peptides based on key residues of LIN28 involved in let-7 binding. A series of new bifunctional conjugates based on the chromenopyrazole scaffold clicked to the peptides were obtained through copper-catalyzed azideÀ alkyne cycloaddition as new LIN28-inhibiting molecules. To improve the potency of the bifunctional conjugates, we capitalized the well-established principle of hotspot analysis in our design and applied the concept used in developing proteinÀ protein interaction inhibitors to develop RBP-targeting molecules. [11] The resulting bifunctional conjugates were further tested for LIN28-inhibitory activities and associated pharmacokinetic properties (Figure 1 C).

Structural modifications based on the chromenopyrazole scaffold
To probe an amenable position for the linker attachment, we first performed structural optimizations centered on the LIN28targeting chromenopyrazole scaffold for which the carboxylic acid on the N-1 substituent was reported as the essential group to maintain the LIN28-inhibitory potency. [10a,12] We focused on four positions on the scaffold for the structural modifications and evaluated them in EMSA (Table 1). First, we changed the position of the carboxylic acid (R 2 ) on the N-1 substituent to determine if this is a suitable position for linker appendage. We observed that both the 4-carboxyphenyl (1/SB1301) and 3carboxyphenyl (2) substituents efficiently disrupted the PRI. Introduction of a 4-sulfophenyl (3) substituent retained activity, whereas the inclusion of a heteroatom in the aromatic fragment was not well tolerated (4). In comparison, compounds without the carboxylic acid (5, 6, 7) did not show any inhibitory activity. Second, replacement of the carbamate group of 1 (R 1 ) with a carbamoyl group (8) retained activity. Methylation on the carbamoyl group of 8 led to compound 9 with a complete loss of activity. Removal of the aromatic 4-(benzyloxy)carbonyl subsitituent and replacement of the piperazine with a piperidine ring led to compound 10 with a small decrease in activity. In comparison with the piperidine 10, the morpholino analogue 11 was inactive. Subsequently, we hypothesized that the linker attachment through an amide bond at the carbonyl position of the 4-(benzyloxy)carbonyl residue could be a feasible approach to maintain the activity. This was tested by comparing the deprotected inactive piperazine precursor 12 with the 4-(2chloroacetyl)piperazin-1-yl derivative 13, which showed a relative inhibition of 60 %, indicating that R 1 is potentially qualified as a linker attachment point. Third, the nitro group at the 8-position (R 3 ) was evaluated as the linker appendage position. Reduction of the nitro group led to the amine compound 14 with greatly reduced inhibitory activity of only 25 % in comparison with that of 1. Addressing the 8-amino group with the similar approach as applied to the R 3 group with a 2-chloroacetamido substituent led to compound 15 with retained, if not improved, inhibitory activity in comparison with that of 1. The position R 3 thus qualifies as a promising linker attachment position. Additionally, we replaced the oxygen of the chromenopyrazole core scaffold with nitrogen to generate the quinolinopyrazole analogs 16 and 17, which showed either greatly reduced activity of 21 % or complete loss of activity. The collective results from the structural modifications indicated that the 8-position on the chromenopyrazole scaffold, as shown in compound 15, is the optimal position to attach the linker for the assembly of the bifunctional conjugates. Linker attachment at the 8-position to replace the 8-nitro group of 1 has the advantage that it can eliminate the cellular liability due to the electrochemical properties associated with the nitro group. [13] To provide the potential binding feature related to the R 3linkage at the 8-position, we performed the docking analysis based on the LIN28-preE-let-7f-1 complex (pdb: 5udz). [14] The optimal docking mode of compound 15 revealed consistent results with that of compound 1 and a series of spirocyclic compounds based on the same chromenopyrazole scaffold. [10j] The carboxylic acid substituent on the N-1 phenyl group engaged with Arg50 via both salt bridge and hydrogen bond interactions, demonstrating the importance of the carboxylic acid substituent. The substituent at the 8-position pointed towards a solvent-exposed channel attached to the surface pocket overlapped with the 5U6A and 14U15A binding sites of let-7 ( Figure S1 and S2). This unblocked solvent-exposed channel can probably explain the feasibility to attach linkers for further conjugation with an affinity-enhancing group.

Design of peptides
To improve PRI perturbation, the concept of bifunctional conjugates was applied. Due to the observation that PRIs and proteinÀ protein interactions have similarities in the nature of the interacting surfaces, we assumed that we could learn from strategies used to address proteinÀ protein interactions to target PRIs ( Figure 2A). [15] The general approach to target proteinÀ protein interactions is to capitalize on the knowledge of hotspot amino acids which are amino acids that contribute significantly more to a binding event than other involved amino acids. [16] The surfaces involved in interacting with macromolecules such as RNAs require hotspots to be evenly distributed to ensure efficient binding. Many synthetic small-molecule drugs obey the rule of five [17] and are therefore not large enough to cover sufficient hotspot residues to achieve strong binding. [18] We started the design for the affinity-enhancing peptide fragment to overcome the limitations of small molecules with a virtual alanine scan. [19] The tool PRI-HotScore scans a designated crystal structure and calculates an interaction score. An interaction score value between 1 and 2 indicates warm spot residues and a value above 2 indicates hotspot residues ( Figure 2B and 2C). [20] Except three warm spots, all identified warm-or hotspot residues on the let-7-binding domain on LIN28 possess either aromatic or positively charged side chains.
The peptides chosen for the bifunctional probes therefore mainly consisted of aromatic and negatively charged side chains to favor stacking and electrostatic interactions between the peptide and the protein surface and to incorporate potential repulsive effects between the negatively charged amino acid side chains of the peptide and the negatively charged RNA backbone. Since RBPs tend to have multiple interacting surfaces with bound RNA substrates and the dynamic RBPÀ RNA interactions involve flexible linkers connect-ing different RNA-binding domains, [21] the accurate prediction of potential interaction surfaces is difficult, as exemplified in the case of the LIN28-let-7 interaction. [22] For this reason, our design did not follow well-established surface analysis based on crystal structures, as was applied for the rational design of proteinprotein interaction inhibitors. [23] Instead, we focused on mimicking the approximate molecular volume-to-charge ratio of the RNA fragments with the designed peptides. The design started by analyzing the volume-to-charge ratio of the bound let-7 RNA in the crystal structure (PDB: 5UDZ). An investigation of the individual volume occupied by the respective nucleotides was already performed (Table S4-S6). [24] Based on the PDB sequence, two volume-to-charge ratios of let-7 were calculated. The first ratio was calculated for the entire RNA sequence and resulted in 306.2 Å 3 per charge. The second ratio of 300.8 Å 3 per charge was calculated based on the fragment occupied in the crystal structure. To establish a first guideline, the goal was to design peptides possessing a volume-to-charge ratio of around 300 Å 3 . Following this threshold, 14 peptides were designed containing either aromatic, negatively charged, or both aromatic and negatively charged side chains based on the predetermined  values for the molecular volume of amino acids. [25] Two peptides consisting of only alanine (34 and 35, Table 2) were used as inactive controls. The purpose of these peptides was to enlarge the selected inhibitor to cover more hotspot amino acids to introduce more chemical modifiability and improved potency to the molecule while relying on the affinity of the small compound.

Synthesis and evaluation of bifunctional conjugates
Upon the successful isolation of the acetylated and alkylated peptide derivatives, we proceeded with the conjugation of peptides and chromenopyrazoles to assemble the bifunctional conjugates. Since aromaticity-induced stacking interaction is one of the main contributing factors for proteinÀ RNA interactions, the copper-catalyzed azideÀ alkyne cycloaddition was used to perform the conjugation with the rationale that the newly formed triazole linker could potentially contribute to RBP-binding. In addition to aromaticity, the triazole group introduced rigidity in contrast to the flexible peptide fragment. The combinations of different PEG-based linkers were employed in the synthesis of the bifunctional molecules. Firstly, one set of 14 bifunctional conjugates was obtained from the chromenopyrazole with azide 39 and peptides modified with alkyne 41 (Scheme 1A). The obtained bifunctional conjugates were evaluated in EMSA in a single dose measurement at 75 μM. Nine peptides (25-32 and 38) were selected for the following synthesis by combining azides 39/40 and alkynes 41/ 42 in different variations to give a library set of 40 bifunctional conjugates (Scheme 1B and 1C). Among the first set of obtained 14 conjugates, six showed inhibition at 75 μM concentration and compound 52 showed significant inhibitory activity in the EMSA (98 % inhibition at 75 μM). Among the remaining bifunctional molecules with longer linkers obtained using azide 40 and alkyne 42, compounds conjugated with the peptide sequence NFQWNY (65, 74, and 83) showed inhibitory activity comparable to that of 1. All four bifunctional conjugates (52, 65, 74, and 83) showed equal inhibitory potency in comparison with C902, which is among the most potent LIN28 inhibitors reported to date (Table 3)

Evaluation of the most potent bifunctional conjugates
The most active bifunctional conjugates 52, 65, 74, and 83, which all feature the aromatic side chain NFQWNY instead of negatively charged side chains, were then evaluated for dosedependent inhibition. All four probes consist of the same peptide sequence and only differ in their respective linkers. Conjugate 52 consists of two short linkers while 83 consists of a short linker on the small-molecule side and a long linker on the peptide side. The evaluation revealed that 52 and 83 were the most potent conjugates (IC 50 : 4.9 μM and 4.0 μM for 52 and 83, respectively).
In comparison, compound 65 that consisted of a long linker on both the peptide and small-molecule component showed an IC 50 of 19.5 μM and compound 74, consisting of a long linker on the small-molecule side and a short linker on the peptide side, showed an IC 50 of 14.9 μM (Figure 3). It is noteworthy that  (Figure 4). Therefore, the results of 83 confirmed that targeting RBPs through a hotspot-focused strategy is a feasible way to improve target affinity.

Stability and permeability evaluation
Due to the inherent disadvantages of peptides in terms of pharmacokinetic properties, the stability and potential perme-  ability of the bifunctional conjugates were evaluated. We tested the membrane permeability of the compounds by employing an immobilized artificial membrane (IAM) column based on the probability to interact with phospholipids and therefore capitalize passive diffusion through the membrane. [26] Compounds with optimal permeability should score a chromatographic hydrophobicity index (CHI) value between 35 and 50. Compounds with a CHI value of more than 35 were found to be   sufficiently adhesive towards phospholipids to enable resorption through passive diffusion. Compounds with a CHI value above 50, especially if containing positively charged moieties, are prone to cause phospholipidosis. The analysis of the peptides alone showed that only the acetylated peptide 25 fit the optimal interval with a CHI value of 46. Fortunately, except 45 (À 13) and 66 (À 12), all small-molecule-peptide conjugates showed CHI values between 35 and 45 ( Figures S117-S121), indicating that the conjugated molecules probably do not suffer from potential permeability issues in contrast to pure peptides. Another concern associated with peptide-based entities is the long-term stability. Degradation by intracellular machineries often limits the application of peptide-based drugs. Thus, a lysate stability assay was performed to investigate the stability of the bifunctional molecules. Nine selected bifunctional conjugates (67-75) were tested in the stability assay in cell lysate and incubated at 37°C for 24 h, followed by HPLC analysis (Table S8 and Figures S122-S139). All conjugates showed decent stability in cell lysate with only minor degree of instability. Compounds 67 and 74 showed the highest instability in cell lysate, as compound 67 decreased by 15 % over the course of 24 h and 74 decreased by 22 % ( Figure 5). All other conjugates showed less than 10 % degradation. The employed peptides have the potential to be modified via stapling, macrocyclization or aminomethylation to further improve stability properties.

Conclusions
This study investigated the strategy of small-molecule-peptide conjugates as bifunctional molecules to disrupt and inhibit the LIN28-let-7 interaction. Starting from the LIN28-inhibiting chromenopyrazole scaffold, an amenable linker attachment position to attach affinity-enhancing fragment was identified after performing structural optimization centered on the chromenopyrazole core. The attached peptides featuring both aromatic and negatively charged residues were rationally designed with a focus on targeting highlighted hotspot amino acids, identified via a virtual alanine scan, in the binding interface of LIN28 and let-7. The peptide constituted of the aromatic sequence NFQWNY was evaluated as the preferred side chain over the ones with negatively charged residues. Analysis of the synthesized small-molecule-peptide conjugates revealed that a short linker fragment between the small molecule and triazole group turned out to be the more efficient. The most potent bifunctional molecule 83 showed a high inhibitory potency (IC 50 : 4.0 μM), thereby confirming the initial hypothesis that a hotspot-focused approach to improve RBP targeting is valid. Additionally, the introduced peptide fragments permit further chemical modifications on the peptide, circumventing the limitations on modifications that were tolerated on the chromenopyrazole scaffold. Besides creating the opportunity to further optimize the affinity towards the target protein, this tailored fragment introduction permits the application of established techniques regarding stability, selectivity, and permeability used in the peptide field. The expansion of an initial proven small-molecule binder introduces new potential for optimization to address both the initial binding site and newly covered hotspot amino acids. A lipid interaction and a lysate stability assay suggested that the inherent disadvantages for peptide-based entities have been attenuated in the obtained small-molecule-peptide conjugates. Considering the scarcity of potent LIN28 inhibitors and the limited discovery approach beyond screening efforts, our strategy using bifunctional conjugates provides a new method to target RBPs via a rational design approach by focusing on hotspots residues at the interface of RBP and RNA interactions.

Experimental Section
Synthetic procedures, compound characterization, and additional figures can be found in the Supporting Information.
LIN28 expression and purification: Human LIN28A (residues 16-187) was expressed in Escherichia coli BL21(DE3). Incubation of the culture was performed at 37°C until the absorbance reached 0.5-0.7 at 600 nm (OD600). IPTG was added to a final concentration of 300 μM. The induction was performed at 18°C overnight. After centrifugation, the bacterial pellet was resuspended in lysis buffer (pH 7.5, 50 mM NaH 2 PO 4 , 300 mM NaCl, 0.1 mM PMSF) and lysed using a Microfluidizer, after which a fresh portion of 0.1 mM PMSF and Triton X-100 (1 % final concentration) were added. The lysate was cleared by centrifugation at 60000 xg and 4°C for 1 h. Immobilized nickel affinity chromatography (HisTrap, GE Healthcare) in buffer containing 50 mM NaH 2 PO 4 (pH 8), 300 mM NaCl and 5 % glycerol was used for the purification of the obtained protein. A maximum concentration of 0.5 M imidazole was used for the gradient elution, the affinity tag was cleaved using His 6 À TEVprotease, and the protease and unspecific binders were eliminated by a second nickel affinity chromatography. LIN28A containing fractions were concentrated and applied to a HiLoad Superdex 75 pg 16/600 column (GE Healthcare) with gel-filtration buffer (pH 7.5, 30 mM NaH 2 PO 4 , 50 mM NaCl, 5 % glycerol, 2 mM β-ME). The purified protein was concentrated and stored at À 80°C.
Electrophoretic mobility shift assay (EMSA): Purified LIN28A (residues 16-187) was incubated with individual compounds and 5 U recombinant ribonuclease inhibitor (Takara Bio) in EMSA reaction buffer (50 mM Tris (pH 7.5), 100 mM NaCl, 10 mM β- mercaptoethanol, 50 μM ZnCl 2 , 2 % DMSO, 0.01 % Tween 20, 12 % glycerol). After incubation at room temperature for 2 hours, preElet-7f-1-Cy3 (mus musculus, purchased from IDT) was added to a final concentration of 5 nM and a reaction volume of 50 μL. The final concentration of LIN28A was 10 nM and compound concentrations up to 75 μM were used. The reaction mixtures were incubated for another 15 minutes, then 10 μL of each reaction was separated in an 8 % polyacrylamide TAE gel at 4°C and 220 V for 1 h using 0.25x TAE as running buffer. Cy3 fluorescence was detected with a ChemiDoc MP (Bio-Rad) and 2 minutes of exposure time. Band intensities were measured with Image Lab (Bio-Rad). IC 50 values were determined by curve fitting with GraphPad Prism 8.
Thermal shift assay (nanoDSF): Measurements were performed with a NanoTemper Prometheus NT.48 nanoDSF instrument for the assessment of melting temperatures of the cold-shock domain (CSD)-compound complex. Compounds (75 μM, 5 % DMSO) were incubated with the LIN28A_CSD (residues 16-126, 30 μM) for 45 min in nanoDSF buffer (30 mM NaH 2 PO 4 (pH 8.0), 50 mM NaCl, 1 mM MgCl 2 ). A temperature scan ranging from 20°C to 90°C with a slope of 1°C/min with an excitation power of 90 % was performed. The ratio of intrinsic tryptophan and tyrosine fluorescence at 350 nm and 330 nm was measured and the first derivative was determined using the software of the instrument, GraphPad Prism 8 and OriginPro 2022b.
Peptide stability assay: Peptide stability was tested in cell lysate prepared from HeLa cells using the freeze-thaw method. Peptides were dissolved in lysate (normalized to 5 mg/mL protein using PBS) at 600 μM and incubated at 37°C. Samples were taken at 0 min and 24 h and mixed in a 1 : 1 ratio with cold MeOH containing 0.05 mg/ mL ethylparaben as an internal standard. Samples were mixed and incubated for 15 min on ice, centrifuged at 14000 rpm at 4°C for 10 min, and the resulting supernatant was carefully removed and analyzed by HPLC.
HPLC based lipophilicity analysis: CHI values were determined using IAM column (IAM.PC.DD2 4.6 × 100 from Regis Technologies, Inc.). A mobile phase of 50 mM ammonium acetate adjusted to pH 7.4 and acetonitrile was used with a flow rate of 1.5 mL/min. Gradients were run from 0 to 85 % acetonitrile in 4.75 min. Then 85 % acetonitrile was kept for 0.25 min and then flushed back to 0 % acetonitrile in 0.25 min. Re-equilibration of the column with pure aqueous buffer was performed for the remaining time of the overall 6 min run. Each run was calibrated with an internal standard to ensure comparability. Readout was either 210 or 254 nm depending on if peptides or bifunctional molecules were measured. CHI values were derived through a calibration plot to evaluate the molecules' lipophilicity and potential for phospholipidosis. [26]