HIF1α-dependent glycolysis promotes macrophage functional activities in protecting against bacterial and fungal infection

Macrophages are important innate immune defense system cells in the fight against bacterial and fungal pathogenic infections. They exhibit significant plasticity, particularly with their ability to undergo functional differentiation. Additionally, HIF1α is critically involved in the functional differentiation of macrophages during inflammation. However, the role of macrophage HIF1α in protecting against different pathogenic infections remains unclear. In this study, we investigated and compared the roles of HIF1α in different macrophage functional effects of bacterial and fungal infections in vitro and in vivo. We found that bacterial and fungal infections produced similar effects on macrophage functional differentiation. HIF1α deficiency inhibited pro-inflammatory macrophage functional activities when cells were stimulated with LPS or curdlan in vitro or when mice were infected with L. monocytogenes or C. albicans in vivo, thus decreasing pro-inflammatory TNFα and IL-6 secretion associated with pathogenic microorganism survival. Alteration of glycolytic pathway activation was required for the functional differentiation of pro-inflammatory macrophages in protecting against bacterial and fungal infections. Thus, the HIF1α-dependent glycolytic pathway is essential for pro-inflammatory macrophage functional differentiation in protecting against bacterial and fungal infections.

required for an inflammatory response while inhibiting Myc-dependent proliferation 19 . However, the significance of HIF1α-glycolytic pathway signals in protecting against different kinds of pathogenic microorganism infections remains unclear.
In this study, we investigated and compared the different macrophage functional effects of bacterial and fungal infections in vitro and in vivo. We found that bacterial and fungal infections exhibited similar effects on macrophage phenotypic differentiation. HIF1α deficiency inhibits pro-inflammatory macrophage functional differentiation in cells stimulated with LPS or curdlan 20,21 (a homopolymer of D-glucose and the major cellular component of fungal extracts of C. albicans) in vitro or in mice infected with Listeria monocytogenes (L. monocytogenes) or C. albicans in vivo by decreasing pro-inflammatory TNFα and IL-6 secretions associated with pathogenic microorganism survival. Alteration of glycolytic pathway activation was required for the differentiation of pro-inflammatory macrophages needed in protecting against bacterial and fungal infections, which should contribute to our understanding of targeted therapy in protecting against different pathogenic microorganism infections in clinical settings.

Materials and Methods
Mice. All animal experiments were performed in accordance with protocols approved by the Animal Ethics Committee of Fudan University (Shanghai, China) or (CLS-EAW-2017-003) Beijing Normal University (Beijing, China). Hif1α flox/flox and Lyz-Cre mice were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). C57BL/6 mice were obtained from the Fudan University Experimental Animal Center (Shanghai, China) or the Weitonglihua Experimental Animal Limited Company (Beijing, China). All mice had been backcrossed to the age of 6-12 weeks. WT control mice were of the same genetic background and, where relevant, included Cre + mice to account for the effects of Cre (no adverse effects due to Cre expression itself were observed in vitro or in vivo).
Bacterial and fungal infection. For bacterial infections, mice were injected intravenously with 1 × 10 5 CFU of L. monocytogenes. At 48 h, mice were killed for analysis, as described previously 16 . Infected liver or spleen samples were fixed in 4% paraformaldehyde, embedded in paraffin and stained with H&E. Spleens were teased with a 32 G needle using RPMI 1640 containing 5% FBS in a petri dish and digested in RPMI 1640 containing collagenase D (400 U/ml; Roche) in a shaking water bath at 37 °C for 35 min. Subsequently, 0.5 ml of 0.1 M EDTA was added, and samples were incubated on ice for 5 min and then transferred to a 50 ml tube by filtering through a 70 µm cell strainer. After a centrifugation step, the supernatant was removed and splenic cells were stained with fluorescent antibodies and macrophages (CD11b + , F4/80 + , TCR − , CD19 − , NK1.1 − , Ly6G − , and 7-AAD − ) were sorted on a FACSAria II (Becton Dickinson). Peritoneal cells were harvested from mouse peritoneal cavities by lavage and the peritoneal exudate macrophages (PEMs; 7AAD − , CD11b + , and F4/80 + ) were sorted on a MoFlo XDP (Beckman Coulter). The mouse macrophages were stimulated with LPS (100 ng/mL) for 6 h for subsequent intracellular staining, ELISA, or RNA isolation. For fungal infections, mice were injected intravenously with 2 × 10 5 of live C. albicans yeast in vivo. After 9 days, mice were killed for analysis. Infected kidney samples were fixed in 4% paraformaldehyde, embedded in paraffin, and stained with H&E. Infected kidney and/or liver samples were collected and CFU counts were determined. Splenocytes were harvested and stimulated with 10 6 of heat-killed C. albicans per ml of cells for 48 h in vitro for subsequent intracellular staining, ELISA, or RNA isolation. Under some conditions, macrophages are isolated from the indicated mice and after an infection in vitro with L. monocytogenes and C. albicans yeast for 10-12 h, we performed intracellular staining, ELISA, RNA isolation, or functional analyses. For drug treatments, cells were incubated either with vehicle, CoCl 2 (200 µM; Calbiochem), 2-ME (2 µM; Calbiochem), or 2-DG (1 mmol/L; Sigma) for 0.5-1 h before stimulation.

Measurement of bacterial and fungal burdens.
On the indicated days after challenge infections, infected livers and kidneys were harvested and homogenized in PBS. For bacterial infections, serial dilutions of homogenates were plated on LB agar plates and incubated at 37 °C for 24 h. For fungal infections, serial dilutions of homogenates were plated on Sabouraud dextrose agar (SDA) plates and incubated at 37 °C for 48 h. Bacterial and fungal colony forming units (CFU) were subsequently counted.
Glycolysis flux assay. Glycolysis in macrophages was determined by measuring the detritiation of [3-3 H]-glucose as described previously 23 . In brief, the assay was initiated by adding 1 μCi of [3-3 H]-glucose (Perkin Elmer), and 2 h later, medium was transferred to microcentrifuge tubes containing 50 μl 5 N HCL. The microcentrifuge tubes were then placed in 20 ml scintillation vials containing 0.5 ml water, and the vials were capped and sealed. 3  NO production. For NO production, culture supernatant was incubated with the Griess reagent (G4410; Sigma-Aldrich), as described previously 24 . Quantitative RT-PCR and Immunoblot. RNA was extracted with an RNeasy kit (Qiagen, Dusseldorf, Germany), and cDNA was synthesized using SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA). An ABI 7900 Real-time PCR system was used for quantitative PCR, with primer and probe sets obtained from Applied Biosystems. The results were analyzed using SDS 2.1 software (Applied Biosystems). The cycling threshold value of the endogenous control gene (Hprt1, encoding hypoxanthine guanine phosphoribosyl transferase) was subtracted from the cycling threshold. The expression of each target gene is presented as the 'fold change' relative to that of control samples, as described previously 25 . Immunolbot was performed with the following antibodies: HIF1α (R&D system) and β-actin (Sigma), as described previously 16 .
Human macrophage functions. For assays of human macrophage functional activities, CD14 + monocytes from healthy human donors (2W-400C; Lonza) were cultured, and monocytes-derived macrophages were differentiated with rHu M-CSF (25 ng/mL; R&D system) for 7 days. Subsequently, live macrophages (7-AAD − and CD11b + ) were sorted with a MoFlo XDP (Beckman Coulter) and then stimulated with LPS for mRNA analysis or ELISAs.
Statistical analyses. All data are presented as the means ± SD. The non-parametric U test was applied for the comparison of means and was used to compare differences between groups. Comparison of the survival curves was performed using the Log-Rank (Mantel-Cox) test. A P value (alpha-value) of less than 0.05 was considered to be statistically significant.

Results
HIF1α-dependent glycolysis is associated with pro-inflammatory macrophage differentiation. The glycolytic activities of macrophages were first investigated using different macrophage polarization conditions. TLR ligands are classically considered to induce pro-inflammatory macrophage differentiation, and IL-4 cytokines are considered to induce anti-inflammatory macrophage differentiation, whereas dectin-1, a c-type lectin specific for β-glucans, induces fungal infection responses [26][27][28] . Therefore, we compared the in vitro adjuvant activity on macrophages of lipopolysaccharides (LPS), a widely used ligand for TLR4, IL-4, and curdlan, a prototypic agonist for dectin-1, and evaluated the macrophages' glycolytic pathway activity alterations. The glycolytic pathway activity of differentiated macrophages was measured by the generation of 3 H-H 2 O from [3-3 H]-glucose. Sorted PEMs were stimulated and the glycolytic pathway activity was determined. As reported previously [29][30][31] , pro-inflammatory macrophage-inducing factor LPS or LPS + IFNγ significantly increased macrophage glycolytic pathway activity, while anti-inflammatory macrophage-inducing factor IL-4 did not significantly alter their glycolytic pathway activity. However, curdlan significantly increased the glycolytic pathway activity of macrophages, but its effect was weaker than that of LPS or LPS + IFNγ treatment groups (Figs 1A and S1). These data showed that fungal infection likely induces weaker glycolytic pathway activity than that of pro-inflammatory macrophage differentiation. As previously reported 32,33 , HIF1α is critically involved in the modulation of glycolytic pathway activity. We then determined the expression of HIF1α in cells. Consistently, HIF1α expression was higher during pro-inflammatory macrophage inducing conditions and with curdlan stimulation, but not during anti-inflammatory macrophage inducing conditions (Figs 1B and S2). These data showed that HIF1α-dependent glycolytic modulation is probably involved in the polarization of macrophages during different kinds of pathogen infections. We next determined whether HIF1α contributes to the alteration of glycolytic pathway activity in differentiated macrophages during different pathogen infections. We used mice with conditional HIF1α deletions. HIF1α flox/flox mice were crossed with LysM-Cre +/+ mice, and as a result, HIF1α is specifically deleted in macrophages (called HIF1α −/− hereafter, Fig. S3). As expected, HIF1α deficiency significantly decreased glycolytic pathway activity of differentiated macrophages stimulated with either LPS or curdlan, and this was also observed in cells challenged in vitro with L. monocytogenes or C. albicans (Fig. 1C,D). Taken together, these data suggested that HIF1α-dependent glycolytic pathway activity is associated with macrophage polarization induced by different kinds of pathogenic microorganism infections.

HIF1α is critical for pro-inflammatory macrophage differentiation in vitro.
To directly investigate how HIF1α affects macrophage polarization differentiation during different pathogen infections, we observed macrophage functional alterations in vitro. In LPS-stimulated cells, HIF1α deficiency significantly inhibited the mRNA expression, secretion and production of the pro-inflammatory cytokine TNFα, IL-6 production, and the production of the anti-inflammatory cytokine IL-10. Since LPS is a classic inducing factor for bacterial infection ( Fig. 2A-C), this indicates that HIF1α is required for the differentiation of pro-inflammatory macrophages during bacterial infection. Curdlan, a specific ligand for Dectin-1, which has been shown to consistently induce fungal infection responses, caused similar and weaker pro-inflammatory macrophage differentiation (Figs 2A-C and S4). Glut1 and SLC2A1 have been used as markers of glycolytic pathway activity in macrophages 19 , and HIF1α deficiency significantly blocked these glycolytic molecules' expression in macrophages (Fig. 2D). This suggested that glycolytic pathway activity is associated with HIF1α-dependent differentiation of pro-inflammatory macrophages.
HIF1α is required for the differentiation of pro-inflammatory macrophages during bacterial infection. To  the liver had higher bacterial CFU in HIF1α −/− than in the WT controls (Fig. 3A). Microscopic and histological observations revealed a more severe pathological inflammation in the liver and spleen of HIF1α −/− mice (Fig. 3B). Consistently, HIF1α deficiency significantly decreased the level of TNFα in the serum and in peritoneal exudates after L. monocytogenes challenge ( Fig. 3C and E). Additionally, splenic and peritoneal macrophages showed decreased TNFα secretion and production in HIF1α −/− mice compared with WT control groups (Figs S5A,B and 3D,F and G). NO production, a marker of pro-inflammatory macrophages, is also lower in HIF1α-deficient peritoneal macrophages (Fig. S6). Furthermore, the glycolytic pathway activity of splenic and peritoneal macrophages was significantly decreased in HIF1α −/− genotypes (Figs S5C,D, S7 and 3H,I). Together, these data showed that HIF1α is required for pro-inflammatory macrophage differentiation during bacterial infection.
HIF1α is required for the differentiation of pro-inflammatory macrophages during fungal infection. We challenged WT and HIF1α −/− mice with C. albicans in vivo and compared their effects with those of L. monocytogenes infection. HIF1α deficiency significantly increased the survival of the fungus (Fig. 4A,B) and revealed a more severe alteration in histopathological inflammation (Fig. 4C). After C. albicans challenge and testing in the serum of mice and the supernatant of cultures from peritoneal macrophages, we consistently observed that HIF1α deficiency significantly decreased the levels and expression of the pro-inflammatory cytokine TNFα and increased the level and expression of the anti-inflammatory cytokine IL-10 (Figs 4D,E and S8A,B). The HIF1α −/− peritoneal macrophages showed significantly decreased glycolytic pathway activity (Figs 4F,G and S8C). Additionally, NO production and iNOS expression, both markers of pro-inflammatory macrophages, were also expressed at lower levels in HIF1α deficient macrophages (Fig. S9). Together, these data showed that HIF1α is required for macrophage differentiation during fungal infection.
Glycolytic activity is involved in the HIF1α-directed macrophage differentiation. To investigate whether glycolytic activity is involved in the HIF1α-directed macrophage differentiation during different pathogenic microorganism infections, we cultured and stimulated peritoneal macrophages from HIF1α −/− , WT, or 2-DG pretreated WT (WT; 2-DG) mice with either LPS or curdlan. LPS and curdlan significantly increased the mRNA expression of HIF1α, while blocking glycolytic pathway activity with 2-DG, a prototypical inhibitor of the glycolytic pathway, by blocking hexokinase, the first rate-limiting enzyme of glycolysis, did not alter the expression of HIF1α (Fig. 5A). However, blocking glycolytic pathway activity with 2-DG treatment significantly inhibited the differentiation of pro-inflammatory macrophages after stimulation with LPS or curdlan in WT macrophages or with different pathogenic microorganism infections. This is consistent with HIF1α deficiency-induced macrophage phenotypic changes (Fig. 5B-D). Furthermore, blocking glycolytic pathway activity with 2-DG treatment in vivo significantly decreased the secretion and expression of the pro-inflammatory cytokine TNFα, and increased the fungus' survival ( Fig. 5E-H). These data suggested that glycolytic pathway activity is involved in the HIF1α-directed macrophage differentiation during different pathogenic microorganism infections.
Pharmacological targeting of HIF1α-dependent glycolytic pathway activity in mouse and human macrophage differentiation. Next, we applied a pharmacological approach to target the HIF1α-glycolytic pathway in both mouse and human macrophages to determine whether we can recapitulate our findings from the infectious inflammation model. Treatment with CoCl 2 , an activator of HIF1α, upregulated HIF1α expression, TNFα secretion, and fungal death. Treatment with 2-ME, an inhibitor of HIF1α, significantly downregulated HIF1α expression, TNFα secretion, and increased fungal survival in mouse pro-inflammatory macrophages (Fig. 6A-C). Blocking glycolytic pathway activity significantly inhibited the secretion of TNFα caused by HIF1α upregulation (Fig. 6B and D). We extended our experiments to the differentiation of human pro-inflammatory macrophages, and largely recapitulated what we observed in genetic and pharmacological targeting of the differentiation of mouse pro-inflammatory macrophages in terms of the alteration in fungal survival, the secretion of the pro-inflammatory cytokine TNFα, and glycolytic pathway activity (Fig. 6E-H). Thus, our data indicated that HIF1α-dependent glycolytic pathway activity mediated an evolutionarily conserved signaling pathway in both mouse and human macrophages during infectious inflammation.

Discussion
Host defenses against infection rely on innate immune cells that sense microbe-derived products through pattern recognition receptors, such as toll-like receptors (TLRs) and c-type lectins, NOD-like receptors, RIG-I-like receptors and cytosolic DNA sensors 36,37 . Macrophages are important first-line innate immune defense cells that upon encountering microbial ligands from bacterial or fungal microorganisms could differentiate into differently polarized cells and cause a series of innate immune defense responses 38 . Previously, our study 19 showed that HIF1α is critical for the differentiation of pro-inflammatory macrophages; however, its significance remains unclear during pathogenic infections, especially since different pathogenic microorganisms cause different macrophage polarization through specific signaling mechanisms. To systemically clarify the effects of HIF1α in the polarization of pro-inflammatory macrophages during different pathogenic microorganism infections, we selected two kinds of pathogens, L. monocytogenes, a classical gram-positive bacterium, and C. albicans, a classical fungal organism, and observed the alteration of macrophage polarization following in vitro and in vivo challenge. Additionally, we examined the role of HIF1α-dependent glycolytic pathway activity on cellular polarization. The results showed that bacterial and fungal infections exhibited similar effects on macrophage phenotypic differentiation. HIF1α deficiency inhibits pro-inflammatory macrophage functional differentiation when cells are stimulated with LPS or curdlan in vitro or when mice are infected with L. monocytogenes or C. albicans in vivo, leading to a decreased secretion of the pro-inflammatory cytokines TNFα and IL-6, secretions associated with pathogenic microorganism survival. Increased survival of pathogenic microorganisms caused a much more severe infectious inflammation. Alteration of glycolytic pathway activation was required for the differentiation of pro-inflammatory macrophages in protecting against bacterial and fungal infections. However, which specific macrophage cell subtype is mainly affected by HIF1α still needs to be investigated. HIF1α deficiency significantly increases the production of IL-10 protein by macrophages following infection but not IL-10 mRNA expression, indicating that IL-10 secretion is regulated by HIF1α through post-transcriptional regulatory mechanisms. Additionally, in the present study, LysM-Cre was used to delete the HIF1α in myeloid cells, at least including macrophages and neutrophils. Therefore, the effect of infection in mice does not fully represent the effect of macrophage in vitro. It has been confirmed that HIF1α is critical in regulating neutrophil survival and recruitment in inflammation 39,40 . So, the effect of neutrophil and other myeloid cell elimination should also be taken into account in the mouse disease model. Thus, combined with the experimental results in vivo and in vitro, the HIF1α-dependent glycolytic pathway is essential for the macrophage functional differentiation in protecting against bacterial and fungal infections (Fig. S10).  E); the sorted splenic macrophages or PEMs were stimulated with LPS for 12 h, the supernatant was collected, and TNFα levels were determined using an ELISA (D and F); the glycolytic pathway activity of macrophages was also determined (H and I). TNFα expression in F4/80 + macrophages from peritoneal exudates was analyzed with FCM. A representative figure is shown in the left image, and the data are summarized in the right image (G). Data are presented as the means ± SD (n = 3-5). One representative experiment of three independent experiments is shown. ***P < 0.001, compared with the indicated groups. Macrophages, in protecting against pathogenic organism infections, have a unique plasticity that allows them to differentiate into differently polarized phenotypes depending on the cytokine microenvironment 41,42 . Pro-inflammatory macrophages are differentiated by type I pro-inflammatory cytokines and microbial products, such as LPS, and express most TLRs, opsonin receptors, inducible NO synthase (iNOS), and secrete the cytokines TNFα and IL-6, among others. In contrast, anti-inflammatory macrophages are differentiated by type II anti-inflammatory cytokines (IL-4 and/or IL-13) and ameliorate type I inflammation 43 . Fungal pathogens, such as C. neoformans, H. capsulatum, P. brasiliensis, Coccidioides immitis, and C. albicans are sensitive to the antimicrobial effects of macrophages. However, many fungi use various mechanisms to interact with macrophages and cause different macrophage inflammatory responses [44][45][46] . In the case of H. capsulatum, macrophages can efficiently phagocytize the fungus but are unable to kill it until activated by cytokines produced by effector T cells 47 . C. albicans and C. neoformans are also thought to mediate NO suppression through the downregulation of macrophage iNOS RNA and usually inhibit pro-inflammatory macrophage phenotypes 44 . Thus, macrophage polarization effects need to be explored for their role in protecting different pathogenic microorganisms during infection. Here, we showed that C. albicans can cause significant pro-inflammatory macrophage polarization and that HIF1α-glycolytic pathway mechanisms are required for the mediation of pro-inflammatory macrophage polarization in protecting against C. albicans fungal infection in vivo and in vitro.
Macrophage biology in humans is less understood compared to that in animal model organisms because of the technical and ethical challenges in obtaining fresh material from human subjects. While several protocols that could differentiate blood monocytes into macrophages (monocyte-derived macrophages) by applying GM-CSF or M-CSF have been available since the 1990s 48,49 , the difficulties of accessing human peripheral tissue leukocytes, including resident macrophages, remains a technical barrier for studying human macrophage biology. To date, most studies have relied on the use of human monocyte-derived macrophages. These in vitro models have allowed us to gain enormous insights into the functions of these cells, but the physiological relevance of some key aspects is still unclear [50][51][52] . Recent comparative biology studies have revealed that certain human and mouse macrophage subsets share some functional and phenotypical characterizations and therefore may represent homologous subsets [51][52][53][54] . For instance, human CD14 + classical monocytes are considered as a homologous subset to murine Ly6C hi monocytes, since M-CSF stimulated human monocyte-derived macrophages displayed characteristics of alternatively activated macrophage (termed M2-like) phenotypes 51,[55][56][57] . In recent years, this has replaced the simplistic view of myeloid precursors giving rise to blood monocytes that, in turn, originate tissue macrophages 58,59 . Similar to human peritoneal macrophages, mouse peritoneal macrophages (F4/80 hi , CD11b hi , and MHCII lo ) are believed to derive from an embryonic precursor [60][61][62] . In the current studies, we have shown that fungal infection drives up glycolysis and promotes pro-inflammatory cytokine production in both human monocyte-derived macrophages and mouse peritoneal macrophages. Importantly, these effects are mediated through the transcription factor HIF1α. Furthermore, our findings showed that pharmacologically targeting HIF1α could direct mouse and human macrophage differentiation in a similar manner. Our studies implicate metabolic programs as promising therapeutic targets for macrophage-driven human diseases.
Although it is known that glycolysis is associated with innate immune cell-mediated inflammation, as well as playing a role in immune-associated diseases, the mechanisms regulating increased glycolysis following pattern recognition receptor (PRR) stimulation particularly in macrophage polarization are unclear, especially in in vivo disease models. HIF1α plays a crucial role in the regulating myeloid cell function in hypoxia and in inflammation more broadly. Conditional knockout mice lacking HIF1α in neutrophils, macrophages and dendritic cells were demonstrated to have profoundly impacted immune functions. Using myeloid-targeted HIF1α knockout mice, Cramer et al. have confirmed a critical role for HIF1α in regulating neutrophil and mononuclear cell phagocytosis 63 . In HIF1α-deficient myeloid cells, this resulted in a reduction of ATP pools, accompanied by profound impairment of bacterial killing. In vivo this correlated with the ablation of ablation of sodium dodecyl sulphate-induced cutaneous inflammation and a reduction in synovial infiltration in an immune complex-mediated inflammatory arthritis model. There are follow up studies that demonstrated that HIF1α promotes myeloid cells (macrophages and/or neutrophils) pro-inflammatory activation and antimicrobial defense in a streptococcal infection model 64 . It showed that HIF1α deficient in myeloid cells decreased the bactericidal activity and exaggerated systemic spread of infection. In the present study, we showed that L. monocytogenes, a classical gram-positive bacterium infection directed a pro-inflammatory macrophage differentiation in vitro and in vivo. Moreover, it has been demonstrated that HIF1α in myeloid cells plays a role in trained immunity and defense against C. albicans fungal infection 32 . Trained monocytes display high glucose consumption, lactate production, and NAD + /NADH ratio, reflecting a shift in the metabolism of trained monocytes with an increase in glycolysis dependent on the activation of mTOR through HIF1α pathway. Inhibition of mTOR, or HIF1α blocked monocyte induction of trained immunity. Mice with a myeloid cell-specific defect in HIF1α were unable to mount trained immunity against bacterial sepsis. Otherwise, there are data that demonstrated that HIF1α plays an important role in antifungal responses 65,66 . Our previous study 19 showed that pro-inflammatory stimulation suppresses Myc-dependent cell proliferation while engaging a HIF1α-dependent transcriptional program to maintain heightened glycolysis in pro-inflammatory macrophage functional differentiation. However, whether HIF1α-dependent glycolysis is critical for macrophage polarization during in vivo infection with different pathogenic microorganisms remains uninvestigated. In the present study, we further showed that HIF1α-dependent glycolytic pathway is essential for the macrophage functional differentiation in protecting against bacterial and fungal infections in vitro and in vivo. (C and G) The CFU was evaluated. (D and H) The Glut1 mRNA expression was determined with qPCR. Data are presented as the means ± SD (n = 3-4). One representative experiment of three independent experiments is shown. *P < 0.05 and ***P < 0.001, compared with the indicated groups.