The Nox2 inhibitor apocynin and the chemical antioxidant N-acetyl-L-cysteine (NAC) were purchased from Sigma (Deisenhofen, Germany). DPI, the p38 MAPK inhibitor SB203580, and the RAC1 inhibitor NSC23766 were obtained from Merck (Darmstadt, Germany). Recombinant human (rh) TGF-β1 was obtained from ReliaTech (Wolfenbüttel, Germany) and was used at a concentration of 5 ng/ml in all experiments.

The present research does not consist of any studies using human participants or animals.

Human PDAC-derived Panc1 and Colo357 cells were cultured as previously described (18). The generation of Panc1 cells stably expressing a dominant-negative mutant of MKK6 (MKK6-Ala) was previously described (19). For the stimulation experiments, cells in normal growth medium [RPMI with 10% fetal calf serum (FCS)] were seeded into 6-well plates (2×10⁵/well), 12-well plates (0.8×10⁵/well) or 24-well plates (0.4×10⁵/well) (all from Nunclon, Thermo Fisher Scientific, Dreieich, Germany) for the migration assays, protein analysis and qPCR, respectively. After cells had reached 80% confluence, they were starved overnight in medium containing 0.5% FCS. Cells were then treated with TGF-β1 for 24 h in the same medium in the presence or absence of the various inhibitors. These were administered to cells 30 min before the addition of TGF-β1.

Total RNA was isolated using PeqGold reagent (PeqGold, Erlangen, Germany) and further processed as recommended by the manufacturer. Synthesis of cDNA was carried out with M-MLV reverse transcriptase (Thermo Fisher Scientific). NOX protein subunit RNAs were either amplified with Taq polymerase (Life Technologies, Darmstadt, Germany) by standard endpoint PCR followed by agarose gel electrophoresis and ethidium bromide staining or by quantitative real-time RT-PCR (qPCR) on a CFX96 instrument (Bio-Rad, München, Germany) using SYBR-Green as previously described (18). The housekeeping gene TATA box-binding protein (TBP) was used for normalization. The human PCR primers used for amplification of NOX and PHOX mRNAs and TBP mRNA are listed in Table I.

Immunoblots were performed as described in detail previously (18,20). The antibodies used were anti-NOX4 (#14347-1-AP, Acris Antibodies), phospho-p38 MAPK (Thr180/Tyr182) (#4370), p38 MAPK (#3104), HSP90α/β (H-114, sc-7947) (all from Cell Signaling Technology, Frankfurt, Germany) and β-actin (Sigma). In some cases, a densitometric analysis of underexposed images was performed using the NIH ImageJ program.

Panc1 cells were seeded in 12-well plates and transfected on the next day serum-free for 4 h with Lipofectamine RNAiMAX (Life Technologies) and 50 nM of either NOX4 (a mix of 4 premade siRNAs, ON-TARGETplus SMARTpool #L-010194-00-0005; Thermo Scientific) (for sequences see Table I), RAC1 siRNA (a mix of 4 premade siRNAs, siGENOME SMARTpool reagent, #M-003560), or a siCONTROL nontargeting siRNA (both from Dharmacon via Biomol, Hamburg, Germany). Following removal of the transfection mixture, cells received standard culture medium and were allowed to recover overnight. Twenty-four hours after the first transfection, cells underwent a second round of transfection. In some experiments which comprised only one round of transfection, cells received Lipofectamine 2000 (Life Technologies) along with expression vectors for RAC1-N17, MKK6-EE (19) or empty vector (pcDNA3), alone or in combination. Forty-eight hours after the last transfection, cells were subjected to chemokinesis assay, immunoblot analysis, or qPCR.

Real-time cell analysis (RTCA) migration assays were performed with xCELLigence^® technology (Acea Biosciences, San Diego, CA, USA, distributed by OLS, Bremen, Germany) as described elsewhere (18,21). In brief, experiments were performed according to the manufacturer's protocol with modified 16-well plates consisting of an upper and a lower chamber separated by a porous membrane (8-µm pores). The CIM-Plate 16 (OLS) is analogous to a Transwell plate, except that the underside of the membrane is equipped with microelectrodes for impedance-based detection of migrated cells. Prior to assembly, the underside of the porous membrane was coated with 30 µl of collagen I (Sigma) to facilitate adhesion of the cells and thereby increase the likelihood of cell-electrode contacts. To start an experiment, the lower chamber was filled with 165 µl of serum-reduced medium (containing 1% FCS plus inhibitors/TGF-β1) and the CIM-Plate 16 was placed in the RTCA DP device and incubated at 37°C in 5% CO₂ for 1 h to equilibrate the medium followed by a measurement step for recording a background signal generated by cell-free media. Cells (30,000–60,000, overnight serum-starved, per well) were mixed and preincubated in 150 ml of culture medium containing 1% FCS and inhibitors/TGF-β1 at the same concentrations as in the lower chamber (chemokinesis setup). After 30 min, TGF-β1 was added and cells were seeded in the wells of the upper chambers. After cell addition, CIM-Plates 16 remained at room temperature in the laminar flow hood for 30 min to allow cells to settle onto the membrane. Phosphate-buffered saline (PBS) was added to the empty space surrounding the wells in order to prevent interference from evaporation. Each condition was performed in quadruplicate with a programmed signal detection (RTCA software version 1.2.1.; OLS) every 15 min for a total of 24–40 h, depending on the cell type and conditions.

The generation of intracellular ROS was detected with H2-DCF-DA (2′,7′-dichlorodihydrofluorescein-diacetate; Molecular Probes, Karlsruhe, Germany) by flow cytometry. H2-DCF-DA is first deacetylated to the non membrane-permeable H2-DCF-DA and then oxidized, emitting a fluorescent signal. For the ROS assay, Panc1 cells were seeded in 6-well plates (2×10⁵/well) with RPMI containing 10% FCS. When 80% confluent, the cells were incubated overnight in growth medium containing 0.5% FCS and on the next day were loaded with 10 mM H2-DCF-DA for 30 min. Following this incubation, the medium was replaced with fresh medium containing TGF-β1 (5 ng/ml) and further incubated for various times. Following trypsinization, cells were centrifuged and resuspended in 500 µl PBS and fluorescence was measured by flow cytometry (FACS; BD Biosciences, Heidelberg, Germany) at a wavelength of 488 nm. For each sample, the specific fluorescence from 10,000 cells was determined using Cell Quest Software from BD Biosciences.

Statistical significance was calculated using the non-parametric Mann-Whitney U test. Data were considered significant at p<0.05.

A study by Hiraga et al showed that NOX4-derived ROS signaling contributes to TGF-β-induced EMT in Panc1 cells (17). In order to test which NOX isoforms are expressed in PDAC-derived cells, we performed semi-quantitative RT-PCR analysis. Notably, Panc1 cells expressed NOX2, NOX4, NOX5, p22^phox, p47^phox, p67^phox and RAC1, while NOX1 and NOX3 were not detected, neither in the Panc1 cell line nor in freshly isolated peripheral blood monocytes used as control (Fig. 1A, left). A quantification of NOX2, NOX4, and NOX5 in Panc1 and Colo357 cells by qPCR revealed that NOX4 exhibited by far the strongest expression of all 3 isoforms with expression levels even higher than those in peripheral blood monocytes (Fig. 1A, right). Panc1 cells have been shown to respond to TGF-β1 with upregulation of NOX4 (17). To ascertain whether NOX4, NOX2 and NOX5 are also induced in other PDAC-derived cells we treated Colo357 cells for 24 h with rhTGF-β1 and determined their expression by qPCR. These cells exhibited a strong upregulation of NOX4 (6.1±3.2, p<0.05), a moderate induction of NOX2 (2.5±1.1, p<0.05) (Fig. 1B), but no induction of NOX5 (data not shown). Due to its higher expression (see Fig. 1A) and the greater response to TGF-β1 (Fig. 1B) and the demonstration that it is a major source for ROS production in Panc1 cells (17) we focussed on NOX4 in all subsequent experiments. Upregulation of NOX4 by TGF-β1 was also noted at the protein level (Fig. 1C). To reveal whether the TGF-β1 effect on NOX4 temporarily correlates with the relatively rapid induction of cell migration by this growth factor (compare Fig. 3), we measured NOX4 expression in Panc1 and Colo357 cells after short stimulation periods. As shown in Fig. 1D, NOX4 expression was strongly induced already after 1 h of TGF-β1 treatment and remained elevated until 8 h after TGF-β1 addition.

Since PDAC cells appear to express the necessary components of the nonphagocytic enzyme, we tested more directly in Panc1 cells whether TGF-β1 was able to generate ROS. Intracellular ROS formation was monitored over a period of 20 min using H2-DCF-DA, a redox-sensitive dye that readily diffuses into the cell where it is freed from the acetate groups by cellular esterases and converted into the highly fluorescent 2′,7′-dichlorofluorescein in the presence of ROS. We observed a rapid, although moderate and transient increase in ROS peaking at 5 min (Fig. 1E). In the presence of the NOX4/flavine oxidase inhibitor DPI, TGF-β1-stimulated ROS production was suppressed (Fig. 1E). Although DPI is not specific for NOX4, high expression of this isoform and low or absent expression of the other NOX isoforms suggests that the observed effect was due to inhibition of NOX4. The data showed that TGF-β1 can induce ROS generation in PDAC-derived cells and, based on its high expression and its responsivity to TGF-β1 suggest that NOX4 accounted for TGF-β1-stimulated ROS production.

We previously showed in a series of studies that TGF-β1 is a powerful driver of random cell migration in PDAC-derived cells (18,21). To evaluate the possibility that TGF-β acts through intermediate production of ROS to control this response, we employed the antioxidant N-acetyl-cysteine (NAC). This drug which has been previously shown to be a specific inhibitor of TGF-β signaling (22) at 0.1–10 mM final concentrations strongly and in a dose-dependent manner downregulated the migratory response to TGF-β1 in both Panc1 and Colo357 cells (Fig. 2). Notably, in Panc1 cells, NAC also inhibited the early and strong rise in migratory activity that was TGF-β1-independent up to ~4 h when the TGF-β effect initially became statistically significant (Fig. 2).

Above we showed that Panc1 cells produced ROS in response to TGF-β1 and that the ROS scavenger NAC can blunt TGF-β1-mediated chemokinesis. TGF-β is known to produce ROS through the activation of RAC1 and we previously demonstrated that RAC1 is activated rapidly (within 5 min) by TGF-β1 in Panc1 cells (19). However, the above data led us to hypothesize that other components of the NADPH oxidase multi-subunit complex, e.g. NOX proteins are involved in mediating the TGF-β1 effect on random cell migration in PDAC cells. Above, we showed that expression of NOX4 was rapidly induced by TGF-β1 (within 1 h) in both Panc1 and Colo357 cells and was closely correlated temporarily with the TGF-β1 effect on cell migration. To verify the involvement of NOX proteins, in particular NOX4, we treated cells with DPI, which at 10 µM strongly inhibited TGF-β1-induced chemokinesis in Colo357 (Fig. 3A, left panel) and Panc1 (Fig. 3A, right panel) cells. Notably, the effect of DPI was generally more pronounced in Panc1 cells compared to Colo357 cells (Fig. 3A). As control, we employed the small molecule NSC23766 which blocks activation of RAC1 by the guanine nucleotide exchange factors TRIO and TIAM1. NSC23766 efficiently reduced TGF-β1-dependent chemokinesis in both Colo357 (Fig. 3B, left panel) and Panc1 (Fig. 3B, right panel) cells.

We complemented these data with a genetic approach using an siRNA to NOX4. Notably, NOX4 silencing in Panc1 cells by siRNA transfection led to a markedly reduced chemokinetic response to TGF-β1 (Fig. 3C, left panel). However, the effect was not as strong as that resulting from siRNA-mediated depletion of RAC1 which was performed as an internal control (Fig. 3C, left panel), or from ectopic expression of RAC1-N17, a RAC1 mutant that is unable to bind GTP and upon ectopic expression can block RAC1 activation in a dominant-negative fashion (Fig. 3C, right panel). As observed above for NAC, NSC23766 (Fig. 3B, right panel), DPI (Fig. 3A, right panel), RAC1-N17 (Fig. 3C, right panel) and NOX4 siRNA (Fig. 3C, left panel) also effectively inhibited the early, TGF-β1-independent phase of strongly increasing migratory activity (here, 0–7 h) in Panc1 cells. TGF-β1-induced chemokinesis was also inhibited by apocynin, a selective inhibitor of NOX2 (23) (data not shown). Together, these data clearly showed that both RAC1 and NOX proteins, particularly NOX4, play an essential role in the regulation of random cell migration by TGF-β1 and are in keeping with our assumption that RAC1 cooperates with NOX proteins to control TGF-β1-driven chemokinesis.

The above results suggest that RAC1 and NOX4 are required for TGF-β1-induced chemokinesis of PDAC cells. Both RAC1 and ROS can activate p38 MAPK; however, whether p38 MAPK is also involved in TGF-β1-induced chemokinesis has not yet been demonstrated. To analyze this, we studied the effect of the p38 MAPK inhibitor SB203580 on TGF-β1-induced random cell migration of Colo357 and Panc1 cells in real-time cell migration assays. To this end, treatment of both cell types with SB203580 (10 µM) moderately and strongly, respectively, suppressed the chemokinetic response to TGF-β1 stimulation (Fig. 4A). Likewise, the migratory activity of Panc1 cells stably expressing MKK6-Ala, a dominant-negative inhibitor of MKK6-p38 MAPK signaling was reduced by 45% (p=0.00104) compared to the empty vector-expressing control cells (Fig. 4B). These data showed that the activation of p38 MAPK is crucial for TGF-β1-mediated chemokinesis and identified p38 MAPK as a downstream target of RAC1 and NOX4 in this process.

To analyze whether inhibition of NOX4 expression/function impacts TGF-β1-induced p38 MAPK activation, we performed phospho-immunoblotting with antibodies associated with p38 MAPK activation (Thr180/Tyr182). As shown in Fig. 5, transient transfection of NOX4 siRNA (Fig. 5, left panel), or RAC1 siRNA as control (Fig. 5, right panel), prevented the phosphorylation of these sites by TGF-β1. However, the expression of NOX4 was not affected by transfection of RAC1 siRNA (data not shown). These data showed that TGF-β1-driven activation of p38 MAPK is dependent on functional NOX4 and RAC1 protein. In addition, it is suggested that although NOX4 appears to function independently of RAC1, both proteins cooperate to induce p38 MAPK activation in order to mediate the TGF-β1 effect on chemokinesis.

The data to date suggest that RAC1 and NOX4 act upstream of p38 MAPK in promoting cell migration. To prove that more directly, we ascertained whether activating p38 MAPK signaling, e.g. by ectopic expression of a kinase-active mutant of MKK6 (MKK6-EE), the upstream activator of p38, overcomes the antimigratory effect of Rac1 inhibition on TGF-β1-induced chemokinesis. To this end, while RAC1-N17 potently inhibited basal and TGF-β1-driven chemokinesis in Panc1 cells, co-transfection of MKK6-EE was able to partially rescue the cells from the decline in migratory activity (Fig. 6). These data provide additional evidence for our contention that p38 MAPK acts downstream of RAC1 and NOX4 in promoting TGF-β1-dependent chemokinesis.