Induction of interferon signaling and allograft inflammatory factor 1 in macrophages in a mouse model of breast cancer metastases

Background: Metastatic breast cancer cells recruit macrophages (metastasis-associated macrophages, or MAMs) to facilitate their seeding, survival and outgrowth. However, a comprehensive understanding of the gene expression program in MAMs and how this program contributes to metastasis remain elusive. Methods: We compared the transcriptomes of MAMs recruited to lung metastases and resident alveolar macrophages (RAMs) and identified a large variety of differentially expressed genes and their associated signaling pathways. Some of the changes were validated using qRT-PCR and immunofluorescence. To probe the functional relevance to metastatic growth, a gene-targeting mouse model of female mice in the C57BL6/J background was used to study allograft inflammatory factor 1 (AIF1, also known as ionized calcium-binding adapter molecule 1 or IBA1). Results: Interferon signaling is one of the most activated pathways in MAMs, with strong upregulation of multiple components of the pathway and a significant enrichment for the gene signatures of interferon-alpha-treated human macrophages. Aif1, an interferon-responsive gene that regulates multiple macrophage activities, was robustly induced in MAMs. Aif1 deficiency in MAMs, however, did not affect development of lung metastases, suggesting that AIF1 indicates MAM activation but is dispensable for regulating metastasis. Conclusions: The drastically different gene expression profile of MAMs as compared to RAMs suggests an important role in promoting metastatic growth. Dissection of the underlying mechanisms and functional validation of potential targets in the profile may provide novel therapeutic strategies for the treatment of metastatic diseases.


Introduction
Tumor-associated macrophages are important in multiple steps of tumorigenesis and progression, including regulation of tumor cell invasion into the stroma, intravasation into blood vessels, seeding at distal organs, angiogenesis, inflammation and immune suppression (Qian & Pollard, 2010). Compared to the abundant knowledge of tumor-associated macrophages at primary sites, the role of the macrophages at metastatic sites, i.e. metastasis-associated macrophages (MAMs), has only begun to emerge in recent years. In mice, lung metastases resulting from mammary cancer have been shown to recruit MAMs that are CD11B + CD11C -, as compared to CD11B -CD11C + resident alveolar macrophages (RAMs) (Qian et al., 2009). MAMs and their monocytic progenitors express CCR2 and respond to the CCL2 signal emitted from metastases (Lu & Kang, 2009;Qian et al., 2011). Once recruited to the metastatic site, MAMs can secrete vascular endothelial growth factor A to facilitate cancer cell extravasation for seeding of the lung (Qian et al., 2011). MAMs also secrete CCL3, which acts in an autocrine manner to augment MAM recruitment and enhance their activity to promote metastatic growth (Kitamura et al., 2015). Despite the significance of these findings, they represent only a patchwork of a global program of MAMs, which remains elusive. To this end, we performed deep RNA sequencing (RNA-seq) in MAMs isolated from micro-dissected lung metastases 11 days after tumor cell inoculation to gain unbiased insight into the global transcriptional program and to learn how MAMs may use this program to regulate metastatic growth in secondary organs.

Ethics statement and animal experiments
All procedures involving mice were conducted in accordance with National Institutes of Health regulations concerning the use and care of experimental animals and were approved by the Albert Einstein College of Medicine Animal Use Committee (animal protocol numbers: 20180202 and 20180205). Mice were housed and maintained in a barrier facility at the Albert Einstein College of Medicine and were kept on 12-hour light/dark cycle and had ad libitum access to chow and water. The number of animals used in each experiment is explained in figure legends. A total of 66 mice were used. No mice were excluded from analyses. All efforts were made to ameliorate harm to animals. Mice carrying lung metastases were terminated by cervical dislocation after anesthesia with isoflurane before showing signs of difficulty breathing. Anesthesia with isoflurane was also performed for intravital imaging as below.
C57BL6/J wild type (WT) mice were purchased from the Jackson Laboratory and were pooled together and were randomly assigned to the non-treated control group or the metastasis group. Lung metastases were induced by intravenous (iv) injection of 1 × 10 6 E0771-LG cells (Kitamura et al., 2015) to the tail vein of syngenic C57BL6/J female mice (6-8 weeks old) and were analyzed on day 11. Aif1 -/-mice were obtained by replacing all coding sequences with a modified LacZ gene in the 129SvEv background and were backcrossed into the C57BL6/J background for 14 generations by NS (Casimiro et al., 2013) and derived from his colony according to genotype.
For bone marrow transplantation (BMT), total bone marrow (BM) cells were extracted from 8-12 weeks old WT or Aif1 -/-female donor mice from their femurs, tibiae and spine by grinding in a mortar. Red blood cells were removed by incubation with RBC Lysis Buffer (Biolegend 420301) for 5 min at 4°C prior to injections. Female WT recipient mice of 4-6 weeks old were irradiated with 10 Gy gamma rays split into two doses with a four-hour interval (Mark I-68A 137Cs irradiator from JL Shepherd and Associates, San Fernando, CA). The random allocation of mice to experimental group (WT vs Aif1-/-) was driven by Mendelian Inheritance. The mice were then iv injected with 1 × 10 7 BM cells from one of the donors the next day. Five weeks after BMT, lung metastases were induced by iv injection of 1 × 10 6 E0771-LG-luciferase-ZsGreen cells (Zheng et al., 2019), and the mice were analyzed on day 10 by in vivo imaging system (IVIS) and histology.
For IVIS imaging, mice were anesthetized with isoflurane (2% v/v with oxygen as the carrier gas) in an inhalation chamber (VetEquipt 911103) before retro-orbital injection with d-luciferin in PBS (GoldBio, LUCK-1G, 1.5 mg/100 µL/20 g mouse). Afterwards, the mice were imaged using IVIS Spectrum In Vivo Imaging System (Perkin Elmer) while the anesthesia was maintained in the imaging chamber. Photon flux (photon/second/cm 2 /steradian) in the lung area was analyzed using Living Image Software (Perkin Elmer, v4.3.1; a possible free alternative is Aura Imaging Software from Spectral Instruments Imaging) and was expressed as relative values with respect to the lowest value in the WT group.
Flow cytometry and cell sorting Lungs were perfused with PBS through the right ventricle, and metastases (<1 mm in diameter) were dissected under a dissection microscope and pooled from 1-5 mice. The tissues were subsequently digested using an enzyme mix of Liberase DL (Sigma-Aldrich 5466202001, 0.52 U/mL), TL (Sigma-Aldrich 5401020001, 0.26 U/mL) and DNase I (Sigma-Aldrich DN25, 150 µg/mL) for 30 min at 37°C. For sorting for RNA-seq or qRT-PCR, transcription inhibitors alpha-amanitin (Sigma-Aldrich A2263, 5 µg/mL) and actinomycin D (Sigma-Aldrich A1410, 1 µg/mL) were also added to the digestion buffer. Flow cytometry was analyzed using a LSRII cytometer (BD Biosciences), and the data were analyzed using Flowjo software (TreeStar, v10; a possible free alternative is Flowing Software 2.5.1 from University of Turku). FACSAria II (BD Biosciences) and Moflo Astrios (Beckman Coulter) were used for sorting. Antibody information is found in Table 1. RNA isolation, qRT-PCR and RNA sequencing Total RNAs were extracted using Picopure RNA Isolation kit (Arcturus KIT0202). RNA sequencing was performed at Beijing Genomic Institute, using the Ovation® RNA-Seq System V2 kit for library construction, pair-end 100 bp and Hiseq 4000, which generated 60-80M reads per sample. The reads were aligned to the mouse reference genome (GRCm38/mm10) using Tophat (v2.0.13) (Kim et al., 2013). Uniquely mapped reads were counted for each gene using htseq-count in the HTSeq package (v0.6.1) with gene models from UCSC RefGene (Anders et al., 2015). FPKM values were generated using Cufflinks (v2.2.1) (Trapnell et al., 2013). Differentially expressed genes were identified using DESeq2 (Love et al., 2014). For qRT-PCR, RNAs were reverse transcribed and amplified using QuantiTect® Whole Transcriptome (Qiagen 207043) before qPCR. Gene expression was normalized to beta-actin. Relative expression is calculated using the formula -ddCt, where Ct stands for threshold cycles. See Table 2 for Taqman gene expression assays.
Unsupervised hierarchical clustering, Gene Ontology, canonical pathway enrichment and gene set enrichment analysis For unsupervised hierarchical clustering, FPKM values were transformed by log2. Without prior grouping, genes of log2 (mean of FPKMs) > 5 and standard deviation > 1.5 were used for clustering. The differentially expressed genes (Extended Data: Table 3 (Zheng et al., 2021)) were analyzed for gene ontology annotation using the DAVID software (v6.8) (Huang et al., 2009) and the canonical pathway analysis using QIAGEN Ingenuity Pathway Analysis (QIAGEN IPA, March 2017 release; a possible free alternative is DAVID v6.8) (Krämer et al., 2014). For gene set enrichment analysis (GSEA), the MAM gene expression data were pre-ranked by the false-discovery rate (FDR) values: for upregulated genes, -log2 (FDR + A); for downregulated genes, log2 (FDR + A), and A = minimal FDR value that is not 0. This pre-ranked gene list was then analyzed in GSEA to interrogate if the three gene sets associated with IFN-alpha treatment of human macrophages (Greenwell-Wild et al., 2009;Liu et al., 2012;Tassiulas et al., 2004) as indicated in Figure 1 showed significant enrichment. GSEA (v3.0) was used for the analysis (Subramanian et al., 2005).

Immunofluorescence and immunohistochemistry
Briefly, lungs were perfused with 1% w/v PFA through the right ventricle, inflated with 2% w/v agarose in PBS via the trachea (for frozen sections) or 1% PFA (for paraffin sections), followed by immersion fixation in 4% w/v PFA for 1 hour at +4°C. For frozen embedding, the tissues were incubated in 25% w/v sucrose in PBS for 6 h or overnight before embedding in OCT Compound (Fisher HealthCare, 4585). Staining was performed in a similar manner to that previously published (Zheng et al., 2019). Briefly, 20-µm frozen sections or 5-µm paraffin sections of the lung were permeabilized with PBS containing 0.3% v/v Triton-X, blocked with PBS containing 0.3% v/v Triton-X, 5% v/v donkey serum, 0.05% w/v sodium azide and 0.2% w/v bovine serum albumin, and incubated with primary antibodies overnight. The sections were subsequently washed and incubated with secondary antibodies conjugated with fluorochromes or horseradish peroxidase (HRP). For HRP-conjugated secondary antibodies, the signal was visualized using ImmPACT® NovaRED® Substrate (Vector Laboratories, SK-4805). For F4/80 immunofluorescence, an additional amplification step using anti-rat-biotin and streptavidin-AF488 was performed. For antigen retrieval in immunohistochemistry of paraffin sections, Tris-EDTA (pH 9.0) (VWR, #K043) for AIF1 was used. See Table 1 for the antibody details.
Quantification of lung metastases was performed similarly as previously described (Nielsen et al., 2001). Briefly, paraffin sections were stained with hematoxylin and scanned with the High Capacity Slide Scanner PANNORAMIC 250 Flash III (3D Histech). Images were analyzed with Fiji (NIH, v2.0.0-rc-64/1.51s), using the Colour Deconvolution and Analyze Particles functions, to quantify the total area of lung metastases and the total lung area. Eighteen mice (nine in each group) were analyzed for histology of lung metastases.

Statistical analysis
Each sample represents an individual mouse or pooled lung metastases obtained from 1-5 mice. Student's t test or twoway ANOVA (for pooling of multiple experiments) were used. Statistical analyses were carried out with Graphpad Prism (version 7). For RNA-seq analysis, multiple hypothesis testing was adjusted using the Benjamini and Hochberg FDR method. Sample sizes were determined empirically.

Gene expression profiling identifies strong activation of interferon signaling in MAMs
To analyze the gene expression profiles of MAMs and RAMs, we dissected metastases from affected lungs 11 days after seeding and normal lungs before cell dissociation for fluorescence-activated cell sorting (FACS) sorting of MAMs and RAMs. Consistent with previous findings, lung metastases were significantly enriched with the CD11B + CD11Csubset of CD45 + F4/80 + LY6C -MRC1 + FSC high macrophages, as compared to the predominant CD11B -CD11C + RAMs in normal lungs ( Figure 1A and 1B). RNA-seq of these two populations revealed distinct gene expression programs between the two, as unsupervised hierarchical clustering completely segregated the two populations ( Figure 1C).
Gene Ontology annotation of the genes differentially expressed by MAMs showed enrichment for cellular responses to cytokine stimulus and reduction in lipid metabolism (Extended Data: Tables 1-3 (Zheng et al., 2021)). To study the biological pathways enriched by the differentially expressed genes in more detail, we performed canonical pathway enrichment analysis using QIAGEN IPA (Extended Data: Table 4 (Zheng et al., 2021)). "Interferon (IFN) Signaling pathway" was the most enriched pathway and was predicted to be significantly activated ( Figure 2A). An independent qRT-PCR analysis of the genes in this pathway confirmed the upregulation of most of the genes ( Figure 2B and 2C). In addition, GSEA revealed that MAMs were strongly enriched for three independent gene sets obtained from IFN-alpha-treated human macrophages ( Figure 2D). These data suggest that increased expression of the pathway components indeed translates into the signaling output of the pathway. Together with the additional significantly altered pathways such as the NF-kappa B pathway and various interleukin pathways (Extended Data:  et al., 2021)) and is in two significantly enriched gene sets that are relevant to the current study -"response to interferon-gamma" and "leukocyte chemotaxis" (Extended Data:  et al., 2013). It integrates various inflammatory stimuli, supports production of particular cytokines, and is critical for survival and pro-inflammatory activity of macrophages (Chinnasamy et al., 2015;Zhao et al., 2013). We verified with an independent qRT-PCR assay the dramatic upregulation of the Aif1 transcript in MAMs ( Figure 3A). This finding was also corroborated by immunofluorescent staining for the AIF1 protein, which showed that approximately 40% of macrophages within metastatic nodules robustly expressed AIF1, whereas little was detected in RAMs in the normal lung ( Figure 3B and 3C).
Next, we tested if induction of AIF1 in MAMs is required for promoting metastatic growth in the lung. To restrict Aif1 gene ablation to the hematopoietic lineage, we performed a bone marrow transplantation assay in which the bone marrow of Aif1 -/-donor mice (Casimiro et al., 2013) was transplanted into WT recipient mice. Lung metastases were induced in these mice for comparison with WT-to-WT recipient mice. Bioluminescent signals emitted from tumor cells that expressed luciferase revealed little difference between the two genotypes ( Figure 3D and 3E). Histological evaluation of the metastases was consistent with the bioluminescent assay ( Figure 3F and 3G). The lack of AIF1-expressing cells in metastases excluded the possibility of inefficient BM engraftment ( Figure 3H). Together, these data suggest that AIF1 indicates MAM activation, but its expression by MAMs is dispensable for regulation of metastasis.

Discussion
In this study, we profiled the transcriptome of MAMs by deep RNA-seq and identified a distinct expression program that are indicative of important metastasis-regulatory functions.
IFN signaling has been shown to have anti-metastatic effects via inhibition of epithelial-to-mesenchymal transition of tumor cells, angiogenesis, intravasation, survival in the circulation, homing to target tissues, and extravasation (Ortiz & Fuchs, 2017). The effects of IFN treatment on MAMs with respect to regulation of metastasis are, however, unknown. Given that ablation of MAMs or inhibition of MAM recruitment suppressed metastatic growth in multiple models (Lu & Kang, 2009;Nielsen et al., 2016;Qian et al., 2011;Qian et al., 2009), it is plausible that activation of IFN signaling promotes the pro-metastatic activity of MAMs, which may therefore limit the favorable effects of IFN-alpha treatment for patients of metastatic diseases (Eggermont et al., 2014). In support of this hypothesis, STAT1, an essential signaling molecule of the IFN pathway, in tumor-associated macrophages regulates the T cell-suppressive activity of tumor-associated macrophages (Kusmartsev & Gabrilovich, 2005). It has been suggested that the difficulty to target the right IFNs at the right dose in the right type of cells might underlie the basis for the limited therapeutic effects of IFNs (Dunn et al., 2006). It is thus interesting to determine in future work if macrophage-specific inactivation of IFN signaling inhibits development of metastasis, and more importantly if MAM-targeting will improve IFN therapy and immunotherapy.
Despite robust induction of the interferon-inducible gene Aif1 in MAMs, we did not see an effect of AIF1 deficiency on development of lung metastases. Similar strong correlation of AIF1-expressing macrophages/microglial cells with malignancy has been previously observed for human gliomas (Deininger et al., 2000). Another study shows that AIF1 is expressed by brain-infiltrating myeloid cells, which enhance metastatic growth by promoting proliferation and reducing apoptosis of tumor cells (Zhang et al., 2015). Functional testing in neither study has been attempted. It is thus interesting to determine whether AIF1 expression in macrophages or microglia plays a functional role in these settings or serves as a marker that indicates cell reprogramming in the tumor microenvironment as identified in our study.
In conclusion, we have identified a distinct gene expression program in MAMs that include strong activation of the IFN pathway and AIF1. Since cancer cells at metastatic sites benefit from these recruited MAMs, the gene expression profile revealed in this study can provide an important platform to uncover novel mechanisms and potential therapeutic targets that are essential for MAMs to promote metastatic growth.
This project contains the following underlying data: • Fig. 1A and B: FCS files for the FACS data and the values for the graph in panel B in XLSX format.
• Fig. 2B: ddCt values for the qPCR analysis of the IFN pathway.
• Fig. 3B and C: Raw images of the fluorescence and values for the graph.
• Fig. 3D and E: Raw IVIS images and values for the graph.
• Fig. 3F: Raw images for histological evaluation of lung metastases and values for the graph.
Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication). This project contains the following extended data: 1. Extended Data Table 1: Fragments per kilobase of transcript per million mapped reads (FPKM) values of MAMs and RAMs. Uniquely mapped reads were counted for each gene using htseq-count in the HTSeq package (v0.6.1) with gene models from UCSC RefGene. Table 2: Differentially expressed genes as identified using DEseq2. Mean FPKM values < 1 in both groups under comparison were excluded from differential expression analysis. Fold change >2, FDR < 0.05. Table 3: Highly differentially expressed genes for IPA analysis. This gene set is obtained with the intersection of Extended Data Table 1 and 2, which is subsequently applied with the following more stringent criteria: log2 (fold change) > 1.5, FDR < 0.03; for upregulated genes, mean FPKMs of MAMs > 20; for downregulated genes, mean FPKMs of RAMs > 40. Table 4: Gene Ontology annotation of the highly differentially expressed genes. Upregulated and downregulated genes from Extended Data Table  3 were analyzed by the GOBP5 analysis of DAVID Bioinformatics Resources.

Open Peer Review
This manuscript titled "Induction of interferon signaling and allograft inflammatory factor 1 in macrophages in a mouse model of breast cancer metastases" by Zheng et al. compares upregulated signaling pathways in MAMs (metastasis-associated macrophages) to those in RAMs (resident alveolar macrophages). The authors performed transcriptomic analyses that highlight the difference of macrophages in lung metastasis from those in normal lung, which led to evaluate the effect of targeting MAM-Aif1 in lung metastases. Although Aif1 deficiency in MAMs does not seem to be really important in lung metastasis, unfortunately, I believe this set of results needs to be shared by the research community to improve the understanding in the field. I only have some minor comments as summarized below: To study the effect of Aif1 deficiency in MAMs, the authors performed a series of BMT of WT and Aif1-/-mice to WT recipient mice, respectively. If conditional Aif1 knockout mouse model was available, it would have been clearer to more specifically target Aif1 only in macrophages. Arent there any Aif1 cKO mice available, and if not/so the authors could discuss a bit in the discussion section? 1.
In Figure 3, the authors represent histological evaluation of the lung metastases of WT-to-WT mice and Aif1-/--to-WT mice which shows the lack of Aif1 expression in metastatic nodules of Aif1-/--to-WT group. In addition to this, have the authors analyzed (or counted) macrophage infiltration per se to lung metastases between these two groups by any chance? 2.
In Results, the authors may want to provide relevant reference(s) to support the following statement: "Consistent with previous findings, lung metastases were significantly enriched with the CD11B + CD11Csubset of CD45 + F4/80 + LY6C -MRC1 + FSC high macrophages, as compared to the predominant CD11B -CD11C + RAMs in normal lungs ( Figure 1A and 1B).

3.
Is the work clearly and accurately presented and does it cite the current literature? Figure 3, the authors represent histological evaluation of the lung metastases of WT-to-WT mice and Aif1-/--to-WT mice which shows the lack of Aif1 expression in metastatic nodules of Aif1-/--to-WT group. In addition to this, have the authors analyzed (or counted) macrophage infiltration per se to lung metastases between these two groups by any chance?

In
Response: Again, an excellent point. Unfortunately, as the F4/80 staining in paraffin sections has not been consistent (note that the F4/80 staining in original Fig. 3B was in frozen sections), we were not able to draw any conclusions. However, we have added new data in the revised version, which showed that Aif1 gene ablation did not affect MAM recruitment to lung metastases by flow cytometry (new Fig. 3E and F).
Although there are some useful points in the discussion, it would be interesting to include some comment on similarities between the transcriptional profile of the murine MAMs and that of the human tumour-associated macrophages from breast and endometrial cancer profiled in the group's 2019 paper in Cancer Cell.
In summary, the data provided can be considered preliminary but are worthy of further investigation.

Is the work clearly and accurately presented and does it cite the current literature? Partly
Is the study design appropriate and is the work technically sound? Yes

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Vincenzo Bronte
Section of Immunology, Department of Medicine, University of Verona, Verona, Italy

Brief summary of the main findings
The manuscript from Wei Zhang et al. describes a gene signature in metastasis associated macrophages (MAMs) isolated from lung nodules of transplantable mouse tumors and compared to resident alveolar macrophages. The interferon (IFN) signaling pathway was highly enriched in MAMs and overlapped with the signature of human macrophages exposed to type I IFN, i.e. IFNalpha. Among the top regulated genes, allograft inflammatory factor 1 (AIF1, also known as ionized calcium-binding adapter molecule 1 or IBA1) was further investigated. Chimeric mice transplanted with Aif1-deficient bone marrow cells did not show differences in the metastatic burden, indicating that AIF1 is not involved in regulating the pro-metastasis activity of MAMs.

Overall evaluation
Even though this manuscript contains a limited number of data, the intriguing observation that type I IFN might have a detrimental effect on tumor immunity is worthy of consideration, especially in view of the limited efficacy of IFN treatment for melanoma patients. These data suggest that the variability in clinical findings might be related to the confounding effects of type I IFN on the metastatic niche.

Limits of the work and suggestions
The current work should be expanded to define whether the IFN signature is also shared by human MAMs, an approach that could also offer the perspective of prioritizing other genes of the IFN-dependent signature and establish the next candidate(s) for in vivo validation studies. However, before fully discarding AIF1, the Authors should confirm molecularly the absence of the gene in MAMs from chimeric mice since the protein detection by immunohistochemistry might not detect a low but functionally relevant amount of AIF1. Finally, experiments in immunodeficient mice will be necessary to sort out the contribution of the IFN-regulated pathway in MAMs independently of their potential effect on adaptive immunity.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Yes

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate? Yes included the following in Discussion (3 rd paragraph) to take into account this possibility: "Although we cannot rule out the possibility that the residual host-derived Aif1 WT BM cells may be sufficient to rescue the impaired function of Aif1-/-cells (if any), it is a less likely event given the magnitude of reduction of AIF1-expressing cells.
On the contribution of the IFN pathway from different cell types. We agree with the reviewer that IFN signaling acts on a broad range of cell types. To this end, we restricted IFN signaling inactivation to macrophages by crossing Ifnar1 f/f mice with Cd169-Cre mice. Unfortunately, as the lab was closed, we did not obtain the desired genotype to test this hypothesis before the close-down. ○ Competing Interests: No competing interests were disclosed.