Novel function of PiT1/SLC20A1 in LPS-related inflammation and wound healing

PiT1/SLC20A1 is an inorganic phosphate transporter with additional functions including the regulation of TNFα-induced apoptosis, erythropoiesis, cell proliferation and insulin signaling. Recent data suggest a relationship between PiT1 and NF-κB-dependent inflammation: (i) Pit1 mRNA is up-regulated in the context of NF-κB pathway activation; (ii) NF-κB target gene transcription is decreased in PiT1-deficient conditions. This led us to investigate the role of PiT1 in lipopolysaccharide (LPS)-induced inflammation. MCP-1 and IL-6 concentrations were impaired in PiT1-deficient bone marrow derived macrophages (BMDMs) upon LPS stimulation. Lower MCP-1 and IL-6 serum levels were observed in Mx1-Cre; Pit1lox/lox mice dosed intraperitoneally with LPS. Lower PiT1 expression correlated with decreased in vitro wound healing and lower reactive oxygen species levels. Reduced IκB degradation and lower p65 nuclear translocation were observed in PiT1-deficient cells stimulated with LPS. Conversely, PiT1 expression was induced in vitro upon LPS stimulation. Addition of an NF-κB inhibitor abolished LPS-induced PiT1 expression. Furthermore, we showed that p65 expression activated Pit1 promoter activity. Finally, ChIP assays demonstrated that p65 directly binds to the mPit1 promoter in response to LPS. These data demonstrate a completely novel function of PiT1 in the response to LPS and provide mechanistic insights into the regulation of PiT1 expression by NF-κB.

which is dependent on the rapid activation of the NF-κB pathway and the subsequent transcription of NF-κB target pro-inflammatory genes such as Tnfα and Il-6 16,20,21 , and other studies indicate that PiT1 expression is regulated by induced or basal activity of NF-κB [22][23][24] . Moreover, Pit1 mRNA levels are increased in the livers of mice when the NF-κB pathway is upregulated due to the deletion of one of its regulators, the Von Hippel-Lindau protein (pVHL) 24 . Thirdly, our group has recently investigated the role of PiT1 in liver regeneration in vivo using the model of liver regeneration following 2/3 rd hepatectomy (PH). During the first hours following PH, mice heterozygous for a deletion in Pit1 (Pit1 +/∆5 ) had lower hepatic Il-6 mRNA levels and lower serum IL-6 compared to control mice. IL-6 is a known NF-κB target gene. Mice with liver-specific Pit1 deletion (the Alb-Cre; Pit1 lox/lox mice) had normal cytokine production during this phase (unpublished data). This led us to hypothesize that the impairment in cytokine production in Pit1 +/∆5 mice may be due to lack of PiT1 in macrophages rather than in hepatocytes. NF-κB is an inducible transcription factor 25 . Since its discovery in 1986 26 , the NF-κB pathway has been shown to be involved in multiple biological functions including cell adhesion, differentiation, proliferation 27 , autophagy, senescence 28,29 , and protection against apoptosis 30 . The most important and evolutionarily conserved role of NF-κB is as a mediator of the immune and inflammatory response 25 . Considering these data, and to elucidate novel aspects of PiT1 function, we sought to investigate whether PiT1 plays a role in the NF-κB-mediated inflammatory response in macrophages and to examine PiT1 regulation by NF-κB.

PiT1 depletion is associated with lower Mcp-1 mRNA and MCP-1 protein levels in vitro.
Murine bone marrow-derived macrophages (BMDMs) were obtained from Mx1-Cre; Pit1 lox/lox and control mice. Mean Pit1 mRNA levels in macrophages, as assessed by RT-qPCR, were reduced by 94.3% ± 0.7 (80 to 98%) in the Mx1-Cre; Pit1 lox/lox mice compared to the controls (Fig. 1A). The mRNA expression and supernatant concentrations of cytokines and chemokines known to be induced by LPS were studied before and after LPS stimulation of the BMDMs for the indicated times. PiT1-deficient macrophages had lower levels of Mcp-1 mRNA (Fig. 1B), and the MCP-1 protein concentration in the supernatant of PiT1-deficient macrophages was lower than in the supernatant of control macrophages following stimulation with 10 ng/ml LPS (Figs 1C and S1D). IL-6 protein levels were also significantly lower in supernatants of PiT1-deficient BMDMs after LPS stimulation than in controls (Figs 1C and S1E). Although not significant, similar decreases after LPS treatment were observed for Tnfα and Il-6 mRNA levels between PiT1-deficient and control BMDMs (Figs 1B and S1B,C). In order to exclude the possibility that our results were caused by a differential expression of LPS receptor TLR4 between PiT1-deficient and control cells, Tlr4 mRNA expression was evaluated and no difference was observed (Fig. S2).
We also studied the LPS-induced cytokine and chemokine response in mouse embryonic fibroblasts (MEFs) isolated from wild-type (WT) and Pit1-KO embryos. MEFs were chosen because these cells do not express endogenous Pit1 due to the gene deletion ( Fig. 1D) and because they express toll-like receptors (TLR) and thus are able to transduce LPS-TLR signaling 31 . As observed in BMDMs, a lower MCP-1 protein concentration was observed in the supernatant of Pit1-KO MEFs following LPS stimulation than in WT cells (Figs 1F and S1H). Although Tnfα and Il-6 mRNA levels appeared to be lower after stimulation in Pit1-KO cells, these results were not significant (Fig. S3A,B) and no differences in TNFα or IL-6 concentrations were observed (Fig. S3C,D). Taken together, these results suggest that PiT1 depletion decreases LPS-induced MCP-1 concentration in vitro.
Vigorous production of pro-inflammatory cytokines such as IL-6 and TNFα is a hallmark of classically activated macrophages, also known as effector or M1 macrophages 32 . We next investigated whether PiT1 depletion could also be responsible for a modification in M2 phenotype. M2 gene Il-10 was expressed at a significantly lower level upon 2 h LPS stimulation in PiT1-deficient BMDMs than in control BMDMs, suggesting an effect of PiT1 depletion on M2 gene expression (Fig. S4).

Loss of PiT1 modulates LPS response in vivo.
Considering the lower overall LPS-induced inflammatory response observed in PiT1-deficient BMDMs and Pit1-KO MEFs, we sought to investigate the effects of PiT1 deficiency in vivo. For this purpose, we performed intraperitoneal injections of LPS (0.5 µg/g) or, as a control, PBS at 24 h before sacrifice of 16 week-old Mx1-Cre; Pit1 lox/lox and control mice. As expected, LPS injection caused an increase in MCP-1, IL-6, and TNFα serum concentrations. As observed in vitro in BMDMs, serum levels of MCP-1 and IL-6 were significantly lower in Mx1-Cre; Pit1 lox/lox than in control mice after LPS injection (Fig. 2). No significant differences were observed for TNFα and IL-1 concentrations between Mx1-Cre; Pit1 lox/lox and control mice. Serum concentrations of IL-4, classically described as a M2-activating cytokine, were not significantly different between Mx1-Cre; Pit1 lox/lox and control mice. The baseline serum IL-10 level was significantly higher in Mx1-Cre; Pit1 lox/lox mice than in control mice; however, no difference was observed between Mx1-Cre; Pit1 lox/lox and control mice in the LPS-stimulated condition. These results show impaired concentrations of MCP-1 and IL-6 following LPS stimulation in the absence of PiT1.
PiT1 influences macrophage function. Wound healing in vitro. MCP-1 plays a major role in the chemo-attraction of monocytes and macrophages during inflammatory conditions 33 and is known to contribute to wound healing 34 . We therefore assayed the effects of PiT1 depletion on the migratory and chemoattractant RT-qPCR analysis of Mcp-1 mRNA expression and (F) ELISA quantification of MCP-1 in the supernatant of WT MEFs (black bars) and Pit1-KO MEFs (white bars) stimulated with 100 ng/ml LPS for the indicated times. Data are means ± S.E.M. of at least three independent experiments. Student's unpaired t-test or an Unpaired t-test with Welch correction for groups with unequal variance was performed; # indicates comparison with the untreated condition; *indicates comparison between control and PiT1-deficient cells; *p < 0.05; **p < 0.01; ***p < 0.001; # p < 0.05; ## p < 0.01; ### p < 0.001. abilities of macrophages in vitro. Wounds generated in PiT1-deficient BMDMs healed more slowly and incompletely than wounds across control BMDM cultures (Fig. 3). This result suggests that PiT1 deficiency has functional consequences on macrophage migration and wound repair abilities perhaps by modulating MCP-1 concentration.  Thioglycollate-induced peritonitis. MCP-1 is necessary for the recruitment of monocytes in several models of experimental peritonitis 35 . Thioglycollate-induced peritonitis is characterized by substantial inflammation and accumulation of inflammatory macrophages 36 . MCP-1 is the primary chemokine required for monocyte recruitment in mouse peritonitis induced with thioglycollate, and the induction of endogenous MCP-1 in this system is highly macrophage-dependent 35 . Peritonitis was induced by injection of 1 ml of 4% thioglycollate brewer medium into the peritoneal cavities of 12-16 week-old mice; controls were dosed with PBS. After 72 h, monocyte peritoneal infiltration was analyzed. The number of cells recruited after thioglycollate injection was significantly higher than after PBS injection (data not shown). We then quantified and sorted peritoneal macrophages by flow cytometry and identified resident and recruited macrophages, which differ by the intensity of CD11b fluorescence. No difference was observed in the number of recruited macrophages between Mx1-Cre; Pit1 lox/lox and control mice after thioglycollate injection (Fig. S5).
PiT1 depletion decreases reactive oxygen species (ROS) production. Inflammatory stimulants such as LPS induce the generation of ROS in macrophages [37][38][39] . ROS are involved in bacterial killing and cytokine production. When we stimulated BMDMs with 1000 ng/ml LPS hydrogen peroxide (H 2 O 2 ) production was higher than in unstimulated cells after 20 h as assessed using the general oxidative stress indicator CM-H2DCFDA. In comparison with control BMDMs, the production of H 2 O 2 in PiT1-deficient BMDMs was significantly lower (Fig. 4A,B), suggesting that PiT1 contributes to ROS production. NOX2 is the NADPH oxidase primarily expressed in macrophages; therefore, we next examined whether PiT1 deficiency correlates with differences in NOX2 activity. NOX2 activation in macrophages was assessed using the bacterial peptide formyl-methionyl-leucyl-phenylalanine (fMLF) to stimulate the cells and a luminol plus HRP-amplified chemiluminescence assay to monitor NOX2 activation. We found that ROS production by PiT1-deficient BMDMs was dramatically decreased compared to BMDMs from control mice (Fig. 4C,D). As ROS are known to participate in killing of pathogens after phagocytosis 40 , we also investigated whether phagocytosis was altered in PiT1-deficient BMDMs. No difference was observed in the phagocytic abilities of PiT1-deficient and control BMDMs (Fig. S6), suggesting that although ROS production is lower in the absence of PiT1, this does not necessarily impair the phagocytosis process in vitro.

LPS-induced PiT1 expression correlates with NF-κB target gene induction.
We next sought to investigate the consequence of LPS stimulation on PiT1 expression. Similar to the increase in Mcp-1, Il-6, and Tnfα, Pit1 mRNA and PiT1 protein expression increased in BMDMs upon LPS stimulation (Fig. 5A,B). The LPS-induced Pit1 expression occurred rapidly (Fig. 5A) in the same time frame as increases in Mcp-1, Il-6, and Tnfα mRNAs and was LPS dose-dependent (Fig. 5C). Similar results were found in MEFs (Fig. 5D,E). These findings show that PiT1 expression is induced by LPS treatment and suggest that Pit1 gene expression might be regulated by transcription factors involved in the LPS-TLR pathway.
In contrast to the effect of LPS on Pit1 mRNA expression, we did not observe any LPS effect on Pit2 mRNA expression in control BMDMs (Fig. S7). Interestingly, Pit2 mRNA expression tended to increase in PiT1-deficient BMDMs following LPS stimulation, and significantly higher Pit2 expression was observed in PiT1-deficient cells compared to control BMDMs following LPS 4 h stimulation, suggesting a possible compensatory effect for the absence of PiT1. Nevertheless, this possible compensation did not prevent the effects of PiT1 depletion on MCP-1 and IL-6 levels indicative of a PiT1-specific response to LPS.

LPS-induced Pit1 expression in macrophages is NF-κB-dependent.
Since Pit1 expression increased following LPS treatment, we chose to study the regulation of Pit1 by transcription factors involved in the LPS-TLR pathway. We focused on NF-κB, since Mcp-1, Il-6, and Tnfα are all known NF-κB target genes. The in silico analysis of the 5-kilobase mouse Pit1 promoter (mPit1p) revealed six putative binding sites for NF-κB and two binding sites for AP1 (Fig. 6A). mPit1p was subcloned upstream of the luciferase gene (mPit1p-LUC) as previously described 13 , and the luciferase activity was measured in HEK293 cells transfected with this vector and a plasmid or combination of plasmids encoding p65, p105, c-JUN, or c-FOS. The expression of p65 increased mPit1p activity by more than 4-fold, and the combination of p105 and p65 increased mPit1p activity by 9-fold ( Fig. 6B), demonstrating that Pit1 expression is regulated by the NF-κB pathway. No increase was observed after cotransfection of mPit1p with AP1 ( Fig. 6C).
To confirm that Pit1 regulation is NF-κB dependent, we stimulated MEFs with 100 ng/ml LPS in the presence of a selective pharmacological inhibitor of NF-κB, BAY11-7082. The NF-κB inhibitor blocked upregulation of Pit1 mRNA (Fig. 6D) and PiT1 protein (Fig. 6E). This confirms that Pit1 is regulated by NF-κB upon LPS stimulation. Similar results were obtained when BMDMs were treated with 10 ng/ml LPS and BAY11-7085 (Fig. S8). Using a chromatin immunoprecipitation (ChIP) assay, we investigated whether there is a direct interaction of NF-κB with the Pit1 promoter. ChIP experiments confirmed the direct binding of p65 to the proximal region of the Pit1 promoter (Fig. 6F).

PiT1 depletion is associated with impaired NF-κB activation.
To better assess how PiT1 influences the expression of NF-κB target genes, we next examined the NF-κB pathway following LPS stimulation in control and PiT1-deficient BMDMs as well as in WT and Pit1-KO MEFs. Whereas control BMDMs and MEFs displayed the expected degradation of IκBα following LPS treatment, PiT1-deficient BMDMs and MEFs showed an impairment in IκBα degradation (Fig. 7A,B), suggesting that the NF-κB pathway may be less activate in PiT1-deficient cells than in WT cells. Consistent with this result, we found a lower p65 signal in the nuclei of Pit1-KO MEFs following LPS stimulation than in WT nuclei (Fig. 7C,D), suggesting that p65 nuclear translocation is impaired in the absence of PiT1.

Discussion
In the present work, and as summarized in Fig. 8, we showed that chemokine Mcp-1 mRNA levels and MCP-1 concentration were impaired in vitro in BMDMs depleted of PiT1 (also known as Slc20a1) and in Pit1-KO MEFs upon LPS stimulation and in vivo in the LPS-induced inflammation model in Mx1-Cre; Pit1 lox/lox mice compared to control mice. Furthermore, major functions such as ROS production and wound healing were impaired in PiT1-deficient BMDMs. Notably, reduced IκB degradation and lower p65 nuclear translocation were observed in Pit1-KO MEFs upon LPS stimulation, suggesting an impact of PiT1 depletion on NF-κB pathway activation. We  found that LPS induces the transcription of Pit1 in an NF-κB-dependent manner. Indeed, the physiological activation of the NF-κB pathway by LPS triggers PiT1 expression. Additionally, the disruption of the NF-κB pathway with pharmacological inhibitors of NF-κB abolishes Pit1 upregulation. Importantly, our study provides the first experimental evidence that p65/NF-κB is able to transactivate the Pit1 promoter: LPS treatment resulted in the recruitment of endogenous p65 to the Pit1 promoter.
Little is known about the regulation of Pit1 transcription. Our group has previously demonstrated that Pit1 is a target gene of EKLF, a transcription factor involved in erythroid differentiation 13 . Although global surveys (microarray and computer-based analysis) identified hundreds of potential NF-κB-responsive genes 41,42 , listed in the following websites: http://www.bu.edu/nf-kb/the-gilmore-lab/ and http://bioinfo.lifl.fr/NF-KB. Pit1 is not among these. The upregulation of Pit1 mRNA in response to diverse treatments known to activate the NF-κB pathway, including TNFα (of pre-adipocytes 22 ), IL1α or PMA (of hematopoietic cell lines 23 ), IGF1 (of immortalized fibroblasts 43 ), and TGFβ or oxidized low density lipoproteins (of chondrocytes 44 ) has been previously described. Interestingly, PiT1, but not PiT2, was identified as a possible NF-κB target gene in the liver after VHL inactivation 24 and is present in the list of genes induced by TNFα 22 , suggesting a PiT1-specific induction of NF-κB mediated inflammation. In the present study, we did not observe an effect of LPS on Pit2 expression. The higher expression of Pit2 that we observed in PiT1-deficient BMDMs following LPS stimulation suggests a compensatory increase in Pit2, which does not, however, prevent the effects of PiT1 depletion on MCP-1 and IL-6 levels. Further experiments will be necessary to determine whether these effects are Pi-transport dependent or not.
We also provide evidence showing that PiT1 is involved in the LPS response in vivo, as demonstrated by the lower amounts of IL-6 and MCP-1 in the serum of Mx1-Cre; Pit1 lox/lox mice after LPS stimulation, which suggests that PiT1 may play a role in innate immunity. Proinflammatory cytokines and chemokines are known NF-κB target genes (http://www.bu.edu/nf-kb/the-gilmore-lab/) and are involved in a wide variety of pathological and inflammatory conditions 33 including rheumatoid arthritis, multiple sclerosis, and glomerulonephritis [45][46][47] .
Our results also provide evidence showing that PiT1-deficient macrophages have reduced ROS production. High levels of ROS can lead to cellular damage, oxidative stress, and DNA damage. In phagocytic cells such as macrophages, ROS production, which is necessary for bactericidal action, is mainly catalyzed by the action of the NADPH oxidase NOX2, a membrane-bound enzyme complex 48 . The NF-κB pathway is a major regulator of ROS production; it fine-tunes the expression of anti-oxidant and pro-ROS genes 49 . ROS also influence the NF-κB pathway 48 . As both PiT1 and NOX2 are expressed at the cell surface, PiT1 and the NOX complex may interact directly. It is also possible that a decrease in NOX2 gene transcription regulated by NF-κB, or expression of NF-κB pro-inflammatory target genes such as Tnfα, which enhance ROS production in an autocrine manner 50 , could explain at least in part the reduced ROS observed in PiT1-deficient cells in the present work.
Importantly, PiT1 was identified as a direct positive modulator of the NF-κB pathway in a global screen 51 . In this study, we found a reduced nuclear translocation of p65 in PiT1-deficient cells, which is consistent with the altered degradation of IκBα in these cells. Although the exact mechanism has not been elucidated, this result suggests that PiT1 interferes with the activation of NF-κB pathway following LPS treatment.
Many of the signaling events leading to cytokine synthesis and release following LPS exposure are now well established [52][53][54][55] . The transcription factor NF-κB is critical for the expression of these inflammatory proteins and is regarded as the master regulator of the immune response 56 . The transcriptional activity of NF-κB is primarily regulated through its sequestration in the cytoplasm by the IκB family proteins 57 . Upon stimulation by a toll-like receptor ligand or TNFα, IκBα is phosphorylated by the IKK complex, which targets it for ubiquitination and subsequent proteasome-mediated degradation. NF-κB dimers then translocate into the nucleus and bind to regulatory regions of target genes 25 . The ubiquitination and proteasomal degradation of NF-κB is critical in the termination of the NF-κB transcriptional response and represents a major limiting factor in the expression of proinflammatory genes [58][59][60] . Ubiquitination and deubiquitination processes are also required at other steps of the NF-κB pathway. Indeed, LPS-induced activation of NF-κB is known to require the activating K63 and M1 ubiquitination of NEMO and TRAFs proteins 25 .
Interestingly, a yeast-two hybrid screen performed by our group identified putative interactions of PiT1 with several ubiquitinating and deubiquitinating enzymes such as ligases UBC9 and SIAH2 and protease USP7. USP7 proteolytically removes polyubiquitin chains from substrates 61 . Among the identified targets of USP7 are tumor suppressor protein P53 62 , its regulator MDM2 63 , and PTEN 64 . Importantly, NF-κB p65 was recently identified as a substrate for USP7 deubiquitinase activity in the nucleus, where USP7 is recruited to NF-κB target promoters and interacts with NF-κB in a DNA-binding-dependent manner 65 . USP7 deubiquitination of NF-κB leads to increased transcriptional activity and expression of target genes in response to TLR and TNFα-receptor activation. In a recent work, we confirmed that USP7 binds to PiT1 and demonstrated that Pit1 deletion inhibited USP7/IRS1 dissociation upon insulin stimulation 15 . This prevented IRS1 ubiquitination and its subsequent proteasomal degradation. Among the other suggested partners of USP7 is RIP1 66 . RIP1 is known to be involved in the LPS-TLR4 pathway, and its ubiquitination is a crucial checkpoint that determines whether NF-κB-mediated transcription is activated. TRIF adaptor protein, indirectly via TRAM, binds to RIP1 and TRAF6 following presented in Supplementary Fig. 9. (F) q-PCR analysis of p65 ChIP experiments. DNA immunoprecipitated with p65 from MEFs treated (black bar) or not (white bar) with 100 ng/ml LPS for 90 min was analyzed by quantitative PCR using primers located along the mouse Pit1 promoter. Negative controls (grey bars) were performed using DNA incubated with beads but without anti-p65 antibody. Cxcl2 was used as positive control. Means ± S.E.M. of three experiments are presented. Student's unpaired t-test or an Unpaired t-test with Welch correction for groups with unequal variance was performed *indicates significant difference from untreated control cells at p < 0.05. receptor activation, which can lead to the activation of NF-κB-and MAPK-mediated signaling 67 . Further work will be needed to decipher the molecular mechanism underlying the impact of PiT1 on the NF-κB pathway and possible interactions with ubiquitinating and deubiquitinating enzymes. Our study does have limitations. Our previous research showed that, in an allelic series of mutant mice, a phenotype was observed only when Pit1 expression was reduced to below 15% of WT levels 4 . Moreover our in vitro studies using siRNA-mediated gene silencing revealed that PiT1-depletion must be very effective to observe phenotype (data not shown). For this reason, we decided to use MEFs or BMDM Pit1-knockout cells instead of cells in which Pit1 was depleted using siRNA. Pit1 gene expression was checked in each experiment involving the Mx1-Cre model. BMDMs expressing Pit1 more than 5% of the controls were excluded. Furthermore, with the Mx1-Cre system, Pit1 is deleted not only in macrophages but also in other organs. This may explain some discrepancies between in vitro and in vivo data. In particular, we observed higher IL-10 serum level in Mx1-Cre; Pit1 lox/lox mice compared to controls, whereas IL-10 was found to be lower in vitro in PiT1-deficient BMDMs. This correlates with the phenotype of Mx1-Cre; Pit1 lox/lox mice characterized by liver inflammation (color change and increased amino acid transferases, data not shown), which suggests that IL-10 is increased in this in vivo model to dampen excessive inflammation. Therefore it is difficult to draw conclusions on IL-10 and to attribute all the in vivo results to PiT1 depletion in macrophages only. Although the Mx1-Cre system is active in numerous organs, we observed consistently lower Mcp-1 mRNA and MCP-1 protein levels in PiT1-deficient macrophages, as well as lower MCP-1 in the serum of Mx1-Cre; Pit1 lox/lox mice. This suggests that the in vivo effects are likely due in part to PiT1 depletion in macrophages.
In summary, we have uncovered a previously unsuspected role for PiT1 in the LPS-mediated inflammatory response and provide mechanistic insights into the regulation of PiT1 expression by NF-κB. We showed that LPS-induced Pit1 upregulation is NF-κB dependent. Thus, PiT1 may enhance the production of pro-inflammatory and chemoattractant mediators and ROS. In future studies, it will be interesting to investigate the physiological impact of NF-κB-dependent PiT1 up-regulation in pathophysiological contexts. Pit1 genetically modified mice were generated as described in 4 . Pit1 lox/+ animals were intercrossed to Pit1 lox/lox mice. Pit1 lox/lox 4 mice were then crossed with Mx1-Cre mice 68 , provided by Dr. Thomas Mercher (INSERM U985, Université Paris XI, Villejuif, France) to obtain the conditional Pit1 strain Mx1-Cre; Pit1 lox/lox . Mx1-Cre; Pit1 lox/lox mice were then bred with Pit1 lox/lox mice in order to obtain controls (Pit1 lox/lox ) and experimental animals (Mx1-Cre; Pit1 lox/lox ) from the same litters. The presence of the Mx1-Cre transgene was confirmed by PCR using the following primers: Forward 5′-AGCCTGGGGGTAACTAAACTGG-3′, Reverse 5′-CCATTGCCCCTGTTTCACTATC-3′. Tissuespecific deletion of Pit1 was induced by three intraperitoneal injections of polyinosinicpolycytidylic acid (pIpC, #tlrl-picw, Invivogen) at 2-day intervals at 5 weeks of age 68 , and mice were analyzed at least 6 weeks after injection to avoid acute side effects due to IFN production. Control mice were also injected with pIpC to normalize for potential sustained effects resulting from pIpC treatment.
Cells and cell culture conditions. For the generation of BMDMs, femurs and tibiae of 12 to 16-week-old mice were flushed with RPMI-1640 medium (Gibco, 61870-010). BMDMs were selected by adhesion to petri dishes after 7 days of differentiation in RPMI-1640 supplemented with 10% fetal bovine serum (FBS) and 20% M-CSF-containing medium 69 . M-CSF-containing medium was obtained by harvesting conditioned medium from L929 mouse fibroblast cells at confluency and at 7 days past confluency. For any further experimental procedures, cells were detached, counted, seeded and cultivated in RPMI-1640 with 10% FBS without L929-conditioned medium. For ROS procedures, RPMI-1640 without phenol red (Gibco, 11835) was used.
Cells were stimulated with complete medium containing lipopolysaccharide (LPS) from E. coli strain O111:B4 (Invivogen, #tlrl-3pelps) at a concentration of 10 ng/ml for BMDMs or 100 ng/ml for MEFs as described previously 31,70 and following preliminary tests suggesting that a 10 ng/ml dose in BMDMs and a 100 ng/ml dose in MEFs were the optimal concentrations to activate the NF-κB pathway (assessed by IκB degradation) and to induce pro-inflammatory cytokine expression in these two different cell types. In experiments using the NF-κB

Isolation of macrophages from the peritoneal exudate. To induce inflammatory exudates, male
Mx1-Cre; Pit1 lox/lox and control mice were injected intraperitoneally at 12-16 weeks of age with 1 mL of 4% brewer thioglycollate medium (Sigma, B2551) or the same volume of PBS. These mice were sacrificed 72 h later, and the peritoneal exudate cells were obtained by injecting 5 mL of ice-cold 0.5% BSA in PBS into the peritoneal cavity. After a gentle massage, the fluid was harvested, and the cells were centrifuged and washed. Peritoneal cells were FACS-phenotyped as described below.

Measurement of ROS production by luminol-amplified chemiluminescence. ROS production was
measured by the luminol-amplified chemiluminescence method. Briefly, BMDMs from Mx1-Cre; Pit1 lox/lox and control mice (2.5 × 10 5 cells) isolated as described above were resuspended in 500 µl of HBSS and pre-incubated 10 min at 37 °C in the presence of 10 µM luminol and 5 U HRP then stimulated with fMLF at a final concentration of 10 −5 M. Chemiluminescence was measured using a luminometer (Biolumat LB937; Berthold), which converts light intensity into counts per minute (cpm).
To correct for transfection efficiency, all cells were cotransfected with 100 ng of the pRL-tk plasmid (Promega, E2241), expressing the Renilla reniformis luciferase. The assay was performed as described previously 13 . The cDNAs encoding p65 and p105 were purchased from Biovalley and subcloned into a pSPORT6 expression plasmid (pSPORT6-p65, pSPORT6-p105). Negative controls were performed by cotransfecting cells either with the mPit1p-LUC construct and an empty expression vector or with an empty pGL3-LUC plasmid and a transcription factor expression vector. An artificial construct containing three NF-κB sites upstream of LUC was used as positive control (NF-κB-LUC) for NF-κB Gene expression analysis. Total RNA was isolated from cells using Nucleospin RNA II columns (Macherey Nagel, 740955.250). RT-PCR amplifications were performed using M-MLV (Invitrogen, 28025-013) according to manufacturer's instructions. Q-PCR was performed using SyBr Green chemistry (Thermo Fischer Scientific).
The Pinin gene was used as the reference gene, and levels were calculated using the 2 −ΔΔCT method and expressed relative to the mean value of untreated samples 71 .
Chromatin immunoprecipitation assay. ChIP assays were performed as described 13 . In brief, the chromatin was fragmented into approximately 500-to 1000-bp fragments by sonication (5 × 1 s, with 0.5 s intervals) at intensity 30% (Branson Digital Sonifier), and then 20 μg of sonicated DNA was incubated overnight at 4 °C with 2 μl of anti-p65 antibody (Cell Signaling, CS8242). The control for non-specific DNA immunoprecipitation was produced by amplifying a fragment of the β-actin gene. The DNA-containing supernatant was analyzed by qPCR. Primer pairs for amplification of the mouse Pit1 promoter region containing NF-κB putative binding sites are detailed in Table 1. Relative occupancy values were calculated by determining the apparent immunoprecipitate efficiency as the ratio of the amount of immunoprecipitated DNA over that of the input sample.

Statistical analyses.
Results are presented as means ± S.E.M. For statistical analyses, significance was tested using the Student's unpaired t-test or an Unpaired t-test with Welch correction for groups with unequal variance, or the Mann-Whitney Rank Sum test when data did not follow a normal distribution. The ANOVA test and Tukey's multiple comparison test were performed for time-courses. In all cases, the level of statistical significance was set at p < 0.05.

Data Availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information Files) or are available from the corresponding author on reasonable request.