Bromodomain and extraterminal (BET) protein inhibition of IgG/IgE production in murine B cells is counter‐balanced by a strong Th2 bias

Abstract Objectives Inhibitors of bromodomain and extra terminal domain (BET) proteins are a new and growing class of anti‐cancer drugs, which decrease oncogene expression by targeting superenhancers. Antibody production is another physiological process relying on superenhancers, and it remains to be clarified whether potential immunomodulatory properties of BET inhibitors might impact humoral immunity and allergy. Methods We thus evaluated humoral immune responses and their Th2 context in vitro and in vivo in mice following treatment with the classical BET‐inhibitor JQ1. We quantified immunoglobulin (Ig) and antibody production by B cells either stimulated in vitro or obtained from immunised mice. JQ1 effects on class switching and activation‐induced deaminase loading were determined, together with modifications of B, T follicular helper (Tfh) and T helper 2 (Th2) populations. JQ1 was finally tested in B‐cell‐dependent models of immune disorders. Results Bromodomain and extra terminal domain inhibition reduced class switching, Ig expression on B cells and antibody secretion and was correlated with decreased numbers of Tfh cells. However, JQ1 strongly increased the proportion of GATA3+ Th2 cells and the secretion of corresponding cytokines. In a mouse allergic model of lung inflammation, JQ1 did not affect eosinophil infiltration or mucus production but enhanced Th2 cytokine production and aggravated clinical manifestations. Conclusion Altogether, BET inhibition thus interweaves intrinsic negative effects on B cells with a parallel complex reshaping of T‐cell polarisation which can increase type 2 cytokines and eventually promote B‐cell‐dependent immunopathology. These opposite and potentially hazardous immunomodulatory effects raise concerns for clinical use of BET inhibitors in patients with immune disorders.


INTRODUCTION
Inappropriate production of pro-inflammatory class-switched IgG or IgE is a major contributor to immunopathology. Besides direct effects, some class-switched antibodies (Abs) also behave as natural adjuvants for T helper (Th) cells 1,2 or, in contrast, play regulatory roles such as blocking Abs of the human IgG4 class. Among immunosuppressive drugs, few can target B-cell function and none directly modulate class switch recombination (CSR). It would thus be of strong interest to identify means to either globally depress CSR or, ideally, to specifically target the production of the most pro-inflammatory classes such as IgE.
Class switch recombination in B cells needs recruitment of activation-induced deaminase (AID) to initiate DNA lesions on target 'switch' (S) regions of the IgH locus, under control of the 3 0 regulatory region (3 0 RR) superenhancer (SE). 3,4 This initiates double-strand breaks within single-stranded DNA structures and R-loops of S regions. Chromatin readers from the bromodomain and extraterminal (BET) family proteins can promote CSR by participating in the repair of broken DNA ends 5 but also likely impact CSR and Ig production by interacting with SEs. We thus wished to evaluate whether a BET protein inhibitor would directly impact CSR via the 3 0 RR SE and what would be the consequences on the humoral response.
Bromodomain and extraterminal proteins are enriched at positions of active promoters, enhancers and, to a higher extent, SE, where they promote recruitment of mediator and RNA polymerase II. 6 SEs and BET proteins such as BRD4 contribute to inflammatory or malignant processes and activate translocated oncogenes. 7,8 Widespread development of BET inhibitors notably includes treatment of solid tumors to reduce post-radiotherapy lung fibrosis, [8][9][10][11] and these drugs have rapidly entered into therapy trials, eventually ahead of science. 12 The impact of BET inhibition on immune responses is notably poorly understood. Li et al. 13 reported that the BET inhibitor JQ1 reduced Th9 responses and might be of interest in controlling airway inflammation. 13 JQ1 also inhibited type 2 innate lymphoid cells and appeared beneficial in a short-term (3-day long) airway allergy model, 14 but the impact of BET inhibition on full-blown allergy remains to be determined.
BRD4 controls expression of Bcl-6, the master regulator of germinal centre (GC) formation, and recruits non-homologous end-joining (NHEJ) factors for late DNA repair during CSR. 5,15 Except for DNA repair, no effect of BET proteins on early steps of CSR is known. 5 Accessibility and germline transcription (GT) of target IgH switch (S) and constant (C) regions are modulated by cytokines and required for CSR. 16 Class switch recombination needs chromatin remodelling and the occurrence of DNA loops for synapsis of S regions before their joining, which is governed by the 3 0 RR SE. 7,8,12 The 3 0 RR is also targeted by locus suicide recombination (LSR). 4 The same locus driven by the same SE can therefore be driven in opposite directions, under strong control by cytokines and T-cell polarisation.
We wished to determine whether BET proteins might modulate humoral responses, either by directly affecting remodelling of the IgH locus or by indirectly modulating the polarity of B/T interactions. We thus evaluated the mechanisms through which a low and non-toxic dose of JQ1 might, on the one hand, affect the B-cell response (3 0 RR function, AID loading, CSR, Ig production), and, on the other hand, impact T-cell function and the global inflammatory or allergic outcome of humoral responses in vivo in mice.

RESULTS
Determination of the non-toxic concentration of JQ1 JQ1 has been widely tested as an anti-cancer agent. It proved effective against mouse tumors in vivo, with a maximum tolerated dose of 50 mg kg À1 per day, and showed antiproliferative effects in vitro. To assess its role on B cells while minimising non-specific effects, we used low doses in the 10-40 nM range for in vitro assays and chose the non-toxic dose of 30-50 mg kg À1 per day for in vivo assays.
We thus validated that the low doses used in vitro did not significantly affect CD19 + B-cell absolute numbers in LPS + IL-4-stimulated cultures (Figure 1a) nor influence the percentage of apoptotic cells, in cultures including up to 40 nM JQ1 (Figure 1b).

Evaluation of in vitro CSR by cell cytometry and ELISA
While absolute numbers of CD19 + cells obtained after in vitro stimulation were not significantly changed in 4-day in vitro stimulation cultures w/ wo JQ1, we looked for qualitative variations in BCR expression and Ig secretion.
To evaluate whether JQ1 modulated classswitching, sorted mouse B spleen cells were stimulated in vitro for 4 days by LPS + IL-4 known to boost CSR and further expression of classswitched IgG1 and IgE. Direct evaluation of class switching in B lymphocytes, by following cellsurface BCR expression after LPS + IL-4 stimulation, showed a strong reduction in the amount of IgG1 class-switched cells observed, with a onefold reduction at 20 nM JQ1, a threefold reduction at 40 nM JQ1 and a reciprocal increase in IgM + CD19 + unswitched cells (Figure 1c).
Parallel ELISA evaluation of Ig secretion in cell supernatants revealed no significant reduction in IgM levels. By contrast, and to a much stronger extent than for BCR expression, secretion of classswitched Ig produced in such conditions (i.e. IgG1 and IgE with LPS + IL-4) decreased for almost all doses of JQ1 tested (Figure 1d).

AID recruitment to S regions and structure of CSR junctions
We measured the loading of AID on target S regions by ChIP experiments in chromatin prepared from B cells stimulated in vitro for CSR (using LPS + IL-4) and observed its drastically reduced recruitment to Sc1 as well as Sɛ regions ( Figure 2a). Part (but not all) of this strong reduction in AID loading might result from decreased expression, since a partial decrease in Aicda gene (encoding AID) transcription was noticed in LPS + IL-4-stimulated cells (Figure 2b).
Sl-Sc1 CSR junctions were PCR-amplified with specific primers, sequenced by next-generation sequencing and analysed as previously described according to their nature and position within S regions.
The most striking change was quantitative, with approximately a fourfold decrease in the number of junctions after JQ1 treatment (Figure 2c).
The repair process after the occurrence of DSBs was, however, unchanged with regard to the use of NHEJ versus the alternate microhomologymediated end-joining (MMEJ) pathway; the strong predominance of NHEJ-mediated blunt junctions, normally observed in CSR, was maintained ( Figure 2d). The unaltered NHEJ/MMEJ ratio indicates that the DNA repair defect induced by JQ1 equally affects the processes of NHEJ and MMEJ. There was also no significant difference in the position of breaks, regarding their distance to the AID-targeted RGYW/WRCY consensus motifs ( Figure 2e).
While AID expression and IgH loading were decreased in the presence of JQ1, DNA breaks were thus less frequent and more focused on canonical AID target sites.

CSR blockade occurs downstream from IgH germline transcription
To examine transcription of CSR targets in LPS/IL-4 in vitro stimulated B cells, we quantified two types of IgH constant (C) gene transcripts, respectively, specific for the 'pre-CSR' (Ic1-Cc1 and Il-Cl germline transcripts originating from unswitched B cells) and the 'post-CSR' stages (i.e. Il-Cc1 and Il-Cɛ switched transcripts). Upon JQ1 treatment, we observed an increase in 'pre-CSR' transcripts that are hallmarks of local IgH accessibility to CSR. By contrast, a specific decrease in IgG1 and IgE class-switched transcripts was seen ( Figure 2f).
IgH locus transcripts are not only found around coding regions and, as for other enhancers, accessibility and activity of the IgH 3 0 RR core enhancer elements can be assessed by evaluating their transcription into so-called 'eRNA'. 3 Comparison of untreated and JQ1-treated LPS + IL-4-stimulated B cells revealed a major dose-dependent reduction in eRNA transcripts from the 3 0 RR hs4 and, to a lesser extent, hs1.2 core enhancer elements (Figure 2g). JQ1 treatment thus did not merely result in globally decreased IgH transcription, but rather modulated the relative amounts of the various IgH transcripts. This indicated decreased activity of      the 3 0 RR superenhancer. It also precisely situated the CSR blockade in between the stages of 'pre-CSR' transcription (increased) and 'post-CSR transcription' (decreased) and pointed to a direct alteration of the recombination step itself.

JQ1 inhibits antigen-specific T-and B-cell humoral responses
To evaluate in vivo modulation of immune responses, we immunised mice with ovalbumin (OVA), which was covalently conjugated to the T- although less markedly than for IgG), suggesting that on a 'per-cell' basis, IgM secretion also tended to be lower in animals receiving JQ1.
We evaluated the extent and nature of the Agspecific B-cell response (see the Methods for precise description of methodology and Supplementary figure 1 for gating strategy). Mice receiving JQ1 showed a significant decrease in total B220 + B cells (Figure 3c and Supplementary figure 4). OVAspecific switched (IgD neg ) B cells appeared higher in frequency ( Figure 3c) but were lower in number in JQ1-treated mice (data not shown), because of the overall decreased total B-cell count.
Among OVA-specific B cells, the frequencies of GC B cells and plasma cells did not significantly differ between both experimental groups ( Figure 3c). In contrast, as observed for immunisation with OVA alone and in vitro in Figure 1, JQ1 impacted isotype switch, as shown by the decrease in frequency of IgG1 + OVAspecific B cells ( Overall, these experiments further show that in vivo JQ1 impacts CSR intrinsically in B cells but also strongly inhibits the development of Tfh cells and enhances Th2 cell polarisation.

JQ1 has contrasting immunomodulatory effects in an ovalbumin-induced asthma model in mice
Using BALB/c mice which had been repetitively immunised i.p. with OVA, we explored whether the occurrence of allergic manifestations triggered by further administration of the immunising ovalbumin antigen could be modulated by JQ1. Four groups of mice received either saline, OVA or OVA + steroid known to prevent the occurrence of allergic asthma (Budesonide 3 mg kg À1 , i.n.), or OVA + JQ1 (50 mg kg À1 , daily i.p injection).
JQ1-treated mice showed reduced systemic OVA-specific IgG1 and IgE responses at day 25 (after primary injection of OVA followed by i.p. Given this apparent dissociation of the humoral and Th2 responses, we decided to analyse the impact of JQ1 on another type of immune disorder known to be T and B cell dependent in more detail. Immunomodulation by JQ1 does not alleviate experimental autoimmune encephalitis JQ1 was shown to have a beneficial effect on the severity of the classical, Th17-dependent and non-B-dependent experimental autoimmune encephalomyelitis (EAE) induced with the immunodominant peptide of myelin oligodendrocyte glycoprotein (MOG). 17 We therefore wished to study whether the immunomodulatory effects of JQ1 might be beneficial in another inflammatory condition known to be B cell dependent: EAE induced in mice by immunisation with recombinant human MOG (rhMOG). 18 After immunisation with rhMOG, when disease onset became evident in some mice (clinical score = 1), mice were treated with vehicle or JQ1 (30 mg kg À1 , daily i.p injection). In contrast to what was observed in Th17-dependent EAE, no significant differences in the severity of B-cell-dependent EAE were observed between vehicle-and JQ1-treated mice (Supplementary figure 3a), but disease onset, however, appeared significantly delayed (Supplementary figure 3b).

DISCUSSION
Class-switched Abs are crucial actors and regulators of immune responses and can be strongly pro-inflammatory, which is notably the case for IgE when produced in excess. Besides some direct pro-inflammatory effects, IgG can also boost cellular responses. 1,2 Drugs targeting CSR and able to modulate the production of classswitched Ig might thus be a grail for therapeutic immunomodulation. Since CSR is dependent upon a SE which is itself bound by BET factors, we wished to evaluate whether the BET-inhibitor JQ1 might be active on the IgH SE which controls CSR, and whether the global effect of BET inhibition on the immune system might alleviate allergic reactions. While JQ1 and related BET inhibitors are actively studied for inhibition of tumor cell growth, they were also shown to inhibit generation of Th17-dependent immune reactions. 7,17 By contrast, and while JQ1 is known to inhibit DNA repair during CSR in B cells, 5 its exact impact on the whole CSR process has not been documented and its global in vivo impact on immune humoral responses and on B-celldependent immune disorders, such as allergic asthma, remains to be explored.
In the present study, we validated that JQ1, at doses up to 40 nM, is not cytotoxic in cultured B cells in vitro, and is well tolerated in mice up to a dose of 50 mg kg À1 per day without major adverse effects. We also confirmed a direct effect of JQ1 on CSR by following BCR expression on the cell membrane of in vitro stimulated B cells. This inhibition also manifested in vivo in immunised mice when specific Ig-producing cells were quantified by ELISPOT, showing an increase in IgM-producing cells vs a decrease in IgG-producing cells. Consistently, an Ig secretion defect was seen in vitro as well as in vivo for class-switched Ig.
We observed various transcriptional changes induced by BET inhibition. IgH constant gene transcripts specific for the pre-CSR stage globally increased while switched transcripts decreased. Noticeably, expression and AID loading on S regions also clearly decreased, suggesting that not only the repair step, but also the initiation of DNA lesions was affected, contributing to the global reduction of CSR junctions observed in DNA recovered from activated B cells. The position of breaks within S regions also became more closely dependent on the proximity of AID consensus motifs and these anomalies altogether indicate a specific B-cell maturation defect targeting the CSR step.
To better understand the overall impact of BET inhibition on the immune response, and on T cells, we immunised animals with 1W1K-OVA, which simultaneously received 50 mg kg À1 per day JQ1. We noticed a decreased generation of OVA-specific IgG-expressing cells by FACS as well as by ELISPOT in JQ1-treated animals, while IgMproducing cells were unaffected or increased. These changes were in agreement with the decreased production of OVA-specific IgG and IgE.
Furthermore, a strong decrease in 1W1K-specific Tfh cells was noticed in the draining lymphoid organs after immunisation. JQ1 treatment, however, had an immune-stimulatory effect on T lymphocytes, favoring Th2 cell polarisation, with increased GATA3 expression and increased production of type 2 cytokines. These data suggest that the in vivo decrease in CSR following BET inhibition affects not only strong B-intrinsic modifications (as observed in vitro, with decreased accessibility of S regions because of decreased AID and 3 0 RR activity as well as decreased repair), but also exerts a complex modulation of the interactions with Tfh and Th2 cells, which are both quantitatively and qualitatively modified during immune responses.
To evaluate the potential interest of these Band T-cell changes in an Ab-dependent immune disorder, we followed specific Ag responses in mice in the ovalbumin sensitisation protocol classically used to induce allergic airway inflammation. Importantly, objective biological markers of allergic asthma triggered by ovalbumin in immunised mice were significantly reduced in mice receiving JQ1, with a global efficiency close to that of steroids, and a notable reduction in all inflammatory cell counts in BALF. Importantly, markers from the T-cell branch of inflammation were significantly affected in parallel, but strikingly showed increased Th2 polarisation with high IL-4 and IL-5 levels in BALF and lungs. Finally, JQ1 had no effect on mucus production but increased airway resistance in the allergic asthma model.
Despite a delayed disease onset, clinical signs did not improve when we explored another model of B-cell-dependent immune disorders: encephalomyelitis induced by rhMOG immunisation in C57Bl/6 mice. It was previously shown in this model that MOG-specific Abs are not pathogenic, but that the disease needed antigen presentation by B cells, so that B celldeficient mice are resistant to rhMOG-EAE. 18 In the light of our observation in EAE-induced mice, the impact of JQ1 on Tfh cell development and on specific cognate B cells was, however, not sufficient to significantly prevent the development of this B/T interaction-dependent disease.
Altogether, both in vivo and in vitro experiments concur to demonstrate that BET proteins contribute to CSR in B cells and can be efficiently targeted by BET inhibitors. The in vitro CSR defect imposed by JQ1 was notably related to decrease functional interactions of AID with S regions and with the 3 0 RR. It translated in vivo by lowered numbers of cells undergoing CSR after immunisation, defective class-switched Ig secretion and reduced development of some (but not all) biological allergy parameters.
The present study with JQ1 underlines the involvement of BET proteins at multiple levels in immune cells during humoral reactions, affecting not only B, but also T cells, which showed strongly reduced Tfh cell generation and increased Th2 cell polarisation. Since the humoral immune response relies on optimal interactions between activated GC B cells and Tfh cells, the opposing effects of JQ1 on both lineages result in a poorly predictable outcome where the apparently decreased Ab reaction is associated with identical or even more severe functional alterations. While BET proteins are often seen just as gene expression activators, it is also important to remember their dual role, since they also interact with repressive factors and BET inhibition reactivates expression of some SWI/SNF-repressed genes. 19 BRD4 can also directly repress some gene promoters, such as the autophagy factor LC3 (Atg8) which is activated by the JQ1 BET inhibitor. 20 In conclusion, BET inhibitors are clearly strong immunomodulators with a broad spectrum of activity and their use in patients affected with Bcell-dependent/Th2-dependent immune disorders might be hazardous. While this study required artificial Ag such as 1W1K-OVA to precisely monitor Ag-specific cells, our observations deserve to be validated for diverse natural antigens or autoantigens. Of note, in another unrelated model, a strong increase in Th2 cytokines was also noticed after BET inhibition in a spinal cord injury model 21 illustrating the complex and eventually de-repressing effects of these inhibitors. While BET inhibitors might be of interest for Th17dependent pathologies, their usage as class switch inhibitors will clearly have to await more specific molecules. Strategies exclusively targeting B cells and production of class-switched immunoglobulins might notably be of interest in IgG and or IgE-dependent immuno-allergic conditions. In addition to our data with T cells, analysis of previous reports indicates that in different settings, BET inhibition was described in some instances as rather immunosuppressive notably by inhibiting IL-2 responses, but shown in other instances to increase either IFN-c or IL-4 production by Th2 cells, while eventually inhibiting FoxP3 expression and Treg function, thereby removing an important brake on Th2 responses. [22][23][24][25] Altogether and despite the strong repressing impact on intrinsic B-cell responses, the conflicting effects on T-cell regulation show that BET inhibition should be considered to have a potential hazardous impact on immunity and immunopathology, thus calling for extensive immune monitoring of current clinical trials in cancer patients.

Mice
Our research received ethical agreement APAFIS no. 16152-2018071717143183v2. Six-to-12-week-old BALB/c or C57BL/6 (depending upon the disease model studied) mice (maintained at 21-23°C with a 12-h light/dark cycle) were used for our experiments.

ELISA and ELISPOT
ELISA was performed on sera and supernatants from in vitro stimulated and JQ1 (10, 20 and 40 nM) treated and untreated sorted primary B cells for the detection of IgM, IgG1 and IgE secretion. Plates were coated with mAbs specific for IgM, IgG1 or IgE (Southern Biotech, Birmingham, AL, USA). Sera or supernatants were added and incubated for 2 h at 37°C. After washing, alkaline phosphatase (AP)-conjugates of goat anti-mouse IgM, IgG1 or IgE (Southern Biotech) were incubated 1 h at 37°C. After washing and addition of AP substrate, absorbance was measured at 405 nm.
Anti-OVA-specific Abs produced in vivo after immunisation were evaluated in sera from JQ1-treated or untreated mice (50 mg kg À1 ). ELISA plates were coated with 10 lg mL À1 ovalbumin. Sera were then incubated for 2 h at 37°C, and plates were treated as above. Anti-OVA IgG and IgM were detected with horseradish peroxidase (HRP)-conjugated anti-mouse IgG or IgM (Southern Biotech).
For ELISPOTs, specific IgG and IgM anti-OVA Ab-secreting cells were quantified using splenocytes from mice sacrificed 9 days after immunisation. Splenocytes were seeded in duplicate at a density starting at 5 9 10 5 per well, followed by twofold serial dilutions in 96-well MultiScreen HTS plates (Millipore, Burlington, MA, USA) coated with 200 lg per well OVA. Cells were incubated overnight at 37°C and then removed by washing with PBS/0.01% Tween. Plates were incubated for 1 h at 37°C with 1 lg per well AP-coupled anti-IgG or anti-IgM. After washing, AP substrate was added. Plates were washed and dried, and images were analysed for spot numbers using Nis-Ar software (Nikon, Tokyo, Japan).

Flow cytometry
Class switch recombination was assessed using the following Abs: anti-mouse IgM-APC (Clone II/

Transcription analysis (RT-qPCR)
After 4 days of in vitro stimulation with or without JQ1, B cells were collected and RNA was extracted to evaluate post-switch and germline transcripts. RNA was prepared using standard techniques. cDNA was synthesised using the High Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific, Waltham, MA, USA). Il-Cl, and Iɣ1-Cɣ1 germline transcripts and Il-Cɣ1, Il-Cɛ post-switch transcripts were quantified. Quantitative PCR was performed using power SYBR green (Applied Biosystems, Foster City, CA, USA) and specific oligonucleotides listed in Supplementary table 1.
RNA extracted from lung tissue was also evaluated by RT-qPCR/SYBR green analysis in order to measure Muc5ac gene expression, as a marker of mucus production. Gene expression was normalised to the expression of mouse Hprt1. Primers used to assess Muc5ac gene expression are listed in Supplementary table 1.

Amplification of Sl/Sc junctions and Ion torrent next-generation sequencing
DNA from LPS + IL-4-stimulated cells (treated with or without 40 nM JQ1) was extracted using the classical phenol/chloroform protocol. Sl/Sc junctions were amplified in triplicate by nested PCR with 100 ng DNA (Phusion HF polymerase; New England Biolabs, Ipswich, MA, USA) using the following primers listed in Supplementary table 1. Each library was prepared using 200 ng PCR2 product. Barcoded libraries with 200-pb read lengths were prepared using Ion Xpress Plus Fragment Library Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Each barcoded library was mixed in equal amounts and diluted to 100 pM. Libraries were run on an Ion PI v3 chip on the Ion Proton sequencer (Life Technologies-Thermo Fisher Scientific). Data analysis was performed using CSReport. 26 ChIP ChIP experiments were performed on LPS + IL-4-stimulated CD43 À spleen cells incubated with or without 40 nM JQ1. Briefly, 15 9 10 6 B cells were cross-linked at room temperature for 15 min in 15 mL PBS with 1% formaldehyde. The reaction was quenched with 2.125 M glycine. After lysis, chromatin was sonicated to 0.5-1 kb using a Vibracell 75043 (Thermo Fisher Scientific). Following dilution in ChIP buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris/ HCl, pH 8.1 and 167 mM NaCl), chromatin was precleared by rotating for 2 h at 4°C with 50 mL 50% protein A/G slurry (0.2 mg mL À1 sheared salmon sperm DNA, 0.5 mg mL À1 BSA and 50% protein A/G; Sigma, St Louis, MO, USA). Cell equivalents (1 9 10 6 ) were saved as input, and the remaining cell equivalents were incubated overnight with anti-AID rabbit polyclonal Abs (kindly provided by P. Gearhart) or control Abs. Immunoprecipitation was with protein A/G. Cross-linking was reversed by overnight incubation (70°C) in In vivo Th immune response in C57/Bl6 mice C57BL/6 (CD45.2 + ) mice were purchased from Janvier (Genest Saint Isle, France). Females (8-12 weeks) were used for experimental procedures. Aluminium hydroxide adjuvant (Alum) was from Thermo Scientific. The T-cell Ag 1W1K (EAWGALANKAVDKA) covalently coupled to OVA was bought from Genecust (Boynes, France). Mice were immunised subcutaneously (s.c.) at the tail base and intraperitoneally (i.p.) with 40 µg 1W1K-OVA in Alum. Mice were then treated daily i.p. with vehicle or JQ1 (50 mg kg À1 ) for 9 days after which draining inguinal and periaortic LN as well as spleens were collected. The extent and nature of the Ag-specific Th cell response were then evaluated by flow cytometry using fluorescent 1W1K-IA b tetramers (obtained from the NIH Tetramer Core Facility, Emory University, Atlanta, GA, USA).

OVA-induced allergic asthma model
BALB/c female mice at 8 weeks of age were immunised i.p. on days 0, 7 and 14 with 20 lg grade V ovalbumin (Sigma) emulsified in 2 mg Alum gel in a total volume of 200 lL. For control mice, 200 lL saline was injected. Mice were challenged at days 21-24 with 10 lg OVA by intra-tracheal (i.t.) administration to provoke allergic asthma with analysis of the allergic response at day 25. Control mice received saline. To compare anti-allergic therapies, mice immunised with OVA received daily either budesonide (3 mg kg À1 , i.n.), JQ1 (50 mg kg À1 , i.p) or vehicle (hydroxypropyl-bcyclodextrin (100 mg mL À1 ), from day À1 until day 24 and 1 h before immunisation or challenge.

Bronchoalveolar lavage and differential cell counts
Bronchoalveolar lavage (BAL) was performed by washing lungs four times with 0.5 mL saline at room temperature. After centrifugation at 400 g for 10 min, supernatants from the first lavage (cell-free BAL fluid) were stored at À80°C for cytokine analysis. Cells were diluted with Turk's solution and counted and 2 9 10 5 cells were centrifuged onto microscope slides (cytospin at 113 g for 10 min, at RT). Airdried preparations were fixed and stained with Diff-Quik (#130832, Medion Diagnostics AG, Merz & Dade, Germany). A total of 200 cells were observed by oil immersion light microscopy to determine the relative percentage of each cell type and absolute number of the differential cell count.

Airway resistance measurement
Mice were anaesthetised by intra-peritoneal injection of ketamine (100 mg kg À1 ; Merial, Lyon, France) and xylasine (10 mg kg À1 ; Bayer, Leverkusen, Germany), paralysed using D-tubocurarine (0.125%, Sigma) and intubated with an 18-gauge catheter. Respiratory frequency was set at 140 breaths per min with a tidal volume of 0.2 mL and a positive end-expiratory pressure of 2 mL H 2 O. Increasing concentrations of aerosolised methacholine (0, 25, 50, 100 and 200 mg mL À1 , Sigma) were administered. Resistance was recorded with a plethysmograph (Buxco, London, UK). Baseline resistance was restored before administering the subsequent doses of methacholine.

Lung homogenisation for evaluation of cytokines and EPO activity
After BAL, the entire lung was perfused with isotonic solution through the right heart ventricle to flush the vascular content and lungs were weighed and frozen at À20°C until use. The right lung was homogenised in PBS containing anti-protease cocktail (# P8340; Sigma) and centrifuged, and the supernatant was aliquoted and stored at À20°C until analysis. EPO activity was determined in lung supernatants by colorimetric assay. Following centrifugation, 100 lL of supernatants was placed in a plate with 50 lL substrate solution, corresponding to 11 mL Tris HCl + 200 mM OPD pellets + 100 lL 30% H 2 O 2 ). After 1-h incubation at 37°C in a shaker, EPO activity was determined as 490 nm absorbance against medium. The right lung postcaval lobe was removed and placed in RNAlater (Thermo Fisher Scientific) for 24 h and snap-frozen for further analysis. The left lung was removed and preserved in 4% formaldehyde for histopathological analysis.

Lung histology and mucus production
After BAL and lung perfusion, the left lung was fixed in 4% buffered formaldehyde (#15225582, Thermo Fisher Scientific) for at least 24 h for standard microscopic analysis. Sections (3-lm) were stained with haematoxylin and eosin (H&E) and periodic acid-Schiff (PAS). Peribronchial infiltrates were assessed by a semi-quantitative score (0-5) by two independent observers on three independent lung sections, and with 10 different fields analysed for each section. Inflammatory cells including eosinophil infiltration were assessed. A quantitative evaluation of mucus production in lungs was obtained in parallel by assessing Muc5ac gene expression using RT-qPCR (as described above) on RNA from lung tissue.

Cytokine measurement
IL-4 and IL-5 concentrations in BALF and lung homogenates were determined by Luminex immunoassay (Millipore) by using MagPix system (Bio-Rad, Hercules, CA, USA) according to the manufacturer's instructions.