A high-sensitivity bi-directional reporter to monitor NF-κB activity in cell culture and zebrafish in real time

Nuclear factor (NF)-κB transcription factors are essential for a number of biological processes such as inflammatory and immune responses, cell growth, apoptosis and development. Their inappropriate activation has been linked to autoimmunity, chronic inflammation and various types of cancer (Ben-Neriah and Karin, 2011; Hayden and Ghosh, 2008; Napetschnig and Wu, 2013).

NF-κB transcription factors are expressed in many cell types. Five NF-κB polypeptides, p65 (RelA), c-Rel, RelB, p50 (encoded by NFKB1) and p52 (encoded by NFKB2), can combine to form 15 different transcription factors through homo- and hetero-dimerization (Hoffmann and Baltimore, 2006). The transcriptional activation of target genes depends on the nuclear translocation of NF-κB dimers. In the absence of stimulatory signals, inhibitory κB protein family members (e.g. IκBα and IκBβ, also known as NFKBIA and NFKBIB, respectively) keep NF-κB transcription factors sequestered in the cytoplasm, thereby preventing their binding to DNA. Activation of different signaling pathways by proinflammatory signals such as pathogen-associated molecular patterns, danger-associated molecular patterns or proinflammatory cytokines (e.g. interleukin-1), results in IκB kinase (IKK) activation. This kinase complex phosphorylates the IκBs, inducing their proteasomal degradation and resulting in release of NF-κB transcription factors and their translocation into the nucleus (Vallabhapurapu and Karin, 2009). All NF-κB transcription factors share the ability to bind to the κB site consensus sequence 5′-GGGRNWYYCC-3′ (where R is a purine, Y is a pyrimidine, W is an adenine or thymine and N is an unspecified base) (Hoffmann and Baltimore, 2006). However, the binding affinity of NF-κB transcription factors to different κB site sequences varies, indicating a relationship between binding affinity, specificity and function (Siggers et al., 2012).

Various assays have been established to determine the activity of the canonical NF-κB pathway. For example, IκBα degradation assayed by western blotting, and analysis of NF-κB binding activity in the absence and presence of IκBα by electrophoretic mobility shift assay (Vancurova et al., 2001). However, these time-consuming methods are restricted to the analysis of small sample numbers. An IκBα-firefly luciferase fusion reporter (IκBα-FLuc) allows IKK activation to be monitored in real time (Gross and Piwnica-Worms, 2005). Non-canonical NF-κB activation cannot be measured using these methods since it occurs independently of IκBα degradation (Hayden and Ghosh, 2004). To assess NF-κB transcription factor dynamics, cell lines that stably express the NF-κB subunit p65 fused to GFP (Bartfeld et al., 2010) are used to quantify the nuclear translocation of p65 in response to stimulation. These assays, however, do not provide direct levels of downstream transcriptional activation by NF-κB. The more commonly used method to assay NF-κB activation is based on reporter genes (e.g. luciferase or β-galactosidase) driven by promoters containing multiple repeats of an NF-κB consensus sequence (Badr et al., 2009; Bowie and O'Neill, 2000; Matsuda et al., 2007; Munoz et al., 1994; Schindler and Baichwal, 1994). The sensitivity and specificity for NF-κB activation varies between these reporters. In this work, we have used a palindromic κB site sequence with high binding affinity for multiple NF-κB proteins and developed a new bi-directional NF-κB-responsive promoter to simultaneously drive the GFP and luciferase reporter genes. We named this reporter pSGNluc. We found that pSGNluc is highly sensitive and inducible in response to various stimuli in cell culture. In addition, we established a transgenic zebrafish line carrying the pSGNluc reporter. We show that GFP and luciferase reporters can be used for different purposes, allowing us to visualize and quantify the NF-κB activity during embryonic development and in response to inflammation in live embryos. In toto imaging of zebrafish embryos carrying pSGNluc revealed a dynamic NF-κB activity during early development. Our data also suggest that Iκbαa (encoded by nfkbiaa), a member of the family of NF-κB inhibitory proteins, regulates the NF-κB activity in the proctodeum and skin.

Most available NF-κB-responsive reporters use multimerized κB site sequences upstream of a minimal promoter to drive a firefly luciferase gene. The sequence and number of κB sites vary between different NF-κB reporters. According to a large dataset of potential κB site sequences for multiple NF-κB dimers (Siggers et al., 2012), the predicted binding affinity of NF-κB dimers to the κB site sequences used to generate these NF-κB reporters is moderate (Fig. 1A). We hypothesized that a binding site with high affinity for a range of different NF-κB dimers might allow us to make a more sensitive reporter. The study by Siggers et al. showed that the sequence 5′-GGGAATTCCC-3′, originally identified as a κB site upstream of the type VII collagen gene (Kon et al., 1999), is able to interact strongly with different combinations of hetero- and homo-dimers of NF-κB subunits. In designing a new reporter plasmid, we took advantage of the palindromic nature of this sequence to create a bi-directional reporter. In this plasmid, which we named pSGNluc, two minimal promoters flank a DNA fragment containing eight tandem copies of the 5′-GGGAATTCCC-3′ sequence (Fig. 1B). GFP was cloned downstream of one promoter to visualize expression in living cells. To enable quantitative analysis of expression levels, luciferase was cloned downstream of the minimal promoter on the opposing side (Fig. 1C). The pSGNluc reporter should therefore simultaneously express both marker genes from the bi-directional promoter, allowing both the spatial and temporal monitoring and quantification of NF-κB activity in living cells.

To examine the efficiency of the new NF-κB-responsive reporter in cell culture, we transfected the pSGNluc vector in HEK293T cells. Transfected, unstimulated cells showed weak or no GFP expression (Fig. 2A), indicating a low background activity of the pSGNluc reporter. Exposure to 10 ng/ml of TNFα (also known as TNF) led to a strong upregulation of GFP expression in transfected cells, as shown by live-cell microscopy and flow cytometry data (Fig. 2A,B).

We also measured the level of luciferase activity of pSGNluc, and compared it to two other NF-κB-driven luciferase reporters, pGL4.32 (Promega) and IgκB-Luci (Munoz et al., 1994) as well as an IL-8 reporter (Bowie et al., 2000). We treated cells transfected with each of these constructs with 10 ng/ml of TNFα and measured the luciferase activity in cell lysates 24 h after treatment. We calculated the inducibility of each of the reporters (measured as the ratio of luciferase with and without treatment) and observed that pSGNluc was induced to the same extent as pGL4.32 when compared to non-treated controls. The IgκB-Luci and IL8-Luc reporters displayed the highest and lowest inducibility levels, respectively (Fig. 2C). All transfected cells showed low levels of background signal without stimulation, although the pSGNluc reporter was expressed at up to 6-fold higher levels than the IL-8 reporter (Table S1). To compare the sensitivity of the NF-κB-responsive reporters, the cells were stimulated with increasing TNFα concentrations. The luciferase level of all four reporters increased in a TNFα-dose-dependent manner after a 24 h treatment. The absolute level of luciferase expressed with the pSGNluc reporter in response to TNFα at 1 ng/ml was 21-, 34- and 73-fold higher than for the IgκB-Luci, pGL4.32 and IL-8 reporters (Fig. 2D,E).

We then determined the kinetics of NF-κB activation in response to 10 ng/ml TNFα. The pSGNluc reporter reached luciferase activity levels that were significantly higher than other reporters tested from 6 h after treatment (Fig. 2F), illustrating an increased sensitivity even at early time points after stimulation.

We also tested the response of the reporter to other inflammatory signals. Nod1 is an intracellular pattern recognition receptor that induces NF-κB activation upon detection of bacterial peptidoglycans (Chamaillard et al., 2003; Girardin et al., 2003). HEK293T cells have low levels of endogenous Nod1 (Viala et al., 2004). In many studies, HEK293T cells are therefore transfected with Nod1 expression plasmids (Masumoto et al., 2006; Zurek et al., 2011) to increase the NF-κB response towards Nod1 stimulation. However, the amount of Nod1 receptor in the cell is crucial and its overexpression can result in autoactivation (Zurek et al., 2011). We tested whether the high sensitivity of pSGNluc would allow the detection of NF-κB activation mediated by the endogenously expressed Nod1. HEK293T cells were transfected with pSGNluc and then treated with L-Ala-γ-D-Glu-meso-diaminopimelic acid (Tri-DAP) (Chamaillard et al., 2003; Girardin et al., 2003). The levels of luciferase activity reached by pSGNluc after Tri-DAP stimulation were 13- to 55-fold higher than those of the pGL4.32, IgκB-Luci and IL-8 reporters (Fig. 3A; Table S2), representing a 24±8-fold (mean±s.d.) increase in luciferase activity in cells carrying the pSGNluc reporter (Fig. 3B). The fact that these inductions were small compared to those achieved by TNFα might suggest that Tri-DAP is a weaker activator of NF-κB signaling. However, when we looked at the GFP expression pattern of transfected cells we observed that, instead of a homogeneous low expression level that could be expected from general weak response, only a fraction of transfected cells responded to Tri-DAP stimulation (Fig. 3C,D). Time-lapse imaging revealed that these cells expressed GFP at similar levels to cells treated with TNFα (Fig. 3E; Movie 1), albeit with a delayed onset. This indicates that the average level of luciferase activity in all cells after treatment was low not because the Nod1 ligand Tri-DAP elicits a weak response, but because only a limited number of cells responded to this treatment.

Overall, cell culture assays showed that pSGNluc is inducible after treatment with inflammatory stimulants to a degree that was comparable with other reporters (measured as fold induction over background expression), but is much more sensitive than other NF-κB-driven luciferase reporters. In particular, we found pSGNluc suitable to monitor NF-κB activation mediated by endogenous Nod1 in HEK293T cells. Furthermore, the bi-directional promoter allows the concomitant monitoring of NF-κB activation by expression of the two reporters, GFP and luciferase, and enables single-cell analysis of NF-κB activation and easier time-resolved analysis.

Members of the NF-κB transcription factor family are expressed almost ubiquitously during embryonic development. In transgenic mice carrying an NF-κB-responsive reporter, the spatial activity of NF-κB signaling has been observed either in fixed tissues (Schmidt-Ullrich et al., 1996) or in live animals with low spatiotemporal resolution (Carlsen et al., 2002). Given that zebrafish NF-κB proteins can bind to the mammalian consensus κB site (Correa et al., 2004), we created transgenic zebrafish carrying the pSGNluc reporter to observe NF-κB activity in the living organism. This transgenic fish line is hereafter called Tg(8×Hs.NFκB:GFP,Luciferase) in accordance with the Zebrafish nomenclature guidelines. We used light-sheet microscopy to monitor the activity of the pSGNluc reporter during early zebrafish embryonic development at high spatiotemporal resolution. Embryos carrying the transgene expressed GFP in a dynamic pattern. Live in toto imaging of the transgenic embryos from mid-gastrula to the beginning of the segmentation stage [approximately from 6 to 12 hours post-fertilization (hpf)] revealed that there already was ubiquitous and weak GFP expression at the shield stage, with the entire animal pole labeled (Fig. 4A). Cells retained this level of expression throughout the gastrula stage and into the segmentation stage (Fig. 4B–E). The first increase in the GFP signal was observed in the nasal vesicle (Fig. 4F) and in the proctodeum beginning shortly afterwards (Fig. 4G). Using confocal microscopy, we observed strong GFP expression in microvillous sensory neurons that are connected to the forebrain (Fig. 4H), which diminished by 2 days post-fertilization (dpf) (Fig. 4I). At 2 dpf, the signal was strongly present in the lateral line (Fig. 4I, yellow arrows), the proctodeum and epithelial cells around the edges of the dorsal and ventral fins. This expression pattern remained at 3 dpf (Fig. 4J). By 5 dpf, cells in the intestinal lining began to express GFP, and expression in the lateral line and fins remained (Fig. 4K, white arrowheads). The onset of expression in the gut coincides with the stage in which the larva begins to feed. After the onset of hematopoiesis, embryos showed a dim GFP signal. A similar level of GFP was also observed in keratinocytes. Zebrafish embryos carrying a previously published NF-κB-responsive reporter also expressed GFP in the skin (Candel et al., 2014; Hatzold et al., 2016). However, the signal in the skin was stronger in embryos carrying the latter reporter than in Tg(8×Hs.NFκB:GFP,Luciferase) embryos (Fig. S1A). In mice, NF-κB proteins are mainly present in basal keratinocytes (Takao et al., 2003), which are mitotically active and provide new cells that gradually undergo differentiation toward the skin surface. To better characterize the GFP signal in the skin, we took advantage of the Tg(krt19:dTomato) fish, in which basal keratinocytes are fluorescently labeled (Fischer et al., 2014). In vivo imaging of double-transgenic Tg(8×Hs.NFκB:GFP,Luciferase; krt19:dTomato) embryos revealed that GFP is expressed at a low level in basal keratinocytes, albeit in a mosaic fashion (Fig. S2). As NF-κB signaling has more than one role in keratinocytes, it remains to be determined whether this expression pattern is associated with keratinocyte homeostasis or inflammation.

To test the specificity of pSGNluc-driven GFP expression as a proxy for NF-κB activation in zebrafish embryos, we interfered with IκBα function. IκBα, an NF-κB inhibitory protein, binds to NF-κB dimers and prevents their translocation into the nucleus (Arenzana-Seisdedos et al., 1997). Knockdown of the zebrafish ortholog iκbαa should therefore enable NF-κB proteins to move to the nucleus and activate transcription of target genes (He et al., 2015). In zebrafish, iκbαa is expressed in the proctodeum and nasal vesicle (Fig. S3). We injected a previously published morpholino (MO) (He et al., 2015) to interfere with Iκbαa function in Tg(8×Hs.NFκB:GFP,Luciferase) embryos. In iκbαa morphants, GFP expression was enhanced in the skin and proctodeum (Fig. 5A–C). Although our RNA in situ hybridization analysis did not detect iκbαa expression in the skin, previous microarray analysis has shown that both the iκbαa and iκbαb (also known as nfkbiab) genes are expressed in the zebrafish skin (Lü et al., 2012). At high doses, injection of the iκbαa morpholino led to severe defects in notochord formation (Fig. 5D). This phenotype is consistent with those previously observed in zebrafish for NF-κB loss- or gain-of-function analysis (Correa et al., 2005, 2004). The tissue-specific increase in GFP levels resulting from knockdown of the NF-κB inhibitor iκbαa suggests that reporter activity in the Tg(8×Hs.NFκB:GFP,Luciferase) line is linked to NF-κB activation, and that the dynamic activity of NF-κB signaling in various tissues during embryonic development can be monitored through live imaging of this transgenic line.

As regulators of the immune and inflammatory processes, NF-κB transcription factors are activated in response to a variety of signals including pathogens, stresses and injuries. We therefore monitored the dynamics of NF-κB activation in response to injuries in Tg(8×Hs.NFκB:GFP,Luciferase) embryos. In a tailfin-wounding assay, GFP expression was upregulated in the area around the wound edge of injured embryos (Fig. 6A; Movie 2). Most epithelial cells at the wound edge exhibited strong NF-κB activation when examined 24 h later. To test the inducibility of the reporter, we first compared the kinetics of NF-κB activation in this assay with those for a previously developed reporter in zebrafish (Banerjee and Leptin, 2014). The Tg(NFκB-EGFP) line, previously used to monitor NF-κB activation upon UV radiation in zebrafish embryos, contains three copies of κB site sequences derived from the IgκB-Luci vector (Banerjee and Leptin, 2014). No GFP is detectable in Tg(NFκB-EGFP) embryos during embryonic development. In response to sterile injury, Tg(NFκB-EGFP) embryos showed weak GFP expression 40 h after wounding. In contrast, Tg(8×Hs.NFκB:GFP,Luciferase) embryos showed a strong increase in GFP expression 40 h after injury (Fig. S1B). Thus, Tg(8×Hs.NFκB:GFP,Luciferase) embryos show a more sensitive response to local injuries than the Tg(NFκB-EGFP).

We also tested the efficiency of the pSGNluc reporter in response to infection. Non-pathogenic Gram-negative Escherichia coli expressing red fluorescent protein were injected into the notochord or the inner ear. No GFP expression driven by pSGNluc was detectable at 3 dpf either in the notochord (Fig. 6B, upper panel) or the inner ear (Fig. 6C, left panel) of untreated animals. Injection of PBS alone resulted in GFP expression in the notochord (Fig. 6B, middle panel) indicating that a sterile injury alone activates NF-κB signaling. However, even higher GFP levels were observed in notochords infected with E. coli when examined 18 h post infection (Fig. 6B, bottom panel). This is in agreement with previous studies showing that microinjection of bacteria in these tissues results in an inflammatory response that includes the recruitment of leukocytes (Levraud et al., 2008; Nguyen-Chi et al., 2014). We also observed an accumulation of GFP-expressing immune cells (Fig. 6C, right panels) in the inner ear suggesting that NF-κB signaling is strongly activated in response to local infection.

One advantage of the Tg(8×Hs.NFκB:GFP,Luciferase) fish compared to previously available zebrafish NF-κB-responsive reporters is the option for quantification of NF-κB activity by bioluminescence. We measured bioluminescence in vivo with a previously published method that is non-invasive and is easily scalable (Lahiri et al., 2014). Briefly, living zebrafish larvae were assayed in 96-multiwell plates by adding beetle luciferin potassium salt solution directly to the water. In the first 4 dpf, concomitant with the broadening of the GFP expression pattern (Fig. 4), luciferase expression increased ∼10-fold (Fig. 7A). Treatment with JSH-23, a drug that has been previously used to inhibit NF-κB activity in zebrafish embryos (He et al., 2015), significantly reduced luciferase levels at 2 dpf (Fig. 7B). Next, we measured the luciferase activity in iκbαa morphants. Morpholino injection resulted in a significant increase in luciferase activity in 2 dpf embryos compared to both uninjected and control-injected embryos (Fig. 7C). These experiments confirmed that the palindromic sequence drives expression of both GFP and luciferase genes, and that the bioluminescence can be used to quantify NF-κB activity in the living fish.

Both in vitro and in vivo data confirmed that pSGNluc is a sensitive reporter for monitoring NF-κB activation, likely due to the high binding affinity of NF-κB proteins for the κB site sequence used in this reporter. One could assume that the strong binding affinity of pSGNluc reporter sites might interfere with the expression of endogenous genes by outcompeting the transcription factor complex, especially in NF-κB target genes that exhibit weak transcriptional activity. To test this possibility in vitro, we performed a competition assay by transfecting HEK293T cells with different amounts of NF-κB-responsive plasmids, and measuring interleukin-8 (IL-8) production and luciferase activity 24 h after stimulation with 10 ng/ml TNFα. We observed a positive correlation between the amount of transfected NF-κB-responsive plasmids and luciferase activity (Fig. S4A). However, increasing amounts of transfected NF-κB-responsive plasmids did not negatively affect IL-8 secretion (Fig. S4B). Surprisingly, even a trend towards higher IL-8 secretion was visible. We next tested whether cytokine expression was negatively affected in the Tg(8xHs.NFκB:GFP,Luciferase) fish. We compared mRNA levels of cytokine genes in transgenic and wild-type siblings before and after inflammatory stimuli. We used UV irradiation to induce inflammation in zebrafish embryos (Banerjee and Leptin, 2014), since NF-κB signaling plays a major role in the induction of several cytokines during the inflammatory responses to this stimulus. As expected, Tg(8×Hs.NFκB:GFP,Luciferase) embryos exposed to UV light showed an increase of NF-κB activity (Fig. 8A). We used quantitative real-time RT-PCR (qPCR) to compare the mRNA level of selected cytokines in transgenic and wild-type siblings (Fig. 8B). The cytokine genes interleukin-10 (il10), il12b and tnfa are direct target genes of NF-κB (Banerjee and Leptin, 2014; Kennedy et al., 1997; Rivas and Ullrich, 1992). Consistent with a previous study (Banerjee and Leptin, 2014), il10 and tnfa, but not il12b, were significantly upregulated in UV-treated embryos (Fig. 8B). The results from qPCR showed that the cytokines were expressed at comparable levels in the transgenic and wild-type siblings. Taken together, these data suggest that the pSGNluc reporter does not significantly interfere with the NF-κB response in cell culture and zebrafish embryos.

We have established a new NF-κB-responsive reporter system. Compared to other NF-κB-driven luciferase reporters, pSGNluc has three major advantages. First, pSGNluc is inducible in response to inflammatory stimuli, such as TNFα, at levels comparable to other available reporters. Second, the reporter is significantly more sensitive to both the proinflammatory cytokine TNFα and the Nod1-specific inflammatory activator Tri-DAP than other commonly used NF-κB reporters. This is likely because the sequence and copy number of the κB site enhance the sensitivity of the NF-κB-responsive reporter (Matsuda et al., 2007). Third, the bi-directional property of the NF-κB-responsive promoter enables simultaneous expression of two reporter genes: GFP expression shows the kinetics and spatial aspects of a response in individual cells in real time, while the luciferase assay allows a quantitative evaluation of the global NF-κB activation throughout all stimulated cells or tissues. This dual reporter system allows simultaneous qualitative and quantitative monitoring of NF-κB activity in response to inflammatory stimuli, the integration of which can provide information beyond on–off activation. For example, the broad GFP expression and high luciferase levels observed in cells in culture after TNFα treatment shows that the majority of transfected cells responded to stimulation. In the case of Tri-DAP exposure, the luciferase measurement suggested that NF-κB signaling was not strongly induced after exposure; however, the mosaic GFP expression after treatment showed that some cells do respond strongly. It is not clear why only a subset of cells respond to Tri-DAP, but we suspect it may be related to differences in drug uptake. Tri-DAP enters epithelial cells via clathrin-mediated endocytosis, then, it requires endosomal processing and an optimal pH for its translocation into the cytosol (Sorbara and Philpott, 2011). This may be an inefficient process, explaining why penetration of Tri-DAP into the cytoplasm of HEK293T cells might not proceed uniformly in all cells.

Two transgenic zebrafish lines carrying NF-κB-responsive reporters have been generated previously (Banerjee and Leptin, 2014; Kanther et al., 2011). Both lines contained constructs with multimerized κB sites to trigger GFP expression. The κB site sequences used in these reporters are predicted to have lower binding affinities for NF-κB proteins than those used in the pSGNluc. One of these reporters (Banerjee and Leptin, 2014) does not show detectable levels of GFP expression during embryonic development and is less responsive to injury than the pSGNluc reporter. The second reporter was used to study the effect of microbiota on NF-κB activity in the intestine (Kanther et al., 2011). This reporter drives GFP expression in the lateral line and intestine during embryonic development, an activation pattern consistent with our own observations for the pSGNluc reporter. However, there are also some differences between the two lines, including the dynamic expression in the olfactory neurons and stronger sporadic expression in the epithelia, as well as the reduced expression in the pharyngeal arches seen for the pSGNluc line. The GFP expression seen with the pSGNluc reporter in the first 30 hpf is consistent with previously described early expression of NF-κB transcription factors (Correa et al., 2004) and reflected published spatiotemporal expression patterns of genes involved in NF-κB signaling at these stages, such as c-rel (Correa et al., 2004) and ikk1 (Correa et al., 2005). This early NF-κB activation matches what is known about the functional requirements for NF-κB signaling during embryogenesis: during gastrulation NF-κB activation is involved in coordinating the cell cycle and mesoendodermal cell movements (Liu et al., 2009) and lack of NF-κB signaling in the notochord leads to embryonic dorsalization and deformities (Correa et al., 2004). The requirement of finely tuned NF-κB signaling during notochord differentiation is also supported by the defects caused by high doses of iκbαa morpholino.

The weak NF-κB activity in keratinocytes and its upregulation in iκbαa morphants and in response to UV irradiation are also in agreement with what is known about the role of NF-κB signaling in keratinocyte inflammation and homeostasis (Wullaert et al., 2011). The skin is exposed to multiple environmental stimuli, and NF-κB signaling is required for the inflammatory response in this organ. Additionally, NF-κB is a regulator of keratinocyte proliferation and growth (Wullaert et al., 2011). The weak GFP signal in the skin is also in agreement with findings in transgenic mice carrying an NF-κB-responsive reporter (Carlsen et al., 2002). In mice, IκBα is expressed in the skin and prevents the nuclear localization of NF-κB proteins. Newborn mice lacking IκBα exhibit increased NF-κB activation in the skin, which triggers keratinocyte proliferation and epidermal hyperplasia (Klement et al., 1996). A similar phenotype was observed in RelA-knockout mice or after overexpression of a dominant-negative mutant version of Iκbα in the epidermal keratinocytes (Seitz et al., 2000; Zhang et al., 2004). The tissue-specific response to iκbαa knockdown as well as to local inflammation suggest that the GFP signal in the Tg(8×Hs.NFκB:GFP,Luciferase) embryos recapitulates the activity of NF-κB signaling in zebrafish embryos.

A novel aspect of our reporter line compared to the existing NF-κB-responsive reporters (Banerjee and Leptin, 2014; Kanther et al., 2011) is that it allows quantification of NF-κB activity by bioluminescence measurement. We showed that induction of luciferase in vivo correlated with the increase in NF-κB activity as measured by GFP expression, and that it responded to the knockdown of the NF-κB inhibitor iκbαa in the same way as GFP did. Our assays measure luciferase activity in each larva individually in such a way that the larva is unharmed and can survive for several days, allowing continuous tracking of reporter activity (Lahiri et al., 2014). This represents a significant improvement over a previously published method to measure NF-κB activity in zebrafish with a luciferase reporter in which lysis of the embryo was required for activity measurement (Alcaraz-Perez et al., 2008). Bioluminescence imaging of NF-κB expression in a luciferase transgenic mouse model has been used as a tool to screen anti-NF-κB drug candidates (Robbins and Zhao, 2011). The zebrafish is a model ideally suited for high-throughput, whole-organism drug screening (Zon and Peterson, 2005) through the use of fluorescence-based analysis both of reporter activity (Wang et al., 2015) and bacterial load (Ordas et al., 2015; Veneman et al., 2013). We therefore anticipate that the zebrafish transgenic pSGNluc will become a valuable tool to study many aspects of NF-κB signaling during development and inflammation in real time. The transgenic reporter could also provide a cost-effective improvement for high-throughput assays in drug discovery.

Multimerized NF-κB-binding sites with the idealized sequence 5′-GGGAATTCCC-3′ separated by 6 bp spacers were generated by oligonucleotide ligation. A fragment containing eight NF-κB-binding sites was inserted in the pSGHluc plasmid (Bajoghli et al., 2004). The resulting plasmid (pSGNluc) contains eight NF-κB sites flanked by two 260 bp long minimal CMV promoters. The bi-directional promoter drives the firefly luciferase gene with a 3′ UTR of the globin gene and an SV40 polyA at one side, and GFP with a bovine growth hormone (bGH) polyA at the other side. The sequence of the pSGNluc plasmid has been deposited at NCBI under the accession number KY129798. The β-galactosidase expression plasmid is described in Philpott et al. (2000). pGL4.32 was purchased from Promega and IL-8 luciferase reporter is described in Bowie et al. (2000). The original name of IgκB-Luci is (IgK)3-conaluc and is described in Munoz et al. (1994).

In vitro analysis was performed as described previously (Zurek et al., 2011). Briefly, cells of the human embryonic kidney cell line HEK293T were plated at a density of 30,000 cells per well in a 96-multiwell plate. Transfection mixes were prepared for triplicate assays. A total amount of 51 ng DNA with 13 ng of the respective NF-κB-responsive vector, 8.6 ng β-galactosidase- or mCherry-expressing plasmid, and 29.4 ng pcDNA3.1 was used per well. HEK293T cells were transfected with DNA mixes using X-tremeGENE^TM 9 DNA transfection reagent (Roche). Unless otherwise indicated, transfected cells were directly stimulated with either TNFα (Invivogen) or Tri-DAP (Invivogen) at the indicated concentrations, and measurements were performed after 24 h. For kinetic measurements, cells were transfected, incubated overnight and stimulated with TNFα at 10 ng/ml for 8, 6, 3, 2 or 1 h before cell lysis. For the measurement of luciferase and β-galactosidase activities, cells were lysed at 24 h post transfection. Cells were lysed in 100 µl lysis buffer (25 mM Tris-HCl pH 8.0, 8 mM MgCl₂, 1% Triton X-100, 15% glycerol) per well. To measure the luciferase activity, 50 µl of the cell lysates were transferred to a white non-transparent 96-multiwell plate and luciferase activity was quantified as relative luminescence units (RLU) in a multiplate reader (Enspire, Perkin Elmer) upon automatic adding of 100 µl luciferase substrate solution [lysis buffer containing 1.3 µM ATP and 770 ng/ml D-luciferin (Sigma)]. To measure β-galactosidase activity, the remaining 50 µl cell lysate was supplemented with 100 µl of 1 mg/ml o-nitrophenyl-β-D-galactopyranosid (ONPG) in 60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄ at pH 7.0 per well and incubated at 37°C for 30 min. Then, absorption was measured in a photometer at 405 nm (620 nm reference). β-Galactosidase activity was used to normalize luciferase activity (nRLU) in each well. Three independent experiments, each performed in triplicate, were conducted for each assay.

To determine IL-8 concentrations, supernatants of HEK293T cells without and with TNFα (10 ng/ml) stimulation were analyzed with the human CXCL8/IL8 DuoSet ELISA kit (R&D Systems) according to the manufacturer's protocol. The IL-8 concentration in unstimulated cells was below the level of detection.

All animal experiments described in the present study were conducted on embryos younger than 5 dpf under the rules of the European Molecular Biology Laboratory and the guidelines of the European Commission, Directive 2010/63/EU. The zebrafish strain used in this study was Danio rerio wild-type TL. The stable transgenic line Tg(krt19:dTomato) was described previously (Fischer et al., 2014). To develop the stable transgenic zebrafish line, pSGNluc plasmid was co-injected with I-SceI meganuclease enzyme in 1× I-SceI buffer (New England, BioLabs) into the blastomere at the one-cell stage as described previously (Aghaallaei et al., 2007). Three zebrafish founders were identified with similar GFP expression patterns. One founder was crossed with the wild-type TL strain and their progeny was used for this work. The transgenic zebrafish carrying the pSGNluc is named Tg(8xHs.NFκB:GFP,Luciferase)^hdb5 in accordance with the Zebrafish Nomenclature Guidelines (https://wiki.zfin.org/display/general/ZFIN+Zebrafish+Nomenclature+Guidelines) and was approved by the Zebrafish Nomenclature Committee.

Antisense morpholino (MO) for zebrafish ikbαa (5′-TGCGGCTCTGTGTAAATCCATGTTC-3′; He et al., 2015) and standard control morpholino were obtained from Gene Tools. They were prepared as 1 mM and 3 mM stock solutions in double-distilled H₂O, respectively. Morpholino at a concentration of 0.3 mM or 1 mM together with 100 mM KCl was injected into the blastomere of zebrafish embryos at the one-cell stage. Non-pathogenic Gram-negative Escherichia coli BL21 expressing HcRed (Aghaallaei et al., 2010) were microinjected into the transgenic pSGNluc embryos at 3 dpf as described previously (Nguyen-Chi et al., 2014).

UV treatment in zebrafish embryos was performed as described previously (Banerjee and Leptin, 2014). Briefly, embryos at 24 hpf (n=50) were sorted into a petri dish and exposed to 24 mJ/cm² of UV. Embryos were maintained in E3 medium at 28°C for 24 h prior to total RNA extraction or luciferase measurement.

JSH-23 (Selleckchem Inc) was diluted to a stock concentration of 10 mM in DMSO, aliquoted and stored at −80°C. For treatment, the drug was added directly to the E3 medium of manually-dechorionated 1 dpf embryos to a final concentration of 100 µM. Luciferase activity was measured 24 h after drug exposure.

For FACS analysis, HEK293T cells were plated in 24-well plates, directly transfected with pSGNluc and pmCherry plasmids in a 1:1 ratio, and stimulated with TNFα or Tri-DAP at the indicated concentrations. 24 h later, cells were trypsinized, resuspended in 5% fetal calf serum (FCS) in PBS and analyzed with the FACSCanto system (BD Biosciences). To image mCherry and GFP expression in vitro, HEK293T cells were plated on glass coverslips, transfected and stimulated as above. After 24 h, stimulated cells were fixed with 4% paraformaldehyde in PBS for 20 min. After a short wash in 1×PBS, fixed cells were mounted with Mowiol4.88 (Sigma) supplemented with bis-benzimide to stain the nuclei. For live-cell imaging, HEK293T cells were plated in compartmentalised glass-bottom petri dishes (Greiner bio-one) before transfection and stimulation. Images were taken on a Zeiss Axiovert 200 M equipped with a Plan-Apochromat 20×0.8 NA M27 objective and AxioCamMR3 camera (Carl Zeiss, Jena) or a Leica DMi8 equipped with a FluotarL 20×0.4 NA objective (Leica) and Orca Flash 4.0LT camera (Hamamatsu).

To image GFP expression in vivo, transgenic pSGNluc zebrafish embryos were anesthetized with tricaine methanesulfonate and mounted in 1.5% low-melting-point agarose. In vivo imaging of the embryos was performed with a Zeiss Lightsheet Z.1 and Zeiss LSM 780 NLO 2-Photon confocal microscopes. To visualize NF-κB activation upon infection or injury, time-lapse experiments were carried out with an Ultraview ERS spinning disk (PerkinElmer) confocal microscope using a 40× water-immersion objective. Images were analyzed with Imaris software as described previously (Bajoghli et al., 2015).

Single embryos were transferred into individual wells of a 96-multiwell plate (Nunc) in 100 µl E3 medium (without methylene blue), supplemented with 0.5 mM beetle luciferin potassium salt solution (Promega), and the plate was sealed using an adhesive Top Seal sheet (Packard). Bioluminescence from each embryo was then assayed at room temperature using a Top-count NXT scintillation counter (2-detector model; Packard).

RNA in situ hybridization of wild-type zebrafish embryos was performed as described previously (Bajoghli et al., 2009) using digoxigenin-labeled RNA antisense probe for iκbαa (accession number, BC068382; nucleotides 230–903) and iκbαb (accession number, BC050175; nucleotides 242–689).

Total RNA was extracted from pools of 25 embryos using TRIzol (Life Technologies) following the manufacturer's protocol. 1 µg RNA samples were treated with 1 µl RQ1 RNase-Free DNase (Promega) before first-strand cDNA synthesis with random hexamer primers and Superscript III Reverse Transcriptase (Thermo Fisher). The first-strand cDNA was directly used as a template in PCR reactions. qPCR was carried out using the SYBR Green kit (Applied Biosystems) on the ABI 7500 Real-Time PCR System. The data were analyzed in Microsoft Excel using the ΔCt method with β-actin as a reference gene for normalization. All primer sequences used in this study were described previously (Banerjee and Leptin, 2014).

Prism software (version 6, GraphPad Software Inc.) was used for graphing and statistical analysis. Unpaired, two-tailed Student t-tests were used to compare the means of different data sets.

We thank the Advanced Light Microscopy Facility (AMLF) at EMBL for continuous support and PerkinElmer and Zeiss for support of the AMLF. We are grateful to Nicholas S. Foulkes (Institute of Toxicology and Genetics, Karlsruhe, Germany) for providing the experimental materials for luciferase measurement in live zebrafish embryos, Stephen A Renshaw (University of Sheffield, UK) for providing the Tg(NFκB:EGFP) fish and to Matthias Hammerschmidt (University of Cologne, Germany) for the Tg(krt19:dTomato) fish. We thank Francesca Peri for hosting our zebrafish, Sinja Kraus and Omnia El Said Ibrahim for experimental help; Yvonne Postma for technical support; Marvin Albert for help with the processing of SPIM data; Joanna Natalia Buffoni and Cornelia Henkel for the care of zebrafish. We thank Sanjita Banerjee for help in the generation of the zebrafish Tg(8×Hs.NFκB:GFP,Luciferase) line.