On-target and off-target effects of novel orthosteric and allosteric activators of GPR84

Many members of the G protein-coupled receptor family, including examples with clear therapeutic potential, remain poorly characterised. This often reflects limited availability of suitable tool ligands with which to interrogate receptor function. In the case of GPR84, currently a target for the treatment of idiopathic pulmonary fibrosis, recent times have seen the description of novel orthosteric and allosteric agonists. Using 2-(hexylthiol)pyrimidine-4,6 diol (2-HTP) and di(5,7-difluoro-1H-indole-3-yl)methane (PSB-16671) as exemplars of each class, in cell lines transfected to express either human or mouse GPR84, both ligands acted as effective on-target activators and with high co-operativity in their interactions. This was also the case in lipopolysaccharide-activated model human and mouse immune cell lines. However in mouse bone-marrow-derived neutrophils, where expression of GPR84 is particularly high, the capacity of PSB-16671 but not of 2-HTP to promote G protein activation was predominantly off-target because it was not blocked by an antagonist of GPR84 and was preserved in neutrophils isolated from GPR84 deficient mice. These results illustrate the challenges of attempting to study and define functions of poorly characterised receptors using ligands that have been developed via medicinal chemistry programmes, but where assessed activity has been limited largely to the initially identified target.

Herein, we set out to establish the detailed pharmacology and usefulness of 2-HTP and PSB-16671 to characterise and define the biology and function of GPR84 in both human and mouse. PSB-16671 produced activation and co-operativity effects that were mediated directly by GPR84 both in cells transfected to express either human or mouse GPR84, and in human and mouse-derived immune cell models. However the ability of PSB-16671 to activate G-proteins in mouse bone marrow-derived neutrophils was instead largely 'off-target' and not mediated by GPR84. These conclusions are based on the inability of a GPR84 antagonist that blocked effects of PSB-16671 at mouse GPR84 in both transfected cells and in the macrophage-like RAW264.7 cell line to do so in bone marrow-derived neutrophils, and because the activity of PSB-16671 is preserved in neutrophils from GPR84 knock-out mice. These studies indicate the challenges and dangers of simply transferring outcomes from medicinal chemistry programmes into physiological studies and outcomes for poorly characterised receptors where little is known about the mode of binding of ligands that are described to have potency and affinity at that specific target. Doing so may result in incorrect assignment of functions to such a receptor.

Results
Human and mouse GPR84 display similar responsiveness to both orthosteric and allosteric agonists. 2-(hexylthiol)pyrimidine-4,6 diol (2-HTP) is a potent agonist of human GPR84 11,13,14 . This ligand has been shown to share an overlapping binding site on human GPR84 with MCFAs such as decanoic acid (C10) that are considered the likely endogenous ligands of the receptor. This is based on effects of both 2-HTP and C10 being eliminated following mutation of arginine172 in the receptor 11 and because, when added concurrently at submaximal concentrations, these two ligands produce additive, rather than co-operative, responses 11,12 . By contrast, 3,3′diindoylmethane (DIM) and related activators of GPR84 bind to a different site as their function is unaffected by mutation of arginine172 11 and at sub-maximal concentrations they produce co-operative effects with either MCFAs or 2-HTP 11 . Recently Pillaiyar et al. 16 identified di(5,7-difluoro-1H-indole-3-yl)methane (PSB-16671) as a more potent analogue of DIM at human GPR84.
GPR84 couples selectively to members of the pertussis toxin-sensitive G i -G protein family 7,8,11 . Activation of these G proteins is particularly amenable to analysis by measuring regulation of binding of guanine nucleotides and their analogues 17 . Initially, therefore, we examined the ability of 2-HTP to promote the binding of [ 35 S]GTPγS in membranes of Flp-In T-REx 293 cells that had been induced to express a human GPR84-Gα i2 fusion protein 11 . 2-HTP promoted binding of [ 35 S]GTPγS in a concentration-dependent manner with pEC 50 = 7.54 ± 0.06 (Fig. 1a). Addition of a range of concentrations of PSB-16671 to cell membranes expressing human GPR84-Gα i2 showed that this ligand can also activate human GPR84 with, in this case, pEC 50 = 6.28 ± 0.05 (Fig. 1b). Equivalent studies using DIM confirmed the higher potency of PSB-16671 as originally reported by 16 (Fig. 1b), but noticeably PSB-16671 also displayed markedly greater efficacy, i.e. greater effect at maximally effective concentrations, than DIM (Fig. 1b). Importantly, no significant stimulation of binding of [ 35 S]GTPγS was produced by either 2-HTP or PSB-16671 when these were added to membranes of cells harbouring human GPR84-Gα i2 but in which expression of the construct had not been induced (Fig. 1a,b). This demonstrates that both ligands act in an 'on-target' manner to directly activate human GPR84. When sub-maximal concentrations of PSB-16671 were co-added with 2-HTP to membranes expressing human GPR84-Gα i2 this resulted in enhanced potency of 2-HTP (Fig. 1c), and fitting of such data to an operational model of allosterism 18 indicated that the extent of co-operativity (α) was greater than 200 fold (log α 2.41 ± 0.17) (Table 1), whilst effects on ligand efficacy were marginal (log β was close to zero) ( Table 1). Such data sets also allowed estimation of the binding affinity of 2-HTP at human GPR84 (pKa = 7.0 ± 0.2) and of PSB-16671 (pKa = 6.02 ± 0.13). Reciprocal experiments in which the effects of increasing concentrations of 2-HTP on the potency of PSB-16671 were assessed (Fig. 1d) resulted in similar conclusions and highly similar estimates of the binding affinities of the two ligands ( Table 1).
As mouse models of GPR84 function and various mouse-derived immune cell lines are central to studies on this receptor we cloned mouse GPR84 from RAW264.7 cells. Following construction into an equivalent Gα i2 fusion protein, stable integration into Flp-In T-REx 293 cells, and doxycycline-induced expression of the construct, membrane preparations were again tested for response to, and co-operativity between, 2-HTP and PSB-16671 (Fig. 2). Outcomes were very similar to those observed when using the human orthologue of GPR84, pEC 50 2-HTP: 7.52 ± 0.11, PSB-16671: 6.53 ± 0.07, DIM: 5.37 ± 0.08 (Fig. 2, Table 2), and once again DIM acted as a partial agonist when compared to PSB-16671 (Fig. 2).
LPS promotes upregulation of the expression and function of GPR84 protein as well as corresponding mRNA. GPR84 is often described as a pro-inflammatory receptor 8 and is expressed by a range of white cell types and cell lines 7,8,10 . With underpinning pharmacological characterisation in place we turned to explore potential effects of these ligands in human THP-1 monocytes. These cells are known to upregulate expression of mRNA encoding GPR84 in response to activation of toll-like receptor 4 (TLR4) by lipopolysaccharide (LPS) 7,19 . Although this has been suggested to imply a pivotal role of GPR84 in monocyte/macrophage activation and host immune response, nothing is known about how this may reflect or be related to changes in  (Fig. 3b(i)). This 3.5 fold increase in receptor density compared to a more extensive 10.0 ± 2.5 fold (mean ± S.E.M., n = 6) increase in GPR84 mRNA levels at this time point as assessed using q-RT-PCR ( Fig. 3b(ii)). To assess if upregulation of expression of GPR84 in LPS-stimulated THP-1 cells had functional consequences we measured the ability of both 2-HTP and PSB-16671 to promote binding of [ 35 S] GTPγS in membranes from these cells. A notable feature in each separate preparation studied was that treatment with LPS resulted in a higher level of basal, ligand-independent binding of [ 35 S]GTPγS (Fig. 3c). Moreover, for 2-HTP, the extent of stimulation of binding of [ 35 S]GTPγS was markedly higher in membranes derived from LPS-pretreated THP-1 cells than those that were not treated with LPS ( Fig. 3d), consistent with the higher measured expression of GPR84 receptor protein. Despite this, no significant increase in potency of 2-HTP was observed (pEC 50 vehicle = 7.89 ± 0.21, LPS treated = 7.79 ± 0.08) (Fig. 3e), suggesting that even after LPS induced up-regulation no substantial receptor reserve was generated for GPR84. Similar results were observed for PSB-16671 (pEC 50 vehicle = 6.82 ± 0.12, LPS treated = 6.94 ± 0.07) (Fig. 3f,g). Although these outcomes were consistent with both 2-HTP and PSB-16671 mediating effects via GPR84 in these cells, both ligands are experimental tool compounds with limited broader current characterisation. To further determine if the effects of these ligands  Table 1 for quantitative analysis.  (Fig. 4a,b). The macrophage cell line RAW 264.7 is also a popular model to study immune cell-controlled release of pro-inflammatory mediators. It is, however, a murine-derived cell line, and differences in production of cytokines such as TNFα and IL-1β have been noted compared to human leukocytes 20 . Expression of mRNA encoding mouse GPR84 was clearly detected in these cells using RT-PCR (Fig. 5a) and indeed, as described earlier, we cloned mouse GPR84 from these cells. As expected from the similar calculated binding affinities and G protein activation characteristics of 2-HTP and PSB-16671 at the cloned human and mouse orthologues of GPR84 (Tables 1 and 2) both ligands were able to stimulate binding of [ 35 S]GTPγS in membranes of LPS-treated RAW 264.7 cells (pEC 50 2-HTP = 7.58 ± 0.08, PSB-16671 = 6.75 ± 0.10) (Fig. 5b) with similar potency as noted for the human GPR84 expressing systems. Moreover, marked co-operative effects were also generated by co-addition of the two ligands (Fig. 5c,d) and estimation of co-operativity and affinity values from these studies were also very similar to those for mouse GPR84 expressed in the Flp-In T-REx 293 cells (Table 2).  Table 2 Table 2. Binding affinity and co-operativity of ligands as activators of mouse GPR84. a Agonist is the compound used to generate concentration-response curve. b Modulator is the compound used in defined concentrations.  The non-competitive GPR84 antagonist compound 107 has lower affinity at mouse than at human GPR84. Compound 107 was able to fully block stimulation produced by both 2-HTP and PSB-16671 in membranes from RAW 264.7 cells in a concentration-dependent manner (pIC 50 versus 2-HTP = 6.50 ± 0.06 and versus PSB-16671 = 6.84 ± 0.09) (Fig. 6a,b). It was also noticeable however that the potency of compound 107 was substantially lower in membranes from RAW 264.7 cells than in equivalent preparations from THP-1 cells. This suggested that compound 107 might have lower affinity at mouse GPR84 compared to the human orthologue. Indeed, when we compared directly the potency of compound 107 to inhibit effects of either 2-HTP or PSB-16671 at cloned human and mouse GPR84 in membranes from Flp-In T-REx 293 cells expressing either human or mouse GPR84-Gα i2 fusion proteins compound 107 was between 20 and 75 fold less potent at the   Table 2 for quantitative analysis.
In mouse bone marrow-derived neutrophils G i activation by PSB-16671 is largely independent of GPR84. Despite this limitation we moved forward to assess function of GPR84 in mouse bone marrow-derived neutrophils because the Immunological Genome Project (Immgen) database (http://www. immgen.org/databrowser/index.html) indicates expression of GPR84 mRNA to be highest in these cells. 2-HTP again acted to stimulate binding of [ 35 S]GTPγS in membranes isolated from such cells (Fig. 7a). The potency of 2-HTP (pEC 50 = 7.89 ± 0.12) (Fig. 7a(i)) was similar to those observed in RAW 264.7 cell membrane preparations and this was potentially consistent with its major target in mouse bone marrow-derived neutrophils also being GPR84. However, the potency of PSB-16671 (pEC 50 = 5.54 ± 0.09) (Fig. 7a(i)) was some 10 fold lower than in RAW 264.7 cell membrane preparations, suggesting that this was not an effect transduced by GPR84. Moreover, although PSB-16671 was not more efficacious than 2-HTP in membranes of either cells transfected to express the mouse GPR84-Gα i2 fusion protein (Fig. 2) or in mouse RAW 264.7 cells (Fig. 5), now in membranes from mouse bone marrow-derived neutrophils stimulation of binding of [ 35 S]GTPγS by PSB-16671 was much greater than produced by a maximally effective concentration of 2-HTP ( Fig. 7a(i)). Although the effect of 2-HTP was completely blocked by compound 107 (Fig. 7a(ii)) the effect of PSB-16671 by contrast was only slightly inhibited by compound 107 (Fig. 7a(ii)). Interestingly, although the parent compound of PSB-16671, DIM, was only able to produce a small fraction of the response generated by PSB-16671, this was instead fully blocked by co-addition of compound 107 (Fig. 7a(ii)). This combination of observations suggested that a large proportion of the effect of PSB-16671 in mouse bone marrow-derived neutrophils must be an 'off-target' effect that is not mediated by GPR84. To test this directly we also isolated bone marrow-derived neutrophils from GPR84 deficient mice (Fig. 7b) and generated membrane preparations. Here, although 2-HTP was unable to stimulate binding of [ 35 S]GTPγS ( Fig. 7c(i)), PSB-16671 maintained capacity to promote a large increase in [ 35 S]GTPγS binding (Fig. 7c(i)) and this was unaffected by the presence of compound 107 (Fig. 7c(i)). In contrast, the effect of DIM observed in wild type bone marrow-derived neutrophils was lacking in those derived from the GPR84 knock-out animals (Fig. 7c(i)), consistent with  .44 ± 0.14) highly similar to preparations from wild type mice (Fig. 7c(ii)).

Effects of PSB-16671 in human neutrophils are mediated predominantly by GPR84. Based on
these observations we also assessed if PSB-16671 might display non-GPR84 mediated effects in human neutrophils. We isolated such cells, generated membrane preparations and again performed a series of [ 35 S]GTPγS binding studies. Here both 2-HTP and PSB-16671 were able to promote [ 35 S]GTPγS binding, and to a similar extent (Fig. 7d). Moreover, for both the orthosteric and the allosteric agonist their effects were prevented by co-addition of compound 107 (Fig. 7d), consistent with being mediated via GPR84. Finally, although PSB-16671 generated a substantial GPR84-independent increase in binding of [ 35 S]GTPγS in mouse bone marrow-derived neutrophils there was potentially a component of this that was sensitive to blockade by the GPR84 antagonist compound 107 ( Fig. 7a(ii)). We thus explored if activation of GPR84 promoted migration of these cells. Interestingly, both 2-HTP and PSB-16671 did so (Fig. 8) and to an extent similar to that produced by the chemotactic peptide fMLP (Fig. 8).
In this assay compound 107 fully blocked the effect of 2-HTP. By contrast, although compound 107 largely prevented migration in response to PSB-16671 a proportion was not blocked by compound 107 (Fig. 8). This suggests that in mouse bone marrow-derived neutrophils activation of GPR84 does indeed promote cell migration. However, there is also a component of the effect of PSB-16671 on cell migration that is GPR84-independent.
The effect of PSB-16671 in mouse bone marrow-derived neutrophils is not via GPR35. Because GPR84 is phylogenetically related to the receptor GPR35 we assessed whether either PSB-16671 or 2-HTP could also activate mouse GPR35. To do so we used a β-arrestin-2 recruitment assay as this has been the mostly widely used approach to explore the pharmacology of GPR35 [21][22][23] . It has previously been noted that PSB-16671 is unable to activate human GPR35 16 . We confimed this and that 2-HTP was also unable to activate human GPR35, whilst the GPR35 agonist lodoxamide 21 did so (Fig. 9a), and with high potency (pEC 50 = 8.33 ± 0.03). Similarly, neither PSB-16671 nor 2-HTP was able to activate mouse GPR35 (Fig. 9b). However, although lodoxamide was able to activate mouse GPR35, it did so with some 500 fold lower potency (pEC 50 = 5.67 ± 0.01) than at the human orthologue (Fig. 9b).

Conclusions
Such outcomes illustrate effectively the challenges in efforts to extrapolate the use of ligands that have received limited characterisation during medicinal chemistry structure-activity studies 16 into broader cell and tissue studies and, potentially, between species orthologues.

Discussion
Studies on poorly characterised GPCRs, including those activated by fatty acids of varying chain length, are frequently restricted by the limited tool box of pharmacological ligands that are available 24 . This has, for example, greatly hindered efforts to better understand roles of free fatty acid receptor 3 24 . Equally, roles of other GPCRs that have few available pharmcologial tool compounds, such as members of the adhesion class of receptors, are poorly understood despite many being potentially interesting targets in areas such as cancer 25 . Even within the numerically predominant, class A family of receptors that are related to rhodopsin a substantial number remain as 'orphans' 2 or have been the subject of a limited number of studies. Whilst such areas of ligand deficit can offer many opportunities for medicinal chemists to produce improved ligands when initial chemical starting points become available, a major challenge is then to understand ligand selectivity and potential 'off-target' effects before such compounds can be used with confidence in physiological settings. In recent studies designed to develop analogues of 3-3′-diindoylmethane, Pillaiyar et al. 16 , showed that that the higher potency ligand di(5,7-difluoro-1H-indole-3-yl)methane (PSB-16671) was selective for GPR84 in that it did not activate the long chain free fatty acid receptors FFA1 and FFA4 or the phylogenetically related receptor GPR35. Moreover, although DIM is known to have effects at the arylhydrocarbon receptor, PSB-16671 lacked activity at this target 16 . It is clearly impractical to randomly investigate all possible 'off-target' effects of a small molecule ligand, but we were alerted to the possibility that PSB-16671 might produce certain non-GPR84-mediated effects in studies using mouse bone marrow-derived neutrophils. In membranes derived from both Flp-In T-REx 293 cells transfected to express a form of mouse GPR84 and, more importantly, in murine RAW264.7 cells, PSB-16671 and the orthosteric GPR84 agonist 2-HTP appeared to have similar efficacy, because at maximally effective concentrations these two compounds increased levels of [ 35 S]GTPγS binding to similar levels. However, in mouse bone marrow-derived neutrophils the stimulation produced by PSB-16671 was markedly higher than produced by 2-HTP. To examine this more directly we assessed whether effects of both 2-HTP and PSB-16671 would be blocked by the non-competitive GPR84 antagonist compound 107, which has previously been shown to antagonise effects of both the MCFA decanoic acid and of DIM at human GPR84 11 . However, although compound 107 did block the effect of 2-HTP almost completely at 10 μM it only marginally inhibited the effect of PSB-16671 in membranes of mouse bone marrow-derived neutrophils. In previous studies we have shown that compound 107 has high affinity at human GPR84 but this has not previously been assessed at the mouse orthologue. When we compared directly the potency of compound 107 to antagonise effects of either 2-HTP or PSB-16671 at human versus mouse GPR84 in cells in which the receptor was expressed as the corresponding Gα i2 fusion protein, it was clear that although compound 107 is, indeed, an effective antagonist of the mouse receptor, it is substantially less potent than at the human orthologue. Moreover, although some 20 fold less potent in blocking effects mediated by the orthosteric agonist 2-HTP at the mouse receptor this was even more extensive when considering PSB-16671 which, as an allosteric agonist at both human and mouse GPR84, binds at a different site than 2-HTP. These were important observations and require consideration because antagonists at certain poorly characterised receptors can display very marked selectivity between species orthologues. For example, CID-2745687 (methyl-5-[(tert butylcarbamothioylhydrazinylidene)methyl]-1-(2,4-difluorophenyl)pyrazole-4-carboxylate) is described in both academic publications and supplier catalogues as an 'antagonist of GPR35′. However, although it is indeed a high affinity antagonist of human GPR35 24 it has been shown to lack activity at both the rat and mouse orthologues 23 . Similarly, although the clinically trialed FFA2 receptor antagonist GLPG0974 is a high affinity antagonist of the human receptor, this compound also lacks significant activity at rodent orthologues 26 .
To further validate the non-GPR84 mediated effect of PSB-16671 in mouse bone marrow-derived neutrophils we isolated such cells and generated membranes from a GPR84 knock-out mouse line 27 . Here, although 2-HTP lacked activity, PSB-16671 produced stimulation of [ 35 S]GTPγS binding almost as effectively as in samples from wild type mice and, once more, this effect was not blocked by compound 107. It was also notable that the potency of PSB-16671 in membranes of bone marrow-derived neutrophils from both wild type and GPR84 deficient mice was almost 10 fold lower than in membranes from RAW264.7 cells, suggesting that it has a further molecular target at which it displays somewhat lower affinity than at mouse GPR84. This illustrates the importance of performing careful pharmacological studies in which the effects of poorly characterised ligands are measured across a range of concentrations, and not simply at a single high concentration. The non-GPR84 molecular target of PSB-16671 in mouse bone marrow-derived neutrophils remains unclear. As the [ 35 S]GTPγS binding assay reports on activation of heterotrimeric G proteins, and predominantly on members of the G i -subfamily 17 , the site is likely to be a highly expressed G i -coupled GPCR. Database information, e.g. the Immgen database (http://www.immgen. org/databrowser/index.html) indicates that such cells express a wide range of GPCRs, with various chemokine receptors, e.g. CXCR4, CXCR2 and CCR1, as well as other receptors associated with chemotaxis e.g. formyl peptide receptor 1 (FPR1) and the complement component 5a receptor 1 (C5a 1 ), being particularly highly expressed. However, a number of orphan GPCRs including GPR27, GPR64, GPR107 and GPR141 are also highly expressed at the mRNA level. In time it will be interesting to explore whether PSB-16671 is able to activate one of more of these receptors, either as an orthosteric or an allosteric agonist. Similarly, efforts have been made to catalogue GPCR expression in RAW264.7 cells 28 and phorbol ester-stimulated THP-1 cells 29 and assessment of such profiles may help define the range of potential targets for activation by PSB-16671 that should be prioritised.
A further key feature of these studies was the ability to measure directly levels of expression of GPR84 in human tissues and human-derived cells. Although it is well appreciated that treatment of cell lines such as THP-1 monocytes with TLR4 receptor activators such as LPS results in a time-dependent upregulation of mRNA encoding GPR84 7,10 it has not previously been possible to assess how this might relate to altered levels of expression of the receptor itself. Herein, by using the specific binding of [ 3 H]9543 we were able to do so. Interestingly, although levels of GPR84 receptor protein were upregulated some 3.5 fold and this resulted in higher G protein activation in response to maximally effective concentrations of either 2-HTP or PSB-16671, such treatment did not enhance the potency of either of these ligands. One potential means to improve biological function of an agonist ligand is to increase levels of its receptor in response to a challenge or stress such that a maintained concentration of the endogenous agonist now produces greater effect. However, these results indicate that the LPS-induced upregulation did not generate a marked GPR84 receptor reserve in THP-1 cells. As such, whilst the effect of LPS is likely to allow greater levels of maximal signal from endogenous activators of the receptor, it will not result in improved sensitivity of the cell to such activators.
Recently, clinical trials with ligands that have affinity at GPR84 have explored the potential of blocking this receptor for the treatment of both ulcerative colitis 5 and idiopathic pulmonary fibrosis 4,30 . Furthermore, suggestions that activation of GPR84 might have beneficial effects in atherosclerotic conditions 6 have begun to provide greater focus on this receptor. In recent times a number of distinct compounds with agonist 6,12,13,31 or positive allosteric modulator (PAM)-agonist 16 activation of GPR84 have been decribed and substantial structure-activity relationships of some of these compounds have been illustrated 6,12,13,16,31 . However, before using compounds from such series to claim direct evidence on further roles for GPR84 and its potential therapeutic indications, a better understanding of potential off-target effects are clearly required.

Methods
Materials. PSB-16671 16  GPR84 knock-out mice. GPR84 (GenBank accession number NM_030720, Ensembl identification number ENSMUSG00000063234) KO mice 27 were provided by Deltagen under a GlaxoSmithKline license agreement. Targeting replaced 257 bp of coding sequence with an IRES?LacZ?poly(A) expression cassette and a positive selection cassette that contains the neomycin phosphotransferase gene driven by the PGK promoter (Neo). The insertion of the LacZ IRESLacZ introduces a premature translational stop signal that deletes the first three predicted transmembrane domain sequences. Disruption of the GPR84 gene was confirmed using PCR and standard agarose gel electrophoresis using the following primers: gene specific (endogenous), 5′-CAGATGCCAACTTCTCCTGCTA-3′; gene specific (targeted, endogenous), 5′-GAGGTAGCGTCCTAGAGCAAT-3′; and Neo (targeted), 5′-GCCTCTGTTCCACATACA-3′.
Mice forming the initial breeding pairs were supplied by GlaxoSmithKline, which consisted of heterozygous (HET) F1 offspring from WT and KO breeding. HET pairs were bred in-house from 8 weeks old to produce litters of mixed genotypes according to Mendelian genetics.
Animal maintenance. Both wild type C57BL/6 (Glasgow) and GPR84 knock out (London) mice were fed Cloning of mouse GPR84. Mouse GPR84 was cloned from the mouse leukaemic macrophage cell line RAW 264.7. RNA was extracted from RAW 264.7 cells using RNeasy mini Kit (Qiagen) followed by cDNA synthesis using QuantiTect Reverse Transcription Kit (Qiagen) according to the manufacturer's instructions. Using this cDNA as template, the FLAG epitope tag (amino acid sequence DYKDDDDK) was incorporated at the N-terminus of mouse GPR84 by PCR reaction using the following primers: Mouse GPR84 sense: 5′ CAT GTT GGA TCC GCC ACC ATG GAC TAC AAG GAC GAC GAT GAT AAG TGG AAC AGC TCA GAT GCC AAC 3′.

Expression of human and mouse GPR84-Gα
Membrane preparation. Membranes were generated from THP-1 or RAW 264.7 cells in the presence or absence of LPS stimulation or Flp-In ™ T-REx ™ -293 cells following doxycycline treatment to induce receptor expression. Cells were centrifuged at 3000 rpm for 5 min at 4 °C. Pellets were resuspended in TE buffer (10 mM Tris-HCl, 0.1 mM EDTA; pH 7.5) containing a protease inhibitor mixture (Roche, Applied Science, West Sussex, UK) and homogenised with a 5 ml hand-held homogeniser. This material was centrifuged at 1500 rpm for 5 min at 4 °C and the supernatant was further centrifuged at 50000 rpm for 45 min at 4 °C. The resulting pellet was resuspended in TE buffer and protein content assessed using a BCA protein assay kit (Pierce, Fisher Scientific, Loughborough, UK).

Isolation of neutrophils from mouse bone marrow.
Tibiae and femurs of GPR84 knock-out and wild type mice were removed and freed of soft tissue. Both epiphyses were removed using sterile scissors and bones were flushed with a syringe filled with ice-cold RPMI 1640. Bone marrow suspension was passed through a 100 µm cell strainer and collected in a sterile 15 mL polypropylene tube for centrifugation at 300 g for 5 min at 4 °C.  β-arrestin-2 interaction assays. To assess potential agonist effects of GPR84 ligands at the phylogenetically related receptor GPR35, C-terminally enhanced yellow fluorescent protein (eYFP)-tagged forms of either human GPR35a or mouse GPR35 were co-transfected with a luciferase tagged form of β-arrestin-2 into HEK293 cells as described previously 21 . 24 hours later cells were treated with varying concentrations of the GPR35 agonist lodoxamine 21 or with 2-HTP or PSB-16671. Bioluminescence resonance energy transfer was then measured 5 minutes after addition of the luciferase substrate coelenterazine-h as in 21 .

Analysis of ligand affinity/co-operativity.
To calculate the allosteric parameters data obtained from allosterism experiments were analysed using an operational model of allosteric modulation described previously 18 where, E is the measured response, [A] and [B] are the orthosteric and allosteric ligand concentration at equilibrium, respectively; K A and K B represent the binding affinities of the two ligands to the receptor; α is the binding cooperativity factor reflecting the allosteric effect on binding affinity of the orthosteric agonist to the receptor, β is the activation cooperativity factor denoting the empirical measure of the allosteric effect on orthosteric efficacy. The values of τ A and τ B measure the ability of the orthosteric and allosteric ligand, respectively to directly activate the receptor. E m is the maximum system response and n denotes the slope factor of the transducer function. To fit the allosterism data globally through this equation, E m and n values were always constrained and other parameters (K A , K B , α, β, τ A and τ B ) were estimated.

Data Availability
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.