Ubiquitin Ligases Siah1a/2 Control Alveolar Macrophage Functions to Limit Carcinogen-Induced Lung Adenocarcinoma

Recognition of the crucial role played by the immune system in controlling tumor initiation, progression, and response to therapy has prompted the development of therapies that exploit the activity of immune cells and mediators (1, 2). The tumor microenvironment (TME) components participate in extensive crosstalk that regulates immune cell activation and tumor infiltration. Within the TME, various immune cells control tumor growth: Some, such as cytotoxic and regulatory T cells and dendritic cells (DC), are well studied (3, 4), whereas the contributions of fibroblasts, endothelial cells, and myeloid cells, including macrophages, to tumor growth control are less well understood (5, 6).

Approximately 85% of lung cancers are non–small cell lung cancers, of which, lung adenocarcinoma (LUAD) and squamous cell carcinoma are the most common subtypes (7). In response to chemical carcinogens (e.g., urethane), mice develop pulmonary adenocarcinoma marked by KRAS mutations and resembling human LUAD (8). Several studies report the importance of the TME to lung cancer development, including immune system components such as macrophages, and their possible contribution to cancer development (9, 10). Two macrophage populations reside in normal lung tissues: Alveolar macrophages (AM), which are in close contact with type I and II epithelial cells in alveoli, and interstitial macrophages (IM), which reside in the parenchyma between the vascular endothelium and the alveolar epithelium (11, 12). AMs are derived from fetal monocytes that populate alveoli after birth and self-renew outside of the bone marrow (BM; ref. 13). Fetal monocytes accumulate in developing mouse lung at embryonic day 14.5 (E14.5) and differentiate into immature fetal AMs (CD11b^int F4/80⁺ CD11c^int SiglecF⁻; refs. 14, 15), which then further differentiate postnatally into immature (CD11b^low F4/80⁺ CD11c⁺ SiglecF⁻) and mature (CD11b^low F4/80⁺ CD11c⁺ SiglecF⁺) AMs. Postnatal differentiation requires signaling via the growth factor GM-CSF and the ligand-activated transcription factor peroxisome proliferator-activated receptor-γ (PPARγ; ref. 16). Although AMs self-renew under physiological conditions, they can differentiate from BM–derived monocytes following an inflammatory insult or infection (13). AM development and function during acute infection or chronic inflammation are modulated by microenvironmental factors, including oxygen tension, glucose supply, exposure to surfactant-rich fluid, and communication with alveolar epithelial cells (17). Previous studies showed inverse correlations between survival of patients with lung cancer and density of tumor-associated macrophages (TAM), their abundance in stroma, and their polarization to a protumorigenic phenotype (10). AMs obtained by bronchoalveolar lavage of patients with lung cancer exhibit reduced phagocytic capabilities (18), but their role in lung cancer progression remains unclear.

The importance of ubiquitin ligases in cancer development and progression is well established (19, 20). Notably, transcript levels of ubiquitin ligases are largely unaltered in pathologies, including cancer, and thus are not traceable in RNA-based databases (19). Furthermore, ubiquitin ligase proteins are expressed at very low levels, given their propensity to autoubiquitinate and self-degrade (21). The Siah1a/2 ubiquitin ligase family has been studied in cancer primarily due to their importance in controlling hypoxia and unfolded protein response signaling (22, 23). Accordingly, Siah1a/2 were shown to function in development of more aggressive forms of prostate cancer and melanoma, among other tumors (24, 25). Notably, although Siah2 is described as an oncogene in most tumors where it is upregulated at the protein levels (including lung cancer), in select tumor types Siah1 levels decrease, suggesting that it may function as a tumor suppressor in some contexts (26–30). Tumor-intrinsic roles of Siah1a/2 have been shown to be complementary in melanoma and prostate cancers (24, 25).

As previously described (31), Siah1a/2 may regulate the TME. Siah2 deletion in mice decreases melanoma tumor growth via Siah2-dependent regulation of T cells, primarily T-regulatory cell (Treg) proliferation, and tumor infiltration (31). Likewise, treatment of Siah2-knockout (Siah2^KO) mice with anti-PD1 therapy effectively blocks development of therapy-resistant (“cold”) melanoma (31). These observations confirm the importance of Siah2 activity in the TME and extend our earlier analysis of Siah2 tumor-intrinsic function showing relative enrichment of macrophage subpopulations in melanomas from Siah2^KO mice (31).

Here, given the homology of Siah1a and Siah2 proteins (32), we analyzed their function in lung in both physiological and pathological settings using mouse models in which Siah1a/2 genes were specifically deleted in macrophages. Siah1a/2^KO AMs exhibited an immature-like phenotype and expressed a profibrotic gene signature enhanced following administration of the lung carcinogen urethane. Moreover, lung tumors were more abundant and larger in urethane-treated Siah1a/2^KO compared with wild-type (WT) mice. Importantly, tumor tissues from patients with LUAD exhibited increased frequency of immature-like AMs as well as comparable profibrotic gene signatures, outcomes more prominent in patients who were smokers compared with non- or ex- smokers. Collectively, these findings suggest that Siah1a/2 serve as gatekeepers of macrophage maturation and function in control of lung cancer development.

All experimental animal procedures were approved by the Institutional Animal Care and Use Committee of Sanford Burnham Prebys Medical Discovery Institute. Siah1a^f/f and Siah2^f/f mice were generated by Cyagen. Lyz2^Cre and Cd11c^Cre mice were purchased from The Jackson Laboratory. All mice were on a C57BL/6 genetic background. For the urethane-induced lung cancer model, 6 to 8-week-old male mice were injected intraperitoneally with urethane (Sigma; 1 mg/g body weight) once a week for 6 weeks, and lungs were collected at either 7 or 24 weeks after the first urethane injection. Male mice, unless specified otherwise, were used to obtain animal data in this article and are representative of at least three independent cohorts.

C57BL/6 recipient mice were lethally irradiated (1,000 rads) and reconstituted by intravenous injection of a total of 1 × 10⁷ BM cells isolated from femurs and tibias of donor mice, consisting of a 1:1 mixture of either WT CD45.1⁺ cells plus WT CD45.2⁺ cells, or WT CD45.1⁺ cells plus Siah1a/2^f/f::Lyz2^Cre CD45.2⁺ cells. Recipients were treated with antibiotics (trimethoprim 8 mg/mL and sulfamethoxazole 40 mg/mL in drinking water) for 3 weeks after injection. Reconstitution was confirmed 6–8 weeks after BM transfer by FACS analysis of peripheral blood samples.

As previously reported (31), tissues were excised, minced, and digested with 1 mg/mL collagenase D (Roche) and 100 μg/mL DNase I (Sigma) at 37°C for 45 minutes. Digests were then passed through a 70-μm cell strainer to generate a single-cell suspension. Cells were washed twice with PBS containing 2 mmol/L EDTA and stained for flow cytometry.

Tissue-derived single-cell suspensions were washed twice with FACS staining buffer, fixed for 15 minutes with 1% formaldehyde in PBS, washed twice in PBS, resuspended in FACS staining buffer, stained with appropriate antibodies, fixed with 1% formaldehyde, and analyzed by flow cytometry. For every flow cytometric antibody used, we used Fluorescent Minus One, to distinguish positive from negative cells. The following antibodies were used and purchased from BioLegend: CD45.2 (clone 104), CD8a (53–6.7), CD4 (GK1.5), CD45.1 (A20), CD11c (N418), CD11b (M1/70), MHC class I (AF6–88.5), CD80 (16–10A1), GR1 (RB6–8C5), CD206 (C068C2), F4/80 (BM8), SiglecF (S17007L), Ly6C (HK1.4), Ly6G (1A8), NK1.1 (PK136), B220 (RA3–6B2), and CD44 (IM7). All data were collected on an LSRFortessa cytometer (BD Biosciences) and analyzed using FlowJo Software (TreeStar).

As previously reported (31), tissues (lungs) were collected and fixed in 4% formalin (overnight at 4°C), washed with PBS, and paraffin-embedded before they were cut into 5-μm-thick sections and subjected to hematoxylin and eosin staining. For immunofluorescence, sections were deparaffinized, rehydrated and washed in PBS. Antigen retrieval was performed in a pressure cooker (decloaking chamber, Biocare Medical) in citrate buffer (pH 6.0). Antibodies against CD206 (B&D Systems, AF2535), CD11c (Cell Signaling Technology, 97858) Ki67 (AbCam, Ab16667), vimentin (Cell Signaling Technology, 5741), Galectin3 (Santa Cruz Biotechnology, sc-23938), and TTF1 (Novus, NBP2–32999) were used for immunostaining by incubating sections overnight at 4°C with antibodies (diluted in Dako antibody diluent). Alexa Fluor 488- or Alexa Fluor 594–conjugated secondary antibodies were added for 2 hours at room temperature (Molecular Probes), and nuclei were counterstained using SlowFade Gold Antifade reagent (Vector) using 4′,6-diamidino-2-phenylindole (DAPI, Vector). Image data were obtained using an Olympus TH4–100 microscope and Slidebook 4.1 software. For quantification, Ki67-, Lag3- and vimentin-positive cells were counted in five random (×40, magnifications) fields per lung. Staining was scored on the basis of intensity scale in which + represents low/medium staining and ++ high intensity.

To stain AMs grown in culture, cells were grown on coverslips and fixed in fixation buffer (4% paraformaldehyde/2% sucrose/PBS) for 20 minutes at room temperature as previously reported (31). Coverslips were then rinsed twice in PBS and permeabilized in permeabilization buffer (0.4% Triton-X and 1% BSA in PBS) for 20 minutes. Primary antibodies β-catenin (Santa Cruz Biotechnology) and Stat3 (Cell Signaling Technology) were applied at 1:250 dilution in staining buffer (0.1% Triton-X and 0.1% BSA in PBS) overnight at 4°C in a humid chamber. Coverslips were washed 5 times (5 minutes each) in wash buffer (0.2% Triton-X and 0.2% BSA in PBS). Secondary antibodies (AlexaFluor secondary 488 or 594, Invitrogen) were applied at 1:250 dilution in staining buffer for 2 to 3 hours at room temperature in a humid chamber in the dark. Before mounting with Vectorshield with DAPI (Vector Laboratories), coverslips were washed two more times in wash buffer. Immunofluorescent analysis was conducted on an Olympus TH4–100 fluorescent microscope using Slidebook V.4.1 digital microscopy.

Fixed and embedded lung tissues were sliced into 5-μm sections to obtain five serial sections per lung. Sections were stained with H&E and the number and size of nodules were determined using a histology microscope. Tumor size (area of nodules) was calculated using the measurement tool provided with cellSens v3.2 imaging software (Olympus America).

Lung tissue was digested as described above to obtain a single-cell suspension. SiglecF⁺ cells were isolated using anti-SiglecF microbeads (Miltenyi Biotec) and plated in nontreated 6-well plates in RPMI-1640 media containing 1x GlutaMAX, 1x pyruvate, 1x penicillin/streptomycin, 10% FBS, and 20 ng/mL recombinant GM-CSF (BioLegend), as previously described (31). Culture medium was replaced after 16 hours of incubation at 37°C and again every 2 days in culture. Cells were collected on days 1 or 6 after plating. Where indicated, AMs were treated overnight with 5 μmol/L of the Wnt/β-catenin inhibitor XAV939 (Sigma) or 1 μmol/L of the Stat3 inhibitor Napabucasin (Thermo Fisher Scientific) before cells were harvested.

As previously reported (31), cells were washed once with PBS at room temperature and resuspended in RIPA buffer (PBS containing 1% NP-40, 1% sodium deoxycholate, 1% SDS, 1 mmol/L EDTA, and phosphatase and protease inhibitors). Tissue samples were homogenized directly in RIPA buffer. Lysates were centrifuged and supernatants were removed and subjected to SDS-PAGE. Proteins were transferred to nitrocellulose membranes (Osmonics), and membranes were blocked with 5% milk in PBS incubated with appropriate primary antibodies, and then incubated with Alexa Fluor–conjugated or horseradish peroxidase–conjugated secondary antibodies. Blots were imaged using an Odyssey detection system (Amersham Bioscience) or a ChemiDoc MP imaging system (Bio-Rad).

RNA preparation and RT-PCR analyses were performed as previously reported (31). Briefly, total RNA was extracted from tissues or cells using TRizol (Ambion) and treated with DNase I. cDNA was synthesized using oligo-dT and random hexamer primers, according to the SYBR Green qPCR protocol (Life Technologies). Total RNA was reverse transcribed using High-Capacity Reverse Transcriptase kits (Invitrogen), according to the manufacturer's protocol. RNA purity and concentration were checked and quantified by reading absorbance at 260 and 280 nm in a NanoDrop spectrophotometer (Thermo Fisher Scientific). qRT-PCR analysis was performed using SYBR Green RT-PCR kits (Invitrogen) on a Bio-Rad CFX Connect Real-Time System or a Roche LightCycler. GAPDH or 18S RNAs served as internal controls. PCR primers were designed using Primer3 and their specificities were checked using BLAST. PCR products were limited to 100–200 bp. Primer sequences are listed in Supplementary Table S4.

Lung tissues were digested as described above and previously reported (31). F4/80⁺ CD11c⁺ SiglecF⁺ cells were purified by flow cytometry and sorted cells were then processed for gene expression analysis. PolyA RNA was isolated using the NEBNext Poly(A) mRNA Magnetic Isolation Module, and bar-coded libraries were constructed using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB). Libraries were pooled and sequenced from a single end (1×75) using an Illumina NextSeq 500 system with the High Output V2 kit (Illumina). RNA-seq samples were sequenced at a depth of 11–16 million reads.

Illumina Truseq adapters and polyA/polyT sequences were trimmed from raw reads using Cutadapt v2.3 (33), and trimmed reads were aligned to mouse genome version mm10 and Ensembl gene annotations v84 using STAR version 2.7.0d_0221 (34), adopting alignment parameters from the ENCODE long RNA-seq pipeline (https://github.com/ENCODE-DCC/long-rna-seq-pipeline). RSEM v1.3.1 was used to obtain gene level estimated counts and transcripts per million (35). FastQC v0.11.5 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and MultiQC v1.8 (36) were used to assess quality of trimmed raw reads and alignment to the genome and transcriptome. To remove poorly expressed genes from downstream analysis, only genes with RSEM estimated counts ≥5 times the total number of samples were retained for differential expression analysis. Differential expression comparisons were performed using the Wald test implemented in DESeq2 v1.22.2 (37). Genes with a Benjamini–Hochberg-corrected P value of < 0.05 and fold-change of ≥1.5 or ≤−1.5 were defined as differentially expressed. Pathway analysis was performed using Ingenuity Pathway Analysis (IPA) software (Qiagen).

Mice were sacrificed 7 weeks after the first urethane treatment. Lungs were digested as described above and single-cell suspensions were washed with 4% FBS in PBS and incubated 20 minutes at 5 × 10⁷ cells/mL on ice with 500 ng/mL Fc block (2.4G2, BD Pharmingen). Cells were then incubated 1 hour on ice with AF700-conjugated anti-CD45.2 mAb (104, BioLegend). For single-cell RNA (scRNA)-seq libraries, live (DAPI-negative) CD45⁺ cells were sorted using a flow cytometer and resuspended in RPMI-1640 medium for counting. Libraries were prepared using Single-Cell 3′ Reagent Kits v3.1, Chromium Single-Cell 3′ Library & Gel Bead Kit v3.1 (10X Genomics, PN-1000196 and PN-1000129) according to the Single-Cell 3′ Reagent Kits v3 (PN-1000269) User Guide (Manual CG000315 Rev C). Libraries were sequenced on a NovaSeq S4 system, using half-full lanes for 2 samples, with a mean read depth of 87,000–128,000 reads per cell.

The 10X Genomics Cell Ranger pipeline v6.1.2 was used to process scRNA-seq datasets. Single-cell gene counts for each sample were generated using cellranger count and mouse genome version mm10 annotated with mouse GENCODE version M28 (38). We obtained a total of 8,522 Siah WT cells (128,000 mean reads per cell, median of 2,674 genes detected per cell) and 6,479 Siah2^−/− cells (103,000 mean reads per cell, median of 2,502 genes detected per cell). Gene count matrices for Siah WT and Siah2^−/− cells were aggregated and normalized for effective sequencing depth using cellranger aggr applying the “–normalize=mapped” parameter.

Aggregated gene count matrices for Siah WT and Siah2^−/− scRNA-seq samples were processed using Seurat v4.0.5 (39) and R v4.0.2. The 10X Genomics cellranger aggr matrix was converted to Seurat object by retaining genes expressed in ≥10 cells and cells expressing ≥200 genes. Only cells expressing <10% mitochondrial genes and <5,000 total genes were retained to remove dead/low quality cells and potential doublets/multiplets, respectively. Integration of Siah WT and Siah2^−/− samples was performed using the sctransform normalization method (https://satijalab.org/seurat/articles/integration_introduction.html#performing-integration-on-datasets-normalized-with-sctransform-1). A total of 3,000 differentially expressed genes was used in the SelectIntegrationFeatures step. The two datasets were integrated using FindIntegrationAnchors(normalization.method) and IntegrateData(normalization.method). Principal component analysis (PCA) components were computed using RunPCA. Cell clusters were computed using RunUMAP(dims = 1:30), FindNeighbors, and FindClusters(resolution = 0.5), resulting in 17 cell clusters. To visualize gene expression and differential expression analysis, the default assay was set to “RNA” and gene counts normalized using NormalizaData. Cluster markers were found using FindAllMarkers. Differential expression analysis comparisons were performed using FindMarkers(test.use = “MAST”). Plots were prepared using Seurat and ggplot2 (39, 40). Pathway analyses were performed using IPA software. RNA-seq and scRNA-seq plots were prepared using ggplot2, ComplexHeatmap and Seurat (39–41). Reclustering of myeloid clusters (clusters 2, 3, 5) was performed by selecting cells from these clusters and repeating the above-described method. For reclustering, the 638 most variably expressed genes, rather than the 3,000 genes used for integration described above, were selected by retaining genes with a normalized expression variance of >0.4.

Raw count (UMI), cell annotation, and sample information were obtained from the Gene Expression Omnibus record (GSE131907; ref. 42). Raw single-cell data were converted to Seurat object using CreateSeuratObject. We kept the “Alveolar Mac,” “mo-Mac,” and “Monocyte” cell types for normal (nLung, 15,338 cells) and tumor (tLung, 6,641 cells) patient samples, which corresponded to 22 lung samples from 12 unique patients. Raw UMI counts were normalized using NormalizeData, and the 571 most variably expressed genes were selected by retaining genes with a normalized expression variance >0.4. Data were scaled using ScaleData and linear dimension reduction was performed using RunPCA. Harmony (43) was used for integration of patient data using RunHarmony. Cell clusters were computed using RunUMAP(reduction = “harmony,” dims = 1:30), FindNeighbors(reduction = “harmony,” dims = 1:30), and FindClusters(resolution = 0.5), resulting in 10-cell clusters. Cluster markers were identified using FindAllMarkers. Differential expression analysis was performed using FindMarkers(test.use = “MAST”). The average expression of 169 upregulated genes in the macrophage profibrotic gene signature (Supplementary Table S3) was computed using AddModuleScore.

RNA-seq and clinical data from The Cancer Genome Atlas (TCGA)-LUAD Pan-Cancer 2018 (44) were downloaded from cBioPortal. Survival analysis was performed in R v4.2.1 using a survival algorithm (45). The survival analysis was enabled using survminer, an R. R package version 3.3–1, https://CRAN.R-project.org/package=survival.; ref. 46) and survival Curves were drawn using “ggplot2.” R package version 0.4.9 (http://www.sthda.com/english/wiki/survminer-r-package-survival-data-analysis-and-visualization), and maxstat (Maximally Selected Rank Statistics) package version 0.7–25 (https://cran.r-project.org/web/packages/maxstat/index.html; ref. 47). Categorization of TCGA-LUAD samples as “high” and “low” gene expressors in TCGA dataset was determined using surv_cutpoint and surv_categorize functions from the survminer package. Survival analyses using gene signatures (multiple genes) were performed by computing the mean expression of the gene signature. Hazard ratios in TCGA-LUAD for each gene were computed in R with the coxph function using the same “high” and “low” gene expression categorizations described above. CODEFACS (COnfident DEconvolution for All Cell Subsets) was developed (48) as a tool for deconvoluting cell type–specific gene expression for each sample from bulk expression data given either precomputed estimates of cell abundance or a cell type–specific signature.

At least 3 samples were used in each experimental group. All experiments were performed at least three times to establish statistical power and reproducibility. Prism v7 software (GraphPad) was used for statistical analysis. Differences between two groups were assessed using a two-tailed unpaired t test. Survival analysis was performed using the Kaplan–Meier method, and survival and hazard ratio P values were computed using the log-rank test. Differences in cell-type compositions between TCGA-LUAD patients were assessed using the Wilcoxon rank sum test.

The data analyzed in this study were obtained from Gene Expression Omnibus (GEO) at GSE131907. Raw and processed RNA sequencing data have been deposited at Gene Expression Ominibus GEO under accession GSE215993 (GSE215920 bulk RNA-seq, GSE215992 scRNA-seq). Raw data generated in this study were generated by performing IHC, Western blots, and qPCR analyses, all are available upon request from the corresponding author.

To establish mice with conditional loss of both Siah1a and Siah2 (Siah1a/2) in myeloid cells, we generated Siah1a^f/f::Siah2^f/f mice and crossed them with myeloid-specific Cre (Lyz2^Cre) mice to ablate Siah1a/2 (Siah1a/2^f/f::Lyz2^Cre; hereafter referred to as Siah1a/2^L-KO). Total macrophage (CD45⁺ CD11b⁺ F4/80⁺) frequency in spleen, heart, kidney, and lungs of Siah1a/2^L-KO and WT mice was similar (Fig. 1A; Supplementary Fig. S1A). We next used markers of macrophage subtypes to further characterize lung macrophages into AMs (CD11b^low F4/80⁺), IMs (CD11b⁺ F4/80⁺), and additional subpopulations by flow cytometry (FACS). Frequency of CD45⁺ cells (Supplementary Fig. S1B) and total IMs (Fig. 1A) was comparable in lungs of WT and Siah1a/2^L-KO mice. In striking contrast, frequencies of both total (CD11b^low F4/80⁺; Fig. 1B and F) and mature (CD11b^low F4/80⁺ CD11c⁺ SiglecF⁺; Fig. 1C, D, and F) AMs were significantly reduced in lungs of Siah1a/2^L-KO relative to WT mice, and conversely, frequency of immature AMs (CD11b^low F4/80⁺ CD11c⁺ SiglecF^low) significantly increased in Siah1a/2^L-KO lungs (Fig. 1E and F). These findings suggest overall that AMs require Siah1a/2 for differentiation and maturation. Of note, most CD11b^low F4/80⁺ macrophages in both WT and Siah1a/2^L-KO lungs were positive for the mannose receptor CD206 (Supplementary Fig. S1C).

In addition to frequency of macrophage subsets, we also examined absolute expression levels of markers by FACS (reported as mean fluorescence intensity, MFI). Notably, expression levels of CD11c and SiglecF were significantly lower, whereas expression of CD11b was significantly higher on mature AMs from Siah1a/2^L-KO mice (Fig. 1G), confirming that Siah1a/2-deficient AMs exhibit a more immature-like phenotype relative to WT mice. As F4/80, MHCI, CD80, and CD206 expressions were comparable on mature AMs from both genotypes (Supplementary Fig. S1D), differences in CD11c and SiglecF levels likely reflect defects in macrophage differentiation induced by Siah1a/2 loss.

Macrophages can present significant differences between genders (49). To determine whether the phenotypes we observed were gender-dependent, we performed FACS analysis of AMs prepared from lungs of Siah1a/2^L-KO or WT male and female mice. Siah1a/2^L-KO mice of both genders exhibited a reduced frequency of total AMs and mature AMs relative to WT mice (Supplementary Fig. S1E). Moreover, the frequency of immature AMs was higher in lungs of Siah1a/2^L-KO relative to WT mice, whereas SiglecF expression in AMs from Siah1a/2^L-KO mouse lung was significantly lower relative to WT, in both male and female mice (Supplementary Fig. S1E and S1F). These observations suggest that the AM phenotype observed in macrophages upon Siah1a/2 ablation occurs independent of gender.

Because a subpopulation of lung IMs reportedly expresses CD11c (50), which is downregulated on mature AMs following Siah1a/2 deletion, we asked whether either abundance of CD11c⁺ IMs or CD11c expression levels on IMs was altered by Siah1a/2 deletion. We observed a small but significant decrease (from 43% to 33%) in CD11c⁺ IM frequency in lungs of Siah1a/2^L-KO relative to WT mice, concomitant with a significant increase in CD11c⁻ IMs (Fig. 1H), although levels of CD11c protein were comparable between genotypes (Supplementary Fig. S1G). We also observed a significant decrease in NK1.1⁺ cell frequency, but not that of other immune cell populations (CD4⁺ T cells, CD8⁺ T cells, B cells or monocytes) in Siah1a/2^L-KO relative to WT lungs (Supplementary Fig. S1H). Overall, these data suggest that Siah1a and/or Siah2 are required for AM differentiation to a fully mature state, and that both had limited effect on CD11c⁺ IM infiltration to the lung.

To assess individual contributions of Siah1a and Siah2 to AM differentiation, we generated mice with macrophage-specific deletion of Siah1a (Siah1a^L-KO) or Siah2 (Siah2^L-KO) alone and compared mature AM populations in lung with those seen in either WT or double-mutant Siah1a/2^L-KO mice. Frequency of mature AMs was comparable in WT, Siah1a^L-KO, and Siah2^L-KO lungs (Fig. 1I and J). We detected a small but significant decrease in CD11c and SiglecF MFI on Siah1a^L-KO mature AMs and a small but significant increase in SiglecF MFI on Siah2^L-KO compared with WT AMs (Fig. 1I and J). Frequency of total lung IMs (CD11b⁺ F4/80⁺) was not altered by Siah1a KO, whereas Siah2 KO resulted in a small but significant decrease. Conversely, respective frequencies of CD11c⁺ and CD11c⁻ IMs significantly increased and decreased following deletion of Siah1a, but not Siah2 (Supplementary Fig. S1I). Taken together, these data indicate that both Siah1a and Siah2 are necessary for AM maturation.

Because Lyz2 Cre is expressed embryonically (13, 16) and Lyz2 is also expressed in neutrophils, we next asked whether Siah1a/2 were required for AM differentiation and maturation during embryonic development and whether Siah1a/2 function in AMs was cell-intrinsic. To this end, we performed comparable analyses of lungs of mice in which Siah1a and/or Siah2 ablation is driven by Cd11c^Cre (referred to as ^C-KO mice), which is most highly expressed in AMs postnatally and is not expressed by neutrophils (13, 16). As seen in lungs of Lyz2^Cre mice, frequencies of total lung CD45⁺ cells were comparable in WT, Siah1a^C-KO, Siah2^C-KO, and Siah1a/2^C-KO mice (Supplementary Fig. S1J). Moreover, similar to Siah1a/2^L-KO phenotypes, frequency of mature AMs and CD11c⁺ IMs decreased and that of CD11c⁻ IMs increased in Siah1a/2^C-KO relative to WT mice (Fig. 1K and M). AMs from Siah1a/2^C-KO mice expressed lower levels of CD11c and SiglecF compared with cells from WT mice (Fig. 1L), as seen in analysis using the Lyz2 driver. Of interest, single-KO of either Siah1a or Siah2 did not alter frequency of mature AMs, CD11c⁺ IMs, or CD11c⁻ IMs in lungs as compared with WT mice (Fig. 1K and M). Within the mature AM population, CD11c expression levels were comparable in WT, Siah1a^C-KO, and Siah2^C-KO cells, whereas only Siah1a deletion promoted a small but significant decrease in SiglecF expression (Fig. 1L). Consistent with observations in Siah1a/2^L-KOmice, single- and double-Siah1a/2^C-KO. had no effect on frequency of DCs or other immune cell populations in lung (Supplementary Fig. S1K), except for NK1.1⁺ cells, whose frequency markedly decreased in Siah1a/2^C-KO relative to WT mice (Supplementary Fig. S1L). These observations suggest that Siah1a and Siah2 expressions in the myeloid population underlie NK1.1⁺ cell recruitment to lungs. Collectively, data derived from Siah1a/Siah2^C-KO mouse models confirm findings obtained seen using the Lyz2 driver and suggest a cell-intrinsic function for Siah1a/2 in AMs maturation and differentiation.

To assess a potential Siah1a/2 function in the transition from fetal to mature AMs, we assessed frequency of immature AMs in E17.5 WT and Siah1a/2-deficient mouse embryos. We observed that the frequency of immature fetal AMs (F4/80^int CD11c^int) was comparable in embryos of both genotypes (Fig. 2A). Although CD11b or F4/80 MFI was similar in both genotypes, the CD11c MFI markedly decreased in Siah1a/2^L-KO versus WT AMs at E17.5, suggesting a delay in differentiation (Fig. 2B). At birth (postnatal day 0, P0), frequency of total AMs (CD11b^low F4/80⁺) was comparable in lungs of Siah1a/2^C-KO and WT mice (Fig. 2C), whereas the frequency of mature AMs decreased (Fig. 2C and D). Siah1a/2 deletion also significantly decreased the MFI of CD11c, SiglecF, and, to a lesser extent, F4/80 on mature AMs (Fig. 2D and E). Although lung IM frequency was comparable between genotypes, frequencies of CD11c⁻ and CD11c⁺ IMs increased and decreased at E17.5, respectively, following macrophage-specific Siah1a/2 deletion (Fig. 2F). These data strongly suggest that Siah1a/Siah2 deletion suppresses AM maturation but not development of immature AMs in fetal lung.

As noted, AMs self-renew and proliferate under normal physiological conditions but can be replenished from BM-derived monocytes under inflammatory conditions or after BM transplant (50). To determine whether Siah1a/2 are required for AM reconstitution from the BM, we established a BM chimeric mouse model in which BM cells from CD45.2⁺ WT or CD45.2⁺Siah1a/2^L-KO mice were coinjected with an equal number of CD45.1⁺ BM cells into irradiated C57BL/6J recipient mice, which were then assessed for AM reconstitution in lung by FACS 8 weeks later. The frequency of CD45.2⁺ mature AMs was approximately 70% in mice reconstituted with CD45.2⁺ WT BM but only 50% in mice reconstituted with Siah1a/2^L-KO BM (Fig. 3A and B). Also, levels of CD11c and SiglecF expression were markedly lower, whereas F4/80 expression was unchanged, on CD45.2⁺ mature AMs from mice reconstituted with Siah1a/2^L-KO relative to WT BM (Fig. 3C), suggesting overall that AM repopulation and maturation are impaired by Siah1a/2 loss. Frequency of CD11c⁺ IMs was also lower in mice reconstituted with Siah1a/2^L-KO (∼30%) compared with WT (∼37%) BM (Fig. 3D), whereas reconstitution of CD11c⁻ IMs and other immune cells was unaffected (Fig. 3E).

Results reported above suggest that Siah1a/2 loss primarily affects regeneration of AMs and CD11c⁺ IMs from BM. To confirm this, we repeated the mixed BM experiments using Cd11c^Cre mice. As seen in Lyz2^Cre mice, approximately 70% and 30% of mature AMs were reconstituted by transfer of WT and Siah1a/2^C-KO BM, respectively (Supplementary Fig. S2A). Similarly, CD11c expression was significantly lower on AMs derived from Siah1a/2^C-KO relative to WT BM (Supplementary Fig. S2B), but we observed no differences in frequencies of CD11c⁺ or CD11c⁻ IMs (Supplementary Fig. S2C). These data confirm the importance of Siah1a and Siah2 expressions in maturation of AMs from BM-derived monocytes.

Given that AM maturation requires Siah1a/2, we next asked what genes/pathways underlie this activity using RNA-seq analysis of sorted AMs (CD11c⁺ F4/80⁺ SiglecF⁺) from lungs of WT, Siah1a/2^L-KO, and Siah1a/2^C-KO mice. PCA showed similar patterns of gene expression in cells from both KO genotypes and a clear segregation between WT and Siah1a/2^KO cells along the main component PC1 (Fig. 4A). Analysis of genes differentially expressed (DEG) in Siah1a/2^L-KO and Siah1a/2^C-KO relative to WT AMs revealed similar patterns of gene expression in AMs from both KO genotypes (Supplementary Fig. S2D). In total, 565 upregulated and 363 downregulated DEGs were identified in AMs from Siah1a/2^L-KO relative to WT mice (Fig. 4B; Supplementary Fig. S2D). Given that AMs from Siah1a/2^L-KO mice exhibit predominantly immature phenotypes and that GM-CSF/PPARγ signaling is implicated in AM differentiation and maturation, we searched for changes in these pathways among DEGs. Although we observed a small but significant increase in Csf2rb and Tgfbr2 mRNA levels in Siah1a/2^L-KO relative to WT AMs, expression of all other GM-CSF/PPARγ pathway components were comparable in both genotypes (Supplementary Fig. S2E). Likewise, immunofluorescence staining for PPARγ in primary lung AMs cultured 1 day in GM-CSF showed a similar degree of PPARγ nuclear localization in cells of both genotypes (Supplementary Fig. S2F), suggesting that Siah1a/2 loss likely does not alter GM-CSF/PPARγ signaling during AM maturation.

To identify pathways dysregulated by Siah1a/2 loss, we performed IPA of DEGs identified in AMs. Among signaling pathways most activated by Siah1a/2 loss were STAT3, TGFβ, the acute phase response, and actin cytoskeleton signaling (Fig. 4C). Accordingly, the gene signature of Siah1a/2^L-KO AMs was markedly proinflammatory (Fig. 4D). Of note, most components of the inflammatory signature seen in AMs (such as Serpine1, Arg1, Chil3 and S100A8; Fig. 4D, highlighted in red; Supplementary Table S1) have been observed in bleomycin-treated mice, which exhibit lung fibrosis (51) associated with increased Stat3 and β-catenin activity in AMs (52). Indeed, AMs derived from lungs of patients with pulmonary fibrosis exhibit a Stat3 activation signature (53). Interestingly, conditional β-catenin ablation in macrophages alleviates bleomycin-induced fibrosis in mice (54), whereas β-catenin activation is associated with increased inflammation (55) and GADD45G expression, the latter implicated in slowing cell proliferation (56). We also identified DEGs functioning in proliferation in RNA-seq analysis of our genetic Siah1a/2^L-KO mouse derived lung samples (Fig. 4E), consistent with the reduction in mature AM frequency seen in Siah1a/2^L-KO and Siah1a/2^C-KO relative to WT mice (Fig. 1C and K). Finally, we confirmed downregulation of Siah1 and Siah2 mRNA levels (Supplementary Fig. S2G) and altered expression of genes involved in proliferation in primary AMs isolated from Siah1a/2^L-KO relative to WT mice (Fig. 4G; Supplementary Fig. S2H).

RNA-seq analysis also revealed increased expression of β-catenin and Stat3 targets within the inflammatory signature of AMs from Siah1a/2^L-KO versus WT mice (Fig. 4D). Indeed, Siah1a/2-deficient AMs exhibited increased levels of β-catenin transcripts (Fig. 4D) and increased nuclear localization of β-catenin protein, the latter indicative of activation (Fig. 4F). Analysis of cultured primary Siah1a/2^L-KO AMs confirmed markedly increased expression of the β-catenin targets Serpine1, Gadd45 g, and Axin2, an increase blocked by treatment with the β-catenin inhibitor XAV939 (Fig. 4G). Likewise, induction of Stat3 target genes (such as Marco and S100A8) seen in Siah1a/2^L-KO AMs was antagonized by treatment with the Stat3 inhibitor Napabucasin (Fig. 4H). Treatment with either XAV939 or Napabucasin blocked increases in expression of the inflammatory marker Chil3 (Fig. 4I), suggesting that both Stat3 and β-catenin contribute to inflammatory phenotypes. Increase in metabolic signaling associated with increased β-catenin activation was also seen in Siah1a/2^KO AMs (Supplementary Fig. S2I). Together, these data indicate that phenotypes observed upon Siah1a/2 deletion in AMs can be linked to pro-inflammatory and profibrotic gene signatures, which are associated with elevated β-catenin and Stat3 signaling (Fig. 4D).

Given AM abundance in lung in physiological conditions, we asked whether Siah1a/2 loss in AMs would alter lung cancer development and/or progression. To do so, we injected WT or Siah1a/2^L-KO mice weekly for 6 weeks with urethane, a carcinogen that induces KRAS mutation and promotes lung cancer development (57). Lungs of Siah1a/2^L-KO mice, collected 24 weeks after the first urethane injection, harbored a greater number of tumor nodules over a larger area than did lungs of urethane-treated WT mice (Fig. 5A–C). Immunohistochemical analysis revealed a significant increase in the number of Ki67-positive cells in lung lesions of Siah1a/2^L-KO relative to WT mice, suggestive of increased cancer cell proliferation (Supplementary Fig. S3A). Immunohistochemical analysis of AMs (CD11c⁺ round-shaped cells) confirmed AM infiltration in lesions of both WT and Siah1a/2^L-KO lungs (Supplementary Fig. S3B and S3C). Of interest, AM infiltration was observed both in areas of epithelial hyperplasia (Supplementary Fig. S3B) and in adenomas (Supplementary Fig. S3C): In both cases AMs appeared to be in close contact with epithelial cells, suggesting cell–cell communication via either secreted factors, membrane receptors or channels. The frequency of mature AMs in lungs of tumor-bearing WT mice was lower than frequencies seen in normal WT mice (12% vs. 22% of CD45⁺ lung cells, Fig. 5D vs. Fig. 1C), and even fewer AMs were detected in lungs of tumor-bearing Siah1a/2^L-KO mice (4% of CD45⁺ lung cells, Fig. 5D). In contrast, there were no changes in the number of total lung CD45⁺ cells in lungs of tumor-bearing WT and Siah1a/2^L-KO mice (Supplementary Fig. S4A).

Of note, although immature AMs were virtually undetectable in lungs of normal WT mice (Fig. 1E and F), their abundance increased in WT mice after urethane treatment and even more so in urethane-treated Siah1a/2^L-KO mice (Fig. 5E and F). In the mature AM population of Siah1a/2^L-KO mice, urethane administration significantly decreased expression levels of CD11c, SiglecF and the costimulatory protein CD80 as compared with mature AMs seen in urethane-treated WT mice (Fig. 5G). In contrast, although CD11c⁻ IM frequency slightly increased in tumor-bearing Siah1a/2^L-KO mice, we noted a significant decrease in CD11C⁺ cells relative to those seen in WT mice (Fig. 5H and I). Expression of MHCI, CD206, CD11c, and CD80 was comparable in lung CD11c⁺ IMs from Siah1a/2^L-KO and WT mice (Supplementary Fig. S4B), suggesting that both genotypes exhibit a comparable degree of activation despite the lower frequency of CD11c⁺ IMs seen upon Siah1a/2 depletion. Frequencies of B, CD4⁺ T, and CD8⁺ T cells in lungs of WT and Siah1a/2^L-KO mice were also comparable, but frequencies of CD8⁺ CD44⁺ T cells and activated (B220⁺) NK1.1⁺ cells decreased following Siah1a/2 loss (Supplementary Fig. S4C). Together, these data suggest that the presence of macrophages lacking Siah1a/2 enhances tumor cell proliferation and suppresses T- and NK cell activity, allowing tumors to escape an immune response and favoring their progression.

To identify signaling pathways potentially altered in immune cells by Siah1a/2 deletion in lung macrophages, we performed scRNA-seq of CD45⁺ cells sorted from lungs of WT or Siah1a/2^L-KO mice, 1 week after the final urethane treatment. scRNA-seq analysis of WT (n = 7,884) and Siah1a/2^L-KO (n = 6,035) CD45⁺ cells identified 17 immune cell clusters with distinct gene expression patterns (Fig. 6A and B), including NKs (clusters 1 and 13), neutrophils (cluster 12), B cells (cluster 15), T cells (clusters 0, 4, 6, 7, 8, and 10), myeloid cells (clusters 2, 3, 5, and 11), plasmacytoid DCs (cluster 9), conventional DCs (cluster 16), and basophils (cluster 14). T-cell populations were distributed into terminally exhausted effector (cluster 0), effector (cluster 6), and proliferating effector (cluster 10) CD8⁺ cells, effector (cluster 7) and Treg (cluster 4) CD4⁺ cells, and CD4⁻ and CD8⁻ CD3⁺ cells (cluster 8; Fig. 6A and B).

Siah1a/2 deletion in macrophages markedly changed frequencies of immune cells in lung relative to WT mice, including increases in neutrophils (cluster 12; 2.45% vs. 1.15%) and decreases in CD8⁺ effector/proliferative cells (cluster 10; 2.93% vs. 4.46%; Supplementary Fig. S5A). Within the myeloid population of Csf1r⁺ cells (clusters 2, 3, and 5), cluster 2 was enriched for AM markers (Apoe, Mrc1/CD206, Cd72, and Itgax; Fig. 6B). Cluster 2 was also positive for genes indicative of monocytic origin (Ly6c2, Fcgr1, and H2-a) and may represent a mix of resident AMs (mature and immature-like) and AMs of monocytic origin (Fig. 6B). Analysis of gene expression (Fig. 6B) mapped monocytes to cluster 3 (Ly6c-expressing) and macrophages to cluster 5 (Hp, Ear2, Ace, Chil3, and Fn1).

IPA analysis identified upregulation of genes associated with fibrosis and tumor progression within AMs of Siah1a/2^L-KO macrophage cluster 2 (Fig. 6C). Accordingly, in pathologic states, AMs showing an immature-like state reportedly exhibit a unique inflammatory signature similar to that seen in fibrotic macrophages (51).

Increased expression of glycolytic genes, which is linked with a profibrotic phenotype (58), within cluster 2, was also consistent with increased macrophage activity (Fig. 6D). Additional differences identified by IPA between macrophages of both genotypes included increased mitochondrial respiration (cluster 5; Supplementary Fig. S5B). Macrophages possessing an immature-like phenotype were enriched upon Siah1a/2 ablation in cluster 2, as reflected by increased Itgam (CD11b) expression concomitant with decreased Itgax (CD11c) and Cx3cr1 expressions (Fig. 6E). Similar to AMs from cluster 2, Siah1/2-deleted monocytes (cluster 3) exhibited increased expression of genes associated with fibrosis (Fig. 6F). Consistent with inflammatory and fibrotic phenotypes, Siah1a/2-deficient AMs isolated from lungs of urethane-treated mice exhibited increased nuclear localization of β-catenin and Stat3, which is linked to both inflammation and fibrosis, indicative of activation (Supplementary Fig. S5C and S5D).

Among transcripts upregulated in Siah1a/2^L-KO AM cluster 2 were Plaur, Lgals3, Ccl9, and Cebpb (Fig. 6C), all implicated in immunosuppression. Of interest, Lgals3 reportedly suppresses T- and NK cell immune function (59) and promotes tumor growth (60), consistent with phenotypes observed in our mouse model (Supplementary Fig. S3A and S4C). Correspondingly, immunofluorescence analysis confirmed increased Galectin3 (Lgals3) protein levels in AMs (CD11c⁺ round-shaped cells) derived from Siah1a/2^L-KO relative to WT lung lesions (Supplementary Fig. S6A). Increased protein levels of vimentin, whose expression is upregulated in cluster 2 macrophages (Fig. 6C), were confirmed by immunofluorescence of AMs (CD206⁺ round-shaped cells) from Siah1a/2^L-KO as compared with WT lungs (Supplementary Fig. S6B). Of note, CD206/CD11c colocalization was seen in round-shaped cells from both Siah1a/2^L-KO and WT lungs (Supplementary Fig. S6C), consistent with our observations in FACS analysis (Supplementary Fig. S1C). These data suggest that Siah1a/2^L-KO macrophages release factors, such as Galectin3, conducive to immunosuppression and tumor proliferation.

Our initial analyses of the three myeloid clusters (2, 3, 5; Fig. 6A) did not separate resident from immature AMs; therefore, we reclustered these groups and obtained nine myeloid subclusters expressing genes, albeit at different levels, associated with a monocytic origin (Fig. 7A and B). Three subclusters (1, 2, and 3) exhibited characteristics of monocyte populations, and four (5, 6, 7, and 8) consisted of more differentiated resident macrophages (Fig. 7C), of which, subclusters 5 and 8 consisted of mature AMs. We also identified a unique group of T cells positive for F4/80 and CD11c (cluster 4; Fig. 7C). Finally, subcluster 0 expressed AM markers but at a level lower than that seen in mature AMs (clusters 5 and 8), suggesting that cluster 0 contains immature AM (Fig. 7C).

Siah1a/2^L-KO mice showed a decreased proportion of AMs (subclusters 5 and 8) relative to WT mice (Supplementary Fig. S7A), whereas immature AMs (subcluster 0) showed increased Itgam levels and decreased Itgax and Cxc3r1 levels (Fig. 7D), confirming our FACS and gene expression analyses. Finally, we observed increased expression of genes implicated in glycolysis (Supplementary Fig. S7B), which is reported in profibrotic AMs (58) and coincides with increased expression of genes associated with fibrosis (Fig. 7E), in Siah1a/2-deficient AMs of cluster 0. These data indicate that Siah1a/2 loss in AMs promotes protumorigenic and profibrotic phenotypes. LY6C protein frequency in AMs, immature AMs and IMs was largely comparable in WT and Siah1a/2-ablated cells (Supplementary Fig. S7C). This latter finding excludes the possibility that the immature-like state of Siah1a/2-deleted AMs is attributable to dysregulated recruitment of BM-derived monocytes.

To assess the clinical relevance of our findings, we examined expression of genes significantly upregulated (72 genes) in immature AMs of Siah1a/2^L-KO compared with WT mice (Supplementary Table S2) in two TCGA-LUAD patient datasets. Of the 72-gene signature identified in mice, 62 unique genes were mapped in both human datasets, and patients were then stratified into two groups based on low and high expression of these genes. Survival analysis identified a significant association between high expression of the 62-gene signature and poor prognosis (Fig. 7F), a trend comparable in female and male populations (Supplementary Fig. S7D). Second, prediction of cell abundance based on bulk RNAseq deconvolution (48) of TCGA-LUAD patients identified increased infiltration of myeloid cells (CD14⁺) that exhibited relatively high average expression of the 62-gene signature (Fig. 7G). To determine which of 62 genes is linked with poorer survival in the TCGA-LUAD dataset, we calculated individual hazard ratios for these genes, which identified genes involved in glycolysis, plasmin generation, and extracellular matrix degradation as associated with worse survival (Supplementary Fig. S7E). Notably, the same genes set are expressed in TAMs and are linked to cancer progression (61, 62).

We next asked whether the 62-gene signature identified in the cluster of immature-like AMs in lungs of Siah1a/2^L-KO mice was expressed by macrophages in LUAD tumor tissues. To do so, we performed scRNA-seq data from macrophages and monocytes derived from tumor and nonmalignant lung tissues from 12 patients with metastatic LUAD, followed by data reclustering (42). This analysis identified 10 subclusters (Fig. 8A), of which, five (0, 1, 2, 8, and 9) expressed markers specific to human AMs (PPARγ, FABP4, MARCO, MCEMP1, and CXCL16; refs. 63, 64), and low levels of the monocytic marker CD14 (Fig. 8B). Subcluster 2 expressed the lowest levels of the 5 AM markers and highest levels of CD14 (Fig. 8B), similar to patterns seen in immature-like AMs of monocytic origin derived from our mouse model (Fig. 7C). Notably, immature AMs (cluster 2) and monocyte-macrophage clusters (3 and 6) showed highest expression of CTSB (in clusters 2 and 3) and TREM2 (in clusters 2 and 6; Fig. 8B), proteins associated with cancer metastasis and immunosuppression, respectively (65, 66). Notably, immature AMs (cluster 2) and monocyte–macrophage clusters (3 and 6) were enriched in lung tumor tissues as compared with normal lung tissue from patients with LUAD (Supplementary Fig. S8A and S8B). We also observed increased expression of the monocytic marker CD14 and decreased expression of AM markers in tumor versus normal tissues (Fig. 8C; Supplementary Fig. S8C). Subclusters 2, 3, and 6 from tumor tissue exhibited highest expression of proinflammatory/profibrotic genes (such as CCL2, CCL3, SEPP1, and PLA2G7; Fig. 8C) relative to normal lung tissue, confirming findings in the genetic mouse model. These observations suggest that immature AMs present in LUAD tissue are potentially linked with protumorigenic functions (Fig. 8C; Supplementary Fig. S8C). As confirmation, we compared the 62-gene signature in mouse subcluster 0 with genes expressed in monocytes and macrophages from tumor and normal lung tissue from patients with LUAD and observed the greatest enrichment of this gene-set in immature-like AMs from tumor tissues (subcluster 2; Fig. 8D).

Given that TAMs exhibit an immature-like phenotype and a profibrotic signature, we assessed expression of 169 genes upregulated in AMs obtained from patients with pulmonary fibrosis (Supplementary Table S3; ref. 67). We validated a profibrotic signature in monocytes and macrophages from tumor and normal lung tissue of patients with LUAD. Indeed immature-like AMs in subcluster 2 and in monocyte/macrophage cluster 3 of human tumor tissue exhibited the most pronounced enrichment of a profibrotic signature (Fig. 8E). These observations provide independent support in patients with LUAD for the enhanced profibrotic signature seen in immature-like AMs initially identified in our genetic mouse model.

As smoking is associated with an increased risk of lung cancer and mortality (68), we compared expression of the profibrotic gene signature identified in patients with pulmonary fibrosis to gene expression signatures seen in macrophages from normal lung and tumor tissues of patients with LUAD who were current, ex-, or never-smokers. We observed strikingly increased expression of the profibrotic signature in macrophages (mostly in subclusters 2 and 3) from lung tumor tissues of current and ex-smokers, relative to never smokers (Fig. 8F). No differences among current-, ex- and never smokers were detected in expression of the profibrotic signature in myeloid cells obtained from normal lung tissues (Supplementary Fig. S8D). These data overall confirm the relevance of a urethane-induced lung cancer model in Siah1a/2^KO mice to human lung cancer and highlight the importance of macrophages in conferring a profibrotic signature, one highly pronounced in smokers and ex-smokers (Fig. 8G).

Although the function of macrophages in lung inflammation is well known, mechanisms used by these cells to alter lung homeostasis and promote pathological transformation remain poorly understood. Here, we demonstrate that macrophage-specific deletion of Siah1a/2 ubiquitin ligases promotes accumulation of AMs in an immature-like state and induces expression of profibrotic and proinflammatory gene signatures. The use of independent but complementary Lyz2^Cre- and Cd11c^Cre-driven genetic models confirmed our principal finding that Siah1a and Siah2 play crucial roles in AM differentiation and maturation and in regulating expression of key gene signatures relevant to pulmonary inflammation, fibrosis, and cancer.

The major effects of macrophage-specific Siah1a/2 ablation were restricted to AMs, which are the most abundant population of lung macrophage. Although AM differentiation is thought to be primarily regulated by GM-CSF/PPARγ and TGFβ signaling (13, 16, 69), our data revealed that Siah1a/2 contributes to AM maturation independently of the GM-CSF/PPARγ pathway. Siah1a/2 were shown to modulate AM activity via activation of Stat3 and β-catenin signaling pathways, conferring a protumorigenic phenotype. Notably, Stat3 plays an important role in the host response to injury and is activated in AMs during lung injury (70), as seen in patients with lung fibrosis (53). Although Stat3 was previously reported to bind to Siah proteins (71), the significance of this interaction in the context of inflammation remains unclear. Complementing effects of Stat3 in AMs is β-catenin, a critical factor in cytoskeletal organization and cell migration, which was reported to be a Siah substrate (72). Elevated β-catenin signaling is implicated in differentiation (54) of AMs, and is observed in AMs from lungs treated with bleomycin (53) and in TAMs (55). Likewise, mice with macrophage-specific Siah1a/2 ablation exhibited an immature-like phenotype with proinflammatory and -fibrotic signatures associated with elevated Stat3 and β-catenin activity. Indeed, these changes resemble lung fibrosis in humans, a condition associated with increased frequency of immature AMs (51). Previous studies have pointed to the importance of Siah2 in regulating Hippo and HIF pathways (22, 73), which are implicated in macrophage polarization or protumorigenic activities. We confirmed Siah1a/2 control of AM differentiation and inflammatory signaling in a model of urethane-induced pulmonary adenocarcinoma. Notably, lung tumors in these mice were more abundant and larger in Siah1a/2^L-KO as compared with WT mice, exhibited increased numbers of tumor-associated exhausted T cells, a decreased frequency of activated NK cells, and increased expression of genes implicated in tumor proliferation and immunosuppression.

Tissue-resident AMs self-renew under normal physiological conditions but are replaced by monocyte-derived AMs following infection or injury (12, 13). Immature AMs are reportedly enriched in several lung pathologies, including infection and fibrosis (50, 51), but not in lung cancer. Here, we demonstrate that, during lung cancer development, the pool of tissue-resident AMs is replaced by monocyte-derived AMs that exhibit immature-like phenotypes and are marked by a profibrotic/proinflammatory gene signature, distinct from that seen in resident AMs.

Analyses of myeloid cells from lungs of patients with LUAD were consistent with findings in the mouse model and confirmed a link between the profibrotic signature seen in TAMs and smoking. Furthermore, high expression of a 62-gene signature upregulated in Siah1a/2^KO immature-like AMs was associated with poor survival of patients with LUAD. An independent AI-based deconvolution of scRNA-seq datasets from patients with LUAD showed that the same gene signature is associated with higher infiltration of CD14⁺ myeloid cells and worse survival. These independent analyses highlight the importance of monocyte/macrophage infiltration for lung cancer progression.

Overall, this study has uncovered new roles for Siah1a/2 ubiquitin ligases in the differentiation of AMs and revealed how these factors impact lung inflammation, fibrosis, and cancer. Our findings highlight novel mechanisms underlying macrophage regulation in lung cancer and identify a select macrophage population that could be monitored for treatment stratification or targeted as therapy for LUAD.

Z.A. Ronai reports grants from NCI during the conduct of the study, as well as reports a patent for initial IP submitted pending. Z.A. Ronai is cofounder and nonpaid scientific adviser of Pangea Biomed LTD. E. Ruppin is cofounder (divested) and nonpaid scientific adviser of Pangea Biomed LTD and MedAware. No disclosures were reported by the other authors.

M. Scortegagna: Data curation, formal analysis, supervision, validation, investigation, writing–original draft, project administration, writing–review and editing. Y. Du: Data curation. L.M. Bradley: Writing–review and editing. K. Wang: Data curation, software. A. Molinolo: Pathological analyses. E. Ruppin: Data curation, software. R. Murad: Data curation, software. Z.A. Ronai: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The authors thank members of the Ronai laboratory for discussion. The authors thank SBP-shared resources, supported by P30CA030199 grant, for their help with the analyses (Vivarium, FACS, and Bioinformatics). Support by R35CA197465 (to Z.A. Ronai) is gratefully acknowledged.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.