Myoglobin Protects Breast Cancer Cells Due to Its ROS and NO Scavenging Properties

Myoglobin (MB) is an oxygen-binding protein usually found in cardiac myocytes and skeletal muscle fibers. It may function as a temporary storage and transport protein for O2 but could also have scavenging capacity for reactive oxygen and nitrogen species. In addition, MB has recently been identified as a hallmark in luminal breast cancer and was shown to be robustly induced under hypoxia. Cellular responses to hypoxia are regulated by the transcription factor hypoxia-inducible factor (HIF). For exploring the function of MB in breast cancer, we employed the human cell line MDA-MB-468. Cells were grown in monolayer or as 3D multicellular spheroids, which mimic the in vivo avascular tumor architecture and physiology with a heterogeneous cell population of proliferating cells in the rim and non-cycling or necrotic cells in the core region. This central necrosis was increased after MB knockdown, indicating a role for MB in hypoxic tumor regions. In addition, MB knockdown caused higher levels of HIF-1α protein after treatment with NO, which also plays an important role in cancer cell survival. MB knockdown also led to higher reactive oxygen species (ROS) levels in the cells after treatment with H2O2. To further explore the role of MB in cell survival, we performed RNA-Seq after MB knockdown and NO treatment. 1029 differentially expressed genes (DEGs), including 45 potential HIF-1 target genes, were annotated in regulatory pathways that modulate cellular function and maintenance, cell death and survival, and carbohydrate metabolism. Of these target genes, TMEFF1, TREX2, GLUT-1, MKNK-1, and RAB8B were significantly altered. Consistently, a decreased expression of GLUT-1, MKNK-1, and RAB8B after MB knockdown was confirmed by qPCR. All three genes of interest are often up regulated in cancer and correlate with a poor clinical outcome. Thus, our data indicate that myoglobin might influence the survival of breast cancer cells, possibly due to its ROS and NO scavenging properties and could be a valuable target for cancer therapy.


INTRODUCTION
Myoglobin (MB) is a cytoplasmic, oxygen-binding heme-protein mainly found in cardiac myocytes and skeletal muscle fibers These genes are often up regulated in cancer and correlate with a poor clinical outcome where it functions as a mobile oxygen reservoir. In times of high oxygen demand, MB is able to provide mitochondria with oxygen through its reversible O 2 storage and transport function (1,2). Besides these long-known functions, newer evidence showed that MB is a capable regulator of nitric oxide (NO) homeostasis in muscle cells by either scavenging (3) or producing NO (4,5). Other attributed functions of MB now include scavenging of reactive oxygen species (ROS) (6), and binding and transport of fatty acids and acylcarnitines (7,8). To date it is known that MB is not only strongly expressed in muscle tissue and myocytes, but also endogenously at lower levels in different forms of malignant tumors, like prostate cancer, nonsmall cell lung cancer, renal cell carcinoma, and breast cancer (9)(10)(11)(12). Kristiansen et al. analyzed the appearance of MB in a large comprehensive study of invasive and non-invasive breast cancer cases and found~40% of invasive carcinomas being at least moderately positive for MB expression. MB was preferentially detected in hormone receptor-positive tumors and Kaplan-Meyer analysis revealed a better prognostic outcome for patients with high MB levels (12). MB mRNA was furthermore detected at different levels in a subset of human breast cancer cell lines, like MDA-MB-468 (high-MB) or MCF-7 (low-MB) (12). In breast cancer cell lines, MB RNA and protein levels were inducible by prolonged hypoxia, which was in part attributed to the transcriptional activation by the hypoxia-inducible factors-1 and -2 (HIF-1, HIF-2) (13). MB transcription is thereby started from a hypoxia-inducible alternative upstream promoter region that is different from the myocytic standard promoter and could be silenced by estrogen hormone treatment (14,15). Studies showed a lower cellular O 2 uptake in breast cancer cells after MB knockdown (KD) as well as a decreased mitochondrial dehydrogenase activity, leading to the assumption that the occurrence of MB in tumor cells is not mainly linked to its main function as an O 2 storage and transport protein (13). Regulation and function of MB could therefore be fundamentally different between muscle and tumor tissue.
One of the main activators of cellular responses to hypoxia are HIFs, which induce the transcription of target genes mainly involved in angiogenesis, erythropoiesis and glycolysis, and also play an important role for the cellular adaptation of the tumor microenvironment (16,17). HIFs (HIF-1, HIF-2, and HIF-3) are heterodimeric transcription factors consisting of an oxygen sensitive alpha-subunit that is degraded by the proteasome under normoxic conditions, and a nuclear, unregulated betasubunit (18). Hypoxic regions are a common feature of many solid tumors (19) with a poorer prognosis because of a higher aggressiveness and progression of the tumor, and resistance to irradiation and chemotherapy (20,21). Here, HIF-1a serves as a marker for hypoxia correlated with malignant properties and increased patient mortality in many different types of cancer (21). In mamma carcinomas, HIF-1a-positivity is associated with higher proliferation rates and tumor size, higher grading status, shorter metastasis-free intervals, and negative hormone receptor status (22). Of note, not all breast cancer samples expressing HIF-1a are necessarily hypoxic. Vleugel et al. showed partially diffuse HIF-1a staining patterns in all regions of the tumor, indicating alternative activation pathways of HIF-1a (23). Indeed, there is a general link between HIF-1a stabilization and other inductors besides hypoxia. Induction of HIFs can occur through inflammatory cytokines or growth factors (24,25) as well as interaction with the NO signaling pathway. NO is thereby only indirectly affecting HIF, but can interact with and inhibit oxygen-sensor enzymes like prolylhydroxylases, which normally target the HIF alpha subunit for proteasomal degradation under normoxic conditions (26). Other activation mechanisms of HIF response include the interaction with H 2 O 2 and other ROS (27,28).
In 2020, breast cancer was the most frequently diagnosed cancer in women worldwide with many known risk factors, and the most common cause of cancer-related death in women in 110 countries (29). To date the exact role of MB in tumors in general and breast carcinoma in particular is not fully understood. A positive correlation between MB and HIF-1a target gene expression was shown in breast, prostate, and non-small cell lung cancer (9,10,12,30). By using an unbiased transcriptomics approach in combination with a siRNA-induced knockdown of MB and hypoxic versus normoxic incubation, a follow-up study reported an influence of MB on key metabolic and regulatory processes in these cancer cell lines. The RNA-Seq data suggested a functional impact of MB on ROS and NO homeostasis, correlating MB positively with increased HIF-1a and p53 signaling (31).
To test and extend the proposed working models, we investigated how MB influences HIF-1a and NO signaling pathways and cell survival by using the high-MB breast cancer cell line MDA-MB-468 in a monolayer and tumor spheroid cell culture model. When comparing wildtype (WT) cells to MB knockdown (KD) cells achieved by transient siRNA treatment, we found an increased central necrosis in tumor spheroids after MB KD, indicating that a lack of MB stimulates cell death. Furthermore, MB KD caused higher levels of HIF-1a protein after treatment with NO and higher reactive oxygen species (ROS) levels in the cells after treatment with H 2 O 2 . As HIF-1 is a transcription factor with several hundred target genes, we performed RNA-Seq analysis after MB KD and NO treatment to identify altered mRNA expression patterns. Of 45 potential HIF-1 target genes that were annotated in regulatory pathways that modulate cellular function and maintaining, cell death and survival, and carbohydrate metabolism, we identified a significant down regulation for GLUT-1, MKNK-1 and RAB8B which was verified by qPCR. All three genes are often up regulated in cancer and correlate with a poor clinical outcome (32)(33)(34). Based on our data, myoglobin therefore might influence the survival of breast cancer cells, possibly due to its ROS and NO scavenging properties and could be a valuable target for cancer therapy.

Cell Culture Under Normoxia and Hypoxia
For all experiments, the human mammary epithelial cell line MDA-MB-468 (ATCC HTB-132, verified by Leibniz Institute DSMZ, Braunschweig) was used as monolayer culture or in a three-dimensional tumor spheroid model. Cells were cultured in DMEM (Gibco) supplemented with 10% FCS (Sigma) and 1% Penicillin/Streptomycin (Gibco) under standard conditions (37°C, 5% CO 2 ) in water-saturated room air (21% O 2 ; normoxia). Depending on the experimental set-up, hypoxic cultivation was performed for up to 72 h under water-saturated 1% O 2 with 5% CO 2 and balanced N 2 at 37°C using the InvivO 2 400 hypoxia workstation + gas mixer Q (Baker Ruskinn).

Generation of Spheroids
Tumor spheroids of MDA-MB-468 cells were generated as described elsewhere (35). Briefly, cells were seeded in 6-well plates at 1x10^5 cells/ml and incubated for 24 h under standard conditions. Cells were harvested and mixed with 5% ice-cold reconstituted basement membrane (rBM, MatrigelTM) before transferring 200 μl of the cell suspension/2x10^4 cells into each well of 0.5% poly-HEMA coated 96-well plates. Spheroid formation was initiated by 10 min of centrifugation (1000 rpm, 4°C) and the spheroids then incubated for 24 h until beginning of experiments.

Central Necrosis
24 h after initial seeding, tumor spheroids were incubated for 24 h, 48 h or 72 h under normoxic or hypoxic conditions and the central necrosis of the spheroids was measured. For this, spheroids were treated for 4 h with 0.5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) for the detection of dead cells. To test the effects of different stress-inducing substances on the formation of a central necrosis, spheroids were furthermore treated with H 2 O 2 , NO or a combination of both (H 2 O 2 + NO) 24 h after initial seeding and incubated for 4 h under normoxic conditions before DAPI staining. As one further condition, spheroids were irradiated with 5 Gy using an X-ray machine (X-rad 320, PXI) operated at 320 kV and 12.5 mA with a 1.65 mm aluminum filter to mimic irradiation therapy of tumor cells. Fluorescence was measured with a Zeiss Axiovert 200M microscope and DS-Ri1 camera system using the NIS-elements F 3.0 imaging software. DAPI signal was analyzed as pixel/field of view using ImageJ for each tumor spheroid.

S-Nitrosoglutathione (GNSO) Exposure
S-Nitrosoglutathione (GNSO) was synthesized as a 1 mM stock solution as previously described (36) and used as a source for bioavailable NO. For experiments, spheroids were cultivated as described or monolayer cells were seeded at 100.000 cells/ml in 6-well plates and incubated for 24 h under either normoxic or hypoxic conditions before addition of 500 μM GNSO. After 0 -6 h of further normoxic or hypoxic incubation, respectively, samples were taken for further RNA or protein analysis.
When introducing the transient MB knockdown via RNA interference, cells were seeded for 24 h as mentioned above, followed by transfection with the respective siRNA for another 24 h under normoxic conditions before starting the GNSO treatment and sample uptake. HiPerFect (Catalog No. 301704; Qiagen, Hilden, Germany) was used as transfection reagent orientated towards the protocol provided. Briefly, 2.3x10^5 cells were seeded on a 6well plate and incubated for 24 h under standard conditions. For transfection, medium was changed to serum-free conditions and cells incubated with the transfection complex (100 μl serum-free medium + 10 μl siRNA + 12 μl HiPerFect transfection reagent, left at room temperature for 10 min for complex formation) for 20 min at room temperature. Next, medium was directly changed to serum-containing medium again and monolayer cells were left incubated depending on the experimental set-up. For spheroids, spheroid formation was introduced 4 h after transfection according to the protocol above. MB knockdown efficiency was validated by qPCR with efficiencies of up to 90% for both monolayer and spheroid cultures under both normoxic and hypoxic incubation conditions.

pC1-HyPer3 Plasmid
2x10^5 cells were seeded into WillCo-dish ® glass bottom dishes (WillCo Wells B.V, Amsterdam, Netherlands) and first transfected with siRNA according to the protocol above. After 24 h of incubation under standard conditions, the second transfection took place using the ViaFect transfection reagent (Promega) and 2 μg of pC1-HyPer3 plasmid, a genetically modified H 2 O 2 probe [Addgene plasmid # 42131 (37)] for detection of ROS. The transfection complex (200 μl serum-free medium + 4.3 μl plasmid DNA + 4 μl ViaFect transfection reagent) was formed for 13 min at room temperature before adding to the cells without changing of medium. After 24 h of further standard incubation, cells were prepared for ROS measurements as mentioned below.

ROS Measurements
For performing live cell imaging, the medium of the doubletransfected cells (MB knockdown + pC1-HyPer3 plasmid) was changed to FluoroBrite ™ DMEM (Gibco). Before measurement, ROS formation was introduced by addition of 100 μM H 2 O 2 to the cells directly under the microscope (Nikon Eclipse Ti). Fluorescence was measured using a 565/40 nm and 480/30 nm filter set and the mean intensity as pixel/field of view was analyzed using ImageJ. Data are represented as quotient of 565/480.

RNA Extraction, Reverse Transcription and Quantitative Real-Time PCR (qPCR)
Cells were lysed using a 4 M guanidinium thiocyanate solution and RNA extracted using the RNeasy Mini Kit (Qiagen GmbH, Germany) according to the protocol provided. A total of 1 μg RNA was transcribed into cDNA using oligo-dT-nucleotides (Life Technologies, USA) as primers and M-MLV reverse transcriptase (Promega). Quantitative real-time PCR (qPCR) was performed using the MESA Green qPCR MastermixPlus for SYBR ® ASSAY (Eurogentec) on an iCycler iQ5 multicolor real-time PCR detection system (Bio-Rad) using the following program: initial denaturation for 10 min at 95°C, then 45 cycles of denaturation (15 sec, 95°C), and elongation (60 sec, 60°C). Gene expression was analyzed using the 2 -ΔΔCT (cycle threshold) method with ribosomal protein S16 (rP) serving as a reference gene. Please refer to Table 1 for respective primers.

Protein Extraction and Western Blots
For protein extraction, cells were harvested with lysis buffer (0.5% NP40, 150 mM NaCl, 10 mM Tris pH 7.9, 2 mM EDTA, 10% protease inhibitor) and total protein concentration was measured with a bovine serum albumin (BSA) standard and the DC Protein Assay kit (Bio-Rad). For Western blotting, 40 μg of each protein extract was electrophoresed on a 7.5% SDS-PAGE gel before transfer to a nitrocellulose membrane. The membrane was blocked with 5% nonfat dry milk in 0,1% TBS-T for 1 h at room temperature before incubation with primary antibody at 4°C overnight. Secondary antibodies were incubated for 1 h at room temperature. Antibodies are listed in Table 2. For visualization, the ECL AdvanceTM Western Blotting Detection Kit (GE Healthcare) was employed. Signals were detected on a photo film. If applicable, membranes were stripped in Restore ™ PLUS Stripping Buffer (Thermo Scientific). Quantification of protein signals was performed using ImageJ.

Subcellular Fractioning
MDA-MB468 and LNCaP cells were cultured for 72 h at normoxia or hypoxia. About 5x10^7 cells were washed with PBS (PAA) supplemented with protease inhibitor and harvested from culture flasks with a cell scraper (Sarstedt). After 10 min centrifugation at 500xg at 4°C, subcellular fractionation was conducted with the Qproteome Cell Compartment kit (Qiagen). Protein samples were precipitated in four volumes of ice-cold acetone and solved in PBS. Protein concentration was determined by bicinchoninic acid assays (BCA), using the BCA Protein Assay reagent (Pierce) with BSA standard dilutions for quantification. Samples were electrophoresed on 12% SDS polyacrylamide gels and blotted to a nitrocellulose membrane. Membranes were incubated in 5% non-fat dry milk for 1 h and washed in 0.1% PBS-T for 5 min. Primary antibodies were applied ( Table 3) overnight at 4°C. Secondary antibodies ( Table 3) were incubated for 1 h at room temperature. ECL solution was applied for time frames varying from 45 sec to 5 min to detect light emission via photo film (Kodak) exposure. Membrane stripping was conducted with Mild Stripping buffer (Abcam).

RT 2 Profiler PCR Array
Cells were seeded at 100.000 cells/ml in 6-well plates and incubated for 24 h under normoxic conditions before transfection with the respective siRNA for another 24 h under normoxia. For the array, cells were then treated with 500 μM GNSO for 3 h before cell lysis and RNA extraction. With the RT² Profiler ™ PCR Array Human Hypoxia Signaling Pathway (Qiagen), the expression of 84 genes related to hypoxic signaling was analyzed in a 96-well plate qPCR format according to the protocol provided.

RNA-Seq
For RNA-Seq analyses, GNSO-treated (500μM) or untreated KD and WT cells were raised under normoxia or hypoxia (1% O 2 for 6 h). RNA was extracted (see above), quality-checked on an Agilent RNA 6000 nano Bioanalyzer Chip and quantified with AGT TGG TGC TGA ACG TCT GG GGT GAC CCT TAA AGA GCC TGA Ribosomal protein S16 (rP) AGA TGA TCG AGC CGC GC GCT ACC AGG GCC TTT GAG ATG GA

Statistical Analyses
All data except for RNA-Seq analyses was analyzed using GraphPad Prism 7 software (GraphPad Software Inc., San Diego, USA, RRID : SCR_002798). All graphs are shown as mean +/-standard deviation. Statistical significance was calculated by unpaired Student's t-test or two-way ANOVA with post-hoc Sidak's multiple comparisons test when having two categorical factors to compare. P values < 0.05 were considered statistically significant.

MB Is Found in the Cytoplasm of Breast Cancer Cells Under Both Normoxia and Hypoxia
Although it is known that MB is a cytoplasmic hemoprotein in cardiac myocytes and skeletal muscle fibers, the exact cellular distribution of MB in the more recently analyzed tumor cell lines has not been confirmed. As new studies emerged linking MB to oxidative stress and interactions with the transcription factor hypoxia-inducible factor-1a (13,31) in tumor cells, we analyzed the cellular distribution of MB in the high-MB breast cancer cell line MDA-MB-468 under both normoxia (NOX, 21% O 2 ) and hypoxia (HOX, 1% O 2 ). Subcellular fractionation of normoxic or hypoxia-raised MDA-MB-468 cells followed by Western Blot detection of MB and different cell compartment-specific marker proteins revealed that MB was consistently located in the cytoplasm. It was found neither in the nucleus, nor in the mitochondria, membrane or cytoskeleton fractions ( Figure 1). The same was seen for LNCaP cells, a human prostate adenocarcinoma cell line (Supplementary Figure 1). Although MB is thus highly unlikely to directly modulate gene regulation inside the nucleus, the globin might nevertheless be capable of modifying the activation of other transcription factors, which are transiently located in the cytoplasm.

MB Is Important for Cancer Cell Survival
As tumor cells in general and MB are both influenced by cellular oxygen levels, we checked the expression of MB under NOX compared to HOX conditions in the MDA-MB-468 monolayer cells. MB mRNA expression was significantly increased up to three times after 24 h of hypoxic incubation (Figure 2A). To further elucidate the effects of MB in hypoxia, we used transient RNA interference to generate MB knockdown (KD) cells. First, we looked for HIF-1a mRNA expression, which remained stable under both normoxia and hypoxia, independent of MB levels ( Figure 2B). One major characteristic of solid tumors is the formation of a central necrotic, hypoxic core, and an outer proliferating rim zone. We therefore investigated the influence of hypoxia and MB on this central necrosis by generating tumor spheroids, which mimic the tumor architecture in vitro. DAPI uptake revealed that the central core region of the spheroids showed a significant increase in dead cells after MB KD compared to WT spheroids, which was stable over 72 h of culturing. Under HOX conditions, this effect was still significantly visible after 24 h of incubation but diminished over 72 h of hypoxia ( Figures 2C, D). Although after 24 h of hypoxia the central necrosis was initially higher as compared to NOX conditions, prolonged hypoxia led to a significant reduction of the overall central necrosis for both MB WT and KD cells ( Figure 2D). Thus, MB significantly protected cells from dying especially under normoxic conditions.

MB Protects Cells From ROS
To further mimic tumor microenvironmental stress besides sole hypoxia, we treated the tumor spheroids with different substances inducing oxidative stress. NO is a free radical involved in many physiological and pathophysiological processes. It was also shown to interact with MB (38) and can regulate HIF-1a accumulation and activity (26,39 (Figures 3A, B). We furthermore measured ROS levels in tumor spheroids after addition of H 2 O 2 with help of a genetically encoded H 2 O 2 probe using live cell microscopy. As seen in Figure 3C, ROS levels were significantly increased in MB KD spheroids in contrast to WT spheroids. Therefore, MB protected cells from oxidative stress, possibly by its ROS scavenging capabilities.

MB Knockdown Elevates HIF-1a Stabilization After NO Treatment
As we found an association between MB and reduced cell death after NO exposure, and since NO can affect HIF-1a in different ways, we investigated the relationship between NO exposure, MB KD and HIF-1a levels in more detail. In normoxic MDA-MB-468 monolayer WT cells, NO addition did not elicit changes in HIF-1a mRNA levels ( Figure 4A), but a significant stabilization of HIF-1a protein 2 h and 4 h after addition, as shown by Western blot analysis. This transient effect was reduced to basal HIF-1a levels again after 6 h ( Figure 4B). Cells that had already been exposed to hypoxia for 24 h before treatment with NO kept their high HIF-1a protein level and showed no further enhancement through NO addition ( Figure 4B). After we saw a clear effect of NO on HIF-1a stabilization under normoxic incubation conditions, we next tested whether knockdown of MB would affect the action of NO. Western blot analysis revealed a transient HIF-1a accumulation for both WT and KD cells after NO treatment, which peaked at 3 h after addition ( Figure 4C). Thereby, KD cells showed a higher, but not significant, tendency in HIF-1a levels compared to WT cells that was also visible for cells treated with dimethyloxalylglycine (DMOG), a prolylhydroxylase-inhibitor that is commonly used as a positive control of HIF-1a stabilization ( Figure 4C). Next, we addressed the influence of MB KD on the degradation status of HIF-1a. By using the same experimental set-up as mentioned above, we investigated the status of hydroxylated HIF-1a with the help of a specific HIF-1a-OH-proline antibody in Western blot analysis. Whereas hydroxylated HIF-1a protein levels were transiently increased after addition of NO in WT cells, levels of HIF-1a-OH in KD cells remained constant over time and significantly lower compared to WT cells after 4 h of NO incubation ( Figure 4D). Conclusively, NO treatment led to HIF-1a stabilization that is elevated through a knockdown of MB accompanied by a reduced HIF-1a hydroxylation.

MB Knockdown Leads to the Induction of Hypoxia Signaling Genes After NO Treatment
To analyze, if the elevated HIF-1a stabilization has also effects on hypoxia target genes, we performed a RT² Profiler ™ PCR Array for Human Hypoxia Signaling Pathways (Qiagen) for WT and MB KD cells. Cells were treated with NO as stimulator for HIF-1a accumulation for 3 h as this time point showed the highest differences between both conditions. In KD cells, a majority of genes showed an increased expression when compared to WT cells ( Figure 5 and Supplementary Table 1). Thus, MB knockdown not only elevated HIF-1a protein stabilization after NO treatment, but consequently also induced the expression of HIF-1 dependent genes involved in hypoxic signaling.

MB Impacts on HIF-1 Target Genes Correlated to Cancer Survival After NO Treatment
Since HIF-1 is a transcription factor with several hundred target genes, we used RNA sequencing to identify altered mRNA expression patterns upon treatment with NO in combination with a MB knockdown in contrast to NO-treated WT cells. The differentially expressed genes (DEGs) were analyzed by Ingenuity pathway analysis (IPA) to gain insight into their molecular mechanisms through annotations of diseases, biological functions, and interaction networks. Figure 6A shows the significantly enriched annotations classified into three categories: molecular and cellular function, physiological system development and function, and disease and disorders. We further identified the interaction networks and found networks of cell death and survival, cardiovascular system development and function, and cancer that are relevant for breast cancer development. Specifically, we found 1029 DEGs of the cancer network that were altered by the combination of MB KD and NO treatment ( Figure 6B). In order to identify whether these DEGs of the cancer category were HIF-1 target genes, we checked 587 known and potential HIF-1a target genes (41). Overlap analysis revealed 45 predicted targets that were changed by the combination of MB KD and NO treatment, in which 11 genes were increased and 34 were decreased in expression. Subsequently, five genes were identified as genes of interest with a fold change >1.5 and a p-value <0.05, namely MKNK1, RAB8B, SLC2A1/GLUT-1, TMEFF1, and TREX2 ( Figure 6B). We validated the effects of MB knockdown in combination with NO treatment on the expression of these genes by qPCR. To see general effects of NO on gene regulation, we furthermore included WT and MB KD cells without NO treatment as controls. NO treatment alone showed a tendency for upregulation of HIF-1 target genes, especially seen for RAB8B ( Figure 6C). In addition, we found that all five genes were down regulated in MB KD cells compared to WT cells upon NO treatment, with SLC2A1/GLUT-1, MKNK1, and RAB8B being significantly affected ( Figure 6C). Overall, we saw changes of gene expression in fields important for cancer metabolism and outcome that were highly affected by MB knockdown and NO treatment.

DISCUSSION
Myoglobin is an oxygen-binding protein primarily located in the cytoplasm of striated muscles, namely cardiac myocytes and skeletal muscle fibers. Besides, MB expression was also found to a lower extent in different solid tumors and human cancer cell lines from colon, breast, kidney, lung, and prostate (9)(10)(11)(12)42). transport from the sarcolemma to the mitochondrial surface (1,2,43). MB can furthermore act as an NO dioxygenase or nitrite reductase, depending on the oxygenation status of the iron atom of the heme-group (3)(4)(5), or as a scavenger of ROS (6) to reduce oxidative stress in the cells.
The exact role of MB in cancer cells is not fully understood so far, although high expression of MB in tumors is correlated to a better patient prognosis (12). Early findings of Galluzzo et al. in a xenograft mouse model showed reduced hypoxia, minimal HIF-1a levels, and a homogeneously low vessel density in tumors after MB overexpression (44). In cell culture experiments, Kristiansen and colleagues found MB to be hypoxia-inducible and involved in controlling oxidative cell energy metabolism and mitochondrial activity (13). Additional findings showed an inhibition of breast cancer cell proliferation due to an induction of mitochondrial fusion (45), and a positive correlation of MB and HIF-1a signaling and effects on ROS and NO homeostasis in a transcriptomics approach (31).

HIF-1a Abundance Under MB Knockdown and NO Treatment
In our study, we therefore focused on the interplay between MB and HIF-1a and different oxidative-stress inducing agents in the high-MB breast cancer cell line MDA-MB-468. Consistent with other studies (13,14), MDA-MB-468 cells showed a strong induction of MB expression under 1% hypoxia ( Figure 2A). As the function of MB in carcinoma cells seems to be different from those in muscle cells and because of rising evidence of a functional correlation between MB and transcription factor HIF-1a, we verified the cellular localization of MB in breast carcinoma cells. We found MB to be exclusively expressed in the cytoplasm of MDA-MB-468 cells, independent of their oxygenation status (Figure 1), consistent with the current knowledge of MB as a cytoplasmic protein in muscle cells. Therefore, MB unlikely has a direct influence on the activity of the transcription factor itself. A possible interaction needs to take place in the cytoplasm. This fits to the specific regulation of HIF-1 stabilization on a post-translational level, where under normoxic conditions the alpha subunit is targeted for proteasomal degradation by hydroxylation through prolylhydroxylases in the cytoplasmic compartment. In line with these findings, HIF-1a mRNA expression did not show alterations between WT and MB KD cells under both NOX and HOX conditions ( Figure 2B).
NO is a molecule important for the regulation of many processes in inflammation and oxidative responses but is also known for its diverse effects in cancer. Depending on cancer type, concentration, and timing, NO can have either anti-tumor or carcinogenic effects. For breast cancer, NO seems to have a rather cancer-promoting role and the NO pathway is also interacting with steroid hormones in different ways (46). HIF-1a stabilization is not only achieved through hypoxic stimuli, but also through NO signaling as shown in different studies and for different cell lines (47)(48)(49)(50). It is therefore imaginable that tumorassociated NO modulates HIF-1a abundance in tumors. Moreover, opposite effects of HIF-1a stabilization were visible under hypoxia, where NO treatment led to a decrease in HIF-1a activity. Berchner-Pfannschmidt and colleagues proposed a bimodal system, where in the early phase, NO inhibits PHD activity that leads to HIF-1a accumulation, whereas in the late phase, increased PHD levels reduce HIF-1a through a negative feedback loop (39). When we studied HIF-1a stabilization after NO exposure in the MDA-MB-468 cells, we found an induction of HIF-1a protein levels which transiently peaked around 2 h to 3 h after GNSO addition ( Figure 4B) and returned to basal levels after 6 h. HIF-1a mRNA levels remained constant over time ( Figure 4A). In addition, when we cultured the cells under hypoxia, NO treatment led to no further stabilization of HIF-1a as HIF-1a levels were already very high, but also led to no degradation as observed for U2OS osteosarcoma cells ( Figure 4B) (39). Of note, though we applied a similar experimental protocol we used a different cancer cell line that may react differently or show different kinetics. Because oxygenated MB can scavenge NO, we compared HIF-1a levels between WT and MB KD cells after NO treatment. We found that the MB knockdown elevates the HIF-1a stabilization after NO treatment ( Figure 4C), accompanied by decreased hydroxylation of HIF-1a ( Figure 4D) which fits to the mode of action of NO as an inhibitor of prolyl hydroxylases. necrosis is an important hallmark of aggressiveUnder physiological conditions, MB therefore seems to metabolize NO that is then to some extent missing for inhibition of PHDs and an enhanced HIF-1a stabilization.

Expression of HIF-1a Target Genes Under MB Knockdown and NO Treatment
Not only abundance of HIF-1a was altered through exposure to NO and knockdown of MB but also target gene expression ( Figure 5). These findings show that MB is indeed capable of NO scavenging in breast carcinoma cells and can thereby influence HIF-1 activity. We therefore took a deeper look into altered gene expression levels via RNA-Seq analysis and searched for correlations between genes of the cancer category and potential HIF-1 target genes. We found that MB affects HIF-1 target genes correlated to cancer survival with up regulated but also down regulated examples ( Figures 6A, B). Surprisingly, when verified by qPCR, we found a down regulation of GLUT-1, MKNK-1, and RAB8B after MB KD ( Figure 6C). These genes are often up regulated in cancer and correlate with a poor clinical outcome (32)(33)(34). GLUT-1 is a very prominent HIF-1 target gene that is normally up regulated through HIF-1. As we saw an extended HIF-1 stabilization after MB KD and NO treatment, we would also have expected mostly up regulations of HIF-1 target genes, as seen in the Qiagen array ( Figure 5). However, this was not corroborated by RNA-Seq ( Figure 6C). This discrepancy needs to be further addressed.

Interplay Between MB Knockdown, Oxidative Stress and Cancer Cell Survival
Tumor necrosis is an important hallmark of aggressive cancers and associated with poorer patient prognosis in different types of cancer (51)(52)(53)(54). By using tumor spheroids, which mimic the 3D architecture of a tumor in vitro, we looked for the abundance of the central necrotic core with or without MB expression under both normoxia and 1% hypoxia. We found that MB is important for cell survival as MB knockdown led to a significant increase in the central necrotic area of the spheroids ( Figure 2D). This was especially true for normoxic conditions, but also visible under hypoxia. The central necrotic core and surrounding rim zone of a tumor are known for their lack of oxygen.  (55). When we treated the spheroids with irradiation, NO or H 2 O 2, the knockdown of MB led to greater central necrosis compared to WT cells ( Figure 4B). Overall necrosis was thereby diminished upon NO treatment but enhanced through H 2 O 2 .
In general, effects of NO on cell death depend on cell type, cellular redox status and NO dosage (56). In our tumor cell model, NO had a beneficial effect on cell survival. H 2 O 2 is frequently used in cell culture to mimic oxidative stress. Next to other ROS, like superoxide, H 2 O 2 is prone to induce cell death, for example through TNF-a signaling, in different cells (57)(58)(59), although its role in cancer development is ambivalent (60). In the MDA-MB-468 tumor spheroids, H 2 O 2 reduced cell survival. By using a H 2 O 2 sensor plasmid, we measured ROS levels in the spheroids after addition of H 2 O 2 and found a significant difference between WT and MB KD spheroids, with enhanced ROS levels after MB KD ( Figure 4C). MB is therefore not only able to scavenge NO but also H 2 O 2 in breast cancer cells, leading to beneficial effects on cell survival under oxidative stress. Nevertheless, high-MB breast tumors have a better prognostic outcome, which may be related to better responsiveness to chemo-or radiotherapy.

CONCLUSION
Our study is limited because we studied only one breast cancer cell line. Certainly, more studies are necessary to better understand the role of NB in breast cancer disease. In that respect, it is interesting that in patients with lung adenocarcinoma, high MB levels were linked to a poorer patient prognosis (10), indicating cancer-specific differences of MB function. Furthermore, HIF-1a is considered to have either pro-or anti-apoptotic effects depending on for example the severity of hypoxia. Besides activation of target genes inducing apoptosis itself, HIF-1a shows contrary interactions with the tumor suppressor protein p53 that is also linked to apoptosis (61,62).In their transcriptome study, Bicker and colleagues found GO-terms associated with negative as well as positive p53-mediated regulation of apoptosis in the MDA-MB-468 cells, linking MB to this mechanism (31). Off note, MDA-MB-468 cells exhibit a gain-offunction mutation of p53 with R273H that also be taken into account (63). Since our study contained experiments on necrosis and cell survival, it would be beneficial to address apoptosis further and the interplay of MB, HIF-1a and p53 in this process in the MDA-MB-468 breast cancer cells.
Taken together, our study is in agreement the work of Flonta and colleagues who proposed several possible mechanisms of MB function in tumor cells, that is as regulator of oxidative stress, O 2 transporter to overcome hypoxia, or by regulation of HIF-1a (42). Indeed, our study found evidence for all three assumptions in breast cancer cells, illustrating the diverse modes-of-action of MB in cancer metabolism and progression. Although precise mechanisms need to be further elucidated, our data indicate that myoglobin plays an important role in survival of breast cancer cells, possibly due to its ROS and NO scavenging properties and might be a valuable target for cancer therapy.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ebi.ac. uk/metagenomics/, PRJEB45683.

AUTHOR CONTRIBUTIONS
AH performed and analyzed the experiments, unless otherwise stated. TQ analyzed part of the data and wrote the manuscript. TN and ABi performed cellular localization experiments. ABi performed RNA-Seq experiments and analysis. ABe conducted the ROS measurements. TH analyzed part of the data and supported the project. JF supervised and supported the project. TS conceived the project and designed the experiments. All authors discussed the results and commented on the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
The study was supported by the Deutsche Forschungs gemeinschaft DFG Ha2103/10-1 (ABi and TH).