Identification of Maleimide-Fused Carbazoles as Novel Noncanonical Bone Morphogenetic Protein Synergizers

Abstract

Morphogenic signaling pathways govern embryonic development and tissue homeostasis on the cellular level. Precise control of such signaling events paves the way for innovative therapeutic approaches in the field of regenerative medicine. In line with these notions, bone morphogenic protein (BMP) is a major osteogenic driver and pharmacological stimulation of BMP signaling holds supreme potential for diseases and defects of the skeleton. Efforts to identify small-molecule modalities that activate or potentiate the BMP pathway have primarily been focused on the canonical signaling cascade. Here, we describe the phenotypic identification and development of specific carbazolomaleimides 2 as novel noncanonical BMP synergizers with submicromolar osteogenic cellular potency. The devised chemical tools are characterized to specifically regulate Id gene expression in a SMAD-independent, yet highly BMP-dependent fashion. Mechanistic studies revealed that GSK3 inhibition and increased β-catenin levels are partly responsible for this activity. The utility of the new BMP synergizer profile was further exemplified by showing how the synergistic action of canonical and noncanonical BMP enhancers additively amplifies BMP-dependent osteogenic outputs. Carbazolomaleimide 2b serves as a new and unique pharmacological tool for the modulation and study of the BMP pathway.

1

Physiological processes of tissue development and homeostasis are orchestrated through a complex interplay of various growth factors and morphogenic signaling pathways. Precise pharmacological manipulation of these cell fate decisions provides potential therapeutic opportunities for difficult-to-treat diseases. The discovery of suitable small-molecule modalities remains notoriously challenging, but the reemerging field of phenotypic drug discovery (PDD) holds great potential to identify novel targets and mechanisms on the way to develop potent and selective growth factor mimetics.¹ The key requirement is a biological system that models cellular response, pathway dynamics, and feedback loops in an authentic cellular context. We have recently disclosed a stem-cell-driven screening platform that recapitulates bone morphogenetic protein (BMP) signaling during early embryogenesis.² With the identification of Chromenone 1 and CGS-15943 by successful screening and deconvolution, we already yielded two potent modulators of BMP signaling with distinct mechanisms and molecular targets (Figure 1).^2,3 Herein, the discovery of an interconnection between PI3K/CK1 pathways and BMP-SMAD dynamics underlined the power of PDD to not only identify new small-molecule agents but to unravel unprecedented processes of the underlying pathway biology.

Figure 1.

BMP pathway modulation by mechanistically distinct chemical probes. FK506 derivative oxtFK sequesters the inhibitory FK506 binding protein 12 (FKBP12) from BMP receptor type 1 (BMPRI) to enhance phospho-SMAD signaling in the canonical BMP pathway. CGS-15943 selectively inhibits casein kinase 1 (CK1) and phosphatidyl inositol-3 kinase (PI3K) isoforms, thereby amplifying SMAD abundancy and signaling. Chromenone 1 (target unknown, no kinase inhibition) and PD407824 (target: checkpoint kinase 1, CHK1) increase the availability of co-SMAD4 by downregulating the competing TGFβ-SMAD pathway. DIPQUO inhibits glycogen synthase kinase 3 (β) (GSK3β), leading to increased β-catenin levels and transcriptional regulation of pro-osteogenic genes. Created with biorender.com.

BMPs are members of the transforming growth factor-β (TGFβ) superfamily and possess key functions in physiological regulation from embryo development to adulthood.⁴⁻⁶ Despite their singular and name-giving osteoinductive activity, only recombinant BMP-2 (rBMP-2) is currently available for bone healing and regeneration in the clinic. Pharmacokinetic (PK) shortcomings necessitate supraphysiological dosing, resulting in severe adverse effects, considerable treatment cost, and ultimately limiting the clinical use of BMP.⁷ In contrast, small-molecule pathway activators promise improved PK properties and simultaneously offer chemically defined and reproducible tools for stem cell technology.

The canonical BMP pathway is triggered by ligand binding, leading to hetero-oligomerization of type I and type II BMP receptors (BMPRI, BMPRII) (Figure 1).⁸ The constitutively active BMPRII phosphorylates and activates BMPRI which subsequently phosphorylates receptor-regulated SMADs1,5,9 (R-SMADs) in the Mad homology (MH) 2 domain.⁹ SMADs are intracellular effectors of the TGFβ (SMAD2,3) and the BMP (SMAD1,5,9) pathways. They consist of the MH1 and MH2 domains connected via a linker. MH1 is the DNA-binding domain, whereas MH2 contains a conserved SXS motif that is phosphorylated by BMPRI.^9,10 The linker region is essential in the regulation of activity as it contains multiple phosphorylation and binding sites, for example, for ubiquitin ligases. Posttranslational modification of the linker can either enhance or attenuate signaling.^11,12 Phosphorylated R-SMADs associate with SMAD4 and these heterooligomers translocate to the nucleus where they interact with other transcription factors or bind to DNA directly.¹³ Besides the SMADs, there are less extensively studied noncanonical pathways transmitting BMP signals, including the TGFβ-activated kinase 1 (TAK1)-extracellular signal-regulated kinases (ERK) cascade as well as the PI3K-AKT axis.¹⁴⁻¹⁷ However, canonical and noncanonical BMP signaling are by no means separate pathways, but rather intricately intertwined.^11,17 Key target genes of BMP are the inhibitors of DNA binding (Id) which are essential mediators for the regulatory BMP functions.^18,19 Id1–4 are helix–loop–helix (HLH) proteins lacking the characteristic basic region of the bHLH transcription factors. Thus, they are unable to bind to DNA directly. Containing the HLH region, however, they can associate with members of the bHLH family, forming nonfunctioning dimers.²⁰ As a result, they act as dominant negative regulators of bHLH proteins like E proteins or myoblast determination protein 1 (MyoD). This way Id proteins are controlling key cell fate decisions.²¹ Moreover, additional mechanisms have been described that facilitate the degradation and posttranslational modification of target proteins.²²

Over the past decade, significant efforts have been made to identify small-molecule BMP mimetics (Figure 1). Mechanistically they differ in their ability to enhance BMP signals, and for many of the compounds, their target(s) and mechanism of action are not yet known, such as Isoliquiritigenin(23) and Ventromorphins.²⁴ Genuine pathway activators that fully mimic BMP ligands do not currently exist. However, FK506 analogue oxtFK(25) acts on the receptor level and increases canonical SMAD phosphorylation by enhancing BMPRI kinase activity through sequestration of inhibitory FK506 binding protein 12 (FKBP12). BMP potentiators sensitize the pathway by increasing the cellular response to subsequent BMP stimuli. This can be achieved by downregulating the TGFβ-SMADs, in turn increasing the availability of the common SMAD4, as shown for the checkpoint kinase 1 (CHK1) inhibitor PD407824(26) (PD) and more recently—in a kinase-independent fashion—for Chromenone 1.² BMP synergizers do not directly enhance canonical BMP signaling but act toward increased cellular BMP output. For example, the osteogenic DIPQUO(27) promotes differentiation by targeting the well-known Wnt effector glycogen synthase kinase 3 β (GSK3β). Wnt signaling is critically involved in osteogenesis and has been shown to enhance bone healing in vivo.^28,29 Several osteogenic BMP target genes, like alkaline phosphatase (Alp) and Runt-related transcription factor 2 (Runx2), have been identified to be commonly regulated by Wnt signaling.^30,31 In fact, a large variety of genes is coordinately controlled by β-catenin and SMAD regulatory motifs.³² Pathway crosstalk additionally takes place through the direct interaction of β-catenin-T cell factor/lymphoid enhancer factor (TCF/LEF) and SMAD protein complexes in the nucleus, adding another layer of interdependent regulation and context sensitivity.³³⁻³⁵

Herein, we describe the phenotypic identification, development, and functional characterization of maleimide-fused carbazoles (i.e., carbazolomaleimides) as a unique class of highly potent osteogenic BMP synergizers. The carbazolomaleimides are potent inducers of an osteogenic gene program by specific regulation of Id genes, a process at least partly mediated via GSK3β inhibition.

2. Results and Discussion

2.1 Discovery of Osteogenic 4-Indoyl Maleimides Enables Design of Novel BMP Modulators

Employing our recently disclosed phenotypic screening platform,^2,3 we identified a cluster of 3-aryl- and 4-indoyl-substituted maleimides from an in-house compound library, which were initially reported as angiogenesis inhibitors.³⁶ Here, the compounds struck our interest for their osteogenic induction potential (Alp activity, C2C12 cells). Initial analysis of a set of analogues under screening conditions allowed for early assessment of the structure–activity relationship (SAR) (Figure 2A). Substitution of the 3-aryl moiety with methoxy groups (R^1–3) drastically increased the potency and efficacy and was deemed necessary (see 1c–f and 1g). Methyl substitution of either indole or maleimide nitrogen (R⁴ and R⁵, see 1e and 1f) appeared to slightly increase osteogenic activity but the effect was negligible. A methyl group in R⁶ (1g) had little effect on activity, while a phenyl residue (1h) triggered pronounced cytotoxicity. Interestingly, bridging indole at R⁶ and the ortho-position of the 3-aryl moiety (i.e., benzo[a]-annulated pyrrolo[3,4-c]carbazole 2a) conferred highest activity and potency in osteogenic differentiation.

Figure 2.

Analysis of 4-indoyl maleimide cluster with discernable SAR informs pharmacophore fusion for BMP modulator design. (A) SAR-based hit validation revealed key structural features. 1d and 2a showed the most favorable profiles with high potency and clear dose dependency in osteogenic differentiation of C2C12 cells. Dose–response curves in 0.01, 0.05, 0.1, 0.5, 1, 5, and 10 μM. tox. = toxicity, BMP = 50 ng/mL BMP-4. (B) Fusing pharmacophore features from structurally related PD407824 with identified 4-indolyl maleimide 1d and carbazolomaleimide 2a to yield 2b and 1j as potential new BMP modulators.

This is in sharp contrast to their reported anti-angiogenic activity with the benzo[a]-annulated carbazoles being >10-fold less potent on kinase insert domain receptor/vascular endothelial growth factor receptor 2 (KDR/VEGFR2) and almost completely inactive on angiogenesis in an in vitro chick embryo model.³⁶ In fact, the most potent angiogenesis inhibitors derived from ring-opened derivatives. Substitution of R⁷ (i.e., formal position 11 in benzo[a]pyrrolo[3,4-c]carbazoles) with a methoxy group further decreased VEGFR2 inhibition activity.³⁶ In the herein-assessed context of BMP-triggered osteogenesis, an R⁷ methoxy group had no influence on bioactivity (see 1i). However, this position carries a hydroxy substituent in the structurally related BMP sensitizer PD407824. Hence, we reasoned to fuse these characteristic pharmacophoric features for the design of novel and potentially improved BMP-active compounds (Figure 2B): Compound 2b keeps the pyrrolo[3,4-c]carbazole scaffold of both PD407824 and 2a but adds the hydroxyl group from PD407824 and the trimethoxyphenyl annulant from 2a. Compound 1j was designed as a ring-opened derivative of 2b in order to evaluate the requirement of this annulation for BMP activity.

2.2 Synthesis of 4-Indoyl Maleimides (1) and Carbazolomaleimides (2)

Unsymmetrical diarylmaleimides 1 are accessible by condensation of aryl glyoxylates 3 with arylacetamides 4 or via Suzuki–Miyaura cross-coupling of maleimide bromides 5 and 7 with arylboronic acids (9) (Scheme 1A).³⁷⁻³⁹ While cross-couplings permit convenient syntheses due to commercially available starting materials, maleimide condensation reactions are often favored when N-unsubstituted maleimides are desired, largely owing to the typically much better yields.

Scheme 1. Synthesis of New 4-Indoyl Maleimides and Annulated Carbazolomaleimides.

(A) Overview of general synthetic routes toward unsymmetrical diarylmaleimides 1 and carbazolomaleimides 2. (B) Synthesis of 1d, 1j and 2a, 2b: (i) MOMCl, Cs₂CO₃, acetone, room temperature (rt), 16 h; (ii) pyridine, tetrahydrofuran (THF), 0 °C, 3 h; (iii) 1 M KO^tBu, THF, 0 °C, 16 h; (iv) para-toluenesulfonic acid (TsOH), MeOH/THF, rt, 16 h; (v) 365 nm (12 W), acetone, rt, 6–8 h; (vi) (1) SOCl₂, dimethylformamide (DMF) (cat.), THF, 40 °C, 30 min, (2) NH₃ (25% in H₂O).

Therefore, we prepared the required glyoxylates 3a and 3b and 2-(3,4,5-trimethoxyphenyl)acetamide (4a) as building blocks (Scheme 1B). For the synthesis of the envisioned hydroxy-functionalized diarylmaleimide 1j and carbazolomaleimide 2b, 5-hydroxyindole 6b was protected with MOM-Cl in a base-mediated nucleophilic substitution reaction with a yield of 65% (Figure 3B). Further glyoxylation of indoles 8 and 11 was performed according to literature procedures, furnishing maleimide precursors 3a and 3b in moderate yields (62–67%).⁴⁰ After several unsuccessful attempts to condensate indole glyoxylates 3a and 3b and phenylacetamide 4a, we identified KOH as the cause for low yields (<20%). KOH is often found as an impurity in solid KO^tBu and likely catalyzed a quick decomposition of the glyoxylates. Using KO^tBu in solution, eventually allowed reproducible preparation of the maleimides and moderate yields (45–46%).

Figure 3.

Osteogenic activity profile of 4-indoyl maleimides 1d and 1j and carbazolomaleimides 2a and 2b reveals potent BMP synergy. (A) Dose–response profiles of all compounds in the presence of low-dose BMP-4 (7.5 ng/mL) in C2C12 osteoblast differentiation (relative ALP activity) compared to high-dose BMP-4 (50, 100 ng/mL). Data normalized to dimethyl sulfoxide (DMSO) (7.5 ng/mL BMP-4), mean ± standard deviation (SD), n = 3. (B) Osteogenic activity of all compounds is highly BMP-dependent. Data normalized to 7.5 ng/mL BMP-4, mean ± SD, n = 3. (C) Expression profile of osteogenic marker genes in C2C12 verifies authentic BMP-like phenotype for 2b. Cells were treated with BMP-4 (7.5 ng/mL) and 2b (0.3 μM) or 2a (5 μM) for 72 h, BMP-4 (100 ng/mL) served as positive control, data normalized to DMSO (7.5 ng/mL BMP-4) and depicted as log 2-fold change, mean ± SD, n = 2. (D) Representative images of C2C12 cells after osteogenic differentiation show characteristic osteoblast “cobblestone” morphology for 2b (0.3 μM) and 2a (5 μM), identical to high-dose BMP-4 (100 ng/mL). (E) Differentiation induction inhibits proliferation. Cell numbers during C2C12 differentiation were measured 24 and 72 h after treatment with BMP or compound (together with 7.5 ng/mL BMP-4). Cells treated with DMSO (without BMP) were used as control (dotted line = 24 h, dashed line = 72 h). The green area indicates an increase in total cell numbers. BMP, 2b, and 2a reduce the proliferation of C2C12 cells as a function of differentiation induction. Data are shown as mean ± SD, n = 3. (F) 2b specifically regulates Id3 expression. Short-term regulation of BMP target genes Id1, Id3, and Runx2 in C2C12 cells by 2b was monitored over 24 h after treatment with BMP, 2b, or DMSO (both together with 7.5 ng/mL BMP-4).

To access carbazolomaleimides (i.e., benzo[a]pyrrolo[3,4-c]carbazol-1,3-diones 2a and 2b), methoxymethyl (MOM)-protected 3-aryl-4-indoyl maleimides 1d and 1k were cyclized in acetone via light-induced (λ = 365 nm, P = 12 W) 1,6-π-electrocylization reaction in high yields (77–82%).^41,42 In a last step, both desired hydroxy derivatives 1j and 2b were readily available after removal of the MOM group using para-toluenesulfonic acid (72–83%).

2.3 Phenotypic Profiling Validates BMP-Dependent Osteogenic Activity

Biological profiling of the newly synthesized compounds 2b and 1j revealed a conversed impact of the hydroxy group on osteogenic activity. The activity of 1j was reduced compared to 1d, showing only very little efficacy. On the other hand, 2b was ∼10-fold more potent than 2a with a half-maximal induction of Alp activity at 130 nM (Figure 3A). Interestingly, 2b displayed a bell-shaped dose–response curve with its peak around 0.3 μM after which the activity decreased. All tested compounds were highly BMP-dependent and did not induce Alp activity above background levels without the addition of low-dose BMP (Figure 3B).

These results suggest that BMP signaling is not activated directly but rather synergistically enhanced. Alp expression is an early but not exclusive marker of osteogenic phenotypes and we have recently reported how solely focusing on Alp activity might lead to false positives.^2,43 Hence, we verified the authenticity of the observed phenotype via expression of osteogenic marker genes osterix (Osx), osteocalcin (Ocn), collagen type I α 1 (Col1a1), Runx2 and Alp as well as morphological characterization. All osteogenic markers were upregulated on the messenger ribonucleic acid (mRNA) level in response to 2b treatment in a comparable fashion to high doses of BMP (i.e., 100 ng/mL) (Figure 3C). To our surprise, 2a only induced Alp expression, whereas other markers were unchanged compared to DMSO. Because of its lower potency in the C2C12 Alp assay, 2a had to be dosed 15-fold higher at 5 μM compared to only 0.3 μM for 2b. This likely exacerbated counterproductive off-target effects, thus impairing correct differentiation. BMP-dependent osteogenic differentiation generates a characteristic “cobblestone-like” phenotype that can be visually assessed.²⁴ Indeed, 2b- and 2a-induced morphology was strikingly similar to that induced by high-dose BMP, even though 2a did not drive osteogenic marker expression (Figure 3D).

Together, these observations confirmed BMP-like osteogenesis for 2b and stress the importance of multiplexing phenotypic readouts. A general hallmark of cell differentiation induction is the associated and typically opposing regulation of proliferation.^44,45 Accordingly, BMP reduces the proliferative capacity of C2C12 cells in a dose-dependent manner and completely stops proliferation at 100 ng/mL BMP-4 (Figure 3E). For small-molecule BMP enhancers, we expected a similar effect where the strongest impact should be observed in concentrations corresponding to the highest BMP effect. Indeed, 2b reduced the proliferation of C2C12 cells in a dose-dependent manner and as a function of its BMP activity. An application of 0.3 μM 2b induced Alp activity almost as strong as 100 ng/mL BMP-4 (Figure 3A) and simultaneously also reduced proliferation to a comparable extent (Figure 3E). At higher concentrations of 2b, overall cell numbers began to decrease and proliferation was halted completely (toxicity >1 μM). Notably, in this concentration range Alp activity also subsided, possibly indicating prevailing off-target effects. 2a influenced the proliferation of C2C12 cells in the same manner but at higher concentrations, corresponding to its lower potency in the C2C12 Alp assay.

To uncover the mechanism behind the induction of osteogenic differentiation we monitored the expression of direct BMP target genes Id1, Id3, and Runx2 from 2 to 24 h after treatment (Figure 3F). Id1 and Id3 responded rapidly to BMP, with their highest induction already reached after 2 h. Runx2 levels were also elevated, although to a lower extent and remained rather constant over the observed time. During osteoblastogenesis, Runx2 is a central regulatory element and it interacts directly with proteins like BMP-SMADs, Menin, and Smurf1 to facilitate differentiation such that the expression of Runx2 is precisely regulated explaining the overall low induction window.⁴⁶Id3 was the only gene that responded to 2b differently than to DMSO with a time-dependent, moderate increase visible after approximately 15 h. Id genes and Runx2 have been described to be direct targets of BMP-dependent SMAD signaling.^18,47 However, noncanonical signaling events, including tyrosine kinases and the ERK-mitogen-activated protein kinase (MAPK) cascade, are known to regulate their expression as well.⁴⁸⁻⁵⁰ The fact that Id1 expression was not affected strongly suggested that the canonical BMP pathway cannot be the main mechanism behind the osteogenic properties of the carbazolomaleimides 2.

To this end, rational pharmacophore design led to 2b, a highly osteogenic compound with submicromolar cellular activity in C2C12 differentiation. Morphological characterization validated an authentic osteogenic phenotype with enhanced markers Alp, Osx, Ocn, and Runx2. Differentiation was induced in synergy with BMP and did highly depend on active BMP signaling.

2.4 Carbazolomaleimides 2 Trigger a Unique Gene Program in Part through GSK3β Inhibition

To verify that canonical BMP signaling was not affected by treatment with compound series 2, phospho-SMAD dynamics were monitored after varying preincubation times. SMAD1 was not affected under any tested condition (Figure 4A). PD407824 enhances BMP-SMAD signaling indirectly by downregulation of competing TGFβ-SMAD2. Based on our pharmacophore fusion approach (Figure 2B), we also tested whether SMAD2 was targeted but observed no change in protein levels (Figure S2). In contrast to Id1, Id3 is positively regulated through Wnt/β-catenin signaling.⁵¹ The predominant pharmacological strategy to activate the Wnt/β-catenin pathway is by means of small-molecule GSK3β inhibitors such as the commonly used CHIR99021 (CHIR).⁵² To test whether the herein disclosed carbazolomaleimides 2 also function through this mechanism, we tested their GSK3β inhibition in vitro. 2b and 2a were tested at 0.5 μM, and both potently inhibited GSK3β to 96.0 ± 0.4 and 97.3 ± 0.04%, respectively (n = 2). To functionally validate these results, we monitored β-catenin accumulation and nuclear translocation following compound treatment in C2C12 cells. Both compounds caused a rapid increase in cytosolic and nuclear β-catenin levels, confirming that they activate β-catenin signaling in living cells (Figure 4C). Interestingly, in our experience typical GSK3β inhibitors can induce Alp activity in C2C12 cells, but they do not require an active BMP pathway. Furthermore, the resulting phenotype does not match BMP-dependent osteogenic differentiation. CHIR as well as the recently identified DIPQUO strongly induced Alp activity with and without addition of BMP; in contrast, 2b and 2a did not (Figures 3A,B and 4B).^27,52 Since its discovery, CHIR has emerged as a gold standard to activate Wnt/β-catenin signaling. With an IC₅₀ of 6–7 nM in vitro, it is a potent GSK3α/β inhibitor. Surprisingly though, in most complex cellular systems CHIR requires concentrations ∼1000-fold above its IC₅₀ to show biological activity.^27,52−54 Thus, despite its universally accepted application as a tool compound to probe Wnt/β-catenin signaling, it is disputable whether observed biological effects (including the herein presented data) can be attributed solely to GSK3 inhibition.⁵⁵ Moreover, we compared the phenotypes induced by 2b and CHIR, with and without BMP. The induction of Alp by CHIR was clearly visible on the mRNA level, but none of the other osteogenic markers were increased. On the contrary, expression of Osx, Ocn, and Runx2 displayed a tendency to be slightly inhibited (Figure 4D). Additionally, the increase in Alp expression by CHIR was independent of the presence of BMP, suggesting unrelated differentiation pathways for CHIR- and BMP-dependent Alp induction. On the other hand, 2b produced an expression profile similar to CHIR when used without BMP, but crucially, enhanced the expression of all osteogenic markers when BMP was administered simultaneously. Alp expression was increased without BMP but was further amplified by co-treatment, strongly indicating BMP synergy. Morphological comparison revealed clear differences between GSK3β inhibitor- and BMP-dependent phenotypes, supporting the quantitative polymerase chain reaction (qPCR) data (Figures 4E and S3).

Figure 4.

Carbazolomaleimides 2 exhibit a unique osteogenic profile, partly through targeting GSK3. (A) 2a and 2b do not influence canonical SMAD signaling: C2C12 cells were preincubated with compounds for 0 h, 2 h (without BMP-4), or 24 h (together with 7.5 ng/mL BMP-4) followed by stimulation with 7.5 ng/mL BMP-4 for 30 min after which cells were lysed and nuclear phospho-SMAD1 quantified. Data normalized to DMSO, 25 ng/mL BMP-4 served as positive control, mean ± SD, n = 3. (B) GSK3β-inhibitors induce ALP activity in C2C12 cells in the absence of BMP-4: Dose–response curves for CHIR and DIPQUO in the presence or absence of BMP-4 (7.5 ng/mL BMP-4). Data normalized to DMSO (7.5 ng/mL BMP-4), mean ± SD, n = 3. (C) 2a and 2b enhance cytosolic and nuclear levels of β-catenin: Immunoblotting 3 and 20 h after treatment with CHIR (5 μM), 2a (5 μM), 2b (0.3 μM), or 25 ng/mL BMP-4. Data normalized to untreated cells, mean ± SD, n = 3. (D) 2b (0.3 μM) induces expression of osteogenic markers in a BMP-dependent fashion, while CHIR enhances Alp expression selectively and independent of BMP: Cells were treated with or without BMP-4 (7.5 ng/mL) and 2b (0.3 μM) or CHIR (1 and 5 μM) for 72 h, BMP-4 (100 ng/mL) served as positive control. Data normalized to DMSO (7.5 ng/mL) and depicted as log 2-fold change, mean ± SD, n = 3. (E) Representative images of C2C12 cells after osteogenic differentiation show a distinct morphology for GSK3β-inhibitors (all 5 μM) that does not resemble BMP-4-induced differentiation. (F) 2b induces a specific gene program: Short-term regulation of BMP target genes in C2C12 cells by 2b (0.3 μM), 2a (5 μM), CHIR (1 and 5 μM), and DIPQUO (1 and 5 μM) was measured after 24 h of treatment with compounds and 7.5 ng/mL BMP-4. Data depicted as log 2-fold change mean ± SD, n = 4.

To shed light onto the apparently distinct mechanisms, we took a more detailed look at the regulation of direct BMP target genes Id1,2,3. 2b showed specific upregulation of Id2 and Id3, whereas expression of Id1 was slightly inhibited (Figure 4F). CHIR downregulated all three Id genes. This was in agreement with the divergent phenotype that CHIR induced compared to BMP. Surprisingly, DIPQUO (at 5 μM) triggered a slight upregulation of all Id genes suggesting mechanistic differences between “classic” GSK3 inhibitors and DIPQUO in osteogenic differentiation.

Wnt signaling through β-catenin accumulation followed by nuclear translocation is recognized as a major player in cell fate decisions. This holds also true for osteoblastogenesis. BMP and Wnt share several target genes like Alp and Runx2 and act synergistically in osteogenic differentiation.^56,57 Salazar et al. reported that Wnt alone is insufficient and requires BMP signaling to form functioning osteoblasts.⁵⁸ This is coherent with CHIR failing to induce an osteogenic phenotype in our experiments. It could be speculated that in lower dosage and in combination with different levels of BMP CHIR would in fact direct C2C12 differentiation toward an osteogenic fate. Interestingly, Id2 and Id3 expression are both increased by Wnt/β-catenin signaling, independently of SMADs.^51,59,60

On the other hand, Id1 is not upregulated by Wnt/β-catenin signaling. In fact, overexpression of Wnt3a in C2C12 cells has previously been shown to reduce Id1 expression and this was mediated by the direct interaction of β-catenin and SMAD1,4 complexes.⁶¹

To this extent, 2b showed GSK3β inhibition, functionally increased β-catenin signaling, and synergistically regulated Id2 and Id3 expression. While Id proteins share some functional redundancy, they are less conserved outside the HLH region, permitting distinct functions and binding partners for individual Id members.⁶² Additionally, Peng et al. found that while expression of Id genes was necessary for osteogenic induction, their overexpression abolished differentiation.⁶³ This suggests dual roles for Ids in the modulation of proliferation versus differentiation and that precise regulation is required to facilitate osteogenic differentiation. In this context, specific roles for individual Id genes have not clearly been established, but downregulation of Id1 with simultaneous expression of Id2,3 appears to amplify osteogenic differentiation.

Together, our data outlines a mechanism through which 2b acts in a BMP-dependent synergy to induce efficient osteogenic differentiation by inducing a unique gene program by specific regulation of Id genes. This effect was in part exerted by GSK3β inhibition and enhanced β-catenin signaling, yet decisive differences distinguishing typical GSK3β inhibitors were observed concerning the mechanism as well as the induced phenotype. Importantly, while canonical BMP signaling was not influenced, 2b acts in a strictly BMP-dependent fashion.

2.5 Carbazolomaleimide 2b Potently Synergizes with Canonical BMP Potentiator Chromenone 1

Chromenone 1 efficiently potentiates BMP signaling through downregulation of the TGFβ-SMAD pathway, in turn increasing canonical BMP-SMAD outputs. While the molecular target of Chromenone 1 is not known, kinase inhibition was excluded as a possible mechanism. This profile makes Chromenone 1 the best available tool to selectively enhance canonical BMP signals without interfering with noncanonical kinase cascades. On the other hand, 2b efficiently enhanced BMP-dependent osteogenic differentiation without affecting the SMAD pathway. We reasoned that combining these distinct BMP modulators would yield a synergistic efficacy on osteogenesis. We performed two-dimensional dose–response experiments and used SynergyFinder to detect synergistic effects according to the zero interaction potency (ZIP) model.⁶⁴

The ZIP model assumes noninteraction for the given compounds and calculates the delta score as a semiquantitative indicator for the additional effects of the combination compared to the individual compounds.⁶⁵ Indeed, the combination of Chromenone 1 with either 2b or 2a resulted in highly overadditive effects and increased potency for both compounds (Figure 5A,B). The highest synergy scores (i.e., delta scores) were observed for 0.5–1 μM Chromenone 1 and 2–3 μM 2a or around 0.3 μM 2b, respectively (Figure S4). The most synergistic concentrations almost doubled the Alp activity compared to the sum of the individual compounds.

Figure 5.

Chromenone 1 and carbazolomaleimides 2a and 2b synergistically drive osteogenic differentiation. (A) Two-dimensional dose–response experiments in C2C12 cells revealed a highly potent synergy between canonical potentiator Chromenone 1 and 2a and 2b: Osteogenic differentiation (Alp activity) was increased beyond the level of 100 ng/mL BMP-4 (indicated as a dotted line). Data normalized to DMSO (7.5 ng/mL BMP-4), mean, n = 3. (B) Synergistic effects exceed additive effects of co-administered compounds: Dose-dependent activity values for Chromenone 1 and 2b or 2a were subtracted (=compound without Chromenone 1 forms a flat line at “0”). Increasing concentrations of Chromenone 1 (0.05, 0.2, 0.5, or 1 μM) were added. Every value above 0 indicates the overadditive effects of the co-treatments. Calculation of the EC₅₀-values from the two-dimensional dose–response experiments demonstrated increased potency (reduced EC₅₀) when combining 2a or 2b with Chromenone 1. This effect was especially pronounced for 2a, which has an approximately 10-fold higher EC₅₀ than 2b. Error bars indicate 95% confidence interval.

Combination therapy is common practice in metabolic disorders, infectious disease, and the treatment of cancer.⁶⁶⁻⁶⁸ The application of synergistic drugs has the potential to increase the efficacy and reduce the required dosage of individual drugs, therefore increasing safety margins. Chromenone 1 and carbazolomaleimide 2b clearly address distinct, unique targets and mechanisms within BMP-driven cell fate modulation. Their synergistic profile nicely demonstrates how targeting different signaling pathway branches can drastically enhance downstream output beyond the capabilities of the individual modulators.

3. Conclusions

Small-molecule mimetics or stimulators of morphogenic cytokine pathways are challenging to find and to develop further. Precise modeling of the underlying complex pathway regulations is key to identifying high-quality chemical probes and new pharmacological agents. Still, several screening campaigns have already yielded BMP-stimulating modalities and contributed to the understanding of pathway dynamics and regulators. However, past efforts have largely focused on the amplification of canonical (SMAD) BMP, possibly owing to the convoluted nature and context sensitivity of noncanonical signaling events. In this work, we describe the phenotypic identification and development of carbazolomaleimides 2 as noncanonical BMP synergizers with submicromolar osteogenic potency that act in a SMAD-independent fashion. GSK3β inhibition and as a result, enhanced β-catenin signaling, appeared to be a major mediator for the amplified osteogenic differentiation. Notably, we observed striking differences between BMP- and solely β-catenin-driven C2C12 differentiation. Compound 2b exhibited a very distinct activity profile compared to conventional GSK3 inhibitors: It highly depends on BMP ligand input and selectively modulates the expression of specific Id genes to trigger an osteogenic gene program, resulting in the regulation of Alp, Osx, Ocn, and Runx2. The 2b-induced cellular phenotypes closely resembled those of high-dosed BMP.

To the best of our knowledge, such a SMAD-independent, noncanonical, yet BMP stimuli-dependent synergizer mechanism has not been described for small-molecule BMP modulators. This profile perfectly qualifies for combination therapy, which was ultimately demonstrated by the highly synergistic induction of osteoblastogenesis using 2b and a canonical BMP potentiator. Simultaneously targeting the canonical and noncanonical branch through synergistic pharmacological tools, might prove a superior dual therapeutic approach to direct cell fate in a more holistic manner and will be interesting to investigate for BMP-relevant disease states in the future.

4. Experimental Section

4.1 Chemistry

4.1.1. General Information

All commercially available compounds were used as provided without further purification: CHIR99021 (#1386, AxonMedchem), 6-bromo indirubin-3′-oxime (BIO) (#3194, Tocris), and DIPQUO (#SML3335, Sigma-Aldrich). Solvents for chromatography were technical grade. Analytical thin-layer chromatography (TLC) was performed on Macherey-Nagel silica gel polyester plates with F-254 indicator, visualized by irradiation with UV light. Flash column chromatography was performed using silica gel columns (particle size 0.015–0.050 mm) provided by Interchim (Montluçon, France) or Teledyne Isco (Axel Semrau GmbH & Co KG, Sprockhövel, Germany). Semipreparative high-performance liquid chromatography (HPLC) was performed on a PuriFlash4250 using RP-columns (Uptisphere Strategy C18-HQ) both by Interchim. Irradiation was performed in a Biometra UV chamber (Analytik Jena GmbH & Co KG, Jena, Germany) without stirring in an open glass dish at λ = 360 nm (12 W). ¹H NMR and ¹³C NMR were recorded on Bruker Avance III 400 (400 MHz, Software: Bruker TopSpin 3.6.0) spectrometers in DMSO, CDCl₃, or acetone-d₆. Data are reported in the following order: chemical shift (δ) in ppm; multiplicities are indicated s (singlet), d (doublet), t (triplet), m (multiplet), and m_c (centered multiplet); coupling constants (J) are given in hertz (Hz). High-resolution mass spectra were recorded on a micrOTOFII (Software: Bruker Compass for otofSeries, otofControl Version 3.4) mass spectrometer. Chemical yields refer to isolated pure substances unless otherwise noted. The purity of the synthesized compounds was determined by HPLC via calculating the percentage of the product peak integral relative to the sum of all observed peak integrals at 254 nm. The HPLC purity for all synthesized compounds was >95% for biological testing, unless otherwise noted. The detailed physical/spectroscopic data can be found in the extended experimental section of the Supporting Information.

4.1.2. General Procedure A: Indole Acylation⁴⁰

The corresponding indole (1 equiv) was dissolved in dry pyridine (1.3 equiv) and dry THF under a nitrogen atmosphere. Ethyl 2-chloro-2-oxoacetate (2 equiv) was added dropwise over 10 min at 0 °C and the suspension was stirred at 0 °C. After 2.5–5 h the solids were filtered off and worked up accordingly.

4.1.3. General Procedure B: Maleimide Formation³⁶

The corresponding indole-3-ethyl glyoxylate (1.4 equiv) as a suspension in dry THF (40 mL) was added dropwise to a mixture of molecular sieve (3 Å, 4 g) and 2-(3,4,5-trimethoxyphenyl) acetamide (1 equiv) in dry THF (20 mL) under a nitrogen atmosphere at 0 °C. Over a period of 20 min, a solution of KO^tBu (1 M in THF, 3.8 equiv) was added dropwise. After completing addition, the purple mixture was stirred at room temperature for 16 h. The reaction was quenched with NH₄Cl (saturated aqueous solution), filtered and the filtrate extracted with EtOAc (3 × 100 mL). Combined organic phases were dried over Na₂SO₄ and concentrated in vacuo. The crude product was purified by flash chromatography to afford the corresponding maleimides as colorful solids.

4.1.4. General Procedure C: MOM Deprotection

MOM-protected indole (1 equiv) and TsOH·H₂O (10 equiv) were dissolved in MeOH/THF (1:1) and stirred at room temperature for 18 h. After completion, the reaction was quenched with NaHCO₃ (saturated aqueous solution) and extracted with EtOAc (2 × 20 mL). The crude product was purified by flash chromatography to afford the corresponding alcohols as a colorful solid.

4.1.5. General Procedure D: 1,6-Electrocyclization

The corresponding maleimide was dissolved in acetone and irradiated at λ = 360 nm (12 W) in a UV chamber until complete consumption (3–8 h) of the starting material. The crude product was purified by flash chromatography to afford the corresponding carbazoles as colorful solids.

4.1.5.1. Synthesis of 5-(Methoxymethoxy)-1H-indole (11)

5-Hydroxyindole (10) (500 mg, 3.75 mmol) and Cs₂CO₃ (2.44 g, 7.50 mmol) were suspended in acetone (18 mL) under a nitrogen atmosphere. To the suspension, MOMCl (285 μL, 3.75 mmol) was added dropwise over a period of 5 min and the reaction mixture was stirred at room temperature for 16 h. The reaction was quenched by the addition of NaHCO₃ (saturated aqueous solution). After extraction with dichloromethane (DCM) (3 × 30 mL), the combined organic phases were dried over Na₂SO₄. The crude was purified by chromatography (5–50% EtOAc/Cy) to obtain 11 as a colorless oil. The product turns brown after storing for more than 48 h. Yield: 411 mg (2.32 mmol, 62%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 10.98 (br. s, 1 H), 7.31–7.26 (m, 2 H), 7.16 (d, ⁴J = 2.3 Hz, 1 H), 6.80 (dd, ³J = 8.6 Hz, ⁴J = 2.3 Hz, 1 H), 6.33 (m_c, 1 H), 5.13 (s, 2 H,), 3.39 (s, 3 H) ppm. ¹³C NMR (101 MHz, DMSO-d₆, 300 K): δ = 151.0, 132.3, 128.4, 126.5, 113.1, 112.3, 106.6, 101.3, 95.6, 55.7 ppm. LC-MS (ESI): m/z (%) = 200.0 (25) [M + Na]⁺, 178.0 (27) [M + H]⁺, 146.0 (100) [M – OCH₃]⁺.

4.1.5.2. Synthesis of Ethyl 2-(5-(Methoxymethoxy)-1H-indole-3-yl)-2-oxoacetate (3b)

Ethyl 2-(5-(methoxymethoxy)-1H-indole-3-yl)-2-oxoacetate (3b) was synthesized from 11 (200 mg, 1.13 mmol) and pyridine (546 μL, 6.78 mmol) following General Procedure A. The yellow solid was washed with water (30 mL), and the crude was purified by chromatography (5% MeOH/DCM) to obtain 3b as a yellow solid. The solid turns red after heating at 40 °C. Yield: 150 mg (540 μmol, 48%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 12.3 (br. s, 1 H), 8.37 (s, 1 H), 7.79 (d, ⁴J = 2.4 Hz, 1 H), 7.46 (d, ³J = 8.8 Hz, 1 H), 7.00 (dd, ³J = 8.8 Hz, ⁴J = 2.4 Hz, 1 H), 5.21 (s, 2 H), 4.35 (q, ³J = 7.1 Hz, 2 H), 3.40 (s, 3 H), 1.33 (t, ³J = 7.2 Hz, 3 H) ppm. ¹³C NMR (101 MHz, DMSO-d₆, 300 K): δ = 179.0, 163.7, 153.7, 138.5, 132.2, 126.3, 114.8, 113.5, 112.3, 107.1, 94.7, 61.6, 55.5, 14.0 ppm. LC-MS (ESI): m/z (%) = 300.0 (100) [M + Na]⁺, 278.1 (96) [M + H]⁺, 246.0 (14) [M – OCH₃]⁺.

4.1.5.3. Synthesis of 2-(3,4,5-Trimethoxyphenyl)acetamide (4a)

2-(3,4,5-Trimethoxyphenyl)acetamide (4a) was synthesized according to the literature. Spectroscopic data are in agreement with the literature.⁴¹

4.1.5.4. Synthesis of 3-(5-(Methoxymethoxy)-1H-indole-3-yl)-4-(3,4,5-trimethoxyphenyl)-1H-pyrrole-2,5-dion (1k)

1k was synthesized from 4a (200 mg, 889 μmol) and 3b (394 mg, 1.42 mmol) following General Procedure B. The crude product was purified by chromatography (5–30% EtOAc/Cy) to obtain 1k as red crystals. Yield: 175 g (400 mmol, 45%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 11.84 (d, ³J = 2.4 Hz 1 H), 11.03 (s, 1 H), 7.98 (d, ³J = 3.2 Hz, 1 H), 7.35 (d, ³J = 8.7 Hz, 1 H), 6.79 (dd, ³J = 8.8 Hz, ⁴J = 2.3 Hz, 1 H), 6.71 (s, 2 H), 6.01 (d, ⁴J = 2.6 Hz, 1 H), 4.69 (s, 2 H), 3.67 (s, 3 H), 3.42 (s, 6 H), 3.19 (s, 3 H) ppm. ¹³C NMR (101 MHz, DMSO-d₆, 300 K): δ = 172.5, 172.2, 152.3, 150.9, 138.0, 132.0, 131.9, 131.8, 127.5, 125.7, 124.3, 113.5, 112.7, 107.7, 104.1, 94.5, 60.0, 55.6, 55.1 ppm. LC-MS (ESI): m/z (%) = 407.2 (100) [M – OCH₃]⁺, 461.2 (30) [M + Na]⁺.

4.1.5.5. Synthesis of 3-(5-Hydroxy-1H-indole-3-yl)-4-(3,4,5-trimethoxyphenyl)-1H-pyrrole-2,5-dion (1j)

1j was synthesized from 1k (56 mg, 128 μmol) following General Procedure C. The crude product was purified by chromatography (5–100% EtOAc/DCM) to obtain 1j as orange crystals. Yield: 42 mg (106 μmol, 83%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 11.65 (d, ²J = 2.6 Hz, 1 H), 10.97 (s, 1 H), 8.59 (s, 1 H), 7.87 (d, ²J = 3.1 Hz, 1 H), 7.23 (d, ³J = 8.69 Hz, 1 H), 6.74 (s, 2 H), 6.61 (dd, ³J = 8.7 Hz, ⁴J = 2.4 Hz, 1 H), 5.71 (d, ⁴J = 2.2 Hz, 1 H), 3.68 (s, 3 H), 3.41 (s, 3 H) ppm. ¹³C NMR (101 MHz, DMSO-d₆, 300 K): δ =172.6, 172.3, 152.1, 151.2, 138.2, 132.0, 131.4, 130.6, 126.8, 125.4, 124.8, 112.3, 112.0, 107.6, 106.6, 103.7, 60.2, 55.4 ppm. HRMS (ESI): m/z calcd for C₂₁H₁₈N₂O₆Na [M + Na]⁺: 417.1062, found: 417.1064.

4.1.5.6. Synthesis of 5,6,7-Trimethoxy-11-(methoxymethoxy)benzo[a]pyrrolo[3,4-c]carbazole-1,3(2H,8H)-dion (2c)

2c was synthesized from 1k (100 mg, 228 μmol) in acetone (300 mL) following General Procedure D with an irradiation time of 3 h. The crude product was purified by chromatography (30–100% EtOAc/Cy) to obtain 2c as an orange solid. Yield: 82 mg (188 μmol, 82%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 11.83 (s, 1 H), 11.09 (s, 1 H), 8.60 (d, ⁴J = 2.5 Hz, 1 H), 8.29 (s, 1 H), 7.81 (d, ³J = 8.8 Hz, 1 H), 7.26 (dd, ³J = 8.8 Hz, ⁴J = 2.5 Hz, 1 H), 5.27 (s, 2 H), 4.20 (s, 3 H), 4.00 (s, 3 H), 3.96 (s, 3 H), 3.46 (s, 3 H) ppm. ¹³C NMR (101 MHz, DMSO-d₆, 300 K): δ = 171.7, 170.5, 154.6, 151.4, 148.6, 141.4, 139.0, 135.6, 127.9, 123.9, 120.9, 116.9, 116.6, 113.2, 113.2, 111.5, 110.1, 100.1, 95.2, 61.6, 61.0, 55.9, 55.5 ppm. LC-MS (ESI): m/z (%) = 437.2 (100) [M + H]⁺, 405.1 (34) [M – OCH₃]⁺, 391.3 (100) [M – CH₂OCH₃]⁺.

4.1.5.7. Synthesis of 11-Hydroxy-5,6,7-trimethoxybenzo[a]pyrrolo[3,4-c]carbazole-1,3(2H,8H)-dion (2b)

2b was synthesized from 2c (49 mg, 92 μmol) following General Procedure C. The crude product was purified by chromatography (5–100% EtOAc/DCM) to obtain 2b as orange crystals. An analytical sample was purified by semipreparative reversed-phase high-performance liquid chromatography (RP-HPLC) (H₂O/acetonitrile (ACN), 10 → 90%). Yield: 28 mg (66 μmol, 78%). ¹H NMR (400 MHz, DMSO-d₆, 300 K): δ = 11.68 (s, 1 H), 11.02 (s, 1 H), 9.19 (s, 1 H), 8.33 (d, ⁴J = 2.4 Hz, 1 H), 8.27 (s, 1 H), 7.69 (d, ³J = 8.7 Hz, 1 H), 7.01 (dd, ³J = 8.7 Hz, ⁴J = 2.4 Hz, 1 H), 4.17 (s, 3 H), 3.99 (s, 3 H), 3.95 (s, 3 H) ppm. ¹³C NMR (400 MHz, DMSO-d₆, 300 K): δ = 171.8, 170.7, 154.4, 151.8, 148.5, 141.3, 138.8, 133.9, 128.0, 123.8, 121.2, 116.1, 115.9, 113.2, 113.0, 111.5, 108.0, 100.1, 61.6, 60.9, 55.9 ppm. HRMS (ESI): m/z calc. for C₂₁H₁₆N₂O₆Na [M + Na]⁺: 415.0906, found: 415.0891.

4.1.5.8. Synthesis of Ethyl-2-(1H-indol-3-yl)-2-oxoacetate (3a)

Ethyl-2-(1H-indol-3-yl)-2-oxoacetate (3a) was synthesized according to the literature (General Procedure A).⁴⁰ Spectroscopic data are in agreement with the literature.³⁶

4.1.5.9. Synthesis of 3-(1H-Indole-3-yl)-4-(3,4,5-trimethoxyphenyl)-1H-pyrrole-2,5-dion (1d)

3-(1H-Indole-3-yl)-4-(3,4,5-trimethoxyphenyl)-1H-pyrrole-2,5-dion (1d) was synthesized according to the literature (General Procedure B). Spectroscopic data are in agreement with the literature.³⁶

4.1.5.10. Synthesis of 5,6,7-Trimethoxybenzo[a]pyrrolo[3,4-c]carbazole-1,3(2H,8H)-dion (2a)

2a was synthesized from 1d (190 mg, 502 μmol) in acetone (250 mL) following General Procedure D with an irradiation time of 7.5 h. The crude product was purified by chromatography (30% EtOAc/Cy) to obtain 2a as an orange solid. An analytical sample was purified by semipreparative RP-HPLC (H₂O/ACN, 10 → 90%). Spectroscopic data are in agreement with the literature.³⁶ Yield: 146 mg (388 μmol, 77%).

4.2. Cell Culture

For routine culture and experiments cells were kept in a humidified incubator at 37 °C with 5% CO₂. C2C12 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (#11960, Gibco) supplemented with 10% fetal bovine serum (FBS) (#10270106, Gibco) and 1% GlutaMAX (#35050061, Gibco). The cells were passaged every 2–3 days before a confluency of approximately 70% was reached to avoid spontaneous differentiation.

4.3. Murine Osteoblast Differentiation Assay

C2C12 cells were seeded at 2000 cells per well in 384-well plates (#781080, Greiner) in DMEM (#11960-044, Thermo Fisher) supplemented with 6% heat-inactivated FBS and 1% GlutaMax. The cells were grown for 24 h, and compounds or DMSO were added in quadruplicates or sextuplicates. Unless stated otherwise, BMP-4 (#314-BP-050, R&D Systems) was added to the wells (7.5 ng/mL) directly after compound addition and cells were incubated for another 72 h. ALP activity was determined via fluorescence read-out. 4-Methylumbelliferylphosphate (#M8168, Sigma-Aldrich) was dissolved in phosphate-buffered saline (PBS) and diluted to a final concentration of 0.5 mM in lysis buffer (Tris base 100 mM, NaCl 250 mM, MgCl₂ 25 mM, Triton X-100 1%, pH 9.5), and after aspiration of the media, 25 μL of this dilution were added per well. Plates were shaken for 2 min and 10 μL of an ethylenediaminetetraacetic acid (EDTA) solution in lysis buffer was quickly added (final EDTA concentration in each well was 35 mM). Plates were shaken again for 1 min and briefly centrifuged. Fluorescence measurement was performed with a Tecan Spark (Tecan) (λ_Excitation 364 nm, λ_Emission 448 nm).

4.4. Quantitative Reverse Transcription Polymerase Chain Reaction (PCR)

In general, cells were handled and treated analog to the osteoblast differentiation assay. C2C12 cells were plated in 12-well plates (#83.3921, Sarstedt) at 8 × 10⁴ cells per well, incubated for 24 h, and treated as indicated. RNA was isolated 72 h after treatment or at the indicated time points. The cells were washed once with PBS and lysis, and RNA isolation was performed using the RNeasy Mini Kit (#74106 and #79256, Qiagen) according to the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized using the qScript cDNA SuperMix (#95048-100, Quantabio) on a C1000 Thermal Cycler (BioRad). qPCR was performed with Takyon SYBR Master Mix (#UF-NSMT-B0701, Eurogentec) on a CFX Opus 96 (BioRad). Primers were purchased from IDT (Integrated DNA Technologies) sequences are listed in Table S1. To determine relative gene expression, Cq values were normalized to GAPDH expression according to the ^ΔΔCq method. To produce more accurate results, experimental data shown in Figure 5F was normalized to GAPDH and Actb. The observed effects were small and normalization to more than one housekeeping gene produces more reliable results.⁶⁹

4.5. Proliferation Assay

C2C12 cells were seeded and treated as described for the osteoblast differentiation assay in 384-well plates (#781091, Greiner). 24 or 72 h after treatment media was removed, and the cells were fixed with 4% paraformaldehyde (PFA) in PBS for 15 min, stained with Hoechst33342, washed three times with PBS, and imaged with an ImageXpress Micro XL (10× objective, Molecular Devices). Images were analyzed using the custom module editor in Metaxpress 6.

4.6. Immunoblotting

The general blotting procedure was conducted as described in detail before.² In brief, C2C12 cells were grown and treated in 100 mm dishes (#83.3902 Sarstedt). The cells were collected and subcellular fractionation was performed to separate nuclear and cytosolic protein fractions. The protein concentration of each lysate was determined using Pierce BCA Protein Assay Kit (#23227 and #23215, Thermo Scientific). The samples were separated on polyacrylamide gels containing 0.5% trichloroethanol after separation gels were illuminated for 5 min on a UV table to allow for fluorescence-based quantification of total protein per lane in the gels and on the blotting membrane.⁷⁰ Signals from specific antibody staining were measured with an Intas ChemoStar (INTAS Science Imaging Instruments), densitometrically quantified, and normalized to total protein per lane. Quantification and analysis were performed with LabImage 1D software (Kapelan Bio-Imaging). Antibodies used: GAPDH (#PA5-85074, Invitrogen), Histone H3 (#ab1791, Abcam), phospho-SMAD1 (#13820 Cell Signaling), SMAD2 (#5339, Cell Signaling), β-catenin (#D10A8, Cell Signaling), Goat-anti-rabbit-horseradish peroxidase (HRP) (#ab97051, Abcam).

4.7. Kinase Inhibition

In vitro kinase inhibition assay against GSK3β was performed in duplicates at the indicated concentrations by Reaction Biology (Malvern, Pennsylvania). The ATP concentration was 10 μM, and a 10-point dose–response curve of staurosporin was used as an internal quality control.

4.8. Statistical Analysis

Statistical analysis between groups was performed via one-way analysis of variance (ANOVA) with a confidence level of 95%, followed by Dunnet’s multicomparison test. ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.001.

Glossary

Abbreviations

ACN

acetonitrile

ALP

alkaline phosphatase

BIO

6-bromo indirubin-3′-oxime

BMP

bone morphogenetic protein

BMPR-I/-II

BMP receptor type 1/type 2

cat.

catalytic

cDNA

complementary DNA

CHIR

CHIR-99021

CHK1

checkpoint kinase 1

CK1

casein kinase 1

cWnt

canonical Wnt

DCM

dichloromethane

DMEM

Dulbecco’s modified Eagle’s medium

DMF

dimethylformamide

DMSO

dimethyl sulfoxide

ERK

extracellular signal-regulated kinases

FBS

fetal bovine serum

FKBP12

FK506 binding protein 12

GSK3 (β)

glycogen synthase kinase 3 (β)

HRP

horseradish peroxidase

Id

inhibitor of DNA binding

MAPK

mitogen-activated protein kinase

MOM

methoxymethyl

mRNA

messenger ribonucleic acid

Ocn

osteocalcin

Osx

osterix

PBS

phosphate-buffered saline

PD

PD407824

PDD

phenotypic drug discovery

PI3K

phosphoino-sitol-3 kinase

TsOH

para-toluenesulfonic acid

rBMP-2

recombinant BMP-2

RNA

ribonucleic acid

RT-qPCR

reverse transcriptase-quantitative polymerase chain reaction

RunX2

runt-related transcription factor 2

SAR

structure–activity relationship

SD

standard deviation

TAK1

TGFβ-activated kinase 1

TGFβ

transforming growth factor β

THF

tetrahydrofuran

ZIP

zero interaction potency

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.3c00103.

• Spectroscopic data for all synthesized compounds, NMR spectra, and HPLC trace of final test compounds; supporting information on immune-blotting (Figure S1) and influence on SMAD2 protein levels (Figure S2); additional images supplementing the images displayed in Figure 4E (Figure S3); synergy scores calculated for the data displayed in Figure 5A (Figure S4); and primers used for RT-qPCR (Table S1) (PDF)

Author Contributions

D.S. conceived and designed the project. D.R. and T.F. performed the biological and biochemical experiments. S.H. performed the chemical syntheses. D.R., S.H., and D.S. analyzed the data, discussed the results, and prepared the manuscript.

The authors declare no competing financial interest.