Small-Molecule Probes as Pharmacological Tools for the Bone Morphogenetic Protein Signaling Pathway

Abstract

The bone morphogenetic protein (BMP) pathway is highly conserved and plays central roles in health and disease. The quality and quantity of its signaling outputs are regulated at multiple levels, offering pharmacological options for targeted modulation. Both target-centric and phenotypic drug discovery (PDD) approaches were applied to identify small-molecule BMP inhibitors and stimulators. In this Review, we accumulated and systematically classified the different reported chemotypes based on their targets as well as modes-of-action, and herein we illustrate the discovery history of selected candidates. A comprehensive summary of available biochemical, cellular, and in vivo activities is provided for the most relevant BMP modulators, along with recommendations on their preferred use as chemical probes to study BMP-related (patho)physiological processes. There are a number of high-quality probes used as BMP inhibitors that potently and selectively interrogate the kinase activities of distinct type I (16 chemotypes available) and type II receptors (3 chemotypes available). In contrast, only a few high-quality BMP stimulator modalities have been introduced to the field due to a lack of profound target knowledge. FK506-derived macrolides such as calcineurin-sparing FKBP12 inhibitors currently represent the best-characterized chemical tools for direct activation of BMP-SMAD signaling at the receptor level. However, several PDD campaigns succeeded in expanding the druggable space of BMP stimulators. Albeit the majority of them do not entirely fulfill the strict chemical probe criteria, many chemotypes exhibit unique and unrecognized mechanisms as pathway potentiators or synergizers, serving as valuable pharmacological tools for BMP perturbation.

Chemical probes are indispensable tools to interrogate the complex interactions in cellular biology. The function of an individual target within vast molecular networks can be investigated by selective manipulation. This enables the discovery of uncharted biochemical and systems biological processes, thereby sparking the development of innovative therapeutic approaches.^1,2 Particularly, in the context of morphogenic signaling pathways, where genomic approaches struggle with high lethality in knockout models, the conditional and reversible control offered by small molecules is invaluable.³

Prerequisite for high-quality chemical probes are meticulous design and optimization combined with extensive functional validation. A proper chemical probe binds to its target selectively and with high affinity and has a well-characterized mode of action in a complex system, and crucially, inactive structural analogues as essential negative controls are available.⁴⁻⁶ Notably, the attributes of a chemical probe are not equivalent to the requirements for a drug candidate. The clinical efficacy of approved drugs often relies on interactions with multiple targets, and ultimately, identifying the target and mechanism of action is not a necessity for approval by the authorities.^4,7

The development of chemical probes can focus on already known proteins of interest, building on previously identified small-molecule binding partners or based on available structural information (i.e., target-centric). However, this approach requires detailed knowledge of the underlying biology and is limited to already identified targets. To expand the druggable space, phenotypic screening has re-emerged as the most powerful tool in the discovery of previously unrecognized targets and mechanisms involved in the investigated biology (i.e., target-agnostic).⁸ A limiting factor to this approach is the tedious process of target deconvolution following hit generation.⁹ Identifying the relevant target(s) from a complex biological system is streamlined by recent advances in multi-omics-based target identification technologies.¹⁰⁻¹² However, facilitating this process requires highly selective chemical tools and, importantly, negative controls to restrict the otherwise long list of non-selective and low-affinity binders.¹³ In the past 20 years, both target-centric and phenotypic drug discovery approaches have been applied to identify small-molecule modulators of the bone morphogenetic protein (BMP) pathway.

The BMP pathway is a core regulator of cell fate throughout life and highly conserved among various species.^3,14 Dysregulation of the pathway causes severe and life-threatening diseases that currently lack causative therapy.^15,16 Attenuated BMP signaling is associated with acute kidney injury (AKI) and its progression to chronic kidney disease (CKD) as well as osteoporosis.¹⁷⁻¹⁹ In fact, BMP’s pronounced osteoinductive activity is therapeutically harnessed to address skeletal trauma, with recombinant human BMP-2 (Dibotermin-α) being approved in the European Union for the treatment of spinal fusions and tibia fractures.^20,21 Furthermore, loss-of-function mutations in the pathway were identified as the main cause of pulmonary arterial hypertension (PAH).^22,23 In contrast, hyperactive BMP signaling causes the heterotopic ossification observed in fibrodysplasia ossificans progressiva (FOP).²⁴ Moreover, divergent roles of BMP exist in a number of malignant disease states.²⁵⁻²⁷

BMP signal transduction, in brief, is initiated by extracellular ligand binding to transmembrane receptors, leading to heteromeric complex formation and phosphorylation of the type I receptor (BMPR-I) by the constitutively active type II receptor (BMPR-II) (Figure 1A).^28,29 The activated type I receptors phosphorylate SMAD1,5,9 proteins that form complexes with SMAD4 and translocate to the nucleus where they drive expression of various genes, including the inhibitor of DNA binding (ID) family and osteogenic regulators like Runt-related transcription factor 2 (RUNX2).³⁰ Additionally, BMP signals through non-canonical kinase cascades, like mitogen-activated protein kinases (MAPKs), that are interlinked with SMAD signaling and integrate signaling events of adjacent pathways.^31,32 The canonical cascade is regulated on multiple levels. For instance, FK506 binding protein 12 (FKBP12) binds to the GS domain of the type I receptors and inhibits phosphorylation by the type II receptor in the absence of ligand.²⁹ Further downstream, the SMAD ubiquitination regulatory factor 1 (SMURF1) ubiquitinates phosphorylated SMADs (pSMAD), labeling them for proteasomal degradation to terminate their activity.³³ SMURF1 binding to SMADs is regulated through phosphorylation events in the SMAD linker region by MAPKs and glycogen synthase kinase 3 (GSK3).^34,35

Figure 1.

Overview of the BMP signaling pathway and BMP modulators. (A) BMP inhibitors target receptor type I and type II kinase activities. BMP stimulators have divergent targets and mechanisms, largely centered around canonical (BMP-SMAD) signaling transduction. (B) Literature analysis (SciFinderⁿ and SciteAI): Left panel depicts the cumulated number of journal publications of small-molecule BMP stimulators (blue) and inhibitors (orange); right panel illustrates the chemical diversity of BMP modulators based on the number of reported chemotypes. Hatched area demonstrates chemotypes present in both categories. MoA, mode of action. See Methods for SciFinderⁿ and SciteAI search parameters and Figures S1–S3 for an overview of all selected chemotypes. Created with biorender.com.

There are currently no small-molecule BMP modulators approved for clinical use despite ever-increasing efforts in their development (Figure 1B). In the following sections we assemble and classify BMP inhibitors and stimulators based on their target and mechanism. A comprehensive summary of available biochemical, cellular, and in vivo activities of the most relevant BMP modulators is provided, along with recommendations on their preferred use as chemical probes and pharmacological tools to study BMP-related (patho)physiological processes. For detailed discussions on the clinical relevance and pharmacology of BMP therapeutics, the reader is referred to recent literature.^16,21,36,37

1. BMP Inhibitors

The development of small-molecule BMP inhibitors has mainly focused on BMP type I receptors (Figure 1B). Despite the large overall number of published inhibitors, most share the same core 6-phenylpyrazolo[1,5-a]pyrimidine scaffold, resulting in a total chemical diversity of 16 distinct BMPR-I inhibitor chemotypes (Figure 2, Table 1). Only recently in 2023 have BMPR-II targeting inhibitors been introduced to the field, critically expanding the available set of chemical tools to study the BMP pathway in health and disease (Figure 3, Table 2).

Figure 2.

Chemical structures of BMPR-I inhibitors and usage frequency of selected compounds. (A) Most inhibitors feature a 6-phenylpyrazolo[1,5-a]pyrimidine scaffold or present closely related derivatives such as the macrocyclic OD36 or 7-phenylimidazo[1,2-a]pyridine LJ000328. The 3,5-diphenylpyridine chemotype has been introduced with derivatives exhibiting prime potencies and ALK selectivities along with the availability of structurally closely related inactive control compounds (i.e., M4K2234). (B) SciFinderⁿ reference analysis revealed that 6-phenylpyrazolo[1,5-a]pyrimidines are by far the most frequently cited and utilized BMPR-I-inhibitor chemotypes, followed by the 3,5-diphenylpyridines. For compounds cited <50 times, publications have further been categorized for compound usage in biological assays vs their total citations. Not shown: Inhibitors with only one publication.

Table 1. Bioactivity Profiles of BMP Type I Receptor Kinase Inhibitors^a.

compounds	main targets	biochemical IC50	cell-based IC50	kinome profile	references
6-Phenylpyrazolo[1,5-a]pyrimidines: Dorsomorphin Derivatives
Dorsomorphin	ALK1	kinase assay	BRE-luc assayb	70 kinase panel	2008(42)
(Compound C,	ALK2	22 nM VEGFR2	(C2C12)	(1 μM)	2013(50)
BML-275)	ALK3	68 nM ALK2	110 nM ALK2	MELK	2013(47)
	ALK6	74 nM BMPR2	172 nM ALK3	PHK	2007(64)
	AMPK	95 nM ALK3	309 nM ALK1	DYRK1A
	BMPR2	103 nM TGFβR2	CAGA-luc assayb	DYRK3
	TGFβR2	106 nM ALK1	(HEK293T)	ERK8
	VEGFR2	235 nM AMPK	980 nM ALK5	MNK1
		235 nM ALK6	2412 nM ALK4	Lck
		17090 nM ALK5
		25740 nM ALK4
LDN-193189	ALK1	kinase assay	BRE-luc assayb	198 kinase panel	2008(45)
(DM-3189,	ALK2	<5 nM ALK3	(C2C12)	(100 nM, 1 μM)	2013(50)
Compound 13)	ALK3	13 nM ALK1	11 nM ALK2	SIK2	2013(47)
	ALK6	41 nM ALK2	11 nM ALK3	ABL1
	TGFβR2	60 nM ALK6	23 nM ALK1	MAP4K4
	VEGFR2	140 nM TGFβR2	CAGA-luc assayb	451 kinase panel
		215 nM VEGFR2	(HEK293T)	(1 μM)
		565 nM ALK5	213 nM ALK5	RIPK2
		1122 nM AMPK	238 nM ALK4	ZAK
		1825 nM ALK4		KIT
		3845 nM BMPR2		ACVR2B
				MEK5
				SIK2
DMH-1	ALK1	kinase assay	BRE-luc assay	/	2010(46)
(VU0364852)	ALK2	<5 nM ALK3	(C2C12)		2013(50)
	ALK3	27 nM ALK1	100 nM		2013(47)
	ALK6	48 nM ALK6	BRE-luc assayb
		108 nM ALK2	(C2C12)
		9622 nM ALK4	230 nM ALK2
			317 nM ALK3
			378 nM ALK1
			CAGA-luc assay
			(HEK293T)
			503 nM ALK4
			658 nM ALK5
LDN-212854	ALK1	kinase assay	BRE-luc assayb	198 kinase panel	2013(47)
	ALK2	1 nM ALK2	(C2C12)	(100 nM, 1 μM)
	ALK3	2 nM ALK1	16 nM ALK2	ABL1
		86 nM ALK3	100 nM ALK1	PDGFRß
		2113 nM ALK4	166 nM ALK3	ABL2
		9276 nM ALK5	CAGA-luc assayb	SIK2
			(HEK293T)	MAP4K4
			1684 nM ALK4	451 kinase panel
			2103 nM ALK5	(1 μM)
				ABL1
				PDGFRβ
				KIT
				ZAK
				MEK5
				RIPK2
				TGFβR2
ML347	ALK1	kinase assay	BRE-luc assay	(10 kinase panel:	2013(50)
(1LWY,	ALK2	32 nM ALK2	(C2C12)	kinase assay data)
VU0469381,		46 nM ALK1	152 nM	(0–100 μM)
LDN-193719)		9830 nM ALK6
		10800 nM ALK3
		19700 nM VEGFR2
		inactive:
		ALK3
		ALK5
		AMPK
		BMPR2
		TGFβR2
Compound 23	ALK1	kinase assay	BRE-luc assay	252 kinase panel	2018(51)
(ALK2-IN-2)	ALK2	9 nM ALK2	(C2C12)	(1 μM)
		51 nM ALK1	8 nM	ARK5
		698 nM TGFβR2		DDR1
		2370 nM ACVR2A		MAP4K5
		3130 nM ACVR2B		RIPK2
		6439 nM ALK3		TNIK
		13140 nM ALK6
		18330 nM ALK4
		18650 nM ALK5
		>10000 nM BMPR2
OD36	ALK1	kinase assay	BRE-luc assayb	366 kinase panel	2019(52)
	ALK2	22 nM ALK2 R206H	(C2C12)	(100 nM)
	ALK2	47 nM ALK2	308 nM ALK2 wt	RIPK2
	R206H	Kd:	3958 nM ALK2 R206H	SIK2
	ACVR2B	37 nM ALK2
		90 nM ALK1
		90 nM ACVR2B
		680 nM TGFβR2
		3000 nM ALK3
		3000 nM ALK4
		3000 nM ALK5
		3000 nM ACVR2A
3,5-Diphenylpyridines: K02288 Derivatives
K02288	ALK1	kinase assay	BRE-luc assayb	200 kinase panel	2013(53)
(K02288a)	ALK2	1 nM ALK2	(C2C12)	(100 nM, 1 μM)	2013(47)
	ALK3	2 nM ALK1	100 nM	ABL1
	ALK6	6 nM ALK6	BRE-luc assayb	ARG
		34 nM ALK	(C2C12)	MAP4K4
		220 nM ActRIIA	225 nM ALK2	MINK
		302 nM ALK4	237 nM ALK3
		321 nM ALK5	440 nM ALK1
			CAGA-luc assayb
			(HEK293T)
			693 nM ALK5
			812 nM ALK4
LDN-214117	ALK2	kinase assay	BRE-luc assay	200 kinase panel	2014(54)
(Compound 10)		24 nM ALK2	(C2C12)	(100 nM, 1 μM)
		3000 nM ALK5	100 nM (BMP6)	TNIK
			960 nM (BMP2)	RIPK2
			1022 nM (BMP4)	ABL1
			CAGA-luc assay	MAP4K4
			(HEK293T)	MAP4K5
			16000 nM (TGFβ1)
M4K2117	ALK1	kinase assay	NanoBRET assay	374 kinase panel	2020(55)
(Compound 14k)	ALK2	7 nM ALK2	(HEK293)	(1 μM)	2021(56)
	ALK3	>5000 nM ALK5	29 nM ALK2	DDR1
	ALK6		CAGA-luc assay	TNIK
			(HEK293)
			3326 nM ALK5
M4K2234	ALK1	kinase assay	NanoBRET assay	375 kinase panel	2020(57)
(Compound 26b)	ALK2	7 nM ALK1	(HEK293)	(1 μM)	Structural
	ALK6	14 nM ALK2	13 nM ALK2	TNIK	Genomics
		41 nM TNIK	83 nM ALK1		Consortium
		88 nM ALK6	536 nM ALK3		(theSGC.org)
		168 nM ALK3	1628 nM ALK6
		1660 nM ALK4	7932 nM ALK5
		1950 nM ALK5	8424 nM ALK4
Diverse Scaffolds
LJ000328	ALK1	kinase assay	/	369 kinase panelc	2020(60)
(U000328,	ALK2	5 nM ALK3		ABL2	2020(59)
Compound 19b)	ALK2	6 nM ALK2 R206H		TNIK
	R206H	11 nM ALK2 wt		DDR1
	ALK3	29 nM ALK6		SIK1
	ALK6	166 nM ALK1		SIK2
		569 nM ALK4		LIMK1
		587 nM TGFβR2
		1267 nM ALK5
		7450 nM PDGFRβ
		13410 nM KDR
		13590 nM AMPK
		>50000 nM BMPR2
OD16	ALK1	Kd:	BMP9-induced	96 kinase panel	2021(61)
	ALK2	17 nM ALK2	pSMAD1,5 response	(100 nM)
	ALK3	20 nM ALK1	(EA.hy926)
	ACVR2A	55 nM ALK3	80 nM
		110 nM ACVR2A
		220 nM ACVR2B
		270 nM ALK4
		360 nM ALK5
		380 nM TGFβR2
Ullrich 23	ALK1	mobility shift assay	BRE-luc assay	468 kinase panel	2022(62)
	ALK2	8 nM ALK2 R206H	(HEK293)	(1 μM)
	ALK2	12 nM ALK2 wt	44 nM (BMP6)	NDR1
	R206H	120 nM ALK1		PRKR
		280 nM ALK3
		2200 nM ALK5
		2100 nM ALK6
Compound 16	ALK1	kinase assay	/	(56 kinase panel:	2022(63)
	ALK2	1 nM ALK2		kinase assay data)
		19 nM ALK1
		100 nM ALK5
		176 nM FGFR3
		901 nM ALK3
		1068 nM KDR

^aSmall-molecule BMPR-I inhibitors are sorted according to their chemical scaffold. If applicable, compound synonyms are provided in parentheses. The main targets are listed alphabetically. IC₅₀ values from biochemical and cell-based assays were rounded to nanomolar digits and listed in ascending order. The kinome profile data includes kinases with inhibition values >80% and excludes BMP receptors, with the concentration used given in parentheses.

^bReporter gene assay with overexpression of constitutively active receptors.

^c% inhibition value and tested compound concentrations unknown.

Figure 3.

Chemical structures of BMPR-II inhibitors. The recently disclosed first-in-class BMPR2 (i.e., one of the three type II receptors) inhibitors CDD-1653 and Compound 8a feature a 2,4-diaminopyrimidine core scaffold but are classified as distinct chemotypes based on their putatively different hinge-binding motifs.^65,66

Table 2. Bioactivity Profiles of BMP Type II Receptor Kinase Inhibitors^a.

compounds	main targets	biochemical IC50	cell-based IC50	kinome profile	references
CDD-1653	BMPR2	kinase assay	BRE-luc assay	403 kinase panel	2023(65)
(Compound 7n)		3 nM BMPR2	(HEK293T)	(1 μM)
		>1000 nM ALK1	6920 nMb	JAK2
		inactive:		JAK3
		ALK2		PIK4CB
		ALK3		TYK2
		ALK4
		ALK5
		ALK6
		TGFβR2
		ActRIIA
		ActRIIB
Compound 8a	BMPR2	ADP Glo assay	/	468 kinase panel	2023(66)
		506 nM BMPR2		(1 μM)
				FLT3 (D835 V)
				GSK3A

^aIf applicable, compound synonyms are provided in parentheses. IC₅₀ values from biochemical and cell-based assays were rounded to nanomolar and are listed in ascending order. The kinome profile data includes kinases with inhibition >80% and excludes BMP receptors, with the concentration used in brackets.

^bThe authors explain the high IC₅₀ value with pronounced selectivity for BMPR2 over the other type II receptors (i.e., ActRIIA and ActRIIB) but not with low overall cellular activity. Cells might compensate the inhibition of one type II receptor via enhanced signaling through the other two. It is difficult to draw conclusions about cellular engagement of BMPR2.

Here, we summarize the inhibitors’ profiles in terms of their biochemical and cellular BMP activities, with a particular emphasis on the in vitro target selectivity. Although several elegant in cellulo assay setups have been introduced in the past years, in our view, they are associated with shortcomings that limit clear-cut statements on target selectivity profiles: 1) Assays employing the overexpression of constitutively active receptors generally tend to show reduced selectivity compared to in vitro biochemical assays. A bias attributed to the expression constructs or the overall assay system cannot be excluded. 2) Nano-BRET assays nicely assess and validate in cellulo target engagement. However, the investigated inhibitor competes against another small molecule (i.e., a “tracer” with unknown kinase selectivity profile), making it difficult to compare numerical IC₅₀ values quantitatively. We included cellular BMP inhibition data from plain BRE reporter assays and overexpression setups but assessed selectivity primarily based on in vitro data. Comparing cellular data with biochemical potencies serves to provide a more overall view on cellular activity, highlighting physicochemical properties such as permeability.

1.1. Type I Receptor Kinase Inhibitors

In recent years, the investigation of new small-molecule BMP inhibitors has mainly centered on BMP type I receptors, with a particular focus on compounds with improved ALK2 selectivity. Activating mutations in ALK2 (e.g., R206H) are crucial in the formation and progression of some rare diseases, notably FOP or diffuse intrinsic pontine glioma (DIPG).^38,39 This extensive pathway activation results in the disruption of inhibitory FKBP12 binding and enhanced SMAD1,5,9 phosphorylation, in the absence of activating ligands.⁴⁰ To further improve our fundamental understanding of these diseases, the availability of high-quality chemical tools as reference compounds is critical. These qualities include not only their inhibitory potencies but also ALK and overall kinome selectivity as well as physicochemical and pharmacokinetic properties. Other ALK receptors, apart from ALK2, also seem to play roles in distinct disease pathologies.¹⁶ For instance, it is established that BMP signaling via ALK3 is involved in processes related to muscle growth and atrophy.⁴¹ To this end, no dedicated efforts were made to develop suitable ALK3 inhibitors to validate the significance of ALK3 in these disease contexts and its potential as a therapeutic target.

1.1.1. 6-Phenylpyrazolo[1,5-a]pyrimidines: Dorsomorphin Derivatives

The development of BMP inhibitors began in 2008 with the 6-phenylpyrazolo[1,5-a]pyrimidine Dorsomorphin (Figure 2A, Table 1), which was discovered in a pioneering work from high-content imaging screening of 7500 small molecules in zebrafish that monitored the development of the embryonic dorsoventral axis.⁴² The compound was originally described as an AMPK inhibitor under the name Compound C and is currently among the most frequently used reference compounds (Figure 2B).⁴³ Notably, due to the lack of ALK and overall kinase selectivity, Dorsomorphin should not be used as a chemical probe to study BMP biology. However, medicinal chemistry efforts produced a series of optimized derivatives (Figure 2A). Replacement of Dorsomorphin’s piperidinylethoxy substituent by piperazine and the 4-pyridine by 4-quinoline led to LDN-193189 with improved cellular potency (10–100-fold) and pharmacokinetic features (e.g., plasma t_1/2 = 1.6 h).^44,45 In order to further reduce off-target effects, structure–activity relationship (SAR) studies furnished DMH-1, which achieved high selectivity for ALK1/2 over ALK3, ALK4/5 (TGFβ, transforming growth factor β), AMPK, KDR/VEGF, and PDGFRβ via installation of an isopropoxy substituent.⁴⁶ It should be noted that DMH-1 exhibited profound ALK selectivity in biochemical kinase assays, while cell-based data for BMP versus TGFβ activities suggested only moderate selectivity (Table 1).⁴⁷ Considering these observations, biochemical selectivity data should be interpreted with caution.⁴⁸ However, a few more interesting Dorsomorphin-derived compounds have been developed. The improved compound LDN-212854 displayed enhanced affinity for ALK2 (IC₅₀ = 1.3 nM) and ALK1 (IC₅₀ = 2.4 nM), with submicromolar cellular potency and selectivity over ALK4/5.⁴⁷ In addition, a full kinome profile is available, an essential chemical probe criterion for newly introduced kinase inhibitors.⁴⁹ Such data are missing for ML347, which also showed potent ALK2 (IC₅₀ = 32 nM) and ALK1 (IC₅₀ = 46 nM) inhibition in a biochemical assay.⁵⁰ The change from quinoline to naphthalene in the 3-position and the insertion of a sulfonamide substituent led to the highly ALK2-selective Compound 23 (IC₅₀ = 9 nM) with only low inhibition of ALK1.⁵¹ Another promising development is the macrocyclic OD36, a bridged pyrazolo[1,5-a]pyrimidine scaffold that potently inhibits the FOP mutant ALK2 R206H (IC₅₀ = 22 nM), wild-type ALK2 receptor (IC₅₀ = 47 nM), and ALK1 while only modestly affecting two further kinases (kinome profile at 100 nM).⁵²

Based on the herein reviewed compound profiles, we recommend the Dorsomorphin derivatives LDN-212854 and OD36 as suitable pharmacological tools. They appear to exhibit the most favorable chemical probe characteristics among the pyrazolo[1,5-a]pyrimidine chemotype of BMP inhibitors, particularly with regard to their ALK1/2 selectivity.

1.1.2. 3,5-Diphenylpyridines: K02288 Derivatives

In addition to the Dorsomorphin derivatives, a distinct prototype approach emerged based on the 3,5-diphenylpyridine scaffold (Figure 2A). K02288 was originally identified from a 2000 compound kinase-directed library screen against a purified human kinase panel.⁵³ It inhibits the BMP type I receptors in a low nanomolar range (ALK2 IC₅₀ = 1.1 nM) but lacks selectivity for BMP versus TGFβ in a cell-based assay (Table 1).^47,53 Its poor aqueous solubility⁵⁴ is a further disadvantage for broad use of K02288 as a pharmacological tool. Efforts to overcome these shortcomings yielded LDN-214117 as a potent and ALK2-selective K02288 derivative (IC₅₀ = 24 nM).⁵⁴ In this study, selectivity and potency against TGFβ receptors were significantly improved by replacing the 2-amino by a methyl group of the hinge-binding motif as well as the 3-phenol group by 4-piperazine. Kinome profiling identified TNIK, RIPK2, ABL1, MAP4K4, and MAP4K5 as potential off-targets.⁵⁴ Several years later, M4K2117 was devised by replacing the piperazine with N-methylpiperidine and shifting the methyl group from the 2- to the 4-position in the pyridine.⁵⁵ This strategy endowed M4K2117 with prime potency and ALK2 selectivity in vitro (IC₅₀ = 7 nM) with only a few off-targets, like DDR1 and TNIK in its kinome array. Moreover, this ALK2 inhibitor was designed to achieve a superior bioavailability and brain-penetration profile for the prospective treatment of DIPG.⁵⁵M4K2117 was shown to cross the blood–brain barrier in rats, demonstrated by positron emission tomography (PET) neuroimaging.⁵⁶ However, the most advanced 3,4-diphenylpyridine is M4K2234, with key modifications of the previously unaltered trimethoxyphenyl substituent, which exhibits comparable inhibitory potency for ALK1 (IC₅₀ = 7 nM) and ALK2 (IC₅₀ = 14 nM).⁵⁷ Notable assets are the very clean kinome activity profile (e.g., only off-target is TNIK at 1 μM), effective in cellulo target engagement (NanoBRET IC₅₀ = 13 nM, 83 nM against ALK2 and ALK1, respectively), and excellent pharmacokinetic performance in vivo.⁵⁷ The Structural Genomics Consortium (SGC) listed M4K2234 together with its full kinome profile and a structurally closely related negative control compound M4K2234NC.⁵⁸

Together, the M4K series of compounds combine excellent chemical probe qualities among the 3,4-diphenylpyridine chemotype of BMP inhibitors, with the set of M4K2234/NC as prime pharmacological tool compounds.

1.1.3. Miscellaneous Heterocyclic Scaffolds

Beyond the Dorsomorphin and K02288 chemotypes, a collection of compounds from diverse scaffolds has been added to the class of BMPR-I inhibitors in recent years. The 7-phenylimidazo[1,2-a]pyridine LJ000328 (Figure 2A) shows a recognizable potency for ALK3 (IC₅₀ = 5 nM) and the FOP mutant ALK2 R206H receptor (IC₅₀ = 6 nM) as well as adequate selectivity against TGFβ receptors (Table 1).^59,60 However, it also interacts with the kinases ABL2, TNIK, DDR1, SIK1, SIK2, and LIMK1. Moreover, cellular data are currently missing to conclusively evaluate its suitability as a tool compound.

Another macrocyclic compound, alongside the previously mentioned pyrazolo[1,5-a]pyrimidine OD36, is the ALK1/2-selective OD16. It was identified by a chemocentric approach based on the Nanocyclix technology. OD16 presents commendable potency and selectivity for ALK2 (K_d = 17 nM) and ALK1 (K_d = 20 nM) over TGFβ receptors in vitro.⁶¹ Advantageously, no significant off-target inhibition was observed in a kinase profiling assay. OD16 was also extensively characterized in various zebrafish assays, demonstrating anti-tumor angiogenesis activity.⁶¹

2-Aminopyrazine-3-carboxamide Ullrich 23 (Figure 2A) presents an SAR-optimized hit from a kinase inhibitor library screening. It demonstrates high affinity for the FOP mutant ALK2 R206H (IC₅₀ = 8 nM) as well as the ALK2 wild-type receptor (IC₅₀ = 12 nM) and is even selective over several other BMP type I receptors. The kinome data indicates high overall target selectivity since only two additional kinases were inhibited. Ullrich 23 was validated to inhibit BMP signaling in the submicromolar range in cellulo (IC₅₀ = 44 nM) and exhibits sufficient aqueous solubility (420 μM at pH 6.8), membrane permeability, and microsomal stability for promising use in FOP models in vivo.⁶²

Compound 16 (Figure 2A) was developed from a SAR series of novel pyrazolo[4,3-d]pyrimidines as a remarkably potent, selective, and orally bioavailable inhibitor of ALK2 in rats.⁶³ In fact, its pyrazolopyrimidine scaffold is reminiscent of the first-in-class BMP inhibitor Dorsomorphin. Nevertheless, a detailed kinome profile for extended characterization is still lacking for final evaluation of chemical probe qualities compared to those of other available BMPR-I inhibitors.

Interestingly, in a collaborative effort with academia and pharmaceutical companies, the SGC has developed another BMPR-I inhibitor: MU1700 is claimed to not inhibit any off-targets while exhibiting high potency against ALK2 (IC₅₀ = 6 nM) and ALK1 (IC₅₀ = 13 nM) in biochemical kinase and cellular NanoBRET assays. MU1700 is reported to enter the central nervous system in an in vivo model. Moreover, a structurally related negative control is available (i.e., MU1700NC). Given that these data did not yet undergo a peer-review process, we have not included it in the current Review.

1.2. Type II Receptor Kinase Inhibitors

Although an extensive collection of inhibitors targeting BMP type I receptors has been built up in the past two decades, selectively targeting BMPR-II remained challenging. The very first selective BMP type II receptor inhibitors were not disclosed until very recently in 2023 (Figure 3, Table 2). The pharmacological utility of BMP type II receptor-specific interrogation might not be apparent yet, but the availability of selective chemical tools is certainly valuable to perform fundamental mechanistic studies. Moreover, BMP signaling via the type I (ALK) receptors requires BMP type II receptors, as described above, and could offer an alternative strategy for therapeutic intervention of ALK-related diseases.

2,4-Diaminopyrimidine CDD-1653 was discovered from a DNA-encoded chemical library screen and shows excellent potency for BMPR2 (IC₅₀ = 3 nM), does not affect the type I receptors to a relevant extent, and exhibits a very good overall kinome selectivity profile.⁶⁵ Additionally, two inactive 2,4-diaminopyrimidine analogues (Compounds 7m/p) were identified during SAR studies that could potentially serve as negative controls. The cellular activity of CDD-1653 currently cannot be estimated, as selective inhibition of one type II receptor seems to be compensated by the other type II receptors.⁶⁵

The latest BMPR2 inhibitor is the macrocyclic pyrazole Compound 8a,⁶⁶ which was designed by macrocyclization of a promiscuous, aminopyrazole-based kinase inhibitor.⁶⁷ In addition to investigating the BMPR2 affinity (IC₅₀ = 506 nM), the overall kinome selectivity was assessed and suggested only FLT3 and GSK as relevant off-targets. However, cellular GSK isoform inhibition (NanoBRET assay) revealed that these kinases are only affected at doses beyond the biochemical on-target IC₅₀.⁶⁶ Additional data from cell-based assays would be desirable in order to assess the suitability of Compound 8a as a chemical probe.

Together, CDD-1653 and Compound 8a both qualify as early probes to interrogate BMPR2 function in BMP biology as one of the three BMP type II receptors. They provide promising starting points for the design and development of BMPR-II inhibitor tool compounds in the future.

1.3. Clinical Candidates

Several BMP inhibitors have been translated into clinical trials, such as for the treatment of heterotopic ossification (HO) relating to FOP or various forms of anemia, while most of them were not originally developed in a BMP-specific context.¹⁶ Consequently, these compounds also hit other targets. Current clinical candidates include KER-047 (Keros Therapeutics),⁶⁸INCB00928 (Incyte),⁶⁹BLU-782 (Blueprint Medicines), BCX9250 and BCX9499 (BioCryst Pharmaceuticals), Saracatinib (AZD0530; Astra Zeneca),⁷⁰ a RARγ agonist Palovarotene,⁷¹E6201,⁷²Momelotinib,⁷³Itacnosertib,⁷⁴ and Mubritinib (TAK-165) to name just a few.

2. BMP Stimulators

BMPs are the most effective osteoinductive cues. Hence, they are highly attractive therapeutic modalities for bone regeneration from diverse etiologies.^20,21,75 Recombinant human BMP-2 (Dibotermin-α, InductOs) is in clinical use to support treatment of difficult fractures but requires administration at doses 10,000-fold above physiological concentrations, which is expensive, inefficient, and ultimately a safety concern (e.g., heterotopic ossification).^21,76 Beyond bone health, there are additional medical indications where BMPs or BMP pathway stimulation is therapeutically explored, including PAH, kidney diseases, and specific malignancies.^23,27,37,77 In the latter context, for instance, a BMP-based treatment of tumor-associated cachexia emerged as a promising therapeutic approach for cancer patients (>50%) suffering from this muscle atrophy during disease progression.⁷⁸ Again, the benefit-to-risk ratio needs to be considered quite carefully.

Notably, BMPs also found widespread entry into stem-cell-based biotechnology processes, mostly for directed differentiation of human induced pluripotent stem cells (hiPSCs) toward desired progenies, e.g., mesoderm-derived cell types.^79,80 Such hiPSC-derived cells were found to be integral for the generation of safety pharmacology assay platforms in preclinical drug development. Similarly, BMPs have supported the construction of 3D humanized bone models from primary cells.^81,82 Thus, BMP is a valuable factor for xeno-free, chemically defined, and GMP-compliant cell and tissue production.

In light of the described limitations and concerns of using BMPs, small-molecule modalities as BMP stimulators present highly attractive alternatives to BMP-based biologicals, as they potentially exhibit superior immunological, physicochemical, and pharmacokinetic profiles.

2.1. General Considerations for Pharmacological Stimulation of the BMP Pathway

Genuine activation of growth factor signaling pathways with small molecules presents an intrinsically challenging endeavor. Most signaling pathways require the formation of multimeric receptor complexes that is usually mediated by ligand manifolds larger than drug-like small molecules. The activation of granulocyte-colony-stimulating factor signaling by SB247464 is one of the few examples where a small molecule facilitates formation of a dimeric receptor complex and downstream signaling.^83,84 A similar approach proved successful for the development of thrombopoietin signaling activation, with the marketed Eltrombopag and Lusutrombopag.^85,86 Heterotetrameric BMP receptors are even more complex, and it is possible that effective BMP signaling activation on this level might be reserved for peptide modalities. To this end, the augmentation of growth factor pathways via small-molecules largely takes place further downstream and is, in most cases, typically based on in-depth knowledge of underlying pathway biology. For example, activating Wnt/β-catenin signaling is usually accomplished by inhibition of GSK3β and, thus, reduced β-catenin degradation.^87,88 A more recent discovery introduced Molidustat, which increases erythropoietin expression through inhibition of hypoxia-inducible factor prolyl hydroxylase in patients with anemia.⁸⁹

BMP stimulator modalities show fundamental differences in their individual modes of action. In an attempt to standardize their nomenclature, we have distinguished them in different categories primarily based on their mechanisms (Figure 1A). However, it should be noted that, for a number of the small molecules, neither target nor mechanism is known, hampering a final classification. “BMP mimetics” act on the extracellular level and mimic the action of ligands on their receptors. These are exclusively represented by protein or peptide modalities, which are beyond the scope of this Review. To provide a brief overview, a variety of BMP mimetic peptides were designed based on crystal structures of BMPs or BMP receptors.⁹⁰⁻⁹³ Notably, THR-184 was tested in phase 1 and phase 2 clinical trials for the prevention of cardiovascular-associated acute kidney injury, where it demonstrated safety but no significant efficacy.^90,94 “BMP activators” also act on the receptor level; they do not mimic ligands but rather enhance the activation of downstream signaling effectors. As the largest and most diverse set, we categorized “BMP potentiators” that increase cellular responsiveness to BMP, typically dependent on at least low-level basal pathway activity. These can be understood to increase the tension of the inactive pathway, resulting in enhanced or prolonged activation upon ligand binding. Lastly, we defined those small molecules that do not necessarily modulate the BMP pathway directly but amplify cell fate in a BMP-dependent direction as “BMP synergizers”.

Historically, the identification and characterization of novel BMP stimulators have focused on the canonical BMP/SMAD pathway. This might be attributed to the convoluted nature of the non-canonical signaling cascade and the fact that non-canonical effectors like MAPKs are by no means selectively involved in BMP signaling. While SMADs are well established as the primary intracellular effector system of BMPs, it should be noted that the SMAD cascade does not fully reflect the diversity of BMP ligands and that the diverging effects individual ligands can have in different models are not always completely understood.^22,95,96 Moreover, depending on the cellular context of the observed assay system, modulation of the same target or off-target or crosstalk with adjacent pathways can exert disparate effects. In osteogenic assays, the cooperation between Wnt and BMP is well documented, whereas in tissue patterning and organ formation, the relationship is more complex and dependent on the probed time frame.⁹⁷⁻⁹⁹ Thus, efficacy of BMP stimulators can vary greatly between biological systems, and mechanistic deconvolution should be prioritized early in small-molecule development campaigns to circumvent apparently contradicting results.

Compared to BMP inhibitors, the overall data quantity (and oftentimes quality) for BMP stimulators is quite limited. Systematic comparative studies permitting a more direct comparison between distinct compounds are missing. In fact, many BMP stimulators have hardly been utilized after their initial identification. Herein, we aim to give a comprehensive overview of current BMP stimulators and recommend suitable pharmacology tools based on their qualities as chemical probes. The latter includes an extensive summary of their available biological data (Table 3 and Table 4).

Table 3. Bioactivity Profiles of BMP Stimulators Targeting SMURF1 and FKBP12^a.

SVAK-12
discovery: target-centric in silico screen				target: SMURF1
MoA: attenuated pSMAD1,5,9 degradation
system			readout	application, concentration	additional information	references
in cellulo	C2C12 (mouse)	pathway specific	luciferase reporter	EC50 2.6 μM; highest activity at 7.3 μM	cytotoxic above 7.3 μM	2011(100)
		functional	Alp, Ocn (mRNA); ALP (activity)	EC50 1.5–5.3 μM; highest activity at 7.3 μM		2013(121)
in vivo	rat	functional	ectopic bone formation, fracture healing	local application, 25–1000 μg per disc

A01*
discovery: target-centric in silico screen				target: SMURF1
MoA: attenuated pSMAD1,5,9 degradation
system			readout	application, concentration	additional information	references
in vitro	SMURF1-SMAD1 binding assay	pathway specific	reduced SMURF1-SMAD1 binding	1 μM	interaction of SMURF2 with SMAD2,3 not disrupted	2014(101)
in cellulo	C33A-2D2 (human)	pathway specific	stabilization of SMAD1,5 (protein)	2–25 μM	control experiments with CHX, MG132, siRNA-KD
	C2C12 (mouse)	pathway specific	S206-pSMAD1, reduced SMAD1,5 ubiquitination (protein)	2–10 μM	SMAD2,3, SMAD4, ING2, RhoA levels not reduced (protein); RUNX2, MEKK2 stabilized (protein)
		functional	Alp, Ocn, Col1a1, (mRNA); ALP (activity)	2 μM	inactive without BMP addition
	HEK293 (human)	pathway specific	reduced SMURF1-SMAD1 binding	2 μM
		pathway specific	luciferase reporter
	MC3T3-E1 (mouse)	functional	Alp, Ocn, Col1a1 (mRNA); ALP (activity)	2 μM	inactive without BMP addition
	ROS17/2.8 (rat)	functional	ALP (activity)
in cellulo	PASMC (primary, human)	pathway specific	pSMAD1,5,9 (protein)	10–50 μM	increased pSMAD1,5,9 signaling attenuated migration of BMP signaling deficient cell population	2016(122)
		functional	cellular migration rates	50 μM
in cellulo	MCF7, T47D (human)	–	attenuated ERα signaling and ERα protein levels	10 μM	no BMP context	2018(109)
in cellulo	N1E-115 (mouse)	–	inhibited SMURF1 dependent downregulation of RhoA	0.01 μM	no BMP context	2020(110)
in vivo	mouse	functional	enhanced BMP effect on oligodendrocyte differentiation	0.1 μM		2021(123)
in cellulo	T24 (human)	functional	reduced wound healing/migration	5–10 μM		2022(124)

B06*
discovery: target-centric in silico screen				target: SMURF1
MoA: attenuated pSMAD1,5,9 degradation
system			readout	application, concentration	additional information	references
in vitro	SMURF1 ubiquitination assay	pathway specific	attenuated SMURF1 ubiquitination			2017(102)
in cellulo	C2C12 (mouse)	pathway specific	stabilized S206-pSMAD1, reduced SMAD1,5 ubiquitination (protein)	1–10 μM	controlled with SMURF1 KD
		functional	ALP (activity)	2 μM
	HEK293 (human)	mechanistic	interaction between SMURF1 and SMAD1 not interrupted; no influence on SMURF2 or SMURF2-SMAD2,3 interaction	2 μM
			increased RUNX2, SMAD2,3, SMAD4, ING2 (protein)		regulation of various SMURF1 substrates

CPAA
discovery: target-centric in silico screen				target: SMURF1
MoA: attenuated pSMAD1,5,9 degradation
system			readout	application, concentration	additional information	references
in cellulo	primary osteoblasts (mouse)	mechanistic	mutational binding mode studies; DARTS	5–10 μM	cells taken from population with enhanced SMURF1 activity; no effect on SMURF2-SMAD2 binding, SMAD2 ubiquitination, pSMAD2,3, pERK or p-p38	2018(103)
		pathway specific	reduced binding of SMAD1 to SMURF1, reduced ubiquitination of SMAD1 and RUNX2	2.5–10 μM
		functional	Ocn, Runx2 (mRNA); pSMAD (protein); ALP (activity); Alizarin staining	2.5–10 μM
in vivo	mouse (or primary osteoblasts isolated after treatment)	pathway specific	reduced binding of SMAD1 to SMURF1, reduced ubiquitination of SMAD1 and RUNX2	local application of 2–4 mg per sponge	population with enhanced SMURF1 activity
		functional	pSMAD1 (protein); Ocn (mRNA); RUNX2 activation (Assay-kit)	local application of 2–4 mg per sponge
		functional	various osteogenic markers and enhanced spinal fusion	local application of 2–4 mg per sponge, systemic application 10 μmol/kg intravenous	systemic efficacy was shown for (DSS)6-CPAA, a bone targeting fusion probe, but not for CPAA alone. Experiments were controlled with teriparatide and inactive (DSS)6 derivates

FKVP
discovery: design from FK506 as lead structure				target: FKBP12
MoA: sequestration of inhibitory FKBP12 from type I BMP receptors
system			readout	application, concentration	additional information	references
unclear in cellulo			competed NFATc2 dephos-phorylation conditioned by FK506	0.001 μM FK506 vs 10 μM FKVP	FK506 inhibits NFAT2c dephosphorylation, competed by FKVP	2019(116)
in cellulo	HEK293 (human)	target specific	competed FKBP12 binding to ALK1,2,3,6	0.2 μM
	Jurkat (human)	pathway specific	luciferase reporter (NFAT)	0.0001–10 μM	Dose-dependently reduced by FK506 but not FKVP
		pathway specific	luciferase reporter	0.0001–1 μM	competed by LDN193189, not competed by noggin, attenuated by FKBP12 KO
		pathway specific	pSMAD1,5,9 (protein)	0.0001–1 μM	competed by LDN193189, attenuated by FKBP12 KO; no effect on pSMAD2,3
	HUVEC (human)	pathway specific	pSMAD1,5,9 (protein); ID1 (mRNA)	0.2 μM-1 μM
in vivo	rat	functional	enhanced wound healing	0.1 mg/kg	type II diabetes rat model
in cellulo	NmuMG (mouse)	functional	luciferase reporter	0.1 μM	activated BMP reporter but not TGFβ reporter	2020(77)

oxtFKb
discovery: phenotypic screen identifies FK506 analogue cluster (C2C12, BRE-luc)				target: FKBP12
MoA: sequestration of inhibitory FKBP12 from type I BMP receptors
system			readout	application, concentration	additional information
in vitro	SPR	target specific	Dissociation constant (Kd): 4 × 10–10 M		FK506: 7 × 10–10 M; no binding between oxtFK and Calcineurin	2021(117)
in cellulo	C2C12 (mouse)	target specific	luciferase reporter	EC50 0.035 μM
	HK-2 (human)	pathway specific	pSMAD1,5,9, ID1 (protein); SMAD6, ID1 (mRNA)	0.002–1 μM
	MES13 (mouse)	functional	reduced Col1a, fibronectin (mRNA)	0.25–1 μM	BMP signaling reduces TGFβ induced fibrosis
	podocytes (mouse, primary)	functional	reduced apoptosis (CAS3,7 activity) in response to stress	1 μM	renal protection model
in vivo	mouse	pathway specific	pSMAD1,5,9, ID1 (protein)	10–100 mg/kg subcutaneous	isolated PBMCs and liver tissue
		functional	blood urea nitrogen, serum creatinine, histopathology scores	30 mg/kg intraperitoneal	renal protection model

^aCompounds were ordered chronologically. According to the underlying biological assay systems, individual read-outs, and measured time points, experimental data was classified into the categories “pathway specific” or “functional”. Pathway-specific experiments show a more direct impact on the BMP pathway (e.g., short-term activation of pSMAD1,5,9 signaling), whereas functional read-outs refer to long-term experiments that capture BMP-dependent outcomes (e.g., differentiation). For brevity, general cytotoxicity studies are not included. The listed dose range refers to data availability, not to the actual activity range of the compound; EC₅₀ and IC₅₀ values were only stated when specifically mentioned by the authors. Subsequent studies utilizing these compounds are referenced, and data is briefly summarized.

^bAdditionally, oxtFK was extensively profiled in vitro and in vivo, to characterize safety, pharmacokinetics, and off-target profile.

Table 4. Bioactivity Profiles of Heterogeneous BMP Stimulator Chemotypes from PDD Approaches^a.

Isoliquiritigenin
discovery: BRE-luc screen (5600 compounds)				target: unknown
MoA: unknown
system			readout	application, concentration	additional information	references
in cellulo	C33A-2D2 (human)	pathway specific	luciferase reporter; pSMAD1,5, ID1,2, pERK (protein)	5–28 μM; luciferase reporter EC50 8.6 μM		2013(125)
	C2C12 (mouse)	pathway specific	pSMAD1,5, ID1,2 (protein)	1–10 μM	in C2C12 pSMAD1,5 and ID1,2 levels were not elevated
		functional	ALP (activity)	1–10 μM
in vivo	zebrafish	functional	ventralized phenotype	5–10 μM
in vivo	zebrafish	pathway specific	pSmad1,5 (protein)			2016(143)
		functional	rescue of BMP dependent phenotype (cartilage formation)	5 μM
in cellulo	primary bronchial fibroblasts (human)	pathway specific	pSMAD1,5 (Protein)	25 μM	TGFβ-1 induced pSMAD was influenced	2020(144)
		functional	rescue of TGFβ-1 induced phenotype	25 μM
in cellulo	PSCs (human)	functional	cardiac mesoderm induction	10 μM	induction was not enhanced by Isoliquiritigenin	2021(145)

KM11073
discovery: not disclosed				target: unknown
MoA: unknown
system			readout	application, concentration	additional information	references
in cellulo	C2C12 (mouse)	pathway specific	pSMAD1,5,9, p-p38 (protein)	10 μM	pSMAD1,5,9 was not enhanced	2015(127)
		functional	Alp, Bmp2,4,6,7,9 (mRNA); ALP (activity)	10 μM, ALP activity 0.1–30 μM	inactive without BMP addition; controlled with p38 inhibitors (attenuated effect) and RAS, PI3K, AKT inhibitors (effect not influenced)
in vivo	zebrafish	pathway specific	runx2, bmp2, opn, alp (mRNA)	1 μM
		functional	enhanced skeletal development	1 μM
	mouse	functional	calvarial bone formation	local application of sponge with 5 μL of 2.5 or 5 mM

PD407824
discovery: BRE-luc screen (4,000 compounds)				target: CHK1
MoA: enhanced SMAD4 availability for BMP-SMAD through reduced SMAD2 levels
system			readout	application, concentration	additional information	references
in cellulo	C2C12 (mouse)	pathway specific	luciferase reporter; Id1,2 (mRNA); pSMAD1,5,9 (protein)	0.01–100 μM, EC50 0.12–12.3 μM (reporter); 1–10 μM (mRNA, protein)	EC50 depending on BMP level; inactive without BMP	2016(133)
		mechanistic	reduced p21, reduced SMAD2,3 with increased linker phosphorylation, pSMAD1,5,9 increased, increased SMAD1,5,9-SMAD4 binding (all protein); Id2 (mRNA)	10 μM	Id2 expression experiment controlled with CHIR124 (CHKi) (similar effect to PD), Wee1i (inactive),PKCi (inactive), CDKi (attenuated PD effect), CDK9 KO, p21 KO, CHK1+p21 or CDK9 dual KO (confirmed mechanism)
		functional	ALP (activity); Alp, Col1, Opn, Ocn (mRNA); Alizarin staining	0.02–2 μM (ALP activity); 1 μM (mRNA, Alizarin staining)	inactive without BMP
	hESC-H1/H9 (human)	mechanistic	reduced p21, increased SMAD2,3 linker phosphorylation, pSMAD1,5,9 increased (all protein); ID2 (mRNA)	10 μM
	hESC-H1/H9 (human)	functional	ID2, Brachyury (mRNA); cardiac differentiation (various markers); cytotrophoblast differentiation (various markers)	0.1–2.5 μM
in cellulo	KhES-1 (human)	functional	pSMAD1,5,9; enhanced retinal differentiation	1 μM	inactive without BMP; CHK1 inhibition might be involved irrespective of BMP pathway	2022(146)

Ventromorphins*
discovery: BRE-luc screen (643,432 compounds)				target: unknown
MoA: unknown
system			readout	application, concentration	additional information	references
in cellulo	C33A-2D2 (human)	pathway specific	luciferase reporter; pSMAD1,5,9, pERK (protein)	0.003–50 μM, EC50 < 1 μM (reporter);		2017(128)
	C2C12 (mouse)	functional	morphology (visual); Affymetrix gene expression array	6.25–25 μM	25 μM of selected ventromorphins show partial overlap in gene expression with low dose BMP4
in vivo	zebrafish	pathway specific	bmp2b, eve1, szl, vent, vox, chd (in situ hybridization)	6.25 μM
		functional	ventralized phenotype	0.1–50 μM
in cellulo	primary mouse cortex cells	functional	rescue of Tamoxifen induced phenotype	10 μM	not in typical BMP context; SJ000291942 was tested for rescue of a Tamoxifen induced phenotype and showed little effect	2020(147)
in cellulo	primary rat lens epithelial cells	functional	rescue of TGFβ-2 induced lens EMT	25–80 μM	BMP4,7 rescued the TGFβ-2 induced phenotype, ventromorphins did not. Ventromorphins increased pERK1,2 (protein)	2021(148)

DMP-PYT
discovery: phenotypic screen, RUNX2 activity (12,259 compounds)				target: unknown
MoA: in part through activation of canonical Wnt signaling
system			readout	application, concentration	additional information	references
in cellulo	C2C12 (mouse)	pathway specific	luciferase reporter (BMP and Wnt); pSMAD1,5,9, β-catenin, p-p38 (protein)	10 μM	pGSK3β, pJNK was not increased; pSMAD was antagonized by LDN-193189; p-p38 independent of BMP	2017(139)
		functional	ALP (activity); Alp, Osx, Runx2, Ocn (mRNA); OSX, WNT3a, BMP2,4,6,7 (protein)	5–10 μM	increased activity together with BMP
	MC3T3-E1 (mouse)	pathway specific	nuclear localization of β-catenin (protein); enhanced TCF-DNA binding (ChIP)	10 μM
		functional	ALP (activity); Alizarin staining; Alp, Runx2, Bmp4 (mRNA)	10 μM	antagonized by LDN-193189
in vivo	zebrafish	pathway specific	enhanced pSMAD1,5,9, β-catenin (protein)
		functional	enhanced ossification	2.5–10 μM

DIPQUO
discovery: phenotypic screen, ALP activity (47,000 compounds)				target: GSK3β
MoA: in part through activation of β-catenin signaling
system			readout	application, concentration	additional information	references
in vitro	kinase assay	mechanistic	GSK3β inhibition	10 μM	multiple kinases inhibited over 70%	2019(140)
in cellulo	C2C12 (mouse)	mechanistic	CETSA	10 μM	GSK3β stabilization	2021(141)
		pathway specific	p-p38 (protein)	2.5–10 μM	slight suppression of p54 (JNK); no activity on PI3K/Akt, ERK, pSMAD2,3,1,5,9; cotreatment of “subthreshold” DIPQUO (5 μM) with GSK3β inhibitor
		pathway specific	nuclear β-catenin accumulation (protein)	0.5–10 μM
		functional	ALP (activity); Runx2, Osx, Alp, Ocn, Oa (mRNA); RNA seq. (ingenuity pathway analysis)	0.1–30 μM, EC50 6.27 μM (ALP activity); 10 μM	mechanistical control via MKK3,6 overexpression, JNK inhibition, p38 activation (U46619), all did not induce C2C12 differentiation; p38 KD attenuated ALP induction; cotreatment of “subthreshold” DIPQUO (2.5 μM) with GSK3β inhibitor; controlled with XAV-939, OE of constitutive active GSK3β
	MSCs (human)	functional	Alizarin staining	10 μM	cotreatment of “subthreshold” DIPQUO (5 μM) with GSK3b inhibitor
	RAW264.7 (mouse)	functional	osteoclast activation	5 μM	osteoclastic differentiation was slightly activated, note that 5 μM is below the EC50 of DIPQUO for osteoblastic differentiation
	BRITER (mouse)	pathway specific	p-p38 and pMKK3,6 (protein)	10 μM	no activity on TAK1, ASK1, MLK3, MEKK3, TPL2, TRAF6
	HEK293 (human)	pathway specific	TCF/LEF luciferase reporter	10 μM
in vivo	zebrafish	functional	Alizarin staining, fin regeneration model	15 μM	controlled with inactive analogues

Sb4*
discovery: BRE-luc screen (63,608 compounds)				target: unknown
MoA: unknown
system			readout	application, concentration	additional information	references
in cellulo	HEK293 (human)	pathway specific	luciferase reporter	0.0005–10 μM, EC50 0.074 μM	remained active in the presence of noggin or LDN-193189	2019(129)
		pathway specific	pSMAD1,5,9 (protein); ID1,2 (mRNA)	1–10 μM	attenuated decay; no effect on pSMAD2,3, pJNK, pERK1,2, p-p38, pTAK1
	PRECs (mouse)	pathway specific	pSMAD1,5,9 (protein)	0.1–0.3 μM
in cellulo	MSCs (human)	functional	pSMAD1,5,9 (protein); Alizarin staining; ALP, RUNX2 (mRNA and protein)	0.1 μM	complex and long treatment scheme, osteogenic phenotype induced by KGN (autophagy modulator), inhibited by KGN+3MA (autophagy inhibitor), rescued by sb4	2022(149)
in cellulo	KhES-1 (human)	functional	retinal differentiation	1–5 μM	no activity shown; PD also tested in this model	2022(146)
in cellulo	RB20 (mouse ESCs)	functional	Blimp1, Prdm14, Tfap2c, Nanos3, Stella (mRNA level)	200 μM	germ-cell production model; very high concentrations used	2022(150)

SY-LB-35*
discovery: pharmacophore-based design				target: unknown
MoA: unknown
system			readout	application, concentration	additional information	references
in cellulo	C2C12 (mouse)	pathway specific	pSMAD1,5,9, pAKT, pPI3K, pERK, pJNK (protein); PI3K activity (after IP)	0.01–10 μM	effects not dose dependent, attenuated by dorsomorphin	2022(130)
		functional	cell viability and number; cell cycle shift toward G2/M and S phase	0.01–10 μM	effects not dose dependent

Chromenone 1
discovery: phenotypic screen, cardiogenic mesoderm patterning (7,000 compounds)				target: unknown
MoA: enhanced SMAD4 availability for BMP-SMAD through reduced SMAD2 levels
system			readout	application, concentration	additional information	references
in vitro	kinase assay	mechanistic	no kinase inhibited	1 μM	484 kinases tested	2022(137)
in cellulo	ESCs (mouse)	functional	attenuation of DMH1 induced differentiation	0.02–5 μM, IC50 0.2 μM
	C2C12 (mouse)	pathway specific	SMAD2 reduction, pSMAD1,5,9 after preincubation (protein)	1–5 μM
in cellulo	C2C12 (mouse)	pathway specific	Id1, Id2 (mRNA); qPCR profiler	1–5 μM	qPCR profiler shows enhanced BMP target-gene expression and inhibited TGFβ target-gene expression
		functional	ALP (activity); morphology; Runx2, Alp, Ocn, Col1a1, Osx (mRNA)	0.01–5 μM, EC50 0.34 μM (ALP activity)
	SaOS-2 (human)	functional	von Kossa staining	0.025 μM
	HEK293 (human)	pathway specific	luciferase reporter	0.1–5 μM	BRE signal enhanced, SBE signal attenuated

CGS-15943*
discovery: phenotypic screen, cardiogenic mesoderm patterning (1,408 compounds)				target: CK1δ,ε and PI3Kα,γ
MoA: enhanced SMAD1 levels
system			readout	application, concentration	additional information	references
in vitro	kinase assay	mechanistic	kinase inhibition, including IC50 values	0.001–20 μM	408 kinases tested, additionally full length kinase assays including inhibition kinetics	2022(138)
	crystal structure	target specific	crystal structure of CGS-15943 bound to CK1δ (PDB 7NZY)
in cellulo	ESCs (mouse)	pathway specific	Id1 (mRNA)	2.5 μM
		functional	attenuation of DMH1 induced differentiation	0.02–5 μM, IC50 0.5 μM	controlled with Adenosin modulators, DNA-PKi
	C2C12 (mouse)	pathway specific	qPCR profiler	5 μM	enhanced BMP dependent gene regulation
		pathway specific	pSMAD1,5,9 (protein); Id1 (mRNA)	5 μM	no influence on direct signaling
		functional	pSMAD1,5,9 (protein); Id1 (mRNA)	1–10 μM	after preincubation enhanced pSMAD1,5,9 signaling and Id1 expression
		functional	ALP (activity); morphology; Runx2, Alp, Ocn, Col1a1, Osx (mRNA)	0.01–10 μM	inactive without BMP; controlled with Adenosin modulators, AhR agonist, DNA-PKi, PI3Ki, CK1i, PI3Ki-CKi synergy experiments
	SaOS-2 (human)	functional	von Kossa staining
	HEK293 (human)	pathway specific	luciferase reporter	0.05–10 μM	BRE activated, SBE, TopFlash, GLI not influenced; inactive without BMP; controlled with Adenosin modulators, AhR agonist, PI3Ki, CK1i
	U2OS (human)	functional	cell painting assay	10 μM	similar profile to PIK93 (PI3Ki)
in vivo	zebrafish	functional	ventralized phenotype; eve1, vox, gsc (in situ hybridization)	1–50 μM

Compound 2b
discovery: phenotypic screen, cardiogenic mesoderm patterning				target: GSK3β
MoA: in part through activation of β-catenin signaling
system			readout	application, concentration	additional information	references
in vitro	kinase assay	mechanistic	GSK3β inhibition	0.5 μM		2023(142)
in cellulo	C2C12 (mouse)	pathway specific	Id2,3 (mRNA); SMAD2,3,1,5,9, β-catenin translocation (protein)	0.3 μM	short-term Id1, Runx2 expression not influenced; SMAD signaling not influenced; controlled with CHIR, DIPQUO
		functional	ALP (activity); Runx2, Osx, Alp, Ocn, Col1a1 (mRNA); morphology	0.01–10 μM, EC50 0.13 μM (ALP activity); all others tested at 0.3 μM	inactive without BMP; controlled with CHIR; synergy Chromenone 1
		functional	proliferation inhibition	0.03–10 μM

^aCompounds were ordered chronologically. According to the underlying biological assay systems, individual read-outs, and measured timepoints, experimental data was classified into the categories “pathway specific” or “functional”. Pathway-specific experiments show a more direct impact on the BMP pathway (e.g., short-term activation of pSMAD1,5,9 signaling), whereas functional read-outs refer to long-term experiments that capture BMP-dependent outcomes (e.g., differentiation). General cytotoxicity studies were not included. The listed dose range refers to data availability, not to the actual activity range of the compound; EC₅₀ and IC₅₀ values were only stated when specifically mentioned by the authors. Subsequent studies utilizing these compounds are referenced, and data is briefly summarized.

2.2. BMP Stimulators from Rational, Target-Centric Approaches

Small-molecule modulators of the BMP pathway could be rationally developed based on well-studied mechanisms of two major players in the BMP/SMAD signaling cascade, i.e., E3-ubiquitin ligase SMURF1 and FKBP12 (Figure 1A).

After phosphorylation of SMADs by the BMP receptors, the E3-ligase SMURF1 limits the activity of pSMADs by marking them for proteasomal degradation. Inhibiting SMURF1 activity can thus be reasoned to potentiate cellular pSMAD response following pathway initiation through BMP. The first structurally disclosed SMURF1 inhibitor was 1,3,5-triazine SVAK-12 (Figure 4A), which does not qualify as a chemical probe due to its low-potency promiscuous bioactivity (Table 3).¹⁰⁰ Employing structural information on the SMURF1-SMAD1 interaction enabled the targeted development of a small series of compounds (e.g., A01, Figure 4A) that specifically disrupt the recognition of phosphorylated serines in the SMAD1 linker by the SMURF1-WW1 domain.¹⁰¹ The same researchers later demonstrated the feasibility of targeting other domains of SMURF1 with B06 (Figure 4A), which specifically interrupts ubiquitin transfer from E2 ligases to the SMURF1-HECT domain.¹⁰² Similarly identified from molecular docking studies, the chalcone derivative CPAA (Figure 4A) was recently studied in vivo and elegantly demonstrated feasibility of targeting SMURF1 for osteoporosis treatment.¹⁰³ However, chalcones exhibit promiscuous bioactivity against a wide variety of molecular targets, strictly limiting their utility as chemical probes.^104,105

Figure 4.

Chemical structures of BMP stimulators from rational, structure-guided approaches. (A) Targeting SMURF1: SVAK-12, CPAA, A01, and B06 have been discovered from in silico screens against ubiquitin E3 ligase SMURF1, attenuating the degradation of BMP-specific pSMADs, leading to amplified signaling output. (B) Targeting FKBP12: Macrolides FKVP and oxtFK sequester inhibitory FKBP12 from BMPR-I and have been derived from FK506 as more selective, non-immunosuppressive analogues.

On a cautionary note, it should be mentioned that SMURF1 is not an exclusive E3 ligase for SMADs but has additional targets that are not necessarily linked to BMP. Among these are the bone regulator RUNX2, TNF receptor associated factor 4 (TRAF4), Ras homologue family member A (RhoA), as well as the estrogen receptor α (ERα).^106−109 While reduced degradation of RUNX2 might be beneficial in a pro-osteogenic context, the effect mediated by enhanced stability of other SMURF1 targets should also be considered. In fact, A01 was already successfully employed as a chemical tool to modulate ERα and RhoA signaling in dosages overlapping with BMP activity.^109,110 The extent of these effects likely depends on the mode of SMURF1 inhibition, making A01 a more suitable tool than B06, which targets SMURF1 functionality in a more fundamental manner.

On the receptor level, it was recognized that binding of FKBP12 prevents the activation of type I receptors by blocking the GS domain.¹¹¹ The pro-osteogenic effect of FKBP12 binding immunosuppressants has been known for some time, and the underlying mechanism was ultimately uncovered as sequestration of the inhibitory FKBP12, leading to enhanced type I receptor phosphorylation and activity.^112−115 To avoid undesired immunosuppression, FK506 analogues FKVP and oxtFK (Figure 4B) were rationally developed by altering the calcineurin binding domain of FK506 toward selective BMP activators.^116,117 FKBP12 was validated as a viable target to stimulate BMP signaling by comparing distinct FKBP inhibitor chemotypes as well as performing FKBP12 knockout studies in vitro.^116,117oxtFK was eventually devised from sophisticated SAR studies: A compound collection was designed and synthesized by late-stage functionalization of FK506 or FK520, targeting the terminal alkene via Heck reaction or cross-metathesis as well as vital functionalities for the 3D conformation of the calcineurin binding domain.¹¹⁷ Both FKVP and oxtFK were highly potent, with EC₅₀ values in the nanomolar range in different cellular assays, and were successfully tested in models for wound healing and renal protection in vivo (Table 3). Interestingly, in spite of FKBP12 binding similarly to TGFβ type I receptors, FK506 and FKVP showed no activation of TGFβ signaling.¹¹⁶oxtFK was additionally profiled in an extensive assay panel characterizing promiscuity, safety, and pharmacokinetics and demonstrated an overall favorable profile and pronounced target selectivity.¹¹⁷

At the moment, oxtFK provides the best-characterized pharmacological tool for the direct activation of BMP/SMAD signaling. Unfortunately, it is not yet commercially available. In addition, it should be noted that selectivity and safety concerns exist from the fact that FKBP12, just as other members of the FKBP family, engages multilaterally with a variety of different binding partners with different physiological roles.^118,119 These interactions might also be influenced when targeting the FK506 binding site and potentially cause adverse effects not linked to BMP, especially in complex in vivo models. Noteworthy is the recent development of bicyclic [4.3.1] aza-amides as highly potent FKBP inhibitors,¹²⁰ a chemotype that had previously been employed (i.e., DBD) to validate FKBP12 as a viable target.¹¹⁷ Their synthetic accessibility enables more straightforward functionalization compared with FK506 derivatives as complex macrolides.

2.3. Target-Agnostic Discovery Screens Yielding a Heterogeneous Group of BMP Stimulators

The previously exemplified strategies toward novel stimulator modalities for the BMP pathway underscored that a profound knowledge of the biology of key pathway regulators supports and accelerates their rational development and in vivo translation. Albeit promising candidates have been developed, it is desirable to expand the druggable target space for effective BMP stimulation beyond SMURF1 and FKBP12. In this regard, phenotypic drug discovery (PDD) campaigns provide a powerful means of unbiased identification of novel drug candidates, targets, and modes of action (MoA) that would not have been discovered by target-centric approaches (TDD). In the field of BMP stimulator discovery, mostly conventional phenotypic screening systems were employed, including BMP-responsive element luciferase (BRE-luc) or other BMP-dependent reporter assays (e.g., alkaline phosphatase, ALP, activity) in immortalized cell lines.

For many BMP stimulators identified from such screens, the underlying mechanisms are not very well understood, and the responsible target(s) have mostly not been unraveled. Hence, they constitute a heterogeneous group of diverse chemotypes with varying degrees of biological characterization. Here, small-molecule candidates entirely lacking target or MoA hypotheses were not assigned to any specific category.

Isoliquiritigenin was among the first described BMP stimulators (Figure 5A), discovered from a phenotypic screen of 5600 commercially available drugs and chemicals with known bioactivities in a human cervical carcinoma clonal cell line (C33A-2D2) with BRE-luc readout (Table 4).¹²⁵Isoliquiritigenin is a frequently used tool and reference compound for BMP stimulation. However, as a chalcone it is well recognized for promiscuous biological activity and polypharmacology and, thus, does not qualify as a chemical probe,^104,126 similar to the above-described CPAA (Figure 4A) as a BMP activator. Quinoline KM11073,¹²⁷Ventromorphins,¹²⁸ and the benzoxazole Sb4(129) form a group with similarities concerning their degree of biological characterization (Figure 5A). They all showed BMP activity in different test systems, and Ventromorphins and KM11073 were also characterized in vivo. Their cellular targets remain elusive, and the MoAs have not been determined. Ventromorphins and Sb4 were utilized by other groups following their publication, although with mixed success and partly in concentrations strongly deviating from their original description (Table 4). SY-LB-35 was more recently disclosed, and while the pharmacophore-fusion approach is an interesting concept,¹³⁰ functional characterization in a BMP-dependent context and determination of pathway selectivity are missing.

Figure 5.

Chemical structures of BMP stimulators from target-agnostic discovery efforts. While PDD approaches introduced a variety of novel chemotypes, target identification remains a major challenge. Only a few of the selected BMP stimulators have been mechanistically characterized. (A) Compounds with unknown targets and MoA, (B) BMP potentiators, and (C) BMP synergizers with distinct targets and MoAs.

The carbazolomaleimide PD407824 (Figure 5B) is a known pan-kinase inhibitor^131,132 and was identified as a BMP stimulator from a cell-based screen of 4000 compounds in C2C12 myoblasts using a BRE-luc reporter for readout.¹³³ It was characterized as a BMP sensitizer (herein classified as “potentiator”) via a mechanism that involves TGFβ/SMAD inhibition. More specifically, checkpoint kinase 1 (CHK1) inhibition by PD407824 reduced p21 activity, in turn activating cyclin-dependent kinase 9 (CDK9), which reduced the TGFβ-SMAD2 stability. As a consequence, the availability of co-SMAD4 increased for the BMP-SMADs. This indirect BMP potentiator mechanism was dissected with small-molecule inhibitors of involved kinases and known PD407824 off-targets, along with knockout models for the involved proteins. A shortcoming remains that, especially in the context of stem cell differentiation, the individual contributions of indirect BMP sensitization and CHK1 inhibition cannot be clearly separated. CHK1 and WEE1 are nanomolar targets of PD407824 and core regulators of cell cycle progression (Table 4).^134,135 Inhibiting this checkpoint might cause genomic instability in rapidly replicating cells like pluripotent stem cells at the onset of differentiation.¹³⁶ In fact, we have observed pronounced cell toxicity already at low doses of PD407824 in all our BMP-dependent assay setups that limited its use as a reference compound.^137,138

In an attempt to introduce a novel approach for probing the BMP pathway, our group built on lessons learned from developmental biology: A phenotypic yet target-agnostic, BMP-dependent, HTS-compatible assay during mesoderm patterning of embryonic stem cells was devised to identify novel BMP stimulator chemotypes.¹³⁷ Using this morphogenic stem cell-based assay, a chemical diversity screen of 7000 small molecules was performed, followed by a panel of secondary BMP-dependent assays. These stringent hit filtering and validation criteria furnished 4H-chromen-4-one Chromenone 1 (Figure 5B) as a novel osteogenic BMP stimulator. Chromenone 1 appeared to share a similar MoA to PD407824, as it potentiated BMP signaling outputs via negative TGFβ feedback loops but exhibited higher potency and efficacy and low toxicity. Strikingly, a kinome scan showed no activity against any of the tested kinases (including CHK1) (Table 4). The molecular target of Chromenone 1 has not been identified yet, but the data hint toward an unrecognized regulatory mechanism of SMAD stability. Thus, Chromenone 1 exhibited a unique MoA via kinase-independent negative TGFβ feedback that enhanced the nuclear BMP-SMAD signaling outputs. In addition, SAR studies with 29 analogues provided structurally related inactive derivatives (e.g., Chromenones 2 and 3) as suitable negative control probes for mechanistic studies. However, it should be kept in mind that both PD407824 and Chromenone 1 inhibit the TGFβ pathway by nature of their mechanism and, therefore, do not selectively perturb BMP.

In view of such intrinsic pathway selectivity concerns, the recent discovery of triazolo[1,5-c]quinazoline CGS-15943 (Figure 5B) as a new-in-class BMP amplifier (here classified as “potentiator”) presents a significant leap forward.¹³⁸ It was discovered from the same stem-cell-based PDD platform as Chromenone 1. A combination of cellular mechanistic studies, chemical biology methodology, and holistic target deconvolution (e.g., transcriptomics, cell painting assay, CPA, and kinome profiling) established a unique MoA of dual targeting of casein kinase 1 (CK1) and phosphatidylinositol 3-kinase (PI3K) isoforms (Table 4). As a consequence, CGS-15943 and optimized, more potent analogues stimulated cellular BMP outputs via enhanced and sustained availability of SMAD1, which is strictly dependent on a minimal BMP input. This was shown in a variety of BMP-dependent systems and species, including in vivo in zebrafish (Table 4). Importantly, CGS-15943 exhibited a clean kinome profile, had no effects on TGFβ/SMAD, Wnt/β-catenin, and sonic hedgehog pathways, and did not target any of the well-recognized BMP effectors that are associated with pathway promiscuity or toxicity-related liabilities (e.g., GSK3β, CHK1). SAR studies with >50 derivatives established key pharmacophoric features and provided several inactive analogues (e.g., Compound 4a) as potential negative controls.

Notably, no direct link between these CK1/PI3K isoforms and BMP-SMAD regulation was reported before, highlighting the strength of PDD to uncover new targets and unrecognized polypharmacology signatures. Although selectivity was assessed within selected developmental signaling pathways, it should be considered that CK1 and PI3K are involved in a variety of cellular processes. Hence, depending on the biological system, BMP-independent effects caused by their inhibition should be reflected and, where necessary, controlled with isoform-selective inhibitors.

Together, a number of distinct BMP stimulator chemotypes have been introduced that share a MoA which leads to potentiation of BMP-dependent signaling outputs. Among these, PD407824, Chromenone 1, and CGS-15943 have been most comprehensively characterized. In view of chemical probe criteria, we believe Chromenone 1 and CGS-15943 are well suited as decent pharmacological tools. Their very distinct MoAs present an asset for studying BMP signaling potentiation under (patho)physiological conditions.

2.4. Synergizers Indirectly Enhance BMP-Dependent Outcomes

As BMP synergizers we selected compounds that do not directly target the BMP pathway but enhance BMP-dependent cellular processes, primarily osteogenesis.

DMP-PYT (Figure 5C) resulted from an approach aiming to develop osteogenic compounds that target BMP and Wnt signaling simultaneously.¹³⁹ This targeted polypharmacology is an interesting concept; however, it is unclear whether addressing both pathways concomitantly is beneficial over individual pathway regulation. Wnt and BMP are finely tuned, and simultaneously hitting the therapeutic range of both pathways with one compound might pose difficulties.

DIPQUO(140,141) and Compound 2b(142) are GSK3 inhibitors, thereby activating canonical Wnt signaling (Figure 5C, Table 4). As indicated above, the cooperation between Wnt and BMP in osteogenesis is well established, and thus, the pronounced pro-osteogenic effect of these compounds is not surprising.^98,99 Imidazolequinolinone DIPQUO was discovered from a phenotypic screen in C2C12 cells (47,000 commercial diversity set, ALP readout). Carbazolomaleimide Compound 2b was designed from a pharmacophore fusion approach of PD407824 and a 3,4-diarylmaleimide hit cluster.¹⁴² The latter was identified from the same PDD platform that furnished Chromenone 1 and CGS-15943. A comparative analysis of different GSK3 inhibitors showed striking phenotypic differences and contrasting dependency on BMP.¹⁴² The underlying mechanisms driving these distinct cell fates were not unraveled but were hypothesized to derive from the differing regulation of non-canonical BMP effectors. Notably, while DIPQUO was capable of stimulating BMP-dependent osteogenesis in the complete absence of BMP, Compound 2b strictly required minimal BMP input for synergistic efficacy and did not act via the canonical effectors (i.e., BMP-SMAD-Id1 axis).

Therefore, Compound 2b is currently the only SMAD-independent, non-canonical, yet BMP stimuli-dependent synergizer chemotype. The utility of this new synergizer profile was further underlined by demonstrating how non-canonical (Compound 2b) and canonical (Chromenone 1) BMP potentiators additively amplified osteogenic efficacy in a BMP-dependent manner.¹⁴² However, it should be noted that Compound 2b can only be used as a tool for this purpose within a narrow, submicromolar dose range. A kinome profile is missing, but the chemotype suggests pan-kinase activity that should be carefully considered when performing mechanistic studies with this compound.

3. Conclusion

Small-molecule probes can serve as valuable and versatile tools in pharmacology to investigate molecular signaling pathways in health and disease. However, their quality is decisive for devising meaningful data that can be translated to coherent molecular networks.^4,5 In biomedical research study design, active chemical probes, corresponding negative controls, and their dosing should be carefully chosen each time in view of the individual application and biological context.¹⁵¹ Unfortunately, probes with promiscuous, non-specific activities are still frequently used, simply because they have been established at times when no higher quality probe was available.¹⁵² Therefore, it is vital to constantly explore the availability of optimized probes for a given target or the introduction of novel probe modalities with unprecedented modes of action and targets in a specific field. Additionally, probes should be distinguished from drug candidates: Key features of a chemical probe are potency, selectivity, a defined target and MoA, as well as availability of inactive derivatives as negative controls.^4,153 The demonstration of in vivo efficacy is neither necessary nor sufficient to define a chemical probe, although it might support in vivo target validation in a disease context.^4,154

In this Perspective, we made an effort to consider such chemical probe aspects for the selection of proper small-molecule modulators of the BMP signaling pathway. These have been derived from both target-centric and target-agnostic (PDD) approaches, the latter of which turned out to critically expand the druggable target space of BMP stimulators.

After the discovery of Dorsomorphin as the first small-molecule BMP inhibitor in 2008, continuous efforts were largely centered on type I receptors and led to a number of highly potent and selective derivatives from this chemotype. While Dorsomorphin itself cannot be recommended as a chemical probe, LDN-212854, OD36 and Compound 23 qualify as suitable pharmacological tools among the pyrazolo[1,5-a]pyrimidine BMP inhibitor chemotype, particularly with regard to their ALK1/2 selectivity and biochemical and cellular potency. However, a total of 16 distinct BMPR-I chemotypes have been introduced to date, from which individual candidates serve as excellent probes. Among these, the 3,4-diphenylpyridine chemotype has been developed toward an attractive probe set, with M4K2234 and its inactive negative control M4K2234NC as prime pharmacological tools (listed by the SGC). M4K2234 exhibits excellent biochemical and cellular potency, selectivity for ALK1/2, and an overall clean kinome activity profile. From the remaining BMPR-I inhibitor chemotypes, we consider 2-aminopyrazine-3-carboxamide Ullrich 23 as another advanced and well-suited chemical probe that has been comprehensively profiled in vitro.

The availability of these probes with distinct chemotypes is an asset for their use as pharmacological tools in comparative mechanistic studies. However, the field still lacks inhibitors that are selective for the ALK3 and ALK6 receptor isoforms. Similarly, there have only been two inhibitors recently reported that target BMPR-II: CDD-1653 and Compound 8a potently and selectively inhibit BMPR2 as one of the three type II receptors. Their utility as pharmacological tools remains to be shown since BMPR2 perturbation under (patho)physiological conditions is underexplored.

In comparison to the development of BMP inhibitor probes, stimulators appear to lag behind. There is currently no “gold standard” stimulator modality for universal application. This might be accounted for by the intrinsic challenge of developing modulators that activate or enhance biological processes (e.g., enzymes, ion channels, protein–protein interactions, or entire signaling cascades). A profound knowledge of the underlying regulators and effectors is required for rational development. There is also not one ultimate target or mechanism that confers stimulation of the BMP pathway. However, several options exist, depending on where in the signal transduction cascade the pathway is targeted. Herein, we classified available small-molecule BMP stimulators into “activators”, “potentiators”, and “synergizers”.

Two regulators in the BMP/SMAD signaling cascade have emerged to target the BMP pathway for stimulation. These are SMURF1, which controls cellular SMAD1 levels via proteasomal degradation, and FKBP12 as the “physiological brakes” of BMPR-I, blocking its activation. Rational, structure-guided approaches could be utilized for the development of chemical probes that interact with these BMP effectors. A01 is a potent SMURF1 inhibitor that can be used to enhance and prolong pSMAD1,5,9 activity in the presence of active BMP signaling (i.e., “potentiator” mechanism).

Macrolides oxtFK and FKVP are currently the only probes with a well-defined molecular mechanism that can directly activate BMP signaling at the receptor level. They potently sequester the inhibitor FKBP12, leading to enhanced type I receptor phosphorylation and activity with nanomolar efficacy in different cellular assays. Interestingly, their MoA was demonstrated to be selective against TGFβ signaling and proved to be effective in models for wound healing and renal protection in vivo. Although the calcineurin-sparing FKBP12 inhibition by design prevented undesired immunosuppressant activity, it is not clear whether targeting FKBPs will be associated with other off-target liabilities. Still, oxtFK presents the best-characterized pharmacological tool for the direct activation of BMP/SMAD signaling.

All other reported small-molecule BMP stimulators were derived from unbiased PDD screening campaigns in an attempt to identify new probes or drug candidates along with unrecognized targets and MoAs. For the majority of BMP stimulators identified from such screens, the underlying mechanisms were poorly characterized, largely lacking target hypotheses. This circumstance makes it difficult to recommend preferred tools, as they all do not completely fulfill the strict criteria that define high-quality chemical probes. However, among the most comprehensively profiled BMP stimulators that potentiate signaling outputs in a BMP-dependent fashion, Chromenone 1 and CGS-15943 are decent tools. Both have shown potent BMP potentiator efficacy in a number of BMP assays in vitro and in vivo but with distinct MoAs: Chromenone 1 enhances canonical BMP/SMAD signaling via a negative TGFβ/SMAD feedback in a kinase-independent fashion. CGS-15943 also amplifies BMP signaling output but through the enhanced and sustained availability of BMP-SMADs without affecting TGFβ or other related developmental signaling pathways. Dual targeting of CK1/PI3K isoforms conferred this activity while overall kinome selectivity is high. Importantly, inactive derivatives are available as negative controls from SAR studies for both of these BMP potentiator chemotypes. Their very distinct MoA presents an asset for studying BMP signaling under (patho)physiological conditions.

The BMP synergizers DIPQUO and Compound 2b are not selective for the BMP pathway but can synergistically enhance BMP-dependent cellular outcomes. They act, at least partially, through adjacent signaling pathways that act cooperatively with BMP. Thus, their applicability strongly depends on cell and tissue context.

4. Methods

4.1. Number of Chemotypes

For the selection of BMP modulators, this Review focused on peer-reviewed substances. Patent literature was not included. Names of chemotypes were chosen by determination of the key pharmacophoric scaffold in places on the basis of available (predicted) crystal structures in complex with the protein of interest. In other cases, names suggested by the original authors were used.

4.2. Literature Analyses

Referring to Figure 1B, for the determination of total BMP modulator publications, we conducted two separate literature analyses on July 25, 2023, using the search engines SciFinderⁿ and SciteAI. Search parameters were chosen manually by scanning the abstracts and titles of the herein mentioned BMP modulators for distinct keywords. Each keyword was selected individually in order to exclude promiscuous keywords leading to false positive results. Thus, some of our cited literature could not be adequately addressed due to missing or promiscuous keywords present in their abstracts (e.g., DIPQUO¹⁴⁰ or SMURF inhibitor¹⁰³). The selected keywords were used for both search engines.

For SciteAI, the numbers of publications were read out manually. Reviews, comments, and clinical trials were excluded using the ‘publication_type’ filter. Resulting publications were not filtered further to avoid user bias. Keyword plurals were included in cases in which they were excluded by the algorithm.

SciFinder literature analysis was performed using SciFinder’s “advanced search”. Reviews, comments, and clinical trials were excluded using the publication exclude function. Numbers of publications were read out using the publication year diagram. Resulting publications were not filtered further to avoid user bias.

4.3. SciteAI

Query input for stimulators:

• (abstract:BMP) AND (abstract:“small molecule” OR “synthetic compound” OR “small molecules” OR “synthetic compounds”) AND (abstract:“activator*” OR “potentiator*” OR “potentiators” OR “sensitizer*” OR “sensitizers” OR “synergizer*” OR “stimulator*” OR “amplifier*” OR “activation of BMP” OR “enhances BMP” OR “increase BMP signal*”)

Query input for inhibitors:

• (abstract:BMP) AND (abstract:“small molecule” OR “synthetic compound” OR “small molecules” OR “synthetic compounds”) AND (abstract:“inhibitor of BMP” OR “inhibitor of bone morphogenic protein” OR “inhibition of BMP” OR “inhibition of bone morphogenic protein”)

Note: A small fraction of relevant literature was excluded by the search algorithm, even though all specified keywords were included. In some cases, BMP modulators were found in the inhibitor search results for being mentioned in the abstract or for false categorization by the algorithm. These false positives could not be sorted out using the NOT operator.

SciFinder

Query input for stimulators:

• (abstract:BMP) AND (abstract:“small molecule” OR “synthetic compound” OR “small-molecule” OR “synthetic compounds”) AND (abstract:“activator” OR “potentiator” OR “sensitizer” OR “synergizer” OR “stimulator” OR “amplifier” OR “activation of BMP” OR “enhances BMP” OR “increase BMP signal”)

Query input for inhibitors:

• (abstract:BMP) AND (abstract:“small molecule” OR “synthetic compound” OR “small-molecule” OR “synthetic compounds”) AND (abstract:“inhibitor of BMP” OR “inhibitor of bone morphogenic protein” OR “inhibition of BMP” OR “inhibition of bone morphogenic protein”)

Note: In some cases, SciFinder was unable to correctly address the keyword “small molecule” and used false adaptations of the keyword, leading to a higher number of results.

SciFinder was utilized for the research conducted in Figure 2B. Initially, the corresponding compound was searched for in the subcategory “Substances”, taking into account its CAS Registry Number. Then, linked references were filtered by “Document Type” - “Journal”. Subsequently, the references listed were individually reviewed to determine whether the compounds were used in assays within those publications. This research was carried out on June 22, 2023.

Glossary

Abbreviations

AKI

acute kidney injury

ALP

alkaline phosphatase

BMP

bone morphogenetic protein

BMPR-I/-II

BMP receptor type I/type II (protein family)

BMPR2

BMP receptor type 2 (protein encoded by the BMPR2 gene)

BRE-luc

BMP-responsive element luciferase

CDK9

cyclin-dependent kinase 9

ChIP

chromatin immunoprecipitation

CHK1

checkpoint kinase 1

CK1

casein kinase 1

CKD

chronic kidney disease

CPAA

2-(4-cinnamoylphenoxy)acetic acid

DIPG

diffuse intrinsic pontine glioma

ERα

estrogen receptor α

ERK

extracellular signal-regulated kinases

FKBP12

FK506 binding protein 12

FOP

fibrodysplasia ossificans progressiva

GFP

green fluorescent protein

GSK3 (β)

glycogen synthase kinase 3 (β)

hiPSC

human induced pluripotent stem cells

HO

heterotopic ossification

Id

inhibitor of DNA binding

JNK

c-Jun N-terminal kinases

MAPK

mitogen activated protein kinase

MoA

mode of action

mRNA

messenger ribonucleic acid

NFAT

nuclear factor of activated T-cells

Ocn

osteocalcin

Osx

osterix

PAH

pulmonary arterial hypertension

PD

PD407824

PDD

phenotypic drug discovery

PET

positron emission tomography

PI3K

phosphoino-sitol-3 kinase

(PI3K)i

(PI3K) inhibitor

pSMAD

phospho-SMAD

qPCR

quantitative polymerase chain reaction

TDD

targed-based drug discovery

RhoA

Ras homology family member A

RunX2

runt-related transcription factor 2

SAR

structure activity relationship

SGC

structural genomics consortium

SMURF1

SMAD ubiquitination regulatory factor 1

TAK1

TGFβ-activated kinase 1

TFs

transcription factors

TGFβ

transforming growth factor β

TRAF4

TNF receptor associated factor 4

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.3c00170.

• Chemical structure directory of all included BMP inhibitors (Figure S1 and S2) as well as BMP stimulators (Figure S3) (PDF)

The authors declare the following competing financial interest(s): D.S. is inventor of patent BMP-MIMETICS (WO 2019/042889).