Monoterpenoid fluorophthalimide IMiDs that lack human cereblon binding: synthesis and anti-proliferative, anti-angiogenic and anti-inflammatory activities

2.1. Chemistry

Melting points (uncorrected) were measured with a Fisher-Johns apparatus. ¹H NMR, ¹⁹F NMR, and ¹³C NMR spectra were obtained on a Varian Mercury Plus 400 spectrometer at 400 MHz, 377 MHz, and 101 MHz, respectively, at room temperature. GC/MS (m/z) spectra were measured on an Agilent Technologies (Santa Clara, CA, USA) 7890B GC equipped with an HP-5MS column and a 5977B mass-selective ion detector in the electron-impact mode. LC/MS spectra were recorded on an Agilent 6125-LC-MS Single Quad system. Elemental analyses were performed by Atlantic Microlab, Inc. (Norcross, GA, USA). All isopinocampheylamine, myrtanylamine, bornylamine, and 2-aminonorbornane hydrochloride were purchased from Sigma-Aldrich (St Louis, MO, USA) and used as received. X-ray crystallography analyses were performed in the Small Molecule X-ray Facility, Department of Chemistry, Johns Hopkins University.

2.2. Cell Culture

Experimental compounds and comparators were prepared in DMSO (Sigma, St Louis, MO, USA), which was ultimately diluted to either 0.1 % or 0.5 %.

2.3. Assays

2.4. Molecular modeling monoterpenoid-based IMiDs docking with human cereblon

A molecular docking study was conducted on the crystal structure of human cereblon in complex with DDB1 (damage specific DNA binding protein 1) and lenalidomide (PDB ID - 4TZ4) utilizing the Glide program, which is part of the Schrödinger molecular modeling package, with extra precision (XP). The human cereblon protein structure was selected to improve the comprehensiveness and reliability of docking predictions by incorporating structural diversity, accuracy, variability in binding sites, and redundancy in validation. This PDB entry contains the native ligand lenalidomide within its structure. The structure was constructed using the protein preparation wizard in the Schrödinger module. This process included the addition of hydrogen atoms and the removal of water molecules beyond 5 Å from the binding sites. Active site grid were generated utilizing the Receptor Grid Generation tool within the Glide module (Maestro, Schrödinger Release 2023–2). The docking experiment involved the re-docking of Lenalidomide. LigPrep (Schrödinger) was employed to assess the chiralities of lenalidomide and subsequently convert it into a 3D structure (LigPrep, Schrödinger Release 2023–2). The OPLS_2005 force field was utilized to facilitate ionization and tautomeric states. The Schrödinger molecular modeling package Glide, specifically the grid-based ligand docking with extra precision (XP) tool, was utilized to enhance the accuracy of molecular docking. Eight monoterpenoid-based IMiDs were then docked at the active site of lenalidomide to analyze the binding and associated interactions by applying the same ligand preparation protocol.

2.5. Statistical analysis

Multiple within-subject comparisons were conducted using a Dunnett’s test (with Bonferroni correction for multiple comparisons, when appropriate). The statistical analysis was performed using GraphPad Prism 7 (San Diego, CA, USA), with a predefined significance level of p < 0.05 for each assessment. All presented data are expressed throughout the text as the mean ± standard error of the mean (SEM). The number of samples per group (n) is provided within each Figure legend.

Chemistry (synthesis):

As illustrated in Schemes 1 and 2, either tetra- or di-fluorophthalic anhydrides were condensated with selected monoterpenoid moieties, such as N-isopinocampheyl (as in 1,5,6), N-bornyl (as in 3), N-norbornyl (as in 4,8), and N-myrtanyl (as in 2,7) groups and, ultimately, were converted to a variety of novel bridged-ring agents (Scheme 1: compounds 1–4; Scheme 2: compounds 5–8).

The synthesis of compounds 1–8, with representative examples and chemical characterization, is described below (Methods). Notably, for select compounds, their three-dimensional structure and stereochemical equilibrium were confirmed by X-ray crystallography and are described below and in the Supplemental information (Appendix A. Supplementary data) [29]).

Human cereblon concentration-dependent binding of compounds 1–8:

Human CRBN, an evolutionary conserved 441- or 442-amino acid protein that was initially reported as a gene product associated with a mild mental disability [31], was later identified as the primary target of thalidomide [10–13]. CRBN, additionally, is a substrate recognition receptor for Cullin 4 RING E3 ubiquitin ligase, which is part of the ubiquitin-proteasome system (UPS) - a major pathway accountable for the degradation of >80 % of proteins in eukaryotic cells [32]. Thalidomide and related IMiDs bind to human CRBN, triggering the recruitment of non-native substrates to the E3 ubiquitin ligase and their consequential proteasomal degradation [32]. The premise underpinning our generation of monoterpenoid-based IMiDs is that the generated bridged-like ring structure present, in leu of the glutarimide ring of thalidomide and clinical analogues, would provide complexity and size to, thereby, preclude the alignment and binding of these compounds into the shallow surface thalidomide binding domain present in the C-terminal of human CRBN. A bead-based AlphaScreen luminescence assay was hence used to evaluate the binding of 8 monoterpenoid-based IMiDs to human CRBN in which pomalidomide was used as a positive control.

As evident from the concentration-dependent plots of compounds shown in Fig. 1, a potent interaction was found between pomalidomide (used as a positive control) and human CRBN as evaluated by a cereblon/BRD3 binding FRET assay (IC₅₀ binding values between 3.743 and 4.42 μM). In contrast, the CRBN binding of monoterpenoid-based IMiDs 1–8 possessed IC₅₀ values in the range of 43 to >100 μM (Fig. 1), which are indicative of negligible binding (i.e., it is in line with a background level of binding).

Our results align with studies of Boichenko and Hartmann [22] that defined the basic requirements for IMiD ligand binding within the human CRBN thalidomide binding pocket. Replacing the glutarimide moiety with a variety of other structures, they determined that at least one carbonyl group is essential for human CRBN binding, and this should be in the 6′-position in relation to the structure of thalidomide or pomalidomide. Furthermore, in both 6- and 5-membered rings, imides provided potent binding and that lactams and cyclic rings with the functional group R–NH–CO–R′ would bind [22]. In contrast, rings lacking these features and, in particular, 7-membered or larger rings would minimally bind. In this regard, the bridged-like ring structures of monoterpenoid-based IMiDs 1–8, have greater complexity and size than glutarimide, and lack the key structural features considered essential to human CRBN binding (i.e., a nitrogen containing ring and/or carbonyl moiety), which is consistent with their high IC₅₀ values – indicative of negligible/poor CRBN binding.

Molecular modeling evaluation of human cereblon interaction:

To clarify the structural basis underpinning the substantial differences in the measured IC₅₀ binding values between compounds 1–8 and a classic glutarimide-containing IMiD, molecular modeling studies were performed on the structure of human CRBN co-crystalized with lenalidomide (1:1) [33]. An initial evaluation of docking accuracy was performed by re-docking the co-crystal ligand lenalidomide (PDB entry: 4TZ4 [33]) into its ligand-binding site, and then calculating the RMSD (Root Mean square Deviation value) between the native and docked conformations. The proximity of the redocked ligand into its original position indicated that the grid box dimensions (20 × 20 × 20 Å; 0.1 Å spacing) constructed around the CRBN target were adequate for docking a native-like ligand from the monoterpenoid-based IMiDs dataset. The structure of the glutarimide moiety, along with essential polar and hydrophobic pocket-lining residues within the docking grid, confirmed that the intermolecular interactions established by lenalidomide with human CRBN protein residues are compatable with human CRBN interaction and binding.

Specifically, Fig. 2 illustrates the docked conformation of the co-crystallized ligand lenalidomide with human CRBN protein. The optimal docking pose, exhibiting a binding energy of −11.69 kcal/mol (Table 1), was identified, and demonstrated a RMSD of 0.12 Å (an indication of close replication to the previously reported docking [33]). The re-docked pose of lenalidomide within the thalidomide binding domain of human CRBN included four hydrogen bonds (with Asn351, His378, Trp380) and one π-π stacking interaction (His353) (Fig. 2A and B). In direct comparison to this, compound 5 exhibits the highest binding energy of −3.92 kcal/mol of the evaluated monoterpenoid-based IMiDs 1–8 (Table 1), and only one hydrogen bond interaction with Cys343 was evident (Fig. 2C and D). Notably, no interactions were evident between the monoterpenoid ring conformation and human CRBN. This lack of interactions was confirmed in our molecular modeling of the other 7 monoterpenoid-based IMiDs. Hence, the binding energy scores of compounds 1–8 are substantially different from that of lenalidomide in relation to human CRBN binding, and align with the negligible experimentally measured binding of compounds 1–8 to human CRBN (Fig. 1).

Anti-angiogenesis activity in human umbilical vein endothelial cells:

The process of angiogenesis, resulting in the generation of new blood vessels from the preexisting vascular network, plays a crucial role in the growth, metastasis, and survival of tumors. This includes hematological cancers, such as MM [34,35], in which monoclonal plasma cells within the bone marrow, similar to other tumor cells, secrete pro-angiogenic factors that stimulate the sprouting of new vessels from nearby capillaries to induce vascular development and, ultimately, to support tumor growth. Furthermore, time- and tissue-dependent angiogenesis is undoubtedly critical in embryonic and fetal development, and may, in part, underpin the teratogenicity of thalidomide [2, 14,36–38]. The angiogenesis process involves the proliferation and migration of host endothelial cells, and their rearrangement to form tube-like structures that, ultimately, develop into functional capillary vessels to augment tumor oxygen and nutrient delivery. Such vessels within tumors are often ill-formed, leaky, and tortuous and, thereby, permit metastatic spread [39,40]. Targeting angiogenesis as a cancer therapy has hence become an intense area of scientific focus to inhibit cancer progression and induce regression. Human umbilical vein endothelial cells (HUVECs) generate hollow tube-like structures when cultured on biological gels, such as ECMatrix, and tubule formation can then be used as a simple in vitro measure of angiogenesis (the lattice assay: Fig. 3A).

A previously optimized lattice assay [17,19] was hence used for compound screening at 10 μM to evaluate anti-angiogenic activity. The known anti-angiogenic agent CPS49 (30 μM), together with thalidomide (100 μM), was assessed alongside, which demonstrated potent and a lack of activity, respectively, consistent with prior studies [17,19,20,23,25]. As shown in Fig. 3, whereas tetrafluorinated compounds 1–4 blocked tube formation, their difluorinated equivalents, 5–8, lacked anti-angiogenic action. Notably, the anti-angiogenic activity of non-CRBN binding compounds 1–4 is in line with the demonstration by Beedie et al. [17] of an IMiD-mediated cereblon-independent anti-angiogenic mechanism in siRNA-mediated CRBN knockdown HUVEC cells in which a loss of CRBN does not eliminate thalidomide’s anti-angiogenic activity [17]. In concert with this, enhanced CRBN binding does not appear to correlate with the anti-angiogenic effects of thalidomide analogues from in silico studies [18], reinforcing the presence of CRBN-independent anti-angiogenic pathways. Also of interest, CPS49, which was used as a positive control in the HUVEC assay consequent to its known anti-angiogenic actions, is a highly fluorinated IMiD with a modified glutarimide ring – similar to but, albeit, structurally different from compounds 1–4. However, whereas CPS49 has been described to generate thalidomide-like induced limb damage in chicken embryos [41], compound 3 (also known as tetrafluorobornylphthalimide (TFBP) [21]) lacked adverse activity when similarly administered to chicken embryos [21].

Anti-proliferative activity against lenalidomide sensitive and resistant multiple myeloma cells:

The development of drug resistance is a primary mechanism underpinning relapse during cancer treatment [42]. In this light, the anti-proliferative activity (cytotoxicity) of compounds 1–8 was evaluated in two human MM cell lines: one sensitive to the widely used IMiD, lenalidomide (MM.1S), and the other resistant to it (U266 R10R). As shown in Fig. 4, tetrafluorinated compounds 1–4 substantially inhibited the cell viability of lenalidomide sensitive (A) and resistant (B) tumor cells. In contrast, their difluorinated analogues, 5–8, substantially lacked activity across both MM cell types. This result is in line with the anti-proliferative activity of CPS49 and the actions of the novel compound Gu1215, both of which are heavily fluorinated IMiDs [20]. Notably, CPS49 has been reported to be a redox-reactive thalidomide analogue that induces cancer cell cytotoxicity by elevating reactive oxygen species [43]. This and other CRBN-independent mechanisms may underpin the anti-proliferative actions of compounds 1–4.

Anti-inflammatory actions in LPS-challenged RAW 264.7 cells:

Our prior studies have demonstrated that IMiDs can effectively mitigate LPS-mediated inflammation in RAW 264.7 cells – a monocyte/macrophage-like cell line that is often compared to peripheral macrophages and CNS microglia [21,30,44,45]. As shown in Fig. 5, all compounds in the present study significantly mitigated inflammation (reducing LPS-elevated nitrite and/or TNF-α levels), with the exception of compound 8. Notably, however, tetrafluorinated IMiDs 1–4 proved to be more potent than difluorinated 5–8, with the former inducing a decline in pro-inflammatory markers at a concentration of 1 μM whereas the latter required 10 μM (Fig. 5 upper and lower panels, respectively). This greater anti-inflammatory potency of the more fluorinated IMiDs 1–4, vs. 5–8, is intruiging in the light of the reported anti-inflammatory actions of phthalimide [46,47], as well as of monoterpenoids [27,28] and warrants further evaluation.

Of note, the anti-inflammatory activity of compounds 3 and 4 is in accord with our prior evaluation of the agents in LPS challenged RAW 264.7 cell cultures in which both agents demonstrated reductions in nitrite and TNF-α at concentrations as low as 100 nM [21] (in that study compounds 3 and 4 are called TFBP and tetrafluoronorbornylphthalimide (TFNBP), respectively [21]). This in vitro activity successfully translated into in vivo anti-inflammatory action in rats challenged with intraperitoneal LPS, a classic preclinical model of systemic and neuroinflammation [48]. Specifically, administration of each compound (at doses equivalent to 12.5 and 25 mg/kg pomalidomide) significantly mitigated proinflammatory markers in plasma and brain, and compound 3 (TFBP) significantly mitigated traumatic brain injury associated neuroinflammation and improved recovery of injury-induced motor coordination and balance impairments [21]. In support of the existence of CRBN-independent anti-inflammatory IMiD pathways [49,50], thalidomide, lenalidomide, and pomalidomide have been demonstrated to mitigate LPS-mediated inflammation in both normal and CRBN knockout mice [49]. Furthermore, the anti-inflammatory actions of IMiDs are retained in an ulcerative colitis model in both normal and humanized CRBN mice [51].

In synopsis, we describe a new series of monoterpenoid-based IMiDs 1–8 that lack binding to human CRBN, the primary drug target of thalidomide, lenalidomide, and pomalidomide that largely accounts for the valuable anticancer and at least some of the adverse teratogenic actions of these clinically used IMiDs [10–17]. Representative bicyclic monoterpenes are not only found in numerous plants and essential oils, but have been increasingly applied in drug development and possess a wide range of reported pharmacological activities, including antibiotic resistance modulation, antimicrobial, anti-Leishmania, antioxidant, anticoagulant, and analgesic effects, in addition to anticancer and anti-inflammatory actions [27,28,52,53]. Tetrafluorination of the phthalimide moiety resulted in compounds 1–4 that possessed promising anti-angiogenic and anti-proliferative (cytotoxic) actions, particularly in lenalidomide resistant U266 R10R MM cells. These compounds additionally proved more potent than their difluorinated equivalents in a cellular anti-inflammatory screen, and separate studies on compounds 3 and 4 have demonstrated that such efficacy successfully translates to in vivo models [21]. Specifically, 3 significantly lowered systemic and brain markers of inflammation in rodents challenged with LPS (a widely used model to screen for anti-inflammatory agents), and mitigated neuroinflammation induced by traumatic brain injury to, thereby, improve recovery [21]. Determination of CNS multiparameter optimization (MPO) scores [54] (Table 1) suggests that compounds 1–4 possess drug-like features (with values > 4.0 [54]). These MPO scores align with those for thalidomide (4.8), lenalidomide (5.4) and pomalidomide (4.8) that remain widely used clinical drugs in cancer aand other disorders [7, 8,55,56]. Whereas the mechanism of action of current clinically available IMiDs is reported to be largerly mediated via CRBN [10–16], the actions of compounds 1–4 clearly must be CRBN-unrelated and proteomic studies are ongoing to define the potential drug targets. In this light, compounds 1–4 not only warrant further pharmacological investigation as potential new drugs to mitigate MM and inflammatory disorders, but also as pharmacological tools to differentiate and characterize CRBN-dependent and -independent pathways of IMiD action.