Functional Analysis and Transcriptomic Profiling of iPSC-derived Macrophages and Their Application in Modeling Mendelian Disease

Abstract

Rationale

An efficient and reproducible source of genotype-specific human macrophages is essential for study of human macrophage biology and related diseases.

Objective

To perform integrated functional and transcriptome analyses of human induced pluripotent stem cell-derived macrophages (IPSDM) and their isogenic PBMC-derived macrophages (HMDM) counterparts and assess the application of IPSDM in modeling macrophage polarization and Mendelian disease.

Methods and Results

We developed an efficient protocol for differentiation of IPSDM, which expressed macrophage-specific markers and took up modified lipoproteins in a similar manner to HMDM. Like HMDM, IPSDM revealed reduction in phagocytosis, increase in cholesterol efflux capacity and characteristic secretion of inflammatory cytokines in response to M1 (LPS+IFN-γ) activation. RNA-Seq revealed that non-polarized (M0) as well as M1 or M2 (IL-4) polarized IPSDM shared transcriptomic profiles with their isogenic HMDM counterparts while also revealing novel markers of macrophage polarization. Relative to IPSDM and HMDM of control individuals, patterns of defective cholesterol efflux to apoA-I and HDL₃ were qualitatively and quantitatively similar in IPSDM and HMDM of patients with Tangier disease (TD), an autosomal recessive disorder due to mutations in ATP-binding cassette transporter A1. TD-IPSDM also revealed novel defects of enhanced pro-inflammatory response to LPS stimulus.

Conclusions

Our protocol-derived IPSDM are comparable to HMDM at phenotypic, functional and transcriptomic levels. TD-IPSDM recapitulated hallmark features observed in HMDM and reveal novel inflammatory phenotypes. IPSDM provide a powerful tool for study of macrophage-specific function in human genetic disorders as well as molecular studies of human macrophage activation and polarization.

INTRODUCTION

Macrophages are the most plastic cells of the hematopoietic system and represent a critical cell type at the intersection of metabolism, immunity and cardiovascular diseases.¹ An imbalance in lipid metabolism and a maladaptive immune response driven by the accumulation of cholesterol-laden macrophages in the artery wall is the hallmark of atherosclerosis.² Experimental human macrophages have been mainly derived from two sources: tumor-derived cell lines, e.g., U937, THP-1 cells, and primary cells, e.g., peripheral blood mononuclear cells (PBMC)-derived macrophage (HMDM). The former are endowed with unlimited replicative potential, but are karyotypically abnormal and phenotypically immature and do not provide the opportunity for genotype-specific studies. HMDM are reflective of tissue macrophages and provide genotype-specific tools, but do not self-renew, and are refractory to genetic manipulation.

Induced pluripotent stem cells (iPSCs) provide an unlimited source of subject genotype-specific cells and are a powerful platform for disease modeling, drug screening and cell therapeutics.³ The differentiation of human iPSCs to macrophages (iPSC-derived macrophages, or IPSDM) offers a powerful alternative system for using a bank of previously-created genotype-specific iPSCs to derive terminally differentiated, karyotypically normal, and genetically consistent human macrophages. This approach provides a potential tool for study of macrophage-specific functions of genomic loci for human disease and novel mechanisms of human macrophage biology in homeostasis, disease states and in therapeutic translation. However, there is some uncertainty as to the functional and molecular fidelity of iPSC-differentiated cells relative to their primary counterparts.^{4, 5} It is unclear, for example, whether IPSDM respond to pathophysiological stress and genetic perturbations in a similar manner to their respective primary cells. Understanding the morphological, functional and transcriptional characteristics of IPSDM is essential.

Here we describe a method for high-throughput generation of macrophages from iPSC. We show that IPSDM, like HMDM, can be polarized in vitro to functionally and molecularly distinct M1 and M2 subtypes, demonstrate that IPSDM are comparable to their primary isogenic counterparts by phenotypic, functional, secretome, and transcriptional profiling. Finally, we use IPSDM from subjects with Tangier disease (TD) to demonstrate the fidelity of IPSDM, relative to primary HMDM, for gene and disease-specific macrophage functional studies in humans.

METHODS

Supplemental Information includes supplemental methods, 4 figures, and 39 tables can be found in the online supplemental materials.

PBMC to macrophage differentiation (HMDM) and polarization

PBMC collected using BD VACUTAINER® Mononuclear Cell Preparation Tube were cultured in macrophage culture media, 20% FBS in RPMI 1640 media with 100 ng/ml human macrophage colony-stimulating factor (M-CSF), for 7 days on BD Primaria^™ tissue culture plate to induce macrophage differentiation.⁶ Polarization was obtained in the presence of M-CSF by 18–20 hour incubation with 20 ng/ml IFN-γ and 100 ng/ml LPS for M1-like polarization, or 20 ng/ml IL-4 for M2-like polarization as we described.⁶

Subject-specific iPSCs derivation, culture and maintenance

All human protocols for this work were approved by the University of Pennsylvania Human Subjects Research Institutional Review Board. Generation and characterization of subject-specific PBMC-derived iPSCs using Sendai viral vectors were performed by the iPSC Core Facility at Penn’s Institute of Regenerative Medicine as described.⁷

Differentiation of iPSCs to macrophages (IPSDM)

Embryoid bodies (EBs) were generated by culturing small aggregates of feeder-depleted iPSCs in Corning ultra-low attachment multiwell plate in StemPro-34 media supplemented with different cytokine cocktails as summarized in Online Table I. Since D8, macrophage culture media was used to enrich for myeloid precursors. At D15, single cells were transferred to BD Primaria^™ tissue culture plate for expansion and maturation.

Cholesterol efflux

HMDM and IPSDM were labeled by 6 μCi [³H]-Cholesterol for 24 hours, followed by 14–16 hour equilibration with or without treatment of the liver X receptor (LXR) agonists, 10 μM 9-cis-Retinoic acid (9cisRA) and 5 μg/ml 22-hydroxycholesterol (22OH), that upregulate ATP-binding cassette transporters A1 and G1 (ABCA1 and ABCG1). In some experiments, polarization was obtained during equilibration to determine the effects on cholesterol efflux capacity. Efflux to apolipoprotein A-I (apoA-I) (10 μg/ml) or high-density lipoprotein-3 (HDL₃) (25 μg/ml) was performed for 4 hours.⁸ Efflux was presented as the percentage of counts recovered from the medium in relation to the total counts (sum of medium and cells), with subtraction of efflux to medium with no acceptors, which was comparable among groups.

RNA-Seq library preparation and sequencing

RNA samples were extracted using All Prep DNA/RNA/miRNA Universal Kit (Qiagen, Valencia, CA). With a minimum of 300 ng input RNA, libraries were prepared using the TruSeq RNA Sample Preparation Kit (Illumina, San Diego, CA), followed by 100 bp paired-end sequencing on an Illumina’s HiSeq 2000 as we described.⁹

Alignment of RNA-Seq reads

As we described,⁹ RNA-Seq data were aligned to the hg19 reference genome using STAR 2.3.0e ¹⁰ with default options. Analyses were based on filtered alignment files. Mapping statistics are summarized in Online Table II. RNA-Seq data are available from the NCBI Gene Expression Omnibus (GEO) under the accession numbers GSE55536.

RNA-Seq data analysis and bioinformatics

Transcript abundance was measured in FPKM (fragments per kilobase of transcript per million fragments mapped) using Cufflinks 2.1.1.¹¹ Differential expression (DE) was tested with Cuffdiff, using the RefSeq annotation.¹² Genes with a false discovery rate (FDR)-adjusted P value <0.01 and a fold change (FC) >2 were considered differentially expressed. Multidimensional scaling (MDS) was done with Euclidean distance based on log₁₀ (FPKM + 0.1) in R programming languages.¹³ To visualize the overall sample to sample relationship within the data set, we performed co-regulation analysis (CRA) based on Pearson’s correlation coefficients by using BioLayout Express^{3D.14, 15} FPKM values were normalized using method from Anders and Huber¹⁶ separately for M0 (non-polarized macrophages), M1, M2 samples in HMDM and IPSDM. Normalized FPKM values were transformed by log₁₀ (FPKM + 0.1) on heat map.¹⁶ Heat maps illustrating expression patterns of DE genes were generated by using ggplot2 in R.¹⁷

Statistical analysis

Data were analyzed with GraphPad Prism 6 (GraphPad Software, San Diego, CA) and were presented as mean±SD. Statistical differences between groups were determined using Mann-Whitney test and Wilcoxon tests because non-parametric tests are more appropriate for small sample sizes. For analysis of gene ontology (GO) pathways in RNA-Seq data, significant enrichment was declared at FDR-adjusted P values <0.05 using Benjamini and Hochberg method.¹⁸ Enrichment analysis was performed in BiNGO plugin using Biological Process category and visualized in Cytoscape.¹⁹

RESULTS

Generation of subject-specific iPSCs

Subject demographics of PBMC-derived iPSC lines used in this study are listed in online Table III. As shown in online Figure I, iPSCs expressed typical pluripotency markers, maintained a normal karyotype, and exhibited silencing of exogenous transgenes beyond passage 10.

Efficient differentiation of human iPSCs to macrophages

We established a stepwise protocol for efficient IPSDM differentiation (Figure 1). With this approach, commitment to hematopoiesis was achieved at stage 2 in serum-free media containing different cytokine cocktails (Figure 1 and Online Table I). At D8, single cells emerged from the culture exhibited over 90% of CD43⁺/CD34⁺ hematopoietic cells. Around 40% of single cells expressed CD41 and CD235 and possessed restricted erythro-megakaryocytic potential but CD45⁺/CD18⁺ myeloid progenitors remained absent. From D8 to D15 in RPMI+ 20% FBS + 100 ng/ml M-CSF, a significant amount of CD45⁺/CD18⁺ myeloid progenitors emerged from the culture, and the CD41⁺/CD235⁺ lineage was lost. Single cells in suspension were harvested at D15 and were plated for adherent culture at about 0.1–0.15 X 10⁶/well of 6-well plate) for macrophage maturation. From D15 to D22, CD45⁺/CD18⁺ myeloid cells showed progressive maturation toward mature macrophages assessed by morphological changes (Figure 1). This protocol produces an almost pure (>95% CD45⁺/CD18⁺), high-yield macrophage population (up to 2×10⁷ of CD45⁺/CD18⁺ differentiated macrophages per 6-well plate of confluent iPSCs) and is highly consistent across the multiple lines/clones.

Figure 1. Schematic figure of IPSDM differentiation.

A stepwise protocol was used to differentiate iPSC to macrophages (IPSDM). Data shown are representative of 6 separate experiments.

IPSDM characteristics (Figure 2, A and C) recapitulated those of primary HMDM (Figure 2, B and D). Specifically, IPSDM exhibited macrophage-like morphology (both by phase-contrast imaging and May-Grünwald-Giemsa staining), phagocytosis of Alexa Fluor (AF) 594-labeled zymosan particles, uptake of Dil-label acetylated-low density lipoprotein (Ac-LDL). Immunofluorescence staining also demonstrated expression of CD68 and MCP-1 in IPSDM (Figure 2A). Flow cytometry revealed that IPSDM acquired expression of myeloid/macrophage markers CD18, CD11b, CD11c, CD14, CD16, CD115 and other well-defined macrophage markers (CX3CR1 and CCR2). Importantly, neither IPSDM (Figure 2C) nor HMDM (Figure 2D) were positive for markers of dendritic cells (CD1a and CD83), T-lymphocytes (CD3) or B-lymphocytes (CD19), indicating the specificity in myelomonocytic development.

Figure 2. IPSDMs are comparable to HMDMs at morphological and immunophenotypic levels.

(A) and (C) IPSDM exhibited macrophage-like characteristics as well as expression of macrophage markers; (B) and (D) show morphologic and phenotypic similarity between HMDM and IPSDM. Data are representative of 6 separate experiments. In (C) and (D), isotype controls were shown as gray shaded and open histograms respectively.

Taken together, these results showed that IPSDM and HMDM shared comparable morphological and phenotypic characteristics and that our protocol for IPSDM differentiation to macrophages is consistent, efficient and scalable.

Human IPSDM share functional characteristics with HMDM

Macrophages are plastic and efficiently respond to a variety of stimuli which results in a spectrum of macrophage phenotypic modulation that is observed in vivo during physiological and pathological stresses.¹ Within this spectrum, the classical M1-associated stimuli, LPS and IFN-γ, as well as M2-associated stimulus, IL-4, represent the stereotypic extremes of M1 and M2 axis in macrophage activation,^{15, 20} with the general perception that M1 macrophages are inflammatory and M2 macrophages promote tissue repair and metabolic homeostasis. In this study, HMDM and IPSDM were polarized to M1 and M2 states as shown in Figure 3A.

Figure 3. Functional phenotypes in response to M1 (LPS+IFN-γ) and M2 (IL-4) polarization are comparable between IPSDM and HMDM.

(A) Schematic protocol for HMDM and IPSDM polarization. (B) M1-HMDM and M1-IPSDM showed lower phagocytosis capacities (n=7 age/sex-matched subjects). HMDM and IPSDM derived from the same subject (Control-6) have comparable phagocytosis capacities (data represent mean of triplicate experiments). (C) ABCA1 and ABCG1 mRNA were upregulated in M1-HMDM and suppressed in M2-HMDM. M1-HMDM showed a clear pattern of increased efflux to apoA-I and HDL₃. IPSDM resembled HMDM in polarization-induced change in ABCA1/ABCG1 mRNA expression and cholesterol efflux. Each dot in the plots represents one subject. In addition, HMDM and IPSDM derived from Control-6 showed comparable efflux levels. (D) Representative cytokine array determining the relative level of 36 human cytokines and chemokines in culture media. HMDM and IPSDM derived from the same subject (Control-6) demonstrate similar cytokine profiles and response to polarization stimuli. (E) Cytokine array data of IPSDM and HMDM from 5 age/sex matched subjects are quantified by Image J and visualized.

Macrophage phagocytic activity is an essential early function in tissue remodeling and clearance of pathogens and dying cells in infection and inflammation. Using the CytoSelect Phagocytosis Assay kit (Cell Biolabs), we found that IPSDM were highly phagocytic, showing comparable engulfment of enzyme-labeled zymosan prepared from yeast as that of HMDM derived from the same subject (Figure 3B, Control-6, Caucasian, male. Relative to non-polarized and M2-IPSDM and M2-HMDM, both M1-IPSDM and M1-HMDM had reduced (~50%) phagocytic capacity (Figure 3B) consistent with reports that LPS treatment inhibits phagocytosis in human macrophages.²¹

Efflux of cellular cholesterol to lipoprotein acceptors reduces macrophage cholesterol accumulation in arterial neointima limiting atherosclerosis progression and promoting disease regression.² Efflux pattern of [³H]-labeled cholesterol to both apoA-I and mature HDL₃, in vivo acceptors for macrophage free cholesterol via ABCA1 and ABCG1 respectively, were almost identical in IPSDM and HMDM of the same subject (Figure 3C, Control-6). Cholesterol efflux capacity has not been reported in human polarized macrophages. During polarization, ABCA1 and ABCG1 mRNA were markedly upregulated in M1-HMDM (~ 6-fold and 20-fold respectively) and significantly reduced in M2-HMDM (Figure 3C). Although not statistically significant using non-parametric Wilcoxon tests in our small sample (n=4 subjects), M1-HMDM showed clear patterns of increased efflux to apoA-I (by ~2-fold, P=0.12) and HDL₃ (by ~40%, P=0.12) while M2 macrophages had small trends toward reduced cholesterol efflux to apoA-I (Figure 3C). IPSDM resembled HMDM in polarization-induced change in ABCA1/ABCG1 mRNA expression and cholesterol efflux (Figure 3C). The mechanism of polarization-related changes in ABCA1 and ABCG1 expression is not fully elucidated, but previous literature has reported that LPS induced ABCA1 expression through LXR-independent mechanisms in THP-1 monocytes.²² These findings support fidelity of IPSDM, relative to primary HMDM, in important macrophage cholesterol metabolic functions.

Basal and polarization-dependent secretion of cytokines and chemokines were determined in culture media (Figure 3D–3E) using semi-quantitative human cytokine array (R&D, Minneapolis, MN). Isogenic M0-IPSDM and M0-HMDM showed remarkably similar secretome profiles (Figure 3D, left panels). Polarization to M1-IPSDM resulted in secretion of multiple cytokines, e.g. IL-6, IP-10, RANTES, and TNF-α, in a pattern almost identical to that in M1-HMDM (Figure 3D, middle panels). In contrast to M1, secretome profile of M2-IPSDM, which mirrored that in M2-HMDM, differed very little from their non-polarized macrophage counterparts (Figure 3D, left vs. right panels). These secretome findings were generalizable across HMDM and IPSDM of five age/sex-matched subjects shown by heat map in Figure 3E.

Overall, these studies suggest secretome, phagocytic and efflux similarities between IPSDM and HMDM in both non-polarized and polarized states, and support a novel finding of loss of phagocytosis potential but increase in ABCA1/ABCG1 expression and cholesterol efflux capacity during M1 polarization (Figure 3B and 3C).

Human IPSDM share transcriptome characteristics with HMDM

An important issue in the field is to what extent somatic cells differentiated from iPSC lose hallmarks of pluripotent iPSC and adopt the transcriptomic characteristics of the somatic cells. To address this question in our IPSDM, we performed RNA-Seq (~130 million reads/sample, 95% mapping rate to the reference genome, ~70% reads uniquely mapped and filtered, see Online Table II) and compared transcriptome profiles of iPSC, M0-IPSDM, M1-IPSDM and M2-IPSDM (including duplicate clones for individual iPSC lines) as well as M0-HMDM and M1-HMDM and M2-HMDM of three healthy individuals Control-1, Control-3 and Control-4 (Figure 4A for schematic experimental design, and Online Table III for subject demographics).

Figure 4. RNA-Seq transcriptome analysis of iPSC, IPSDM and HMDM.

(A) Schematic figure of RNA-Seq design and data analysis (n=3 subjects, Control-1, Control-3 and Control-4, with two iPSC clones per subject). (B) Multidimensional scaling (MDS) and (C) co-regulation analysis (CRA) confirmed the distinct transcriptome profile of iPSC from HMDM and IPSDM. M1 polarization profoundly affects the transcriptional profile while M2 polarization results in more subtle expression changes. (D) IPSDM showed expression of macrophage markers and the absence of markers of pluripotency and other hepatopoietic cells. (E) and (F) show the total number of genes (FPKM > 1% expression of all genes) and DE genes in iPSC vs. IPSDM, and IPSDM vs. HMDM. The top 10 gene ontology (GO) terms ranked by FDR-adjusted P values for DE genes are illustrated. The size of the circle is inversely correlated to FDR-adjusted P values of enrichment in the respective GO term. The thickness of the lines corresponds to similarity between two GO terms. Scatterplot (F) suggests the strong Pearson’s correlation (r=0.85) of transcriptome between HMDM and IPSDM. Green, DE genes with higher expression in M0-HMDM; Red, DE genes with higher expression in M0-IPSDM; Blue, non-DE genes.

Marked transcriptome changes during iPSC transition to IPSDM

MDS (Figure 4B) and CRA (Figure 4C) revealed that IPSDM had very different transcriptome profiles relative to their precursor iPSC but displayed very similar profiles to that of isogeneic HMDM. Differentiation to IPSDM had profound effects on the global transcriptome. DE analysis (Cuffdiff) revealed 6,305 DE genes (3,470 upregulated, 2,835 downregulated, FDR-adjusted P <0.01, FC >2) (Figure 4E). iPSC to IPSDM transition resulted in 100 to >4,000-fold decreased expression of key pluripotency genes including LEFTY1, DNMT3B, and LIN28A. (online Table V), with a concomitant increase in the expression of macrophage genes including CD14, CD33, CD68, and CCL2, which were induced at levels comparable to that observed in HMDM (Figure 4D and on-line Table IX). Analysis of upregulated genes showed an over-representation of expected GO terms using BiNGO plugin ¹⁹ associated with immune response, defense response, response to wounding, and inflammatory response (Figure 4E, on-line Table VIII).

Transcriptome similarities between HMDM vs. IPSDM

At >1^th percentile FPKM expression level, we detected 13,418 expressed genes in HMDM and 13,471 genes in IPSDM (Figure 4F) with significant overlap (~99%) and strong correlation (Pearson’s r=0.85). DE analysis of HMDM vs. IPSDM revealed 89% genes to have similar expression (11,948 non-DE genes out of a total number of 13,331 genes expressed in both HMDM and IPSDM, online Table IX) with ~12% of genes (1,610) to be differentially expressed (FDR-adjusted P < 0.01, FC > 2). Of the 1,610 DE genes, 1,019 genes had higher expression in IPSDM, while 591 genes had higher expression in HMDM (Figure 4F, online Table X and XI). Genes expressed at higher levels in IPSDM include typical fibroblast markers (PDGFRA, PDGFRB, LOX, THY1, FGF1, TIMP1, TIMP3, ZEB1, CDH2), and several genes that encode collagen and extracellular matrix (COL11A1, COL3A1, COL1A1, COL1A2 and DCN), but importantly, no markers of undifferentiated mesenchymal stem cells e.g., GNL3 (online Table X). Of 591 genes with lower expression in IPSDM, 100 (20.5%) belonged to the immune response GO term (FDR = 8.64E-36 for enrichment) ¹⁸ (online Table XIII). This group of genes included several members of human leukocyte antigen (HLA) system corresponding to MHC protein complex, as well as a few of established M1 polarization markers. Importantly, though expressed at lower levels in M0-IPSDM vs. M0-HMDM, most of these chemokines/cytokines (e.g. CCL5, CXCL9 and CXCL10) and HLA genes (e.g. MHC class I, HLA-A, HLA-B, HLA-C, etc.) were markedly upregulated during M1- polarization of IPSDM to levels comparable to that in HMDM-derived M1 lines (online Table IV).

Transcriptome changes during M1 or M2 polarization in HMDM and IPSDM

MDS and CRA (Figure 4B–4C) demonstrated that polarization to M1-type macrophages markedly altered the transcriptional profile of both IPSDM and HMDM precursors. DE analysis in M0-HMDM vs. M1-HMDM and M0-IPSDM vs. M1-IPSDM revealed 1,931 and 1,643 upregulated DE genes respectively (FDR-adjusted P <0.01, FC >2); (online Table XIV–XV), with marked overlap between HMDM and IPSDM during M1 polarization (1,114 genes) (online Table XVI). Amongst these overlapped DE genes were key prototypic markers of M1 polarization, such as surface molecules CD86, the cytokine/chemokine genes, CXCL9, CXCL10, TNF, CCL5, IL6, IL8, and genes encoding intracellular protein GBP1 to GBP5, which were inducible by IFN-γ (online Table XVI). Consistent with our cytokine and chemokine secretion data (Figure 3D–3E) and earlier microarray datasets showing modest difference between IL-4 derived M2-HMDM and their M-CSF differentiated non-polarized HMDM precursors,²³ RNA-Seq of M2-HMDM and M2-IPSDM revealed quite subtle transcriptomic changes during M2 polarization of HMDM and IPSDM (Figure 4B–4C, on-line Table XXII–XXIII), suggesting that M-CSF differentiated macrophages are already shifted towards the M2 transcriptome and phenotype. Yet the overlapping DE genes (n=85) during M2-HMDM and M2-IPSDM polarization did include several classic M2 polarization markers, such as the F13A1, CLEC4A, MRC1 (online Table XXIV). Scatterplots of Log₂ FC of DE genes revealed a slightly stronger correlation between HMDM and IPSDM in M1 polarization (r = 0.80, Figure 5A), than that in M2 polarization (r = 0.71 Figure 5B). To eliminate noise from genes expressed at low levels and genes with small or non-significant FC, only genes with FPKM >1% expression and also defined as DE genes in either HMDM or IPSDM were included in the analysis. (Refer to online Table XXX to XXXV for the list of genes.)

Figure 5. IPSDM recapitulates transcriptome signature of M1 (LPS+IFN-γ) and M2 (IL-4) macrophage polarization.

(A) and (B) Venn diagrams show the total number and the overlapped genes in M1 and M2 macrophages in HMDM and IPSDM. Scatterplots illustrate the correlation of Log₂ FC of DE genes in M0 vs. M1 (A) and M0 vs. M2 (B) polarization between HMDM and IPSDM. Only genes with FPKM >1% expression and also defined as DE genes in either HMDM or IPSDM were plotted. Genes with the same direction of change were highlighted in red for up-regulation and green for down-regulation. Genes with opposite direction of change were highlighted in purple. The DE genes of M1 vs. M2 are separated into those expressed at higher levels in M1 (C), or in M2 (D). Venn diagrams of the number of common DE genes in HMDM and IPSDM, and GO enrichment analysis of these common DE genes are performed and the top 10 GO terms ranked by FDR-adjusted P values are visualized. (E) The heat maps depict expression profiles of subsets of well-established macrophage polarization markers for selected M1-enriched (top) and M2-enriched (bottom) genes in HMDM and IPSDM.

Known and novel macrophage polarization markers

We found higher expression of 1,790 and 1,623 genes (Figure 5C, HMDM and IPSDM, respectively; FDR-adjusted P <0.01, FC >2) in M1 vs. M2 macrophages with 1,082 overlapped genes (online Table XXXVI). GO analysis of these DE genes (online Table XXXVIII) showed enrichment for genes involved in immune response, defense response, inflammatory response, and response to wounding, suggesting the expected upregulation of host defense and inflammatory response programs in M1 polarization (Figure 5C). In contrast, 2,072 and 1,552 genes (Figure 5D) were identified as being elevated in M2 vs. M1 macrophages with 1,221 overlapped genes (online Table XXXVII). Representative GO terms (online Table XXXIX) include oxidation reduction, cellular respiration and small molecule metabolic process, suggesting distinct metabolic phenotypes in M2-polarized vs. M1-polarized macrophages (Figure 5D). Heat map visualization of well-established M1 and M2 like macrophage markers showed similar expression profile in both HMDM and IPSDM (Figure 5E). We also identified potential novel markers for M1 and M2 polarization by ranking the absolute value of Log₂ FC of genes with FPKM > 5% expression in M1 vs. M2-HMDM and M1 vs. M2-IPSDM, and the common top DE genes were classified by known functional categories and visualized by heat map (online Figure II).

Overall, RNA-Seq revealed distinct transcriptome profile of iPSC from HMDM and IPSDM, suggested M1 polarization profoundly affects the transcriptional profile while M2 polarization results in more subtle expression changes, confirmed known markers for M1 and M2 macrophage polarization and identified novel genes related to specific polarization programs that were common to both HMDM and IPSDM.

IPSDMs reproduce macrophage cholesterol defects in Tangier disease (TD)

Loss of function mutations in the ABCA1 transporter underlies the rare Mendelian disorder Tangier disease (TD).²⁴ Generation of iPSC from TD patients provides an opportunity to examine the reproducibility of the TD phenotypic defects in IPSDM, and to probe additional functional impact of ABCA1 deficiency in the IPSDM system. We recruited two TD individuals, TD-1, compound heterozygote at S2046R/K531N, and TD-2, homozygous for the E1005X/E1005X truncation mutation. The sibling of TD-1, heterozygous at K531N (Hetero-1) was recruited as a family comparator. Age, race and sex matched healthy subjects, Control-1 for TD-1 and Control-2 for TD-2, were also recruited.

IPSDM and HMDM of TD patients had similarly abolished cholesterol efflux to apoA-I and equally impaired efflux to HDL₃ (Figure 6A). The heterozygote ABCA1 mutation carrier had an intermediate defect in cholesterol efflux consistent with the presence of one functional allele (Figure 6A). Despite small sample sizes (n=3 replicates for TD-1/Hetero-1/Control-1 HMDM), it is apparent that efflux to apoA-I in TD-HMDM was completely abolished (Figure 6A, upper panel). Further, LXR agonists upregulated efflux in both Hetero-1 and Control-1 HMDM but failed to induce any cholesterol efflux of TD-1 HMDM (Figure 6A, upper panel). Relative to Control-2 IPSDM, the cholesterol efflux defect to apoA-I and HDL₃ in TD-2 IPSDM (Figure 6B) was almost identical to that observed in TD-1 studies, and was consistent in longer efflux studies (20-hour) and with higher concentration of HDL₃ (50 μg/ml) (online Figure III). Thus, IPSDMs recapitulate the key cellular defect in TD macrophages in a highly reproducible manner.

Figure 6. TD (Tangier disease)-IPSDMs recapitulate hallmark phenotypes of impaired cholesterol efflux and reveal novel inflammatory phenotype in TD-IPSDM.

(A) In HMDM, efflux of [³H]-labeled cholesterol to apoA-I was completely abolished, and efflux to HDL₃ was impaired in TD-1 and partially impaired in the heterozygote sibling. LXR agonists enhanced efflux in the Hetero-1 and age/sex/race-matched healthy control, but not in the TD patient. The TD-1 IPSDM faithfully reproduced the impaired cholesterol efflux function. (B) IPSDM of TD-2 and age/sex/race-matched Control-2 also showed abolished efflux to apoA-I and impaired efflux to HDL₃. The bar graph represents mean±SD of triplicate experiments for HMDM, and quadruplicate experiments of two clones per iPSC line of each subject for IPSDM. * P<0.05 vs. TD -9CisRA/22OH; # P<0.05 vs. TD +9CisRA/22OH; & P<0.05 for -9CisRA/22OH vs. +9CisRNA/22OH. (C) TD-HMDMs showed marginal increase in baseline expression of IL1B, IL8 and TNF, and are hypersensitive to LPS stimulus as evidenced by greater increases in inflammatory gene expression. IPSDM reproduce the enhanced inflammatory response in TD-HMDM. The ΔCts represent the mean cycle threshold for target genes relative to ACTB as the internal reference. n=4 replicates of Control-2 and TD-2.

ABCA1 deficiency in human IPSDMs leads to increased inflammatory response

In addition to its canonical role in mediating cholesterol efflux and reverse cholesterol transport, ABCA1 has been implicated in the interface of inflammation and cholesterol metabolism. Indeed, peritoneal macrophages isolated from macrophage-specific Abca1^−/− mice are hypersensitive to LPS stimulus and showed higher expression of inflammatory genes, including Il1b, Il6 and Tnf.²⁵ Macrophage-specific Abca1^−/−/Abcg1^−/− bone marrow-transplanted Ldlr^+/− mice after 10 to 11 weeks of Paigen diet showed increased inflammatory gene expression in macrophages in atherosclerotic lesions.²⁶ However, data on inflammatory response in human ABCA1 deficiency are lacking. Due to the limited availability of blood samples from subjects with rare Tangier disease, TD-IPSDMs provide a unique opportunity to study novel phenotypes in human ABCA1-deficiency. In LPS-primed (100 ng/ml for 4 hours) macrophages, inflammatory gene expression (IL1B, IL8, TNF and CCL5) was increased in both HMDM and IPSDM of control and TD subjects, but with a markedly greater response in TD-HMDM and TD-IPSDM relative to their controls (Figure 6C). In contrast, baseline expression of IL1B, IL8 and TNF, but not CCL5 in TD-HMDM was only marginally increased compared to Control-HMDM and did not differ between Control- and TD-IPSDM. Thus, IPSDM reveal inflammatory phenotypes in human TD-HMDM that may be of pathophysiological and clinical relevance. These findings also suggest the utility of IPSDM in screening for novel yet subtle macrophage defects in Mendelian-inherited disorders.

DISCUSSION

Despite a substantial appreciation for the dual function of macrophages in innate immunity and lipid metabolism, understanding of human macrophage biology has been hampered by the lack of reliable and scalable models for cellular and genetic studies. IPSDM, as an unlimited source of subject genotype-specific cells, not only can shed light on the molecular physiology of macrophages in complex cardiometabolic disorders, but also permit deeper insights into consequences of rare genetic variants associated with monogenic Mendelian disorders, where primary samples are often not readily accessible. In the current work, we combined human iPSC technology, RNA-Seq, and a Mendelian genetic disorder to gain new insights into the application of IPSDM in cell-specific functional disease modeling of human genetic disorders and in human macrophage biology. Using human iPSCs derived by our Sendai virus reprogramming protocol,⁷ We present 1) a relatively simple protocol that uses fewer cytokines for rapid, high-throughput generation of IPSDM of high purity and functional homogeneity.^{27, 28} 2) comparable morphological, functional and transcriptome profiles in IPSDM and HMDM, 3) remarkably similar in vitro functional plasticity and polarization in phagocytosis and cholesterol efflux capacity and inflammatory cytokine secretion in IPSDM and HMDM, 4) almost identical hallmark cholesterol-metabolism phenotypes in IPSDM and HMDM of TD patients, and 5) novel insights via IPSDM into macrophage inflammatory response in human ABCA1 deficiency. Thus, our IPSDM system paves the way towards large-scale applications in functional, disease and therapeutic modeling of macrophages in innate immunity and cellular cholesterol homeostasis in human.

We present the first high-resolution transcriptome map of isogenic HMDM and IPSDM during differentiation and polarization. These data highlight the fidelity of our IPSDM but also reveal challenges as well as opportunity for application to complex mechanistic studies. First, IPSDM lost expression of pluripotency markers, had remarkably distinct gene expression profiles relative to precursor iPSCs, and had similar gene expression as HMDM. Second, although our protocol yields ~ 95% pure CD45⁺/CD18⁺ myeloid cells, a CD45-negative population accounting for ~5% of the differentiated cells may have contributed to a minor difference in gene expression between HMDM and IPSDM. This might impact the use of IPSDM in macrophage modeling of certain genes associated with diseases, such as for PDGFD in coronary heart disease, FGF1 in cardiac hypertrophy, DCN in sudden cardiac arrest, and TIMP3 in age-related macular degeneration.²⁹ Thus, specific customization of the IPSDM protocol may be required to suit each question addressed. Our study represents a starting point and provides a resource to assist investigators in determining the appropriateness of the current IPSDM protocol and whether modifications are required. We note also the potential to eliminate the ~5% CD45-negative cells by fluorescence-activated cell sorting or magnetic beads separation when a highly purified macrophage population is needed. Third, and of some concern, there were 591 genes expressed at lower levels in IPSDM vs. HMDM, and these were enriched in immune response and defense response genes. Many of these, however, were upregulated during polarization to levels comparable to M1- or M2-HMDM (online Table IV) suggesting that polarization leads to more complete convergence of ISPDM and HMDM transcriptomes. Nevertheless, a few members of MHC protein complex class II (DOA, DP, DQ, and DR) and chemokines (CCL1, CCL18, and CCL22 etc.) were expressed at generally lower levels in IPSDM (online Table IV), suggesting that optimized protocols are required for specific applications. Fourth, consistent with functional and secretome similarities, expression profiles of IPSDM- and HMDM-derived M1 lines were highly correlated with each other and both were dramatically different to their respective IPSDM and HMDM precursors. This suggests particular value of the IPSDM system in modeling inflammatory and M1-type macrophage functions. Similarly, there was substantial transcriptomic overlap between IPSDM- and HMDM-derived M2 lines that included classical M2 polarization markers. Fifth, through RNA-Seq we identified many new genes modulated during polarization in both HMDM and IPSDM thus revealing novel, and potentially regulatory, polarization markers that warrant further study. Finally, the pairwise Pearson’s correlation of mRNA expression in all M0-HMDM and M0-IPSDM pairs (on-line Figure IV) reveal that the correlation between isogenic HMDM and IPSDM was quite high (ranging from 0.83 to 0.92). Yet, the correlation across HMDMs of the three subjects in our study was also high (0.90 to 0.96). This sample is too small to draw inference of within and across subject correlations but larger samples and diverse populations will result in larger variation between subjects. We emphasize, however, that the high expression correlation between isogenic HMDM and IPSDM is in support of the fidelity of IPSDM for human macrophage functional genomics studies.

We demonstrate the specific utility of IPSDM for interrogation of functional impact of human genetic variation in macrophages. IPSDM faithfully recapitulate the hallmark macrophage cholesterol homeostasis defects in TD, both in a qualitative and quantitative manner. Deficiency of ABCA1 in IPSDM also revealed a novel phenotype, of hypersensitivity to inflammatory stress, in human TD-IPSDM. Whether this relates to FC toxicity, as proposed in rodent cells,^{25, 30} or distinct signaling actions in human TD macrophages is uncertain but our IPSDM system provides a renewable tool for molecular studies to address the underlying mechanism(s). These observations provide support for the utility of IPSDM in defining subtle macrophage phenotypes in Mendelian disorders and for ongoing studies of common genetic variation for complex diseases such as atherosclerosis. This work lays the groundwork for large-scale use of genome editing technologies,³¹ in conjunction with iPSC-macrophage differentiation in cells of isogenic background, for modeling of genetic effects on human macrophage functions in Mendelian disorders (e.g., gene editing in isogenic wild-type IPSDMs to study novel inflammatory phenotypes in TDs) as well as in complex-disease macrophage genomics at the intersection of metabolism, immunity and cardiovascular diseases.

In summary, we describe an efficient protocol for generation of human IPSDM, establish functional and transcriptomic fidelity of IPSDM relative to HMDM, demonstrate equivalent plasticity in IPSDM and HMDM polarization in vitro, and recapitulate hallmark macrophage lipid phenotypes while revealing novel inflammatory defects in IPSDM of TD patients, a Mendelian disorder of macrophage function. Through high-resolution RNA-Seq of HMDM and IPSDM under non-polarized, M1-like and M2-like conditions, we also provide novel insights into the human macrophage function and transcriptome and identify gene clusters for further study in macrophage polarization and biology. This work suggests genome-to-phenome fidelity at the individual level during IPSDM differentiation and polarization. The IPSDM protocol provides a unique tool to study the macrophage-specific functions of novel genomic loci for human disease, to execute gene-editing strategies to prove causality, and to advance clinical and therapeutic translation of human genomic discoveries.

Novelty and Significance.

What Is Known?

• Macrophages represent a critical cell type at the intersection of metabolism, immunity and cardiovascular diseases.

• The differentiation of human induced pluripotent stem cells (iPSCs) to macrophages (iPSC-derived macrophages, or IPSDM) provides a ready source of subject genotype-specific cells with potential applications in disease modeling, drug screening and cell therapeutics.

What New Information Does This Article Contribute?

• We present the first high-resolution transcriptome map of isogenic human peripheral blood mononuclear cell-derived macrophages (HMDM) and IPSDM during differentiation and polarization.

• M1 activation of HMDM and IPSDM is associated with an apparent coordinate reduction in phagocytosis and increase in cholesterol efflux capacity as well as characteristic secretion of inflammatory cytokines whereas M2 activation of HMDM and IPSDM is characterized by a preservation of phagocytosis and trend toward reduced cholesterol efflux capacity with a distinct pattern of cytokine secretion.

• Tangier disease (TD) subject-specific IPSDM recapitulate hallmark macrophage lipid phenotypes of impaired cholesterol efflux, while revealing novel inflammatory phenotypes of enhanced inflammatory responses to lipopolysaccharide (LPS) stimulus.

iPSC technology offers a promising approach in vitro disease modeling and drug screening; however, it is unclear to what extent iPSC-differentiated cells adopt the functional and transcriptomic characteristics of their primary somatic cells. We developed a protocol for rapid, high-throughput generation of IPSDM of high purity and functional homogeneity. IPSDM, like their primary isogenic HMDM, can be polarized in vitro to functionally and molecularly distinct M1 and M2 subtypes, with in vitro functional plasticity and polarization in phagocytosis and cholesterol efflux capacity, secretome, and transcriptome profiles similar to native cells. TD-IPSDM showed identical hallmark phenotypes of impaired cholesterol efflux in HMDM of TD patients, while revealing novel insights into the enhanced inflammatory response to LPS stimulus due to human ATP-binding cassette transporter A1 deficiency. This work represents a significant step forward for the utilization of IPSDM in defining subtle macrophage phenotypes in Mendelian disorders, lays the groundwork for large-scale use of genome editing technologies in conjunction with IPSDM to model genetic effects of rare Mendelian mutations as well as novel genomic loci for complex diseases in human macrophages, and to advance clinical and therapeutic translation of human genomic discoveries.

Nonstandard Abbreviations and Acronyms

ABCA1

ATP-binding cassette transporter A1

ABCG1

ATP-binding cassette transporter G1

Ac-LDL

acetylated-low density lipoprotein

apoA-I

apolipoprotein A-I

CRA

co-regulation analysis

DE

differential expression

EB

embryoid body

FBS

fetal bovine serum

FC

fold change

FDR

false discovery rate

FPKM

fragments per kilobase of transcript per million fragments mapped

GO

gene ontology

HDL₃

high-density lipoprotein-3

HMDM

human PBMC-derived macrophage

IFN-γ

interferon-gamma

IL-4

interleukin-4

iPSC

induced pluripotent stem cells

IPSDM

iPSC-derived macrophage

LPS

lipopolysaccharide

LXR

liver X receptor

M-CSF

macrophage colony-stimulating factor

MDS

multidimensional scaling

PBMC

peripheral blood mononuclear cells

TD

Tangier disease