Comparison of the enzymatic and cellular profiles of clinical JAK inhibitors for the treatment of myelofibrosis

Key Points

• •Ruxolitinib is the most selective and potent of the JAKinibs indicated for treating myelofibrosis.
• •Other JAKinibs show broader kinase inhibition, potentially contributing to off-target toxicities.

Visual Abstract

Abstract

Four Janus kinase inhibitors (JAKinibs), ruxolitinib, fedratinib, pacritinib, and momelotinib, are indicated for myelofibrosis. All inhibit JAK2, but effects on additional kinases vary markedly, shaping their specific clinical pharmacology. Published comparative inhibitory profiles and cellular pharmacology data are incomplete and were determined here. Inhibitory potency and relative selectivity of JAKinibs were determined by full kinome profiling. JAKinib effects on signaling and cell growth were measured using JAK2-dependent and JAK2-independent cell lines. Ruxolitinib was the most potent JAK2 inhibitor (half-maximal inhibitory concentration [IC₅₀] at physiological [1mM] ATP, 2.9nM), followed by fedratinib (17nM), momelotinib (29nM), and pacritinib (39nM), and the most selective. The other JAKinibs inhibited multiple additional kinases beyond the few typically reported for each drug in the literature. Interestingly, pacritinib and momelotinib showed only modest activin A receptor type 1 (ACVR1) inhibition at clinical doses and physiological ATP concentrations, suggesting the role of ACVR1 inhibition in mediating their anemia-related benefits may be overestimated. The observed inhibitory levels are unlikely to translate to a clinically meaningful reduction in hepcidin levels. In JAK2-dependent cellular assays, ruxolitinib showed most potent inhibition of STAT5 phosphorylation (IC_50, 14nM), followed by momelotinib (201nM), pacritinib (421nM), and fedratinib (669nM), and was the only JAKinib that had minimal impact on growth in JAK2-independent cell lines. In conclusion, ruxolitinib was the most potent, and selective JAKinib with no relevant effects in JAK2-independent cells. In contrast, fedratinib, pacritinib, and momelotinib inhibited many other kinases and inhibited cell growth by JAK2-unrelated mechanisms at clinically relevant concentrations.

Introduction

Myelofibrosis (MF) is a rare myeloproliferative neoplasm characterized by abnormal hematopoiesis and formation of fibrotic tissue in bone marrow.1, 2, 3 The clinical phenotype is characterized by splenomegaly, constitutional symptoms, and a variety of blood cell alterations, including anemia.² MF is also associated with high incidence of vascular complications and 10% to 20% incidence of leukemic transformation.^2,4,5 Gain-of-function mutations in 1 of the 3 genes, JAK2, calreticulin, and myeloproliferative leukemia, serve as primary drivers of myeloproliferative neoplasms, capable of independently initiating and promoting disease. These mutations lead to constitutive activation of JAK2/signal transducer and activator of transcription (STAT) 5 signaling, resulting in elevated production of proinflammatory cytokines and overactivation of multiple downstream pathways that promote cell growth and survival.^1,2,6

Four JAK inhibitors (JAKinibs) have been approved by the US Food and Drug Administration for the treatment of MF: ruxolitinib, fedratinib, pacritinib, and momelotinib. Although all 4 JAKinibs target JAK2, their selectivity for other kinases varies and distinct phenotypic profiles have been reported, attributed to these differential off-target effects.⁷ Primary targets for each drug are often assigned as follows: ruxolitinib, JAK1/JAK2; fedratinib, JAK2/fms-like tyrosine kinase 3 (FLT3); pacritinib, JAK2/FLT3/interleukin 1 (IL-1) receptor–associated kinase/activin A receptor type 1 (ACVR1); and momelotinib, JAK1/JAK2/ACVR1.^8,9 These drugs are also known to target other kinases in addition to those mentioned earlier.^10,11 Importantly, with respect to momelotinib, it has been reported that the major metabolite, M21, contributes significantly to treatment effects. M21 presents with metabolite-parent ratios ranging from 1.4 to 2.1, demonstrates ∼40% of the pharmacological activity of the parent compound, and like momelotinib, is also described in the literature as an inhibitor of JAK1/2 and ACVR1.^11,12

Notably, most published data on the kinase inhibitory profiles of the approved JAKinibs were determined under differing assay conditions, often testing a limited number of kinases, and ultimately preventing meaningful comparisons.¹³ Interpreting biological effects arising from differences in the enzymatic and cellular properties of the JAKinibs requires standardization and optimization of the assay conditions used. Intracellular adenosine triphosphate (ATP) concentration, cell membrane permeability, and plasma protein binding need to be considered when translating biochemical potencies of kinase inhibitors to pharmacological effects.¹⁴ For ATP-competing compounds (type 1 inhibitors), such as the 4 JAKinibs used to treat MF, ATP concentration in the assay has a profound impact on measured half-maximal inhibitory concentration (IC₅₀) values.¹⁵ Nevertheless, previous reports of IC₅₀s for JAKinibs often do not take into account differences in ATP concentrations used for these determinations.¹³ For example, seminal studies on JAK2 potency for ruxolitinib, fedratinib, pacritinib, and momelotinib report IC₅₀s of 2.5nM, 3nM, 23nM, and 18nM, respectively. These studies were performed using widely varying or undisclosed ATP concentrations (1mM, kinase Km, 0.15μM, not reported, respectively) and different methods; for example one was a binding assay, the others used different enzymatic assays.16, 17, 18, 19 Moreover, ATP concentrations must not only be consistent across biochemical assays, they must also be biologically relevant to be applicable to the clinical environment.

Recently, the first head-to-head comparative analysis of approved JAKinibs using in vitro, ex vivo, and in vivo models was reported by Kong et al.²⁰ Although this study directly compared the JAKinibs across multiple assays including RNA-sequencing and mass cytometry analyses, the study did not evaluate their full biochemical profiles. Additionally, the use of a single concentration of each JAKinib in ex vivo assays (eg, a colony-forming unit [CFU] assay), may have limited the ability to interpret the findings with regard to the different potencies of these inhibitors.

As more JAKinibs gain regulatory approval, there is an increasing need for head-to-head studies to evaluate differences in kinase inhibition profiles and cellular effects under identical conditions. Such comparative data are essential to help clinicians personalize treatment decisions. A deeper understanding of the relative efficacy and safety of each JAKinib, based on both on (JAK2) and off (non-JAK2)-target effects will allow for more precise matching of therapeutic potential and risk profiles to individual patient needs. Here, we compare inhibitory potency and kinome-wide selectivity profiles of the JAKinibs approved for treatment of MF as well as M21,¹² and compare their effects on signaling pathways and cell proliferation in myeloid and other human cell types.

Methods

Enzymatic activity and binding assays

JAKinib potency was determined using JAK1, JAK2, JAK3, and tyrosine kinase 2 (TYK2) enzymatic assays in a time-resolved fluorescence resonance energy transfer format in the presence of 1mM ATP (supplemental Methods). In vitro, recombinant cyclin-dependent kinase (CDK) and ACVR1 activities against a peptide substrate were assayed using the time-resolved fluorescence resonance energy transfer–based LANCE Ultra Kinase Assay (PerkinElmer; supplemental Methods). JAK2 binding affinities for JAKinibs were determined using surface plasmon resonance (SPR) (supplemental Methods).

Full kinome profiling

JAKinibs (100nM) were evaluated in 372 assays of 356 unique wild-type (WT) kinases in the presence of 10μM ATP using kinase profiling radiometric HotSpot kinase activity panel assays performed at Reaction Biology Corporation (Malvern, PA). IC₅₀ values for each JAKinib in the presence of 100μM ATP were determined for each kinase inhibited by ≥50%.

Quantification of phosphorylated STAT5 (pSTAT5) inhibition by individual JAKinibs using Baf3 JAK2 WT-EPOR or JAK2 V617F-EPOR cells

Baf3 JAK2 WT-erythropoietin receptor (EPOR) or JAK2 V617F-EPOR cells, cultured in starvation media (RPMI-1620 medium supplemented with 0.5% fetal bovine serum) were treated with serial dilutions (0.1nM to 10μM) of each JAKinib or dimethyl sulfoxide (DMSO) control (supplemental Methods).

Quantification of pSTAT5 in response to individual JAKinibs using SET2 (JAK2 V617F) cells

SET2 cells (DSMZ, Braunschweig, Germany), a megakaryoblastic cell line from a patient with human essential thrombocythemia carrying the JAK2 p.V617F mutation, were treated with serial dilutions of JAKinibs (0.1nM to 25μM; supplemental Methods).

Assessment of cytotoxicity of JAKinibs in various cell lines and in CD34⁺ cells in liquid culture

JAKinib cytotoxicity was assessed in human SET2 (DSMZ), F36-P (DSMZ), TF1 (ATCC), EOL1 (DSMZ), MV-4-11 (ATCC) and MOLM-16 (DSMZ) cell lines treated with serial dilutions of JAKinibs (0.5nM to 10μM; supplemental Table 1). Viable cells were detected using CellTiter-Glo reagent (Promega, Madison, WI) by luminescence measured using a PHERAstar Microplate Reader (BMG Labtech, Cary, NC). Cell viability was calculated and JAKinib cytotoxicity was reported as percentage inhibition relative to DMSO only (0% inhibition) and assay medium only (100% inhibition).

Cytotoxicity assessments of JAKinibs on normal hematopoietic cell differentiation, using CD34⁺ cells in liquid culture assays were performed using multilineage flow cytometry (supplemental Methods).

Ex vivo expansion and viability and lineage commitment analysis of CD34⁺ hematopoietic stem and progenitor cells

Human cord blood CD34⁺ hematopoietic stem and progenitor cells isolated from healthy control individuals were used to assess the impact of treatment with JAKinibs on CD34⁺ self-renewal and lineage commitment by flow cytometry using clinically relevant JAKinib concentrations corresponding to the respective steady-state peak plasma concentrations (C_max,ss; Table 1; Table 2; supplemental Methods).

Table 1.

JAKinib free drug plasma concentrations calculated from clinical trials

JAKinib	Clinical dose (mg)	Cavg.ss (nM)	Cmax.ss (nM)	2× Cmax.ss (nM)
Ruxolitinib21,22	25 mg bid	11	35	70
Fedratinib23	400 mg qd	171	275	550
Pacritinib24	200 mg bid	202	213	427
Momelotinib25	200 mg qd	30	104	208
M21	N/A	52	182	364

Values correspond to free concentrations in plasma, estimated from total concentrations in plasma, of each JAKinib at steady state and human protein binding (97% for ruxolitinib, 92% for fedratinib, 99% for pacritinib, and 91% for momelotinib) as reported from clinical trials for the recommended daily dose, per the US FDA-approved prescription labels. Where multiple dosing regimens are available, the maximum daily dose is included here. M21 concentrations were calculated using 1.75 ratio over momelotinib based on the prescription label (“The mean M21 to momelotinib ratio for AUC ranged from 1.4 to 2.1”).

AUC, area under the concentration-time curve; bid, twice daily; C_avg,ss, average steady-state plasma drug concentration during multiple-dose administration; maximum steady-state plasma drug concentration during a dosage interval; FDA, Food and Drug Administration; N/A, not available; qd, once daily.

Table 2.

JAKinib C_max concentrations used in this study

JAKinib	0.25× Cmax free drug (nM)	0.5× Cmax free drug (nM)	Cmax free drug (nM)	2× Cmax free drug (nM)	4× Cmax free drug (nM)
Ruxolitinib	8.7	17.4	35	70	139
Fedratinib	69	138	275	550	1100
Pacritinib	53	107	213	427	853
Momelotinib	26	52	104	208	416
M21	46	91	182	364	728
Momelotinib + M21	72	143	286	572	1144

Human protein binding (96% for ruxolitinib, 88% for fedratinib, 100% for pacritinib, and 86% for momelotinib) data were generated at Incyte Corporation, using standard methodology. M21 concentrations were calculated using 1.75 ratio over momelotinib based on the prescription label (“The mean M21 to momelotinib ratio for AUC ranged from 1.4 to 2.1”).

C_max, maximum plasma concentration.

CFU assay

WT cord blood CD34⁺ cells (Stemcell Technologies) were seeded at 250 cells/mL in MethoCult H4035 (Stemcell Technologies) with ruxolitinib, pacritinib, fedratinib, momelotinib, M21, and a combination of momelotinib + M21 at concentrations corresponding to 0.5-4× the respective clinical C_max (Tables 1 and 2; supplemental Methods). Each condition was plated in triplicate. Colonies were counted after 14 days using STEMvision (Stemcell Technologies).

Intracellular signaling in hematopoietic stem and progenitor cells evaluated using cytometry by time of flight

Effects of JAKinibs on intracellular signaling pathways were assessed at concentrations corresponding to the respective clinical C_max values (Tables 1 and 2; supplemental Methods).

Results

Profiling of JAKinib activity against JAK family enzymes

The JAKinibs ruxolitinib, fedratinib, pacritinib, momelotinib, and M21 were tested side by side to determine their potency to inhibit JAK family enzymes using JAK1, JAK2, JAK3, and TYK2 enzymatic assays. Assays were performed at 1mM ATP to reflect physiological ATP levels.²⁶ JAK inhibitory potency varied across the JAKinibs, with ruxolitinib identified as the most potent JAK2 inhibitor (Table 3). Primary JAK target for the 4 parental drugs was JAK2, whereas M21 was most potent against TYK2. Ruxolitinib, momelotinib, and fedratinib also significantly inhibited JAK1.

Table 3.

Inhibition of JAK family kinases by JAKinibs

Activity assay	Enzymatic activity assay IC50 (nM) GeoMean (GeoSD), 1mM ATP
	Ruxolitinib	Fedratinib	Pacritinib	Momelotinib	M21
JAK1	4.3 (1.1)	99 (1.0)	921 (1.1)	130 (1.2)	271 (1.0)
JAK2	2.9 (1.1)	17 (1.1)	39 (1.1)	29 (1.1)	118 (1.0)
JAK3	540 (1.1)	4291 (1.3)	1551 (1.1)	1277 (1.2)	2941 (1.0)
TYK2	22 (1.2)	761 (1.1)	380 (1.2)	53 (1.2)	70 (1.2)

GeoMean, geometric mean; GeoSD, geometric standard deviation.

As expected from the enzymatic assay, ruxolitinib demonstrated the greatest binding affinity for JAK2 catalytic domain by SPR, with a dissociation constant of 0.125nM, followed by fedratinib, momelotinib, and pacritinib (3.6, 4.6, and 5.3 times less potent than ruxolitinib, respectively). Binding profiles of the 4 JAKinibs to the catalytic domain of JAK2 assessed by SPR are shown in supplemental Figure 1.

Full kinome profiling of JAKinibs

Comprehensive kinase profiling of JAKinibs was performed using 356 unique WT kinases to further assess selectivity for JAK2 inhibition. First, each JAKinib was tested at a concentration of 100nM with 10μM ATP. This low ATP concentration increases the sensitivity of the assays. Ruxolitinib demonstrated greatest selectivity, inhibiting only 12 kinases by ≥50%, whereas M21 was least selective, inhibiting 57 kinases. Fedratinib, pacritinib, and momelotinib inhibited 21, 40, and 44 kinases by ≥50%, respectively (supplemental Table 3).

Overall, a total of 83 kinases showed ≥50% inhibition by any JAKinib in the previous assay and were then tested separately for dose response against each JAKinib at 100μM ATP. Ruxolitinib again demonstrated greatest selectivity for JAK2 inhibition compared with the other JAKinibs. Dendrograms based on IC₅₀s for each JAKinib and for all kinases tested in this comprehensive analysis are shown in Figure 1A. Fedratinib, pacritinib, momelotinib, and M21 inhibited many kinases outside the JAK family, with a similar potency to their JAK2 inhibition (Figure 1B; Table 4). In particular, fedratinib, pacritinib, momelotinib, and M21, but not ruxolitinib, exhibited strong inhibition of FLT3 (with IC₅₀s ranging from 6.2nM to 82nM vs 2067nM for ruxolitinib) and KIT kinases (IC₅₀s ranging from 0.9nM to 386nM vs >10 000nM for ruxolitinib). Pacritinib and momelotinib inhibited ACVR1 with IC₅₀s of 175nM and 381nM, respectively (Table 4).

Figure 1.

JAKinib selectivity profiles in the presence of 100μM ATP showing IC₅₀s of kinases previously demonstrated to be inhibited by ≥50% by ≥1 JAKinib in initial profiling (at 100nM in the presence of 10μM ATP). (A) Dendrograms show IC₅₀s for JAKinibs against all kinases previously demonstrated to be inhibited by ≥50% inhibition by ≥1 JAKinib in initial profiling; (B) JAKinib inhibition of kinases implicated in cellular function. Only IC₅₀ values within 100-fold of JAK2 inhibition are shown; labels mark JAK family kinases, other kinases with strong biology in hematopoiesis, and, on the right side of every panel, kinases essential for cell proliferation according to DepMap.²⁷

Table 4.

IC₅₀ (nM) at 100μM ATP within 40-fold of the IC₅₀ for JAK2 for each JAKinib

Ruxolitinib		Fedratinib		Pacritinib		Momelotinib		M21
Kinase	IC50 (nM)	Kinase	IC50 (nM)	Kinase	IC50 (nM)	Kinase	IC50 (nM)	Kinase	IC50 (nM)
JAK1†	1.1	JAK2∗	11	KIT	0.9	MYLK4	7.2	CDK2/cyclin A	1.4
JAK2∗	1.5	FLT3†	11	CSF1R	3.9	JAK2∗	11	CDK2/cyclin O	2.0
TYK2	3.8	SIK2	21	TRKC	4.8	TYK2	11	CDK2/cyclin A1	3.7
TRKC	36	FAK	21	ROS1	5.7	CDK4/cyclin D1	11	ROCK2	11
JAK3	58	DDR1	23	FLT3†	6.2	CDK2/cyclin A	11	MYLK4	14
FLT3	2067	JAK1	37	IRAK1†	6.2	ROCK2	21	ROCK1	20
KIT	>10 000	AURKC	48	JAK2∗	15	SIK2	24	CDK4/cyclin D1	21
		TRKA	58	TNIK	17	NEK5	24	TYK2	30
		ARK5	66	HIPK4	33	ERK7	27	CDK15/cyclin A2	31
		TYK2	71	CDK4/cyclin D1	44	TBK1	30	CDK4/cyclin D2	31
		ACK1	71	CDK9/cyclin T1	48	IKKE	31	CDK4/cyclin D3	36
		BMX	88	ARK5	53	CDK6/cyclin D1	35	CDK1/cyclin B	38
		ROS1	89	MAK	54	JAK1†	36	CDK6/cyclin D1	39
		TRKC	90	TRKA	58	FLT3	45	ERK7	44
		SIK1	105	DDR1	71	LRRK2	54	CDK4/cyclin D2	45
		ABL2	122	BRK	71	LIMK1	63	NEK5	51
		SRC	128	TYK2	71	NEK3	73	JAK2∗	60
		FYN	136	YES1	73	ARK5	90	TBK1	73
		MUSK	140	PKCA	75	NEK1	110	IKKE	80
		YES1	151	LOK	78	MARK4	117	JAK1†	81
		ZIPK	196	CLK1	86	JAK3	120	FLT3	82
		KIT	201	BMX	88	CDK1/cyclin B	121	CDK3/cyclin E	83
		AXL	208	AXL	98	IKKB	124	CDK9/cyclin K	85
		MYLK4	212	ACK1	103	SIK3	125	MAK	85
				NEK1	110	PRKD1	157	PRKD1	116
				ALK1	117	DRAK1	180	CDK3/cyclin E2	124
				ABL2	122	IRAK1	201	CDK2/cyclin 2	126
				SRC	128	SIK1	214	LIMK1	133
				LCK	134	MAK	226	NEK1	160
				FER	147	MARK2	235	CDK9/cyclin T2	164
				LIMK1	159	CDK3/cyclin E	239	CDK16/cyclin Y	183
				MAP4K2	165	KIT	241	PKCB2	193
				JAK3	175	PRKD2	255	CDK9/cyclin T1	197
				ACVR1†	175	MARK1	256	NEK3	198
				CLK2	182	PRKD3	287	CDK17/cyclin Y	205
				SIK2	209	ROCK1	312	PKCD	208
				CDK16/cyclin Y	212	ACK1	320	SIK2	228
				CDK19/cyclin C	223	MARK3	328	LRRK2	267
				CDK6/cyclin D1	286	CDK9/cyclin T1	342	PRKD2	306
				LRRK2	288	AURKC	379	CDK1/cyclin E	316
				JAK1	298	ACVR1†	381	JAK3	329
				FYN	314	CDK16/cyclin Y	398	IKKB	353
				IKKE	359			CDK1/cyclin A	373
				MINK	380			KIT	386
				TRKB	401
				MYLK4	457
				TBK1	574

For M21, the IC₅₀ values shown are within 40 fold of the JAK2 IC₅₀ for momelotinib; for ruxolitinib, FLT3 and KIT data are shown for reference.

Highlighted in bold are kinase targets currently assigned to every JAKinib in the medical literature.

^∗JAK2 is the key target to treat myelofibrosis.

^†Other kinases typically associated in the literature with the pharmacology of each drug.

Given the importance of ACVR1 in pathobiology of anemia in MF,^28,29 enzymatic activity of JAKinibs against ACVR1 was tested at 3 different ATP concentrations to allow contextualization of previous data. Results are shown in Table 5 and plotted in supplemental Figure 2, illustrating the effect of ATP concentration on IC_50. The results are consistent with previous reports of a Michaelis constant for ATP in the low micromolar range for ACVR1.³⁰ IC₅₀s reported by Reaction Biology Corporation at 100μM ATP were validated at 1mM ATP for some additional kinases with distinct importance in cell viability (eg, CDKs, FAK) and hematopoietic stem cell biology (KIT, FLT3; supplemental Table 4).

Table 5.

Comparison of IC₅₀s for ACVR1/ALK2 by JAKinibs

ATP (μM)	Expected IC50/Ki ratio∗	IC50 (nM)					Data generation	Reference
		Ruxolitinib	Fedratinib	Pacritinib	Momelotinib	M21
3	1.2×	>10 000	111	6	21	47	Incyte	This study
NR†	NR	>1 000	273	17	53	ND	Reaction Biology Corporation	29†
NR‡	NR	6 100			8.4		Thermo Fisher	28
100	7.3×	>10 000	1575	88	269	703	Incyte	This study
100	7.3×	>10 000	3351	175	381	ND	Reaction Biology Corporation	This study
1000	63×	>10 000	8934	644	2184	5163	Incyte	This study

IC₅₀ (nM) is the GeoMean of ≥3 replicates of data generated at Incyte Corporation, and the average of 2 repeats for data generated at Reaction Biology Corporation.

ACVR1, activin A receptor type 1; ALK, activin receptor-like kinase-2; GeoMean, geometric mean; Ki, inhibitory constant; Km, Michaelis constant; ND, not determined, NR, Not reported.

^∗Calculated using Cheng-Prusoff equation³¹ assuming ACVR1 Km (ATP) of 16μM.³⁰

^†ATP concentration not reported by Oh et al.²⁹ Assays were performed by Reaction Biology Corporation, in an assay with 2 potential ATP concentrations (10μM and 100μM). Given the values reported here at 100μM ATP, we infer that the ATP concentration used to calculate the IC₅₀s in the assays in Oh et al was 10μM.

^‡ATP concentration not reported by Asshoff et al.²⁸ Assays were performed by SelectScreen Biochemical Kinase Profiling Service (ThermoFisher Scientific. Waltham, MA). ATP concentrations used in the SelectScreen activity assays were within twofold of the experimentally determined apparent Michaelis constant value for each kinase; therefore, ATP concentration can be inferred to be ∼16μM.

Two independently produced samples for each JAKinib (synthesized by Incyte Chemistry, and by MedChem Express) were analyzed to further validate our study findings. Each pair of samples showed similar results in both single-point full kinome inhibition assays (supplemental Table 3), and IC₅₀ determination assays (supplemental Figure 3).

Effect of JAKinibs on pSTAT5 signaling

On-target effects of JAKinibs on pSTAT5 were evaluated in cell-based assays using JAK-2 dependent (ie, JAK2 activity required for survival) Baf3 JAK2 WT and Baf3 JAK2 V617F cell lines expressing EPOR. STAT5 is located downstream of JAK2 and requires erythropoetin (EPO) to activate the dimeric EPOR in Baf3 JAK2 WT cells to undergo phosphorylation but is constitutively phosphorylated in oncogenic JAK2 V617F cells. No selective inhibition of JAK2/STAT5 signaling was observed between the 2 cell lines for any of the JAKinibs tested (Figure 2A-B). However, ruxolitinib demonstrated ∼19-, 23-, and 14-fold higher potencies than fedratinib, pacritinib, and momelotinib respectively, and a >159-fold higher potency than M21 for inhibiting pSTAT5, as expected from inhibition of JAK2 reported in Table 3.

Figure 2.

Percent inhibition of pSTAT5 by JAKinibs in Baf3 and SET2 cells. Percent inhibition of pSTAT5 in Baf3 JAK2 WT (A) or Baf3 JAK2 V617F-harboring cells (B) and in SET2 cancer cells (JAK2 V617F) (C) following treatment with JAKinibs at indicated concentrations for 2 hours. Relative inhibition of pSTAT5, normalized to DMSO (no inhibition), and 10μM ruxolitinib (100% inhibition), is shown. Geometric mean (SD) IC₅₀s are shown. GeoMean, geometric mean; GeoSD, geometric standard deviation; pSTAT5, phosphorylated signal transducer and activator of transcription 5.

Effects of JAKinibs on pSTAT5 signaling were also investigated using the JAK2 V617F-dependent SET2 cancer cell line. Once again, ruxolitinib exhibited the most potent IC₅₀ response (14nM), followed by momelotinib (201nM; 14.5-fold less potent than ruxolitinib), pacritinib (421nM; 30.3-fold less potent than ruxolitinib), fedratinib (670nM; 48.2-fold less potent than ruxolitinib), and M21 (2042nM; 146.9-fold less potent than ruxolitinib; Figure 2C). Ruxolitinib also demonstrated greatest potency in suppressing the phosphorylation of STAT5 in a variety of on-target activity assays using HEK-Blue reporter lines, and weakest off-target activity in assays assessing JAKinib inhibition of phosphorylated c-KIT and pSTAT5 using c-KIT– and FLT3-containing Baf3 cells (supplemental Figures 4 and 5).

Effects of JAKinibs on cell viability

On- and off-target effects of JAKinibs on cellular viability were evaluated by testing antiproliferative effects of the drugs, on 3 JAK2-dependent acute myeloid leukemia (AML) cell lines (SET2, F36P, and TF1) and 3 JAK2-independent (ie, JAK2 activity not required for survival) AML cell lines (MV4-11, EOL1, and MOLM16), as defined by Dependency MAP (DepMap).²⁷ Dose-response curves are shown in Figure 3. Ruxolitinib exhibited the most potent inhibition of JAK2-dependent cell lines (IC₅₀ range, 25nM-94nM), with the largest separation (>100-fold) with the drug effect on non-JAK2–dependent cell lines (3.2μM-9.7μM). The other JAKinibs showed no separation in antiproliferative effect across both JAK2-dependent and JAK2-independent AML cell lines.

Figure 3.

Cytotoxicity assessment on JAK2-dependent and non-JAK2–dependent AML cell lines. Quantification of the effects of JAKinibs on the viability of JAK2 dependent cells (on-target: SET2, F36P, TF1) and JAK2 independent cells (off-target: EOL1, MV4-11, MOLM16) in assays using a CellTiter-Glo assay following treatment for 72 hours with JAKinibs at indicated concentrations. C_max,ss and C_avg,ss per drug are indicated with vertical lines. Values are provided in supplemental Table 2. AML, acute myeloid leukemia; C_avg,ss, average steady-state plasma drug concentration during multiple-dose administration; C_max.ss, maximum steady-state plasma drug concentration during a dosage interval; GeoMean, geometric mean.

Additional non-JAK2–dependent cell lines including HEK293 to evaluate general human cellular toxicity, HepG2 to screen for hepatotoxicity, and human lung fibroblasts and human umbilical vein endothelial cells (HUVECs) were tested to further assess cellular toxicity in human primary cells (supplementary Figure 6). Ruxolitinib did not demonstrate any impact on cell viability against nonmyeloid cell lines (HEK293, HepG2, fibroblasts, and HUVECs). In contrast, other JAKinibs inhibited cell viability in these cell lines although with weaker potency than was observed above for effects on the AML cell lines (Figure 3).

Assessment of cytotoxic effect of JAKinibs on WT hematopoietic cell differentiation

JAKinib impact on self-renewal, and lineage commitment, of healthy CD34⁺ cells was evaluated using defined medium supplemented with hematopoietic cytokines including TPO and EPO. Under these conditions, CD34⁺ cells either self-renew or differentiate into mature lineages (Figure 4). Multi lineage flow cytometry was used to quantify total number of live cells, myeloid cells (CD11b⁺), neutrophils (CD13⁺/CD16⁺), mature megakaryocytes (CD41⁺, CD42b), and erythroid cells (CD71⁺) following treatment with ruxolitinib, pacritinib, fedratinib, momelotinib, M21, and a combination of momelotinib + M21 at concentrations associated with 0.5-2× the clinical C_max (Tables 1 and 2). Clinically relevant concentrations of each JAKinib were selected to reflect steady-state plasma exposures (C_max) achieved in patients at approved therapeutic doses, thereby enabling translational assessment of their effects under physiologically meaningful conditions. Treatment with JAKinibs caused a reduction in both overall viability of CD34⁺ cells, and numbers of all 4 differentiated cell types at concentrations equivalent to clinical C_max (Figure 4). Overall, impact on viability and lineage commitment was more pronounced with pacritinib and fedratinib vs ruxolitinib and momelotinib, with more modest effects seen with cells exposed to ruxolitinib, momelotinib, and M21 (Figure 4). Despite comparable reductions in viability following ruxolitinib or momelotinib exposure at these clinical doses, ruxolitinib selectively and strongly impaired both neutrophil and erythroid lineage differentiation of CD34⁺ cells, while momelotinib displayed more limited inhibition. This distinction was most apparent for inhibition of erythroid lineage differentiation of CD34⁺ cells consistent with the increased potency of ruxolitinib for JAK2 inhibition (shown in Figure 2).

Figure 4.

Impact of JAKinibs on WT CD34⁺ cell viability and ex vivo differentiation. Relative percentage of viable WT CD34⁺ cells, and percentage of differentiated cell types, myeloid (CD11b⁺) cells, neutrophils (CD13/CD16⁺), mature MgK (CD41⁺), and erythroid (CD71+) cells, following treatment of WT CD34⁺ cells with JAKinibs at concentrations associated with 0.5-2× the clinical C_max, shown in Table 2, for 6 days. Relative percentage of cells normalized to DMSO is shown. C_max, maximum plasma concentration; MgK, megakaryocytes.

Effect of JAKinibs on CD34⁺ cell colony-forming potential

A CFU assay was performed to evaluate impact of clinically relevant doses of JAKinibs on differentiation of WT CD34⁺ cells (Figure 5). At concentrations up to 4 times above clinically relevant levels, ruxolitinib had minimal effects on colony-forming potential of normal CD34⁺ cells. In contrast, pacritinib and fedratinib significantly reduced colony formation by ∼60% to 70% at concentrations corresponding to the clinical C_max (Tables 1 and 2). Momelotinib and M21 had only mild effects on colony counts when used separately, while the combination of the 2 molecules showed profound inhibition (∼90%) at concentrations associated with C_max (Figure 5). The CFU media used (H4035) in these assays only supports generation of granulocyte-macrophage progenitor–derived colonies (ie, CFU-GM, CFU-G, and CFU-M). Assessment of the distribution of colony subtypes following treatment with different JAKinibs showed no significant or consistent differences in subtype proportions across biological replicates for any of the JAKinibs tested (data not shown), suggesting that the reduction in colony formation observed reflects general suppression of myeloid progenitor proliferation or survival, rather than lineage-specific inhibition.

Figure 5.

In vitro impact of JAKinibs on WT CD34⁺ cell colony formation. Relative percentage of CD34⁺ colonies formed after incubation with JAKinibs at clinically relevant concentrations, shown in Table 2, for 14 days, normalized to DMSO. Concentrations from left to right correspond to 0nM JAKinib (DMSO control), 0.25-4× the clinical C_max free JAKinib concentration. Data are shown as the average of 3 healthy donor samples. Statistical analysis is by 1-way ANOVA; ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001. C_max, maximum plasma drug concentration.

Effects of JAKinibs on intracellular signaling

To evaluate the effects of maximum clinically relevant doses of JAKinibs on intracellular signaling pathways, phospho-profiling was performed using cytometry by time of flight in CD34⁺ cells from 2 independent cord blood donors treated with the JAKinibs at concentrations corresponding to clinical C_max. Cells were stimulated with IL-3, thrombopoietin (TPO), and interferon alfa, resulting in activation of canonical JAK/STAT signaling nodes, including pJAK2, pSTAT3 (Y705), and pSTAT5 (Y694) (Figure 6).

Figure 6.

Impact of JAKinibs on intracellular signaling using CyTOF. Percentage change in mean fluorescence intensity of proteins involved in JAK/STAT and JAK/STAT independent signaling pathways. Phosphorylation was measured after treatment of WT CD34⁺ cells stimulated with IL-3, TPO, and IFN-α with JAKinibs for 2 hours at the clinical C_max, shown in Table 2. Relative mean fluorescence intensity, normalized to DMSO, is shown. Assays used 2 independent CD34⁺-derived cell lines, which were generated from 2 different healthy cord blood donors. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. C_max, maximum plasma drug concentration; C_max,ss, maximum steady-state plasma drug concentration during a dosage interval; CyTOF, cytometry by time of flight; fed, fedratinib; IFN-α, interferon alfa; MMB, momelotinib; pac, pacritinib; rux, ruxolitinib; STAT, signal transducer and activator of transcription; stim, stimulation; TPO, thrombopoietin.

At concentrations corresponding to clinical Cmax, all 4 JAK inhibitors reduced pJAK2 phosphorylation, although inhibition across downstream and off-target pathways varied by compound (Figure 6). Ruxolitinib and fedratinib demonstrated the most potent and consistent suppression of JAK/STAT signaling, in line with their biochemical as well as cellular potencies against JAK2 at clinically relevant C_max concentrations (Table 3; Figure 2). In contrast, pacritinib and momelotinib showed only modest effects at these maximum clinical doses.

Beyond the JAK/STAT axis, all 4 JAK inhibitors showed pS6 (S235/S236) suppression, and limited inhibition of pCREB (S133), pTBK, and pERK (T202/Y204) likely reflecting incomplete activation of these pathways under the stimulation conditions (Figure 6). Although these data suggest that the JAK2/STAT5 inhibition is limited at clinically relevant C_max concentrations, a previous study by Kong et al²⁰ suggests that at supraphysiologic concentrations, all JAK inhibitors effectively suppress JAK2/STAT5 signaling in hematopoietic progenitors.

Comparison of ACVR1 cellular potency of momelotinib and pacritinib with ACVR1 inhibitor zilurgisertib

Cellular assays were performed to further investigate momelotinib and pacritinib inhibition of ACVR1 kinase activity at cellular levels of ATP using 2 different cell lines. We assessed reduction of phosphorylation of direct target SMAD1 in HeLA cells, and levels of the SMAD target gene hepcidin (required for iron homeostasis) in HUH-7 cells. The IC₅₀ values associated with momelotinib and pacritinib were compared with those of potent and selective ACVR1 inhibitor zilurgisertib.³² In both assays, momelotinib and pacritinib showed weak ACVR1 inhibition in the micromolar range, with IC₅₀ values at least 50- and 10-fold higher, respectively, than observed with zilurgisertib. For momelotinib vs pacritinib vs zilurgisertib, respectively, pSMAD 1/5 IC₅₀ were 5173nM vs 717nM vs 61nM, and hepcidin IC₅₀ were 1081nM vs 294nM vs 29nM. Further assays measuring inhibition of BMP6, and BMP6 + IL6–stimulated hepcidin in HUH-7 cells to investigate inhibition of hepcidin via BMP-SMAD and IL6-JAK-STAT pathways again showed similar trends (Table 6).

Table 6.

Comparison of ACVR1 cellular potency of momelotinib and pacritinib with ACVR1 inhibitor zilurgisertib in 4 separate assays in 2 cell lines (HeLa and HUH-7)

ACVR1 inhibitor	pSMAD1/5 HTRF Assay (HeLa Cells)			Hepcidin assay (HUH-7 Cells - BMP6/IL6 Stimulated)			Hepcidin assay (HUH-7 - BMP6 Stimulated)			Hepcidin assay (HUH-7 Cells - BMP76/IL6 Stimulated)
	IC50 (nM) Geomean	GeoSD	N	IC50 (nM) Geomean	GeoSD	N	IC50 (nM) Geomean	GeoSD	N	IC50 (nM) Geomean	GeoSD	n
Zilurgisertib	61	2.3	8	20	1.2	6	17	1.4	10	99	1.9	10
Momelotinib	5173	2.9	8	1081	1.1	6	863∗	1.8	10	2113	2.0	10
Pacritinib	717	3.0	8	294	1.5	6	305	1.5	10	546	1.6	10

See the supplemental Methods for details.

GeoMean, geometric mean; GeoSD, geometric standard deviation; HTRF, homogeneous time-resolved fluorescence.

^∗In a similar assay measuring hepcidin RNA levels in HepG2 cells Asshoff et al²⁸ reported an IC₅₀ of 650nM ± 200nM.

Discussion

With the identification of the JAK2 p.V617F mutation, a number of type 1 JAKinibs can bind and inhibit JAK2 have been developed in an attempt to alleviate MF symptoms and manage the disease. With Food and Drug Administration approval of 4 of these JAKinibs for treatment of MF, an understanding of their pharmacological properties and effects outside of JAK2 inhibition is essential to understand their therapeutic potential and toxicity risks, and rationalize combination strategies. To date, JAKinib kinase inhibitory profile studies have been mostly determined under differing assay conditions and with limited exploration of their inhibitory profile in the full kinome.^10,16, 17, 18^,20 As shown in supplemental Table 5, published JAK2 IC₅₀ values for each JAKinib were determined using different assays and ATP concentrations, leading to different values and rank order of potency to those presented here. An important limitation in many of these published reports is the assumption that biochemical IC₅₀s, regardless of ATP concentration used in the test, can be extrapolated to clinical pharmacology.¹⁵ Despite these limitations in preclinical knowledge of JAKinibs, the clinical literature has assigned a handful of primary kinase targets to each of these drugs.^8,9 In this study, we generated head-to-head kinome profiling data for the 4 approved JAKinibs (and the major metabolite of momelotinib, M21) and compared their effects on signaling and proliferation in myeloid and other human cell types.

Enzymatic activities were consistent with cellular inhibition of STAT5 signaling in response to different cytokines, with ruxolitinib demonstrating the most potent inhibition among the JAKinibs, followed by fedratinib, pacritinib, and momelotinib. Ruxolitinib also demonstrated the most selectivity, inhibiting the fewest kinases as shown by comprehensive kinase profiling. Notably, fedratinib, pacritinib, and momelotinib, but not ruxolitinib, exhibited high inhibitory potencies against FLT3 and/or KIT kinases in both biochemical and cellular assays. Although these kinases are therapeutic targets in some malignancies, both are critical for self-renewal and differentiation of WT HSCs,^33,34 and their inhibition can lead to hematologic toxicities. Notably, fedratinib (FAK), pacritinib (CDK4), and momelotinib/M21 (CDK2, CDK4, CDK6, and CDK1) inhibited kinases essential for cell survival,²⁷ with potencies similar to JAK2.

The lack of significant ACVR1 inhibition by momelotinib and pacritinib at physiological ATP concentrations (1mM) with IC₅₀s of 2.2μM and 0.6μM, respectively, contrasts sharply with earlier reports citing potent inhibition (Table 5).^28,29 In those studies, ACVR1 potencies were comparable to JAK2 inhibition,^18,19 and formed the basis for proposing ACVR1 inhibition as a mechanistic contributor to anemia benefits in MF. However, those studies were performed at subtherapeutic ATP levels, which undoubtedly overestimates kinase inhibition achievable in cells where ATP is in the millimolar range and will outcompete these drugs.

Our newly determined momelotinib and pacritinib IC₅₀s exceed free plasma drug concentrations expected at therapeutic doses by several fold (Table 1), and are 17- and 75-fold higher than their respective JAK2 IC₅₀s at 1mM ATP (Table 3). These findings were consistent across assays using 2 independently synthesized compound lots. ATP titration experiments further confirmed the inverse relationship between ATP concentration and ACVR1 inhibition potency, and that our IC₅₀ values at 1mM ATP align with previously reported values when ATP concentration differences were accounted for (supplemental Table 4). Moreover, ACVR1 IC₅₀s for pacritinib and momelotinib remained significantly higher than their JAK2 IC₅₀s at fixed ATP concentration (100μM, Tables 4 and 5; 1mM, Tables 3 and 5). Inhibitory effects at physiological ATP were further confirmed in a range of assays across 2 different cell lines, and compared with those of potent and selective ACVR1 inhibitor, zilurgisertib.³² Consistent with findings of the biochemical assays, ACVR1 inhibition IC₅₀ by both drugs was only observed at high nanomolar to low micromolar range in cellular assays (Table 6) and remained weaker than for the selective ACVR1 inhibitor, zilurgisertib.

Collectively, these data, generated under physiological ATP conditions, indicate that pacritinib and momelotinib exhibit minimal ACVR1 inhibition at maximum clinical doses; thus, ACVR1 blockade is unlikely to be the primary driver of their benefits in MF-associated anemia. Furthermore, emerging clinical evidence suggests the mechanism of anemia improvement in MF is multifactorial. In a recent phase 1/2 study in patients with MF-associated anemia, treatment with the potent and selective ACVR1 inhibitor, zilurgisertib, did not suppress hepcidin levels in all patients treated and resulted in only modest anemia improvement, whether used alone or with ruxolitinib.³⁵ This underscores the complexity of anemia pathophysiology in MF and suggests that additional mechanisms may underlie the clinical benefits observed with pacritinib and momelotinib in patients with MF.^8,9,13

In cell-based assays using both AML and nonhematopoietic cell lines, ruxolitinib demonstrated clear selectivity, potently inhibiting signaling and growth of JAK2-dependent cells while sparing cells not dependent on JAK2 for growth (MV4-11, EOL1, MOLM16, HEK293, HepG2, human lung fibroblasts, and HUVECs). By contrast, pacritinib, fedratinib, and momelotinib inhibited growth of JAK2-dependent and -independent cell lines with similar potencies. Studies based on physiologically relevant models showed ruxolitinib also had a milder effect on WT CD34⁺ colony formation and cell viability than fedratinib and pacritinib at the maximum recommended clinically dose including minimal impact on differentiation into mature megakaryocytes, which was markedly inhibited by both fedratinib and pacritinib. Ruxolitinib did however demonstrate the strongest inhibition of erythroid cells, which rely heavily on JAK2/STAT5 signaling.³⁶ Similarly, analysis in CD34⁺ cells of effects on signaling pathways showed inhibition of JAK/STAT pathways by all 4 approved JAKinibs at maximum recommended clinical doses, with ruxolitinib and fedratinib showing the most potent inhibition of downstream markers pSTAT3 and pSTAT5, reflecting their demonstrated potency for JAK2 inhibition.

Overall our findings closely align with those recently reported by Kong et al,²⁰ although their studies employed flat doses of JAKinibs (1μM) which show variable correlations with the maximum recommended doses concentrations. They showed effective suppression of JAK2/STAT5 signaling by all 4 approved JAKinibs, and also concluded ruxolitinib and pacritinib were most potent inhibitors in ex vivo assays using primary patient cells, including CFU assays and patient-derived xenografts.

Small molecules can interfere with the physiological function of many proteins beyond the intended target, potentially causing off-target effects at therapeutic doses. We therefore focused this study on kinase inhibition to assess selectivity, because given the high degree of structural homology within the kinase family, particularly at the highly conserved ATP binding pocket, off-target kinase inhibition is likely in kinase inhibitors such as the JAKinibs tested. We also investigated off-target effects by testing effects of the JAKinibs on the health of cells that do not require JAK2 activity for survival. Based on kinome profiles, the observed effects of fedratinib, pacritinib, and momelotinib on cell health on these non-JAK2–dependent cells could be attributed primarily to their effect on other kinases essential for cell survival; however, effects on other nonkinase proteins could also contribute to their promiscuous cellular pharmacology. For example, fedratinib has also been described as a bromodomain-containing protein 4 (BRD4) inhibitor.^37,38 Importantly, these in vitro effects do not directly predict clinical toxicity as specific organ distribution and other factors in vivo will influence clinical pharmacology in ways that warrant investigation beyond the scope of this study.

In summary, the 4 JAKinibs clinically approved for treatment of MF exhibit distinct kinase inhibition profiles and cellular activities that are comprehensively characterized here, to our knowledge, for the first time, providing a basis for direct comparison across agents. As determined in these analyses, ruxolitinib was the most potent and selective JAK2 inhibitor with the most favorable preclinical risk profile within this class of drugs for treatment of MF.

Conflict-of-interest disclosure: The authors report employment by and stock ownership in Incyte Corporation.