Rheumatoid arthritis (RA) is a painful and disabling autoimmune disorder, affecting about one percent of the population in Western countries [1]. As a main indication the disease comprises chronic inflammation of pheripheral joints, resulting in progressive destruction of articular cartilage and bone [2]. Inflamed tissue is characterized by infiltration of leukocytes, pannus formation and occurrence of aggressive synovial fibroblasts [1]. Enhanced expression of several cytokines or matrix metalloproteinases by these cells promotes pathogenicity [reviewed in [3, 4]]. Moreover, various chemotactic factors are produced or activated in the joint tissue, recruiting even more leukocytes and exacerbating inflammation [5].

Despite major advantages during the last decade, currently available therapeutic approaches for RA have only partial clinical benefit and are associated with considerable side effects. Treatment strategies include anti-inflammatory or immunosuppressive drugs and biologicals, e.g. antibodies against TNF. Recently, prevention of leukocyte infiltration in inflamed tissue by blocking chemokines or chemokine receptors has also been explored but with limited success [6], possibly due to redundancy, enabling efficient leukocyte responses, even when one particular factor is blocked. Therefore current investigations are directed to the suppression of mutual intracellular signaling pathways shared by multiple chemokines. One prominent protein, integrating chemokine signaling in leukocytes, is PI3Kγ, a G-protein-coupled receptor (GPCR) isoform of phosphoinositide 3-kinases [7]. This enzyme was shown to regulate chemotactic responses of neutrophils, macrophages and T cells to several stimuli, including IL-8, C5a and SDF-1α [8]. Furthermore, PI3Kγ is involved in oxidative burst induction in phagocytes [9, 10] and activation of T cells [11, 12]. Thus, ablation of PI3Kγ could prevent both, leukocyte infiltration into joints and autoimmune activation. Indeed, Camps et al. showed a marked suppression of joint inflammation and reduced tissue destruction by inhibition of PI3Kγ activity in the mouse model of collagen-induced arthritis (CIA) [13]. Our data, presented here, prove a role for PI3Kγ in the early phase of murine antigen-induced arthritis (AIA), which is due to decreased phagocyte infiltration into the joint and reduced macrophage activation. However, PI3Kγ-/- mice showed unaltered inflammation at later time points together with normal T cell responses in this model.

Our study aimed at elucidating the role of PI3Kγ in the mouse model of AIA, which resembles many aspects of RA in humans. PI3Kγ is known to control various effector functions of leukocytes, including migration (e.g. extravasation from blood to tissue), oxidative burst, cytokine production and phagocytosis [8–10, 19, 20]. This suggests the kinase as a possbile exacerbating factor in autoimmune disorders, including RA. Accordingly, in the AIA model genetic or pharmacologic inactivation of PI3Kγ results in considerably diminished manifestation of early disease. Joint swelling two days after arthritis induction was only about 60% of the controls and histological scoring in the knockouts revealed only about 40% disease intensitiy compared to wildtype animals. These findings are in line with studies on PI3Kγ in mouse models of CIA or serum-induced arthritis [13, 21] and underline the importance of this molecule in acute inflammatory processes. Quite surprisingly, in the AIA model at later time points, PI3Kγ-/- mice revealed a fully developed disease with knee swelling and histological scoring being comparable between controls and knockouts. These observations were in contrast to both other studies on PI3Kγ in arthritis [13, 21] but may be explained by different requirements for T cells among the three models. Randis et al. [21] induced arthritis by serum transfer from K/BxN mice to PI3Kγ-/- on a C57BL/6 background, a model that was reported to be independent of lymphocytes [22]. Likewise, T cells are dispensable for the development of anti-CII antibody-induced arthritis [23], one of the models investigated by Camps et al. [13]. However, the same group also studied CIA, which requires T cell activation, and found considerably decreased disease symptoms upon treatment of mice with AS-605240. The reason for this discrepancy is not completely clear, but most probably, the relative contribution of innate immunity to CIA is much stronger than in the later (chronic) phases of the AIA model. This might be reflected by the different pattern of paw and knee swelling: while paw thickness in the CIA model steadily increases during the whole observation period, the knee swelling in AIA has a maximum at day 1 and decreases thereafter.

Macrophage activation peaked at day 2 in our experiments, while main T cell responses were observed at days 7 and 21. From this, we hypothesized, that loss of PI3Kγ mostly inhibits phagocyte activation and function, whereas T cells are less or even not affected. Our data in vivo and in vitro support this hypothesis: Effector production by peritoneal macrophages (NO, IL-1β, IL-6) ex vivo was clearly decreased in PI3Kγ-/- mice during the course of AIA. It is important to note, that cytokine production in this experiment was not induced by external stimuli, specifically targeting GPCR receptors and PI3Kγ, but rather reflected the state of in vivo macrophage priming. Thus, our results extend data of several groups, investigating PI3Kγ-dependent activation of macrophages in vitro [9, 10, 20]. In our experiments, C5a induced phosphorylation of Akt in wildtype macrophages in vitro, while no phosphorylation could be observed in PI3Kγ-/- cells. This is in line with previous reports [8, 13] and contradicts to some extend a recent study, showing concerted and partially redundant action of PI3Kγ and PI3Kβ isoforms downstream of GPCR [24]. These issues require further investigation and clarification. Additionally, macrophage migration towards C5a in vitro and infiltration of both, neutrophils and macrophages to the inflamed joint in vivo were diminished in the PI3Kγ-/- mice and migration of wildtype macrophages could be blocked in vitro by using AS-605240. Taken together, innate immunity seems to be largely impaired by loss of PI3Kγ, causing the marked alleviation of clinical symptoms at early days of AIA.

On the other hand, our experiments revealed no changes in T cell activation of PI3Kγ-/- mice. Neither cytokine production in vitro (TNFα, IFNγ, IL-2, IL-4, IL-5, IL-17) was significantly altered, nor was the DTH reaction in vivo. Those results were somewhat surprising, considering several reports on involvement of PI3Kγ in T cell maturation, migration and activation [25–28]. However, most of those results were obtained in vitro, while in vivo data are less clear. Most groups investigated an acute activation of T cells by anti-CD3-antibodies in vitro or migration of naïve T cells towards SDF-1α in a time frame of several hours. Contrary, the immunization period of the AIA model (3 weeks) might provide sufficient time for efficient T cell priming and mounting of effector function. Moreover, in two models of systemic lupus erythematosus, PI3Kγ was shown to control the CD4+ memory T cell reservoir mostly in the context of a second genetic alteration (MRL-lpr or p65^PI3KTg), while single-targeted PI3Kγ-inhibition revealed only modest phenotypes [27, 29]. Additionally, although defective T cell differentiation can be detected in newborn PI3Kγ-deficient mice, differences decrease in mice of about one month of age [12]. Those data suggest an at least partial compensation of PI3Kγ in vivo, dependend on the microenvironment and activation status of the T cell. This premise is supported by recent studies of Ward and colleagues on human T lymphocytes. The activation status or even culture conditions strongly influenced requirement of PI3Kγ for migration in these experiments [30, 31].

The DTH reaction at 24 and 48 hours post challenge was unchanged in PI3Kγ-/- mice (days 8 and 9 of AIA). This again demonstrates a largely unaltered T cell response and suggests a compensation of deficient signaling reactions in PI3Kγ-/- T cells by other proteins. Although defective development of thymocytes or CD4 memory T cells as described by Rodriguez-Borlado et al. [12] and Barber et al. [27] was not addressed by our experiments, this does not seem to primarily influence disease development.

The effects of genetic ablation of PI3Kγ could be mimicked largely by using the inhibitor AS-605240, specific for PI3Kγ. This compound was able to markedly decrease macrophage activation and joint swelling during the first days (acute phase) of AIA, in accordance to Camps et al. [13]. Efficacy of AS-605240 in both models points to a possible impact in human rheumatoid arthritis too and asks for further development of this drug candidate towards clinical use. Although T cell activation was not influenced by the inhibitor, its ability to alleviate clinical parameters specifically during early AIA suggests a potential of AS-605240 or other PI3Kγ inhibitors preferentially in therapy of acute exacerbations of the disease, a hallmark of chronic RA in humans.

In summary, PI3Kγ has its greatest impact on AIA during the first days of the inflammation, when innate immune cells dominate pathogenesis. The enzyme controls production of NO and cytokines by macrophages and migration of macrophages and neutrophils. Deletion or inhibition of PI3Kγ does not influence T cell responses in the AIA model. Accordingly, disease becomes fully developed at later stages, as judged from T cell cytokine production and histological sections of inflamed joints. Thus, influence of PI3Kγ on adaptive immune reactions is rather limited in this model.