Fth1-mScarlet Reports Monocyte State during Lipopolysaccharide-induced Lung Inflammation

Monocytes and macrophages play central roles in host defense and are critical modulators of inflammation. During immune responses, these cells integrate pro- and anti-inflammatory signals within the microenvironment to support immunity to pathogens while minimizing tissue damage (1, 2). Monocytes are recruited to sites of inflammation, where they perform antimicrobial functions, produce cytokines, and differentiate into macrophages that amplify or dampen the inflammatory response and phagocytose pathogens and cell debris (3). In the context of immune dysregulation, inflammatory monocytes and monocyte-derived macrophages often promote pathology, and the frequencies of these cells can predict severe outcomes in human pulmonary infection (4). The pathological effects of infiltrating monocytes have been demonstrated in mouse models, with the loss of CCR2-dependent monocyte infiltration leading to reduced morbidity and mortality independent of pathogen burden (5).

Although the mechanisms driving monocyte differentiation are poorly understood, the growing consensus is that monocytes are not precommitted and that their fate is shaped by diverse microbial and host-derived stimuli (6–8). A recent study has defined a monocyte-to-macrophage fate decision occurring within 24 h in an inflamed environment and identified transcriptional factors that regulate monocyte-derived macrophage versus monocyte-derived dendritic cell differentiation (9). The stimuli driving monocyte differentiation into macrophages are complex and dynamic, resulting in macrophages with diverse transcriptomic and functional states. The binary classification system of proinflammatory M1 macrophages and anti-inflammatory M2 macrophages, often modeled in vitro using LPS+IFN-γ or IL-4+IL-13, respectively, is conceptually useful but does not reflect multidimensional macrophage programs present in complex microenvironments. Macrophages expressing a combination of canonical M1 and M2 markers have been described in inflammatory disease (10, 11), and even in acute LPS-stimulated inflammation and resolution, markers commonly used to distinguish macrophage activation states in vitro are largely coexpressed on individual cells, regardless of the stage of inflammation (12).

Quantification and manipulation of monocyte and macrophage function is limited by a lack phenotypic markers that report their differentiation programs and spectrum of functional states. We leveraged recently published single-cell RNA-sequencing (scRNAseq) analysis of monocyte-derived and tissue-resident alveolar macrophages (AMs) sorted from mouse lungs during homeostasis and LPS-induced inflammation, and we identified the gene Fth1, which encodes the ferritin H chain, as predictive of complex transcriptomic states. In this study, we describe an Fth1-mScarlet mouse that reports monocyte-to-macrophage differentiation and inflammatory state, supporting efforts to interrogate and manipulate regulation of these cell fates.

Using previously published AM scRNAseq data from Mould et al. (12), genes with average unique molecular identifiers (UMIs) less than 1 were removed, and counts were normalized by dividing each cell’s UMI by the sum of the cell’s UMI values and then multiplying those values by 10,000. Any zero values were set to 1, the log was taken, and values were z-score normalized by gene. The dimensionality was reduced by first running principal component analysis (PCA) with six components and then running t-distributed stochastic neighbor embedding (tSNE) with two components on the PCA reduced dataset for the final tSNE visualizations. All dimensionality reduction was done with Python’s scikit-learn package. The same analysis was performed using previously published interstitial macrophage scRNAseq data from Moore et al. (13). For interstitial macrophages, 20% of cells were sampled to use for computing PCA and tSNE visualizations.

Least absolute shrinkage and selection operator (Lasso) regression and partial least squares regression (PLSR) were used to identify genes with expression that is predictive of the broader transcriptional landscape using both AM and interstitial macrophage datasets. For Lasso, data were split into approximately a 90-10 training-validation split to tune the regularization parameter, testing values from 0.01 to 1 by intervals of 0.01. A value of 0.08 was selected to be optimal for prediction performance, and that value was used for the subsequent steps. To identify genes that could predict other genes, 1000 bootstrapped runs were performed with 700 randomly selected cells for each run to predict a single gene’s expression (response gene) from all other genes in the data (explanatory genes). The regression weights were added across the predictions and averaged across bootstrapped runs to get a weight (or ranking) per gene. Two components were used for PLSR predictions, and the same bootstrapping and gene prediction procedures as described above for Lasso were used. For PLSR, the weights were from the gene loadings of the first component, and, like the regression coefficients from Lasso, they represent the predictive power of a given gene for all other genes in the dataset. The same was done for interstitial macrophages.

AMs were subsetted using gene markers of recruited macrophages. AMs expressing Cd14, Ly6c1, and/or Apoe at levels above the 85th percentile for each marker were classified as recruited (Rec AMs), whereas cells that did not meet these criteria were classified as nonrecruited (non-Rec AMs). The Welch t test quantified the significance of the difference in Fth1 expression levels between Rec AMs and non-Rec AMs.

The Fth1-mScarlet strain was produced in the Mouse Embryo Services Core in the Department of Immunology of the University of Pittsburgh, using CRISPR/Cas9 technology directly in C57BL/6J zygotes. Briefly, fertilized C57BL/6J embryos (The Jackson Laboratory) produced by natural mating were microinjected in the pronuclei with a mixture of 0.67 µM EnGen Cas9 protein (New England Biolabs, catalog no. M0646T), a Cas9 guide RNA (21 ng/µl = 0.66 µM), and a donor DNA template. The SpyCas9 single-guide RNA targets the sequence 5′-CGGTGGTCATGGTGGCGGCGGGG-3′ (chr19:9982857–9982879, Mouse Dec. 2011, GRCm38/mm 10 assembly), which overlaps the start codon but cleaves in the 5′ untranslated region (UTR). The Cas9 single-guide RNA was produced as described previously (14). For donor DNA, “Alt-R HDR Donor Blocks” was used; this is a modified double-stranded DNA custom synthesized by Integrated DNA Technologies. The injected zygotes were cultured overnight, and the next day, the embryos that developed to the two-cell stage were transferred to the oviducts of pseudopregnant CD1 female surrogates. Potential founders were genotyped by two overlapping PCR products. PCR 1 produces a fragment of 973 bp with forward primer 1 (5′-CCTCACACTCACACAGGCTC-3′) and reverse primer 1 (5′-TTCGTACTGTTCCACCACGG-3′). PCR 2 produces a fragment of 898 bp with forward primer 2 (5′-CACGAGTTCGAGATCGAGGG-3′) and reverse primer 2 (5′-CATTTTCCAGAAGGGCGCTG-3′). Sanger sequencing confirmed the correct sequence of the allele and the proper integration in the endogenous Fth1 locus. The same PCRs were used for routine genotyping. The Fth1-mScarlet allele was maintained in the heterozygous state.

Lungs, spleens, and thymi were cut into 1-mm³-sized pieces and digested in DMEM containing Liberase (10 µg/ml) and DNase I (10 U/ml) in a shaker for 30 min at 37°C. Lungs were then further dissociated using the program “m_lung_02_01” on the gentleMACS Dissociator (Miltenyi Biotec). The resulting tissue homogenates were filtered through a 70-mm filter. Cells were lavaged from the peritoneal cavity with cold PBS containing 1% FBS and 5 mM EDTA. Bone marrow cells were flushed from the femur and tibia and passed through a 70-μm filter. For lungs, spleens, and bone marrow, RBCs were lysed with ACK Lysing Buffer (Lonza). Livers were mashed through 100-μm filters before undergoing erythrocyte lysis using the Mouse Erythrocyte Lysing Kit (R&D Systems, WL2000). Liver cells were then isolated via purification of mononuclear cells using 45% Percoll centrifugation (Cytiva, 17089101). For tissue phenotyping, single-cell suspensions were stained with Abs including CD11c-BUV737 (BD Horizon, HL3), CD11c-BV421 (Brilliant Violet, N418), Ly6C-BV605 or allophycocyanin (BioLegend, HK1.4), SiglecF-BV421 (BioLegend, S17007L), SiglecF-AF647 (BD Horizon, E50-2440), Ly6G-PerCP-cyanine (Cy)5.5 (BioLegend, 1A8), and CD11b-allophycocyanin-Cy7 (BioLegend, M1/70, CX3CR1-BV785 (BioLegend, SA011F11), CCR2-BV650 (BioLegend, SA203G11), F4/80-BUV395 (BD Horizon, T45-2342), CD64-BV421(Brilliant Violet, X54-5/7.1), CD45-BUV395 (BD Horizon, HI30), CD4-BV605 (Brilliant Violet, RM4-5), CD8a-AF647 (Brilliant Violet, 63-6.7), IA/IE-PCP-Cy5.5 (BioLegend, M5/114.15.2), and B220-allophycocyanin-Cy7 (BioLegend, RA3-6B2). Data were collected with an Aurora spectral cytometer (Cytek) or a BD Fortessa (BD Biosciences) and analyzed using FlowJo software (BD). In some experiments, whole-lung single-cell suspensions were stained with CD11b-allophycocyanin-Cy7 (BioLegend, M1/70), Ly6C-allophycocyanin (Invitrogen, HK1.4), and Ly6G-BV421 (BioLegend, 1A8), and GFP⁺CD11b⁺Ly6G⁻Ly6C⁺ cells were sorted directly into TRIzol LS Reagent (Ambion) on the basis of low versus high mScarlet expression. Cells were sorted using a BD FACSAria II (BD Biosciences).

LPS (20 μg in 50 μl; Escherichia coli 0111:B4 from InvivoGen) was instilled directly into the tracheas of Fth1-mScarlet × Lyz2-GFP mice sedated with isoflurane. Three days later, mice were sacrificed using CO₂ euthanasia. Lungs were harvested and processed as described above.

Lungs were inflated using 1 ml of a mixture of OCT and 4% PFA (1:1), and then suture thread was used to tie off the trachea and each lobe. Lungs were embedded in OCT compound (Tissue-Tek) and placed in −80°C before sectioning. Sections were sliced to 10 μm on a cryostat (Leica CM 1950) and stored at −20°C until imaging. The slides were mounted using ProLong Diamond Antifade Mountant with DAPI and imaged using a Keyence BZX-810 widefield microscope equipped with a 0.75 numerical aperture, 20× objective.

Samples were lysed in TRIzol LS (Ambion), and RNA was isolated with the Direct-zol RNA MicroPrep Kit (Zymo Research). RNA concentration was quantified on a Qubit FLEX fluorometer, and libraries were generated with the Illumina Stranded mRNA Library Prep kit according to the manufacturer’s instructions. Briefly, 25 ng of input RNA was used for each sample. Following adapter ligation, 15 cycles of indexing PCR were completed, using IDT for Illumina RNA UD Indexes. Libraries were normalized and pooled to 2 nM by calculating the concentration on the basis of the fragment size (bp) and the concentration (ng/µl) of the libraries. Sequencing was performed on an Illumina NextSeq 2000 using a P3 200 flow cell. The pooled library was loaded at 750 pM, and sequencing was carried out with read lengths of 2 × 58 bp, with a target of 25 million reads per sample. Sequencing data were demultiplexed by the on-board Illumina DRAGEN FASTQ Generation software (version 3.10.12). The Mus musculus reference genome (mm10) was downloaded from the UCSC genome browser (15), and Rsubread (16) was used to align and obtain raw read counts with that reference. DESeq2 (17) was used for read normalization and differential gene expression calculations. Genes that were expressed in at least one sample and were below the 90th percentile of the coefficient of variation across samples were included in further analysis. The p values for differential expression were corrected using the Benjamini-Hochberg method for multiple hypothesis testing, where genes with adjusted p values less than 0.05 were considered differentially expressed. RNAseq heatmaps were clustered using hierarchical clustering using Euclidean distance and average linkage. Data are available via Gene Expression Omnibus GSE278058.

Monocytes were isolated from Fth1mScarlet bone marrow using the EasySep Mouse Monocyte Isolation Kit per the manufacturer’s instructions. Isolated monocytes were plated at 1 × 10⁶ cells/ml in complete DMEM and allowed to rest for 2 h. Cells were stimulated with 100 nM Kdo2-Lipid A (KLA) (Avanti Polar Lipids) or 60 ng/ml of M-CSF (R&D Systems) for 1, 2, or 4 d. Cells were then stained for flow cytometric analysis.

Toward our goal of generating a model system that reports macrophage inflammatory state, we sought to identify genes that can be quantified and used to extrapolate broader transcriptomic signatures associated with inflammatory function. We leveraged scRNAseq of AMs during intratracheal LPS challenge previously published by Mould et al. (12); in this study, CD64⁺F4/80⁺ macrophages were isolated from the bronchoalveolar lavage, sorted, and sequenced at days 0 (homeostasis), 3 (peak neutrophil inflammation), and 6 (resolution of lung inflammation). We used multiple computational approaches to identify candidate genes whose expression is predictive of gene signatures in single cells across a range of inflammatory macrophage states. We ran Lasso (L1-penalized linear regression) using all data apart from a single gene as “explanatory genes,” ranking the explanatory genes according to their ability to predict the expression of the excluded gene (response gene). Lasso weights (regression coefficients) were summed for each response gene prediction and averaged across bootstrapped runs. As an alternative approach to Lasso, which reduces the dimensionality of the explanatory genes before making predictions, we employed a latent factor approach by way of PLSR. We used the same prediction process as with Lasso, except instead of using the regression coefficients as weights, we used the loadings of the first latent factor. The loadings rank the genes by the amount they contribute to latent factor 1, which is the component most predictive of the response gene. We assessed genes most predictive of the overall transcriptomic signatures by comparing the weights of each gene per method (Fig. 1A). For highly weighted gene candidates, we compared the average gene expression over the course of the LPS response (Fig. 1B). Fth1 was the only candidate in the top five weighted genes for both Lasso and PLSR and was highly specific to peak inflammation (Fig. 1A, 1B). We noted that both Fth1, which encodes the ferritin H chain, and Ftl1, which encodes the ferritin L chain, were highly predictive of the transcriptomic state of single cells during LPS-induced lung inflammation. These genes are well appreciated to be associated with a proinflammatory macrophage phenotype (18), and recent studies have demonstrated that FTH1 plays a critical role in macrophage activation of nitric oxide synthase 2 and protects against iron-induced oxidative stress in response to LPS (19). We found that the expression of Fth1 captures much of the transcriptomic space (Fig. 1C, 1D). In contrast to AMs, performing comparable analysis on scRNAseq of interstitial macrophages during intratracheal LPS challenge, using data previously published by Moore et al. (13), shows that Fth1 does not predict interstitial macrophage transcriptomic profiles; Fth1 was not highly weighted by Lasso or PLSR, and Fth1 expression did not associate with specific cell clusters (Supplemental Fig. 1A, 1B). Together, these data supported selection of Fth1 for the generation of a reporter mouse to investigate myeloid cell states during lung inflammation.

To make the Fth1-reporter strain, the bright monomeric red fluorescent protein mScarlet-I (20) was knocked in to the endogenous Fth1 locus to drive the expression of this gene under all the native regulatory elements. Because homozygous FTH1 deficiency is embryonic lethal in mice (21), the mScarlet-I coding sequence was introduced in-frame immediately after the start codon and fused to FTH1 using a T2A peptide (22) (Fig. 2A), which can induce ribosomal skipping during translation (23). Thus, a single amino acid of the T2A peptide remains attached to the N-terminal of FTH1, so that the knocked-in allele still expresses functional FTH1 protein. In addition, these mice were crossed to an existing Lyz2-GFP reporter strain (24); Lyz2-GFP is highly expressed in neutrophils, monocytes, and macrophages. Generating the Fth1-mScarlet × Lyz2-GFP reporter strain allowed us to assess Fth1-mScarlet expression across myeloid compartments.

We assessed Fth1-mScarlet × Lyz2-GFP reporter expression in the lung at steady state and found that high Fth1-mScarlet protein expression was restricted to CD45⁺GFP⁺ cells, although Fth1-mScarlet expression was observed at lower levels in CD45⁻, CD45⁺GFP⁻, and CD45⁺GFP⁺ cellular compartments (Fig. 2B, Supplemental Fig. 2A). We confirmed that high Fth1-mScarlet expression in the lung was limited to GFP⁺ cells using immunofluorescence imaging (Supplemental Fig. 2B). We went on to explore the Lyz2-GFP⁺ population and identified AMs as the only Fth1-mScarlet high cells in the steady state (Fig. 2C). Another lung Lyz2-GFP⁺ myeloid population expresses considerably lower levels of Fth1-mScarlet (Fig. 2B, 2C). Of the GFP⁻ populations, CD11c⁺MHCII⁺ dendritic cells expressed the highest Fth1-mScarlet but only demonstrated levels comparable to the low Fth1-mScarlet expression seen in non-AM Lyz2-GFP⁺ cells (Supplemental Fig. 2C). We also quantified Fth1-mScarlet in additional steady-state tissues (Figs. 2C, 2D; Supplemental Fig. 2D–2F); we noted a substantial population of CD45⁻GFP⁻ cells expressing intermediate Fth1-mScarlet in the bone marrow and a small population of Lyz2-GFP⁺ expressing high Fth1-mScarlet in the thymus. Further gating showed these Lyz2-GFP⁺Fth1-mScarlet high cells to be F4/80⁺ thymic macrophages (Supplemental Fig. 2D, 2F). Peritoneal macrophages expressed intermediate levels of Fth1-mScarlet (Fig. 2C, 2D), whereas Fth1-mScarlet expression was low in F4/80⁺ splenic and liver macrophages (Supplemental Fig. 2D, 2F). Our data suggest that Fth1-mScarlet expression can be used to discriminate myeloid populations in the lung and will have tissue-specific utility in visualizing resident macrophages.

To evaluate the effectiveness of Fth1-mScarlet in reporting macrophage inflammatory state, we administered intratracheal LPS as performed in the scRNAseq study that was used to identify Fth1 as our reporter gene (Fig. 1A–1D). LPS was intratracheally administered to Fth1-mScarlet × Lyz2-GFP mice, and, 3 d later, lungs were analyzed by flow cytometry. Although Fth1-mScarlet expression significantly increased in AMs in response to LPS, the observed difference in reporter expression was only 2-fold (Fig. 3A, 3B). This raised concerns that this small difference in fluorescence above high steady-state expression may limit the utility of this reporter in distinguishing macrophage state. In contrast, we observed a 10-fold increase in the average Fth1-mScarlet fluorescence in Lyz2-GFP⁺CD11b⁺Ly6G⁻ myeloid cells following LPS administration, with substantial variation in Fth1-mScarlet expression across this population 3 d after LPS challenge (Fig. 3A, 3B). We did not observe Fth1-mScarlet induction in Lyz2-GFP⁻ populations in the lung in response to LPS-induced inflammation (Supplemental Fig. 2A). We speculated that Fth1-mScarlet expression may reflect changes in monocyte state.

We sought to better understand monocyte heterogeneity in Fth1-mScarlet expression by investigating the transcriptomic signatures associated with low versus high Fth1-mScarlet expression. Using the same intratracheal LPS administration model, on day 3, we sorted Lyz2-GFP⁺CD11b⁺Ly6C⁺Ly6G⁻ monocytes on the basis of low versus high Fth1-mScarlet expression (Supplemental Fig. 3A), as well as low Fth1-mScarlet monocytes from untreated mice, for RNAseq analysis. Hierarchical clustering of genes that were differentially expressed between any two groups revealed sets of genes associated with LPS treatment, such as cluster 8 (Fig. 4A). However, a larger number of genes cluster on the basis of Fth1-mScarlet expression. For example, clusters 2 and 3 had higher expression in the Fth1-mScarlet low groups, whereas cluster 5 had higher expression in the Fth1-mScarlet high group, regardless of LPS treatment (Fig. 4A). We compared these gene clusters with previously reported gene sets that were associated with monocytes versus AMs during lung inflammation. In brief, Misharin et al. performed bulk RNAseq analysis on FACS-sorted monocytes, interstitial macrophages, monocyte-derived AMs, and tissue-resident AMs at multiple time points during bleomycin-induced lung fibrosis and defined clusters associated with monocyte-to-AM differentiation; Misharin et al. cluster K1 increased from monocyte-to-AM differentiation (most highly expressed in AMs), whereas K2 was highly expressed in monocytes and decreased with differentiation (25). We found that these monocyte and AM-associated gene sets (Misharin et al. K2 and K1) were overrepresented in Fth1-mScarlet low and Fth1-mScarlet high genes, respectively (Fig. 4A). Fth1-mScarlet low gene cluster 2 had statistically significant overlap with monocyte-associated genes, and Fth1-mScarlet high gene cluster 5 had statistically significant overlap with AM-associated genes (Fig. 4B). The opposite pairing of gene set did not show significant overlap (Supplemental Fig. 3B). Direct comparison of genes differentially expressed between the Fth1-mScarlet low and Fth1-mScarlet high again demonstrates the presence of macrophage-associated genes within the genes increased in high Fth1-mScarlet cells, whereas monocyte-associated genes primarily overlap with genes more highly expressed in the Fth1-mScarlet low group. Differentially expressed genes between the LPS-treated and nontreated Fth1-mScarlet low groups were predominantly upregulated in the treated group and represented both the monocyte- and the AM-associated genes (Supplemental Fig. 3C), perhaps representing the initial transitionary state following LPS-induced inflammation. Together these results are consistent with Fth1-mScarlet reporting monocyte differentiation into AMs in response to lung inflammation.

Given that the Lyz2-GFP⁺CD11b⁺Ly6C⁺Ly6G⁻ population that we analyzed by RNAseq includes both monocytes and cells at varying stages of monocyte-to-macrophage differentiation, we further explored these myeloid subpopulations by flow cytometry. Ly6C-hi classical monocytes, present in both untreated and LPS-treated mice, expressed high surface levels of CX3CR1 (Fig. 5A, left panels), in agreement with a report describing high CX3CR1 surface staining in classical monocytes, despite low CX3CR1-EGFP expression in commonly used CX3CR1 reporter mice (26) The Lyz2-GFP⁺CD11b⁺Ly6C⁻ population, present in both steady-state and LPS-treated lung, also expressed CX3CR1 (Fig. 5A), consistent with interstitial macrophage phenotype (27). This population showed only a modest increase in Fth1-mScarlet (Supplemental Fig. 1C), in line with our analysis of scRNAseq data from Moore et al. (13), showing that Fth1 does not predict interstitial macrophage transcriptomic profiles during LPS-induced inflammation.

At day 3 after LPS administration, within the Lyz2-GFP⁺CD11b⁺Ly6C⁺Ly6G⁻ population, we noted a subset of cells expressing CD11c, a marker that is upregulated as recruited monocytes differentiate into AMs (28). These CD11c⁺ cells could be at varying stages of the monocyte-to-macrophage differentiation process and are labeled as “mono-mac” in Fig. 5. Although classical Ly6C⁺CX3CR1⁺CD11c⁻ monocyte expression of Fth1-mScarlet was increased in LPS-treated mice, CD11c⁺ cells had the highest levels of Fth1-mScarlet reporter activity (Fig. 5B, 5C). We note that Fth1-mScarlet is also increased in Lyz2-GFP⁺CD11b⁺Ly6G⁺ neutrophils in LPS-induced inflammation (Fig. 5C); as indicated, Ly6G⁺ neutrophils were excluded during gating and sorting for the analysis of Lyz2-GFP⁺CD11b⁺ cells shown in Figs. 3–5. Finally, within populations of differentiating monocytes in LPS-treated mice, we observed an inverse relationship between Fth1-mScarlet and CCR2 at the single-cell level (Fig. 5D). Thus, our data suggest that Fth1 reporter activity is increased in myeloid populations recruited during lung inflammation and that high Fth1-mScarlet expression reports as monocytes differentiating into macrophages, as evidenced by increased CD11c and decreased CCR2 expression.

In light of our findings, we returned to the AM scRNAseq data from Mould et al. (12), which were used to identify Fth1 as highly predictive of the transcriptomic state of AMs following intratracheal LPS challenge. In these data, we defined recruited AMs (Rec AMs) on the basis of expression of any combination of Cd14, Ly6c1, or Apoe, consistent with analysis by Mould et al. (12) (Fig. 5E). Comparing the Rec AMs with cells not meeting these criteria, or “non-recruited” AMs, we found that Fth1 expression was significantly higher in Rec AMs (Fig. 5F); this is also apparent by comparing the Rec AM distribution with the tSNE showing Fth1 gene expression in Fig. 1D. Together with our other results, increased Fth1 gene expression in Rec AMs that were recently differentiated from monocytes (still expressing Cd14, Ly6c1, or Apoe) is consistent with Fth1 reporting monocyte-to-macrophage differentiation.

Intratracheal LPS may induce monocyte differentiation in response to direct monocyte sensing of LPS, consistent with evidence that TLR ligation directs monocyte-to-macrophage fate (6). Thus, in evaluating the monocyte and macrophage state information reported by Fth1-mScarlet, it is challenging to uncouple monocyte-to-macrophage differentiation from inflammatory function. To assess the ability of TLR4 ligation to regulate Fth1-mScarlet expression in vitro, we stimulated bone marrow monocytes or bone marrow–derived macrophages (BMDMs) with Kdo-2 Lipid A, the stimulatory portion of LPS. A small increase in BMDM expression of Fth1-mScarlet with KLA was noted, with a contrasting small decrease with canonical M2 cytokines IL-4 and IL-13, suggesting that Fth1 is associated with inflammatory state (Fig. 6A). However, the relatively small difference in fluorescence limits the utility of this reporter in distinguishing macrophage state, as seen in AMs (Fig. 3A). Comparable to our in vivo studies, TLR4 ligation induced a robust change in Fth1-mScarlet in bone marrow monocytes, with levels rapidly increased starting at day 1 and continuing to increase over 4 d (Fig. 6B, 6C). In parallel, we cultured bone marrow monocytes with M-CSF to compare TLR4-induced levels with differentiation in a noninflammatory context. In response to M-CSF, we observed a delayed increase in Fth1-mScarlet expression, starting at day 2 and increasing more steeply by day 4 of the culture, reaching only 40% of the Fth1-mScarlet fluorescence intensity compared with TLR4 stimulation (Fig. 6B, 6C). We noted that although Ly6C was downregulated in the M-CSF culture conditions, as expected in differentiating monocytes, Ly6C remained high in Fth1-mScarlet high LPS-stimulated monocytes (Fig. 6B). Thus, Fth1-mScarlet levels are positively associated with both macrophage differentiation and inflammatory state.

Our results suggest that Fth1-mScarlet expression reports monocyte differentiation into inflammatory macrophages during LPS-induced lung inflammation. Although monocyte differentiation is shaped by a variety of stimuli, including microbial products, cytokines, and metabolites (6–8), our data suggest that Fth1-mScarlet will be more robust in the context of proinflammatory stimuli. This is consistent with reports that Fth1 expression is sensitive to inflammatory stimuli (29). Alternatively, even in the same microenvironment, heterogeneity in Fth1 induction may be associated with distinct monocyte and macrophage functions, consistent with scRNAseq analysis showing considerable transcriptomic heterogeneity in monocyte-derived macrophages during lung inflammation (12). Macrophages depend on Fth1, the heavy subchains of ferritin, to sequester iron within cells to modulate iron availability as a host defense mechanism (29). FTH1 is essential for macrophage production of NO, IL-6, and IL-1b in response to LPS (19, 30), and mice with myeloid-specific deletion of FTH1 were impaired in their ability to clear Salmonella typhimurium (31). Given these FTH1-dependent mechanisms, Fth1-mScarlet expression is likely to correlate with proinflammatory and antimicrobial monocyte and macrophage functions. Our data with both AMs and BMDMs suggest that Fth1-mScarlet expression is also associated with inflammatory function in macrophages. However, the uniformly high Fth1-mScarlet expression in these cells prior to stimulation, with the relatively small shift in mScarlet expression, limits our ability to analyze a range of reporter levels. Toward the goal of reporting transcriptomic profiles, we speculate that an Fth1 reporter using a fluorescent protein engineered to have a short half-life would improve the dynamic range of this reporter in macrophages to support quantification of inflammatory states.

In this study, we show that Fth1-mScarlet expression reports differentiation of recruited monocytes into macrophages during LPS-induced lung inflammation. Given the complexity of Fth1-mScarlet regulation, we speculate that the value of this tool in dissecting macrophage differentiation from inflammatory programs will be enhanced when combined with other reporter systems. The potential of an Fth1-mScarlet dual reporter has been briefly highlighted in this work, using Fth1-mScarlet × Lyz2-GFP, which supported identification of resident macrophages in multiple tissues. Existing monocyte reporters lines, such as CX3CR1-GFP (32) or CCR2-CreER-GFP (33), may support interpretation of Fth1-mScarlet dynamics in the context of stimuli inducing both macrophage differentiation and inflammatory functional programs. Efforts toward understanding monocyte and macrophage states may benefit from tissue-specific approaches, such as comparing levels of CD11c-YFP (34) and Fth1-mScarlet, given that monocyte-derived AMs express CD11c (35) and CD11c reflects the monocyte origin of salivary gland macrophages (36). Using dual reporter systems, cells with varying levels of the two fluorescent proteins can be sorted and analyzed by RNAseq to establish corresponding transcriptional signatures. Finally, given that ferritin has emerged as a biomarker for a variety of inflammatory diseases (37–39), this tool will be valuable in efforts to understand pathological mechanisms and optimize treatments for diseases in which ferritin levels have been associated with disease. In summary, generation of the Fth1-mScarlet reporter model is a step toward addressing an important lack of tools to support investigation of monocyte and macrophage differentiation programs and functional states in vivo.

The authors have no financial conflicts of interest.

We acknowledge Chunming Bi and Zhaohui Kou of the Mouse Embryo Services Core, University of Pittsburgh, Department of Immunology, for microinjection of zygotes and production of the Fth1-mScarlet strain. We also thank the United Flow Core for cell sorting, the Health Sciences Sequencing Core at UPMC Children’s Hospital of Pittsburgh for library generation and sequencing, the Center for Research Computing for providing computing resources, and the Division of Laboratory Animal Resources for animal care.