Contrasting functional responses of resident Kupffer cells and recruited liver macrophages to irradiation and liver X receptor stimulation

In the murine liver, there are two major macrophage populations, namely resident Kupffer cells (resKCs) with phagocytic activity and recruited macrophages (recMφs) with cytokine-producing capacity. This study was performed to clarify the functional differences between these two populations, focusing on their susceptibility to radiation and response to stimulation via liver X receptors (LXRs), which are implicated in cholesterol metabolism and immune regulation. Liver mononuclear cells (MNCs) were obtained from C57BL/6 (WT) mice with or without 2 Gy irradiation, and the phagocytic activity against Escherichia coli (E. coli) as well as TNF-α production were compared between the two macrophage populations. To assess LXR functions, phagocytosis, TNF-α production, and endocytosis of acetylated low-density lipoprotein (LDL) were compared after synthetic LXR ligand stimulation. Furthermore, LXRα/β knockout (KO) mice and LXRα KO mice were compared with WT mice. Irradiation decreased intracellular TNF-α production by recMφs but did not affect the phagocytic activity of resKCs. In vitro LXR stimulation enhanced E. coli phagocytosis by resKCs but decreased E. coli-stimulated TNF-α production by recMφs. Phagocytic activity and acetylated LDL endocytosis were decreased in both LXRα/β KO mice and LXRα KO mice, with serum TNF-α levels after E. coli injection in the former being higher than those in WT mice. In conclusion, resKCs and recMφs exhibited different functional features in response to radiation and LXR stimulation, highlighting their distinct roles liver immunity and lipid metabolism.


Introduction
Various innate immune cell types can be found in the liver, including natural killer (NK) cells, natural killer T (NKT) cells, and macrophages. Among these, the liver resident macrophages, Kupffer  Macrophages are thought to be involved in various aspects of lipid metabolism. One of their major functions in this context is the phagocytosis of excessive lipoproteins in serum. However, aberrant phagocytosis results in the internalization of excessive lipoproteins and the subsequent formation of foam cells, which exacerbate atherosclerosis and NASH [20,21]. In these reports, CD36 is a primary scavenger receptor for oxidized low-density lipoprotein (LDL) and involves in this foam cell formation in NASH. Thus, endocytosis of denatured LDL and CD36 expression are the important functional criteria which evaluate the implication of macrophage in metabolic diseases.
In this study, we sought to elucidate how irradiation and LXR affect the functions of liver recMφs and resKCs, including bacteria phagolytic activity, cytokine production, and LDL endocytosis.
The Ethics Committee of Animal Care and Experimentation, National Defense Medical College, Japan, approved all requests for animals and the intended procedures for the present study (Permission number: 18064). All experiments were performed in accordance with the relevant guidelines and regulations. Every effort was made to minimize animal suffering, and there were no unexpected deaths.

Isolation of liver leukocytes
Mice were euthanized under deep anesthesia through inhalation of 5% isoflurane, and blood samples were obtained from the inferior vena cava into a heparinized syringe. Mice were perfused with 10 ml saline to remove blood from organs, and the liver was obtained. Liver nonparenchymal leukocytes were isolated via Percoll density gradient centrifugation, as previously described [8]. Briefly, the liver was minced with scissors and digested with 0.05% collagenase (Wako, Osaka, Japan). The digested liver was dispersed in medium by pushing it with syringe rubber on a stainless steel mesh (30 μm wire thickness and 72 μm pore size, AMICHU corporation, Osaka, Japan). Non-parenchymal MNCs were isolated from hepatocytes via Percoll density gradient centrifugation (33%), and red blood cells were removed using RBC lysis buffer.

Flow cytometry
Liver leukocytes were incubated with Fc-blocker (BD Biosciences, Franklin Lakes, NJ, USA) to prevent non-specific binding and labeled with APC-CD45 (Thermo Fisher, Waltham, MA USA) and PerCP-eFluor710-Ly-6G (Thermo Fisher). To exclude non-immune cells and neutrophils, CD45 + Ly6G − mononuclear cells (MNCs) were gated. FITC-F4/80 and PE-CD11b (Thermo Fisher) were used to define each macrophage population. For the analysis of PE-Ly6c, PE-CD36 Ab (Thermo Fisher), and PE-MerTK (R&D Systems, Minneapolis, MN, USA), PerCP-Cy5.5-CD11b was selected to identify each cell population. The cells with the highest expression of CD11b in addition to high FS and SS were defined as neutrophils and were excluded from the analysis. In the case of CLEC4F (C-type lectin domain family 4 F) staining, cells with or without permeabilization of cell membranes were labeled with a polyclonal goat anti-mouse CLEC4F antibody (R&D Systems) and an FITC-donkey polyclonal anti-goat IgG antibody (R&D Systems). PE-F4/80 (Thermo Fisher) was used for macrophage detection. Similarly, after permeabilization of the cell membrane (BD Biosciences), intracellular TNF-α was labeled with FITC-anti-TNF (Thermo Fisher). The cells labeled with fluorescent antibodies were analyzed using a Novocyte flow cytometer (Agilent, Santa Clara, CA, USA).

Bacteria phagocytic activity and acetylated LDL endocytosis
For the analysis of phagocytic activity, a total of 2.5×10 9 FITC-E. coli (particles/kg, Thermo Fisher) were injected into mice, and liver MNCs were obtained 15 min after injection. The cells were labeled with PE-F4/80, PerCP-Cy5.5-CD11b, and APC-CD45 for the identification of resKCs and recMφs. The phagocytic activity of each cell population was evaluated based on the FITC signal intensity of non-injected control mice.
For the analysis of phagolytic activities, pHrodo 1 E. coli (Thermo Fisher) was used in vitro. Its fluorescence emission increases as the pH of the surrounding environment becomes acidic, enabling the detection of functional phagolysosomes in phagocytic cells. Liver MNCs obtained from collagenase-treated livers (pretreated with 1 μM T0901317 or DEMSO for 18 h) were cultured with pHrodo 1 E. coli (50 μL per 5 × 10 5 cells) at 37˚C for 1 h, harvested by pipetting on ice, washed with medium, and then labeled with FITC-F4/80, PerCP-Cy5.5-CD11b, and APC-CD45. The phagolytic activity of each cell population was determined based relative to the PE signal intensity after being kept on ice for 1 h instead of at 37˚C.
To assess the in vivo endocytosis of modified-LDL, 100 μL of DIO (green fluorescence)acetylated LDL (Thermo Fisher) or control acetylated LDL was injected intravenously into mice, and, 20 min after injection, the endocytosis by resKCs and recMφs was evaluated via flow cytometry.

LXR ligand pretreatment of liver MNCs and their cytokine production after LPS stimulation
Hepatic MNCs were isolated from WT mice with collagenase treatment and incubated with T0901317 (1 μM) or control (DEMSO) for 18 h, followed by stimulation with LPS (1 ng/mL). After 45 min of LPS stimulation, mRNA was extracted from adherent cells using an RNeasy mini kit (Qiagen, Venlo, the Netherlands) and subjected to reverse transcription with Super-Script III (Invitrogen), followed by quantitative PCR analysis with LightCycler 480 SYBR Green I (Roche, Basel Switzerland). For the analysis of inflammatory cytokine production, liver MNCs isolated without collagenase treatment (lacking resKCs) were cultured with T0901317 or control (DEMSO) for 18 h and were subsequently stimulated with LPS. The supernatant of each well was collected after 6 h for TNF-α and 12 h for IL-12. The concentration of each cytokine was measured using ELISA (TNF-α ELISA kit; BD Biosciences, Total IL-12 ELISA Kit, Thermo Fisher).

Radiation exposure
Radiation exposure was performed two days before sacrifice and immune cell isolation. The condition was set to 2 Gy, which is a relatively low dose, in order to prevent adverse effects on other organs as much as possible. The MBR-1503R (HITACHI, Tokyo, Japan) irradiation device was used at a dose rate of 0.4 Gy per minute. The physical condition of mice, including food intake, body weight, and moving activity, did not change after irradiation.

Statistical analyses
Data are presented as the mean ± SE. We performed one-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons, unpaired Student's t-tests, using the GraphPad software.
We previously reported that CD68 is a marker of resKCs, with recMφs having intracellular CD68 and activated recMφs expressing surface CD68 [3,7]. In contrast to CD68, activation of recMφs via LPS injection into mice did not induce CLEC4F according to the previous report [23]. These results indicate that F4/80 and CD11b in combination with Ly6c and CLEC4F may be a robust markers for the classification of resKCs and recMφs.

Differences in radiosensitivity between resKCs and recMφs
To further clarify the functional differences between resKCs and recMφs [7], low-dose irradiation (2 Gy) was performed in WT mice. The distribution and functions of resKCs and recMφs were examined 2 days after irradiation. Irradiation decreased the total number of liver MNCs compared to control mice without irradiation (6.2±0.4×10 6 vs. 2.3±0.3×10 6 cells, n = 13, 6)  (Fig 2A and 2B) among liver MNCs. ResKCs, but not recMφs, showed potent in vivo phagocytosis of FITC-E. coli (15 min after i.v. injection), and irradiation did not significantly decrease their phagocytic activity (Fig 2C and 2D, n = 5). In contrast, when liver MNCs from mice with or without irradiation were stimulated with LPS in vitro, a certain proportion of recMφs (32.9±7.9%), but not resKCs, produced intracellular TNF-α, while irradiation significantly decreased the proportion of intracellular TNF-α-positive recMφs (14.8±1.4%) (Fig 2E and 2F, n = 5, p<0.05). These results suggest that irradiation impaired the inflammatory function of recMφs but did not substantially affect the phagocytic activity of resKCs.

Synthetic LXR ligand (T0901317) pretreatment in vitro enhanced the phagocytic activity of resKCs but decreased LPS-induced TNF-α production by recMφs
Liver MNCs were isolated after collagenase treatment and pretreated in vitro with T0901317 or control (DEMSO) overnight (18 h). Thereafter, liver MNCs were cultured with FITC-E. coli for 1 h. MNCs were harvested from the plates and analyzed via flow cytometry. The results showed that liver resKCs pretreated with T0901317 had more potent phagocytic activity against E. coli than the control resKCs (Fig 3A and 3B, green lines and columns, n = 7). RecMφs exhibited weaker phagocytic activity against FITC-E. coli than resKCs. After T0901317 pretreatment, the difference in their phagocytic activity was not statistically significant (Fig 3A and 3B, blue lines and columns, n = 7). In contrast, the mRNA expression of TNF in T0901317-pretreated liver MNCs following LPS stimulation in vitro was significantly lower compared to that of control liver MNCs (Fig 3C left panel, n = 5 in each). Consistently, hepatic MNCs pretreated with T0901317 produced less TNF-α after LPS stimulation than control hepatic MNCs (Fig 3C, right panel, n = 14 in each). Since hepatic MNCs harvested without collagenase treatment included virtually no resKCs, TNF-α was mainly produced by LPS-stimulated recMφs. However, T0901317 pretreatment did not affect IL-12 production by recMφs (Fig 3D, n = 9 in each group). These results suggest that LXR signaling enhanced the phagocytic function of resKCs but inhibited the inflammatory activity of recMφs.

LXRα/β KO mice showed a decreased population of resKCs with impaired phagocytic activity in parallel to an increased recMφs population with enhanced TNF-α production
LXRα/β KO mice had higher proportions of recMφs and lower proportions of resKCs than WT mice (p<0.05) (Fig 4A and 4B, n = 4). ResKCs of LXRα/β KO mice exhibited significantly lower phagocytic activity against FITC-E. coli at 15 min after injection compared to those of WT mice, while the phagocytic activity of LXRα KO mouse resKCs did not differ significantly compared to that of WT mouse resKCs (Fig 4C and 4D, n = 3 each). LXRα/β KO mice had much higher plasma TNF-α levels at 1 h after live E. coli injection than WT mice (Fig 4E) (n = 3 in each group), suggesting that recMφs in the liver and other organs were activated and had a high TNF-α production capacity. Plasma IL-12 levels in LXRα/β KO mice at 12 h after 80 low CD11b high cells were recMφs (blue dots)., and F4/80 high CD11b low cells were defined as resKCs (green dots).

PLOS ONE
Different responses of two subsets of liver macrophages to irradiation and liver X receptor stimulation viable E. coli injection were also significantly higher than those in WT mice, although the difference was not as pronounced as observed for TNF-α levels (Fig 4F).

Decreased phagolytic activity of resKCs against E. coli in LXRα KO mice
To further explore the role of LXRα in macrophage phagocytic function against E. coli, we assessed functional phagolysosome formation in resKCs using pHrodo E. coli. Interestingly, the in vitro positivity for pHrodo E. coli in resKCs from LXRα KO mice was significantly lower compared to that in WT mice (Fig 5A and 5B, n = 5 in each group), suggesting that LXRα is required for the phagolytic activity of phagolysosomes in resKCs. In addition, only a small proportion of recMφs were positive for pHrodo E. coli, suggesting that even though a certain proportion of recMφs phagocytosed E. coli (Fig 3A), they may not kill E. coli as effectively as resKCs.

Decreased modified-LDL endocytosis by resKCs in LXRα KO and LXRα/β KO mice
At 20 min after the intravenous injection, endocytosis activity of resKCs for acetylated LDL was blunted in LXRα KO mice and severely impaired in LXRα/β KO mice, when compared to WT mice (Fig 6A lower panels and Fig 6B green columns). These results suggested that LXR regulates not only the phagocytosis of bacteria, but also that of lipids. In contrast, recMφs in each group of mice endocytosed a small amount of acetylated LDL (Fig 6A upper panels and Fig 6B blue columns, n = 5). These results suggested that LXRα deletion may impair LDL endocytosis (as well as bacteria phagolysis capacity) in resKCs and subsequently reduce the hydrolysis of cholesterol esters and triglycerides in the phagolysosomes of resKCs [25], suggesting that lipogenesis was impaired in LXR KO mice.

Increased CD36 + resKCs in LXRα KO and LXRα/β KO mice
The proportion of CD36 + resKCs in WT mice was higher than that of recMφs (Fig 7A, n = 6). Unexpectedly, the number of CD36 + resKCs in LXRα KO and LXRα/β KO mice (n = 4, each group) was significantly higher than that in WT mice (Fig 7A) (p<0.01). Since LXR is a nuclear receptor implicated in the expression of CD36 (oxidized LDL receptor) [26], it was somewhat unexpected that the number of CD36 + resKCs in LXRα and LXRα/β KO mice was higher than that in WT mice. However, peroxisome proliferator-activated receptor γ (PPARγ) is considered the main inducer of CD36 [27]. Therefore, CD36 expression may be upregulated to compensate for the impaired LDL uptake by resKCs in LXR KO mice.

Discussion
The present study provides insight into the function of resKCs and recMφs, with LXRs affecting these two subsets in contrasting manners. LXR stimulation potentiated the phagocytic mice (n = 13 and 6). (C) Phagocytic activity of recMφs (upper panels) and resKCs (lower panels) against intravenously injected FITC-E. coli in mice with or without irradiation. The percentage of FITC-E. coli-positive cells was calculated relative to non-injected control mice (gray area in each panel). (D) Bar graph of E. coli phagocytosis by recMφs (blue columns) and resKCs (green columns) in control and irradiated mice (n = 5 in each group). (E) Intracellular TNF-α production in recMφs (upper panels) and resKCs (lower panels) stimulated with LPS in vitro. Liver MNCs from control or irradiated mice were stimulated with LPS in vitro for 1.5 h, and intracellular TNF-α was labeled (n = 5 in each group). (F) Bar graph of intracellular TNF-α expression in recMφs (blue columns) and resKCs (green columns) in control and irradiated mice (n = 5 in each group). Data are presented as the mean ± SE ( � p <0.05, unpaired Student's t-test).
https://doi.org/10.1371/journal.pone.0254886.g002 function of resKCs but suppressed TNF-α production by recMφs. LXRs may also be involved in lipid metabolism through regulating LDL endocytosis by resKCs. Taken together, LXRs modulated the function of both subsets in a sophisticated manner to maintain host defense against infections and metabolic dysregulation. The in vitro TNF-α production by MNCs with or without T0901317 pre-treatment. Adherent cells were obtained after incubation with LPS for 45 minutes, mRNA was isolated, and the relative expression TNF mRNA was compared between the two groups (left panel, n = 5 in each group). After incubation with LPS for 6 h, culture supernatant was collected, and TNF-α concentration was measured and compared (left panel, n = 14 in each group). (D) IL-12 production by liver MNCs stimulated with LPS for 12 h with or without T0901317 pre-treatment (n = 9 in each group). Data are presented as the mean ± SE ( � p <0.05, unpaired Student's t-test). https://doi.org/10.1371/journal.pone.0254886.g003

PLOS ONE
Different responses of two subsets of liver macrophages to irradiation and liver X receptor stimulation Since the identification of resKCs and recMφs [3], we have realized that the immunological phenomena in the liver cannot be elucidated without distinguishing the roles of these two subsets. ResKCs are essential for anti-bacterial immunity through their phagocytic activity, whereas recMφs are essential for anti-tumor immunity through IL-12 production [2,3,7]. Consistently, the depletion of resKCs by clodronate liposome treatment rendered mice extremely susceptible to bacterial infections, whereas liver anti-tumor immunity was not compromised at all (rather, it was enhanced) as the number of recMφs was increased, presumably due to resKCs producing MCP-1 prior to apoptosis [2]. Non-lethal irradiation was previously reported to transiently decrease recMφs, but not resKCs [2]. In the current study, 2 Gy irradiation diminished TNF-α production in LPS-stimulated recMφs but did not affect resKC phagocytic activity against E. coli. Consistently, the depletion of recMφs by repeated low-dose irradiation (1 Gy/week for 8 weeks) greatly decreased plasma TNF-α levels and remarkably improved NASH in fibroblast growth factor-5-null mice fed a Western diet [7]. In previous reports, radiation could induce TNF-α secretion, which was in turn implicated in radiationinduced fatal organ inflammation [28]. In these experiments, the amount of radiation was mice (n = 3), and LXRα/β KO mice (n = 3). Mice were injected with FITC-E. coli, and the phagocytic activity of resKCs (green line) was analyzed 15 min after injection. ResKCs from a mouse without E. coli injection were used as negative controls (gray line). (D) Bar graph of the phagocytotic activity of resKCs (green columns) in WT, LXRα KO, and LXRα/β KO mice (n = 6, 3, 3 in each). Data are presented as the mean ± SE ( � p <0.05, one-way ANOVA followed by Tukey's multiple comparisons) (E, F) WT mice and LXRα/β KO mice were injected with live E. coli, and plasma TNF-α and IL-12 levels were measured at the indicated time points after E. coli injection (n = 3 in each group). Data are presented as the mean ± SE. ( � p <0.05, unpaired Student's t-test).
LXR ligand pre-treatment was reported to enhance host immunity against bacterial infections [30]. In addition, LXRα/β KO mice, as well as LXRα KO mice, but not LXRβ KO mice, were reported as susceptible to bacterial infections [30]. In line with these observations, LXRα KO but not LXRβ KO mice fed a Western diet were susceptible to hepatic cholesterol overload and NASH [22,31]. These findings suggest that LXRα, rather than LXRβ, is preferentially involved in the phagocytosis of bacteria and lipid endocytosis. Furthermore, LXR ligands reduced TNF-α production by Kupffer cells (presumably recMφs) in mice injected with LPS, attenuating LPS-induced liver injury [32]. In line with these observations, we recently reported

PLOS ONE
that the number of liver recMφs was increased in LXRα/β KO mice, and LPS-induced liver injury was exacerbated in mice due to the increased numbers of recMφs and associated TNF-α production [12]. Of note, the immunological alterations in LXR-deficient mice may be attributed to the indirect effects of LXR deficiency, such as compensatory reactions. Further studies in which synthetic ligands are administered to LXR-knockout mice are therefore needed to address this issue.
We recently reported that LXRα KO mice fed a Western diet had significantly increased numbers of liver TNF-α-producing recMφs and developed NASH, showing higher liver and plasma cholesterol levels but not liver triglyceride levels compared to WT mice fed a Western diet, in addition to lower plasma triglyceride levels [12]. Taken together with the current results, it is possible that LXRα signals are involved in cholesterol metabolism and lipogenesis in resKCs and suppress inflammation, including NASH-associated inflammation, by inhibiting TNF-α-producing recMφs. Therefore, LXRα deletion may impair LDL endocytosis in resKCs and subsequently reduce the hydrolysis of cholesterol esters and triglycerides in the phagolysosomes of resKCs [25,33], suggesting that lipogenesis is impaired in LXRα KO mice. However, since cholesterol efflux and elimination are impaired in the resKCs and hepatocytes of LXRα KO mice fed a Western diet due to the ABC transporter deficiency [12], cholesterol may be stored in resKCs as well as hepatocytes. Thus, LXRα dysfunction could underpin the decrease in lipogenesis and increase of cholesterol accumulation, especially within the liver (hepatocytes and resKCs).
It should be noted that the number and function of resKCs and recMφs may be influenced by various background factors, such as age, dietary status, chronic inflammation, and sex. These should be considered during research. In humans as well as mouse experimental models, metabolic syndrome and tissue damage usually occur in elderly individuals (�20 weeks old in mice and �40 years old in humans) and progress with aging [34][35][36]. Therefore, we used aged mice (30-50 weeks of age) in the current study.
The LXR-enhanced phagolytic activity of resKCs and decreased TNF-α production by recMφs are both beneficial for host protection against bacterial infections and TNF-α-mediated metabolic disease (diabetes, atherosclerosis, NASH, etc.). Notably, the administration of LXR ligands also activates lipogenesis by enhancing hepatic fatty acid synthesis through the upregulation of sterol regulatory element-binding protein 1c (SREBP-1c) and carbohydrate response element-binding protein (ChREBP), resulting in hypertriglyceridemia and hepatic steatosis [37][38][39]. These adverse metabolic effects impede the application of LXR ligands in clinical practice. A previous report revealed that LXR ligand treatment could ameliorate inflammatory reactions while preventing hypertriglyceridemia [40]. An approach for upregulating endogenous LXR ligand desmosterol was recently published [41]. It could suppress inflammatory reactions while ameliorating fat synthesis via SREBP inhibition. Based on these reports, LXR stimulation is a promising strategy for the treatment of various inflammatory diseases. Therefore, the elucidation of its effect on immune cells and metabolism, with a focus on different macrophage populations, is essential for future immuno-metabolism research.