Hemorrhage-activated NRF2 in tumor-associated macrophages drives cancer growth, invasion, and immunotherapy resistance

Microscopic hemorrhage is a common aspect of cancers, yet its potential role as an independent factor influencing both cancer progression and therapeutic response is largely ignored. Recognizing the essential function of macrophages in red blood cell disposal, we explored a pathway that connects intratumoral hemorrhage with the formation of cancer-promoting tumor-associated macrophages (TAMs). Using spatial transcriptomics, we found that NRF2-activated myeloid cells possessing characteristics of procancerous TAMs tend to cluster in perinecrotic hemorrhagic tumor regions. These cells resembled antiinflammatory erythrophagocytic macrophages. We identified heme, a red blood cell metabolite, as a pivotal microenvironmental factor steering macrophages toward protumorigenic activities. Single-cell RNA-Seq and functional assays of TAMs in 3D cell culture spheroids revealed how elevated intracellular heme signals via the transcription factor NRF2 to induce cancer-promoting TAMs. These TAMs stabilized epithelial-mesenchymal transition, enhancing cancer invasiveness and metastatic potential. Additionally, NRF2-activated macrophages exhibited resistance to reprogramming by IFN-γ and anti-CD40 antibodies, reducing their tumoricidal capacity. Furthermore, MC38 colon adenocarcinoma–bearing mice with NRF2 constitutively activated in leukocytes were resistant to anti-CD40 immunotherapy. Overall, our findings emphasize hemorrhage-activated NRF2 in TAMs as a driver of cancer progression, suggesting that targeting this pathway could offer new strategies to enhance cancer immunity and overcome therapy resistance.

these cancer cells with differentially pretreated BMDMs.For mCherry-4T1 spheroids, data are plotted as the mean ± 95% CI (n=10) of the red fluorescence integrated across the spheroid area.For GL261-Luc spheroids, data are plotted as the mean ± 95% CI (n= 10) of the spheroid area, and luminescence of each spheroid was measured at the end of the study after adding D-luciferin to the cell culture medium (ANOVA with Tukey-Kramer posttest corrected for each comparison, GL261-L vs. BMDMs IFNγ p 0.97, GL261-L vs. control BMDMs p 0.30, control BMDMs vs. IFNγ BMDMs p 0.10, heme BMDMs vs. heme + IFNγ p 0.31, further comparisons p< 0.0001).
Supplementary Figure 5 Supplementary Figure 5 (related to Figure 8)

A. Experimental workflow of lung metastasis model B.
Approximately 750 mixed spheroids per condition were collected from microwell plates on day 4 post-spheroid formation and injected i.v.into Rag2 −/− γc −/− mice.Lungs were harvested 21 days after injection.Representative lung paraffin sections were stained with H&E and for GFP.Scale bar = 5 mm.C. Whole-lung GFP fluorescence intensity was integrated across the lung image area (gray dots= dorsal view, red dots = ventral view).ANOVA with Tukey-Kramer posttest corrected for multiple comparisons.

Data and code availability
Sequencing data are publicly available (GEO accession code GSE237612).

TCGA data analysis -Regression modeling
We used Xena browser (1) to explore data in The Cancer Genome Atlas (TCGA) Pan-Cancer (PANCAN) database (https://www.cancer.gov/tcga)and to plot survival curves for CD163 and SPP1 expression, including all solid cancer types excluding lymphoma and leukemia.The log2(count+1)-transformed and batch-corrected gene expression data were exported from the database and further analyzed in JMP 15 (SAS Institute Inc.).To make sure that the expression data of our marker genes were not excessively zero-inflated, we confirmed approximately normal distribution of the log2(count+1)-transformed data using the Distribution function combined with Fit normal and Normal quantile plot functions before computing multiple linear regression models (ordinary least squares method) using CD163 and SPP1 as the response variables and the tissue microenvironmental factors as predictors.In Figures 1A and   1B, we report the overall fit of the model (r 2 ) and the coefficient for each predictor gene with the respective p-value.

Animals
C57BL/6J (JAX TM strain) mice were obtained from Charles River Laboratories.

Cell lines and primary cultures
Tumor cell line culture
Cultures were treated on day 3 with 300 μM heme.For inflammatory polarization, IFNγ (10 ng/ml, PeproTech) or LPS (10 ng/ml, Sigma) was added on day 6 for 24 hours.For in vitro anti-CD40 stimulation, FGK45 (1 μgr/ml, InVivoPLus) was crosslinked with goat anti-rat immunoglobulin G (0.5 μgr/ml, BioLegend) for 30 min at room temperature before addition to the culture medium for 24 hours.The BM cells were harvested for analysis on day 7 from the temperature-responsive cell culture plates after cooling to room temperature.Cells were washed twice in PBS and centrifuged (300g, 10 min) before processing.For experiments involving conditioned medium from tumor cells, BMDMs were seeded after washing at the end of the differentiation period in 12-well plates (TPP) in MC38 conditioned medium overnight and lysed in RNA lysis buffer 1% β-mercaptoethanol for transcriptome analysis.

Heme preparation for cell culture
Hemin (heme-chloride) was obtained from Frontier Scientific (Newark).Batches were tested endotoxin-free and prepared as heme-albumin for cell treatments as described (5).

Quantification of spheroid growth and invasion
Single spheroids were imaged in the cell culture incubator with an IncuCyte S3 instrument (Sartorius).Green or red fluorescence and phase contrast images of the spheroids were acquired every 4 hours for seven to ten days.The area and fluorescence intensities of the images were measured using the IncuCyte Spheroid Software Module (Sartorius).Data are reported as spheroid fluorescence intensity integrated across the spheroid area (for tumor cells expressing a fluorescent protein) or as spheroid area.For the spheroid invasion assay, a mask based on the invading cell area was created automatically with the IncuCyte Spheroid Software Module.
Multispheroids were scanned using a Zeiss Axio Observer Z1 microscope.The spheroid area was quantified manually in QuPath (6) (v0.3.0), and fluorescence intensity was measured using QuPath's intensity feature plugin.The spheroids were detected in the EGFP/FITC channel using a set pixel size of 1.26 µm.

High-resolution spheroid imaging in glass bottom microwell plates
Spheroids were transferred from the microwell plate (Axonlab) to a flat glass-bottom plate (TPP) in FluoroBrite™ DMEM on day 4 post-spheroid formation and embedded into Cultrex extracellular matrix (Bio-Techne).Spheroid morphology and cancer cell invasion into the matrix was visualized after 24 hours by cell type-specific fluorescence with a Leica SP8 laser scanning microscope.3D representations from Z-stacks were rendered in Imaris software (Oxford Instruments).Z-slices were stacked for 2D projections using Adobe Photoshop.

Metabolic flux analysis
Spheroids were transferred from a microwell plate (Axonlab) on Geltrex (Thermo Fisher) matrix-coated Seahorse cell culture plates on days 4, 8, and 10 after spheroid formation.BMDMs were harvested on day 7 from the temperature-responsive cell culture plates after cooling to room temperature and replated into Seahorse 24-well plates.The mitochondrial function (oxygen consumption rate) and glycolysis (acidification rate) of spheroids or BMDMs were measured using a Seahorse XF24 extracellular flux analyzer and the Cell Mito Stress Kit (Agilent Technologies) according to the instructions provided by the manufacturer.

3D cell viability assay
ATP concentration was measured in single spheroids with the CellTiter-Glo® 3D cell viability assay on days 4, 8, and 10 post-spheroid formation, according to the instructions provided by the manufacturer (Promega).Luminescence was measured with an infinite M200 Pro plate reader (Tecan).

RBC isolation and RBC-ghost preparation
C57BL/6J mice were anesthetized with isoflurane, and blood was collected by cardiac puncture.
RBCs were then washed, centrifuged (4000g, 30 min, 4°C), and resuspended in PBS.For RBC-ghost preparations, RBCs were lysed in 50 ml ultrapure water for 30 min on ice, centrifuged (3000g, 20 min, 4°C), and resuspended in PBS.The procedure was performed twice for each sample.The solutions were stored at 4°C for 12 hours.

Matrigel plug preparation and digestion
Matrigel (Corning) was thawed overnight on ice in a cold room (4°C) and mixed with RBCs (RBC-heme) or RBC-ghost membranes at a 1:1 ratio.In fibroblast growth factor (FGF)-enriched plugs, Matrigel was mixed with 500 ng/ml FGF (R&D Systems).Matrigel plugs were explanted seven days post-injection.Excised Matrigel plugs were mechanically disrupted, placed in pyruvate-free DMEM with 1.2 mM calcium chloride dihydrate and 1 mg/mL collagenase IV (Stemcell), and incubated on a shaker at 37°C for 60 min.Cells were passed through a 70 µm cell strainer, and RBCs were lysed with 1x RBC lysis buffer (BioLegend).The samples were then washed, centrifuged (450g, 5 min, 4°C), and resuspended in 1% FBS in PBS for flow cytometry measurement or in 0.1% bovine serum albumin (BSA) in 2 mM EDTA PBS for macrophage isolation and transcriptome analysis.For spectrophotometry with spectral-deconvolution and size exclusion chromatography (SEC), RBC-heme plugs were digested on days 1, 4, and 5 postimplantation, centrifuged (10000g, 5 min, 4°C), and the supernatant analyzed stored at -80 °C until analysis.

Macrophage isolation from digested Matrigel plugs
Anti-rat IgG Dynabeads (Invitrogen) were washed and incubated with rat anti-mouse F4/80 IgG2a antibodies (BD Biosciences) and CD11b IgG2b antibodies (BioLegend) at a ratio of 3.33 µg of antibody per 50 µl of Dynabeads.Single-cell suspensions from Matrigel plugs were incubated with anti-F4/80-coated (for RT-qPCR experiments) or anti-F4/80 and anti-CD11b-coated Dynabeads (for scRNA-seq experiment) on a shaker at 4°C for 30 min.After incubation, a positive selection of Dynabead-bound single-cell suspensions was performed on a DynaMag magnet (Invitrogen) with three washing steps, as suggested by the manufacturer's instructions.

Spectrophotometry and spectral deconvolution
For spectral deconvolution of hemoglobin and hemoglobin metabolites (i.e., bilirubin), the absorption spectra of supernatants from digested RBC-heme Matrigel plugs and Matrigel alone were measured on a microvolume UV-Vis spectrophotometer (NanoDrop One, Thermo Fisher) using the manual mode.The absorption spectrum from 190 to 850 nm was acquired with a 0.5 nm resolution.We measured from each sample 2 µl according to the manufacturer's instructions.Then, the acquired absorption spectrum was deconvolved using reference spectra for oxy-hemoglobin, met-hemoglobin, and bilirubin based on a non-negative least-squares method (7).Before deconvolution, the biological background from the Matrigel was subtracted.

Size exclusion chromatography (SEC-HPLC)
After digestion, the supernatant was subjected to SEC to quantify free heme and hemoglobin in the RBC-heme plugs.Each supernatant was additionally measured after adding haptoglobin (CSL Behring) and hemopexin (CSL Behring) to assess the biologically available heme.Intact hemoglobin is complexed within haptoglobin, while biologically available heme is complexed within hemopexin, resulting in a mass shift and change in the elution profile.Therefore 200 µl of supernatant was mixed with 10 µl haptoglobin 1-1 (105 mg/ml), followed by 10 µl hemopexin (92 mg/ml).Then 10 µl of each sample was separated on a YMC-Pack Diol SEC column ( 4.6 mm ID, 30 nm, S-3 µm, 300 x 4.6 mm, DL30S03-3046WT) connected to a Gilson 307 HPLC Pump operated in an isocratic mode.Ammonium nitrate (0.2M, pH 7.4) with a flow of 0.5 ml/min was used as a mobile phase, and the elution profile was measured at 414 nm using a spectrophotometer (Jasco UV-970 Intelligent UV/VIS Detector) (8,9).
For anti-CD40 antibody experiments, the mice were treated intravenously on day 7 with an agonistic anti-CD40 antibody (20 mg/kg, InVivoPlus, clone FGK4.5) or an isotype control antibody.Mice were euthanized after 24 hours, and plugs were collected.F4/80 + macrophages were recovered from digested Matrigel plugs, and gene expression was analyzed by RT-qPCR.

Lung metastasis model in mice
Approximately 750 spheroids were collected from microwell plates (equal to the content of one macro well) at different time points post-spheroid formation and injected intravenously into the tail vein of C57BL/6J or Rag2 −/− γc −/− mice.Three weeks postinjection, the lungs of anesthetized mice were perfused with PBS through the right ventricle and the trachea and collected for whole organ fluorescence imaging with a Zeiss Discovery V8 stereomicroscope and histology.For the metastasis experiments shown in Figure 9H, metastasis were manually counted to enhance sensitivity and specificity in the low disease burden range, to assess the effects of anti-CD40 treatment.

Tumor growth model in mice
Once confluent, GFP-MC38 tumor cells were harvested using 5 mM EDTA (Gibco) (4 min at 37°C) and washed twice in PBS.MC38 cells (2 x 10 6 ) in culture medium were mixed with Geltrex (Thermo Fisher) and injected subcutaneously into the mouse flanks.Agonistic anti-CD40 treatment (20 mg/kg, InVivoPlus, clone FGK4.5) or an isotype control antibody was administered intravenously seven days after tumor cell injection.In some experiments, animals were rechallenged two days after the first treatment.Mice were euthanized, and tumors were collected two or three days after antibody administration.GFP fluorescence was measured immediately, and tumors were then fixed in formalin (10%) and stored at room temperature.

Immunohistochemistry and iron staining
Nonheme iron staining: Tissue paraffin sections were incubated with Perl's iron reagent containing 5% potassium ferrocyanide and 2% hydrochloric acid for 60 min at room temperature, after which they were rinsed in deionized water.Sections were then incubated with 3% hydrogen peroxide and methanol for 20 min at room temperature.
GFP staining: Tissue sections were incubated overnight with a goat anti-GFP antibody (Abcam) diluted 1:1000, followed by a biotinylated horse anti-goat secondary antibody (Vector) diluted

scRNA-seq sample preparation
Single-cell suspensions were multiplexed according to the experimental design and then used for GEM generation following the 10X Genomics protocol (CG000388 Demonstrated Protocol Chromium Next GEM Single Cell 3ʹ v3.1 (Dual Index) with Feature Barcode technology for Cell Multiplexing Rev B) targeting a cell recovery rate of 10,000 cells.GEM generation and library preparation were performed according to the 10X Genomics protocol (CG000388).Ready-made libraries were sequenced at the Functional Genomics Center Zurich (FGCZ) on an Illumina NovaSeq 6000 system.Downstream analysis was performed in Python (version 3.8.6)with Scanpy (1.7.2) (14).

Quality Control and Preprocessing
To assess the quality of the cells, the following covariates were considered: number of genes expressed in a cell (n_genes_by_counts), number of counts per cell (total_counts), and percentage of mitochondrial RNA (pct_counts_mt).Cells that expressed fewer than min_genes or more than max_genes were filtered out.Cells with a percentage of mitochondrial RNA greater than max_pct_mt were considered dead and removed from the analysis.Genes that were expressed by fewer than min_cells cells were excluded.See below for the cutoff values used in each experiment.The count data were normalized by an algorithm based on deconvolving size factors from cell pools implemented in the R package scran (calculateSumFactors) (15) and log(x+1) (sc.pp.log1p) transformed, yielding normalized expression values.

Data integration
Multiplexed samples were merged into one dataset by simple concatenation.Additionally, samples from different experiments were integrated using the harmony algorithm (sc.external.pp.harmony_integrate) (16) after normalization and PCA.

Dimensionality reduction and clustering
For dimension reduction, the following steps were performed using the Python package Scanpy: identifying highly variable genes (sc.pp.highly_variable_genes), performing PCA using highly variable genes (sc.tl.pca), computing the neighborhood graph (sc.pp.neighbors) and computing the UMAP (sc.tl.umap).The cells were clustered using Leiden clustering (sc.tl.leiden), which depends on the neighborhood graph.The resolution of the Leiden clustering was chosen so that a biologically meaningful number of clusters was produced.

Cell type annotation and functional classification
To identify cell types, we analyzed the expression of marker genes and other differentially expressed genes (sc.tl.rank_genes_groups with method = 'wilcoxon').
Depending on the experiment, we analyzed these genes separately or by scoring gene sets (sc.tl.score_genes).GSEA was performed to assess functional and biological process-related differences between clusters or conditions.First, genes were ranked using the output of the Wilcoxon rank-sum test (rank = -log10(adj.p value)*sign(logfoldchange)) and then fed to the GSEA algorithm implemented in the Python package gseapy (17) (gseapy.prerank),resulting in a normalized enrichment score (NES) and a false discovery rate (FDR) per gene set.For transcription factor analysis, gseapy.enrichrwith the TRRUST_Transcription_Factors_2019 gene set database was used.

Quality Control and Preprocessing
Genes that were detected in fewer than 50 spots were excluded from further analysis.Counts  Experiment related to Figure 5 Spheroids of GFP-MC38 and GFP-MC38 mixed with BMDMs that were pretreated with heme were grown in microwell plates for four, eight, and ten days.9000 spheroids per condition were pooled and digested as aforementioned for 40 min in a water bath at 37°C with gentle shaking and subsequently resuspended using a 1 ml pipet to disrupt the digested spheroids into single cells.Cells were centrifuged at 300g for 5 min at 4°C and resuspended with PBS + 0.04% BSA.
Single-cell suspensions were counted and transferred to 2 ml safe lock tubes and directly used Experiment related to Figure 7 BMDMs from one conditional Keap1 KO mouse (control) and one WT littermate mouse (control and heme-treated) were resuspended and transferred into 15 ml falcon tubes.Tubes were filled up with RPMI 1% BSA + 2mM EDTA and centrifuged at 300g for 10 min at room temperature.
The supernatant was discarded, and cells were taken up in 1 ml RPMI + 1% BSA + 2mM EDTA for counting.1 x 10 6 cells per condition were taken for cell staining using lipid tags following 10X Genomics protocol (CG00391; Demonstrated Protocol Cell Multiplexing Oligo Labeling for Samples with >80% Viable Cells; Rev B).Cell Multiplexing Oligos (CMO) labels B308 -B311 (3' CellPlex Kit Set A) were used to label the four samples.
After labeling, cell suspensions were counted again, and the samples were pooled according to the pooling calculations in the appendix of the labeling protocol (CG00391) in equivalent ratios.
Experiment related to Figure 6 Preprocessing, Dimensionality reduction, Clustering Functional classification of clusters by means of DGE and GSEA using the MSigDB_Hallmark_2020 database (Fig. 6E) Visualization of the functional class by scoring gene sets (Fig. 6G) Experiment related to Figure 7 Preprocessing, Dimensionality reduction Visualization of expression values of marker genes (Fig. 7E) Transcription Factor Analysis of PCA loadings using the (Fig. 7F)

1: 500 .
All immunohistochemical sections were rinsed in 0.1 M phosphate buffer, pH 7.4, and incubated with diaminobenzidine (DAB, Abcam) for 2-5 min.After incubation, sections were washed in deionized water and lightly counterstained with Gill No. 2 hematoxylin (Sigma).Multiplexed immunofluorescence staining: Paraffin-embedded microtome sections were stained for immunofluorescence analysis using the Opal 4-Color anti-Rabbit Manual IHC Kit (Akoya Biosciences) as instructed by the manufacturer.The following primary antibodies were used: rabbit anti-mouse F4/80 antibody solution (1:1000, Cell Signaling), rabbit anti-mouse HMOX-1Sequencing-based workflows and data analysisBulk RNA sequencing RNA was extracted from BMDMs using the RNeasy Micro kit (Qiagen) according to the manufacturer's protocol, including on-column DNase I treatment.RNA quality was validated with an Agilent Technologies 2100 Bioanalyzer using an RNA chip, and only samples with an RNA integrity number (RIN) of > 9 were used for sequencing.cDNA libraries were generated from the RNA samples using the Illumina TruSeq RNA stranded kit following the manufacturer's instructions.Libraries were amplified by PCR (total of 15 cycles), and the quality and concentrations of the libraries were determined using an Agilent Fragment Analyzer with DNA High-Sensitivity Chips.The libraries were pooled in equimolar amounts and sequenced in an Illumina NovaSeq 6000 sequencer (single-end 100 bp) with a depth of approximately 20 million reads per sample.
were normalized (sc.pp.normalize_total) and log(x+1) (sc.pp.log1p) transformed.Tumor segmentationClustering (sc.pp.neighbors, sc.tl.leiden) and differential gene expression of the clusters (sc.tl.rank_genes_groups with method = 'wilcoxon') was performed to identify tumor cells.The projection of the identified tumor clusters on the spatial image yielded the tumor outline.Oxidative stress and Nrf2 scoring Spots were scored (sc.tl.score_genes) for oxidative stress using a gene set extracted from the list of differentially expressed genes of the RBC-heme plug (cutoffs: log2FC > 2, adj.p-val < 0.001).Genes that are activated by NRF2 were obtained from the TRRUST version 2 database(18).

Figure-by-
Figure-by-figure details of sample preparation and data analysis (sequencing data)Sample preparation for scRNA-seqExperiment related to Figure3Macrophage-enriched single-cell suspensions from RBC-heme and RBC-ghost Matrigel plugs were used for GEM generation following 10X Genomics protocol (CG000388 Demonstrated Protocol Chromium Next GEM Single Cell 3ʹ v3.1 (Dual Index) with Feature Barcode technology for Cell Multiplexing Rev B) targeting a cell recovery rate of 10'000 cells for each condition.
Fig. 3 500 8000 12.5 20 NA 5000 80000 no Fig.4 500 10000 15 20 NA NA no Fig. 5 200 0 10000 15 20 45 NA no Nrf2 -/-and WT littermates were obtained from Professor Yuet Wai Kan (University of California, San Francisco).Rag2 −/− γc −/− mice were obtained from the SwImMR.All breeding colonies were housed and bred in the specific pathogen-free (SPF) animal facility at the Laboratory Animal Services Center (LASC) of the University of Zurich in individually ventilated cages.Males and females mice aged 7-12 weeks were used for all experiments, and all experiments with mice were performed according to animal experimentation licenses approved by the Swiss Federal Veterinary Office.For all studies, mice were randomly allocated to treatment groups, and the investigators were blinded to allocation during experiments and outcome assessment.