IKKα regulates human keratinocyte migration through surveillance of the redox environment

During the early tissue formation stage of the cutaneous wound repair process, re-lining of wounds by basal keratinocytes restores an intact epidermal barrier (Arwert et al., 2012; Rieger et al., 2014). This step involves dynamic physiological and molecular changes to initiate migration over the provisional matrix and proliferation of keratinocytes from the surrounding epidermis (Usui et al., 2005; Watt, 2002). A number of soluble factors modulate wound repair (Gault et al., 2014; Gurtner et al., 2008; Niethammer, 2014), including the small reactive oxygen species (ROS) hydrogen peroxide (H₂O₂) (Loo et al., 2011; Niethammer et al., 2009). The H₂O₂ signal transduction cascade is evolutionarily conserved and acts via receptors, protein kinases, structural components and downstream transcription factor-dependent, post-translational and genomic mechanisms (Gough and Cotter, 2011). H₂O₂ is an immediate wound attractant signal for Drosophila hemocytes (Moreira et al., 2010), and H₂O₂ gradients within the wounded epithelium are crucial for phagocytic cell migration (Niethammer et al., 2009; Tauzin et al., 2014) through direct oxidation of the Src-family kinase member Lyn (Yoo et al., 2011). H₂O₂ also promotes the regrowth of peripheral sensory axons and their reinnervation of healing skin (Rieger and Sagasti, 2011), and it is essential for tail regeneration of Xenopus tadpoles (Love et al., 2013). Addition of exogenous H₂O₂ to wounds accelerates wound closure in mice (Loo et al., 2012; Roy et al., 2006) and in keratinocyte culture injury models (Loo and Halliwell, 2012; Loo et al., 2011; Pan et al., 2011). Thus, H₂O₂ appears to be critical for wound repair; yet the molecular understanding of this physiological function remains elusive.

At low concentrations, H₂O₂ functions as a second messenger whereby it oxidizes cysteine thiols with low pKa in signaling proteins, which are often found in catalytic domains of signaling enzymes (Claiborne et al., 1999). Oxidation of cysteine thiols stimulates the formation of sulfenic acid (sulfenylation), a highly unstable metabolite that can rapidly convert into other metabolites, such as sulfinic and sulfonic acid, or nitrosothiol (Leonard et al., 2009). A common metabolic process in which sulfenic acid participates is the promotion of intramolecular disulfide bond formation. This alters protein conformation and modulates the activation or inactivation of enzymes (Leonard and Carroll, 2011; Stone and Yang, 2006; Truong and Carroll, 2012a). As a consequence, H₂O₂ can modulate phosphorylation cascades within cells, often by activating kinases and deactivating phosphatases (Claiborne et al., 1999; Gough and Cotter, 2011).

The inhibitor of κB kinase α (IKKα) (encoded by CHUK) represents a candidate for oxidative regulation of H₂O₂ during migration. First, IKKα is structurally related to its homolog IKKβ (encoded by IKBKB), which has been shown to be oxidized by arsenite at a conserved cysteine residue in the kinase domain (Kapahi et al., 2000). Second, we have recently shown that IKKα is essential for keratinocyte migration in vitro and acts downstream of H₂O₂ (Lisse et al., 2016). Third, IKKα has a non-conventional role in repressing EGF and HBEGF promoter activity to control epithelial differentiation and prevent tumor formation during skin homeostasis (Hu et al., 1999, 2001; Liu et al., 2008; Marinari et al., 2008; Sil et al., 2004). For example, ablation of IKKα leads to increased keratinocyte proliferation and impairment of terminal differentiation (Hu et al., 2001; Sil et al., 2004). In addition, IKKα plays an important role in human cancers, as its dysfunction has been reported in squamous cell carcinomas of the skin (Liu et al., 2006). Studies have focused primarily on the role of IKKα during epidermal terminal differentiation, but whether it is involved in other cellular activities, such as migration, has however not been examined. In addition, it is unclear whether the keratinocyte-specific molecular and biological functions of IKKα can be modulated by oxidation, similar to IKKβ. Given our previous findings implicating IKKα in migration, we speculated that IKKα might serve as an intracellular surveilling protein to promote either migration or differentiation depending on the redox environment. We hypothesized that oxidation of IKKα can promote the de-repression of EGF promoter activity, which may stimulate migration via EGFR signaling.

To investigate cell-autonomous molecular mechanisms regulating H₂O₂-induced keratinocyte migration, we utilized a human epidermal keratinocyte line (HEK001) (Sugerman and Bigby, 2000). We first validated that HEK001 cellular responses were similar to those in primary keratinocytes. We found that these cells differentiated after treatment with high (2 mM) Ca²⁺ concentrations (Fig. S1A). PCNA (proliferation marker), KRT14 (basal keratinocyte marker) and AXIN2, an inhibitor of the Wnt pathway, were expressed during the first 7 days in vitro (DIV). KRT10 (early differentiation marker) and IVL (late differentiation marker) showed a lower level of expression by 24 h, which gradually increased until day 7 in vitro. These markers seemed to increase further when cells were treated with high Ca²⁺ concentrations. The increase in PCNA and KRT14 expression ≥7 DIV suggests that this cell line shows aberrant differentiation characteristics. Regardless, 7-DIV cultures, when treated with high Ca²⁺ (2 mM), showed increased stratification (Fig. S1B) and aggregation of terminally differentiated lipid-forming keratinocytes when assessed with Crystal Violet staining (Fig. S1C). 7-DIV cultures, moreover, failed to close scratch wounds even after 72 h (Fig. S1D), which correlated with reduced H₂O₂ formation at the scratch wound, as assessed with pentafluorobenzenesulfonyl-fluorescein (HPF) (Fig. S1E,F). These findings show that migration is initiated in undifferentiated keratinocytes that resemble more closely a progenitor-like population (at 2 DIV), consistent with in vivo characteristics of basal keratinocytes.

Using in vitro scratch wound assays prior to day 7 in vitro, we observed rapid formation of H₂O₂ (∼30 min) after single-pulse labeling with HPF, with strongest signal formation in two or three cell layers at the scratch margin (Fig. 1A,B). For our studies, characterization of ROS signal-to-noise ratios showed that 1 µM but not 4 µM was critical to detect differential H₂O₂ levels, whereas higher concentrations resulted in toxicity (Fig. S2A-C). Migration onset occurred around ∼6 h and closure was typically seen between 12 and 18 h, which was batch-dependent (Fig. 1C). Closure coincided with downregulation of H₂O₂ in cells at the scratch margin. Migration was blocked by treatment with 10 µM of the general ROS inhibitor diphenyleneiodonium (DPI) (Fig. 1D left panel), but this was not caused by cellular toxicity, as cells were viable when assessed with 6-carboxyfluorescein diacetate (6-CFDA) (Fig. 1D, right panel). Also, the NADPH oxidase inhibitor apocynin blocked scratch repair (Fig. 1E), confirming our results. To validate that H₂O₂ levels were reduced upon inhibitor treatment, we assessed cells that had been pre-treated with DPI and HPF for 30 min before scratching, resulting in a significant reduction of H₂O₂ at the scratch margin (Fig. 1F). Thus, DPI does not decrease cell viability (Fig. 1D) but impairs migration (Fig. 1D,E). We next analyzed whether addition of exogenous H₂O₂ at concentrations ranging from 0.001-100 µM enhanced scratch closure. While low concentrations increased scratch repair, high H₂O₂ levels were inhibitory (Fig. 1G). We further performed micromass assays to test whether keratinocytes at the periphery of the micromass foci, which we assumed to be less differentiated than those in the center (Bedal et al., 2014; Stott et al., 1998), generated H₂O₂ similar to scratch margin cells. This showed elevated H₂O₂ levels at the outer edges compared with confluent regions where less H₂O₂/HPF signal was measured (Fig. 1H,I). We further utilized human primary keratinocytes to assess H₂O₂ production upon scratching and found a similar response as in HEK001 cells (Fig. S2D). Quantification of the migration distance after H₂O₂ treatment using 0.1 and 100 µM H₂O₂ showed increased migration at low levels but not high levels, which blocked migration (Fig. S2E). These findings suggest a concentration-dependent role for H₂O₂ in keratinocyte migration.

To dissect out the specific role of H₂O₂ (as opposed to other ROS) during scratch closure, we co-administered DPI and exogenous H₂O₂. We first assessed the time until scratch closure and found that DPI alone decreased closure within 18 h by 73%, whereas low H₂O₂ concentrations restored closure to near endogenous rates (Fig. 1G). To monitor bona fide migration, we utilized Transwell chambers and observed a similar increase in migration with addition of low H₂O₂ concentrations (Fig. 1J,K). Low-level (0.1 µM) H₂O₂ significantly promoted migration in the presence of DPI, but migration was reduced in DPI-treated cells and those treated with 100 and 1000 µM H₂O₂. We also performed Transwell assays after addition of the proliferation inhibitor mitomycin C using various concentrations of H₂O₂ to see whether proliferation accounts for the observed increase in cell number (Fig. 1L). This confirmed that the observed increase was related to bona fide migration. Thus, the presence of low H₂O₂ concentrations specifically induces keratinocyte migration. Finally, to assess specific NADPH oxidases involved in H₂O₂ production, we monitored NOX1, NOX4 and DUOX1 transcription using quantitative PCR (qPCR). This revealed that NOX4, but not DUOX1 and NOX1, was significantly upregulated after scratch wounding at 12 h post scratch while remaining low in unscratched cells (Fig. 1M). Taken together, low concentrations of H₂O₂ promote migration of HEK001 keratinocytes, which leads to delayed upregulation of NOX4 mRNA.

We previously performed RNAseq and Ingenuity^® Upstream Regulator analyses to identify genome-wide transcriptional networks mediated by H₂O₂ and showed four major upstream networks with notable cutaneous associations (i.e. involving H₂O₂, EGF, FOXO1 and IKKα) following treatment of zebrafish with H₂O₂ (Lisse et al., 2016). Further sub-categorization of each network to define enriched pathways within them (Fig. S3) showed genes in each of these networks that functionally cluster into overlapping categories, including cell migration, defense response and wound repair. Genes involved in migration and which were differentially regulated in our data set included FOXO1, IGFBP-1, HMOX1, HSPA1L and ITGB4. All of those showed a dose-dependent increase in gene expression following H₂O₂ treatment of HEK001 cells (Lisse et al., 2016).

To investigate these networks further in H₂O₂-dependent keratinocyte migration, we pharmacologically blocked the function of EGFR, FOXO1 and IKKα–NF-κB. EGFR has a demonstrated role in migration and metastasis (Jones and Rappoport, 2014), and can be modulated through either direct oxidation (Paulsen et al., 2012) or phosphorylation (Loo et al., 2011). As expected, EGFR inhibition completely blocked injury-induced scratch repair (Fig. 2A,B). We next analyzed FOXO1 involvement in H₂O₂-dependent migration. Consistently, we observed an impairment of injury-induced cell migration after FOXO1 inhibition (Fig. 2A,B), but H₂O₂ production at the scratch margin was unaffected 30 min after scratching in FOXO1-inhibited HEK001 cells (Fig. 2C), suggesting an upstream regulatory function of H₂O₂. We did, however, observe a decrease in H₂O₂ levels at the unclosed scratch margin after 12 h of inhibitor treatment compared to controls. FOXO activity is tightly controlled by post-translational modifications, which mediate FOXO nuclear localization and degradation. We therefore assessed its subcellular localization after scratching and in the presence of DPI. Immunofluorescence studies showed rapid FOXO1 nuclear localization in keratinocytes at the scratch margin, which was attenuated by DPI (Fig. 2D). These results suggest that H₂O₂ regulates FOXO1 activity and nuclear localization/function during early wound repair. FOXO1 seems to modulate H₂O₂ levels at later stages, consistent with previous findings showing that it controls oxidative stress levels during wound re-epithelialization (Ponugoti et al., 2013).

We next assessed the role of IKKα in H₂O₂-dependent migration. Given that IKKα is best known for its regulation of the nuclear factor κB (NF-κB) pathway, which is implicated in oxidative stress responses (Gloire et al. 2006), we first determined if NF-κB was activated. Pharmacological inhibition of canonical NF-κB signaling by blocking RelA (p65) nuclear translocation (Shin et al., 2004) failed to impair scratch closure (Fig. 2A,B). We did, however, observe transient NF-κB1 p50 and p105 (two cleavage products encoded by NFKB1) nuclear localization at 6 h following scratch wounding (Fig. 2E). H₂O₂ treatment however did not promote nuclear localization (Fig. 2F), suggesting alternative functions in scratch margin cells, possibly induced by the oxidant-dependent phosphoinositide 3-kinase (PI-3K) ‘survival’ pathway (Sonoda et al., 1999). Also staining of NF-κB2 p52 and p100 (two cleavage products encoded by NFKB2) was predominantly cytoplasmic upon scratch wounding and H₂O₂ treatment (Fig. 2G,H). Thus, H₂O₂ does not appear to regulate NF-κB signaling during scratch wound repair.

Given the absence of NF-κB involvement and our findings that IKKα is a predicted target of H₂O₂, we wanted to determine whether IKKα plays an NF-κB-independent role in scratch repair. Using the IKKα and IKKβ inhibitory compound, Wedelolactone, we found that scratch wound repair was impaired only when using high inhibitor concentrations of 25 and 50 µM (Fig. 2A,B). To further investigate IKKα functions, we assessed its (sub)cellular localization using immunofluorescence staining. We observed its predominant expression in keratinocytes that were adjacent to the scratch margin, where it was localized in the cytoplasm and perinuclear regions (Fig. 3A,B). Nuclear localization was however promoted with DPI (Fig. 3A,B). In addition, nuclear IKKα was also predominant after treatment with 2 mM Ca²⁺ to promote differentiation (Fig. S4A), and in cells that had been cultured for prolonged periods (7 DIV) following treatment with DPI (Fig. S4B) and after EGF withdrawal (Fig. S4C). Following H₂O₂ treatment of unscratched keratinocytes, staining was predominant in the perinuclear regions but nuclear or diffuse in untreated cells (Fig. 3C). At a high H₂O₂ concentration (500 µM), we found IKKα staining to be more diffuse in the cytoplasm and occasionally nuclear (Fig. 3C), suggesting that IKKα localization is regulated by H₂O₂ in a concentration-dependent manner. To further establish that IKKα is involved in H₂O₂-dependent repair, we assessed scratch closure with and without exogenous H₂O₂ treatment using siRNA against IKKα. IKKα mRNA and protein was specifically and significantly knocked down (>60%) following siRNA transfection (Fig. S4D-F). This correlated with the absence of cell migration following scratching (Fig. S4G). In the presence of endogenous and exogenous H₂O₂, control-siRNA-transfected keratinocytes showed closure rates comparable to those of untransfected cells, whereas IKKα knockdown impaired scratch closure under both conditions (Fig. 3D). Overall, these findings indicate that IKKα is regulated by H₂O₂ and essential for scratch wound repair.

Having shown the importance of IKKα during H₂O₂-dependent scratch repair, we next asked whether IKKα is directly oxidized by H₂O₂ at a conserved cysteine residue in the kinase domain, as has been shown for the structurally related kinase, IKKβ (Kapahi et al., 2000). We first determined global protein oxidation following 2 h H₂O₂ treatment and upon scratch wounding using the oxidation (sulfenylation)-selective compound, dimedone, followed by immunofluorescence staining for sulfenic acid (S-OH)–dimedone complexes (Paulsen et al., 2012; Truong, 2012a) (Fig. 4A). First we analyzed unscratched keratinocytes that had been treated with exogenous H₂O₂ (0.1 µM) as control. Interestingly, this revealed cysteine oxidation specifically in cells that appeared to have adopted a migratory phenotype, as indicated by the formation of lamellipodia at the leading edge (Fig. 4B). Also the length:width ratio of sulfenylated cells (Fig. 4C) increased, indicating a migratory fate (Rieger et al., 2009) consistent with cell detachment from neighboring cells (Fig. 4D). Phalloidin staining further revealed the formation of stress fibers in cells following low H₂O₂ treatment, which was uncommon in untreated cells (Fig. 4E). These findings suggest that H₂O₂ either selectively diffuses into cells to induce protein oxidation and migration, or it is degraded in a subpopulation of keratinocytes that do not become migratory. We next assessed scratch-dependent sulfenylation, which was predominant in the cytoplasm and perinuclear regions of cells at the wound margin (Fig. 4F, top row). The localization of sulfenylated proteins was comparable to that of the H₂O₂-dependent localization of IKKα (Fig. 3A,C), indicating that IKKα may be a sulfenylation target. Intriguingly, sulfenylation was most prominent after ∼2 h, indicating that this process might be highly regulated. To validate sulfenylation specificity, we further treated keratinocytes with DPI and then performed dimedone staining, which showed diminished signals at the scratch margin (Fig. 4F, bottom row). To test whether sulfenylation is required for scratch repair, we treated keratinocytes with dimedone, which specifically and irreversibly reacts with sulfenylated cysteine residues (Seo and Carroll, 2011). Addition of dimedone at various concentrations before scratch wounding impaired wound closure in a dose-dependent manner (Fig. 4G). We found that the highest dimedone concentration (100 mM) promoted cellular toxicity, as assessed with 6-CFDA, and apoptosis, as assessed with Annexin-V staining (Fig. 4H,I).

To assess IKKα sulfenylation, we utilized DYn-2, a chemically modified dimedone, which through click-chemistry allows for biotinylation and isolation via biotin–streptavidin interactions (Paulsen et al., 2012; Truong, 2012a) (Fig. 4A). Treatment for 5 min with H₂O₂ at 10 µM, a concentration that promotes cell migration, increased proteome-wide sulfenylation when assessed with streptavidin-tagged horseradish peroxidase (HRP) in western blots (Fig. 5A, upper panel). The isolated sulfenylated fraction was further probed for IKKα, showing a 19-fold increase in sulfenylated IKKα after treatment with H₂O₂ compared to basal (untreated) levels that presumably contained endogenously oxidized and unoxidized IKKα (Fig. 5A, lower panel). GAPDH served as positive control due to its known function as an oxidoreductase (Truong and Carroll, 2012b), which was oxidized under both basal and H₂O₂ conditions (the latter showing a 2-fold increase). These results demonstrate that H₂O₂ directly oxidizes IKKα.

We next asked whether cysteine oxidation of IKKα is required for keratinocyte migration. We had already generated two Ikkα–tdTomato reporter constructs (wild type and mutated cysteine) using zebrafish Ikk1 (encoded by chuk), and therefore assessed these constructs in HEK001 cells. The Ikk1 sequence is largely conserved within the kinase domain when compared with human IKKα (Fig. 5B). Following transfection into HEK001, we monitored the localization of both constructs. The wild-type (full-length) form localized to the cytoplasm and became largely perinuclear following H₂O₂ treatment, whereas DPI treatment led to subnuclear staining (Fig. 5C). This is consistent with the cytoplasmic and nuclear localization of human IKKα under these treatment conditions when detected using antibody staining (Fig. 3). Intriguingly, cysteine-mutated Ikk1 localized to the nucleus with and without the presence of H₂O₂ (Fig. 5D,E), consistent with the idea that H₂O₂-dependent oxidation promotes cytoplasmic localization of IKKα, which is abrogated when the kinase-specific cysteine residue is mutated. To further determine the role of IKKα oxidation in migration, we monitored the motility of transfected cells over time. Because transient transfections precluded the formation of confluent monolayers and thus scratch assays, we performed assays with unscratched cells at ∼60-70% confluence. To validate this approach, we treated some wells with 0.1 and 100 µM H₂O₂ and compared HEK001 migration under these conditions. This showed a migration increase after low- but not high-level H₂O₂ treatment, which was also seen in scratched cells (Fig. 1). Comparisons of Ikk1-transfected cells showed that the average migration distance was highest in keratinocytes that had been transfected with full-length Ikk1–tdTomato in the presence of low H₂O₂ concentrations (Fig. 5F). Keratinocytes that had been transfected with cysteine-mutated Ikk1 (C179A) tagged with tdTomato migrated, however, in a manner similar to that of keratinocytes treated with high concentrations of H₂O₂ (100 µM), or DPI. Untreated control cells traveled intermediate distances, possibly due to effects related to reduced confluence (Fig. 5F). These results suggest that IKKα cysteine oxidation and cytoplasmic localization is required for cell migration, and they further indicate that IKKα may serve as an innate surveilling protein of the redox environment to regulate keratinocyte migration and differentiation.

Evidence so far suggests an important function for oxidized IKKα in keratinocyte migration. How this translates into downstream functions that promote migration is unclear. This could be mediated, in part, by the cis-repressive transcriptional binding activity of IKKα to the EGF promoter region (Sil et al., 2004), which regulates autocrine loops within keratinocytes (Liu et al., 2008). To assess the role of oxidized IKKα in EGF promoter activity, we performed chromatin immunoprecipitation (ChIP) using H₂O₂-treated keratinocytes. Using ENCODE and genome assembly for human coordinates (GRCh38; GenBank assembly accession GCA_000001405.15), we designed primers to target the predicted EGF promoter region in which IKKα may bind (Fig. 6A, left panel). ChIP results showed that endogenous IKKα bound to the EGF promoter region under unstimulated (no H₂O₂ treatment) conditions (Fig. 6A, right panel). In contrast, binding was reduced upon treatment with 10 µM H₂O₂, consistent with the idea that oxidation promotes EGF promoter activity by reducing IKKα nuclear localization. RNA polymerase II, a general marker for precursor RNA synthesis, showed reciprocal binding activity at the EGF promoter region. Consistent with these findings, siRNA knockdown of IKKα led to a >1.6-fold increase specifically in EGF transcripts, whereas IKKβ and IKKγ (IKBKG) transcripts remained unaffected (Fig. S4E), validating a functional relationship between IKKα and EGF. Similarly, we observed highly elevated EGF transcript levels upon treatment with low exogenous H₂O₂ (0.1 µM) concentrations but, intriguingly, not when cells were treated with high concentrations (100 µM) (Fig. 6B).

Since growth factor levels are regulated through rapid decay of their transcripts (Tang et al., 1997), we also investigated the potential role of H₂O₂ on EGF mRNA stability. Following pre-treatment with actinomycin D for 1 h, which produced steady-state EGF mRNA levels (Fig. 6C, 0 min time point), we observed within 30 min a transient increase and subsequent degradation of EGF mRNA at low H₂O₂ concentrations but not in the absence of H₂O₂. We also did not see this effect when assessing a known H₂O₂-independent control transcript, OVOL2 (Marinari et al., 2008) (Fig. 6C). This rapid effect of H₂O₂-dependent EGF mRNA accumulation may occur through the activation of non-genomic upstream signaling regulators – i.e. ERK1 and ERK2 (ERK1/2; also known as MAPK3 and MAPK1, respectively) (Nagashima et al., 2014), which we also observed to be upregulated at the scratch margin (Fig. 6D).

Our findings suggest that oxidized IKKα regulates EGF promoter activity, which is likely to regulate EGFR signaling given the activation of its downstream effector ERK1/2 at the scratch wound. However, EGFR could also be additionally activated by other mechanisms, such as oxidation in an EGF-dependent manner, as shown previously in squamous carcinoma cells (Paulsen et al., 2012). We therefore investigated H₂O₂-dependent EGFR activation using immunofluorescence staining and western blot analysis. This showed rapid transient phosphorylation of EGFR at residue Y1173 and internalization at the keratinocyte scratch margin, which dissipated under DPI treatment (Fig. 6E). We also assessed another previously shown EGFR phosphorylation site (Y1068) in squamous carcinoma cells (Paulsen et al., 2012), which did not show a difference in phosphorylation with or without H₂O₂ treatment (data not shown). Phosphorylation of EGFR at Y1173 also occurred under EGF withdrawal, suggesting wound-dependent autophosphorylation that could be mediated by endogenous EGF and/or H₂O₂ production (Fig. 6F). Quantitative western blot analyses revealed a ∼3-fold increase in phosphorylation of EGFR at Y1173 (normalized to total EGFR) after 1 h of treatment with low H₂O₂ concentrations (0.1 µM), and a ∼28-fold decrease with DPI (Fig. S4H). These findings indicate that EGFR activity in HEK001 cells is regulated by H₂O₂.

Long-term HEK001 cultures (i.e. 14 DIV) resulted in deregulated proliferative (PCNA) and basal markers (KRT14) in the presence of high Ca²⁺, as observed for other keratinocyte cell lines previously (Micallef et al., 2009). This hyperproliferation may be due to Ca²⁺ resistance, response to hypotonic stress or a lack of differentiation factors besides Ca²⁺ in the medium. Furthermore, it is possible that these differences also reflect heterogeneous culture preparations, as it is unclear whether HEK001 is a pure keratinocyte cell line. Since our studies were performed for relatively short time periods, these late changes should not have influenced our findings. We further show that similar to human lung epithelial cells, primary epithelial cells produce scratch-induced H₂O₂ and show increased migration upon H₂O₂ treatment, similar to HEK001 cells. Thus, our findings are also likely to be relevant to other epithelial cells types and to the in vivo context.

We assessed enzymes that are possibly required for H₂O₂ production given various reports in the literature. For instance, Drosophila and zebrafish embryos, and human lung and neonatal foreskin epithelial cells produce H₂O₂ via DUOX (Choi et al., 2014; Niethammer et al., 2009; Razzell et al., 2013; Wesley et al., 2007). NOX1 and NOX4, and the common subunit P22PHOX (encoded by CYBA) are, in contrast, activated in cancer cells (Ha et al., 2016; Ito et al., 2016) and have been associated with keratinocyte-specific activity (Chamulitrat et al., 2004; Nam et al., 2010; Sun et al., 2016). While DUOX1 and NOX1 expression remained unchanged during scratch wound repair, NOX4 expression was significantly upregulated, although relatively late – after 12 h. It is possible that the mRNA expression data may not fully reflect protein expression levels within cells since alternative regulatory mechanisms could be at play. For instance, regulation of subunits that stabilize Nox enzymes or an increase in RNA and/or protein stability may be sufficient to generate H₂O₂ over the course of multiple hours. At present, however, our studies do not permit conclusions about NOX4 involvement given that the expression kinetics of NOX4 do not match the expected H₂O₂ production profile after scratch wounding.

Our findings show that IKKα is an oxidation target of H₂O₂. IKKα is mostly known for its kinase-dependent regulation of NF-κB signaling, such as during cellular responses to stress (Israel, 2010). In the differentiated epidermis however, IKKα has additional kinase- and NF-κB-independent nuclear repressor functions (Fukazawa et al., 2010) to maintain skin homeostasis (Descargues et al., 2008; Hu et al., 2001; Liu et al., 2008). Our results are in line with these findings, suggesting that NF-κB independent mechanisms are involved in oxidation of IKKα, leading to keratinocyte migration. Previous studies have shown that the nuclear function of IKKα is associated with production of an unidentified soluble factor termed keratinocyte differentiation-inducing factor (kDIF), which can promote terminal differentiation (Descargues et al., 2008; Hu et al., 2001). In parallel, nuclear IKKα is also known to promote keratinocyte differentiation through cis-repression of Egf in mice (Liu et al., 2008). Our study, for the first time, demonstrates that the EGF repressive activity can be blocked by H₂O₂ through promotion of IKKα cytoplasmic accumulation that stimulates keratinocyte migration. We observed increased IKKα at localized subcellular regions of keratinocytes at the scratch margin. At this point, it remains to be shown how IKKα is regulated by H₂O₂. For instance, a steady-state flow of IKKα between the cytoplasm and nucleus might be present under homeostatic conditions, and oxidation of cytoplasmic IKKα may lead to cytoplasmic accumulation and reduced nuclear flow. Alternatively, nuclear IKKα might be oxidized, leading to active nuclear export. Indeed, it has been shown previously that H₂O₂ can be detected in the nucleus of mammalian cells and in whole organisms, such as Caenorhabditis elegans (Dickinson et al., 2011). Cytoplasmic sequestration of IKKα indicates alternative molecular functions that are independent of NF-κB signaling. It is known that cysteine oxidation can change protein conformation and enzymatic activity. Hence, in addition to dissociation from the EGF promoter, H₂O₂-modified IKKα may regulate signaling pathways using its kinase function, which would be in line with the observation that IKKα enzymatic inhibition impaired scratch closure, yet endogenous NF-κB inhibition did not. It is possible that oxidized IKKα actively sequesters NF-κB in the cytoplasm, which remains to be investigated. Alternatively, oxidized IKKα could potentially regulate other migration-promoting target genes through cytoplasmic sequestration of nuclear co-repressors (Fernández-Majada et al., 2007).

Our data shows that low but not high exogenous H₂O₂ concentrations induce EGF mRNA expression, correlating with the cytoplasmic accumulation of IKKα and the induction of migration. Our findings further suggest a concentration-related regulation of IKKα sulfenylation, which could be mediated by the availability of intermediate enzymes that control for the sulfenylation reaction. This would be in line with our finding that strongest sulfenylation was observed about 2 h after scratch wounding, despite H₂O₂ being produced as early as ∼30 min after wounding. Thus, this suggests that initially there is IKKα-independent EGFR signaling, which is consistent with the rapid transient H₂O₂-dependent activation of EGFR via phosphorylation of residue Y1173, and the temporal stabilization of EGF messenger RNA by H₂O₂. The stabilization could be achieved through lack of degradation of the messenger RNA. Based on these findings, we propose a model in which, initially, EGFR signaling is activated by direct oxidation, whereas oxidation of IKKα subsequently induces EGF expression through activation of the EGF promoter. This might ultimately lead to sustained EGFR signaling that is required for persistent migratory activity of keratinocytes (Fig. 6G). Further studies are required to determine the relationship between IKKα oxidation and EGFR activity, and the role of EGFR oxidation in basal keratinocyte migration. It is known that EGFR is overly active in many cancers and that IKKα genetic/epigenetic downregulation (Maeda et al., 2007; Park et al., 2007) and cytoplasmic sequestration (Marinari et al., 2008) can trigger oncogenic pathways, of which the regulators of sequestration remain unknown. Interestingly, to our knowledge, loss of IKKα is associated with hyperproliferation but not tumor metastases. By contrast, IKKα expression itself is known to promote a metastatic phenotype in some tumors (Luo et al., 2007). Thus, our findings may have even broader implications, as elevated ROS levels exist in tumors (Liou and Storz, 2010) and low but not high H₂O₂ levels may act directly on IKKα in the cytoplasm to regulate metastasis.

HPV16-transformed human epidermal keratinocytes [HEK001; American Type Culture Collection (ATCC), CRL-2404] were maintained in keratinocyte-serum free (KSF) medium (Gibco-Brl 17005-042) supplemented with 5 ng/ml human recombinant EGF, low CaCl₂ (0.06 mM) and 2 mM L-glutamine (without bovine pituitary extract). Cells were incubated under 5% CO₂ and a 92% humidified atmosphere at 34°C, and seeded (4×10⁴ cells/cm²) in tissue culture plates pre-coated with type I collagen (Gibco-Brl, R-011-K).

For RNA and scratch analyses, cells were cultured in 12-well plates (Corning, 3513) and 8-chamber glass-bottom dishes (In Vitro Scientific, C8-1.5H-N), respectively. The culture medium was changed every 2-3 days and the dishes were confluent by 2 DIV post-seeding.

Keratinocyte differentiation was initiated with 2 mM CaCl₂ and monitored for ∼2 weeks. The scratch assay was used to evaluate cell migration and wound recovery (Goetsch KP, Niesler CU, 2011). Cells were grown to confluence, replaced with EGF-lacking medium for 12 h, refreshed with complete medium, and glass Pasteur pipettes were used to make vertical scratches along the surface of the vessels. Wells were immediately washed with PBS to avoid re-attachment of disassociated cells.

Cells were treated with H₂O₂ and inhibitors as indicated in the text and in Materials and Methods under ‘Antibodies and chemical inhibitors’. For western blots, after H₂O₂ treatment, cells were lysed in RIPA buffer supplemented with phosphatase and proteinase inhibitors, and analyzed using the NuPage system (Invitrogen). For RNA expression analyses, wounds were generated using a comb-like device, and RNA was subsequently purified. For EGFR immunofluorescence studies (see antibody staining), cells were treated without EGF for >12 h and then scratched followed by supplementation with complete medium. For some assays, cells were pre-treated for 1 h with DPI to eliminate endogenous ROS.

Primary epidermal keratinocytes (ATCC; PCS-200-010) were grown in KSF medium, and scratch assays and treatment with H₂O₂ were performed according to HEK001 cell protocols.

HEK001 were pre-treated for 30 min with the H₂O₂ sensor HPF (Cayman Chemical, CAS: 728912-45-6), scratched and then detected by confocal imaging. HPF was maintained as a 1 mM stock in DMSO. Initially, the optimal signal:noise ratios were empirically determined using 0.1, 1, 4 and 10 µM concentrations. Each condition was tested in four separate wells with four replicate images/well. Cell viability and apoptosis were monitored using the 6-CFDA kit (ThermoFisher) and Annexin-V–Cy3 kit (APOAC, Sigma).

Assays were performed using Millicell with hanging Transwell polyethylene terephthalate inserts (8 μm pore size, Millipore, PIEP12R48) for impedance-based detection of migrated cells. 200 µl of serum-free medium that contained 5×10⁴ 12 h-unstimulated cells was added to the upper compartment, while the lower compartment was filled with 750 µl KSF medium with or without EGF, H₂O₂ or inhibitors. After 12-24 h, non-migratory cells remaining in the upper chamber were removed with cotton swabs. Migratory cells on the lower surface of the Transwell membrane were fixed (4% PFA) and stained with Giemsa/DAPI. Cells were imaged on an FV1000-confocal microscope (Olympus) and counted using ImageJ.

For mitomycin C experiments, a stock solution of mitomycin C (Sigma, M0503) was made fresh in sterile water. A working concentration of 10 µg/ml mitomycin C was determined by performing bioactivity assays. For migration assays, cells were co-cultured using Boyden Transwell chambers along with mitomycin C and H₂O₂, and the number of migrating cells post 12 h were quantified using DAPI staining.

HEK001 cells were expanded in T-25 flasks, trypsinized and resuspended in 1 ml of complete medium. Cells were transferred to 1.5 ml tubes and pelleted by centrifugation at 0.2 g. The medium was aspirated and cells resuspended in KSF-medium to perform cell counts. Cells were resuspended in a final volume [=number of wells×volume per well (in 24-well plates)] where the seeding volume was 30 µl per well at 0.5×10⁶ cells. Cells were placed at the center of each well, carefully placed into a CO₂ incubator and allowed to attach for 2-4 h. 1 ml of KSF+EGF was carefully added to each well. The medium was replaced every other day without disturbing the centered micromass. The H₂O₂ detection assay was performed after 4 days in culture.

Total RNA was purified using the RNeasy mini kit (Qiagen). RNA was reversed-transcribed using Superscript III reverse transcriptase (Invitrogen) and equal amounts of poly-dT and random hexamer oligonucleotides. Gene expression was normalized to human ACTB mRNA and analyzed using the comparative CT Livak method (Livak and Schmittgen, 2001) using BrilliantII SYBR^® Green qPCR Master Mix (Agilent) (see Table S1 for primer sequences).

For EGF mRNA stability assessment, cells were pre-treated with actinomycin D (2.5 µg/ml; Sigma-Aldrich) for 1 h, followed by treatment with 0.1 µM H₂O₂ before RNA extraction. Both untreated and H₂O₂-treated samples with actinomycin were compared to the starting samples on graphs.

Antibodies against human ERK1 (phosphorylated at T202 and Y204) and ERK2 (phosphorylated at T185 and 187) (Abcam, ab50011), EGFR (phosphorylated at Y1068; Abcam, ab32430), EGFR (ThermoFisher, MA5-12880), EGFR (phosphorylated at Y1173; Abcam, ab32578), sulfenic acid modified cysteine/2-thiodimedone (Millipore, AB330), forkhead box O1 (LSBio, LS-C123562), IKKα (Abcam, ab4111), NF-κB (p52 and p100 subunits) (ThermoFisher, PA5-27340) and NF-κB (p50 and p105 subunits) (ThermoFisher, PA1-14284) were used at a 1:200 dilution for immunofluorescence studies. AlexaFluor488-conjugated anti-rabbit IgG (Molecular Probes, Invitrogen) and Cy3-conjugated anti-mouse IgG (Millipore) were used as secondary antibodies at a 1:1000 dilution. Inhibitors were kept as stock solutions in DMSO for FOXO1 (Millipore, 344355; IC₅₀=33 nM), DPI (Sigma-Aldrich, 2926), NF-κB (Santa Cruz Biotechnology, JSH-23; IC₅₀=7.1 μM), IKK Wedelolactone (Millipore, 401474; IC₅₀=10 μM), EGFR (Millipore, 324673; IC₅₀=20 nM) and apocynin (Santa Cruz Biotech, sc-203321; IC₅₀=10 μM). DYn-2 was a kind gift from Kate Carroll (Scripps Institute).

HEK001 cells were fixed in 4% PFA (15 min) and permeabilized in 0.25% Triton X100 (10 min) at room temperature. Cells were blocked with 1%BSA with 10% goat serum and 0.1% Tween-20 in PBS (30 min) and incubated with primary antibodies (1:200) in blocking buffer overnight at 4°C. Cells were washed and labeled with appropriate secondary antibody (1:1000) and DAPI or Hoechst 33342 (1:5000). To detect sulfenylated proteins, keratinocytes were incubated for 2 h in DMSO or 2 mM dimedone (5,5-dimethyl-1,3-cyclohexanedione) (Cayman Chemical), scratch wounded or treated with H₂O₂ and further incubated for 2 h. Samples were fixed and stained with a polyclonal anti-dimedone antibody (EMD Millipore, 07-2139). The cells were washed and stained with anti-rabbit AlexaFluor488-conjugated secondary antibody (Life Technologies) for detection on a confocal microscope. To stain 7-DIV cells that have been differentiated with 2 mM Ca²⁺, cells were incubated in 0.2% Crystal Violet with 10% ethanol for 10 min, followed by imaging on an inverted stereomicroscope (Zeiss, Germany). Phalloidin–AlexaFluor647 (ThermoFisher) was used to assess the actin cytoskeleton. Cells were grown to confluency and either left untreated or treated with 0.1 and 1 µM H₂O₂ for 2 h. Cells were fixed in 4% PFA (15 min) and washed in PBS. Phalloidin staining was performed according to the manufacturer's protocol. The cells were washed in PBS and immediately imaged on an FV1000 confocal microscope (Olympus).

The alkyne probe DYn-2 [4-(pent-4-yn-1-yl)cyclohexane-1,3-dione; 5 mM] (gift from Kate Carroll, The Scripps Institute, San Diego, CA, and purchased from Cayman Chemical, 11220) was added to HEK001 cells for 1 h at 37°C. The cells were lysed in modified RIPA (200 U/ml catalase, EDTA-free protease and phosphatase inhibitors) and pre-cleared of endogenous biotinylated proteins using pre-equilibrated NeutrAvidin resin (Thermo, 29202). Probe-labeled proteins were detected using click-chemistry (Invitrogen, C10276) with solubilized 4 mM biotin azide (Invitrogen, B10184). Proteins were precipitated using methanol and electrophoresed. For whole proteome detection, the membrane was blotted with streptavidin–HRP and loading antibodies [Santa Cruz Biotech, IKKα (B-8), sc-7606; GeneTex, GAPDH, 124503]. For detection of specific oxidized proteins, biotin-azide-labeled proteins were immunoprecipitated using NeutrAvidin agarose beads, eluted with 8 M guanidine-HCL, dialyzed using a Slide-A-Lyzer^® MINI dialysis device (Thermo Scientific; 88401), concentrated using Amicon Ultra(3 K) centrifugal filters (Millipore, 500324) and re-probed for western analyses.

A 50 µM stock of Silencer^® Select validated siRNA oligonucleotides against human IKKα (Locus ID: 1147) (Ambion^®, Life Technologies; siRNA ID: s3077) was resuspended in nuclease-free water and stored at −80°C. For IKKα mRNA knockdown validation, HEK001 cells were grown to 60-70% confluence [1.6×10⁵ cells/well (12 well-plate)], and siRNAs (10-20 pmol) were transfected into cells for 24 h using the Lipofectamine^® RNAiMAX reagent and methodology (Life Technologies). Gene expression (qPCR) was investigated after 48 h transfection (see Table S1 for primers). Validation of protein knockdown was also performed at this time using an anti-IKKα (Abcam, ab4111) antibody in western blots. For scratch assays using siRNA-transfected HEK001, cells were trypsinized and then re-plated at full confluence in 8-well chambers at 48 h post-transfection. siRNA-transfected cells were scratched 12 h after re-plating and processed accordingly (two biological replicates performed in triplicates).

HEK001 cells were grown to 80% confluence and cultured for 12 h without EGF. EGF was reintroduced with or without 10 µM H₂O₂ for 30 min. In vivo crosslinking of chromatin was performed with 1% formaldehyde and quenched with glycine. Digestion and isolation of chromatin-bond DNA by immunoprecipitation (ChIP) was performed using a magnetic ChIP kit (26157, Pierce) based on the manufacturer's protocol. To assess physical binding of components to genomic sequences, RNA polymerase II (included in ChIP kit) and IKKα (sc-7606, Santa Cruz Biotech), ChIP-grade antibodies were used for immunoprecipitation. Subsequent PCR quantification was performed using SYBR™ green (600882, Agilent) with primers recognizing the EGF promoter and intergenic negative control DNA sequences (Table S1). The current genome assembly for human coordinates (GRCh38; GenBank assembly accession GCA_000001405.15) was used to design primers. The negative control was used to normalize for DNA content to calculate the enrichment of the regulatory EGF region according to the Livak method.

Zebrafish Ikk1 (Chuk) cDNA was purchased from Open Biosystems (Clone ID: 5914247, GE Dharmacon). Full-length (FL)-Ikk1 was amplified with Advantage 2 Taq Polymerase (Clontech) using forward primer 5′-GCTAGCTGTCAATATGGAGAAACCCCCT-3′ and reverse primer 5′-TAATGGATTTGGTACCGACAAACGCGCTGATTTA-3′. The amplified PCR products were cloned into pCR Blunt II-TOPO using the Zero Blunt^® TOPO^® PCR Cloning Kit (Life Technologies). To generate tdTomato fusion constructs, pCRTOPO-FL-Ikk1 and ptdTomato-N1 vector (Clontech) were digested with KpnI and NheI, and ligated. The Ikk1(C179A)–tdTomato variant was generated from FL-Ikk1–tdTomato using site-directed mutagenesis (GeneArt^® Site-Directed Mutagenesis PLUS Kit, LifeTechnologies). The constructs were subsequently cloned into a plasmid containing the zebrafish tp63 promoter (pT2KXIG_tp63:AcGFP, courtesy of Gromoslav Smolen, Harvard University, Cambridge, MA). First, GFP was removed via digestion with BamHI and NotI, and both sites were Klenow blunt-ended. The Ikk1–tdTomato plasmids were digested with SnaBI and SfoI and ligated into pT2KXIG_tp63 to generate tp63:FL-ikk1–tdTomato and tp63:ikk1(C179A)–tdTomato. HEK001 were transfected using Nanofectin (PAA, Q050-005) when 60% confluent in 8-chamber vessels. It was determined that 0.5 µg of DNA per 0.6 µl of transfection reagent was optimal.

Laser scanning confocal microscopy images were obtained using an inverted Olympus FV1000 or Zeiss LSM510 unit. z-stacks obtained from 1-3 µm/slice per sample and a speed of 12.5 (µs/pixel) were projected. All other parameters (e.g. pinhole diameter, gain, laser intensities) were kept constant. The series of projected ‘z-axis’ images were used to calculate average fluorescent intensity profiles per channel using the line series and box analyses tools in the Fluoview software v. 4.1 (Olympus). A minimum of 16 evenly distributed ‘scratch’ lines were averaged per image, or six evenly distributed boxes were averaged per image to generate average (background) fluorescence intensities. The fluorescence intensity was never saturated (maximum 4096 intensity level) during imaging. Samples within the same well were compared; i.e. scratched regions were compared to unscratched regions, and the background was corrected between the scratch margins and averaged unscratched ‘margin-free’ regions within the same well and vicinity. The line tool was used to measure length:width ratios for most elongated sulfenylated cells (a total of ∼60 per group). Cell–cell distances and fluorescence intensities were measured in Imaris (Bitplane) software using the line tool and spots function, respectively. Migration of Ikk1-transfected HEK001 cells was documented with time-lapse imaging for 8 h using an incubation chamber.

Statistical comparisons were made using Prism 6 software (GraphPad). As indicated in the figures, the unpaired Student's t-test with a 95% confidence interval was used to compare the means of two unmatched groups, assuming that the values followed a Gaussian distribution. For multiple comparison tests of three or more groups, one-way ANOVA with an α value=0.05 (95% confidence interval) and Tukey's multiple comparison post hoc tests were utilized to compare the means of each column. Significance is denoted with asterisks in the figures: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

We thank Elizabeth A. Brochu for technical assistance. We further thank Dr Kate Carroll (Scripps Institute) for providing reagents and intellectual assistance. We also thank Dr Andrew Chisholm (University of California San Diego) for critical review of the manuscript.