Connecting transcriptional and functional macrophage heterogeneity in atherosclerosis

Macrophage accumulation in the artery wall is a hallmark of atherosclerosis, and these immune cells play key roles in the initiation, progression and eventual rupture of atherosclerotic plaques.¹ Plaque macrophages have various origins: they can be recruited into the tissue,² can proliferate after being embryonically implanted³ or can even derive from transdifferentiated smooth muscle cells.⁴ Moreover, in-vitro imaging has established stage dependent recruitment of different types of monocytes and macrophages into the lesion site.⁵ Not surprisingly, emerging single cell technologies have unveiled that the plaque macrophage pool is vastly heterogeneous, containing a plethora of transcriptionally distinct sub-populations.^6–8 Connecting this transcriptional heterogeneity to functional phenotypes and in vivo cell behavior is a new challenge in understanding the role of macrophage dynamics in the pathogenesis of atherosclerosis.

In this issue of Circulation Research, McArdle et al.⁹ employ a multi-pronged approach combining single cell RNA sequencing (scRNA-Seq), bulk RNA sequencing, flow cytometry, microscopy and intravital imaging to connect the transcriptional heterogeneity identified in aortic plaque macrophages to specific functional behaviors. First, in early atherosclerotic plaques of Apoe^–/– mice fed 12 weeks of western diet, scRNA-Seq was performed on plaque leukocytes and macrophages were identified by genomically gating for Cd68 and Adgre1 (F4/80) expressing cells. K-means clustering revealed 4 distinct macrophage populations, which differentially expressed Cx3cr1 (cluster 1 and 3) and Itgax, the gene encoding for Cd11c (cluster 3 and 4). Pathway analysis revealed cluster associations with inflammatory macrophages (Cluster 1, Cx3cr1⁺), apoptotic macrophages (cluster 2, Cx3cr1^–/Cd11c^–), migrating and proliferative macrophages (cluster 3, Cx3cr1⁺, Cd11c⁺), and migrating and inflammatory phagocytic macrophages (cluster 4, Cd11c⁺). Based on these clusters, the authors generated Cx3cr1^GFP-Cd11c^YFP-Apoe^–/– reporter mice to allow for the isolation and phenotyping of these 4 distinct populations: GFP⁺ (corresponding to Cx3cr1⁺ cells in cluster 1), GFP^–YFP^– (corresponding to double negative cells in cluster 2), YFP⁺ (corresponding to Cd11c⁺ cells in cluster 3), and GFP⁺YFP⁺ (corresponding to Cx3cr1⁺Cd11c⁺ cells in cluster 4). Flow cytometric analysis of common macrophage cell surface markers showed that ~80% of the Cx3cr1 (GFP⁺) or CD11c (YFP⁺) cells were F4/80+ macrophages, and 60% of the double negative cells also expressed F4/80. Among the F4/80+ cells in each group, differing levels of another commonly used macrophage marker, CD64, were observed, suggesting that these markers do not parallel each other. However, MHC Class II could be used as a marker to cluster macrophages into subpopulations, suggesting that MHC class II may be an important determinant of aortic macrophage phenotype.

To begin to ascribe functional and behavioral phenotypes to the identified clusters, the investigators performed two-photon microscopic imaging of aortic explants of Cx3cr1^GFP-Cd11c^YFP-Apoe^–/– reporter mouse fed a western diet for 6–10 weeks. These studies revealed that the macrophages in aortic plaques assumed a variety of morphologies, with CX3CR1- GFP⁺ cells appearing larger than CX3CR1-GFP⁺CD11c-YFP⁺ cells, and CD11c-YFP⁺ cells appearing more spherical with fewer extensions than the other two groups of macrophages. Next, using intravital imaging, the investigators showed that these macrophage phenotypes persisted in advanced plaques of Cx3cr1^GFP-Cd11c^YFP- Apoe^–/– fed 5 months of western diet feeding, and showed distinct patterns of movement in the intima. All CX3CR1-GFP⁺CD11c-YFP⁺ macrophages showed a distinctive “dancing on the spot” movement pattern in which the cells extended their processes without any net migration. Like GFP⁺YFP⁺ cells, the majority of CX3CR1-GFP+ cells showed a dendritic-like cell shape and dancing behavior, whereas CD11c-YFP⁺ cells were round in shape and showed interstitial motility. While previous studies of macrophage morphology have linked cellular shape and dynamics to inflammatory polarization, with round macrophages linked to an M1-like state and elongated macrophages linked to an M2-like phenotype¹⁰, FACS sorting and deep RNA-sequencing of the 4 macrophage clusters failed to connect such functional differences with the transcriptional signatures. Interestingly, despite their dissimilar shape and movement, integrated pathway analysis showed that the gene signatures of Cx3cr1-GFP⁺CD11c-YFP⁺ and CD11c-YFP⁺ macrophages were both enriched for genes involved in cell protrusion and microtubule dynamics, cell migration, extracellular remodeling, and the complement pathway, while GFP+ macrophages showed activation of organismal death and phagocyte infiltration pathways. Analysis of eight categories of differentially expressed genes related to atherosclerosis showed differences in expression of integrin, scavenger receptor, chemokine/chemokine receptor, cytokine, cytoskeletal, and extracellular matrix related genes among the 4 macrophage clusters, but none that provided tangible clues to the specific functions of the dancing macrophages in atherosclerotic plaques. These findings highlight the prevailing challenges of integrating datasets generated using traditional cellular markers, functional phenotypes and transcriptional signatures to better understand the diverse roles of macrophages in atherosclerosis.

The recent surge of interest in illuminating macrophage heterogeneity in atherosclerosis has generated a number of bulk and scRNA-seq datasets available for comparison with the current study. First, using bulk RNA-seq transcriptomes of atherosclerotic and healthy mouse aortas published by Habernicht and colleagues¹¹, McArdle et al showed that CX3CR1-GFP+, CD11c-YFP+ and CX3CR1-GFP+CD11c-YFP+ gene signatures were each significantly enriched in the transcriptome of atherosclerotic aortas, with genes associated with CX3CR1-GFP+ macrophages specifically enriched in aortas exhibiting tertiary lymphoid organ formation. Interestingly, CX3CR1-GFP+ macrophage gene signatures were also more enriched in the transcriptome of non-foamy plaque macrophages than lipid-laden plaque macrophages reported by Randolph and colleagues¹², and in transcriptome of macrophages from regressing plaques than progressing plaques published by Fisher and colleagues¹³. Interestingly, genes upregulated in CD11c-YFP+ macrophages were enriched in the transcriptome of macrophages from progressing plaques¹³, which correlated with increased expression of the proinflammatory cytokines Il12b and Il6 in this subset, suggesting a potential pro-atherosclerotic function for CD11c-YFP+ macrophages. Attempts were also made to compare CX3CR1-GFP+, CD11c-YFP+ and CX3CR1-GFP+CD11c-YFP+ gene signatures with 3 recently published single cell RNA-seq datasets from atherosclerotic mice. Despite differences in the number and naming of macrophage clusters defined in each scRNA-Seq study, these comparisons yielded interesting insight into the CX3CR1-GFP+, CD11c-YFP+ and CX3CR1-GFP+CD11c-YFP+ macrophage subsets. The CX3CR1-GFP⁺ macrophage gene signature was most similar to the inflammatory macrophage cluster reported by Cochain et al.⁶, macrophage clusters involved in NF-kB signaling and cytokine signaling reported Kim et al.¹², and clusters described as chemokine^hi and IFN signature^hi by Lin et al⁷. The CD11c-YFP⁺ macrophage gene signature showed similarities with the Trem2^hi macrophage cluster reported by Cochain et al^{6, 7} and with a cluster associated with metabolic pathways defined by Kim et al¹². The gene signatures of CX3CR1-GFP+CD11c-YFP+ macrophages, which were predicted to be proliferative and migratory, were found to be enriched in resident-like⁶ and stem-like⁷ macrophage clusters defined by Cochain and Lin, respectively. Despite these similarities, many of the scRNA-Seq defined clusters showed transcriptional overlap with more than one of the CX3CR1-GFP+, CD11c-YFP+ and CX3CR1-GFP+CD11c-YFP+ macrophage subsets, again highlighting the difficulty of distilling complex macrophage phenotypes identified using different methods to easily defined subsets. Moreover, the gene signature of double negative macrophages did not match any of the bulk or the single cell RNA-Seq transcriptomic clusters, leaving their phenotype and function unresolved.

In summary, the work of McArdle et al provides an exemplar for how researchers can begin to integrate various technologies to elucidate distinct cellular transcriptional and functional phenotypes to better understand the behavior and role of macrophages and other immune cells in vivo. In the atherosclerosis field, vast amounts of information have been generated about the phenotypic complexity of macrophages within the plaque microenvironment using conventional techniques such as flow cytometry, immunohistochemistry and bulk RNA-Seq. A major challenge currently facing the field is integrating these findings with emerging scRNA-seq and scATAC-seq datasets and developing standardized nomenclature for the multitude of cell phenotypes that will emerge. This complexity will no doubt grow as future studies begin applying these integrated approaches to understand how macrophage subpopulations and their phenotypes shift in response to environmental cues present within the plaque (e.g. hypoxia, necrosis, oxidative stress, cholesterol) and across disease pathogenesis, as is now warranted.