Single cell transcriptomics based-MacSpectrum reveals novel macrophage activation signatures in diseases - PubMed (original) (raw)

Single cell transcriptomics based-MacSpectrum reveals novel macrophage activation signatures in diseases

Chuan Li et al. JCI Insight. 2019.

Abstract

Adipose tissue macrophages (ATM) are crucial for maintaining adipose tissue homeostasis and mediating obesity-induced metabolic abnormalities, including prediabetic conditions and type 2 diabetes mellitus. Despite their key functions in regulating adipose tissue metabolic and immunologic homeostasis under normal and obese conditions, a high-resolution transcriptome annotation system that can capture ATM multifaceted activation profiles has not yet been developed. This is primarily attributed to the complexity of their differentiation/activation process in adipose tissue and their diverse activation profiles in response to microenvironmental cues. Although the concept of multifaceted macrophage action is well-accepted, no current model precisely depicts their dynamically regulated in vivo features. To address this knowledge gap, we generated single-cell transcriptome data from primary bone marrow-derived macrophages under polarizing and non-polarizing conditions to develop new high-resolution algorithms. The outcome was creation of a two-index platform, MacSpectrum (https://macspectrum.uconn.edu), that enables comprehensive high-resolution mapping of macrophage activation states from diverse mixed cell populations. MacSpectrum captured dynamic transitions of macrophage subpopulations under both in vitro and in vivo conditions. Importantly, MacSpectrum revealed unique "signature" gene sets in ATMs and circulating monocytes that displayed significant correlation with BMI and homeostasis model assessment of insulin resistance (HOMA-IR) in obese human patients. Thus, MacSpectrum provides unprecedented resolution to decode macrophage heterogeneity and will open new areas of clinical translation.

Keywords: Diabetes; Inflammation; Macrophages; Metabolism; Obesity.

PubMed Disclaimer

Conflict of interest statement

Conflict of interest: The authors have declared that no conflict of interest exists.

Figures

Figure 1

Figure 1. scRNA-seq profiles of cultured M0, M1, and M2 BMDMs and ATMs from lean and obese mice.

(A) scRNA-seq scheme. Cultured M0, M1, and M2 BMDMs were barcoded separately and processed together for scRNA-seq. ATMs from lean and obese mice were barcoded separately and processed together for scRNA-seq. (B and C) t-SNE clustering of scRNA-seq from 6979 M0, 4736 M1, and 6391 M2 BMDMs (B), and combined data from BMDMs (M0, M1, and M2) and 1710 lean and 1758 obese ATMs (C). (D) Expression of macrophage lineage markers (left panel), and well-known M1 (middle panel) and M2 markers (right panel) by t-SNE plots.

Figure 2

Figure 2. Generation of the macrophage polarization index (MPI).

(A) rm1 versus rm2 contour plots of 4736 M1 and M2 BMDMs using the top 500 most differentially expressed (absolute fold change) and significantly changed genes (FDR-adjusted P < 1 × 10–10) (PSG). (B) Polarization axis is the regression line of transcriptomes to M1 (rm1) or M2 gene sets (rm2). (C) MPI is calculated as l = P – P0. (D) Macrophage distributions along the MPI scale. (E) PSG-enriched pathways. (F) Heatmap of PSG. Known markers for M1 or M2 are indicated; complete gene list is in Supplemental Table 1. (G and H) MPI density distributions of human PBMC-derived macrophages stimulated with IFN-γ (GSE82227) (67) (G) and murine BMDMs after Salmonella exposure (GSE65528) (83) (H).

Figure 3

Figure 3. Comparison of BMDM scRNA-seq profiles.

(A) Similarity analyses of individual cells in M0, M1, and M2 BMDM samples were calculated using whole transcriptomes. An equal number (4736) of cells were randomly selected from each population; rows and columns represent individual cells from the 2 populations being compared, and the color of their crossing point represents adjusted correlation coefficient r; higher (yellow) and lower (blue) r suggest higher and lower similarity, respectively. (B) Bulk similarities between the M0, M1, and M2 BMDM populations as in A. Distances between populations are indicated next to the connecting lines; longer distances indicate more different from each other.

Figure 4

Figure 4. Generation of the activation-induced macrophage differentiation index (AMDI).

(A) Pathways significantly enriched in the 435 AMDSGs. (B) Macrophage distributions along the AMDI scale. (C and D) Single-cell trajectory of M0, M1, and M2 BMDMs (4736 cells/sample) with Monocle (30) using PSGs and AMDSGs colored by BMDM population (C) or pseudotime (D). (E) Heatmap showing smoothened relative expression of PSGs and AMDSGs along pseudotime progression, from M0 to M1 and M2 branches of the single-cell trajectory. Known signature genes are indicated. Hierarchical clustering generated 6 expression groups: groups I and III enriched in M1, group II enriched in M2 or M1 and M2, groups V and VI enriched in M0, group IV enriched in M0 or M2.

Figure 5

Figure 5. MacSpectrum characterization of visceral ATM subsets from lean and obese mice.

(A) Macrophage subsets on the MacSpectrum plot were designated as A,“M2-like”; B, “M1-like”; C, “transitional M1-like”; and D, “preactivation”. (B) MacSpectrum plot of 1710 lean and 1758 obese ATMs with percentages calculated for each region (A, B, C, or D). (C) Pathways enriched in lean (L) and obese (O) region-specific (A, B, C, or D) subpopulations. (D) Heatmap of ATM signature genes identified using MacSpectrum. Each gene was plotted as one row along MPI; all genes arranged by hierarchical clustering.

Figure 6

Figure 6. MacSpectrum identified unique gene sets associated with diabetes conditions in obesity.

(A) Heatmap showing expression of the 23 gene candidates identified using MacSpectrum in blood monocytes of 18 obese patients before and after bariatric surgery. (B and C) Correlation coefficients (r) of the 23 gene candidates with patients’ BMI (B) and HOMA-IR (C). (D and E) Correlation (D) of PRKAG2, ERCC1, AMOTL2, and BDH2 with HOMA-IR and BMI and their microarray-determined relative expression (E) in visceral adipose tissue CD14+ cells from 12 obese patients with (Dia) or without diabetes (No Dia) (GSE54350). Data represent mean ± SEM. *P < 0.05, **P < 0.01 by 2-tailed Welch’s t test.

Figure 7

Figure 7. Scheme showing the data processing pipeline of MacSpectrum.

References

    1. Davies LC, Jenkins SJ, Allen JE, Taylor PR. Tissue-resident macrophages. Nat Immunol. 2013;14(10):986–995. doi: 10.1038/ni.2705. - DOI - PMC - PubMed
    1. Martinez FO, Gordon S. The M1 and M2 paradigm of macrophage activation: time for reassessment. F1000Prime Rep. 2014;6:13. - PMC - PubMed
    1. Mosser DM, Edwards JP. Exploring the full spectrum of macrophage activation. Nat Rev Immunol. 2008;8(12):958–969. doi: 10.1038/nri2448. - DOI - PMC - PubMed
    1. Nathan CF, Murray HW, Wiebe ME, Rubin BY. Identification of interferon-gamma as the lymphokine that activates human macrophage oxidative metabolism and antimicrobial activity. J Exp Med. 1983;158(3):670–689. doi: 10.1084/jem.158.3.670. - DOI - PMC - PubMed
    1. Stein M, Keshav S, Harris N, Gordon S. Interleukin 4 potently enhances murine macrophage mannose receptor activity: a marker of alternative immunologic macrophage activation. J Exp Med. 1992;176(1):287–292. doi: 10.1084/jem.176.1.287. - DOI - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources