Interaction of DJ-1 with Lyn is essential for IgE-mediated stimulation of human mast cells

Abstract

Background

DJ-1 is a redox-sensitive protein with multiple roles in cell homeostasis whose levels are altered in mast cell (MC)-related disorders. However, whether DJ-1 can regulate human MC function is unknown.

Objective

We sought to investigate the potential role of DJ-1 in the responses of human MCs to antigen stimulation.

Methods

DJ-1 was silenced in human CD34⁺-derived and LAD2 MCs using lentiviral sh-RNA constructs. Release of β-hexosaminidase, PGD₂ and GM-CSF and changes in reactive oxygen species (ROS) levels were measured after FcεRI engagement. Enzymatic assays, sucrose density gradient centrifugation, immunoprecipitations, dot and Western blots, and confocal imaging were performed for signaling, cellular localization and co-association studies.

Results

DJ-1 knockdown substantially reduced mediator release as well as Lyn and Syk kinase activation and signaling by mechanisms which appeared largely unrelated to DJ-1 antioxidant activity. Upon FcεRI activation, non-oxidized rather than oxidized DJ-1 translocated to lipid rafts where it associated with Lyn, an interaction that appeared critical for maximal Lyn activation and initiation of signaling. Using purified recombinant proteins, we demonstrated that DJ-1 bound to Lyn directly but no other Src kinases, and this interaction was specific for human but not mouse proteins. In addition, DJ-1 reduced SHP-2 phosphatase activity by scavenging ROS thus preventing Syk dephosphoryation and perpetuating MC signaling.

Conclusion

We demonstrate a novel role for DJ-1 in the early activation of Lyn by FcεRI that is essential for human MC responses and which provides the basis for an alternative target in allergic diseases therapy.

Graphical Abstract

INTRODUCTION

Antigen-dependent activation of mast cells (MCs) occurs primarily through FcεRI, the high-affinity receptor for IgE, which causes the release and generation of vasoactive and inflammatory mediators of allergic inflammation^{1, 2}. Understanding the mechanisms that control IgE dependent MC activation is critical to the development of small molecular weight inhibitors of MC responses and to understanding aberrant regulation. It is currently understood that FcεRI aggregated by antigen coalesce within specialized microdomains of the plasma membrane (lipid rafts), and this causes the activation of Src family tyrosine kinases (SFKs), among which Lyn is the most critical for the initial stages of signaling elicitation³. Activated Lyn then phosphorylates the β and γ chains of the FcεRI, which respectively leads to recruitment of Lyn and recruitment/activation of spleen tyrosine kinase (Syk), an obligatory step for further propagation and divergence of signaling cascades leading to the release of bioactive mediators^3–5. Targeting of these receptor-proximal signaling events is preferable since they are common to all antigen-induced MC responses.

We have recently demonstrated altered serum levels of DJ-1, an evolutionary conserved protein reported to protect cells against oxidative damage, in patients with atopic dermatitis patients; and a biphasic regulation in mastocytosis where DJ-1 was induced by IL-6 and implicated in the expansion of neoplastic MCs^{6, 7}. Although DJ-1 is abundantly expressed in human MC lines^{6, 7}, it is unknown whether DJ-1 can influence FcεRI-mediated activation of human MCs.

As will be shown, we find that absence of DJ-1 in human MCs (HuMCs) results in marked inhibition of FcεRI–induced degranulation, cytokine and eicosanoid production by mechanisms which appear independent of its antioxidant activity. We further show a direct interaction of DJ-1 with Lyn in lipid rafts and that this association is required for Lyn activation, initiation of human MC signaling and functional responses. The interaction with Lyn is specific for non-oxidized rather than oxidized DJ-1 and unique to the human homologs, since such interaction was not observed between murine DJ-1 and Lyn proteins. Our data thus details a critical here-to-fore unknown function for DJ-1 in human MC responsiveness as a novel activator of Lyn and which may be of therapeutic interest in developing strategies to target the MC compartment.

METHODS

Human Mast Cell Cultures

Primary HuMCs were derived from CD34⁺ peripheral blood progenitors isolated from healthy volunteers after informed consent under a protocol (NCT00001756) approved by the Institutional Review Board of the National Institute of Allergy and Infectious Diseases (NIH). HuMCs and human LAD2 MCs were cultured as described^{8, 9}. For activation, HuMCs and LAD2 cells were sensitized overnight with biotinylated IgE (0.1 μg/ml), washed and the IgE-bound FcεRI crosslinked with 100 ng/ml streptavidin (SA).

Knockdown of DJ-1 by sh-RNA

Viral particles containing non-target (SHC002) or DJ-1-targeted short hairpin RNAs (sh-RNAs) (TRCN0000004921) in a pLKO1 vector were prepared as described¹⁰ and used to transduce HuMC or LAD2. Cells were selected in media containing 0.2 μg/ml puromycin for 5 days¹⁰.

Measurement of ROS and DJ-1

ROS levels were measured with OxiSelect^™ Intracellular ROS Assay kit and DJ-1 in supernatants of stimulated cells were measured by ELISA as described⁷. Alternatively, DJ-1 and oxidized DJ-1 were detected by Western blot⁷.

SHP-1 and SHP-2 phosphatase and in vitro kinase assays

Phosphatase activities of SHP-1 and SHP-2 in immunoprecipitates from SA-stimulated LAD2 cells were measured by a Duoset IC Phosphatase Assay. Kinase activities in immunoprecipitates of Lyn, Fyn, Fgr or Syk from cell lysates were determined by using an ELISA-based Universal Tyrosine Kinase Assay Kit.

Isolation of lipid rafts

LAD2 cells (2.0 × 10⁷) were lysed in 2 ml of lysis buffer containing 0.5% Brij 58¹¹ for 30 min on ice. Lysates were mixed with an equal volume of 80% (w/v) sucrose in buffer. A sucrose gradient was prepared and processed by ultracentrifugation at 270,000g for 18 h at 4°C. Fractions were collected from the top to the bottom of the gradient for Western blot analysis.

Confocal microscopy

Non-stimulated and SA-stimulated HuMC or LAD2 cells were seeded into 8-well Lab-Tek chamber slides and centrifuged at 600 rpm for 5 min. After washing, cells were fixed with 4% formaldehyde and permeabilized with 0.1% Triton X-100 for 10 min. Cells were then incubated 1 h with 1% BSA before staining. Confocal images were obtained with a Leica SP5 X confocal laser scanning microscope.

Statistical analysis

Data were expressed as mean ± SEM of values from 3 or more independent experiments. 1-way ANOVA was used to determine statistically significant differences between groups using Prism 6 software (Graph Pad Software, San Diego, CA, USA). Statistical significance was indicated as follows: *P <0.05 and **P <0.01.

A detailed description of the methods used in this study can be found in Supplementary Methods section.

RESULTS

Non-oxidized DJ-1 translocates to the plasma membrane early after FcεRI activation

In examining the role of DJ-1 in HuMCs, we first verified that activation via FcεRI resulted in a rapid increase in intracellular production of ROS which was maximal by 30 min, slowly declining thereafter (Fig. 1A), as reported¹². Total intracellular levels of DJ-1 did not change but DJ-1 was visibly oxidized by 30 min at Cys106, the major residue responsible for its antioxidant function (Fig. 1B, upper panels). Intracellular DJ-1 oxidation plateaued by 1 h, coinciding with both DJ-1 secretion into the media and the decline in intracellular ROS levels (Fig 1A and Fig 1B, lower panels). Therefore, a significant antioxidant effect of DJ-1 was only apparent 30 min to 1 h after FcεRI activation.

Figure 1. Non-oxidized DJ-1 localizes to the plasma membrane early after mast cell activation.

(A) Changes in intracellular ROS and release of DJ-1 into the extracellular media following stimulation with 100 ng/ml streptavidin (SA) of HuMCs sensitized with biotinylated IgE measured by ELISA. (B) Immunoblots showing changes in total DJ-1 and oxidized DJ-1 (at Cys106) inside cells or secreted to the media, as indicated. Quantification of blots is shown as bar graphs representing the average of 3 independent experiments. Loaded amounts of lysates (intracellular DJ-1) and of supernatants (extracellular) are not equivalent in “per cell” basis, being the latter 5 times more concentrated than those for the intracellular (see methods). (C) Confocal microscopic analysis of the intracellular localization of DJ-1 (green) and oxidized DJ-1 (red) in non-stimulated or SA-stimulated HuMCs for 0, 10, 60, and 240 min. Data are from a representative experiment of three independent experiments.

Additionally, within 10 min after activation, a portion of DJ-1 localized to the plasma membrane and this translocation was independent of any significant oxidation in Cys106 as determined by Western blot analysis and confocal microscopy (Fig. 1B and C). In fact, oxidized DJ-1 appeared distinctly excluded from the plasma membrane at this time point. The redistribution of non-oxidized DJ-1 into the membrane was transient since after 1 h most of the total cellular DJ-1 had been oxidized and appeared in the cytosol or associated with organelles (Fig. 1B and C). By 4 h, oxidized DJ-1 moved to the plasma membrane coinciding with the secretion of DJ-1 into the extracellular media (Fig. 1A). Thus, DJ-1 secretion appears to be dissociated from degranulation which is completed within minutes of activation. Since the oxidized form of DJ-1 is preferentially secreted¹³, the late appearing membrane-localized and now oxidized DJ-1 may well represent molecules in transit to the extracellular space. Of note, the amounts of DJ-1 secreted into the media were only a fraction of the cellular DJ-1 content as indicated in the legend to Fig. 1. The temporal changes in ROS as well as DJ-1 oxidation and localization were reproduced in the LAD2 MC line (Supplementary Fig. 1). Overall, these data indicate that the distinct rapid and transient translocation of the non-oxidized DJ-1 to the plasma membrane is unrelated to its scavenger function and may have functional consequences.

DJ-1 is essential for HuMC responses by mechanisms greatly independent of its antioxidant activity

After demonstrating early translocation of DJ-1 to the plasma membrane, we explored whether loss of DJ-1 by shRNA knockdown would have an effect on FcεRI-induced responses. We found that DJ-1 silencing with lentivirus-mediated shRNA resulted in a >75% reduction in DJ-1 (Fig. 2A and Supplementary Fig. 2A and B) without affecting the expression of KIT and FcεRI (Supplementary Fig. 2C). DJ-1 knockdown markedly inhibited FcεRI-mediated degranulation in HuMCs both in the absence (Fig. 2B, blue line) or presence (Fig. 2B, red lines) of stem cell factor (SCF) which enhances FcεRI dependent degranulation¹⁴. Degranulation induced by thapsigargin, which causes cytosolic Ca²⁺ increases by inhibiting Ca²⁺ uptake into intracellular stores or by GPCR agonists (Figure 2C) was not affected by the loss of DJ-1 (Fig. 2C), consistent with specificity on FcεRI signaling. FcεRI-mediated release of prostaglandin D₂ (PGD₂) and GM-CSF (Fig. 2D and F), but not thapsigargin-induced PGD₂ release (Fig. 2E), were also substantially reduced with DJ-1-knowdown compared with non-target shRNAs. DJ-1 knockdown had similar effects on FcεRI dependent responses in LAD2 MCs (Supplementary Fig. 3A–D).

Figure 2. DJ-1 expression is required for normal FcεRI–induced human mast cell responses.

(A) Reduced DJ-1 protein (left panel) and mRNA expression levels (right panel) in HuMCs transduced with either DJ-1 sh-RNA or non-targeted sh-RNA constructs. (B, C) Effect of DJ-1 knockdown (empty circles) in HuMC degranulation induced by SA (B) or the indicated concentrations of thapsigargin, complement component 3a (C3a) and Compound 48/80 (C). In B, IgE-sensitized HuMCs were stimulated with various concentrations of SA in media with (red lines) or without (blue line) 100 ng/ml SCF. Degranulation was measured as β-hexosaminidase activity released into the extracellular medium and expressed as percentage of the total cellular β-hexosaminidase activity. (D, E) Effect of DJ-1 knockdown in the production of PGD₂ secreted into the culture medium induced by the indicated concentrations of SA (D) or by thapsigargin (E) six hours after stimulation. PGD₂ was determined by ELISA. (F) Effect of DJ-1 knockdown in the production of the cytokine GM-CSF induced by SA. HuMCs were stimulated for 6 h and the amounts of GM-CSF released into the media determined by ELISA. All values are means ± SEM from 3 independent experiments. *P < 0.05, **P < 0.01.

Pretreatment of HuMCs with the general ROS scavenger 2,2,6,6-tetramethylpiperidin- 1-oxyl (TEMPO) prior to MC stimulation reversed the augmented intracellular ROS levels in DJ-1 knockdown cells (Fig. 3A) but did not restore the inhibition of degranulation (Fig. 3B), or PGD₂ (Fig. 3C) and GM-CSF (Fig. 3D) secretion by DJ-1 shRNA silencing. Similarly, the antioxidant TEMPO did not restore FcεRI-dependent responses in LAD2 cells (Supplementary Fig. 3E). Higher concentrations of TEMPO (300 μmol/L) did not have significant additional effects on ROS levels over 100 μmol/L TEMPO (Fig. 3A and Supplementary Fig. 3F) but induced 20% cell death and thus this concentration was not used in the study. The results using concentrations of TEMPO that markedly reduce ROS levels (albeit not totally), are consistent with the conclusion that the defect in FcεRI–induced mediator release in DJ-1 knockdown HuMCs, is largely (about 40%) unrelated to increased intracellular ROS levels due to the loss of DJ-1 antioxidant activity.

Figure 3. The effect of DJ-1 knockdown in FcεRI -induced human mast cell responses is largely independent of its effect on ROS.

(A) Confocal microscopy images showing intracellular ROS levels in HuMCs transduced with either lentiviral DJ-1 shRNA or non-target (NT) shRNA and pretreated or not with TEMPO for 10 min prior to SA (100 ng/ml) stimulation for 30 min. NS, non-stimulated. Right panels show determinations of ROS levels in HuMC after pretreatment with 30, 100 or 300 μmol/L TEMPO 10 min before stimulation with the indicated concentrations of SA. (B–D) Effect of the antioxidant TEMPO at the indicated concentrations on SA-induced degranulation (B), PGD₂ (C) and GM-CSF (D) secretion in HuMCs transduced with either lentiviral DJ-1 shRNA or non-target shRNA control. IgE-sensitized cells were pretreated with TEMPO for 10 min and then stimulated with SA. Degranulation (B) was measured as percentage of β-hexosaminidase released into the media 30 min after stimulation, and the amounts of PGD₂ (C) and GM-CSF (D) secreted into the culture medium were determined 6 hours after activation. All values are means ± SEM from 3 independent experiments. *P < 0.05, **P < 0.01.

Elevated ROS caused by knockdown of DJ-1 result in enhanced activity of SHP-2 but not SHP-1

DJ-1 has been reported to alter the function of the tyrosine Src homology 2 domain-containing phosphatase 1 (SHP-1) and SHP-2 in mouse bone marrow-derived MCs (mBMMCs), and negatively regulating FcεRI-induced activation⁶. In marked contrast, in human LAD2 cells, FcεRI induced activation of SHP-1 was not affected by DJ-1 knockdown (Fig. 4A). Further, in LAD2 cells, pretreatment with TEMPO (Fig. 4B), or incubation with H₂O₂ in vitro did not alter activation of SHP-1 (Fig. 4C). This contrasts with mBMMCs, where inhibition of SHP-1 activity by DJ-1 deficiency was reversed by the antioxidant TEMPO and was concluded to be the main mechanism involved in the hypereactive phenotype found in DJ-1 deficient mBMMC⁶.

Figure 4. The effect of DJ-1 knockdown on SHP-2 activation is mediated through increased ROS levels.

LAD2 cells transduced with lentiviral DJ-1 shRNA or non-target shRNA were sensitized with IgE and then stimulated with 100 ng/ml SA for 15 min. SHP-1 (A) and SHP-2 (D) were immunoprecipitated from the cell lysates and their phosphatase activities measured as the amount of free phosphate released from a synthetic phosphopeptide in 30 min. (B,E) Effect of ROS removal by TEMPO in the changes in SHP-1 (B) and SHP-2 (E) activity by DJ-1 knockdown. LAD2 cells were pretreated with 100 μmol/L TEMPO for 10 min before SA stimulation and the phosphatase activities determined as in A and D. (C,F) Effect of incubation with H₂O₂ for 20 min on SHP-1 (C) and SHP-2 (F) activities from non-stimulated (NS) and SA-stimulated (SA) LAD2 cells. All values are means ± SEM from 3 independent experiments. *P < 0.05. SA, streptavidin stimulated; NS, non-stimulated.

SHP-2 activation was, in contrast, enhanced in DJ-1 knockdown human LAD2 MCs (Fig. 4D), an effect shared with DJ-1 deficient BMMCs⁶. The enhanced activation of SHP-2 was fully restored by treatment with TEMPO (Fig. 4E), indicating this effect is ROS-mediated. Further confirming this conclusion, H₂O₂ induced a dose-dependent increase in SHP-2 activity in immunoprecipitates after FcεRI ligation (Fig. 4F). These results indicate that while the role of DJ-1 as a negative regulator of SHP-2 activation is conserved in human and mouse MCs, DJ-1 plays no role in the regulation of SHP-1 in human MCs. Therefore, the antioxidant component of DJ-1 function in human MCs is restricted to the control of SHP-2 and this, together with the lack of effect on SHP-1, may help explain the opposing mediator release effects between human and mouse MCs.

Knockdown of DJ-1 inhibits FcεRI-induced Lyn activation mostly independently of its effects on ROS

Since non-oxidized DJ-1 translocated to lipid rafts early after ligation of FcεRI and the activation of SFKs including Lyn, Fyn, and Fgr are the most receptor-proximal events^{3, 4, 15–17}, we explored the possibility that reduced DJ-1 expression affects their activity. The activation of Lyn kinase was significantly impaired in DJ-1-silenced LAD2 cells (Fig. 5A, left panel). However, the activities of Fyn, and Fgr (Fig. 5A, middle and right panels) were not affected. The inhibition of Lyn by DJ-1 knockdown was mostly independent of its regulation of ROS levels since it was only partially reversed by treatment with TEMPO (Fig. 5B). In agreement, addition of H₂O₂ to immunoprecipitated Lyn only mildly reduced its Lyn tyrosine kinase activity (Fig. 5C). As in LAD2 cells, DJ-1 knockdown also impaired the activation of Lyn in primary HuMCs and showed little dependency on ROS levels (Supplementary Fig. 4). These results also contrasted with the effect of DJ-1 deficiency or knockdown in mBMMC⁶, where Lyn was not found to be affected, and suggest that DJ-1 is an important contributor to Lyn activation in HuMCs but not mBMMCs, an effect that is in general unrelated to DJ-1 antioxidant activity.

Figure 5. DJ-1 knockdown suppresses Lyn activation mostly independently of its effects on ROS.

(A) Effect of DJ-1 knockdown in the activity of Src family kinases Lyn, Fyn and Fgr after SA stimulation of MCs. LAD2 cells transduced with lentiviral DJ-1 shRNA or non-target shRNA were sensitized with IgE and then stimulated with 100 ng/ml SA for the indicated times. Lyn, Fyn, and Fgr were immunoprecipitated and their kinase activities measured. (B) Effect of ROS removal by TEMPO in the loss of Lyn activation by DJ-1 knockdown. LAD2 cells were pretreated with 100 μmol/L TEMPO for 10 min before SA stimulation. (C) Dose response effect of H₂O₂ on the activity of immunoprecipitated Lyn from non-stimulated and SA-stimulated LAD2 cells. Lyn activity from immunoprecipitated Lyn samples was assayed in vitro in the presence of different concentrations of H₂O₂ for 20 min. All values are means ± SEM from 3 independent experiments. *P < 0.05, **P < 0.01. SA, streptavidin-stimulated; NS, non-stimulated.

DJ-1 is critical for activation of Syk and Syk-dependent phosphorylation events independently of its effects on ROS

Syk activation by Lyn is critical for early signaling events mediated by FcεRI¹⁸. Consistent with the effect on Lyn, Syk activity in immunoprecipitates was significantly reduced by DJ-1 knockdown in LAD2 MCs after FcεRI stimulation (Fig. 6A) and this effect was only partially restored by treatment with TEMPO (Fig. 6B). In agreement with the reduction in Syk activity, we observed reduced phosphorylation of Syk in tyrosines 525/526 and 352 (Fig. 6C, upper panel) that was minimally reversed with TEMPO (Fig. 6C, lower panel). Furthermore, knockdown of DJ-1 also substantially reduced phosphorylation of Syk-dependent targets including the adaptor linker for activation of T cells (LAT) and the downstream phosphorylation of PLCγ1, JNK and ERK (Fig. 6D). However, in agreement with the lack of effect on Fyn activity, Akt phosphorylation, which is downstream of Fyn activation¹⁶, was not significantly affected. Of note, only a small fraction of Lyn is needed to initiate signaling early after FcεRI engagement^19–21 and thus increased Lyn activity may not be readily detectable by immunoprecipitations and in vitro kinase assays until later times as receptor clusters and signaling complexes enriched in Lyn become enlarged ^{6, 19–22} (Fig. 5A).

Figure 6. DJ-1 knockdown suppresses Syk activation and downstream signals.

(A) Effect of DJ-1 knockdown in SA-induced activation of Syk. LAD2 cells transduced with lentiviral DJ-1 shRNA or non-target shRNA were sensitized with IgE and then stimulated with 100 ng/ml SA for the indicated times. Syk was immunoprecipitated and its activity in the immunoprecipitaes measured using the ELISA-based Tyrosine Kinase Assay Kit. (B) Involvement of ROS in DJ-1 knockdown-induced effect on Syk activity. Cells were treated with TEMPO (100 μmol/L) for 10 min prior to SA stimulation and Syk activity from the indicated lysates was determined as in A. Values are means ± SEM from 3 independent experiments. *P < 0.05, **P < 0.01. (C,D) Effect of DJ-1 knockdown on SA-induced signaling. Phosphorylation of Syk, LAT, PLCγ1, Akt, Jnk and Erk1/2 on LAD2 cells treated as in A was assessed by Western blotting using specific antibodies for the indicated proteins. Blots are representative of three experiments.

Collectively, the data are consistent with an essential role for DJ-1 in the propagation of FcεRI-mediated Lyn-Syk signaling and human MC functions by mechanisms that fundamentally differ from its effects on mBMMC.

Non-oxidized DJ-1 is required for proper phosphorylation and activation of Lyn in lipid rafts

Since DJ-1 was rapidly translocated to the plasma membrane where Lyn triggers signaling after FcεRI engagement, we investigated whether Lyn and DJ-1 colocalize and associate after stimulation. We detected DJ-1 in immunocomplexes with Lyn within 3 min of FcεRI crosslinking in both primary HuMCs (Fig. 7A) and LAD2 MCs (Fig. 7B), consistent with the finding that DJ-1 is critical for degranulation, which occurs within 2–3 min. However, association of DJ-1 and Lyn was maximal at 7 min and remained so for up to 15 min (Fig. 7A and B, upper panels). The persistence of the association DJ-1/Lyn after degranulation is ending suggests roles for DJ-1 in responses additional to degranulation, such as cytokine and PGD₂ release which are late MC responses. Confocal microscopic analysis confirmed the redistribution of a portion of DJ-1 to the plasma membrane and colocalization with Lyn early after stimulation (Fig. 7A and B, lower panels).

Figure 7. After FcεRI crosslinking, DJ-1 interacts with Lyn in lipid rafts and enables Lyn phosphorylation.

(HuMCs (A) and (B) LAD2 cells were stimulated with 100 ng/ml SA for 0, 3, 7, and 15 min and Lyn was immunoprecipitated (upper panels). Presence of DJ-1 in the immunoprecipitates was demonstrated by Western blotting. Confocal images show the localization of DJ-1 (in green) and Lyn (in red) before and after SA stimulation of HuMCs (A) and (B) LAD2 cells. (C) Distribution of Lyn and DJ-1 in sucrose density gradient centrifugation fractions from non-stimulated (NS) or SA-stimulated (SA) LAD2 cell lysates. (D) Lyn was immunoprecipitated from 1% Triton X-100 soluble and insoluble fractions of IgE-sensitized LAD2 cells stimulated or not with 100 ng/ml SA for 7 min, and the presence of DJ-1 or oxidized DJ-1 determined by Western blotting. (E) LAD2 cells transduced with lentiviral DJ-1 shRNA or non-target shRNA were stimulated with 100 ng/ml SA for 7 min. Lyn was immunoprecipitated from Triton X-100 soluble and insoluble fractions and DJ-1 or Lyn localization and Lyn phosphorylation determined by Western blotting. Quantification of phospho-Lyn and Lyn for are shown. The values are means ± SEM from 3 independent experiments. **P < 0.01.

Since Lyn is enriched in lipid rafts after MC stimulation, an environment important for Lyn-mediated phosphorylation of FcεRI³, we suspected that DJ-1 may also re-localize to these rafts, as was described in astrocytes after LPS activation²³. Fractionation studies did indicate that a substantial portion of DJ-1 translocated from the cytoplasm fractions (fractions 8–11) into lipid rafts (fractions 4–6; as defined by the marker protein flotillin-1 and Lyn), thus colocalizing with Lyn following stimulation (Fig. 7C). Consistent with the confocal images shown in Fig. 1C, the majority of DJ-1 that translocated into lipid raft fractions was not oxidized (Fig. 7C). Furthermore, DJ-1 was detected in immunocomplexes with Lyn only in detergent-insoluble fractions (lipid rafts) after MC stimulation but not in detergent-soluble fractions (Fig. 7D) and the immunoprecipitated DJ-1 associating with Lyn in these fractions was not oxidized in Cys106 (Fig. 7D).

The association of Lyn with DJ-1 in lipid rafts after FcεRI engagement did not appear to be required for the translocation of Lyn into these domains since silencing of DJ-1 did not prevent enrichment of Lyn in lipid rafts (Fig. 7E). However, it did markedly inhibit FcεRI-mediated Lyn phosphorylation (Fig. 7E), indicating that DJ-1 facilitates Lyn activation.

Human non-oxidized DJ-1 directly interacts with human Lyn

Because DJ-1 is reported to directly alter the activity of kinases and tyrosine kinases^24–26, we next determined whether recombinant DJ-1 could induce Lyn activation in vitro. As shown in Fig. 8A, recombinant DJ-1 induced a dose-dependent increase in Lyn activity in Lyn- immunoprecipitates from resting LAD2 cells (Fig. 8A, black bars). However, this induction was not observed when DJ-1 was previously oxidized with H₂O₂ (Fig. 8A, red bars), which together with the observation that oxidized DJ-1 was not present in the plasma membrane or lipid rafts early after FcεRI engagement, indicates that the effect of DJ-1 on Lyn activity is independent of its antioxidant activity.

Figure 8. Human DJ-1 directly interacts with Lyn, but not Fyn or Syk, and enhances its tyrosine kinase activity.

(A) Lyn activity in immunoprecipitates from unstimulated LAD2 cells incubated with or without various amounts of purified human recombinant DJ-1 or oxidized DJ-1 protein (OX-DJ-1) for 30 min. Values are means ± SEM from 3 independent experiments. *P < 0.05. (B,C) Human recombinant Lyn, Fyn, Syk and DJ-1 were immobilized on a nitrocellulose membrane and incubated with a solution of 100 ng/ml human recombinant DJ-1 (B) or human recombinant Lyn (C). After washing, positive interactions were revealed using antibodies against DJ-1 (B) or anti-Lyn (C). C, right panel: positive controls for recognition of doted proteins to the corresponding antibodies. (D) Dot blots demonstrating that only non-oxidized human recombinant DJ-1 (h R DJ-1) interacts with human recombinant Lyn. Blots are representative of three experiments. (E) Model summarizing the findings. After activation of the FcεRI, non-oxidized DJ- 1 is translocated to lipid rafts where it interacts with Lyn facilitating its activity and downstream signaling. In addition, at later times after FcεRI engagement, DJ-1 has a positive effect on MC signaling acting as an antioxidant protein that reduces ROS levels, thus limiting SHP-2 activation and enhancing Lyn activity, both of which favor Syk activation and MC responses.

To test whether Lyn and DJ-1 interact directly, we performed dot blots with recombinant DJ-1 and recombinant tyrosine kinases. Increasing concentrations of either purified recombinant Lyn (Fig. 8B) or DJ-1 (Fig. 8C, left panel) immobilized in nitrocellulose showed immunoreactive dots after incubation with DJ-1 (Fig. 8B) or Lyn (Fig. 8C, left panel). In contrast, no immunoreactive dots were observed when either recombinant Fyn or Syk was employed (Fig. 8B and C, left panel). Spotted recombinant proteins immunoreactingto their corresponding antibodies are shown in Fig. 8C (right panel) as positive controls.

Furthermore, human recombinant oxidized DJ-1 did not show positive interactions, indicating that only non-oxidized DJ-1 is able to interact with Lyn (Fig. 8D). Dot blot assays also indicated that mouse recombinant DJ-1 did not interact with mouse Lyn (Supplementary Fig. 5A) in agreement with the lack of effect of DJ-1 on Lyn activity in mBMMC. Furthermore, this interaction appears to be specific between the human homologs because no immunoreactive dots were observed between mouse DJ-1/human Lyn or human DJ-1/mouse Lyn (Supplementary Fig. 5B and C). Collectively, these results demonstrate a direct interaction between Lyn and non-oxidized DJ-1 that enhances the Lyn activity and appears unique to the human proteins.

DISCUSSION

MCs play a critical role in allergic disorders mainly through the activation of the high-affinity IgE receptor (FcεRI) expressed on the MC surface. When the IgE-FcεRI complex is aggregated by multivalent antigen, the open, active conformation of the Src family tyrosine kinase (SFK) Lyn is favored, causing autophosphorylation of Lyn and transphosphorylation of tyrosine residues in the β and γ subunits of adjacent receptors. The latter phosphorylation creates docking sites for the recruitment of the tyrosine kinase Syk. Syk phosphorylation by Lyn and consequent activation amplifies signaling cascades that result ultimately in the release of vasoactive and inflammatory mediators from the MC^4,3,5. However, how exactly Lyn initiates these events in human MCs is not entirely clear. Here we demonstrate that direct binding of Lyn to DJ-1 in the plasma membrane is a required step for full Lyn kinase activation, transphosphorylation of the γ subunits of FcεRI and consequent responses in human MCs. This is a first report showing that DJ-1, by mechanisms partially independent of its action as a ROS scavenger, is critical for the organization of early events in human MC signaling.

Interest in understanding the biochemical and functional properties of DJ-1 (also known as PARK7) has been fueled by the association of DJ-1 mutations with familial early onset Parkinson’s disease²⁷ and by altered levels in diseases such as cancer, stroke, dermatitis and mastocytosis^{6, 7, 28, 29}. A well-established function of DJ-1 is to protect cells from oxidative damage partly by undergoing self-oxidation thus reducing ROS levels^30–32, although other functions for DJ-1 independent of this feature have also been described^{24, 25, 33, 34}. We confirmed that FcεRI engagement induced ROS generation in HuMCs¹² and found that self-oxidation of DJ-1 30 min after activation occurred concomitantly with a decline in ROS levels, consistent with the antioxidant role for DJ-1. Our study further reveals a requirement of DJ-1 for degranulation, cytokine and eicosanoid responses in HuMCs that appeared to be, however, greatly independent of its antioxidant activity since treatment with an antioxidant only restored by 20% the diminished responses caused by DJ-1 silencing. This is consistent with reports indicating that ROS generation has little effect on FcεRI-mediated MC responses³⁵. Thus, the data supports ROS-dependent and ROS-independent roles for DJ-1 in HuMC responses. We provide evidence that a consequence of DJ-1 antioxidant activity is to restrict SHP-2 activation by removing excess ROS levels which normally enhances SHP-2 phosphatase activity, as was observed in mBMMCs⁶. However, unlike in mBMMCs, DJ-1 had no effect on SHP-1 activity. Restricted SHP-2 activity would result in diminished Syk dephosphorylation and thus sustained Syk activation.

In addition, silencing of DJ-1 expression markedly reduced Lyn activity in HuMCs by a mechanism unrelated to its ROS scavenger activity, as was true for mediator release, and involved a direct, specific association with Lyn. DJ-1 did not similarly associate with Syk nor the SFK Fyn, and Akt signaling downstream of Fyn was unaffected by DJ-1 absence. DJ-1 has been reported to alter the activity of certain receptors and signaling proteins in part by direct association^{36, 37}, in some instances by mechanisms not yet well understood^{24, 25, 33, 38}. Based on our findings we propose a unique dual function for DJ-1 in FcεRI-activated human MCs sequentially regulating Lyn and Syk pathways: initially, when DJ-1 is not oxidized, by directly facilitating Lyn activation; and later, as it becomes oxidized, by reducing ROS levels and SHP-2 activity thus potentially alleviating SHP-2 repression of Syk activity (Fig. 8E).

We provide insights on how and where DJ-1 exerts its regulation of Lyn activation. Similar to other cells^{39, 40}, in non-stimulated HuMCs, DJ-1 appeared localized intracellularly but shortly and transiently after activation, a fraction of non-oxidized DJ-1 translocated to lipid rafts. Lipid rafts are plasma membrane microdomains enriched in cholesterol, glycosphingolipid, and sphingolipids^{41, 42}, believed to generally function as focal points for the consolidation of FcεRI signaling components. Lyn and DJ-1 not only colocalized but formed distinct Lyn/DJ-1 immune complexes in these domains in response to FcεRI ligation. Formation of such complexes did not require adaptor proteins since purified recombinant DJ-1 directly bound recombinant Lyn in vitro, and resulted in enhanced Lyn activity. Although the precise mechanisms for the effect on Lyn activity are not entirely clear they appear to be partly independent of its antioxidant function. It is unlikely that DJ-1 prevents interactions with Cbp/Csk complexes, a feed-back loop that keeps Lyn in an inactive state through phosphorylation of Lyn in Tyr507⁴³, since DJ-1 deficiency did not alter Lyn phosphorylation in Tyr507 or the phosphorylation and levels of Cbp that co-immunoprecipitated with Lyn (Supplementary Fig. 6A). Nevertheless, DJ-1 knockdown blunted total phosphorylation of Lyn and its association with Fcγ subunits (Supplementary Fig. 6A) and recombinant DJ-1 was sufficient by itself to enhance Lyn kinase activity in Lyn immunoprecipitates from unstimulated MCs and of recombinant Lyn in vitro (Fig. 8A and Supplementary Fig. 6B) to suggest that a consequence of Lyn/DJ-1 interaction may be to promote the open, active conformation of Lyn.

The active conformation of Lyn is usually favored by binding of Lyn SH2 and SH3 domains to phosphotyrosine-containing sequences and proline-rich motifs, respectively^{44, 45}. DJ-1 does not contain proline-rich motifs, but it was phosphorylated on tyrosine after antigen stimulation (Supplementary Fig. 6C). The fact that purified recombinant DJ-1 and Lyn interacted in the absence of ATP in our studies suggests that DJ-1 phosphorylation is not necessary for binding, but may stabilize the active conformation of Lyn. However, it appears that a non-oxidized Cys106 is required for the interaction and enhancement of Lyn activity. The interaction between human Lyn and DJ-1 is also species-specific in that it was not observed between the two murine proteins or between human and murine homologs, which suggests an acquired function for DJ-1 in humans despite the overall sequence homology between species. This adds a cautionary note in translating mouse data to humans with respect to MC activation mechanisms.

Collectively, our results shed new light on the regulation of human MC responses after IgE receptor engagement by identifying DJ-1 as an unsuspected player in Lyn activation. This function of DJ-1 requires non-oxidized cysteine residues as well as translocation of DJ-1 to lipid rafts and direct interaction with Lyn that facilitates its kinase activity and initiation of signaling cascades. Thus, further characterization of DJ-1/Lyn interaction may open new avenues of therapeutic modalities in MC-derived diseases. The finding of DJ-1 as an essential regulator of Lyn may also be relevant for other immunoglobulin receptors and/or human immune cells where Lyn function is critical.

KEY MESSAGES.

• Ligation of FcεRI in human mast cells induces the early translocation of non-oxidized DJ-1 to lipid rafts where it interacts with Lyn enabling its full activation, downstream signaling and mast cell responses

• Later after FcεRI activation, the ability of DJ-1 to become oxidized quenches ROS levels and reduces the activity of the phosphatase SHP-2 which permits further Syk-mediated signaling.

• Further characterization of the interaction DJ-1 with Lyn could form the basis for alternative treatments in allergy.

ABREVIATIONS

Cbp

Csk binding protein

Csk

C-terminal Src kinase

FcεRI

High affinity receptor for IgE

GM-CSF

Granulocyte-macrophage colony-stimulating factor

GPCR

-G-protein coupled receptors

HuMCs

CD34⁺-derived human mast cells

MC

Mast cells

mBMMC

Murine bone marrow mast cells

PGD₂

Prostaglanding D2

ROS

Reactive Oxygen Species

SA

Streptavidin

SFK

Src family kinase

Syk

Spleen tyrosine kinase