Cross Talk in HEK293 Cells Between Nrf2, HIF, and NF-κB Activities upon Challenges with Redox Therapeutics Characterized with Single-Cell Resolution

Abstract

Aim: Many transcription factors with importance in health and disease are redox regulated. However, how their activities may be intertwined in responses to redox-perturbing stimuli is poorly understood. To enable in-depth characterization of this aspect, we here developed a methodology for simultaneous determination of nuclear factor E2-related factor 2 (Nrf2), hypoxia-inducible factor (HIF), and nuclear factor kappa-light-chain-enhancer of activated B cell (NF-κB) activation at single-cell resolution, using a new tool named pTRAF (plasmid for transcription factor reporter activation based upon fluorescence). The pTRAF allowed determination of Nrf2, HIF, and NF-κB activities in a high-resolution and high-throughput manner, and we here assessed how redox therapeutics affected the activities of these transcription factors in human embryonic kidney cells (HEK293).

Results: Cross talk was detected between the three signaling pathways upon some types of redox therapeutics, also by using inducers typically considered specific for Nrf2, such as sulforaphane or auranofin, hypoxia for HIF activation, or tumor necrosis factor alpha (TNFα) for NF-κB stimulation. Doxorubicin, at low nontoxic doses, potentiated TNFα-induced activation of NF-κB and HIF, without effects in stand-alone treatment. Stochastic activation patterns in cell cultures were also considerable upon challenges with several redox stimuli.

Innovation: A novel strategy was here used to study simultaneous activation of Nrf2, HIF, and NF-κB in single cells. The method can also be adapted for studies of other transcription factors.

Conclusion: The pTRAF provides new opportunities for in-depth studies of transcription factor activities. In this study, we found that upon challenges of cells with several redox-perturbing conditions, Nrf2, HIF, and NF-κB are uniquely responsive to separate stimuli, but can also display marked cross talk to each other within single cells. Antioxid. Redox Signal. 26, 229–246.

Introduction

Several redox-regulated transcription factors are highly important in both health and disease (3, 4, 28, 42). However, complex, fast, and transient reactions limit current knowledge on how redox-responsive transcription factors may be regulated with regard to each other within the same cellular context, especially upon challenges to cells by redox-perturbing conditions. To enable direct analyses of three distinct transcription factor activities at the same time and in the same cellular context, we here developed a novel fluorescence-based tool named pTRAF (plasmid for transcription factor reporter activation based upon fluorescence), enabling simultaneous analyses of the transcriptional activation from three separate promoter response elements at single-cell resolution.

Innovation.

A new vector called pTRAF (plasmid for transcription factor reporter activation based upon fluorescence) was here shown to enable simultaneous assessment of three different transcription factor activities at the same time, with single-cell resolution and in a high-throughput manner. Using this tool, novel insights into the effects of diverse redox-perturbing conditions on nuclear factor E2-related factor 2, hypoxia-inducible factor, and nuclear factor kappa-light-chain-enhancer of activated B cell regulation were gained at a level of detail that is difficult to obtain using other conventional experimental methodologies.

Herein, we used the pTRAF tool to study activation patterns of nuclear factor E2-related factor 2 (Nrf2), hypoxia-inducible factor (HIF), and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB). These three transcription factors, in particular, are affected by the cellular redox state and are furthermore involved in several diseases, including pathogenesis of cancer, degenerative diseases, inflammatory responses, and ischemia–reperfusion injuries (3, 4, 28, 42). Most cancer forms have constant activation of at least one of these transcription factors correlating to aggressiveness of tumors, drug resistance, and propensity to form metastases (21, 38). Insights into how their responses are coordinated with regard to each other upon different cellular redox stimuli are, nonetheless, rudimentary at best (4, 12, 46).

Nrf2 regulates an array of genes coding for proteins involved in the protection against oxidative and electrophilic stress, as well as metabolic flow (34). In normal cell homeostasis, Nrf2 is constantly ubiquitinated by the E3 ligase Keap-1 and degraded. Upon oxidative or electrophilic stress, Nrf2 is stabilized and translocated to the nucleus, promoting the expression of downstream Nrf2 target genes by binding to the antioxidant/electrophile response element (ARE/EpRE) in target gene promoters (34), including important proteins of the glutathione (GSH) and thioredoxin (Trx) systems (1, 4, 19).

HIF induces genes required for adaptation to hypoxia and is constituted by heterodimers of HIFα and HIFβ subunits (16). In normoxia, HIFα is constantly degraded by the proteasomal system, while upon hypoxia, HIFα becomes stabilized and translocates to the nucleus, where HIF (dimer of HIFα- and HIFβ- subunits) regulates the expression of a line of genes involved in angiogenesis, glucose transport, glycolysis, survival, and cell migration (16, 39).

The transcription factor NF-κB, finally, is a protein complex involved in inflammation and cellular responses to different stresses, including exposure to bacterial or viral antigens (4). Under normal conditions, NF-κB is located in the cytosol bound to its inhibitor, IκB. Upon activation, IκB is phosphorylated and releases NF-κB that thereby translocates to the nucleus to induce its target genes (4, 13, 18, 20). Using the pTRAF methodology, we could here study how different redox-perturbing conditions affect these three transcription factors concomitantly within the very same cellular contexts.

Redox regulatory events are important in the control of cellular responses to diverse stimuli (18). In the cytosol, reductive events typically prevent stabilization and activation of Nrf2, HIFα, and NF-κB (4, 19), while in the nucleus, the reducing Trx system facilitates binding of the transcription factors to target DNA motifs (32, 33). Such different levels of redox regulation may naturally affect several transcription factors in parallel, and because Nrf2 in turn activates several antioxidant redox pathways in cells, the possibilities of complex interplay between these signaling pathways have been implicated (5, 15, 29, 44). We therefore propose that it is important to study these activities simultaneously in the same cells upon use of multiple cellular stimuli, including such that perturb the cellular redox balance, which is the basis for this study.

Results

Construction and validation of the pTRAF^{Nrf2/HIF/NF-κB} vector

To construct the pTRAF tool, we selected previously well-characterized promoter response elements for studies of the three transcription factor activities characterized here (see the Materials and Methods section for details). These response elements were first functionally validated with regular luciferase reporter-based assays in human embryonic kidney cells 293 (HEK293), using cellular stimuli known to be potent inducers of Nrf2, HIF, or NF-κB, respectively (Fig. 1A).

FIG. 1.

Construction and validation of the pTRAF^{Nrf2/HIF/NF-κB} vector. (A) To validate the response patterns of the chosen response elements for analyses of Nrf2, HIF, and NF-κB activation, the commercially available dual-luciferase assay was used. For Nrf2 activity (left), a vector (backbone pGL4.32) was made using antioxidant response elements (ARE; pGL4.Luc^Nrf2), and for HIF activity (middle), a vector with HREs (backbone pGL4.32) was used. For NF-κB activity, the reporter vector pGL4.32 was used. HEK293 cells were transfected for 24 h with any one of these vectors in combination with pGL4.73, a vector that constitutively expresses the luciferase reporter hRluc as a control for transfection efficiency. For validation of the Nrf2 response element (left graphs), cells were transfected with pGL4.Luc^Nrf2 and then treated with 80 μM BHQ for 24 h to analyze the Nrf2 response (significant induction p ≤ 0.05 compared with controls). An HIF response was analyzed in cells transfected with pGL4.Luc^HIF and grown in either 21% O₂ or in 3% O₂ for 24 h (middle graphs; p ≤ 0.05). To study the NF-κB response, cells were transfected with the pGL4.32 vector and treated with 20 ng/ml TNFα for 5 h (right graphs; p ≤ 0.05). Promoter responses are displayed as ratios between the luminescent signals of luc2P and hRluc in the top panels, as well as showing absolute values in RLUs (relative light units, bottom panels), representing mean ± SEM of triplicates from one representative experiment. (B) The fluorescent excitation and emission spectra for the three fluorescent proteins used in this study are shown, mCherry (red fluorescent protein), YPet (yellow), and cyan (CFP) (the fluorescent spectra were adapted from www.tsienlab.ucsd.edu). The right panels list the different filter sets used in our experiments to monitor fluorescence signals with the Operetta^®, Axiovert^®, or fluorescence-activated cell sorting equipment (BP stands for band pass filter and FT for Beam splitter). For further details, see the Materials and Methods section. (C) Detection, validation, and quantification of the three fluorescent proteins constitutively expressed in HEK293 cells. Fluorescence microscopy pictures of HEK293 cells transiently transfected with stoichiometric 1:1 mixtures of two plasmids constitutively expressing single fluorescent proteins under control of SV40 promoters (as indicated), recorded using the Operetta high-content imaging system. Upper panels show overlay pictures (Merge), while lower panels show the individual channels, as indicated. Scale bar in the pictures illustrates 100 μm. Quantification of total fluorescence intensities of two independent experiments performed in triplicates is shown as bar graphs. ARE, antioxidant response element; BHQ, butylhydroquinone; HEK293, human embryonic kidney cells 293; HIF, hypoxia-inducible factor; HRE, hypoxia response element; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; pTRAF, plasmid for transcription factor reporter activation based upon fluorescence; TNFα, tumor necrosis factor alpha; YPet, yellow fluorescent protein for energy transfer. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

In parallel, we selected three different fluorescent proteins (mCherry, YPet, and CFP) having as separate spectra as possible (Fig. 1B) to allow independent detection and quantification within the same cells (40). Distinct detection of these fluorescent proteins was first validated using combinations of plasmids guiding their constitutive expression in the different pairwise combinations (Fig. 1C). With none of the fluorescent proteins affecting the quantification of the other two, the pTRAF^{Nrf2/HIF/NF-κB} plasmid was subsequently constructed (see the Materials and Methods section for nomenclature) having the response elements for Nrf2, HIF, and NF-κB guiding expression of the red (mCherry), yellow (YPet), and cyan (CFP) fluorescent proteins, respectively (see Fig. 2 for the different vectors used in this study, and Supplementary Table S1A for construction information; Supplementary Data are available online at www.liebertpub.com/ars). Because transcriptional responses activated by HIF are typically several-fold lower in absolute amplitude compared with NF-κB activation (Fig. 1A), we chose the brightest fluorescent protein (YPet) as the reporter for HIF activation and the least bright fluorescent protein (CFP) for NF-κB, while Nrf2 responses were assessed with the intermediately fluorescent mCherry protein.

FIG. 2.

Scheme of plasmids constructed and utilized in this project. This scheme illustrates the features of plasmids constructed in this study, with important restriction sites and functional elements indicated. Details on the cloning procedure are described in the Materials and Methods section and primers are summarized in Supplementary Table S1A. The original backbone of each plasmid is given in parentheses under the schematic position of the origin of replication (Ori), and resistance genes are displayed as gray rectangles. Transcriptional directions are shown with arrows and unique restriction sites are indicated by arrowheads. Names of plasmids are given to the left of each panel, followed by their response elements (R.E.) and sizes in base pairs (bp) indicated in parentheses. Response elements and reporter genes are shown as filled rectangles, with the corresponding identities written above each rectangle. (A) The source and sequences of the different response elements for Nrf2, HIF, and NF-κB used in this study. (B) Vectors used to validate the activities of the chosen response elements via transcription of the luciferase gene luc2P (for results, see Fig. 1A). Plasmids shown in (C) were made as an intermediate cloning step for construction of the final pTRAF vectors using the vectors listed in (B) as templates. The luc2P gene was exchanged for mCherry, while NdeI and EcoRI restriction sites were introduced concurrently to facilitate subsequent exchanges of the mCherry gene for either YPet or CFP as displayed in (D). Additionally, the 4xHRE sequence was exchanged for an 8xHRE sequence to increase responses to HIF activation in the fluorescence-based setup. These vectors carry different response element–fluorescent protein combinations of Nrf2-mCherry, HIF-YPet, and NF-κB-CFP and were used for assessments in the further development of the final vectors (results not shown). (E) Control vectors with any of the three fluorescent proteins transcribed from SV40 promoters are shown. These were utilized for verification and calibration in all experiments referred to in this study, with data displayed in Figure 1C and Supplementary Fig. S2A and B. (F) The principal configuration of the final pTRAF^{Nrf2/HIF/NF-κB} plasmid as it was utilized for the experiments in this study. The response element–fluorescent protein combinations of Nrf2-mCherry, HIF-YPet, and NF-κB-CFP are flanked by the indicated unique restriction sites, which can be utilized to freely exchange the individual cassettes for future applications. (G) The pTRAF^3xSV40 control vector is shown, which is similar in size and structure as pTRAF^{Nrf2/HIF/NF-κB}, but has the transcription of the fluorescent proteins driven constitutively by SV40 promoters. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

For further validation of the pTRAF tool, we quantified transcripts of known downstream target genes for Nrf2, HIF, or NF-κB using quantitative reverse transcription-PCR (qRT-PCR) (Fig. 3A, and see Supplementary Table S1B for primer sequences). This analysis confirmed that under our treatment conditions, specific target genes were indeed induced as expected, that is, for Nrf2 (induced with 3 μM sulforaphane, a known Nrf2 inducer), we detected upregulation of hemoxygenase-1 (HMOX1), glutamate–cysteine ligase, modifier subunit (GCLM), and thioredoxin reductase 1 (TXNRD); for HIF (activated by culturing in 1% O₂), vascular endothelial growth factor A (VEGFA) was induced; and for NF-κB activation using tumor necrosis factor alpha (TNFα) treatment, interleukin-8 (IL8) was induced. The mRNA levels encoding transcription factor components of Nrf2 (NFE2L2), HIF-1α (HIF1A), and NF-κB (NFKB1) were also monitored and showed that both NFKB1 and NFE2L2 were induced by TNFα, while the mRNA levels of HIF1A were not (Fig. 3A, see statistics in Supplementary Table S1C). Furthermore, to validate that transfection of the pTRAF^{Nrf2/HIF/NF-κB} plasmid into HEK293 cells did not alter the expected endogenous cellular responses, these transcript levels were compared with those triggered in nontransfected cells treated in the same manner, showing that the expected patterns of responses were seen in both cases (see Fig. 3B for the correlation plot and Supplementary Table S2 for ΔC_t values).

FIG. 3.

mRNA expression of Nrf2, HIF, and NF-κB and their downstream target genes. Quantitative reverse transcription-PCR (qRT-PCR) was used to study the mRNA levels of (A) Nrf2 (NFE2L2) and its downstream target genes (red bars), for example, hemoxygenase 1 (HO1), glutamate–cysteine ligase regulatory subunit (GCLM), and TrxR1 (TXRD1). HIF-1α (HIF1A) and its downstream target gene, Vascular endothelial growth factor (VEGFA) (yellow bars), as well as NF-κB (NFKB1) and its downstream target gene, Interleukin-8 (IL8) (cyan bars). HEK293 cells were exposed to either single treatment or combinatory treatment of 20 ng/ml TNFα and 3 μM sulforaphane for 6 h at 21% or 1% O₂ (hypoxia), respectively. The mRNA expression is displayed as transcript levels of each gene normalized with the housekeeping genes, β-actin (ACTB) and hypoxanthine phosphoribosyltransferase 1 (HPRT1), the data are based on triplicates from one representative experiment. (B) The figure shows qRT-PCR data of downstream target genes for Nrf2, HIF, and NF-κB as a correlation plot between untransfected cells and cells transfected with pTRAF^{Nrf2/HIF/NF-κB}, confirming that the responses of downstream target genes were not significantly altered during transfection with pTRAF^{Nrf2/HIF/NF-κB} (see Supplementary Table S2 for the raw ΔC_t values plotted in this graph). Transfected or untransfected HEK293 cells were exposed to 1 μM auranofin (Nrf2 inducer) or 20 ng/ml TNFα (induces NF-κB) in either 21% O₂ or 1% O₂ (for induction of HIF) for 1, 3, or 6 h to determine downstream target gene induction using qRT-PCR (see Supplementary Table S1B for the selection of target genes and primers and the Materials and Methods section for further details). TrxR1, thioredoxin reductase 1. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

Evaluation of the pTRAF tool using fluorescence microscopy

Transfection of HEK293 cells with the pTRAF^{Nrf2/HIF/NF-κB} reporter vector, with the three response elements introduced into the same vector backbone (Fig. 4A), enabled simultaneous fluorescence microscopy-based detection of the activation of each of the three transcription factors in all transfected cells. The responses were as expected when cells transfected with this plasmid were exposed for 24 h to known inducers of any of the three transcription factors, that is, using 1 μM auranofin for Nrf2 activation (30), 20 ng/ml TNFα for NF-κB (43), or culturing the cells in hypoxia (1% O₂) for induction of HIF (45). Hence, classic Nrf2 induction gave mainly red fluorescent cells, NF-κB induction blue (cyan) cells, and hypoxia resulted in yellow cells, while combinations of these treatments gave combined responses (Fig. 4B). Furthermore, as HEK293 cells have low endogenous activities of Nrf2, HIF, and NF-κB, the nontreated, but still transfected, cells displayed very little fluorescence, illustrating that spontaneous promoter leakage with expression of any of the reporter proteins was small. These results collectively showed that the pTRAF^{Nrf2/HIF/NF-κB} tool fulfilled its intended functionality to enable simultaneous detection of Nrf2, HIF, and NF-κB activation with single-cell resolution. It must, however, be considered that different cell types are likely to display different response patterns with regard to transcriptional activation, which will be reflected through unique readout patterns using the pTRAF tool. As an example, we found that HCT116 colon carcinoma and A375 malignant melanoma cells displayed higher basal activities in untreated controls compared with HEK293 and unique cell type-specific profiles in response to auranofin or doxorubicin treatment (Supplementary Fig. S1). Fully unraveling or characterizing cell type-specific responses was judged to be outside the scope of the present study and thus we here report further results of using the pTRAF tool only with HEK293 cells.

FIG. 4.

The pTRAF^{Nrf2/HIF/NF-κB} plasmid enables simultaneous detection at single-cell resolution of Nrf2, HIF, and NF-κB activation patterns. (A) A schematic picture of the pTRAF^{Nrf2/HIF/NF-κB} plasmid is shown, with response elements for Nrf2, HIF, and NF-κB (boxed, as labeled) made to guide expression of the three fluorescent proteins, mCherry (red), YPet (yellow), and CFP (cyan), respectively. Restriction sites and principles of the reporter cassettes are shown to the right. See Figure 2 for more detailed information. (B) The upper panels show overlay fluorescence microscopy pictures (Merge) of HEK293 cells transiently transfected with pTRAF^{Nrf2/HIF/NF-κB} upon exposure for 24 h to either 1 μM auranofin or 20 ng/ml TNFα, with Control being transfected untreated cells grown in either 21% O₂ or low oxygen tension (1% O₂), as indicated. The lower panels for either 21% O₂ or 1% O₂ show the separate fluorescence channels of the merged picture. Scale bars in the pictures are 50 μm. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

Use of the pTRAF methodology in high-throughput settings

Aiming to utilize the pTRAF reporter in a high-throughput compatible setup, we next adopted its use to the Operetta^® high-content imaging system. The activation patterns of Nrf2, HIF, and NF-κB could thereby be measured in a 96-well format, with cells challenged for 24 h using a selection of redox-perturbing drugs used at low nontoxic concentrations (see experimental setup in Fig. 5A). The HEK293 cells displayed highly reproducible responses between replicate experiments (compare triplicates, lanes 1–4, 5–8, and 9–12, respectively, in Fig. 5B). Concurrent determinations of cell numbers were done using Hoechst staining of the nuclei (Fig. 5C). A straightforward quantification of Nrf2, HIF, and NF-κB responses was obtained by integrating the total fluorescence signals for each treatment with correlation to the number of cells in each well (Fig. 6). Indeed, these results agreed well at the accumulated cell culture level with those obtained using regular reporter systems such as luciferase-based reporter assays (cf. Figs. 1A and 6), with the difference that use of the pTRAF tool allowed for assessment of three separate promoter response element activities in the same experiment.

FIG. 5.

Assessment of Nrf2, HIF, and NF-κB responses using the Operetta high-content imaging system. Studies of the Nrf2, HIF, and NF-κB responses to an array of different drug treatments in either 21% O₂ or 1% O₂, as detected using an Operetta high-content imaging system in a 96-well format. HEK293 cells were seeded, transfected with pTRAF^{Nrf2/HIF/NF-κB} or control vectors, and subsequently treated with different compounds according to the experimental setup shown in the top panel (A). The activation of Nrf2, HIF, and NF-κB responses in single cells was subsequently detected and quantified with the Operetta high-content imaging system. (B) One example of an overlay picture for a readout [see experimental setup in (A) using the Operetta system in 96-well format, with the reproducibility between triplicates being visualized (lanes 1–4, 5–8, and 9–12, respectively]. The viability shown in (C) was assessed by counting the number of nuclei stained with Hoechst in living cells after treatment according to (A). The bars represent mean ± SEM of three separate experiments, with each treatment measured thrice per plate (n = 9). To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

FIG. 6.

Drug-specific induction patterns of Nrf2, HIF, and NF-κB in HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB} analyzed with the Operetta high-content imaging system. Total cellular responses of Nrf2, HIF, and NF-κB analyzed with the Operetta high-content imaging system. HEK293 cells were transfected with pTRAF^{Nrf2/HIF/NF-κB} and the total cellular Nrf2, HIF, and NF-κB responses to a 24-h treatment of indicated drugs in either 21% O₂ (left, solid bars) or 1% O₂ (right, shaded bars) are here represented as the total fluorescence signal for Nrf2 (red), HIF (yellow), and NF-κB (cyan) divided by the total number of cells in each experiment, as summarized in Figure 5C. These data are based on the experimental setup and concentrations described in Figure 5A and represent mean ± SEM from three separate experiments, each made in triplicates (n = 9), see statistics in Supplementary Table S1D. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

A number of different previously characterized Nrf2 inducers were tested, including tert-butylhydroquinone (BHQ), auranofin, and sulforaphane. At low subtoxic concentrations (Fig. 5C), we found that auranofin was the most potent Nrf2 inducer. Furthermore, we found that cobalt, commonly used to induce HIF activation at 21% O₂ through stabilization of HIF-1α (47), also significantly induced Nrf2 at low concentrations. Finally, as a strong activator of NF-κB, we used TNFα and we found that a combination of TNFα with redox-perturbing drugs often influenced the responses of Nrf2, HIF, or NF-κB compared with single drug exposures. Furthermore, HIF was found to be significantly activated by several drugs at both 21% and 1% O₂. These results are all found in Figure 6, with statistical analyses given in Supplementary Table S1D. Some of the redox-perturbing treatments are here analyzed in further detail as follows.

Treatment of HEK293 cells with auranofin, sulforaphane, or the proteasomal inhibitor, AHQ, activates Nrf2 and HIF, but not NF-κB

Both Nrf2 and HIF are directly regulated via proteasomal degradation, while NF-κB is instead regulated through proteasomes via degradation of the NF-κB inhibitor, IκB (4, 39). We therefore wished to study how the proteasome inhibitor, 5-amino-8-hydroxyquinoline (AHQ), suggested as a candidate drug to overcome Bortezomib resistance in cancer treatment (27), might modulate the activities of these three transcription factors. Using HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB}, we found that upon treatment with 5 μM AHQ for 24 h, Nrf2 and HIF were significantly activated at both 21% O₂ and 1% O₂, while the low basal NF-κB activity in these cells was not affected (Figs. 6 and 7 and statistics in Supplementary Table S1D). Interestingly, in combinatory treatment using AHQ together with TNFα, an additive effect was seen in Nrf2 and HIF activation, while AHQ instead clearly dampened the TNFα-induced activation of NF-κB, especially at 1% O₂ (Fig. 7B). Thus, the effects of AHQ are both transcription factor and context dependent, but can still simultaneously affect the activities of the three transcription factors studied here.

FIG. 7.

In-depth studies of the cellular response of Nrf2, HIF, and NF-κB in HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB}. HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB} and exposed to the proteosomal inhibitor AHQ (5 μM), the Nrf2 inducers, auranofin (1 μM), and sulforaphane (3 μM), and/or NF-κB inducer TNFα (20 ng/ml) at (A) 21% (solid bars) and (B) 1% O₂ (HIF inducer; shaded bars). These data are based on Figure 6 and represent mean ± SEM from three separate experiments, each made in triplicates (n = 9), see statistics in Supplementary Table S1D. AHQ, 5-amino-8-hydroxyquinoline. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

We also found that AHQ gave essentially similar effects on Nrf2, HIF, and NF-κB as treatment with low subtoxic doses of either auranofin, known as a thioredoxin reductase 1 (TrxR1) inhibitor (17), or sulforaphane, classically regarded as an Nrf2-inducing agent (10). However, auranofin was also previously shown to induce Nrf2 (26, 30) and sulforaphane was found to inhibit TrxR1 (22), suggesting overlapping biological activities between these compounds. Furthermore, a significant activation of HIF was detected upon TNFα treatment (Figs. 6 and 7). Activation of HIF by TNFα treatment, as found here, was recently shown to occur also in skeletal muscle (37). Our results on AHQ, auranofin, sulforaphane, TNFα, and hypoxia effects on transcriptional responses, as measured here with our new pTRAF methodology, are summarized in Figures 6 and 7. These results illustrate well the added value of concomitantly analyzing the activation of Nrf2, HIF, and NF-κB when assessing effects of redox-perturbing treatments or conditions on the regulation of these transcription factors.

These experiments exemplified how the pTRAF tool can be used for facile monitoring of simultaneous complex responses of three transcription factor activities. The additional advantage of the pTRAF tool is, however, to allow determinations at single-cell resolution, which were next analyzed in further detail.

Assessing transcription factor activities at single-cell resolution

To visualize single-cell responses, triangular scatter plots were constructed to illustrate the degree and spread of responses in the three transcription factor activities (see the Materials and Methods section and Supplementary Fig. S2A, B for more detailed descriptions of how these diagrams were made). This mode of graphic representation of the responses was first validated using the fluorescence reporter genes expressed from constitutively active promoters, either plotting the 4000 most intensely fluorescent cells in a triangle scatter plot (Supplementary Fig. S2C) or plotting them as intensity distribution plots that summarized the profiles of all cells having fluorescence over background (Supplementary Fig. S2D). With the overall brightness of such diagrams adjusted for aggregated total responses of the readouts upon a given treatment, triangular plots were next constructed for cells transfected with pTRAF^{Nrf2/HIF/NF-κB} (Supplementary Fig. S2E) or also, in this case, as intensity distribution plots (Supplementary Fig. S2F). Important observations could thereby be made with regard to the effects of specific treatments on the transcription factor activities in addition to their overall activation patterns on the cell culture level. This is exemplified and discussed next for a selection of specific redox-perturbing treatments (for the whole series of experiments performed in the present study, see Supplementary Figs. S2E, F).

Heterogeneity and stochastic responses in certain redox therapeutic conditions

By assessing the bar graphs in Figure 7 in combination with single-cell responses as displayed in the intensity distribution plots in Figure 8, a more detailed picture of the cellular responses to the treatments becomes apparent. In the upper left corners in both Figure 8A and B are Schemes displaying how to interpret the triangles (for further details, see the Materials and Methods section and Supplementary Fig. S2B). Analyzing single-cell responses in pTRAF^{Nrf2/HIF/NF-κB}-transfected HEK293 cells (displayed as accumulated signals in the form of bar graphs in Fig. 7A, B), it became clear that AHQ, auranofin, or sulforaphane at 21% O₂ predominantly activated mainly one of the three transcription factors, with the responses illustrated as intense red squares in the upper corner of the triangle indicating Nrf2 activation (Fig. 8A). With TNFα exposure, an accumulation of blue (cyan) squares in the right corner shows that most cells mainly expressed NF-κB (Fig. 8A). Cells exposed only to 1% O₂ displayed a strong induction of HIF, as illustrated by the yellow squares in the left corner (Fig. 8B).

FIG. 8.

Single-cell response of Nrf2, HIF, and NF-κB shown as intensity distribution plots correlated to the bar graphs in Figure 7. The triangle plots display the single-cell response of Nrf2, HIF, and NF-κB in HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB}, exposed to AHQ (5 μM), auranofin (1 μM) and sulforaphane (3 μM), and/or TNFα (20 ng/ml) at (A) 21% and (B) 1% O₂ (HIF inducer). The small Scheme graphs on the left in both (A) and (B) are schematic pictures that illustrate how the graphs should be interpreted. The intensity distributions plots combine the total intensity readout data (Fig. 7) with single cell assessment (Supplementary Fig. S2E; for further details on how the plots are constructed, see Supplementary Fig. S2B), with overall brightness reflecting the total response. The color of each dot reflects the relative contribution of each response (Nrf2 is red component, HIF is yellow component, and NF-κB is cyan component), with the brightness reflecting the sum of total intensities of all cells corresponding to that response pattern. Thus, more widespread dots in these intensity distribution plots represent a larger variation in responses between individual cells. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

In certain combinatory treatments, such as TNFα in combination with AHQ, auranofin, or sulforaphane, exposed at either 21% O₂ or 1% O₂, strongly heterogenic responses were found. Hence, individual cells activated the different transcription factors to different extents with regard to absolute activities as well as the ratios between the three transcription factor responses (Fig. 8B). Such stochastic patterns of responses at single-cell resolution were not seen with constitutively active promoters (Fig. 1C and Supplementary Fig. S2B), thus strongly suggesting that the heterogeneity was explained by varying activities within a cell culture at the unique response elements present in the pTRAF^{Nrf2/HIF/NF-κB} vector.

Synergistic activation of NF-κB and HIF using doxorubicin in combination with TNFα

Because all the three transcription factors studied here have implications in cancer pathology, the pTRAF^{Nrf2/HIF/NF-κB} vector may be particularly useful for assessments of the effects of anticancer drugs. As examples of such usage, we here found that both doxorubicin and cisplatin, at low subtoxic concentrations (see maintained cell viability compared with controls in Fig. 5C), lacked activation of Nrf2, HIF, or NF-κB. However, these drug treatments still yielded high synergy with TNFα-induced activation of NF-κB in both 21% O₂ (Figs. 6 and 9 and Supplementary Table S1D) and 1% O₂ (Fig. 6, Supplementary Fig. S2C, D, and Supplementary Table S1D). Moreover, HIF became highly activated by the combination of doxorubicin and TNFα, especially in hypoxia (Fig. 6, Supplementary Fig. S2C, D, and Supplementary Table S1D).

FIG. 9.

A strong synergistic activation of NF-κB and HIF in combinatory treatment of the anticancer drug, doxorubicin, and TNFα. Cellular response of pTRAF^{Nrf2/HIF/NF-κB}-transfected HEK293 cells exposed to clinically used anticancer drugs, doxorubicin and cisplatin, for 24 h at 21% O₂ (for hypoxia, see Fig. 6 and Supplementary Fig. S2E). Upper panel intensity distribution plots of transcriptional responses of Nrf2 (red), HIF (yellow), and NF-κB (cyan) based on single-cell analyses. Red indicates Nrf2-activity, yellow HIF activity, and cyan NF-κB activity, with the brightness reflecting the sum of total intensities of all cells corresponding to that response pattern and lower graphs showing the same data as total fluorescent intensity (I/cell) per cell based on cell number of each cell stained with Hoechst (data from Fig. 6 that are based on three independent experiments in triplicates [n = 9]). To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars

Discussion

In this study, we have investigated simultaneous activation patterns of three redox-regulated transcription factors in a high-throughput manner with single-cell resolution. To enable this type of analysis, we developed and validated a new pTRAF tool for facile simultaneous determination of Nrf2, HIF, and NF-κB activities.

Using the pTRAF^{Nrf2/HIF/NF-κB} vector, we could assess the complex activation patterns of Nrf2, HIF, and NF-κB upon combinatory treatment of cells with different redox-perturbing stimuli. Because many diseases involve combinations of oxidative stress (Nrf2 activation), low oxygen tension (HIF activation), and inflammatory responses (NF-κB activation), we found it worthwhile to use this new methodology for simultaneous studies of these pathways in the exact same cellular context. Indeed, the ability to assess activities at single-cell resolution allows for better insights and understanding of the interplay of these transcription factor activities with each other. Separating and enriching different subpopulations of cells from a given experiment using fluorescence-activated cell sorting (FACS) should enable further in-depth studies of the exact molecular pathways leading to the diverse responses of these transcription factor activities between individual cells of the same cell culture. Because the pTRAF technology is based upon fluorescence, it should indeed be directly amenable for use in FACS (Supplementary Fig. S3).

The pTRAF tool should become useful as a method to assess cell type-specific transcriptional activation patterns of Nrf2, HIF, and NF-κB. In the present study, we validated its use with HEK293 cells. Other cell types, as shown herein with the cancer HCT116 or A375 cell lines, can display other patterns of responses to treatment or other levels of basal activities of Nrf2, HIF, or NF-κB. Use of the pTRAF vector in different cell types, including primary cells, will require successful transfection. As transfection efficiency varies between cell types, viral pTRAF vectors may become useful and should be assessed in future development of the method. However, it should be underlined that even in cells hard to transfect, the fact that three response elements are present in the same plasmid backbone to guide expression of fluorescent reporter proteins, it should be possible to use also in cells transfected with quite low efficiency.

Interactions between Nrf2 and NF-κB, in particular, in different pathological conditions have been much discussed, both in terms of possible molecular mechanisms for the cross talk and pathophysiological consequences (6, 14). Interactions between NF-κB and HIF (2, 5) and between Nrf2 and HIF have also been reported (29). Our study is, however, to our knowledge, the first to simultaneously study the activities of all these three transcription factors, that is, Nrf2, HIF, and NF-κB, at the same time and with single-cell resolution.

Our results revealed significant cross talk between these pathways upon treatment of cells with certain redox-perturbing stimuli. Observing the obvious synergistic HIF induction using the potent Nrf2 inducers, sulforaphane or auranofin, or the NF-κB inducer, TNFα, as well as the synergistic effects of doxorubicin together with TNFα in activation of NF-κB and HIF, these results pave the way for further detailed studies of the molecular mechanisms leading to such cross talk. It may also be that several of the detected pTRAF responses could have been triggered or modulated by a general oxidative stress that may be difficult to tell apart from modulation of dedicated redox signaling pathways converging on specific transcription factors. It is indeed likely that some drug treatments or cell culture conditions can affect the pTRAF readout due to a general stress response, rather than solely through direct interactions with specific signaling pathways. Future use of the pTRAF tool in concert with dedicated modulation of key signaling molecules, such as genetic knockout of specific regulatory genes, should be possible as a future strategy of dissecting the exact pathways that lead to the observed transcriptional responses.

The pronounced cell-to-cell heterogeneity in pTRAF^{Nrf2/HIF/NF-κB} signals under certain, but not all, conditions was an interesting finding. It should be noted that such heterogeneity cannot be detected unless single-cell resolution is enabled. However, stochastic behavior in transcription on the single-cell level has recently been reported also using several other methodologies (7, 9, 23, 24), which thus agrees well with our findings. Further work to understand the molecular determinants for stochastic patterns in transcriptional responses will be important and the pTRAF methodology could in such context prove useful as a tool to enable individual cell identification and enrichment for analyses of cells with certain present profiles of transcriptional activities.

As outlined in the Materials and Methods section, we selected a combination of three fluorescent proteins on the basis of their compatibility with each other with regard to emission and excitation spectra, that is, independent detection profiles based upon fluorescence (40). Additional properties of these fluorescent proteins are also important to consider when interpreting the results with the pTRAF methodology, including maturation times and half-lives (31).

As we utilized the well-characterized mCherry, YPet, and CFP proteins without the use of fusion proteins or additional peptide motifs that can affect their half-lives or subcellular localization, they were upon their synthesis allowed to accumulate in the cytosol. As such, their levels and thereby the overall fluorescent signals were dictated by the respective promoter activities, in combination with inherent maturation and stability profiles of the fluorescent proteins in the mammalian cellular context. Although we did not measure these properties directly, we assume that in our cell experiments also the fluorescent proteins would have preserved maturation times of up to 1–2 h with high stability and turnover times of several days, as shown previously (8, 31). Thus, with our current version of the pTRAF vector, the readouts represent accumulated levels of reporter activities after several hours of cell treatment. As such, we could here use the method to determine how different treatments activated Nrf2, HIF, and/or NF-κB over time courses of several hours. The methodology can therefore in this form not be used for analyses of rapid (within minutes) alterations of promoter activities, and also not for assessment of cessation of promoter activity.

A further development of the pTRAF methodology to allow for detection of dynamic changes in reporter activities, within the minute time range, would clearly require use of high-turnover reporter proteins. That may serve as the basis for future developments of the technique, but a drawback of such constructs would likely be a lower detection limit due to lower total levels of the reporter proteins even after promoter activation.

To conclude, the novel pTRAF methodology was here shown to enable simultaneous assessment of three separate transcription factor responses at single-cell resolution and in a methodology compatible with high-throughput analyses. The pTRAF tool is easy and fast to use, cost-effective, compatible with most types of fluorescent detection setups, and works with live as well as fixed cells. Furthermore, in this study, we evaluated the methodology for analyses of Nrf2, HIF, and NF-κB activation patterns upon exposure of cells to diverse forms of redox therapeutics.

However, we conclude that the same principal methodology should be possible to adapt for studies of many other transcription factors simply by a change of the three transcription factor response elements in the pTRAF vector. In this study, we have used repeats of consensus response elements for human Nrf2, HIF, and NF-κB to guide expression of the reporter proteins. Additional derivatives of such pTRAF construct should be rather straightforward to construct for future studies with other scope. One possibility would be to convert the response elements to become compatible with consensus sequences found in animals for work in animal models, such as rodents, fish, or worms. Another possibility would be to study strengths in activation of different variants of the same response element, for example, studying three variations of a consensus sequence for the same transcription factor, or family of factors. The alternative strategies of using this general approach to study three separate reporter activities in single cells should be many and will hopefully be addressed in future studies.

Materials and Methods

Reagents and cloning templates

The pGL4.32 (Promega) vector, containing NF-κB response elements that drive transcription of the luciferase gene luc2P, was used as a backbone for cloning of the various plasmids used in this study. The pGL4.73 vector (Promega) was used as a transfection control in dual-luciferase reporter assays, according to the manufacturer's instructions, as well as a backbone for cloning vectors that were made to constitutively express fluorescent proteins, as described below. Four repeats of the Nrf2 binding antioxidant response element (ARE: TGCAAAATCGCAGTCACAGTGACTCAGCAGAATCTGAGCCTAGG) sequence from the human NQO1 [NAD(P)H:quinone oxidoreductase 1] gene (25) and either four or eight repeats of the HIF binding hypoxia response element (HRE: GATCGCCCTACGTGCTGTCTCA) sequence from the human EPO (erythropoietin) gene (11) were ordered as oligonucleotides with flanking 3′ NheI site and 5′ BglII sites from BlueHeron. Plasmid preparation kits were from Qiagen and template DNA was prepared from transformed Escherichia coli MachI clones grown in LB medium with appropriate addition of 100 μg/ml ampicillin.

Sequencing of the products from all major cloning steps was performed by GATC Biotech. Enzymes required for the cloning steps were purchased from Thermo Scientific, the fluorescent protein, mCherry, was a kind gift from Dr. Lars Bräutigam of the Prof. Arne Holmgren laboratory, Karolinska Institutet, YPet was purchased from Addgene (cat. no. 14032) (36), and CFP (35) was a kind gift from Prof. Björn Öbrink, Karolinska Institutet. For sequences of all primers used in the cloning steps as described below, see Supplementary Table S1A. All vectors constructed in this study are listed in Figure 2, which also includes a scheme of the final response elements guiding transcription of the pTRAF reporter genes.

Vector nomenclature

Vectors with only one response element or one promoter guiding expression of a reporter gene are named with the reporter gene and response element or promoter in superscript: pGL4.[reporter gene]^{[response element/promoter]}. Plasmids with three response elements guiding transcription of three different fluorescent proteins in the order of mCherry, YPet, and CFP are abbreviated as pTRAF, with the corresponding specific response elements in superscript: pTRAF^{[response element 1, 2, 3]}. The control vector with the red, yellow, and cyan fluorescent proteins driven by individual SV40 promoters is termed pTRAF^3xSV40. See Figure 2 for the different vectors and sequences of the response elements.

Nrf2, HIF, and NF-κB-driven transcription of the luciferase reporter gene luc2P

Using the pGL4.32 vector as backbone, the NF-κB response elements were exchanged with four copies of ARE (Nrf2 response element) and four copies of HRE (HIF response element) to create the vectors, pGL4.luc2P^Nrf2 and pGL4.luc2P^HIF, respectively. For this cloning, the sequences were cleaved with restriction enzymes, NheI and BglII, and individually ligated into pGL4.32 vectors cleaved with the same restriction enzymes (see schematic overview Fig. 2).

Exchanging the luciferase reporter genes in pGL4.luc2P^Nrf2, pGL4.luc2P^HIF, and pGL4.32 with the gene for mCherry and exchanging the four repetitions of HRE with eight copies

The luc2P gene in pGL4.luc2P^Nrf2, pGL4.luc2P^HIF, and pGL4.32 was replaced with mCherry using standard PCR techniques. The pGL4.luc2P^Nrf2, pGL4.luc2P^HIF, and pGL4.32 vectors were thereby used as templates in three separate PCR reactions with the primers, pGL4.32F and pGL4.32R, introducing flanking NdeI and EcoRI sites, while another PCR was performed on the plasmid carrying mCherry using the primers, mCherryF and mCherryR (for primer sequences, see Supplementary Table S1). All the PCR products were purified and cleaved with NdeI and EcoRI, whereupon the cleaved mCherry PCR product was individually ligated into each of the three cleaved pGL4.luc2P products. The newly produced vectors thereby contained an in-frame mCherry gene in the position previously occupied by the luc2P gene. Subsequently, to increase the transcriptional response readout to activation of HIF, the four repetitions of HRE were expanded to eight repetitions of HRE. See Fig. 2 for a schematic drawing of these vectors.

Exchanging the gene for mCherry in pGL4.mCherry^HIF with YPet and in pGL4.mCherry ^NF-κB with CFP

The mCherry gene of pGL4.mCherry^HIF and pGL4.mCherry^NF-κB was subsequently replaced with the genes for YPet and CFP, respectively. For this, PCR reactions were performed on the plasmids containing the YPet and CFP genes (see Supplementary Table S1A for primer details), with the products purified and cleaved with NdeI and EcoRI and subsequently ligated into the corresponding sites of the pGL4.mCherry^HIF and pGL4.mCherry^NF-κB vectors, respectively. For schematic drawings of these vectors, see Figure 2A–D.

Vectors constitutively expressing mCherry, YPet, and CFP

Control vectors constitutively expressing the various fluorescent proteins were cloned by applying similar steps as described above, using the vector pGL4.73 as a template, which contains an SV40 promoter that drives expression of the hRluc gene (Promega). For this, the primers, pGL4_73F and pGL4_32R, were used for a PCR with pGL4.73 as the template, which introduced flanking NdeI and EcoRI sites. These were subsequently utilized to ligate mCherry, YPet, and CFP as inserts, following the procedures described above. The newly produced pGL4.mCherry^SV40, pGL4.YPet^SV40, and pGL4.CFP^SV40 vectors that expressed the various fluorescent proteins under control of the SV40 promoter are schematically shown in Figure 2E.

Construction of the pTRAF^{Nrf2/HIF/NF-κB} vector

The gene regions of pGL4.mCherry^Nrf2, pGL4.YPet^HIF, and pGL4.CFP^NF-κB that encompassed the response elements, as well as the genes for the respective fluorescent proteins, are here referred to as single cassettes. These cassettes of Nrf2-mCherry, HIF-YPet, and NF-κB-CFP were cloned into the pGL4.32 vector backbone, at the same time introducing unique restriction sites flanking each cassette. The Nrf2-mCherry cassette was made to be flanked by Af1ll and XhoI, NF-κB-CFP flanked by AatII and AscI, and with HIF-YPet flanked by MluI and SalI, thus constituting the final pTRAF^{Nrf2/HIF/NF-κB} vector. The cassettes were introduced so that the transcriptional directions of the reporter sequences were alternated to minimize interference between the different cassettes. A schematic structure of the final pTRAF^{Nrf2/HIF/NF-κB} vector is shown in Figures 3A and 2F together with the sequences of the response elements. All final cassettes and the sequences shown in the scheme were fully confirmed by DNA sequencing.

Construction of pTRAF^3xSV40, a vector constitutively expressing mCherry, YPet, and CFP

pTRAF^3xSV40 was constructed as a size-matched control to pTRAF^{Nrf2/HIF/NF-κB} that constitutively expresses mCherry, YPet, and CFP (Fig. 2G). To clone pTRAF^3xSV40, pTRAF^{Nrf2/HIF/NF-κB} was used as a template and all three cassettes were exchanged with SV40 promoter-driven mCherry, CFP, and YPet instead of the response elements. The new SV40-containing cassettes were amplified from the respective constitutively expressing single vectors by using the following primers: for SV40-mCherry: Af1llF and XhoIR; SV40-CFP: AatIIF and AscIR; and SV40-YPet: MluIF and SalIR.

Cell cultures

HEK293 (ATCC) cells were cultured in Eagle's minimum essential medium (No 30–2003; ATCC), HCT116 (human colorectal carcinoma; ATCC) cells were cultured in McCoy's 5A medium (ATCC), and A375 (human malign melanoma; kind gift from Sonja Lain, Karolinska Institutet) cells were cultured in Dulbecco's modified Eagle medium (Lonza, BE12-74F) supplemented with 1 mM sodium pyruvate (Gibco). All media were additionally supplemented with 10% fetal bovine serum (PAA; GE Healthcare) and 100 units penicillin/ml and 100 μg streptomycin/ml (Biochrom). Cells were cultured at 37°C in a humidified atmosphere with 5% CO₂ in either 21% O₂ or in hypoxia experiment 1% O₂ (3% O₂ for luciferase assay).

Drug treatments

Since pathways were studied in this article, low subtoxic concentrations of the following drugs were used: auranofin (1 μM), BHQ (30 μM), 1-chloro-2,4-dinitrobenzene (CDNB; 1 μM), and etoposide (25 μM), which were all diluted in ethanol (EtOH) as 10 mM stocks, and these compounds should be compared with the EtOH control. TNFα (20 ng/ml), AHQ dihydrochloride (5 μM), tert-butyl hydroperoxide (BuOOH; 100 μM), cumene hydroperoxide (CuOOH; 30 μM), cobalt sulfate (cobalt; 100 μM), doxorubicin (0.2 μM; from TEVA Pharmaceuticals), cisplatin (2 μM; from MEDA AB), and S-nitrosoglutathione (GSNO; 80 μM diluted in 0.1 M sodium phosphate buffer pH 7.4, 0.1 mM EDTA, Chelex-treated) should be compared with control (pTRAF-transfected unexposed cells). All compounds were purchased from Sigma-Aldrich if not stated otherwise. Viability assay using Hoechst (33342; Life technologies™)-stained nuclei and the Operetta system were first used to determine subtoxic concentrations of the mentioned drugs, see Figure 5A for final concentrations and Figure 5C for toxicity readouts. Hypoxia experiments were performed at 1% O₂, but since the fluorescent proteins need O₂, for proper folding, the cells were transferred to 21% O₂ for 2 h before fixing or analysis.

Luciferase assays to validate response elements for Nrf2, HIF, and NF-κB

To measure NF-κB responses in HEK293 cells with luciferase as reporter, the pGL4.32 (Promega) vector was used, whereas the newly cloned pGL4.luc2P^Nrf2 and pGL4.luc2P^HIF vectors were used in the same manner for Nrf2 and HIF responses. For luciferase quantification, the dual-luciferase assay was performed according to the manufacturer's instructions (Promega). In brief, one day before transfection, HEK293 cells were seeded at a density of 20,000 cells/well in 96-well plates. Cells were thereupon transfected with 0.05 μg DNA/well (pGL4.luc2P^Nrf2, pGL4.luc2P^HIF, or pGL4.32). Lipofectamine was used as transfection reagent at a ratio of 1:3 luciferase DNA/Lipofectamine. To correct the luciferase signals for transfection efficiencies, cells were cotransfected with a control vector constitutively expressing renilla luciferase (pGL4.73) with 0.001 μg DNA/well. The luminescence signals were measured as RLU (relative light unit) using a Victor™, 1420, Multilabel counter (PerkinElmer).

Quantitative RT-PCR for downstream genes of Nrf2, HIF, and NF-κB

In brief, HEK293 cells were seeded at a density of 500,000 cells/well in six-well plates and cultured overnight. Cells were then either transfected with pTRAF^{Nrf2/HIF/NF-κB}, using 1 μg DNA, 100 μl OptiMEM, and 2 μl TurboFect per well, or were used without transfection. For induction of transcription factors, cells were exposed to 20 ng/ml TNFα or 1 μM auranofin for 1, 3, or 6 h or sulforaphane 3 μM either solely or in combination with 20 ng/ml TNFα for 6 h in either 21% or 1% O₂. After exposure, the cells were washed with phosphate-buffered saline (PBS), trypsinized, centrifuged, and resuspended in 300 μl Buffer RLT (kit from Qiagen) with 1% β-mercaptoethanol and stored at −80°C until mRNA isolation. The mRNA isolation was performed with the RNeasy Mini kit from Qiagen and cDNA purification was done with the high-capacity cDNA reverse transcription kit (Applied biosystems) according to the manufacturers' protocol and stored at −80°C until analysis.

Quantification of gene expression was performed in two different settings; either in duplicate using the Maxima SYBR Green qPCR Master Mix (Fermentas), with detection on an Applied Biosystems 7500 real-time PCR system (Applied Biosystems), or in triplicates using SYBR Green (Bio-Rad), with detection on the Bio-Rad CFX Optical System (Bio-Rad). The reactions used were at 95°C for 2 min succeeded by 40 cycles at 95°C for 15 s and 60°C for 1 min or 95°C for 3 min succeeded by 40 cycles at 95°C for 10 s, 60°C for 30 s, and 72°C for 30 s, respectively, both programs were followed by melt curve analysis. Relative gene expression quantification was based on the comparative threshold cycle method (2^−ΔΔCt) using GAPDH or HPRT and ACTB as endogenous reference genes. Primers are listed in Supplementary Table S1B and a complete list of ΔC_t values for each transcript in the correlation plot (Fig. 3B) can be found in Supplementary Table S2.

Analyses with the Operetta high-content imaging system

HEK293 cells were seeded in collagen I-coated 96-well plates (Biocoat 354649, BD) at a density of 20,000 cells/well one day before transfection. The cells in each well were transfected with 0.02 μg DNA, 5 μl OptiMEM, and 0.04 μl TurboFect diluted in 55 μl media overnight, followed by exposure to different toxicants for 24 h (see model plate setup and concentrations Fig. 4A). To prepare the samples for analysis, the cultures were incubated with 40 ng/ml Hoechst for 30 min to stain the nuclei and subsequently fixated in 2% ice-cold paraformaldehyde for 10 min at RT. The wells were covered with PBS and plates were stored in the refrigerator until analysis. The Columbus (PerkinElmer), IgorPro (WaveMetrics), ImageJ, and Excel (Microsoft) computer programs were used to analyze the data obtained with the Operetta high-content imaging systems. An AxioVert 40 CFL (Zeiss) fluorescence microscope was also used to monitor the cellular responses.

The experiments using HCT116 and A375 cells were performed in a similar manner as HEK293 cells, except that cells were seeded at a density of 30,000 cells/well, cultured overnight, and transfected for 6 h before being exposed to 1 μM auranofin and 0.2 μM doxorubicin for 17 h. The Columbus program was used to analyze all the data, including fluorescent pictures, and calculate the intensity per cell. Transfection efficiencies in our study were ∼40% (HEK293 and HCT116 cells) or 70% (A375 cells).

For single-cell quantification, seven fields of view for each cell culture well, covering both edges and center, were recorded in bright-field microscopy as well as four channels for fluorescence to detect the mCherry, YPet, CFP, and Hoechst signals. Excitation and emission filters were set to optimize the signals with minimized leakage between the channels (see Fig. 1B, C, and Supplementary Fig. S2B). The excitation and exposure times were fixed for each channel, with all samples analyzed with the same settings. Determinations of fluorescence signals were subsequently performed by preset algorithms, as described next.

Automated data analyses of single-cell fluorescence signals

Subtraction of the smooth and continuous background was first performed for all fluorescence pictures using the Sternberg's rolling ball method (41) as supported by the ImageJ software. Single-cell identification as well as quantification of cell numbers was performed using the bundled Columbus software (PerkinElmer). For this, individual cells were identified based upon Hoechst staining of nuclei and next the corresponding cytosols were defined using a maximum intensity z-projection of mCherry, YPet, and CFP to ensure that all cells showing responses above background with any of the three fluorescent proteins were analyzed. The intensities of these signals were quantified as total signals within the defined cell area. The resulting single-cell data altogether covering the number of identified cells, the area of each cell, and its fluorescence signal intensities were subsequently exported from the Columbus software and further analyzed using the Excel, IgorPro, and Prism (GraphPad) computer programs.

Background cutoff fluorescence levels were defined using cells transfected with vectors constitutively expressing single fluorescent proteins (using two of the three pGL4.mCherry^SV40, pGL4.YPet^SV40, and pGL4.CFP^SV40 plasmids at a time). The background was subsequently estimated via the channel of the fluorescent protein that was not expressed and defined as the mean fluorescence intensity (intensity of a single cell divided by its cell area) where 99.9% of all cells in this channel were below background of at least 100,000 cells. This definition ensured a consistent correction for background fluorescence intensity that also corrected for potential leakage between channels. Only cells displaying signals of at least one fluorescent protein above background were considered for further analysis. To determine accumulated responses on the cell culture level, all single-cell signals within an experiment were summarized (after subtraction of background) and corrected for total cell numbers, as described in the main text.

Single-cell scatter plots

pTRAF^{Nrf2/HIF/NF-κB} carries the three response elements on the same vector, which can therefore be used as their own internal transfection control. Thus, direct comparison between the responses of the three different transcription factors could be done within the same cells, thereby yielding estimations of the relative expression of the three fluorescent reporter proteins. For scatter plots, total expression responses for each single cell were first determined by summarizing the fluorescence intensities of their mCherry, YPet, and CFP signals. Subsequently the relative fractions of mCherry, YPet, or CFP in relation to the total responses were calculated. Finally, scatter plots were made to visualize the distribution of the three fluorescence signals with regard to each other, using the 4000 most intense cells for each sample (for further description on how these scatter plots were made, please see Supplementary Fig. S2A, C).

Intensity distribution triangle plots

Intensity distribution plots were also made to better visualize the combination of the total fluorescence intensity readout of an experiment (Fig. 6) with the distribution of the signals as seen in the single-cell scatter plots (Supplementary Fig. S2D). For this, all of the responding cells in the analyzed experiments were included as the basis for scatter plots. These plots were subsequently transformed to intensity distribution plots, whereby the whole scatter plot area was divided into 210 smaller areas, with each small area of the plot thereby assembling cells having similar relative combinations of mCherry, YPet, and CFP levels (for further description of this methodology, please see Supplementary Fig. S2B). All the intensity distribution plots were based upon the average signals of three independent experiments, each made in triplicates.

Multichannel FACS analyses

The transcriptional response of Nrf2, HIF, and NF-κB could also be monitored in living HEK293 cells transfected with pTRAF^{Nrf2/HIF/NF-κB} using FACS analysis. Please see Supplementary Figure S3 for results and further details of the methodology.

Statistics

A two-tailed t-test was used for statistical analysis. For the qRT-PCR data in Figure 3A, see p-values in Supplementary Table S1C, and the p-values of the transcriptional response shown in Figures 6, 7, and 9 can be seen in Supplementary Table S1D.

Abbreviations Used

AHQ

5-amino-8-hydroxyquinoline

ARE

antioxidant response element

BHQ

tert-butylhydroquinone

CDNB

1-chloro-2,4-dinitrobenzene

CFP

cyan fluorescent protein

FACS

fluorescence-activated cell sorting

GSH

glutathione

HEK293

human embryonic kidney cells 293

HIF

hypoxia-inducible factor

IL8

interleukin-8

IκB

inhibitor of κB

NF-κB

nuclear factor kappa-light-chain-enhancer of activated B cells

Nrf2

nuclear factor E2-related factor 2

PBS

phosphate-buffered saline

pTRAF

plasmid for transcription factor reporter activation based on fluorescence

qRT-PCR

quantitative reverse transcription-PCR

RLU

relative light units

TNFα

tumor necrosis factor alpha

Trx

thioredoxin

TrxR1

thioredoxin reductase 1

YPet

yellow fluorescent protein for energy transfer

Acknowledgments

The authors are thankful to Birgitta Wester at the MTC core facility of Karolinska Institutet for help with the FACS and to Qing Cheng, Division of Biochemistry, MBB, Karolinska Institutet, for valuable tips of cloning strategies. The authors acknowledge the funding received from the Swedish Society for Medical Research (SSMF), the Swedish Research Council, the Swedish Cancer Society, the Swedish Medical Society, and Karolinska Institutet.

Author Disclosure Statement

No competing financial interests exist.