Multiplex Imaging for Cell Phenotyping of Early Human Atherosclerosis

Abstract

Background

Previous studies using animal models and cultured cells suggest that vascular smooth muscle cells (SMCs) and inflammatory cytokines are important players in atherogenesis. Validating these findings in human disease is critical to designing therapeutics that target these components. Multiplex imaging is a powerful tool for characterizing cell phenotypes and microenvironments using biobanked human tissue sections. However, this technology has not been applied to human atherosclerotic lesions and needs to first be customized and validated.

Methods and Results

For validation, we created an 8‐plex imaging panel to distinguish foam cells from SMC and leukocyte origins on tissue sections of early human atherosclerotic lesions (n=9). The spatial distribution and characteristics of these foam cells were further analyzed to test the association between SMC phenotypes and inflammation. Consistent with previous reports using human lesions, multiplex imaging showed that foam cells of SMC origin outnumbered those of leukocyte origin and were enriched in the deep intima, where the lipids accumulate in early atherogenesis. This new technology also found that apoptosis or the expression of pro‐inflammatory cytokines were not more associated with foam cells than with nonfoam cells in early human lesions. More CD68⁺ SMCs were present among SMCs that highly expressed interleukin‐1β. Highly inflamed SMCs showed a trend of increased apoptosis, whereas leukocytes expressing similar levels of cytokines were enriched in regions of extracellular matrix remodeling.

Conclusions

The multiplex imaging method can be applied to biobanked human tissue sections to enable proof‐of‐concept studies and validate theories based on animal models and cultured cells.

Nonstandard Abbreviations and Acronyms

IL‐1β

interleukin‐1β

MYH11

myosin heavy chain 11

L‐FC

leukocyte foam cells

PIT

pathologic intimal thickening

SMA

α‐smooth muscle actin

SMC

smooth muscle cell

SMC‐FC

SMC foam cells

TNF‐α

tumor necrosis factor α

Clinical Perspective.

What Is New?

• For the first time, multiplex imaging using PhenoCycler technology was customized and validated to determine cell phenotypes in the tissue context of human atherosclerotic lesions, which is difficult to do using conventional microscopy and single‐cell RNA sequencing.

• Applying multiplex imaging in biobank samples proved the human relevance of studies based on animal models and cultured cells: the association between inflammation and cell phenotypes.

What Are the Clinical Implications?

• Connecting cell phenotypes with their microenvironment in human atherosclerotic lesions can be achieved, which is important for designing anti‐inflammatory therapies for secondary prevention of adverse cardiovascular events.

• Using this flexible and scalable approach, a small amount of biobanked human samples can unveil the association between risk factors and disease progression, identifying therapeutic targets and providing insights into the pathophysiology of atherosclerosis.

Cardiovascular disease is the leading cause of death and the costliest disease globally. Eighty‐five percent of these deaths are due to heart attack and stroke, the outcomes of atherosclerosis. ¹ Despite tremendous efforts in preclinical research, mechanisms learned from cell and animal models do not always apply to human disease. A proof‐of‐concept study using human atherosclerotic lesions is needed to confirm that observations in animal studies or cultured cells apply to human atherogenesis. Our previous review has summarized the opportunities and challenges of using archived human atherosclerotic lesions for proof‐of‐concept studies. One of the challenges is the lack of spatial characterization tools to identify the connection between cell phenotypes and lesion progression. ²

An example of a challenge in atherosclerosis research is defining the roles of smooth muscle cells (SMCs) in disease development. Allahverdian et al. reported that >50% of the foam cell population in human coronary lesions is of SMC origin. ³ One may hypothesize that they can become apoptotic and pro‐inflammatory, and contribute to the formation of a necrotic core, just like leukocyte‐derived foam cells (L‐FC). Single‐cell RNA sequencing (scRNA‐seq) has revealed 3 phenotypes of SMCs from human carotid lesions: contractile, transitional, and de‐differentiated. ⁴ , ⁵ Although this powerful high‐throughput method can provide cell phenotypes in the entire lesion, the analysis algorithm uses classic cell lineage markers to annotate different cell clusters. There is no reference gene expression data set to specifically define foam cells. A more straightforward way is to use a lipid dye and visualize foam cells in tissue sections. ⁶ However, due to limitation in the number of cell markers that can be included in 1 tissue section by conventional tissue staining, it has been difficult to characterize the features (eg, expression of pro‐inflammatory cytokines) of SMC‐derived foam cells (SMC‐FC) to predict their roles in lesion progression.

Another challenge is in interpreting the presence of inflammatory cytokines in atherogenesis. For example, the inflammatory cytokine interleukin‐1β (IL‐1β) has typically been considered proatherogenic by inducing more SMC migration into lesions. ⁷ , ⁸ However, incorporation of SMCs into the athero‐protective fibrous cap also requires IL‐1β. As shown in animals with established fibroatheroma, blocking IL‐1β resulted in lesion destabilization, ⁹ suggesting that an understanding of the spatial distribution of IL‐1β and its effect on different cell phenotypes and structural progression is needed to confirm whether it is an appropriate therapeutic target. Unanswered questions include the following: Are inflammatory cytokines associated with SMC transdifferentiation, which may contribute to the formation of SMC‐FC? Are they associated with inflammation of L‐FC and extracellular matrix remodeling? Again, scRNA‐seq cannot address these region‐specific questions.

In this pilot study, we aim first to build a framework for using PhenoCycler technology to characterize human atherosclerotic lesions and identify foam cells that likely originate from SMCs and leukocytes. Using a small number of samples, we will then use this framework to test theories derived from cell culture and animal studies in human subjects, such as the roles of inflammatory cytokines.

Methods

Data Availability

The data that support the findings of this study appear in the article. Raw imaging data and other technical data are available from the corresponding author upon reasonable request.

Study Approval

Human coronary lesions were obtained from explanted hearts donated by heart transplant patients. These studies were conducted with the approval of the Providence Health Care Research Ethics Board (H21‐01069). Informed consent was obtained from all subjects through the Bruce McManus Cardiovascular Biobank at the Centre for Heart Lung Innovation, University of British Columbia. The investigation adhered to the principles outlined in the Declaration of Helsinki.

Human Samples

Coronary artery tissues were obtained from 5 patients diagnosed with dilated cardiomyopathy (n=3), hypertrophic cardiomyopathy (n=1), or giant cell myocarditis (n=1). These tissues were subjected to Oil Red O staining and Movat's pentachrome staining to identify early coronary lesions enriched with foam cells. Disease stage of lesions was categorized by a pathologist, based on the modified American Heart Association classification. ¹⁰ Nine early lesions classified as pathologic intimal thickening (PIT) and 1 early fibroatheroma were selected for imaging with the PhenoCycler System (Akoya Biosciences). When multiple tissue sections from 1 patient were imaged, these sections were considered different biological samples because they were taken from different PIT lesions at distinct branches of the coronary vessels.

Validation of Self‐Conjugated Antibodies

CD45, cleaved caspase 3, IL‐1β, and tumor necrosis factor α (TNF‐α) carrier‐free antibodies were self‐conjugated to oligonucleotide barcodes following standard procedures detailed in the manufacturer's guidelines and previous studies. ¹¹ , ¹² The sequence of each oligonucleotide barcode is unique to the conjugated antibody (Table S1). Myosin heavy chain 11 (MYH11) antibody was purified using a BSA removal kit (Abcam, CAT #ab173231) and then conjugated to a barcode. To validate that the conjugation process did not change the antibodies' capability to recognize target proteins, we visually compared the staining pattern of adjacent tissue sections, 1 stained with the conjugated antibodies per the manufacturer's protocol and the other with the original antibodies using regular immunofluorescence. Formalin‐fixed paraffin‐embedded sections of human atherosclerotic lesions were dewaxed and rehydrated. Antigen retrieval was performed by boiling the samples in 10 mmol/L sodium citrate, 0.05% Tween buffer (pH 6.0) for 30 minutes using a pressure cooker. Subsequently, slides for conjugated antibodies were stained using a PhenoCycler staining kit (Akoya Biosciences, Cat #7000008). Adjacent sections were blocked with 10% goat serum for 60 minutes and incubated with the original primary antibodies at 4° C overnight. The target markers were visualized with Alexa Fluor secondary antibodies (Invitrogen) and imaged with a Zeiss LSM880 inverted confocal microscope. Conjugated antibodies showing a consistent staining pattern with the original antibodies were considered validated for PhenoCycler imaging.

Multiplex Tissue Staining and Imaging

OCT‐embedded frozen human coronary lesions 8‐μm thick were stained using all barcoded antibodies (Table S1) following the manufacturer's protocol (Akoya Biosciences, Cat #7000008). Multiplex imaging was achieved by visualizing barcoded antibodies through several imaging cycles. Each cycle started by adding 3 types of reporters that carry 1 of the following fluorescent tags: Alexa Fluor 488, Atto 550, or Alexa Fluor 647, as well as oligonucleotides with complementary sequences to specific barcodes attached to the antibodies. In this way, fluorescent images of 3 antibodies were captured at a time and then the reporters were stripped before the next cycle to visualize a new set of 3 different antibodies. To detect foam cells, the fluorescent lipid dye BODIPY 493/503 was used as previously described. ⁶ This dye was added into the reporter stock solution on the last imaging cycle of the PhenoCycler so that its unremovable fluorescence staining would not interfere with signals of any other 488‐conjugated antibodies. All imaging cycles were performed using the PhenoCycler fluidic system (Akoya Biosciences). We used the Keyence BZ‐X800 fluorescence microscopy (20× objective) to acquire tile images.

Image Analysis

Cell segmentation: Images were processed with the PhenoCycler processor (version 1.8.3.14) to align imaging cycles and stitch image tiles. Processed images were analyzed using the HALO image analysis software (Indica Labs) HighPlex FL module. Nuclear identification was carried out using the DAPI channel with a nuclear contrast threshold of 0.52, segmentation aggressiveness of 0%, and minimum nuclear roundness of 0% to detect nuclei of various shapes. The cell segmentation algorithm was benchmarked using 3 different methods. Method 1 assigned any signals within a 2‐μm radius to the closest nucleus and considered them to belong to that cell. Method 2 used a 5‐μm radius. Method 3 considered signals that overlap with (within 0‐μm radius) both the nucleus and a cytoplasmic protein marker as signals inside of 1 cell. Each method was compared against manual counting.

Cell annotation: Cell phenotypes were defined using a combination of inclusion and exclusion of different classic cell lineage markers (CD45: leukocyte, MYH11: SMC, CD31: endothelial cell) and other protein/dye markers within a 2‐μm radius for cell segmentation. Leukocytes were defined as CD45⁺MYH11⁻CD31⁻ cells and smooth muscle cells as MYH11⁺CD45⁻CD31⁻ cells to exclude the possibility that cells in close proximity were misinterpreted as transdifferentiated cells. Subgroups of these 2 cell types included additional criteria of being BODIPY⁺ (foam cell) and BODIPY⁻ (non–foam cell), cleaved caspase 3⁺ (apoptotic), and cleaved caspase 3⁻ (nonapoptotic). For cytokines, the intracellular fluorescence intensity of the cytokine above the geometric mean (top 50%), calculated in both SMCs and leukocytes, was considered as high, and below the geometric mean as low.

Spatial visualization and analysis: We performed infiltration analysis of foam cells by measuring foam cell density in five 50‐μm‐thick adjacent layers, starting from the endothelial cell layer of the blood vessels, and going down to the deep intima before reaching the medial layer. CD68⁺ SMCs were separately quantified in the intima and the media, as divided by the internal elastic lamina.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism and JMP5. The normality of distribution was assessed using the D'Agostino & Pearson and Shapiro–Wilk normality tests (α=0.05). Normally distributed data were presented as mean±SEM, while non‐normally distributed data were presented as median and interquartile range. Normally distributed data with equal variances were analyzed using paired t tests or 1‐way ANOVA with Dunnett's or Bonferroni's multiple comparison tests. Data that did not meet parametric assumptions were analyzed using the Wilcoxon or Kruskal–Wallis tests. The Friedman test was used for nonparametric matched measurements. Statistical significance was defined as P values <0.05.

Results

Benchmarking PhenoCycler Staining of Human Coronary Lesions

The PhenoCycler imaging panel comprised 8 markers: DAPI, CD45, MYH11, CD68, cleaved caspase 3, IL‐1β, TNF‐α, and BODIPY. BODIPY 493/503 fluorescent lipid staining was incorporated into the staining panel to visualize lipid droplets by PhenoCycler imaging. This dye has been used previously to detect lipids in mouse atherosclerotic tissues. ⁶ We validated that it is compatible with multiplex imaging and that staining adjacent tissue sections with BODIPY and Oil Red O yields similar results (Figure 1A).

Figure 1. Establishing a multiplex imaging method to quantify different types of foam cells in human atherosclerotic lesions.

A, Representative Oil Red O, BODIPY lipid staining, and Movat's pentachrome staining of the adjacent tissue section (scale bar=200 μm). B, Representative images of the 8 PhenoCycler markers used to phenotype foam cells on a single tissue section (red scale bar=500 μm, white scale bar=50 μm). *=Lumen, I=Intima, M=Media, A = Adventitia. C, Performance of 3 cell segmentation methods for quantifying SMC‐foam cells. The diagram shows the schematic detection of the BODIPY signal within an SMC using: Method 1—A 2‐μm radius from nuclei; Method 2—A 5‐μm radius from nuclei; or Method 3—A 0‐μm radius from MYH11 signal (left panel). A graph shows the comparison of Methods 1–3 with manual counting (right panel). Min to Max, Kruskal–Wallis test with Dunn's multiple comparison test (n=9 per group). *, P≤0.05; **, P≤0.001. D, Patient age, sex, and foam cell counts. FC indicates foam cell; L, leukocyte; and SMC, smooth muscle cell.

Validated antibodies, including self‐conjugated antibodies (Figure S1), and the lipid dye maintained their distinct staining patterns when applied simultaneously on 1 tissue section. This is a standard quality check ¹³ to confirm that no artificial overlaps were introduced as a result of carryover signals from earlier imaging cycles (Figure 1B). We used a commercially available HighPlex FL module to analyze the image of human coronary lesions that contain cells of varying sizes and shapes, as well as expected signals (lipids) within the cells and in the extracellular matrix. We used SMC‐FC to optimize the overall analyzed cell radius because SMCs are large lesion cells with elongated shapes (up to 200 μm by length and 5 μm by width). ¹⁴ Compared with round‐shaped cells, such as monocytes, a long radius to compensate for the length of SMCs will increase the chances of counting extracellular signals as intracellular. We found that Method 2 led to an overestimation of SMC‐FC by counting extracellular BODIPY signals as intracellular (Figure 1C). Method 1 used a 2‐μm radius and generated results similar to manual counting. Method 3 is the most stringent algorithm that excludes any extracellular signals, which applies to all lesion cells. Method 1 and Method 3 generated similar results (Figure 1C), but Method 3 relies on the presence of a cytoplasmic marker. Hence, Method 1 was selected for cell segmentation.

Foam Cell Origins and Spatial Distribution

To determine that our imaging method provides results consistent with previous findings that SMC‐FC contribute to the majority of foam cells in human coronary lesions, ³ we quantified the absolute numbers of SMC‐FC and L‐FC in the intima of the 9 PIT lesions and found that the ratio of SMC‐FC to L‐FC is ≈9:1 (Figure 1D). Infiltration analysis also indicated that the number of SMC‐FC increases from the endothelial cell layer to the deep intima of the arterial wall, while L‐FC exhibit the opposite pattern of distribution (Figure 2A). Combining results from both cell count and spatial analysis results, our data confirm that SMCs in the intima are the major reservoir of excess cholesterol in early atherogenesis and the framework we built is reliable to characterize human atherosclerotic lesions.

Figure 2. Multiplex imaging reveals the spatial distribution of different phenotypes of lesion cells.

A, Illustration of areas of interest in the infiltration analysis (a). Oil Red O staining of the adjacent tissue section (a; left panel) marks the orientation and depth of infiltration in the area of interest from the luminal side (red, 0–50 μm) to the deep intima (blue, 200–250 μm) of a PIT lesion (a; right upper panel, scale bar=200 μm). Spatial visualization of SMC‐foam cells (blue dots) and leukocyte‐foam cells (red dots) in this area of interest (a; right bottom panel). A histogram showing the spatial distribution of foam cells in PIT lesions in the defined regions from the luminal side to the deep intima (b). Mean with SEM, 1‐way ANOVA with Dunnett's multiple comparison test (n=9 per group). B, Representative PhenoCycler images of CD68 expression in leukocytes (upper left panels) and SMCs (bottom left panels) in the intima. White arrows point to CD68⁺ cells (scale bar=20 μm). A quantitative analysis comparing CD68 expression in foam (BODIPY⁺) vs. nonfoam (BODIPY⁻) SMCs (bottom right panel) and leukocytes (upper right panel). Median with interquartile range, Wilcoxon matched‐pairs signed‐rank test, ns (n=9 per group). C, A representative spatial plot showing the distribution of CD68⁺ (blue dots) and CD68⁻ SMCs (yellow dots) with Movat's pentachrome staining of the same area (scale bar=200 μm; left panels). I=intima, M=media, A=Adventitia. A quantitative analysis comparing CD68 expression in the intimal vs. medial layers of the coronary arteries (right panel). Median with interquartile range, Wilcoxon matched‐pairs signed‐rank test, P=0.0039 (n=9 per group). D, Percentage of CD68⁺ SMCs in IL‐1β⁺ (left panel) and TNF‐α⁺ (right panel) SMCs that expressed different levels of inflammatory cytokines. Median with interquartile range, Wilcoxon matched‐pairs signed‐rank test, P=0.0039 (left panel), ns (right panel), n=9 per group. E, Percentage of apoptotic cells in SMCs (a, b) and leukocytes (c, d) with different levels of cytokine expression. Median with interquartile range, Wilcoxon matched‐pairs signed‐rank test, P=0.0078 (a, b), ns (c, d), n=9 per group. (e) Representative spatial plots showing SMCs (blue) and leukocytes (red) with high cytokine expression, SMCs and leukocytes with low cytokine expression (yellow) in a PIT (left panels) and fibroatheroma (right panels). The highlighted regions and enlarged Movat's images of adjacent sections indicate the intima regions (dashed lines) included in the spatial analysis and zoomed‐in areas of interest with extracellular matrix remodeling (scale bar=200 μm). FC indicates foam cell; IL‐1β, interleukin‐1β; L or Leu, leukocyte; MYH11, myosin heavy chain 11; ns, nonsignificant; PIT, pathologic intimal thickening; SMC, smooth muscle cell; and TNF‐α, tumor necrosis factor‐α.

Biological Process in Early Atherosclerotic Lesions

Previous studies suggested that the formation of SMC‐FC is associated with the expression of a macrophage marker CD68 in SMCs. ¹⁵ We found that a similar portion of SMC‐FC and SMC non‐foam cells express CD68 (Figure 2B), suggesting that CD68 expression is not solely tied to foam cell formation or taking up cholesterol from the microenvironment. Approximately 12% of MYH11⁺ SMCs in the intima expressed CD68, while <2% in the media were CD68⁺ (Figure 2C). This evidence suggests that the expression of CD68 in MYH11⁺ SMCs is triggered by microenvironment in the intima.

The lesion microenvironment is enriched with pro‐inflammatory cytokines accumulated in the intima during atherogenesis. We found that most of the signal for TNF‐α and IL‐1β resided in SMCs (Figure S2A), but this was not related to foam cell formation since neither SMC‐FC nor L‐FC expressed higher inflammatory cytokines (TNF‐α and IL‐1β) compared with non‐foam cells (Figure S2B). Among IL‐1β⁺ SMCs, more SMCs that highly express IL‐1β gained expression of CD68, a phenomenon not seen in TNF‐α⁺ SMCs (Figure 2D), suggesting the specific connection between the IL‐1β signaling pathway and SMC phenotypes. Compared with leukocytes expressing similar levels of cytokines (Figure 2E, c–d), a trend of increased apoptosis was seen in SMCs that highly expressed either TNF‐α or IL‐1β (Figure 2E, a–b). Instead, highly inflamed leukocytes (TNF‐α^high or IL‐1β^high) were found in regions with extracellular matrix remodeling in PIT and early fibroatheroma lesions. SMCs expressing similar levels of inflammatory cytokines were not enriched here, suggesting a unique role of inflamed leukocytes in driving disease progression to the advanced stage (Figure 2E, e). Combined, our data indicate that inflamed SMCs and leukocytes activated different cellular processes.

Discussion

Multiplex imaging is a new stream of in situ phenotyping technology that aims to increase the output of information obtained from a limited number or amount of tissue biospecimens. Previous applications include locating different phenotypes of cells simultaneously on 1 tissue section, finding cell distribution patterns that are correlated with pathological diagnosis, and inferring cell–cell interactions. ¹² , ¹³ , ¹⁶ , ¹⁷

The first antibody panels and analysis software for the PhenoCycler multiplex imaging platform were tailored for characterizing immune cells. ¹² As far as we know, this technology has never been applied to characterize cells in human atherosclerotic lesions. We selected a well‐accepted lesion cell phenotype, SMC‐FC, to benchmark this new application. This application is of particular interest to atherosclerosis research because SMC‐FC are recognized by the accumulation of intracellular lipid droplets but do not yet have well‐defined gene or protein markers to be easily classified by scRNA‐seq. Another concern in the phenotyping of atherosclerotic lesions is cell trans‐differentiation. As reported in cell‐lineage tracing mouse models, CD31⁺ endothelial cells can express the SMC contractile protein α‐smooth muscle actin (SMA) ¹⁸ and SMCs can express the macrophage marker CD68. ¹⁹ Moreover, leukocytes and endothelial cells can also express SMA both in atherosclerotic lesions and in injured blood vessels. ¹⁸ , ²⁰ , ²¹ , ²² Currently, MYH11 is the most specific SMC lineage marker that is not expressed by other cells. ²³ , ²⁴ Therefore, our criteria to define SMCs included MYH11 and excluded CD31⁺ cells and CD45⁺ cells in close proximity. It may rule out de‐differentiated SMCs that have lost MYH11. However, previous scRNA‐seq data have shown that SMCs do not completely lose MYH11 expression until they trans‐differentiate to a fibrochondrocyte phenotype in advanced mouse and human lesions. ²⁵ Hence, we do not expect an underestimation of MYH11⁻ SMCs. Confirming our expectation, the robust MYH11 protein expression was detected in human PIT lesions. Our definition of SMCs also excluded any CD45⁺ SMCs. Although we cannot completely exclude the presence of CD45⁺ SMCs in human lesions, they are rarely detected in SMC‐lineage tracing mouse models of atherosclerosis by flow cytometry and scRNA‐seq, ⁶ , ²⁶ as well as in scRNA‐seq of human lesions. ²⁷ CD45 was found in >25% of SMC‐derived stem cells in the adventitia after ligation‐induced injury in mice ²⁸ ; however, the multipotency of these stem cells is much higher than that of SMCs in atherosclerotic lesions.

Compared with the previous estimation that more than half of foam cells in advanced lesions of mice are derived from SMCs, ⁶ we observed a higher ratio of SMC to leukocyte foam cells (9:1) in early human lesions. This could be attributed to the intrinsic differences between human and mouse intima. Normal blood vessels from mice have a thin intima with no SMCs. In contrast, humans develop diffuse intimal thickening enriched with SMCs, a phenomenon found in coronary arteries as early as 36 weeks of gestation. ²⁹ During early atherogenesis in humans, lipid accumulation starts in the deep intima where there is little colocalization with macrophages, which are concentrated on the apical side. ³⁰ This distribution suggests that extra lipids trapped in the extracellular matrix of the intima would be surrounded by a large number of SMCs before monocyte infiltration. Our multiplex imaging confirmed that these lipids in the deep intima were taken up by SMCs to become SMC‐FC. The distribution patterns of SMC‐FC and L‐FC also support the assumption that SMCs initiate early foam cell formation in human lesions. ³¹ These concepts cannot be proven without quantitative spatial analysis. Previous studies using light microscopy ³² and multispectral immunofluorescence imaging ¹⁷ rely on cell morphology and minimizing spectral overlap of traditional immunofluorescence microscopy ³³ to characterize lesion cells. These methods are less scalable than the sequential imaging method we use because they (1) are limited by the recognition of morphology or the imaging spectrum of the microscope, and (2) rely on visual inspection instead of computational algorithms to perform quantitative analyses. Therefore, it is easier to scale up our method, including the ability to analyze more samples and obtain more information from a single tissue section.

One interesting observation we have made contradicts the conventional wisdom about foam cells: a lack of pro‐inflammatory cytokine expression in both L‐FC and SMC‐FC. This finding is the first piece of evidence in human atherosclerotic lesions that foam cells do not necessarily upregulate IL‐1β and TNF‐α, as suggested by an earlier transcriptomic study that compared foam and nonfoam leukocytes isolated from the same mouse atherosclerotic lesions. ³⁴ However, the expression of other inflammatory markers in human lesions remains to be determined.

The concept of CD68⁺ SMCs in human atherosclerotic lesions has been debated for half a decade since an early study found that 18% of SMA⁺ human lesion cells also express the traditional macrophage marker CD68. ³ Arguments were that SMA⁺CD68⁺ cells could be either macrophages that have engulfed SMCs or macrophages that have trans‐differentiated to express SMA. For the first time, we have estimated that 12% of MYH11⁺ SMCs in human PIT lesions co‐expressed CD68. Compared with previous studies that used SMA, our estimation is more specific to the SMC population (MYH11⁺CD45⁻CD31⁻), excluding leukocytes or myofibroblasts in the SMA⁺ cell population. Several theories have been proposed to explain why SMCs co‐express CD68: the on and off of transcription factors, such as Klf4 and Oct4, ⁴ in response to microenvironmental stimuli and/or the clonal expansion of SMC stem‐like cells followed by exposure to environmental cues. ³⁵ Our data suggest that the expression of CD68 by MYH11⁺ SMCs is associated with the expression of IL‐1β, a downstream product of NLRP3 inflammasome activation. Previous in vitro studies have shown that NLRP3 inflammasome activation induced IL‐1β and CD68 expression, as well as cell death in SMCs. ³⁶ Exposure to IL‐1β activated SMC phenotype switching via the IL‐1β‐STAT3 axis. ³⁷ Moreover, IL‐1β is the therapeutic target in the CANTOS (Canakinumab Anti‐inflammatory Thrombosis Outcome Study) clinical trial, which aimed to reduce the risk of cardiovascular events with canakinumab. ³⁸ Inhibiting phenotype switching and cell death could explain the beneficial effect of canakinumab on SMCs.

Our pilot study showcases the advantages of multiplex imaging in addressing questions that cannot be answered by conventional microscopy and scRNA‐seq. However, our framework has its limitations: (1) The design of the antibody panel is hypothesis‐driven, which is ideal for a proof‐of‐concept study but not for exploring unknown mechanisms; and (2) The analysis is also hypothesis‐driven, which is user‐friendly for any known cell phenotypes but cannot perform unsupervised clustering to find unknown phenotypes (as with QuPath or MAV). ³⁹ , ⁴⁰ Hence, our current framework is more suitable for clinical pathology. Multiplex imaging data provide a snapshot of disease status that can infer associations. To prove any causal relationships between cells and their microenvironment still requires mechanistic studies of atherosclerosis at various stages. For example, the unique location of highly inflamed leukocytes could initiate the formation of early fibroatheroma by secreting matrix metalloproteinases. ⁴¹ Proving the roles of these leukocytes will require the addition of metalloproteinase in the antibody panel and imaging of early fibroatheroma. From the perspective of designing anti‐inflammatory therapies for secondary prevention of adverse cardiovascular events, the perfect case is to target inflammatory factors that have atherogenic effects on all cell phenotypes in advanced stages. Therefore, understanding the molecular and biological profiles of cells at the disease state when patients are treated is critical for predicting therapeutic outcomes.

Conclusions

Multiplex imaging has unique advantages of characterizing cell phenotypes in atherosclerotic lesions and connecting phenotypes with microenvironment, which are difficult for conventional microscopy and scRNA‐seq. With the help of multiplex imaging and biobank human specimens, proof‐of‐concept studies will accelerate the mobilization of knowledge learned from cell culture and animal studies to therapeutic solutions.

Sources of Funding

This study was supported by the Canadian Institutes of Health Research (PJT‐180260 to Y.W.; Canada Graduate Scholarship Master's Award #6556 to M.E.; Canada Graduate Scholarship Master's Award to K.S.), Government of Canada's New Frontiers in Research Fund to Y.W., Wellcome Leap (HOPE Program Award to K.S.), the Heart and Stroke Foundation of Canada (New Investigator Award to Y.W.), the Michael Smith Health Research British Columbia (SCH‐2022‐2648 to Y.W.), and the Fondation Leducq (‘PlaqOmics’ 18CVD02 to N.J.L. and Y.W.). G.A.F. is supported by a Michael Smith Health Research/Providence Research Health Professional Investigator Award.

Disclosures

None.

Supporting information

Data S1