Protein Phosphatase 2A Catalytic Subunit α (PP2Ac) Plays a MyD88-dependent, Central Role in the Gene-Specific Regulation of Endotoxin Tolerance (ET)

As a Core Component of MyD88 Interactome in ET Macrophages PP2Ac is Chronically Activated

First, by using our AACT-based quantitative proteomic approach(Chen et al., 2000) with modifications for interactome screening (Figure 1A) we dissected the MyD88-interacting complexes assembled in RAW cells under different inflammatory states including (1) no stimulation (N), (2) challenging with a single high LPS dose (LPS-responsive, NL), (3) priming with a low LPS dose (LPS-tolerant, T), and (4) challenging T cells with a high-dose LPS (TL). Compared to NL cells, T or TL cells showed ET-specific phenotype of immunosuppression and resistance to apoptosis (Figure S1A). Through phenotypic interactome analysis (Figure 1A), we found that MyD88 interacts with different sets of proteins in NL vs TL macrophages (detailed proteomic data will be reported elsewhere): together with many negative immune regulators(Liew et al., 2005) including a negative TLR regulator Flii (Wang et al., 2006), PP2Ac was found recruited into the MyD88 interactome specifically in ET cells (Figure 1B and Figure S1B). Because approximately 20% of the components in TL-specific MyD88 interactome contain the domains interacting with PP2Ac (Figure S1C) that is generally considered as a suppressor of pro-inflammatory kinases (Junttila et al., 2008), this inflammation-phenotypic proteomic finding suggested that PP2Ac plays a central, MyD88-dependent, immunosuppressive role during ET.

Given neither expression nor stability of PP2Ac was affected by LPS-induced inflammation (Figure S2A), we compared the PP2Ac activity in RAW cells under different inflammatory conditions. Compared to naïve (N) cells, while its activity was little changed under NL, PP2Ac was highly activated with a prolonged stimulation and was sustained under TL (Figure 1C), indicating the activity-based, inflammation-phenotypic function of PP2Ac. To clarify the MyD88-dependence of PP2Ac activation, we measured PP2Ac activity in paired wild-type (WT) and MyD88-depleted (MyD88^−/−) (Figure S2B) BMDMs under different inflammatory states induced by a variety of TLR agonists. Similar to RAWs, PP2Ac activity showed a 2-fold increase in TL WT BMDMs but less than 10% changes in MyD88^−/− BMDMs under N, NL, TL (Figure 1D). The TLR9 agonist CpG, which triggers only MyD88-dependent pathways, induced a greater increase in PP2Ac activity in TL WT BMDMs than that induced by LPS which activates both MyD88-dependent and -independent pathways (Figure 1E), while no differences in PP2Ac activity were observed between WT and MyD88^−/− BMDMs when the TLR3 agonist Poly (I:C) stimulated only MyD88-independent pathways (Figure 1F). These agonist-specific effects indicated the MyD88-dependent activation of PP2Ac in chronic-inflamed cells. Second, to clarify how PP2Ac activity is regulated in tolerant macrophages, we respectively used either LPS-conditioned medium, i.e., that separated from the LPS-pretreated RAWs, or TNFα or ionizing radiation (IR) to trigger inflammation in BMDMs. Under chronic exposure all non-TLR stimuli caused a similar degree of tolerance-specific activation of PP2Ac in both WT and MyD88^−/− BMDMs (Figure 1G). In a coordinate manner, unlike its LPS-inducible nature in WT but not in MyD88^−/− BMDMs (Figure 1H, left), phospho-p65 triggered by non-TLR stimuli showed similar time-dependent induction in both types of BMDMs (Figure 1H, right), indicating that stimuliinduced, secondary, NFκB-regulated production of inflammatory cytokines such as TNFα causes tolerance-specific PP2Ac activation. Because TNFα is a T-class gene and its release is MyD88-promoted (Figure S2B), this secondary effect on PP2Ac activation was therefore more pronounced in the WT BMDMs directly stimulated by the TLR agonists that trigger MyD88- dependent pathways for producing these T-class cytokines. Therefore, only during the chronic inflammation mediated by TLR2/4/9, the sustained PP2Ac activation is MyD88-dependent.

Constitutively Active PP2Ac Regulates Immunosuppression During ET

To determine the correlation of persistently or constitutively activated PP2Ac with immunosuppression, we measured cytokine release from stable RAW lines expressing either shRNA for knocking down PP2Ac (PP2Ac KD) or shRNA for GFP (WT control) (Figure 2A and S2C). Under NL, most cytokines showed LPS-inducible releases in both PP2Ac KD and WT RAWs (Figure 2B). In contrast, compared to the LPS-tolerized WT RAWs showing reduced production of T-class cytokines, the LPS-primed PP2Ac KD cells were hyper-responsive to highdose LPS stimulation and released greater amounts of cytokines. Further, we determined the effect of constitutively active PP2Ac on the expression of 84 mouse genes encoding inflammatory factors, most known to be regulated by NFκB (Figure S2D). Compared with their levels in WT RAWs, 34 out of 84 genes showed more than 30% increases (Table S1), some of which were validated by qPCR (Figure 2B). Also, T-class genes (IL6 or IL10) had dramatically enhanced expression in the PP2Ac KD cells even under TL, reversing their suppressed expression in TL WT cells. In contrast, under NL or TL PP2Ac knockdown had little effect on the LPS-inducible trend of Fpr1 (Figure 2B), a non-tolerizeable gene (Foster et al., 2007), suggesting that constitutivelyactive PP2Ac is involved specifically in T-class gene regulation under ET.

Further, to determine whether the constitutively active form of PP2Ac directly contributes to ET-acquiring, we transfected constructs expressing either WT or a dominant-negative (DN) mutant of PP2Ac(Chang et al., 2011) into the PP2Ac KD RAWs. Under TL, while expression of WT PP2Ac restored the tolerance with reduced IL6, the PP2Ac DN mutant with similar expression failed to suppress this T-class gene (Figure 2C), indicating that constitutively active PP2Ac participates in the gene-specific control of chronic inflammation by selectively suppressing T-class genes.

Constitutively Active PP2Ac Promotes the MyD88-dependent Survival of ET Cells Through Broadly Targeting Pro-inflammatory Kinases and Pro-apoptotic Factors

To identify those PP2Ac-target pathways through which ET is regulated by constitutively active PP2Ac, we profiled the phosphoproteomic changes in PP2Ac KD versus WT RAWs under TL (Figure S3A). Some of PP2Ac-target sites (Figure S3B) that were identified based on their phosphorylation increases in TL PP2Ac-depleted cells were in the components of TL-specific MyD88 interactome (Figure S1B), including Rps6 (a substrate of p70S6K), Rap1 (a MAPK component), Akt1, and a mTORC1/C2 component Raptor, revealing that through sitespecific dephosphorylation constitutively active PP2Ac represses the pro-inflammatory/-apoptotic activity of the PI3K-mTOR-p70S6K-Akt pathway. In addition, we found that constitutively active PP2Ac targeted AP1/c-Jun, the mediator of stimuli-induced apoptosis (Li et al., 2004) (Figure S3C): under TL, in contrast to its decreased abundance in WT RAWs, phosphorylation of c-Jun Ser63 was found significantly increased when PP2Ac was depleted (Figure 2D). Thus, similarly to the anti-apoptotic nature of PP2A (Kong et al., 2004), PP2Ac, when constitutively activated, negatively regulates the pro-apoptotic activity of c-Jun by directly dephosphorylating its pro-apoptotic pSer63.

We next determined how PP2Ac knockdown affects the ET-characteristic resistance to apoptosis. First, we evaluated cell survival/proliferation in PP2Ac KD RAWs under N, NL, TL in the presence of cycloheximide (CHX) that can amplify LPS-induced apoptosis (Suzuki et al., 2004). Compared with the ET-specific increase of the surviving WT cells, PP2Ac knockdown led to reduced colony formation, indicating the increasing apoptotic/dead cells even under TL (Figure 2E). Similarly, in flow cytometry analysis, consistent with what was shown for cytokine release (Figure 2B), PP2Ac KD RAWs showed some degree of LPS-inducible apoptosis/cell death even in non-stimulated cells (Figure 2F), indicating the cell sensitization due to PP2Ac depletion. The NL WT RAWs showed LPS-inducible apoptosis, whereas this induction was suppressed under TL. In contrast, a dramatically increased susceptibility to apoptosis was observed even for T or TL PP2Ac KD cells. The consistent results from both assays (Figure 2E and 2F) indicated that PP2Ac depletion reduces the resistance of ET macrophages to apoptosis.

Further, because MyD88 promotes tumor growth by either modulating particular kinase pathways(Lee et al., 2010) or inhibiting apoptosis(Rakoff-Nahoum and Medzhitov, 2007), we next examined whether the PP2Ac effect on apoptotic propensity is MyD88-dependent. Because PP2Ac is highly abundant and robust, and non-dividing BMDMs have extremely low transfection efficiency and shRNA stability, very low percentages of endogenous PP2Ac KD could be achieved in BMDMs (data not shown). Given that the phosphatase inhibitor okadaic acid (OA) at concentrations <50 nM showed greater specificity in inhibiting the activity of PP2Ac to that of PP2Ac DN (Chang et al., 2011), we used 20 nM OA to mimic PP2Ac KD in the experiments involving BMDMs. Accordingly, we observed similar increases of c-Jun pSer63 in either PP2Ac KD RAW or OA-treated BMDMs, both under TL (Figure 2D and Figure S3C). Additionally, under T or TL, both WT and PP2Ac KD RAWs pre-treated with 20 nM OA showed similar changes in the populations of apoptotic/death cells (Figure 2F, middle versus bottom row). All these similarities in OA treatment and shRNA-mediated PP2Ac KD on promoting apoptosis indicated the specificity of 20 nM OA in inhibiting chronic-active PP2Ac.

We therefore used 20nM OA to investigate the MyD88 dependence of the PP2Ac effect in primary WT vs MyD88^−/− BMDMs. Accordingly, similar to WT RAWs, more WT BMDMs underwent apoptosis/death under NL than those under TL (Figure S3D). Conversely, viability of MyD88^−/− BMDMs was unaffected by either NL or TL. Further, OA-mediated PP2Ac inhibition triggered more apoptosis/death of TL WT but not MyD88^−/− BMDMs. Further, compared to their untreated counterparts, the OA-treated WT BMDMs under TL showed features typical of mitochondrial-mediated apoptosis, including greater cleavage of caspase-3 and PARP, and higher expression of the pro-apoptotic proteins Bak and Bax (Figure 2G). In contrast, such an OA-initiated apoptosis in TL cells was abolished in MyD88^−/− BMDMs. Because c-Jun pSer63 is upstream of caspase-3 activation during apoptosis(Li et al., 2004), thus, analysis of multiple, interconnected pathway-specific apoptotic markers indicated that constitutively active PP2Ac promotes cell survival in a MyD88-dependent manner.

Constitutively Active PP2Ac Dissociates the Signal-promoting TLR4-MyD88 Complex Through Dephosphorylating key Serines in the Toll/IL-1 Receptor (TIR) Domain of MyD88

We next investigated the role of TL-specific PP2Ac-MyD88 interaction (Figure 3A, left) in acquiring ET. First, we found that MyD88 binds to the WT PP2Ac more tightly than to the PP2Ac DN under TL (Figure 3A, right), indicating that MyD88 interacts only with constitutively active PP2Ac. Further, to identify exactly where constitutively active PP2Ac targets MyD88, we employed an AACT-based quantitative proteomics approach where tagged MyD88 expressed in HEK TLR4 stable cells was immunoprecipitated for MS analysis because no phosphorylation on endogenous MyD88 was detected in the phosphoproteomic analysis above. Meanwhile, we used 20nM OA to highlight PP2Ac-target phospho-sites on MyD88. In MyD88 immunoprecipitates from differentially stimulated TLR4 stable cells, with or without 20nM OA treatment, we identified novel phosphorylation sites at Ser242/244 (Figure 3B), and also found that these residues exhibited an LPS-inducible increase in phosphorylation, more so in NL than in TL cells (Figure S4A and S4B). Further, phosphorylation of these sites was increased approximately 30% in the OA-treated TL cells (Figure 3C) compared to NL cells in which OA had little effect, suggesting that constitutively active PP2Ac targets these phosphoserines for dephosphorylation. Thus, we generated single (S242A and S244A) and double (S242A-S244A) non-phosphorylatable mutants of MyD88, and found that constitutively active PP2Ac associated more strongly with WT MyD88 than with any non-phosphorylatable mutant (Figure 3D), thus validating these PP2Ac-recognizing sites in TL cells. Because these serines reside in the TLR4-interacting MyD88 TIR domain that is critical for initiating pro-inflammatory signaling (Wang et al., 2006) we reasoned that these LPS-inducible, PP2Ac-target phosphorylations are essential for stability of the signal-promoting MyD88-TLR4 complex. Thus, we examined the LPS-induced, time-dependent TLR4-MyD88 association, which reached its highest strength at 15 min following LPS stimulation before returning to the basal level 24hr after stimulation (Figure 3E, Left). We then generated an anti-phospho-Ser242 antibody (Figure S4C) to assay the timedependent level of this pro-inflammatory MyD88 phosphorylation (pSer242) that also reached the highest level at 15min and was reduced with prolonged LPS stimulation (Figure 3E, Middle). We then compared binding of TLR4 to WT MyD88 versus each mutant in co-transfected cells stimulated by LPS for 15 min. Compared with WT MyD88, binding of each non-phosphorylatable mutant to TLR4 was not only weaker but also not LPS-inducible (Figure 3E, Right). These results together revealed that site-specific phosphorylation(s) in the MyD88 TIR are directly involved in the pro-inflammatory TLR4-MyD88 interaction, also demonstrating the physiological accuracy of our quantitative proteomic method for identification of these sites first with 20nM OA treatment.

To determine the impact of the PP2Ac-mediated MyD88 dephosphorylation on T-class gene regulation, we transfected WT and the non-phosphorylatable MyD88 mutants into MyD88^−/− BMDMs. Coincident with the weakened binding of MyD88 mutants to TLR4 (Figure 3E), following a high-dose Pam3CSK4 stimulation (MyD88-dependent TLR2 agonist), a 50% reduction in TNFα was observed for the mutant-expressing cells compared to cells expressing WT MyD88 (Figure 3F). Clearly, disruption of the phosphorylation-dependent TLR4-MyD88 complex by constitutively active PP2Ac directly contributes to suppression of T-class genes.

Constitutively Active PP2Ac Co-localizes with MyD88 in the Nuclei of Tolerant Macrophages

Given that constitutively active PP2Ac facilitates the departure of MyD88 from the TLR4 complex through site-specific dephosphorylation that often regulates protein translocation(Deribe et al.), we then determined the ET-specific distribution of MyD88 and PP2Ac. Compared to their subcellular distributions in pro-apoptotic NL cells, both MyD88 and PP2Ac showed coincidently increased presence in the nuclei of the TL, pro-survival WT BMDMs (Figure 3G), in line with previous report that MyD88 was preferentially in nuclei of non-apoptotic cells(Jaunin et al., 1998). Further, immunoblotting revealed that more MyD88 and PP2Ac were retained in the nuclei of LPS-tolerant WT BMDMs than in naïve cells (Figure 3H). In contrast, in the PP2Ac KD cells under TL, MyD88 failed to accumulate in nuclei. Similarly, OA-mediated PP2Ac inhibition reversed survival-characteristic retention of MyD88. Immunofluoresence staining further indicated that more PP2Ac and MyD88 co-localized in nuclei of TL WT macrophages than those in non-stimulated cells (Figure 3I, yellow spots in the nuclei area), all indicating that chronic-active PP2Ac promotes survival-specific translocation of MyD88 to ET nuclei through MyD88 dephosphorylation.

Constitutively-Active PP2Ac Plays a MyD88-dependent Role in Establishing the Gene-specific Chromatin Modifications Responsible for Acquiring ET

We next explored the impact of the nuclear MyD88/PP2Ac accumulation on T-class gene regulation. First, along with the increasing PP2Ac in ET nuclei, the LPS-inducible increase of phospho-p65 was significantly suppressed in TL WT cells (Figure 3G, left lanes 2–4). Either PP2Ac KD or OA treatment dramatically reversed TL-specific suppression of p-p65, indicating chronic-active PP2Ac specifically dephosphorylates p-p65. Additionally, although p-p65 was suppressed, p65 was even higher in TL nuclei, indicating abnormal nuclear retention of non-phosphorylated p65 under ET. We then clarified the impact of PP2Ac-mediated p65 dephosphorylation on NFκB-regulated T-class genes and its MyD88-dependence: in contrast to the MyD88-dependent suppression of IL6/IL10 in the TL WT BMDMs, OA inhibition of PP2Ac activity reactivated their releases under LPS stimulation, also more so with Pam3CSK4 treatment for MyD88-dependent pathways (Figure 4A), all suggesting that, in a MyD88-dependent manner, chronic-active PP2Ac renders NFκB-dependent genes unresponsive by dephosphorylating nuclear p65 in ET macrophages.

Additionally, our phosphoproteomic findings of PP2Ac-target chromatin assembly/remodeling factors such as SWI/SNF (Figure S3B) implied that nuclear translocated, constitutively active PP2Ac directly participate in the gene-specific chromatin modifications. First, because histone H3 Ser10 phosphorylation (H3pS10) reflects the accessibility of NFκB to T-class genes and mediates NFκB-dependent transactivation of these genes (Saccani et al., 2002), we determined whether constitutively active PP2Ac affects this pro-inflammatory epigenetic marker. Compared with its NL-enhanced but TL-suppressed nature in irritated WT cells, H3pS10 remained high in PP2Ac KD cells under NL or TL (Figure 4B), indicating that constitutively active PP2Ac specifically targets H3pS10. Further, the LPS-inducible H3pS10 under NL was fully suppressed in TL WT BMDMs that was reversed by OA pre-treatment, whereas no such OA effect was observed for NL cells (Figure 4C). As well, no detectable change due to OA treatment was observed in MyD88^−/− BMDMs under TL, all indicating that in a MyD88-dependent manner, constitutively active PP2Ac dephosphorylates H3pS10 to selectively restrict accessibility of NFκB to T-class gene promoters.

We then determined the link of the PP2Ac-coordinated MyD88 signaling to two major characteristics of the gene-specific regulation, histone deacetylation and inhibition of chromatin remodeling, at T-class gene promoters under TL (Foster et al., 2007). First, the reversed suppression of IL-6 and IL-10 in the TL WT BMDMs pretreated with a HDAC inhibitor (SAHA) was not observed in the TL MyD88^−/− cells (Figure 4D), indicating that histone deacetylation is MyD88-dependent. Second, levels of both p-Ser10 and acetyl-lysine 9 on H3 showed similar OA-triggered MyD88-dependent changes (Figure 4C), indicating that in the MyD88 interactome chronic-active PP2Ac negatively regulates histone acetylation in a phosphorylation-dependent manner.

In combination, our findings of MyD88-dependent, PP2Ac-mediated p65 dephosphorylation and PP2Ac-target HDAC1 and chromatin remodeling complexes (Figure S3B) suggested that, in a phosphorylation-dependent manner, chronic-active PP2Ac directly regulates TLR4-induced recruitment of chromatin remodeling complexes to T-class gene promoters, a process is inhibited in tolerant macrophages. Thus, we examined the effect of PP2Ac KD on recruitment of an ATP-dependent chromatin remodeling complex, Brg1, to T-class promoters, which was found suppressed under TL(Foster et al., 2007). As revealed by ChIP experiments, either PP2Ac KD or OA-mediated PP2Ac inhibition similarly reversed the ET-specific suppression of Brg1 recruitment to the IL6 and Lipg promoters (Figure 4E). Further, this pro-inflammatory recruitment was heavily suppressed in the TL WT BMDMs but suppressed to a much lower degree in the TL MyD88^−/− BMDMs (Figure 4F). By contrast, while OA had little effect on Brg1 in the NL WT BMDMs, the TL WT BMDMs pre-treated with OA showed greater Brg1 recruitment. However, this OA-initiated Brg1 recruitment was not observed in the TL MyD88^−/− BMDMs. These results indicated that chronic-active PP2Ac co-translocated with MyD88 negatively regulates the chromatin modification/remodeling required for T-class gene transactivation.

Proteomic Identification of TL-specific MyD88 Interactors

The details about sample preparation, nano-LC-MS/MS analysis, AACT-based quantitative proteomic identifications of NL-or TL-specific MyD88-interacting partners are given by EXTENDED EXPERIMENTAL PROCEDURES.

BMDM Stimulation and OA or SAHA Treatment

LPS stimulations of differentiated BMDMs were similar as to RAWs. In Pam3CSK4 stimulations the priming dose was at 50ng/ml and a high dose was 500ng/ml. SAHA was introduced into media 24 hrs before stimulation and 20nM OA was added along with the stimuli.

PP2A Activity Assay

PP2Ac was immunoprecipitated (IP) using anti-PP2Ac antibody with protein A agarose from 0.5–1.0 mg of each lysate of differentially stimulated cells. The phosphatase activity of IP PP2Ac was measured by Phosphatase Assay Kit (Millipore) and normalized based on the amount of PP2Ac in the IP products.

Generation of the RAW stable lines with either shRNA-GFP or –PP2Ac

The lentiviral plasmids pLKO.1 expressing the shRNA-PP2Ac targeting sequences CCAGATACAAATTACCTGTT and CGACGAGTGTTTAAGGAAATA were purchased from Thermo Scientific. To produce virus, pLKO.1-shRNA plasmids were co-transfected into 293T cells with ViraPower Mix (Invitrogen) by jetPRIME™ in vitro DNA & siRNA transfection reagent (Polyplus). Pseudo-virus in supernatants was collected after 48 hrs following transfection and used to transduce RAWs by spinoculation. 48 hrs after transfection, 8µg/ml puromycin was added to select puromycin-resistant clones. Stable clones were maintained in the media containing 4µg/ml puromycin.

Identification of LPS-Inducible, PP2Ac-target Phosphorylations at MyD88

Due to the low abundance of endogenous MyD88 in RAW cells few phosphorylation sites could be identified/quantified for their LPS-inducible changes. Therefore, quantitative proteomic identification of novel phosphorylation sites on MyD88 and the analysis of LPS or OA effects on these sites were conducted on the AACT-tagged HEK 293 cells stably expressing TLR4-MD2- CD14. The details are given by EXTENDED EXPERIMENTAL PROCEDURES.

Retrovirus Infection and Cytokine Assay

Each cDNA of wild-type MyD88 or a MyD88 mutant was inserted into the MIG vector (Wang et al., 2010). Viral particles were generated in 293T cells and infected bone marrow cells 3 times. 5 days after infection, matured BMDM cells were processed for LPS stimulation as described above.

Analysis of LPS-induced Binding between TLR4 and Wild-type or Mutant Myd88

HEK-TLR4 stable cells were co-transfected with both YFP-TLR4 and HA-WT or a mutant MyD88 by using jetPRIME (Polyplus). The cells were stimulated with 1µg/ml LPS for indicated time and harvested. The cells were lysed in the buffer (50mM Tris, pH8.0, 150mM NaCl, 1mM EDTA, 1% Triton, protease inhibitor and phosphatase inhibitors (Sigma). 1mg lysate was incubated with the anti-GFP IgG coupled on Protein A magnetic beads for co-IP and immunoblotting.