Background
The Janus kinase (JAK) family of cytoplasmic protein tyrosine kinases comprises JAK1, JAK2, JAK3, and tyrosine kinase 2 (TYK2). Janus kinases bind to type l and type ll cytokine receptors and transmit extracellular cytokine signals to activate signal transducers and activators of transcription (STATs), which translocate to the nucleus and modulate transcription of effector genes [
1]. Recent advances in the treatment of rheumatoid arthritis (RA) have been made with the use of small molecules that inhibit JAKs, specifically targeting cytokine signaling pathways implicated in RA pathogenesis [
2‐
5].
Baricitinib and tofacitinib are JAK inhibitors (JAKis) that have been approved for the treatment of RA, and other JAKis, including upadacitinib, are in clinical development [
2]. Baricitinib is approved for the treatment of moderately to severely active RA in adults in over 60 countries including European countries, Japan, and the USA. In vitro kinase assays demonstrate that baricitinib is a selective JAK1 and JAK2 inhibitor with moderate activity against TYK2 and significantly less against JAK3 [
6]. Tofacitinib is a potent JAK1 and JAK3 inhibitor but is less active against JAK2 and TYK2 [
7]. Upadacitinib is reported as a selective JAK1 inhibitor [
8,
9]. Distinct cytokine signaling pathways are mediated by varying JAK complexes, indicating that various JAKis may have differing effects on host inflammatory responses, including those that drive RA.
Herein, we sought to study JAKis that have shown clinical efficacy in the treatment of RA and other autoimmune diseases. The objective of the study was to compare the in vitro cellular pharmacology of baricitinib, upadacitinib, and tofacitinib across relevant leukocyte subpopulations, coupled with their in vivo pharmacokinetics (PK), to determine their effects on distinct cytokine pathways, many implicated in RA pathogenesis.
Methods
Leukocyte preparation and experimental design
Whole blood samples from healthy donors (N = 4–12) were apheresed, and leukocyte-enriched fractions containing approximately 400 million leukocytes were transferred to the company Primity Bio (Fremont, CA, USA). Immediately following apheresis, approximately 600,000 cells were plated in 100 μL into 96-well plates and incubated with JAKis using a 7-point dose range from 2 to 10,000 nM (four-fold dilutions from the highest concentration) for 1 h prior to stimulation with cytokines for 15 min at 37 °C. Laboratory procedures ensured that compound incubation and cytokine stimulation times were kept constant across cytokines and donors, including when multiple donor samples were being processed. Baricitinib (Eli Lilly and Company), upadacitinib (synthesized by Eli Lilly), and tofacitinib citrate (ApexBio) were prepared as 10-mM stocks in dimethyl sulfoxide. Eight cytokines were used at a concentration of 30 ng/mL (granulocyte colony-stimulating factor [G-CSF], interferon [IFN]-γ, interleukin [IL]-2, IL-4, IL-6, IL-10, IL-15, and IL-21); three others were used at different concentrations: granulocyte-macrophage colony-stimulating factor (GM-CSF) (15 pg/mL), IFN-α (5 ng/mL), and IL-3 (2 ng/mL). For the first eight cytokines, data were generated in two batches; one batch compared baricitinib to tofacitinib (6 donors) and a second batch compared baricitinib to upadacitinib (6 donors). For the remaining cytokines, data were generated in three batches; one batch compared baricitinib to upadacitinib (6 donors) and two other batches compared baricitinib to tofacitinib (2 donors/batch). The choice of cytokine concentrations and incubation conditions were optimized in order to ensure consistent signaling in alternate cell types and phosphorylated STAT (pSTAT) readouts.
Flow cytometry
After stimulation, cells were fixed, permeabilized, and fluorescence barcoded as previously described [
10]. Samples were combined and stained with fluorochrome-conjugated pSTAT1 (Y701), pSTAT3 (Y705), pSTAT5 (Y694), pSTAT6 (Y641), CD3, CD4, CD20, and CD56 antibodies. Multicolor flow cytometry was used to quantify STAT phosphorylation in gated leukocyte subpopulations, and the signals from each sample were de-barcoded for statistical analysis. For a given case (stimulation, cell type, and pSTAT combination), the half maximum inhibitory concentration (IC
50) was determined if there was a consistent response to the stimulus as described in the “
Statistical analysis” section. The primary pSTAT observed for each stimulus is reported. Leukocyte populations were defined as CD20+ (B cells), CD3+CD4+ (CD4+ T cells), CD3+CD4− (CD8+ T cells), CD3−CD56+ (natural killer [NK] cells), and by forward and side scatter (monocytes).
Statistical analysis
The IC50 values for JAKis were determined by analyzing the mean fluorescence intensity (MFI) of cytokine-stimulated samples in the presence of the designated concentration of compound. For a given case (stimulation, cell type, and pSTAT combination), the MFI for unstimulated and stimulated cells was determined for each donor. To ensure that a biologically relevant signal was induced, concentration-response curves (CRCs) were only analyzed when a consistent response to stimulus was observed as described below. Data were analyzed with a statistical model using an integrated data set that included a model term to account for any systematic batch effects.
Selection of cases for analysis, fitting, and selection of CRC curves
Two sets of criteria for reporting an IC50 value were used: one at the case level and another at the individual curve level. For the CRCs to be estimated for a given case, the case had to satisfy two criteria: (c1) the minimum MFI difference between stimulated and unstimulated reads across all donors was at least 10 fluorescence units and (c2) the p value of the one-sided t test of the null hypothesis stimulated < unstimulated was at most 0.10. Once a case met these two criteria, four-parameter logistic curves were fit to the 7-point curve concentration data, with the top fixed at the stimulated MFI fluorescence (ensuring all curves for the same donor had the same top parameter). Once fitted, the IC50 from a curve was accepted if the following conditions were met: (i1) the R2 of the fitted curve was above 0.65, (i2) the standard error (SE) of the log(IC50) corresponded to a fold change smaller than 8, (i3) the SE of the slope parameter was smaller than 8, and (i4) the difference in the MFI signal between the highest and lowest compound concentrations was larger than both half the distance between the fitted top and bottom parameters for that donor, compound curve, and the stimulated and unstimulated MFI signals for that case. If more than 25% of the curves for a compound were removed, the entire case was removed from further analysis.
Estimation and comparison of mean IC50 values
Statistical analysis and comparison of IC50 values were conducted by fitting statistical mixed effect models to the log(IC50), including a random effect for donor and fixed effects for compound for each stimulation, cell type, and pSTAT combination. In cases where data were collected in multiple batches, a batch-fixed effect was added to the model to account for any systematic differences between batches. Reported IC50 values corresponded to the least squares means of the compound effects from the model. Reported p values were adjusted for multiplicity by using a Bonferroni correction within each stimulant and pSTAT combination.
Pharmacokinetic profiles
The PK profiles of baricitinib were estimated from a two-compartment model with zero-order absorption that was developed using data from patients with RA from phase 2 and phase 3 clinical trials with once daily (QD) 2- or 4-mg doses [
11]. The PK profiles of upadacitinib were obtained from published data using RA scaled healthy volunteer profiles in a dose-proportional manner for QD 15- or 30-mg doses [
12,
13]. The PK profiles of tofacitinib were estimated from a one-compartment model with zero-order absorption using data from patients with RA from phase 3 clinical trials with twice daily 5- or 10-mg doses [
14].
Estimation and comparison of time above IC50 and daily percent inhibition
The individual 4PL CRCs were combined with population PK curves to calculate time above IC50 and average daily percent inhibition. Protein binding effects were accounted for by replacing the in vitro IC50 with an adjusted IC50 value computed by dividing the IC50 value for each donor by the proportion of compound unbound (baricitinib 50%, upadacitinib 54%, and tofacitinib 60%). The time above IC50 was defined as the time the (linearly interpolated) PK concentration was above the adjusted IC50. Protein-bound adjusted CRCs were constructed by replacing the in vitro IC50 value with the adjusted value. The average daily percent inhibition for a subject was obtained by entering the steady-state PK concentrations into the adjusted CRCs, computing the area under this curve, and dividing it by 24 h. Using the individual donor values for time above IC50 and average daily percent inhibition, the same mixed model used to fit the log(IC50) values was used to estimate and compare mean times above IC50 and percent inhibitions. No transformations were undertaken to keep the estimates within the 0–24-h range (for time above IC50) and 0–100% (for average daily inhibition). The values were truncated in a few cases where the estimates fell outside the range.
Discussion
The introduction of multiple JAKis into clinical practice, each with distinct selectivity across the JAK family members as determined by in vitro kinase assays, poses obvious questions as to the relative functional impact of such differences. Herein, we characterized the in vitro cellular pharmacology of baricitinib, upadacitinib, and tofacitinib, coupled to their in vivo PK, to determine their effects (at human oral doses that are approved or included in late phase clinical studies) on distinct cytokine pathways involved in the pathogenesis of RA.
Tofacitinib and upadacitinib, for example, were the most potent inhibitors of the JAK1/3-dependent cytokines tested. Moreover, lower IC
50 values and increased time above IC
50 for tofacitinib and upadacitinib, compared for example with baricitinib, translated to a greater overall inhibition of STAT signaling during the 24-h dosing interval for JAK1/3-dependent cytokines. Inhibition of this particular pathway may have potential impact on lymphocyte activation. Notably, IL-15 and IL-21 induce the maturation and function of NK cells and this may be of relevance to the changes in peripheral blood NK cell numbers reported in JAKi trials in RA [
9,
15‐
19]. Relationships between the durability of STAT inhibition by any given JAKi and overall changes in lymphocyte subpopulations observed in RA clinical trials with these molecules remains to be determined.
Upadacitinib, baricitinib, and tofacitinib inhibited the JAK2/2 or JAK2/TYK2 signaling cytokines IL-3, GM-CSF, and G-CSF, albeit to varying degrees. Upadacitinib proved the most potent inhibitor of IL-3 and GM-CSF (JAK2/2), followed by baricitinib and tofacitinib, while tofacitinib proved the most potent inhibitor of G-CSF (JAK2/TYK2), followed by upadacitinib and baricitinib. Both G-CSF and GM-CSF may contribute to RA pathogenesis through the activation, differentiation, and survival of myeloid cells [
20]. Indeed, blockade of GM-CSF receptor with the humanized monoclonal antibody mavrilimumab was effective in a phase 2 clinical trial in RA [
21]. Baricitinib and other JAKis with activity against JAK2 may attenuate RA pathogenesis in part through inhibiting GM-CSF-mediated cellular responses.
With respect to JAK1/2-dependent cytokine signaling, baricitinib, upadacitinib, and tofacitinib were all inhibitors of IL-6 and IFN-γ (JAK1/2), and IL-10 and IFN-α (JAK1/TYK2) signaling, although differences in potency emerged. Tofacitinib proved the most potent inhibitor of IL-6, IFN-γ, and IL-10 signaling, followed by upadacitinib and baricitinib, while upadacitinib and tofacitinib were the most potent inhibitors of IFN-α signaling. Interleukin-6 is a pleiotropic pro-inflammatory cytokine that contributes to synovial inflammation, articular joint destruction, and some of the systemic features observed in RA [
22]. Baricitinib and other JAKis may be effective in the treatment of RA in part by IL-6 inhibition, which has been validated as a therapeutic target in RA patients by the monoclonal antibody tocilizumab [
23]. Interleukin-10, IFN-α, and IFN-γ are also likely contributors to RA pathogenesis, and inhibition of these responses may contribute to the mechanism of JAKis in RA treatment; lesser inhibition of IL-10 may be desirable as the net effects of this cytokine have been described as anti-inflammatory in RA [
4,
24].
Upadacitinib has been reported to be a selective JAK1 inhibitor [
8,
9]. Data in this study, however, showed that at clinically relevant doses, upadacitinib was the most potent inhibitor among the drugs tested of the JAK2-dependent cytokines IL-3 and GM-CSF. These findings indicate that upadacitinib would also inhibit cytokines other than those primarily dependent upon JAK1 at clinically effective concentrations. Data in this study also demonstrate that tofacitinib has moderate activity against JAK2 and TYK2, in addition to activity against JAK1 and JAK3. Together, these findings suggest that conclusions about potency and selectivity drawn from refined in vitro kinase assays may differ when considering the more biologically relevant concept of signal blockade at the cellular level in the context of circulating drug concentrations at doses used in humans.
There are limitations to the conclusions that can be drawn from this analysis. One potential limitation with this study is that high concentrations of cytokine can right shift the IC50 values if the STAT substrate is limiting. This could potentially be limiting our interpretation of the data for upadacitinib. However, we investigated this phenomenon and adjusted the cytokine concentrations where necessary, but some cytokines were either insensitive to concentration or could not be lowered without compromising the data in an alternate cell type or STAT readout. It would have been valuable to assess the ability of each JAKi to impair the half maximal response concentration (EC50) of these cytokines, but would require extensive tuning of each cytokine in each cell type, which is well beyond the scope of this initial report. Another limitation was that the statistical analysis of this study used a single, average PK profile and did not reflect inter-subject variability. Furthermore, all IC50 values were calculated from PBMCs derived from healthy volunteers and extrapolated where available to RA patient exposure curves. Finally, while this study describes a reproducible cellular test system in which to test the molecules, the drugs were introduced to unstimulated healthy volunteer cells, which were then stimulated. Looking at how such molecules perform in previously activated cells may be more relevant to in vivo inflammatory disease conditions. This may warrant future study, for instance using samples from patients with active inflammatory disease.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.