Tumor-associated neutrophils attenuate the immunosensitivity of hepatocellular carcinoma

Abstract

Tumor-associated neutrophils (TANs) are heterogeneous; thus, their roles in tumor development could vary depending on the cancer type. Here, we showed that TANs affect metabolic dysfunction-associated steatohepatitis hepatocellular carcinoma (MASH-related HCC) more than viral-associated HCC. We attributed this difference to the predominance of SiglecFhi TANs in MASH-related HCC tumors. Linoleic acid and GM-CSF, which are commonly elevated in the MASH-related HCC microenvironment, fostered the development of this c-Myc-driven TAN subset. Through TGFβ secretion, SiglecFhi TANs promoted HCC stemness, proliferation, and migration. Importantly, SiglecFhi TANs supported immune evasion by directly suppressing the antigen presentation machinery of tumor cells. SiglecFhi TAN removal increased the immunogenicity of a MASH-related HCC model and sensitized it to immunotherapy. Likewise, a high SiglecFhi TAN signature was associated with poor prognosis and immunotherapy resistance in HCC patients. Overall, our study highlights the importance of understanding TAN heterogeneity in cancer to improve therapeutic development.

Graphical abstract

Figure 1.

Neutrophils are pathogenic in MASH-related HCC. (A) Kaplan–Meier plots comparing HCC patients from various datasets with high (top 25%) and low (top 25%) TAN signature scores. (B) The etiologies of HCC patients in A. (C) Kaplan–Meier plots comparing different groups of TCGA-LIHC patients with high and low TAN signature scores. (D and E) Mice bearing NRAS/AKT or cMYC/sgp53 HCC tumors were sacrificed at 4 or 7 wk after tumor induction, respectively. Representative images of AFP staining (left) and neutral lipid staining (right) on liver sections. The scale bar represents 100 µm (D). Representative FACs plots and percentages of CD11b⁺Ly6G⁺ neutrophils among CD45⁺TILs (n = 4–5/group) (E). (F) NRAS/AKT HCC mice (top panel) and cMYC/sgp53 HCC mice (bottom panel) were treated with αLy6G or isotype control mAbs for 2 wk. Representative images of HCC livers (left); liver-to-body weight ratio (middle) and tumor nodule counts (right). Scale bar represents 1 cm. (G) Tumor volume of RIL175 orthotopic HCC mice fed with MCD diet (MCD + RIL175; top) or Chow diet (CD + RIL175; bottom), following αLy6G or isotype control mAb treatments. Each symbol represents one mouse. Data are mean ± SEM (E–G) and represent two to three independent experiments (D–G). *P < 0.05, ***P < 0.001 by Kaplan–Meier analysis (A and C) and unpaired Student’s t test (F and G). NA, not available; T, tumor; NT, non-tumor.

Figure S1.

Characterization of murine models reveals MASH-HCC features in NRAS/AKT HCC and MCD + RIL175 models. (A–C) Representative H&E liver sections. The HCC tumor region is bounded by the dotted line. Scale bar represents 100 µm. (D) Untargeted lipidomic LC-MS analysis of the various types of HCC tumors and naïve liver. (E) Representative FACs plots and percentages of Ly6G⁺CD11b⁺ neutrophils (TANs) among CD45⁺TILs from cMYC/sgp53 tumors after 2 wk of αLy6G or isotype control mAbs treatment (top). Quantity of TANs per gram of tissue (bottom). (F and G) Gating strategies for tumor infiltrating lymphocytes (F): TANs (i), DCs (ii), CD8⁺T cells (iii) and their cytokines (iv), and tumor cells (G) from all the HCC models used. Each symbol represents one mouse. Data are mean ± SEM and represent two independent experiments (E). ****P < 0.0001 by unpaired Student’s t test (E). T, tumor; NT, non-tumor; TG, triglyceride; DG, diglyceride; PC, phosphatidylcholine; PE, phosphatidylethanolamine; SM, sphingomyelins.

Figure S2.

Neutrophil identification and characteristics. (A) Uniform Manifold Approximation and Projections (UMAPs) (left) and expression of neutrophil lineage markers (right) by the major immune populations identified from the scRNA-seq of the BM and spleen (top), and HCC tumors (bottom) of NRAS/AKT HCC mice. (B) Dot plot of the differently expressed genes by each neutrophil cluster in the BM and spleen (top), and HCC tumor (bottom). (C) RNA-seq analyses were performed on SiglecF^hi and SiglecF^– TANs purified from NRAS/AKT-HCC tumors. A gene set defining SiglecF^hi TANs was generated from their top 100 differentially expressed genes. The distribution of this gene set among TAN subsets. (D) Representative FACs plots and percentage of SiglecF^hi neutrophils detected in the BM, spleen, blood, and tumor of NRAS/AKT HCC mice. (E) CXCR2, CXCR4, CD62L, and dcTRAIL-R1 expression within SiglecF⁻ and SiglecF^hi TANs of NRAS/AKT HCC mice. (F) Functional characterization of BM neutrophils, SiglecF⁻ and SiglecF^hi TANs from NRAS/AKT HCC mice. (G) Representative FACs plots and percentages of SiglecF^hi TANs of RIL175 orthotopic HCC mice fed with MCD diet (MCD + RIL175; left) or Chow diet (CD + RIL175; right) (n = 3/group). (H) Representative immunofluorescence images and proportion of SiglecF^hi TANs (myeloperoxidase [MPO]: red; SiglecF: green; nucleus: blue) in NRAS/AKT HCC tumors after 2 wk of treatment with αLy6G or isotype control mAbs. Scale bar represents 100 µm. (I) Expression levels of genes in the Siglecf-Hi–like TAN signature across various cell types in the tumor of HCC patients (https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA007744). Each symbol represents one mouse. Data are mean ± SEM and represent two independent experiments (D–H). **P < 0.01, ***P < 0.001, ****P < 0.0001 by one-way ANOVA (F) and unpaired Student’s t test (H).

Figure 2.

Protumorigenic SiglecF ^hi TANs are enriched in MASH-related HCC tumors and can be removed through αLy6G mAb injections. (A–C) scRNA-seq analysis of neutrophils from NRAS/AKT HCC mice. UMAP visualization of neutrophil subsets (A). Pearson’s correlations across neutrophil clusters in NRAS/AKT HCC mice and HCC patients (https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA007744) (B). Distributions of SiglecF^high neutrophil signature score derived from Engblom et al. (2017) in TAN subsets (C). (D) Representative FACs plots of SiglecF^hi TANs detected in NRAS/AKT HCC (left) and cMYC/sgp53 HCC (right) tumors. (E) Representative immunofluorescence images of SiglecF^hi TANs (MPO: red; SiglecF: green; nucleus: blue) in NRAS/AKT HCC tumors. Dotted oval shows the tumor region. Scale bars represent 200 and 5 μm. (F) Representative cytospin images of SiglecF⁻ and SiglecF^hi TANs. Scale bar represents 10 μm. (G) Representative FACs plots and quantities of TANs (CD11b⁺ Ly6G^{intracellular+} (top) and TAN subsets (bottom) in NRAS/AKT HCC tumors after 2 wk of αLy6G or isotype control mAbs treatments. (H) Distributions of conventional neutrophil signature and Siglecf-Hi–like TAN signature scores in HCC patients with non-MASH (GSE63898) and MASH (GSE164760) etiologies. (I) Kaplan–Meier plots comparing different groups of TCGA-LIHC patients with high and low Siglecf-Hi–like TAN signature score. Each symbol represents one mouse. Data are mean ± SEM (D and G) and represent three independent experiments (D–G). ***P < 0.001, ****P < 0.0001 by unpaired Student’s t test (D and G-upper panel) and one sample t test (G-lower panel). Data are analyzed by Wilcoxon rank-sum test (H) and Kaplan–Meier analysis (I). hPB, human peripheral blood; hAL, human adjacent liver.

Figure 3.

SiglecF ^hi TANs directly promote tumorigenesis. (A) Pearson’s correlation of tumorigenesis-related gene signatures with Siglecf-Hi–like TAN signature in the TCGA-LIHC dataset. (B–E) Sort-purified SiglecF^– and SiglecF^hi TANs were co-cultured with RIL175 cells. Schematic diagram of the sorting and co-culture strategy (B). Limiting dilution assay: representative images of spheroids formed and stem cell frequency. Scale bar represents 50 μm (C). Clonogenic assay: representative images of colonies formed and colony counts. Scale bar represents 3.5 mm (D). Scratch wound assay: representative images of the wound at indicated time points and percentage of the wound area. Scale bar represents 200 μm (E). (F–H) RIL175 cells alongside SiglecF⁻ or SiglecF^hi TANs were subcutaneously injected into C57BL/6J mice. Schematic diagram of the adoptive transfer strategy (F). Representative images and volumes of tumors at the endpoint. Scale bar represents 1 cm (G). Representative FACs plots and percentage of Ki-67⁺ tumor cells (H). Each symbol represents TANs isolated from two to three mice (D, E, G, and H). Data are mean ± SEM (D, E, G, and H) and represent two to three independent experiments (C, E, G, and H) or are pooled from two independent experiments (D). *P < 0.05, **P < 0.01, ***P < 0.001 by Pearson’s correlation test (A), Pearson’s chi-square test with a 95% confidence interval (C), one-way ANOVA (D and E), and unpaired Student’s t test (G and H).

Figure S3.

Siglecf ^hi TAN removal impairs tumorigenesis. (A) Visualization of Vegfa expression among TILs. (B–E) Profiling of tumor cells from NRAS/AKT HCC mice after 2 wk of αLy6G or isotype control mAbs treatments. scRNA-seq expression of mature hepatocyte and hepatic progenitor markers (B). Representative FACs plots and percentage of Hoechst-effluxing cells (polygon gate) (C). Representative FACs plots and percentage of CD13⁺CD133⁺ CSCs (polygon gate) (D). Stem cell frequency in tumor cells isolated from NRAS/AKT HCC mice 1 wk following αLy6G or isotype control mAbs treatments (E). Each symbol represents one mouse (C and D). Data are mean ± SEM and represent two independent experiments (C–E). *P < 0.05, **P < 0.01 by unpaired Student’s t test (C and D) and Pearson’s chi-square test with a 95% confidence interval (E). MQs, macrophages; DCs, dendritic cells.

Figure 4.

SiglecF ^hi TANs secrete TGFβ. (A) GO analysis of differentially expressed genes between RIL175 cells and those co-cultured with SiglecF^hi TANs. BP, biological process. (B and C) NicheNet analysis of scRNA-seq data from NRAS/AKT HCC mice treated with αLy6G or isotype control mAbs. Prioritized ligands and their target genes driving transcriptomic changes in tumor cells after αLy6G treatment (B). Log fold-change (LFC) in senders of ligands to tumor cells after αLy6G treatment (C). (D) Relative expression of indicated genes among neutrophil subsets from NRAS/AKT HCC mice. (E) Representative FACs plots (left) and percentage of LAP (TGFβ1)⁺ cells among TAN subsets. (F) Spontaneous TGFβ production by SiglecF^– and SiglecF^hi TANs. Each symbol represents one mouse (E) or TANs isolated from two to three mice (F). Data are mean ± SEM and represent two independent experiments (E and F). *P < 0.05, ****P < 0.0001 by unpaired Student’s t test (E and F).

Figure 5.

SiglecF ^hi TANs mediate their protumorigenic functions via TGFβ secretion. (A–C) SiglecF^hi TANs were co-cultured with RIL175 cells, with or without the addition of Vactosertib (10 nM). In vitro limiting dilution (A), clonogenic (B), and scratch wound (C) assays were performed. (D–G) NRAS/AKT HCC tumors were induced in Mrp8^Cre+ve.TGFβ^fl/fl mice (Mrp8^Cre.TGFβ^fl/fl) and Mrp8^Cre−ve.TGFβ^fl/fl littermates (W/T). Representative FACs plots and percentage of Ki-67⁺ tumor cells (D). Representative FACs plots and percentage of CD13⁺CD133⁺ CSCs (E). Representative images of HCC livers and liver-to-body weight ratio. Scale bar represents 1 cm (F). Representative images of AFP⁺ tumors and tumor nodule counts. Scale bar represents 100 μm (G). (H–J) SiglecF^– and SiglecF^hi TANs were sort-purified from NRAS/AKT HCC tumors and subcutaneously injected alongside RIL175_shTgfbr1 cells into C57BL/6J mice. Schematic diagram of adoptive transfer strategy (H). Representative images and volumes of tumors at the endpoint. Scale bar represents 1 cm (I). Representative FACs plots and percentage of Ki-67⁺ tumor cells (J). Each symbol represents TANs isolated from two to three mice (B and C), or one mouse (D–G, I, and J). Data are mean ± SEM and represent two to three independent experiments (A, C–G, I, and J) or are pooled from two independent experiments (B). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by Pearson’s chi-square test with a 95% confidence interval (A), one-way ANOVA (B and C), or unpaired Student’s t test (D–G).

Figure 6.

TGFβ-producing TANs are enriched in MASH-related HCC patients. (A) NicheNet analysis of HCC patient dataset (https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA007744). Prioritized ligands from neutrophils, and their target genes, driving transcriptomic differences in tumor cells between patients with high or low neutrophil infiltration. (B) Distributions of TAN-derived TGFβ production signature in HCC patients with non-MASH (GSE63898) and MASH (GSE164760) etiologies. (C) Representative immunofluorescence images and quantification of TGFβ-producing CD66b⁺neutrophils (CD66b: red; TGFβ: green; nucleus: blue) among all neutrophils in tumors resected from patients with HBV-related HCC or MASH-related HCC. Scale bars represent 50 and 20 μm. Each symbol represents one patient (C). Data is mean ± SEM. **P < 0.01 by Wilcoxon rank-sum test (B) and unpaired Student’s t test (C).

Figure 7.

GM-CSF and linoleic acid activate c-Myc regulon and drive SiglecF ^hi TAN development. (A and B) Relative differential transcription factor activity within each neutrophil subset in NRAS/AKT HCC mice (A) and HCC patients (B). (C) Representative histograms and c-Myc levels within SiglecF^– and SiglecF^hi TANs. (D) c-Myc expression assessed by western blot. Proteins were extracted from TANs subsets isolated from five mice. (E–G) SiglecF⁻ and SiglecF^hi TANs were treated with or without 10058-F4 (100 μM). GSEA analysis of genes defining SiglecF^hi TANs (E). Spontaneous TGFβ production (F). RIL175 cells were s.c. injected with different TAN subsets. Tumor volume at day 8. Scale bar represents 1 cm (G). (H) Representative histograms and neutral lipid levels within SiglecF^– and SiglecF^hi TANs. (I) GC-MS analysis of HCC tumors and naïve liver. (J) Relative scores of M-CSF, G-CSF, and GM-CSF signaling within neutrophil subsets from NRAS/AKT HCC mice. (K and L) BM neutrophils from NRAS/AKT HCC mice were cultured overnight with or without the addition of GM-CSF (20 ng/ml) and the indicated fatty acids (50 μM). c-Myc expression (K) and TGFβ1 production (L). (M–O) BM neutrophils from NRAS/AKT HCC mice were cultured overnight with or without the addition of GM-CSF, linoleic acid, or 10058-F4. Heatmap of genes defining SiglecF^hi TANs (M). Schematic diagram of adoptive transfer strategy (N). Tumor volume (O). Each symbol represents one mouse (C, H, K, L, and O) or TANs isolated from two to three mice (F and G). Data are mean ± SEM (F, G, K, L, and O) and are pooled from two (C, F–H, K, and L) or four (O) independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by Wilcoxon matched-pairs signed rank test (C and H) and one-way ANOVA (F, G, K, L, and O). MFI, median fluorescence intensity; CM, complete media. Source data are available for this figure: SourceData F7.

Figure S4.

Identification of factors contributing to the development of SiglecF ^hi TANs. (A) Quantities of SiglecF⁻ and SiglecF^hi TANs after overnight culture with or without 100 μM 10058-F4. (B) Representative immunofluorescence image and proximity of TAN subsets to lipid droplets (MPO: red; SiglecF: green; BODIPY 493/503: pink; nucleus: blue) in NRAS/AKT HCC tumors. Scale bar represents 10 µm. (C–E) BM neutrophils were isolated from NRAS/AKT HCC-bearing mice and cultured overnight with or without the addition of the indicated fatty acids (50 μM). Representative FACs plots of SiglecF⁺ neutrophils (C). Representative histograms of c-Myc expression (D). TGFβ1 production measured by ELISA following 4 h of LPS (100 ng/ml) stimulation (E). (F) Expression of Csf2 by cellular populations within NRAS/AKT HCC tumors. (G) NRAS/AKT HCC was induced in C57BL/6J or GM-CSF reporter Csf2-iCre-EGFP^+/wt mice. Representative FACs plots of GM-CSF producing cells. (H) Representative FACs plots of SiglecF⁺ neutrophils after 20 ng/ml GM-CSF was included in C. (I and J) Neutrophils from spleen and tumor of NRAS/AKT mice were cultured overnight with or without the addition of GM-CSF (20 ng/ml) or/and linoleic acid (50 μM). c-Myc expression within splenic neutrophils (I) or SiglecF⁻ and SiglecF^hi TANs (J). (K) Pseudotime of TAN development in NRAS/AKT HCC mice, calculated with Monocle 3. (L) SiglecF⁻ or SiglecF^hi TANs isolated from CD45.1⁺ mice were adoptively transferred into NRAS/AKT HCC-bearing CD45.2⁺ recipients. After 2 days, TILs were harvested, donor TANs (CD45.1⁺, polygon gate) and endogenous TANs were assessed for their SiglecF expression (rectangle gate indicates SiglecF^hi cells). (M and N) BM neutrophils were isolated from naïve mice and cultured with and without GM-CSF (20 ng/ml) and linoleic acid (50 μM) overnight. Representative histograms of c-Myc expression (M). TGFβ1 production measured by ELISA following 4 h of LPS (100 ng/ml) stimulation (N). Each symbol represents TANs isolated from two to three mice (A and J), or one mouse (E, I, and N), or one neutrophil from screening liver sections of eight mice (B). Data are mean ± SEM and are representative of two independent experiments (C–E, G, H, and L–N) or are pooled from two independent experiments (A, B, I, and J). *P < 0.05 by unpaired Student’s t test (B). CM, complete media.

Figure 8.

SiglecF ^hi TANs downregulate the antigenicity of tumor cells via TGFβ, and their removal sensitizes HCC mice to anti-PD-1 mAb treatment. (A) Distributions of Siglecf-Hi–like TAN signature score in HCC patients from the IMbrave150 and GO30140 clinical trials. (B) MHCI expression by IFNγ-treated RIL175 cells co-cultured with SiglecF⁻ or SiglecF^hi TANs, with or without Vactosertib (Vac.). (C and D) Antigen processing and presentation signature score was calculated from an imputed tumor-specific gene expression profile (tumor antigen presentation signature) in patients treated with Atezolizumab and Bevacizumab from the IMbrave150 and GO30140 clinical trials. Correlation between tumor antigen presentation signature and Siglecf-Hi–like TAN signature scores (C). Distributions of tumor antigen presentation signature score in responder (R) and non-responders (NR) (D). (E and F) OVA-expressing NRAS/AKT HCC was induced in Mrp8^Cre+ve.TGFβ^fl/fl mice (Mrp8^Cre.TGFβ^fl/fl) and Mrp8^Cre−ve.TGFβ^fl/fl littermates (W/T). Representative histograms and OVA/MHCI presentation by tumor cells (E). Tumor cells were co-cultured with OT-1 cells. Representative histograms and proliferation index of OT-1 cells (F). (G and H) NRAS/AKT HCC Mrp8^Cre+ve.TGFβ^fl/fl mice (Mrp8^Cre.TGFβ^fl/fl) and Mrp8^Cre−ve.TGFβ^fl/fl littermates (W/T) were treated with αPD-1 or isotype control mAbs for 2 wk. Representative images of HCC livers and liver-to-body weight ratio. The scale bar represents 1 cm (G). Representative images of AFP⁺ tumors (brown) and tumor nodule counts. Scale bar represents 100 μm (H). (I) NRAS/AKT HCC mice were treated with isotype control mAbs or αLy6G mAbs or αPD-1 mAbs or both at the indicated time. Survival curve (n = 10/group). (J–Q) Mice were sacrificed 2 wk following treatment. Representative images of HCC livers and liver-to-body weight ratio. Scale bar represents 1 cm (J). Representative images of AFP⁺ tumors (brown) and tumor nodule counts. Scale bar represents 100 μm (K). MHCI expression on tumor cells (L). Number of CD8⁺TILs (M). Proportion of PD-1⁺cells among CD8⁺TILs (N). Proportion of terminally exhausted cells (PD-1^hiCTLA4⁺Tox⁺Tcf1⁻) in CD8⁺TILs (O). Frequency of Ki-67⁺ CD8⁺PD-1⁺TILs (P). Percentage of CD8⁺PD-1⁺TILs co-producing IFNγ and TNFα after in vitro anti-CD3/CD28 restimulation (Q). (R–T) CD8⁺T cell depletion abrogated the beneficial effects of αLy6G mAb treatment in anti-PD-1 mAb administrated mice. Survival curve (n = 7–10/group) (R). Mice were sacrificed 2 wk following treatments. Liver-to-body weight ratio (S). Tumor nodule counts (T). Each symbol represents TANs isolated from two to three mice (B), one mouse (E, G, H, J–Q, S, and T), or the mean of tumor cells isolated from three mice (F). Data are mean ± SEM and are pooled from three independent experiments (B), or represent two to three independent experiments (E–T). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by Wilcoxon rank-sum test (A and D), one-way ANOVA (B, G, H, J–Q, S, and T), unpaired Student’s t test (E), two-way ANOVA (F), and Kaplan–Meier analysis (I and R).

Figure S5.

αLy6G mAb treatments downregulate the antigenicity of tumor cells and activates CD8 ⁺ TILs in HCC-bearing C57BL/6J mice, while HCC-bearing Mrp8 ^Cre .TGFβ ^fl/fl mice respond to αPD-1 mAb treatment. (A) SiglecF⁻ and SiglecF^hi TANs were adoptively transferred alongside OVA-expressing RIL175 (left) or RIL175_shTgfbr1 cells (right). OVA/MHCI presentation by tumor cells. (B) Representative histograms and MHCI expression on NRAS/AKT tumor cells following αLy6G or isotype control mAbs treatments. (C) UMAP of tumor infiltrating CD8⁺T cells (CD8⁺TILs) in NRAS/AKT-injected C57BL/6J mice treated with αLy6G or Isotype mAbs. (D) Distribution of marker genes in different CD8⁺TIL subsets. (E) Expression of T cell receptor signaling related genes within the Teff_CD8 subset. (F–J) OVA-expressing NRAS/AKT HCC tumors were used to assess tumor antigen presentation and T cell recognition following αLy6G or isotype control mAbs treatments. Schematic outline of the experiment (F). Representative histograms and OVA/MHCI presentation by tumor cells (G). OVA/MHCI presentation by tumor-infiltrating CD11c⁺MHCII⁺ DCs (H). Proliferation of OT-1 cells 3 days after adoptive transfer into OVA-NRAS/AKT HCC mice (I). Tumor cells from OVA-NRAS/AKT HCC mice were co-cultured with OT-1 cells at the indicated ratios for 3 days. Representative histograms and proliferation index of OT-1 cells (J). (K–P) NRAS/AKT HCC was induced in Mrp8^Cre+ve.TGFβ^fl/fl mice (Mrp8^Cre.TGFβ^fl/fl) and Mrp8^Cre−ve.TGFβ^fl/fl littermates (W/T), followed by treatment with αPD-1 or isotype control mAbs. MHCI expression on tumor cells (K). Number of TCRβ⁺CD8⁺TILs per gram of HCC tissue (L). Proportion of PD-1⁺ cells among TCRβ⁺CD8⁺TILs (M). Frequency of proliferating TCRβ⁺CD8⁺PD-1⁺TILs determined by Ki-67 staining (N). Proportion of terminally exhausted cells (PD-1^hiCTLA4⁺Tox⁺Tcf1⁻) among TCRβ⁺CD8⁺TILs (O). Percentage of TCRβ⁺CD8⁺PD-1⁺TILs co-producing IFNγ and TNFα after in vitro anti-CD3/CD28 restimulation (P). Each symbol represents one mouse (A, B, G–I, and K–P), or the mean of tumor cells isolated from three mice (J). Data are mean ± SEM and represent two to three independent experiments (A, B, and G–P). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by unpaired Student’s t test (A, B, and G–I), two-way ANOVA (J), and one-way ANOVA (K–P).