Repression of interferon β-regulated cytokines by the JAK1/2 inhibitor ruxolitinib in inflammatory human macrophages

Ruxolitinib is a Janus kinase (JAK) 1/2 inhibitor, currently used in the treatment of myeloproliferative neo- plasms. It exerts potent anti-inflammatory activity, but the involved molecular and cellular mechanisms remain poorly understood. In order to gain insights about this point, ruxolitinib effects towards expression of main inflammatory cytokines were studied in human macrophages, which constitute a key-cell type implicated in inflammation. Analysis of mRNA expression of cytokines (n = 84) by PCR array indicated that, among those induced by the pro-inflammatory stimulus lipopolysaccharide (LPS) (n = 44), 61.4% (n = 27) were repressed by 5 μM ruxolitinib. The major inflammatory cytokines, interleukin (IL) 6 and tumor necrosis factor α, were notably down-regulated by ruxolitinib at both the mRNA and protein level. Other repressed cytokines included IL27 and the chemokines CCL2, CXCL9, CXCL10 and CXCL11, but not IL1β. The interferon (IFN) β/JAK/signal transducer and activator of transcription (STAT) pathway, well-activated by LPS in human macrophages as demonstrated by increased secretion of IFNβ, STAT1 phosphorylation, and up-regulation of reference IFNβ-responsive genes, was concomitantly blocked by the JAK inhibitor. Most of cytokines targeted by ruxolitinib were shown to be regu- lated by IFNβ in a JAK-sensitive manner. In addition, counteracting the IFNβ/JAK/STAT cascade using a blocking monoclonal antibody directed against IFNβ receptor resulted in a similar profile of cytokine repression to that observed in response to the JAK inhibitor. Overall, these data provide evidence for ruxolitinib-mediated repression of inflammatory cytokines in human macrophages through inhibition of the LPS/IFNβ/JAK/STAT signalling pathway, which probably contributes to the anti-inflammatory effects of the JAK inhibitor.

Ruxolitinib (also known as INCB018424) is a small drug belonging to the emerging class of Janus kinase (JAK) inhibitors and currently clinically used in the treatment of JAK2 V617F-positive myeloproli- ferative neoplasms, including intermediate or high risk myelofibrosis and polycythemia vera [1,2]. Besides the JAK2 isoform, primarily as- sociated with receptors for the hematopoietic growth factors ery- thropoietin and thrombopoietin, the JAK1 isoform, which plays a major role in the signalling pathway of inflammatory cytokines, is also po- tently inhibited by ruxolitinib [3]. By contrast, the two other members of the JAK family, JAK3 and tyrosine kinase 2 (TYK2), are less sensitive to ruxolitinib [4]. Ruxolitinib exerts anti-inflammatory activity, which may contribute to its beneficial effects towards myeloproliferative neoplasms [5] and is likely also implicated in the suppression by ruxolitinib of the harmful consequences of macrophage activation hemophagocytic lymphohis- tiocytosis [6] and of age-related inflammation [7]. Among the inflammatory cell types targeted by ruxolitinib, macrophage, a key cell type involved in inflammation [8,9], may likely be a major one, even if the exact effects of the JAK1/2 inhibitor towards this cell remain to be clarified.

Indeed, ruxolitinib has been demonstrated to rather increase the inflammatory potential of mouse macrophages stimulated with Toll- like receptor (TLR) 4 agonists such as lipopolysaccharide (LPS) [10], through stimulating secretion of the inflammatory cytokines tumor necrosis factor (TNF) α, interleukin (IL) 6 and IL12, whereas, by con- trast, it suppresses production of IL1β, TNFα, IL6, chemokine (C-X-C motif) ligand (CXCL) 10 and CXCL11 in TNFα-treated human macro- phages [11]. CXCL10 and CXCL11 expressions, but not those of IL1β, IL6 and TNFα, were also repressed by ruxolitinib in rheumatoid ar- thritis synovial macrophages [11]. The reasons for such discrepancies with respect to ruxolitinib effects towards cytokine regulation in mac- rophages remain unknown, but they could notably reflect differences related to the species [12] or to the nature of the inflammatory stimulus and the treatment conditions. To gain insights into ruxolitinib and cy- tokine regulation in inflammatory macrophages, in the present study we have analysed the expression of the main cytokine/chemokines in LPS-stimulated human macrophages in the presence and absence of this JAK1/2 inhibitor. Our data demonstrate that ruxolitinib was able to markedly suppress LPS-mediated induction of various human in- flammatory cytokines, through antagonizing the autocrine interferon (IFN) β-related regulatory signalling pathway secondary triggered by LPS in human macrophages. Such effects of ruxolitinib may contribute in a major way to its anti-inflammatory properties.

2.Materials and methods
Ruxolitinib, AT9283 and TG101209 were provided by Selleckchem (Houston, TX, USA), whereas LPS (Escherichia coli O55:B5) was from Sigma-Aldrich (Saint-Quentin Fallavier, France). Granulocyte macro- phage-colony stimulating factor (GM-CSF, also known as sargramostim) was obtained from Sanofi-Aventis (Montrouge, France), whereas IFNβwas from Peprotech (Neuilly-sur-Seine, France). The blocking anti-human IFNα/β receptor subunit 2 (IFNAR2) monoclonal antibody (clone MMHAR-2) was purchased from Tebu-bio (Le Perray-en- Yvelines, France), whereas those against phosphorylated (Tyr701)-signal transducer and activator of transcription (STAT) 1, nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and phos- phorylated (Ser172) -TANK-binding kinase 1 (TBK1) were from Ozyme (Saint-Quentin-en-Yvelines, France). The anti-rabbit IgG antibody coupled to Alexa Fluor was from Cell Signalling (Leiden, TheNetherlands). Other chemicals and reagents were of the highest avail- able grade.Macrophages were generated from peripheral blood monocytes, as previously described [13]. Briefly, peripheral blood mononuclear cells were first isolated from blood buffy coats of healthy donors through Ficoll gradient centrifugation. Human monocytes were then prepared by a 1 h adhesion step, which routinely obtained > 90% of adherent cells expressing the prototypical monocytic marker CD14 [14].

These monocytic cells were next cultured for 6 days in RPMI 1640 medium,supplemented with 10% (vol/vol) fetal bovine serum, 2 mM L-gluta- mine, 20 IU/mL penicillin and 20 μg/mL streptomycin, in the presence of 400 IU/mL GM-CSF; this protocol led to obtain pure macrophage cultures, with < 1% of contaminating cells [15]. Cultured macrophageswere then used for experiments, notably for treatment with LPS and/or ruxolitinib. Ruxolitinib and other chemicals were routinely used as stock solutions in dimethylsulfoxide (DMSO) and were commonly added 1 h before exposure to LPS, as a pre-treatment. Final con- centrations of DMSO in culture medium did not exceed 0.2% (vol/vol) and control cultures were exposed to the same concentration of solvent as chemical-treated cultures.Human highly-differentiated hepatoma HepaRG cells were culturedin Williams' E medium supplemented with 10% (vol/vol) fetal bovine serum, 20 IU/mL penicillin, 20 μg/mL streptomycin, 5 μg/mL insulin, and 5 × 10−5 M hydrocortisone hemisuccinate; their hepatocytic dif- ferentiation was induced by addition of 2% (vol/vol) DMSO for 2 weeks, as previously described [16].The effect of ruxolitinib on cell viability was analysed using the tetrazolium salt WST-1 reagent, as previously reported [17].Total RNAs were extracted using the TRI reagent (Sigma-Aldrich). The RT2 Profiler™ PCR array “Human cytokines and chemokines” was obtained from Qiagen (Hilden, Germany). This PCR array containsgene-specific qPCR assays for a thoroughly researched set of 84 genes relevant to cytokine/chemokine area as well as for 5 various house- keeping genes, put into 96-well plates (See Table S1 for the complete list of genes investigated in the array). Total RNAs were extracted as described above and then purified using the spin column-based RNeasy Mini Kit (Qiagen). For each condition, an equimolecular pool of RNAs from five independent assays (5 × 80 ng total RNA) was prepared and was reverse transcribed. PCR experiments were next performed and analysed according to manufacturer instructions. Levels of mRNA ex-pression were normalized to mRNA expression of β2-microglobulin,used here as a house-keeping gene, and were commonly expressed as induction factor, i.e., the ratio mRNA level in treatment condition versus that found in control or reference conditions, or as repression factor, i.e., the ratio mRNA level in control condition versus that found in treatment conditions.The putative inducibility by type I IFN (IFNα and IFNβ) of LPS- regulated cytokines was studied online using the interferome database ( [19]. Cytokinewas considered as type I IFN-inducible when at least one condition of type I IFN-treatment reported in the database results in increased mRNA expression by at least a 2-fold factor. It was considered as pu- tatively non-inducible when the criteria of type I IFN-responsiveness described above was not reached or when no data concerning potential regulation by type I IFN was available in the database. The levels of cytokines secreted in culture medium of macrophage cultures were quantified by the enzyme-linked immunosorbent assay (ELISA) using specific human Duoset ELISA Kit (R&D Systems, Minneapolis, MN, USA), according to the instructions of the manu- facturer. Data were finally collected by spectrophotometry using a SPECTROstar Nano microplate reader (BMG Labtech, Champigny sur Marne, France).Macrophages were fixed with a cold solution of acetone for 10 min on ice. Cells were then washed with phosphate-buffered saline (PBS) and incubated in PBS containing 4% bovine serum albumin for 1 h at room temperature. Cells were next incubated with primary antibody against phosphorylated-STAT1, NF-κB or phosphorylated-TBK1 over-night. After two washes with PBS, the primary antibody was detectedwith anti-rabbit IgG coupled to Alexa Fluor from Cell Signalling (Leiden, The Netherlands), whereas nuclei were stained by 4′,6-diami- dino-2-phenylindole (DAPI), for 1 h at room temperature. Immunofluorescence images were finally acquired with a confocal fluorescence microscope LEICA DMI 6000 CS (Leica Microsystemes SAS, Nanterre, France).Quantitative data were routinely expressed as means ± SEM of at least three independent assays. They were statistically analysed using analysis of variance (ANOVA) followed by the Newman-Keuls' post-hoc test. Linear correlation between quantitative data was investigated using the Pearson correlation coefficient. Chi-square analysis was con- ducted to examine association between categorical data. The criterion of significance was p < 0.05. Half maximal inhibitory concentration (IC50) values of ruxolitinib towards cytokine mRNA expression were determined using GraphPad Prism 5.0 software (GraphPad Software, La Jolla, CA), through nonlinear regression based on the four parameter logistic function. 3.Results The RT2 Profiler™ PCR array “Human cytokines and chemokines” was firstly applied to characterise cytokine mRNA expression in pri- mary human macrophages exposed to 10 ng/mL LPS for 6 h, in the absence or presence of 5 μM ruxolitinib. This concentration of rux-olitinib was initially used because it is in the range of in vitro con-centrations previously shown to be fully active towards JAKs [20,21]. This 5 μM concentration of ruxolitinib, as well as those of 1 and 10 μM, did not alter viability of macrophages in response to a 7 h or 24 h ex- posure (Fig. S1). The time of exposure to LPS (6 h) was chosen for the study because it is representative of the peak of inflammatory generesponses, whereas a shorter time (2 h) or a longer time (24 h) of LPS treatment reflect, respectively, early gene response or subsequent re- solution phase [12,22]. The number of genes whose mRNA expression was induced or repressed by LPS or ruxolitinib using a threshold value of 2 is indicated in Table 1. LPS treatment was found to alter mRNA expression of 56.0% of analysed genes, most of them being induced (See Table S2 for the complete list of induction factors for genes whose mRNA expression was increased by LPS). By contrast, ruxolitinib only minimally affected constitutive cytokine mRNA expression, i.e., only 14.3% of analysed genes has impaired mRNA expression in ruxolitinib- exposed macrophages when compared to untreated counterparts (Table 1). However, ruxolitinib more obviously affected cytokine mRNA expression in LPS-exposed macrophages, with a total of 41.6% of targeted genes, most of them being repressed (Table 1). The complete list of repression factors for the genes down-regulated by ruxolitinib in inflammatory LPS-treated macrophages is provided in Table 2; these ruxolitinib-sensitive genes belong to various families of cytokines (i.e., chemokines, interleukins, growth factors…), and include major in-flammatory cytokines such as IL6 and TNFα (Table 2). The number ofthese ruxolitinib-repressed genes (n = 27) represents 61.4% of the total number of genes up-regulated by LPS (n = 44) (Table 1 and Table S2). A notable proportion of LPS-up-regulated cytokines, corresponding to 38.6% (n = 17) of the total number of LPS-up-regulated cytokines (n = 44), was consequently not repressed by ruxolitinib. These rux-olitinib-unresponsive genes notably included some cytokines highly inducible by LPS such as IL12B, GM-CSF, CCL20, CXCL2 and IL1β (Table S2).To validate the results of the PCR array, single qPCR experiments were then performed using macrophages derived from monocytes of various individuals. As shown in Fig. 1, the LPS-mediated up-regulation of various cytokines such as CCL2, CCL7, CCL8, CXCL9, CXCL10, CXCL11, FASLG, IL6, IL7, IL10, IL15, IL27, IFNγ, TNFα and TRAIL werefully confirmed to be repressed by ruxolitinib. Such repression was al-most complete for most cytokines; LPS-mediated inductions of cyto- kines such as TNFα and IL6 were nevertheless only partly repressed by the JAK inhibitor (Fig. 1). Single qPCR assays additionally confirmed the fact that LPS-mediated up-regulations of IL1β and CXCL2 were not impaired by ruxolitinib (Fig. 1). Repression factors due to ruxolitiniband determined from single qPCR assays were finally shown to be highly correlated with those from the PCR array (Fig. S2), thus likely fully validating the data from the PCR array. The repressing effects of ruxolitinib towards mRNA expression of LPS is known to regulate cytokine production through activation of its receptor, i.e., the Toll-like receptor (TLR) 4, and the subsequent activation of two main distinct signalling cascades: (1) a myeloid dif- ferentiation factor 88 (MyD88)-dependent pathway, which notably re- sults in nuclear translocation of the transcription factor NF-κB throughphosphorylation/inactivation of inhibitors of κB (IκBs) and subsequentproduction of NF-κB-responsive cytokines like IL1β, and (2) a MyD88-independent pathway involving the TLR4 adaptor TIR-domain-con- taining adapter-inducing IFNβ (TRIF), which triggers activation/phos- phorylation of TBK1, following by that of the transcription factor in- terferon regulatory factor (IRF) 3, and finally resulting in production ofIFNβ. IFNβ up-regulated by LPS/TLR4/TRIF/TBK1/IRF3 pathway sec- ondary induces various IFN-stimulated genes (ISGs), including cyto-kines, through autocrine activation of its receptor composed of IFNAR1 and IFNAR2 subunits [26–29]. Because the signalling cascade of IFNAR1/2 basically involves the JAK1/STAT1 pathway [30] and be- cause LPS-mediated induction of some cytokines has already beenshown to require STAT1 activity [31], we hypothesised that ruxolitinib may alter LPS-mediated regulation of cytokines through impairing the autocrine effects of IFNβ. To investigate this hypothesis, we beforehand determined whether ruxolitinib efficiently counteracts LPS-triggeredIFNβ signalling pathway in human macrophages. As shown in Fig. 4, LPS up-regulated IFNβ at both mRNA and protein secretion levels in primary human macrophages. Induction of IFNβ mRNA levels was however transient and peaked after a 2 h exposure to LPS (Fig. 4A), whereas IFNβ secretion appeared in a delayed manner after a 4 h LPS- exposure and persisted after a 6 h LPS-exposure (Fig. 4B). This LPS- mediated secretion of IFNβ appeared to be functional, because mRNA expressions of reference ISGs such as MX1, ISG15, protein kinase R(PKR) and IRF7 were markedly induced in response to exposure to LPS (Fig. 5A). The kinetic of these inductions, i.e., up-regulations in re- Primary human macrophages were either untreated (control), or exposed to 10 ng/mL LPS for 1, 2, 4 or 6 h. (A) IFNβ mRNA expression was determined by RT-qPCR. Data are expressed as mRNA fold change compared to untreated macrophages, i.e., the ratio mRNA expression in treated-macrophages versus that found in untreated cells. (B) IFNβ secretion in culture medium was determined by ELISA. Data are expressed as pg/mL. (A,B) Data are the means ± SEM of at least 3 independent experiments. **, p < 0.01; ***, p < 0.001 when compared to control.cytokines were next shown to be dose-dependent with IC50 values ranging from 9 nM (for CXCL9) to 128 nM (for CCL2) (Fig. 2). They were moreover associated with concomitant significant repression of LPS-induced cytokine secretion in supernatants of macrophage cultures for CXCL10, TNFα, IL6 and CCL2 (Fig. 3).Because ruxolitinib may exert JAK-independent effects [23], theeffects of JAK inhibitors structurally unrelated to ruxolitinib, such as AT9283 and TG101209, were studied. As indicated in Fig. S3, these JAK inhibitors used at 5 μM were found to fully repress LPS-mediated mRNA induction of cytokines such as CCL2, CXCL10, IL27 and TRAIL in pri- mary macrophages. Like ruxolitinib (Fig. 1), they also partly decreased LPS-triggered TNFα mRNA induction (Fig. S3).To determine whether ruxolitinib may also counteract pro-in-flammatory factors which do not primarily signal via JAKs, such as IL1βand TNFα, its putative effect towards IL1β- and TNFα-mediated up- regulation of the inflammatory chemokine IL8 was studied in humanhepatoma HepaRG cells, used here as a reference cellular model re- sponsive to inflammatory cytokines [24,25]. As indicated in Fig. S4, ruxolitinib failed to impair IL1β- and TNFα-induced IL8 mRNA ex- pression in HepaRG cells. sponse to 4 h and 6 h exposure to LPS, but not for shorter exposures (1 h and 2 h) (Fig. S5), was moreover fully consistent with that of IFNβ secretion (Fig. 4B). Such LPS-mediated inductions of these ISGs were fully inhibited by ruxolitinib (Fig. 5A), thus suggesting that they re- quired activation of the JAK/STAT signalling. In agreement with thishypothesis, LPS was found to induce phosphorylation of STAT1 and its subsequent translocation into the nucleus, which were fully blocked by ruxolitinib (Fig. 5B). Taken together, these data demonstrate that rux- olitinib was able to fully repress the LPS-mediated activation of IFNβ signalling pathway in primary human macrophages. Ruxolitinib wasadditionally found to significantly, but partly, counteract LPS-mediated up-regulation of IFNβ mRNAs (Fig. 6A) and IFNβ secretion (Fig. 6B) in primary cultures of human macrophages. By contrast, ruxolitinib did not alter LPS-triggered nuclear translocation of NF-κB in macrophages (Fig. S6A) nor LPS-mediated phosphorylation of TBK1 (Fig. S6B).The putative inducibility by type I IFN, like IFNβ, of LPS-regulated cytokines was first analysed using the online interferome database and is indicated in Table S2 for each cytokine. When globally analysed(Table 3), the majority (66.7%) of LPS-induced genes repressed by ruxolitinib was found to be inducible by type I IFN by our in silico approach. By contrast, LPS-induced cytokines not sensitive to rux- olitinib were, for most of them (94.1%), not responsive to type I IFN (Table 3 and Table S2), thus underlining the link between sensitivity to ruxolitinib and response to IFN. To confirm this relationship, we next investigated the direct regulation of various ruxolitinib-sensitive cyto-kines by IFNβ in primary human macrophages. As indicated in Fig. 7, CCL2, CXCL9, CXCL10, TRAIL, IL6 and TNFα mRNA expressions were found to be up-regulated by a 4 h exposure to 50 ng/mL IFNβ, in a JAK/ STAT-dependent manner as demonstrated by the suppressive effects of 5 μM ruxolitinib. By contrast, IL1β, which is not sensitive to ruxolitinib (Fig. 1), was not up-regulated by IFNβ in macrophages (Fig. 7). IL-27, although inhibited by ruxolitinib in LPS-exposed macrophages (Fig. 1), was also not induced by IFNβ in primary human macrophages (Fig. 7). To finally determine whether abrogation of the IFNβ signalling cascade may mimic the activity of ruxolitinib towards cytokine ex-pression, we analysed the effects of a blocking antibody directed against IFNAR2 in LPS-exposed macrophages. As shown in Fig. S7, the use of this blocking anti-IFNAR2 antibody fully reproduced the inhibitory ef- fects of ruxolitinib towards LPS-mediated induction of reference ISGs such as MX1, ISG15, PKR and IRF7, which demonstrates its efficiency for counteracting IFNβ signalling in macrophages. The anti-IFNAR2antibody was next shown to block LPS-mediated up-regulation of var-ious cytokines, including CCL2, CXCL9, CXCL10, CXCL11, IL15, IL27, TNFα and TRAIL (Fig. 8). Such effects of this anti-IFNAR2 antibody were similar to those caused by ruxolitinib, which fully supports rux- olitinib efficiency for counteracting IFNβ signalling in human macro- phages (Fig. 7). By contrast, the anti-IFNAR2 antibody failed to inhibit IFNβ mRNA induction caused by LPS in human macrophages, whereas it concomitantly fully suppressed that of ISG15 (Fig. S8). 4.Discussion In the present study, we demonstrated that the JAK1/2 inhibitor ruxolitinib markedly counteracts the LPS-mediated induction of various cytokines in human macrophages. Such data fully confirm the fact that inflammatory cytokines constitute key-targets for ruxolitinib and sup- port its now well-established, but possibly complex, interactions with between human and mouse macrophages [12], likely supports this hypothesis. Moreover, despite its putative pro-inflammatory activity in mouse macrophages, ruxolitinib has been reported to exert rather po- tent anti-inflammatory effects in various murine models of inflamma- tion, including those related to macrophage activation hemophagocytic lymphohistiocytosis [6], age-related inflammation [7], experimental candida sepsis [36], chronic pancreatis [37] and pancreatic cancer [38]. Such data overall support a dominant anti-inflammatory effect of ruxolitinib in mouse species.The molecular mechanism by which ruxolitinib down-regulates expression of various cytokines in human macrophages most likely in- volves inhibition of the JAK-STAT pathway. Indeed, the IC50 values of ruxolitinib towards LPS-mediated up-regulation of various cytokines are in the 10–100 nM range, which corresponds to concentrationspreviously shown to inhibit the JAK-STAT pathway in cultured cells[4]. Also, structurally-divergent JAK inhibitors such as AT9283 and TG101209 prevented LPS-mediated up-regulation of cytokines. More- over, ruxolitinib efficiently blocked JAK-mediated phosphorylation and subsequent nuclear translocation of STAT1, known to be required for induction of some LPS-regulated genes [39]. Ruxolitinib also counter-acted LPS-mediated induction of references ISGs such as MX1, which are well-known to be induced via IFNβ-mediated activation of the JAK1/STAT1 pathway in inflammatory macrophages [27]. Lastly, ruxolitinib failed to inhibit JAK-independent IL8 mRNA up-regulationin IL1β- or TNFα-treated hepatoma HepaRG cells and it did not impair LPS/TLR4-triggered activation of JAK-unrelated molecular signalling actors, like NF-κB (MyD88-dependent pathway) and TBK1 (MyD88-in- dependent pathway), in macrophages. In fact, the inhibition of the JAK- dependent IFNβ signalling cascade probably contributes in a major way to the nature of the cytokines targeted by ruxolitinib in human mac- inflammatory/immune processes [32,33]. Expression of major in- flammatory cytokines such as TNFα and IL6 was notably reduced in inflammatory human macrophages exposed to ruxolitinib, which may agree with clinical beneficial anti-inflammatory properties of the JAK1/ 2 inhibitor in various human diseases, notably in myeloproliferativeneoplasms [5], graft-versus-host disease [34] and cutaneous diseases [35]. In contrast to our data, ruxolitinib has however been reported to stimulate secretion of TNFα and IL6 in mouse bone-marrow macro-phages [10]. Such a discrepancy may reflect species-specific response ofmacrophages to ruxolitinib. The fact that such divergences in response to inflammatory effectors, including LPS, have already been reported rophages. Indeed, most of the cytokines down-regulated by the JAK inhibitor, including notably CCL2, CXCL9, CXCL10 and TRAIL, were found to be inducible by IFNβ and abrogation of IFNβ signalling using a blocking antibody directed against the IFN receptor IFNAR2 fully mi- micked the effects of ruxolitinib. The JAK inhibitor appears therefore to counteract IFNβ-dependent effects triggered by TLR4/TRIF/IRF3-de- pendent signalling cascade, through inhibition of the JAK1/STAT1 signalling pathway activated by TRIF-induced IFNβ (Fig. 9). Autocrineeffects of IFNβ in human inflammatory macrophages are consequentlyabrogated by ruxolitinib-mediated inhibition of JAK1/2 activity. The partial, but significant, reduction of LPS-induced IFNβ secretion due to ruxolitinib may additionally contribute to its inhibitory activity to- wards the TRIF/IFNβ signalling cascade activated by LPS. The way by which ruxolitinib diminishes, but not abolishes, IFNβ up-regulation in LPS-exposed human macrophages remains to be determined. Impair-ment of a putative autocrine amplification of IFNβ secretion by the cytokine itself can however be discarded, as the use of an anti-IFNAR2 antibody failed to alter IFNβ induction in response to LPS. Similarly, an inhibition of the first steps of the MyD88-independent TLR4/TRIF/TBK1/IRF3 pathway, leading to activation of TBK1, is very unlikely, as ruxolitinib did not inhibit phosphorylation of TBK1 in LPS-treated human macrophages. The JAK inhibitor may therefore be hypothesised to partially decrease activity of IRF3 or to partially inhibit other uni- dentified molecular ways required for proper IFNβ expression. Furtherstudies are likely required to investigate these points.In contrast to the MyD88-independent TRIF/IFNβ pathway, the MyD88/NF-κB cascade activated by LPS is unlikely to be impaired by ruxolitinib (Fig. 9). Indeed, induction by LPS of IL1β, a prototypical cytokine regulated by NF-κB [40], is not altered by the JAK inhibitor, as well as LPS-induced nuclear translocation of NF-κB. In this context, it may be hypothetized that ruxolitinib probably failed to block the well- established NF-κB-mediated induction of IL6 and TNFα in response to TLR4 activation. Partial reduction of the expression of these cytokines by ruxolitinib thus likely results from blockage of IFNβ signalling, knowing that TNFα and IL6 were demonstrated to be inducible by IFNβ in the present study. In addition, inhibition of other JAK/STAK signalling cascades, activated by LPS-induced cytokines distinct from IFNβ, may be implicated. With respect to IL27, its LPS-induced ex- pression is markedly inhibited by ruxolitinib and also by the blocking antibody directed against IFNAR2, thus demonstrating that IFNβ sig- nalling is required for its induction by LPS. IL27 was however not found to be directly inducible by IFNβ itself in human macrophages; this suggests that its up-regulation by LPS may in fact initially involve an unknown IFNβ-dependent factor, which secondary stimulates IL27 ex- pression.The exact contribution of the ruxolitinib-mediated repressions of distinct cytokines to its in vivo anti-inflammatory effects remains to be precisely determined. As already discussed above, the partial repression of IL6 and TNFα expression by ruxolitinib is likely to contribute, owing to the major inflammatory role of these cytokines. Decreased expressionof the chemokines CCL2, CCL7, CXCL9, CXCL10, CXCL11, which act as pro-inflammatory chemoattractants for immune cells like monocytes, macrophages or T lymphocytes [41,42], may also support the anti-in- flammatory activity of the JAK inhibitor. The same hypothesis may be drawn for ruxolitinib-mediated repression of IL15 and IL27, owing to the pro-inflammatory role of these cytokines [43,44]. Interestingly, macrophagic cytokines are repressed by ruxolitinib concentrations inthe 10–100 nM range, which is easily reached in plasma of patientstreated by ruxolitinib [45]. These cytokines may therefore be postu- lated to be efficiently in vivo targeted by administration of the JAK inhibitor. This hypothesis is supported by the fact that plasma levels of IL6 and TNFα have been reported to decrease in patients suffering from myelofibrosis and receiving ruxolitinib [1]. Further studies are required to determine whether other cytokines targeted by the JAK1/2 inhibitorin macrophages and up-regulated in myeloproliferative neoplasms, such as IL7, IL15, CCL2, CXCL9 and CXCL10 [46,47], may be also re- pressed in plasma of ruxolitinib-treated patients.Besides pro-inflammatory cytokines, IL10, a cytokine known to limit and ultimately terminate inflammatory responses [48] and to be in- ducible by type 1 IFN in response to LPS [49], is also repressed by ruxolotinib. This JAK1/2 inhibitor may therefore be considered as ex- erting some opposite effects, i.e., it represses expression of both in- flammatory and anti-inflammatory cytokines in macrophages. The net balance, which has to integrate the effects of ruxolitinib towards other immune cells, such as T lymphocytes [50], dendritic cells [21] and NK cells [20], tilts nevertheless likely in favour of an anti-inflammatory role, as exemplified by the improvement of inflammatory symptoms observed in patients treated by ruxolitinib [5] and the known dominantIFN activities stimulating macrophages for enhanced effector function [51]. In addition, IFNβ has been demonstrated to significantly con- tribute to LPS-induced lethality [28]. Ruxolitinib, through counter- acting IFNβ signalling, may consequently protect against mortality due to bacterial infections.Apart from its potential beneficial anti-inflammatory activity, rux- olitinib may exert deleterious effects on health, through the repression of the TLR4/TRIF/IFNβ signalling cascade. Indeed, IFNβ, like other IFNsubtypes, plays a major role in antiviral response [52,53]. Abrogation of IFNβ effects by ruxolitinib, which is moreover associated with down- regulation of LPS-mediated type II IFNγ expression, may consequently favour viral infection. The fact that administration of ruxolitinib topatients resulted in the development or reactivation of viral infections [54–56], supports this hypothesis. In summary, ruxolitinib was shown to prevent the up-regulation of various pro-inflammatory cytokines in human inflammatory macro- phages, mainly through abolishing LPS-induced activation of the TLR4/ TRIF/IFNβ signalling pathway. Such effects indirectly confirm that IFNβ plays a major role in the response to LPS in human macrophages, as previously established in mouse counterparts [28]. They also fully supports the promising and already well-established anti-inflammatory effects of INCB084550 ruxolitinib.