Hmox1 Upregulation Is a Mutual Marker in Human Tumor Cells Exposed to Physical Plasma-Derived Oxidants

Increasing numbers of cancer deaths worldwide demand for new treatment avenues. Cold physical plasma is a partially ionized gas expelling a variety of reactive oxygen and nitrogen species, which can be harnesses therapeutically. Plasmas and plasma-treated liquids have antitumor properties in vitro and in vivo. Yet, global response signatures to plasma treatment have not yet been identified. To this end, we screened eight human cancer cell lines to investigate effects of low-dose, tumor-static plasma-treated medium (PTM) on cellular activity, immune-modulatory properties, and transcriptional levels of 22 redox-related genes. With PTM, a moderate reduction of metabolic activity and modest modulation of chemokine/cytokine pattern and markers of immunogenic cell death was observed. Strikingly, the Nuclear factor (erythroid-derived 2)-like 2 (nrf2) target heme oxygenase 1 (hmox1) was upregulated in all cell lines 4 h post PTM-treatment. nrf2 was not changed, but its baseline expression inversely and significantly correlated with hmox1 expression after exposure to PTM. Besides awarding hmox1 a central role with plasma-derived oxidants, we present a transcriptional redox map of 22 targets and chemokine/cytokine secretion map of 13 targets across eight different human tumor cell lines of four tumor entities at baseline activity that are useful for future studies in this field.


Introduction
Cancer is the second leading cause of death worldwide [1]. Hence, global research efforts aim at better understanding cancer pathology and the discovery of new therapeutic avenues to combat disease. While there is a plethora of small molecules and biologicals being investigated for antitumor effects, research on a number of technological approaches supporting therapies or targeting cancer is carried as well. This includes studies using nanoparticles [2], photodynamic therapy [3], high hydrostatic pressure [4], acoustic treatment [5], hyperthermia [6], ionizing radiation combinational therapies [7], and cold-physical plasma-based approaches [8]. Cold physical plasma is a partially ionized gas that is generated at body temperature and composed of multiple effectors, such as electric fields, light radiation, ions and electrons, and reactive oxygen and nitrogen species (ROS/RNS) [9]. The later are hypothesized to be the main component responsible for antitumor effects observed with plasma treatment [10]. This includes the inactivation of tumor cells in vitro [11][12][13] and in vivo [14][15][16]. The use of plasma-treated liquids directly injected into tumors or used as lavage in peritoneal carcinomatosis 60 min. The plasma jet was operated with five standard liters per minute of argon (Air Liquide, Berlin, Germany), at a frequency of 1 Mhz, and a total power of less than 3.5 W in the handheld device. After plasma or argon gas treatment of the liquids, a pre-determined amount of double-distilled water was added to the liquids to compensate for evaporation. Media were then stored at −20 • C in aliquots prior to use within seven days. The feasibility of this approach has been documented previously [31][32][33]. For treatment of cells in 60 mm dishes, overnight culture medium was removed, cells were washed with PBS, and 5 mL of PTM or argon gas-treated medium were added to dishes before returning them to the incubator for another 4 h. For the treatment of cells in 96-well plates, overnight culture medium was removed, cells were washed with PBS, and 50 µL of PTM or argon-gas treated medium were added to each well (in quadruplicate per condition).

Live Cell Imaging
For live cell imaging, the 96-well plate was placed in a high content imaging device (Operetta CLS, PerkinElmer, Hamburg, Germany) equipped with a temperature module (37 • C and 5% CO 2 ). The outer rim of the Eppendorf 96-well plate protected the outer wells from excessive evaporation during the 4 h of time lapse imaging. Images were acquired with laser-based autofocus every 15 min with a 20× air objective (numerical aperture 0.4; Zeiss, Oberkochen, Germany) and a 4.7 megapixels scientific complementary metal-oxide-semiconductor camera (sCMOS). Image mode was digital phase contrast (DPC), a label-free visualization method of the cytosolic area of cells. A standardized de-focusing procedure and software algorithm generates contrast-rich cell areas. More than 20,000 images were acquired across all experiments. After flat-field correction, quantification of these images was carried out with Harmony 4.8 analysis software (PerkinElmer, Hamburg, Germany).

Multiplex Chemokine/Cytokine Analysis
Thirteen different immune-relevant targets were investigated using LegendPlex (BioLegend, London, UK) multi-analyte assay, a bead-based sandwich immuno assay. These beads differ in size and fluorescence intensity and thereby allow for the separate quantification of 13 targets in parallel by flow cytometry. The specific chemokines and cytokines quantified were arginase, CC-chemokine ligand (CLL) 17, C-X-C motif chemokine ligand (CXCL) 1, CXCL10, interferon (IFN) γ, interleukin (IL)1β, IL6, IL8, IL10, IL12, tumor growth factor (TGF) β, tumor necrosis factor (TNF) α, and vascular endothelial growth factor (VEGF). The supernatants were taken before harvesting tumor cells for quantitative polymerase chain reaction (qPCR) experiments, allowing for conclusions from one dataset to the other. The experimental procedure was performed in accordance to the supplier's instructions and the data from flow cytometry (using CytoFLEX S, Beckman-Coulter, Brea, CA, USA) were analyzed using LegendPlex 8.0 analysis software (VigeneTech, Carlisle, MA, USA). Absolute concentrations (in pg/mL) were calculated from an asymmetric sigmoidal model from each target's standard curve.

Abbreviation
Full Name Physiological Relevance

CD47
Cluster of differentiation 47 Serves as "don't eat" me signal by binding CD172a on myeloid cells to prevent phagocytosis of malignant and non-malignant cells [34] CRT Calreticulin ICD marker and serves as "eat-me" signal on non-malignant and malignant cells to promote phagocytosis by myeloid cells [35] HLA-ABC Human leukocyte antigen A, B, and C Major histocompatibility complex I (MHC class I), serves to present peptides of endogenous protein to patrolling CD8 + T-cells; important for antitumor immune responses [36] HSP70 Heat-shock protein 70 ICD marker, chaperone, and damage-associated molecular pattern (DAMP) [37] HSP90 Heat-shock protein 90 ICD marker, chaperone, and damage-associated molecular pattern (DAMP) [38] 2.6. Quantitative Polymerase Chain Reaction (qPCR) After 4 h of incubation with control or PTM, the cells were scratched off the dishes and transferred into 1.5 mL tubes (Eppendorf, Hamburg, Germany). After pelleting and suspending in lysis buffer, RNA isolation was performed according to the protocol of RNA isolation kit (RNA Mini Kit; Bio&SELL, Feucht, Germany). The RNA concentration of each sample was measured by using the NanoDrop 2000C (Thermo, Waltham, MA, USA) device was aliquoted into micro-tubes for further experiments. For quantitative polymerase chain reaction (qPCR), 1µg of RNA was synthesized into cDNA, according to the manufacturer's instructions (ThermoFisher, Waltham, MA, USA) using a thermocycler (Biometra, Goettingen, Germany). qPCR was performed in white 96-well V-bottom plates with Sybr Green (BioRad, Munich, Germany) labeled targets over 50-cycles using a Light Cycler 480 machine (Roche, Mannheim, Germany). Fold changes in expression was calculated using the 2 −∆∆Ct method, and normalized against glyceraldehyde 3-phosphate dehydrogenase (gapdh).

Statistical Analysis
Data are from 3-8 independent experiments performed with four technical replicates each. For normalization and calculation with raw data, Excel 2016 (Microsoft, Redmond, WA, USA) was utilized. Statistical analysis was performed with prism 7.05 (GraphPad software, San Diego, CA, USA). Bars show mean and standard error. Statistical comparison was performed either using paired t-test, or one-way or two-way analysis of variances (anova) to compare multiple groups. Levels of significance were indicated, as follows: α = 0.05 (*), α = 0.001 (**), α = 0.001 (***).

Plasma-Treated Medium Reduces the Metabolic Activity and Induces Swelling in Tumor Cells
The kINPen plasma jet (Figure 1a) was used to deposit reactive oxygen and nitrogen species (ROS/RNS) into cell culture medium (plasma-treated medium, PTM). Control or PTM was given to eight different human tumor cells having intrinsic variation in morphology and growth pattern (Figure 1b). After four hours of incubation, the metabolic activity was assessed. A moderate but consistent decreases of metabolic activity was observed among all eight cell lines investigated (Figure 1c). Antioxidants 2018, 7, x FOR PEER REVIEW 5 of 16 Next, we performed live cell time lapse imaging of all eight cell lines for up to 4 h following incubation either with control or PTM. For each well, the total cellular cytosolic area (as measure of accumulated cellular spread of all cells identified per field of view) was quantified to normalize to t = 0 for each sample to monitor changes of time. Except melanoma cells (Figure 2a,b), all cell types showed a significant increase in total cytosolic area (Figure 2c-h). As some these changes were sudden within the first hour after treatment, we attribute this behavior to cellular swelling that is often associated with cellular repair and cycle arrest, which may be linked to the findings of reduced metabolic activity (Figure 1c). This effect was particular pronounced in HT29, SW480, and MCF7, and Next, we performed live cell time lapse imaging of all eight cell lines for up to 4 h following incubation either with control or PTM. For each well, the total cellular cytosolic area (as measure of accumulated cellular spread of all cells identified per field of view) was quantified to normalize to t = 0 for each sample to monitor changes of time. Except melanoma cells (Figure 2a,b), all cell types showed a significant increase in total cytosolic area (Figure 2c-h). As some these changes were sudden within the first hour after treatment, we attribute this behavior to cellular swelling that is often associated with cellular repair and cycle arrest, which may be linked to the findings of reduced metabolic activity ( Figure 1c). This effect was particular pronounced in HT29, SW480, and MCF7, and rather moderate in pancreatic tumor cell lines Capan1 and Panc01. Changes in the osmolality of PTM can be excluded as not all cell types exposed to PTM changed (e.g., SKM and MNT1) and some cell types increased their sum cytosolic area also in control medium (e.g., Capan1 and MDA).
Antioxidants 2018, 7, x FOR PEER REVIEW 6 of 16 rather moderate in pancreatic tumor cell lines Capan1 and Panc01. Changes in the osmolality of PTM can be excluded as not all cell types exposed to PTM changed (e.g., SKM and MNT1) and some cell types increased their sum cytosolic area also in control medium (e.g., Capan1 and MDA).

Baseline and Regulated Transcriptional Level 22 Redox-Related Genes in Eight Human Cancer Cell Lines Following Exposure to Physical Plasma-Derived Oxidants
In further experiments, we sought to profile in all eight cancer cell lines the transcriptional level of 22 redox-related genes that may be associated with the perception of ROS/RNS released by physical plasma into cell culture medium. To gain a better understanding of the differences among all cell lines, a mini redox-map was generated to assess the baseline expression levels of all targets investigated via qPCR (Figure 3a). Using gapdh levels as cell-specific normalization control, endoplasmic Reticulum protein 27.7 kDa (erp27), erp4, nqo1, protein disulfide-isomerase/beta-subunit of prolyl 4-hydroxylase (p4hb), protein disulfide isomerase family A member 3 (pdia3), peroxiredoxin 5 (prdx5), and prdx6, as well as thioredoxin related transmembrane protein 1 (tmx1) and tmx3 showed

Baseline and Regulated Transcriptional Level 22 Redox-Related Genes in Eight Human Cancer Cell Lines Following Exposure to Physical Plasma-Derived Oxidants
In further experiments, we sought to profile in all eight cancer cell lines the transcriptional level of 22 redox-related genes that may be associated with the perception of ROS/RNS released by physical plasma into cell culture medium. To gain a better understanding of the differences among all cell lines, a mini redox-map was generated to assess the baseline expression levels of all targets investigated via qPCR (Figure 3a). Using gapdh levels as cell-specific normalization control, endoplasmic Reticulum protein 27.7 kDa (erp27), erp4, nqo1, protein disulfide-isomerase/beta-subunit of prolyl 4-hydroxylase (p4hb), protein disulfide isomerase family A member 3 (pdia3), peroxiredoxin 5 (prdx5), and prdx6, as well as thioredoxin related transmembrane protein 1 (tmx1) and tmx3 showed a medium to high expression in cells cultured in control medium. Transcripts of glrx1, nxn, and prdx4 were not detected in any cell line. Moreover, the expression of nrf2, nqo1, and hmox1 is shown in comparison to gapdh at higher resolution as these targets are known to be modulated upon oxidative stress (Figure 3b). A relatively low expression of nrf2 in all eight cell lines is observed, whereas nqo1 is mostly highly expressed. The biggest differences of expression between all cell lines investigated can be seen for hmox1.
Antioxidants 2018, 7, x FOR PEER REVIEW 7 of 16 a medium to high expression in cells cultured in control medium. Transcripts of glrx1, nxn, and prdx4 were not detected in any cell line. Moreover, the expression of nrf2, nqo1, and hmox1 is shown in comparison to gapdh at higher resolution as these targets are known to be modulated upon oxidative stress (Figure 3b). A relatively low expression of nrf2 in all eight cell lines is observed, whereas nqo1 is mostly highly expressed. The biggest differences of expression between all cell lines investigated can be seen for hmox1. Having investigated the baseline transcriptional expression levels of 22 redox-related proteins in eight human cancer cell lines, we investigated their up or downregulation following exposure to PTM at 4 h. We did not identify a consistent change with all except one target (hmox1) quantified across all cell lines used in this study (Figure 4a). A mixed response was observed for glrx1, prdx1, and prdx5. A more detailed analysis display of hmox1 data revealed a pronounced upregulation, particularly with HT29, SW480, and MCF7 cells (Figure 4b). As hmox1 is known to be a target of nrf2, correlation analysis was performed between baseline nfr2 expression levels and hmox1 expression levels induced with PTM. A significant inverse correlation was observed between both factors (Figure   erp27  erp44  glrx1  glrx3  gsr  hmox1  nqo1  nrf2  nxn  p4hb  pdia3  prdx1  prdx2  prdx3  prdx4  prdx5  prdx6  tmx1  tmx3  txn  txndc11  Having investigated the baseline transcriptional expression levels of 22 redox-related proteins in eight human cancer cell lines, we investigated their up or downregulation following exposure to PTM at 4 h. We did not identify a consistent change with all except one target (hmox1) quantified across all cell lines used in this study (Figure 4a). A mixed response was observed for glrx1, prdx1, and prdx5. A more detailed analysis display of hmox1 data revealed a pronounced upregulation, particularly with HT29, SW480, and MCF7 cells (Figure 4b). As hmox1 is known to be a target of nrf2, correlation analysis was performed between baseline nfr2 expression levels and hmox1 expression levels induced with PTM. A significant inverse correlation was observed between both factors (Figure 4c). Especially for the three cell lines with high upregulation of hmox1 following PTM exposure, a rather low nrf2 baseline expression was observed. Interestingly, through quantification of the DPC signal after challenging the cells with ROS/RNS, we found a correlation between the expression of hmox1 and the change in cytosolic area (Figure 4d). Hence, a high hmox1 expression is characteristic, together with an increased cell area after PTM treatment.
Antioxidants 2018, 7, x FOR PEER REVIEW 8 of 16 4c). Especially for the three cell lines with high upregulation of hmox1 following PTM exposure, a rather low nrf2 baseline expression was observed. Interestingly, through quantification of the DPC signal after challenging the cells with ROS/RNS, we found a correlation between the expression of hmox1 and the change in cytosolic area (Figure 4d). Hence, a high hmox1 expression is characteristic, together with an increased cell area after PTM treatment.

Modest Changes in the Immunomodulatory Profile in Response to PTM in Eight Human Cancer Cell Lines
Having analyzed the tumor-static effects of PTM with subsequent changes in the transcriptional profile of 22 redox-related proteins, we sought to investigate immunomodulatory effects of PTM treatment. An expression of hmox1 is known to be tightly associated with inflammation [39]. Essential for such immune reaction is the secretion of inflammatory mediators. Therefore, we determined such soluble mediators in the cells supernatant and found that PTM induced a complex secretion profile of chemokines and cytokines in the different cell lines (Figure 5a-m). A significantly higher secretion of the pro-inflammatory mediators CXCL10 (in Capan1), arginase (in MDA), and TNFα (in HT29) was found, as well as a higher secretion of the anti-inflammatory interleukin 10 (in HT29) and the depletion of IL8 (in SKM) and arginase (in SKM, Panc01, HT29, and MDA). However, to evaluate the immunogenic effect of the secretion of different factors a more complex model, including different cell types is needed and our observations suggest a mixed response.

Modest Changes in the Immunomodulatory Profile in Response to PTM in Eight Human Cancer Cell Lines
Having analyzed the tumor-static effects of PTM with subsequent changes in the transcriptional profile of 22 redox-related proteins, we sought to investigate immunomodulatory effects of PTM treatment. An expression of hmox1 is known to be tightly associated with inflammation [39]. Essential for such immune reaction is the secretion of inflammatory mediators. Therefore, we determined such soluble mediators in the cells supernatant and found that PTM induced a complex secretion profile of chemokines and cytokines in the different cell lines (Figure 5a-m). A significantly higher secretion of the pro-inflammatory mediators CXCL10 (in Capan1), arginase (in MDA), and TNFα (in HT29) was found, as well as a higher secretion of the anti-inflammatory interleukin 10 (in HT29) and the depletion of IL8 (in SKM) and arginase (in SKM, Panc01, HT29, and MDA). However, to evaluate the immunogenic effect of the secretion of different factors a more complex model, including different cell types is needed and our observations suggest a mixed response.  Moreover, previous studies indicated that tumor-toxic concentrations of plasma-derived ROS/RNS are capable of inducing the immunogenic cancer cell death (ICD) [40]. This type of cell death promotes antitumor immune responses, and is characterized by release upregulation of a number of molecules on the cell membrane, such as calreticulin (CRT), major histocompatibility complex I (MHC class I; HLA-ABC), and heat-shock proteins (HSP) 70 and 90 [41]. This may be accompanied by a downregulation of CD47, which promotes the phagocytosis of tumor cells [34]. However, only SK-Mel 28 (SKM) cells showed a considerable increase of HSP90 on the membrane of living cells (Figure 6a). For all other surface molecules and cell lines, there were only minor changes that were observed in response to PTM at 4 h when normalized each respective control (Figure 6b). This may be due to the overall mild effect (short plasma treatment time) of the PTM.
lines indicate the target-specific limit of detection (LOD). Data are pooled from eight independent experiments. Red bars = plasma-treated medium (PTM); grey bars = untreated medium (control).
Moreover, previous studies indicated that tumor-toxic concentrations of plasma-derived ROS/RNS are capable of inducing the immunogenic cancer cell death (ICD) [40]. This type of cell death promotes antitumor immune responses, and is characterized by release upregulation of a number of molecules on the cell membrane, such as calreticulin (CRT), major histocompatibility complex I (MHC class I; HLA-ABC), and heat-shock proteins (HSP) 70 and 90 [41]. This may be accompanied by a downregulation of CD47, which promotes the phagocytosis of tumor cells [34]. However, only SK-Mel 28 (SKM) cells showed a considerable increase of HSP90 on the membrane of living cells (Figure 6a). For all other surface molecules and cell lines, there were only minor changes that were observed in response to PTM at 4 h when normalized each respective control (Figure 6b). This may be due to the overall mild effect (short plasma treatment time) of the PTM.

Discussion
In need of a common marker delineating the response of cancer cells to plasma-treated medium (PTM), we studied of growth behavior, immunogenic features, and the expression of 22 redox-related transcripts in eight human cancer cell lines. Besides presenting a small redox map in all cell types with regard to baseline redox-protein transcript expression and chemokine/cytokine profile, hmox1 was identified as key response element in all cancer cell lines following exposure to PTM.
Angiogenesis is a key hallmark of cancer [42]. New blood vessel help support tumor growth by removing metabolic waste and carbon dioxide, and increasing nutrient supply. Induced by VEGF and nrf2 as well as others, heme oxygenase 1 (HO1, hmox1) derived carbon monoxide (CO) possess not only proangiogenic effects, but also anti-inflammatory, antioxidant, and anti-apoptotic properties [43]. HO1 moreover detoxifies heme into ferrous iron, as the former would otherwise contribute to the cytotoxic fenton reaction generating hydroxyl radicals [44]. HO1 has therefore implicated as major target in cancer therapy although its expression is linked to both tumor progression and regress [45]. Oxidative stress is a known inducer of HO1 in human cells [46,47], underlining reactive oxygen and nitrogen species being a major bio-active component in physical plasma-treated liquids. Indeed, both human keratinocytes [48] and the human cell line THP-1 monocytes [49] upregulate hmox1 in response to plasma-derived ROS/RNS.
While the identification of a common hmox1 signature in response to PTM may be useful for further research, the consequences on tumor metastasis in vivo are controversial. In breast cancer cells, we found a strong increase in hmox1 levels after PTM exposure, and such an upregulation was shown to be associated with a significant reduction of invasive properties in MCF7 and MDA-MB-231 cells [50]. Similar observations were made with colon cancer cells [51,52]. As a mechanism, the inhibition of matrix-metalloproteinase (MMPs) through HO1-derived CO was proposed, as MMPs facilitate the degradation of extracellular matrix needed for metastatic spread. However, several reports argue for a role of HO1 in promoting metastatic spread in cancer [53][54][55]. This is linked especially to its constitutive upregulation in many types of tumors as well as studies showing that hmox1 downregulation is associated with better outcome. Specifically, among non-responders to chemotherapy in a cohort of head and neck cancer patients, hmox1 was the most upregulated gene that is identified in a transcriptomic microarray study as compared to complete responders [56].
Hmox1 expression is tightly linked to regulation of immunity and inflammation [39]. Investigating a panel of well-recognized markers for immunogenic cell death (ICD) that are involved in immunological recognition and removal of tumor cells [41,57,58], we identified only heat-shock protein 90 (HSP90) to be upregulated in SK-Mel 28 cells. However, the profound upregulation of HSP90 on SK-Mel 28 cell membrane in response to low-dose oxidative stress may be a valuable research target for tumor immunology. HSP90 is a damage-associated molecular pattern (DAMP) that can facilitate the uptake of tumor cells by dendritic cells to induce antitumor immune responses [59]. Release of DAMPs and enhanced expression of ICD-relevant markers has been shown for plasma-derived ROS/RNS in various type of tumor cells [60][61][62]. Yet, our PTM was rather tumor-static than tumor-toxic, possibly contributing to the lack modulation with all except one marker in almost all cell lines. With respect to the three cell lines (HT29, SW480, MCF7) with the strongest increase in hmox1 levels, a number of chemokines/cytokines/growth factors were all increased in tendency in response to PTM, namely CCL17, IFNγ, IL1β, IL6, IL8, IL10, IL12, TGFβ, TNFα, and VEGF. A correlation between increase of hmox1 and IL8 as well as VEGF was reported before being major angiogenic drivers in tissues [63]. In a murine carcinogenesis model, IL1β models for only non-significantly increased with a significant increase in hmox1 [64], corroborating our findings. HO1 is also described to increase IL10 production [65], as seen in our subset of cell with high hmox1 expression. In general, this subsets had a consistent increase in both pro-inflammatory (CCL17, CCL17, IFNγ, IL1β, IL6, IL12, and TNFα) and anti-inflammatory (IL8, IL10, TGFβ, and VEGF) cytokines, underlining the complex effects that HO1 has on inflammation.
The effect-inducing agent in our study was cell culture medium exposed to cold physical plasma, a partially ionized gas operated at room temperature. A number of reactive oxygen and nitrogen species were so far identified in plasma-treated liquids [66], among them e.g., superoxide anion, singlet oxygen, nitrite, nitrate, peroxynitrite, hydrogen peroxide, hydroxyl radical, nitric oxide, ozone, and atomic oxygen [67][68][69][70][71]. However, most of these species quickly deteriorate to some major long-lived oxidants, which are likely to mediate the main effect that was observed with PTM. Only recently, reports dealt with post-translational modifications on biomolecules and proteins that are present in liquid media [72], which may be recognized by cells and their oxidation-sensitive receptor repertoire. Oxidative stress is sensed through the nrf2/Keap1 pathway [73], which leads to the activation of antioxidative responses elements (ARE), including transcriptional targets, such as vegf, hmox1, bach1, fos, junb/d, maf, and others [74]. While we here identified a consistent upregulation of hmox1 in eight tumor cell types following exposure to PTM, no changes in nrf2 levels were observed. This is consistent with a previous report in human keratinocytes were nrf2 mRNA levels peaked at 20 min following exposure to PTM, whereas that of hmox1 peaked at 3 h [75].

Conclusions
The main aim of this study was to identify a common redox-related transcriptional signature in eight human cancer lines following exposure to physical plasma-treated medium (PTM). We not only identified transcriptional levels of hmox1 as mutual response element but also provided useful tools, such as a 22 target transcriptional map and cytokine/chemokine patterns in all cell lines to further study the role of other redox proteins that signal with plasma-derived ROS/RNS.