This study and protocols were approved by the University of Southern California Institutional Review Board and abide by the Declaration of Helsinki principles.

PBMCs were collected from five healthy donors. Blood was collected in Vacutainer CPT Citrate tubes (BD Biosciences) and processed according to manufacturer’s instructions. Cells were stored in liquid nitrogen until use.

Normal human Kupffer cells (KCs) were commercially obtained from LifeNet Health. Selection criteria included BMI ranging from 19 to 27, fibrosis grade =0, and NAS score < 1.

Normal human neutrophils were obtained from one healthy donor, and cells were isolated using MACSxpress^® Whole Blood Neutrophil Isolation Kit, human (Miltenyi) and used fresh for chemotaxis and NETosis assays.

CD14⁺ monocytes were isolated from five healthy control PBMCs using the CD14 MicroBeads human kit according to the manufacturer’s protocol (Miltenyi). CD14⁺ monocytes were plated at 10⁶ cells/well in 6-well plate in RPMI 1640 with 50ng/mL of M-CSF plus 5% FBS for 5 days at 37°C/5% CO₂. After differentiation, M2-like macrophages were stimulated for 24h with RPMI + 5%FBS media plus PBS (unstimulated control), or 50μg/mL of native low-density lipoprotein (Kalen Biomedical LLC), medium oxidized LDL (Kalen Biomedical LLC), or high oxidized LDL (Kalen Biomedical LLC). Cells were then harvested in Buffer RLT (Qiagen) + 1% 2-Mercaptoethanol (Sigma-Aldrich) and stored at -80°C until processing.

M2-like macrophages RNA was isolated using RNeasy Mini kit (Qiagen) and RNA integrity was analyzed by 2100 Expert Bioanalyzer System (Agilent), using the Agilent RNA 6000 Pico Kit (Agilent). Libraries were simultaneously prepared from extracted total RNA using Illumina Truseq Stranded mRNA library preparation kit according to the manufacturer’s protocol (Illumina). Prepared libraries were sequenced on the Illumina Nextseq500 at 3 million reads per sample at 2x75 cycles. Libraries were prepared and sequenced by the USC Molecular Genomics Core.

The raw sequencing reads were first checked for overall quality and adapter contamination using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) prior to downstream analysis. Reads were then used to qualify transcript abundances with Salmon (19) using the GENCODE version 31 reference including only protein coding genes. The resulting transcript abundances were summarized to gene level counts using functions in the Bioconductor package tximport (20). Significantly differentially expressed genes were identified using the Bioconductor package DESeq2 (21) with a significance threshold of FDR < 0.1 and fold change of at least 1.5 when compared to paired control (PBS) samples.

We used the Ingenuity Pathway Analysis (IPA) (Qiagen_https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis) to determine pathways, and the IPA Biomarker Tool to select the biomarker candidates (22). GeneMANIA (https://genemania.org/) was used to evaluate the strength of the networks, with following criteria: (i) automatically selection of weighting method, (ii) no additional genes or attributes, and considering (iii) co-expression, (iv) co-localization, (v) genetic and (vi) genetic and physical interactions, (vii) common pathways, (viii) shared protein domain and (ix)predicted interaction (23). The platform Gene Ontology (http://geneontology.org) was used to identify biological processes (24), using significance threshold of FDR < 0.1. The software Cluster 3.0 (http://bonsai.hgc.jp/~mdehoon/software/cluster/) was used to hierarchically cluster genes and samples (25), and Java TreeView v11.6r4 (http://jtreeview.sourceforge.net) was used to visualize the clusters and generate cluster figures (26). The RNAseq files generated for this manuscript are all deposited in the Gene Expression Omnibus (GEO) with the access number GSE160200.

Kupffer cells from 5 healthy controls were plated at 10⁵ cells/well in 24 well plates, in RPMI 1640 (phenol-red free) plus 5% FBS, and additional stimuli was added as follows: PBS, 50µg/mL of native LDL or high oxLDL. Cells were incubated for 24h. Supernatant (conditioned media) was saved at -20°C and cells were harvested in buffer RLT + 1% 2-mercaptoethanol. Conditioned media was used for the chemotaxis assay and RNA was used for gene expression analysis.

For the NETosis assay, Kupffer cells from 3 healthy donors were plated at 10⁵ cells/well in 24 well plates, in RPMI 1640 (phenol-red free) plus 0.5% human serum albumin, and additional stimuli was added: PBS, 50µg/mL of native LDL or high oxLDL. Cells were incubated for 18h. Conditioned media were stored at -20°C until use.

Kupffer cell RNA was isolated using RNeasy Mini kit (Qiagen) and cDNA conversion was done using QuantiTect Reverse Transcription Kit (Qiagen). The QuantiTect Primer Assays (Qiagen) selected for this study targeted CXCL8 (Cat# QT00000322), IL10 (Cat# QT00041685), CD163 (Cat# QT00074641), and MRC1 (Cat# QT00012810). For housekeeping gene, we used RRN18s (Cat#QT00199367). Reactions were performed using RT² SYBR Green ROX qPCR Mastermix (Qiagen) for 40 cycles on StepOne™ Real-Time PCR System (Thermo Fisher Scientific). Gene expression from target genes were obtained by delta-delta-Ct method.

For the chemotaxis assay, we used the IncuCyte S3 System. For this assay, we used the supernatant of KCs stimulated for 24h with native LDL or high oxLDL. Incucyte^® Clearview 96-Well Chemotaxis Plates (Essen Biosciences) were coated with Matrigel^® Matrix (Corning). In the bottom chamber, we added 200µL of the following attractants: (a) RPMI 1640 (phenol-red free) + 5% FBS, (b) 1 µM fMLP, (c) 25µg/mL native LDL, (d) 25µg/mL high oxLDL, (e) conditioned media KCs + PBS, (f) conditioned media KCs + native LDL, and (g) conditioned media KCs + high oxLDL. On top chamber, we added 5x10³ neutrophils resuspended in RPMI 1640 (phenol-red free) plus 0.5% human serum albumin. Plates were placed into IncuCyte S3 and images were acquired every 1 hour, for 24 hours. Each condition had 4 replicates.

For the NETosis assay, we used the IncuCyte S3 System. Normal human neutrophils were resuspended to 10⁵ cells/mL in RPMI 1640 (phenol red-free, serum-free) medium with 250nM Incucyte^® Cytotox Green Dye (Essen Biosciences). We added 50µL of cells suspension (2x10⁴) per wells into Incucyte^® Imagelock 96-well plates (Essen Biosciences). Then, we added 50µL of: (a) RPMI 1640 (phenol-red free) + 0.5% HSA, (b) 1 µM PMA, (c) 25µg/mL high oxLDL, (d) KC-PBS conditioned media, (e) KC-nLDL conditioned media, or (f) KC-Hox conditioned media. Each condition had 3 replicates. Plate was placed in the IncuCyte S3, and images were acquired every 10 minutes for 12 hours.

Kupffer cells from 4 healthy donors were plated at 10⁴ cells/well in 96-weel plates in 10%FBS RPMI for 2h. Media was replaced by 0.5% human serum albumin RPMI plus: PBS, 50µg/mL nLDL or 50µg/mL HoxLDL (final concentrations), and 10µg of Incucyte^® pHrodo^® Red E. coli Bioparticles^® (Essen Biosciences) per well. Each condition had 4 replicates. For negative controls, some wells had only cells or bioparticles. Cells were analyzed in the IncuCyte S3, and images were acquired every 10 minutes for 5.5h.

Chemotaxis quantification was performed by Incucyte^® Chemotaxis Analysis Software Module. Phagocytosis and NETosis events were quantified by IncuCyte^® basic analyzer software.

All statistical analyses and graphs were generated using GraphPad Prism version 8 for macOS (GraphPad Software www.graphpad.com). A Two-way ANOVA test was applied for Chemotaxis, Phagocytosis and NETosis. For the qPCR, we used Ordinary one-way ANOVA test.

Human Magnetic Luminex Assay (6-Plex) (R&D systems) was used to measure chemokines in the supernatant of M2-like macrophage cultures after 24h stimulation with PBS or HoxLDL. The supernatant was diluted at 1:2, and the experiment followed the manufacturer’s instructions. The chemokines analyzed were: CCL3/MIP-1 alpha (BR35), CCL4/MIP-1 beta (BR37), CXCL1/GRO alpha/KC/CINC-1 (BR77), CXCL2/GRO beta/MIP-2/CINC-3 (BR27), CXCL5/ENA-78 (BR52), and IL-8/CXCL8 (BR18).

We used the low-pass RNAseq approach to measure the changes of the highly expressed genes in human monocyte-derived macrophages exposed to native, medium oxidized- and highly oxidized LDL. We applied the DESeq2 package (21) to identify differentially expressed genes (DEG), with significance threshold of FDR < 0.1 and fold change of at least 1.5 when compared to paired control (PBS) samples.

Our results showed that the LDL oxidation degree had different effects on the global gene expression of those cells. Cells incubated with highly oxidized LDL displayed the most dramatic changes (Figures 1A, B). Compared to unstimulated (PBS) cells (Figure 1C), native LDL (nLDL) induced 63 genes, with only 43 genes upregulated and 20 genes downregulated (top 40 genes on Table S1). Medium oxLDL (MoxLDL) stimulated 214 genes, with upregulation of 125 and downregulation of 89 genes (top 40 genes on Table S2). In the presence of highly oxidized LDL (HoxLDL), a total of 2032 genes were differentially expressed (top 40 genes on Table S3); surprisingly, more genes were downregulated (1155) than upregulated (877). All stimuli triggered a set of unique genes: 35 genes by nLDL, 48 genes by MoxLDL, and 1858 genes by hoxLDL. The 8 genes commonly modulated by nLDL, MoxLDL, and HoxLDL are related to cell survival, autophagy, response to insulin, cell trafficking and protein folding (Figure 1D).

To identify a gene expression pattern that could distinguish cells stimulated by nLDL, MoxLDL, and HoxLDL, we evaluated the DEG lists (stimuli vs PBS) using the Ingenuity Pathway Analysis (IPA) biomarker tool. Then, the top 40 upregulated genes of each IPA biomarker list (i.e., nLDL vs PBS; MoxLDL vs PBS, and HoxLDL vs PBS) were used to construct gene networks using GeneMANIA (23). Genes that did not show connection by co-expression, co-localization, and genetic/physical interactions were filtered out. The remaining genes were then used to perform hierarchical clustering using Cluster v3 (25).

This network evaluation strategy identified a network of 25 genes in the MoxLDL (Figure S1A) data set and a network of 26 genes using the nLDL biomarkers list (Figure S1B). However, when clustered, these sets of genes did not group the samples according to the stimuli used (Figures S1C, D). In contrast, the network of 25 genes resulting from the HoxLDL dataset analysis showed not only robust network correlation (Figure 2A) but were also able to cluster the HoxLDL samples together (Figure 2B), confirming that the HoxLDL treatment induced a much stronger transcriptional change in the macrophages when compared to other stimuli.

The biological functions triggered by the LDL treatment were identified using the Gene Ontology platform. For these analyses, we used the lists of upregulated genes from each comparison (HoxLDL, MoxLDL, or nLDL vs PBS), and we excluded functions related to gene transcription and translation, and functions displaying less than 5 genes. For the cells treated with nLDL, no significant change was detected, while MoxLDL-treated cells showed an upregulation of lipid metabolism and oxidative stress. As expected, the main biological processes upregulated in HoxLDL-stimulated cells were also related to cholesterol metabolism and cellular stress, but surprisingly, these cells also exhibited significant inhibition of the foam cell formation and apoptotic processes (Figure 2C and Figure S2).

Moving forward with the DEG list obtained from HoxLDL-stimulated M2-like macrophages, we decided to investigate whether the detected transcriptional changes would affect the cells’ phenotype. Our analyses showed that the M2-like macrophages exposed to HoxLDL did not fully relate to the classical M1 and M2 phenotypes.

We observed that HoxLDL induces a transcriptional shift in monocyte-derived macrophages consistent with the M4 macrophages (Figure 3A), with a mixture of M1 and M2 characteristics, as previously described in atherosclerotic plaques (27, 28). Based on the M4 macrophage characterization published by Gouwy et al. (6), which compared monocytes differentiated by CXCL4 and CXCL4L1 to resting M2-like macrophages, we evaluated the expression of M4 markers in M2-like macrophages treated with LDLs. For this purpose, we only considered the genes included in the DEG lists. Overall, HoxLDL-stimulated cells upregulated the expression of CXCL8, CCL2, CCR5, and IL1RN, and relative downregulation of CD14, CD163, HMOX1, MRC1, and IL10 when compared to resting M2-like macrophages. We also noticed the downregulation of several HLA genes, as expected for M4 macrophages, and surprisingly the decrease of TLRs’ gene expression as well (Figure S3A). In keeping with prior studies, these M4 macrophages derived from HoxLDL stimulation display expression patterns similar to CXCL4L1-exposed monocytes (Figure S3B, C). M2-like macrophages treated with MoxLDL or nLDL did not follow the M4-like gene expression pattern.

Furthermore, we evaluated some of these markers on human Kupffer cells (KCs) exposed to HoxLDL. We selected the scavenger receptor CD163 and mannose receptor MRC1 genes, which are commonly expressed by M1 and M2 KCs (29) but drastically downregulated in M4 cells, and the genes IL10 and CXCL8. HoxLDL-stimulated KCs significantly downregulated the transcription levels of IL10, CD163, and MRC1 (Figure 3B), while kept the CXCL8 expression unchanged but still high, similar to monocytes differentiated by CXCL4 (27).

Phagocytosis is one of the crucial functions of macrophages and is significantly attenuated in M4 macrophages differentiated by CXCL4 (7). To analyze if HoxLDL-differentiated KCs’ phagocytic capacity was also impaired, we incubated KCs obtained from healthy livers with PBS, nLDL, or HoxLDL for 5.5 hours in the presence of heat-killed E. coli conjugated to a pH-sensitive fluorescent dye [pHrodo red E-coli bioparticles^®] (Figures 4A, B). At 1h 10min, human KCs treated with HoxLDL significantly decreased their phagocytic capacity by 60% when compared to unstimulated KCs. When compared to nLDL treatment, HoxLDL significantly decreased phagocytosis by 39% at 2h. The inhibition of phagocytosis by HoxLDL was maintained for the entire time course. In contrast, phagocytic capacities of unstimulated KCs and nLDL-stimulated KCs were comparable.

The secretion of cytokines and chemokines is another hallmark function of macrophages, which contributes to the recruitment and activation of other inflammatory cells. Considering that several chemokines produced by M4-like macrophages could recruit neutrophils (Figures S4A, B), we generated conditioned media from normal KCs incubated with PBS, nLDL, or HoxLDL for a live-cell image neutrophil chemotaxis assay. As shown in Figure 5A, conditioned media from HoxLDL-treated KCs (KC-Hox) recruited significantly more neutrophils than media from PBS- or nLDL-stimulated KCs (KC-PBS and KC-nLDL, respectively). The differences were significant at 8h, reaching a peak of migration at 12h. No significant difference was observed between KC-PBS and KC-nLDL recruitment.

In addition to neutrophil recruitment, we also analyzed the influence of the conditioned media on neutrophil extracellular trap formation (NETosis), a process that collectively includes the disintegration of nuclear and granule membranes and externalization of molecules including [e.g., MPO, double-stranded (ds)DNA, histones] (14–18) leading to inflammation-mediated injury. For that purpose, fresh isolated human neutrophils were incubated with serum-free conditioned media obtained from KCs incubated with PBS, nLDL or HoxLDL. NET formation was detected with Incucyte^® Cytotox Green Dye (Essen Biosciences), which yield fluorescence upon binding to the deoxyribonucleic acid (DNA) extruded by the neutrophils. Apoptotic neutrophils were excluded by IncuCyte^® basic analyzer software. KC-Hox triggered NET formation at 20 minutes, achieving a maximum effect at 1h 10min, and lasting for almost 2 hours (Figures 5B, C). KC-PBS and KC-nLDL had comparable results and showed much lower NET formation induction.

Even though it is known that cholesterol crystals may affect the behavior of neutrophils (17), HoxLDL alone did not have any effect on neutrophil recruitment or NETosis initiation (Figures S5A, B).