Cyclo(phenylalanine‐proline) induces DNA damage in mammalian cells via reactive oxygen species

Abstract Cyclo(phenylalanine‐proline) is produced by various organisms such as animals, plants, bacteria and fungi. It has diverse biological functions including anti‐fungal activity, anti‐bacterial activity and molecular signalling. However, a few studies have demonstrated the effect of cyclo(phenylalanine‐proline) on the mammalian cellular processes, such as cell growth and apoptosis. In this study, we investigated whether cyclo(phenylalanine‐proline) affects cellular responses associated with DNA damage in mammalian cells. We found that treatment of 1 mM cyclo(phenylalanine‐proline) induces phosphorylation of H2AX (S139) through ATM‐CHK2 activation as well as DNA double strand breaks. Gene expression analysis revealed that a subset of genes related to regulation of reactive oxygen species (ROS) scavenging and production is suppressed by the cyclo(phenylalanine‐proline) treatment. We also found that cyclo(phenylalanine‐proline) treatment induces perturbation of the mitochondrial membrane, resulting in increased ROS, especially superoxide, production. Collectively, our study suggests that cyclo(phenylalanine‐proline) treatment induces DNA damage via elevation of ROS in mammalian cells. Our findings may help explain the mechanism underlying the bacterial infection‐induced activation of DNA damage response in host mammalian cells.

Cyclo(phenylalanine-proline) (cFP) is produced by various bacteria such as Lactobacillus reuteri, Streptomyces sp. AMLK-335, Vibrio vulnificus, V. cholera, Pseudomonas aeruginosa and P. putida [6,7,[12][13][14][15]. A number of studies have been conducted to investigate the biological function of cFP. For example, cFP demonstrated an antibacterial function against diverse bacteria such as Escherichia coli and P. aeruginosa [16]. In addition, cFP from the L. plantarum strain was shown to be anti-fungal [17]. In some bacteria, cFP functions as a quorum-signal molecule. cFP from V. vulnificus was shown to induce the ompU gene, which is important for the pathogenicity of V. vulnificus [6]. In V. cholera, cFP attenuates the production of the cholera toxins by activating the expression of the regulatory gene leuO [7]. The L. reuteri RC-14 also produces cFP, which inhibits the quorum-sensing system in staphylococci, leading to repression of the expression of staphylococcal exotoxin toxic shock syndrome toxin-1 in the human vagina [15]. CDPs including cFP from P. aeruginosa promote growth of Arabidopsis thaliana seedlings through activation of auxin-regulated gene expression [18]. These results indicate that cFP may be an evolutionally conserved signalling molecule among bacteria or between prokaryotes and eukaryotes.
A few studies have reported the biological effects of cFP on mammalian cell differentiation and metabolism. When HT-29 colon adenocarcinoma cells are treated with cFP, cells are differentiated, most likely because of alternation of gene expression via increased cAMP response element-binding protein (CREB) phosphorylation and histone acetylation [5]. High concentration of cFP induces cell growth arrest and apoptosis through caspase-3 activation and Poly ADP ribose polymerase (PARP) cleavage in HT-29 colon cancer cells [19,20]. Interestingly, treatment of 10 lM cFP isolated from Strepto-

Neutral comet assay
Cells were combined with 1% low melting agarose at a ratio of 1:10 (v/ v) and immediately pipetted on to slide glass. The slide glasses were placed at 4°C for 30 min. and immersed in cold lysis solution (2% sarkosyl, 0.5 M EDTA, 0.5 mg/ml proteinase K, pH 8.0) at 4°C for overnight. After slide glass was washed with 19 Tris Borate EDTA (TBE) buffer at 4°C for 30 min., electrophoresis was performed for 40 min. at 0.6 V/cm. The slide glass was washed with DIW for 5 min. and 70% Ethanol for 5 min. The slide glass was dried and stained with propidium iodide (P4170, Sigma-Aldrich). DNA damage was analysed by examining at least 50 comets using the Comet Score program (TriTek Corporation, Sumerduck, VA, USA).

RNA interference (siRNA)
INT-407 cells were transfected with siRNA against human ATM (SV 1002; Bioneer, Daejeon, Korea) or control siRNA (sc37007; Santa Cruz Biotechnology) using an XtreamGENE siRNA transfection reagent (Roche, Mannheim, Germany). The efficiency of knock down of specific gene was confirmed with real-time PCR.

RNA-Seq
Total RNA was extracted using RNeasy mini kit (Qiagen, Valencia, CA, USA). The quality of the total RNA was evaluated using RNA electro- pherogram (Experion; Bio-Rad, Hercules, CA, USA) and assessing the RNA quality indicator. The resulting mRNA samples were processed for the sequencing libraries using the Illumina TruSeq Stranded mRNA sample preparation kit (Illumina, San Diego, CA, USA) following the manufacturer's protocols. One lane per 6 samples was used for sequencing by the Illumina HiSeq 2500 to generate directional, pairedend 100-base-pair reads. Quality-filtered reads were mapped to the human reference genome sequences, hg19 (UCSC Genome Bioinformatics, https://genome.ucsc.edu) using tophat2 (http://ccb.jhu.edu/software/tophat) [22]. The relative transcript abundance was estimated by counting the fragments per kilobase of exon model per million mapped sequence reads (FPKM) and differential expressed genes were evaluated using cufflinks package (http://cufflinks.cbcb.umd.edu) [23]. Gene Ontology categories with differential expressed genes were analysed by DAVID (http://david.abcc.ncifcrf.gov).

Real-time RT-PCR
First strand cDNA synthesis from total RNA template was performed with PrimeScript II 1st strand cDNA Synthesis Kit (Takara Bio, Otsu, Shiga, Japan). The resulting cDNAs were subjected to real-time PCR with a Stratagene Mx3000P (Agilent Technologies, Santa Clara, CA, USA) using TOP SYBR Green (Enzynomics, Daejeon, Korea). PCR conditions used to amplify all genes were 10 min. at 95°C and 40 cycles of 95°C for 15 sec., 60°C for 40 sec. Expression data were calculated from the cycle threshold (Ct) value using the DCt method for quantification. GAPDH mRNA levels were as used for normalization. Oligonucleotides were described in Table S1.

ROS measurement
After cells were harvested with trypsin-EDTA and washed with PBS, cells were stained with 5 lM dichlorofluorescein diacetate (DCFDA; D6883; Sigma-Aldrich) for 90 min. at 37°C or 5 M MitoSox Red (M36008; Molecular Probes, Eugene, OR, USA) for 30 min. at 37°C in the dark. Flow cytometry was performed with flow cytometer (FACSCalibur; BD Bioscience, San Jose, CA, USA) in the FL1 channel for DCFDA or FL3 channel for MitoSox Red. For cell imaging, cells were stained with 5 lM MitoSox Red for 30 min. at 37°C in the dark. After cells were fixed for 15 min. with 4% paraformaldehyde in PBS, cells were permeabilized with PBST solution (0.5% Triton X-100 in PBS) for 30 min. in the dark. Cells were coverslipped using VECTASHIELD mounting media plus DAPI. Images were acquired with a confocal microscope (Leica TCS SPE; Leica).

Mitochondria membrane potential measurement
Cells were stained with 5 lM Rhodamine 123 (R8004; Sigma-Aldrich) for 30 min. at 37°C in the dark. Flow cytometry was performed with Flow cytometer (FACSCalibur; BD Bioscience, San Jose, CA, USA) in the FL1 channel.

Statistical analyses
All quantitative data are presented as mean AE S.D. for three independent experiments. ANOVA was used for multiple comparisons. Significance values were *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.005.

Results
Induction of DNA damage by cFP treatment in mammalian cells cFP is produced from bacteria, fungi, plants and animals. It may function as an anti-cancer agent by induction of apoptosis or inhibition of DNA topoisomerase I [5,14,19,20]. To investigate the possibility that cFP treatment can induce DNA damage response in mammalian cells. We chose HeLa derivative INT-407 cell line as in vitro model because it has been used in the studies on DNA damage and pathogenesis of Vibrio sp., which produces cFP [6,7,[24][25][26][27][28][29][30][31]. We first measured the level of DNA damage with cH2AX foci formation using an anti-phospho H2AX (S139) antibody. We observed cFP-induced accumulation of cH2AX foci in concentration and incubation time-dependent manner ( Fig. 1A and Fig. S1). In particular, cH2AX foci formation was detected readily at 48 hrs after 1 mM cFP treatment. In contrast, untreated control (CTR) or linear Phe-Pro dipeptides (FP) did not induce cH2AX foci formation in any tested concentrations or incubation times (Fig. 1A). As a positive control, INT-407 cells were treated with 1 lM doxorubicin (Doxo), a DNA damage-inducing agent. To exclude the cell line specific response to cFP treatment, several mammalian cells including U2OS osteosarcoma and Huh7 hepatoma cells were tested. Although there was some variation in concentration and incubation time for cFP treatment, we consistently observed cFP-induced cH2AX foci formation (Fig. S2). We further confirmed cFP-induced DNA damage by Western blot analysis. While phosphorylation of H2AX (S139) was barely detected in linear FP dipeptides-treated INT-407 cells, increased phosphorylation of H2AX (S139) was observed at 48 hrs after 1 mM cFP treatment (Fig. 1B). To detect the physical DSB in cFP-treated INT-407 cells, a neutral comet assay was performed. As expected, we found a significant increase in DSB in 1 mM cFP-treated cells (Fig. 1C). However, we did not observe increased DSB in linear FP dipeptides-treated cells. Collectively, these results suggest that cFP treatment induces DNA damage such as DSB in mammalian cells.
Activation of DNA damage response in cFPtreated INT-407 cells DNA damage triggers a cellular signalling pathway that regulates the cell cycle and DNA damage repair through either ATR-CHK1 or ATM-CHK2 activation. Therefore, we next checked ATR and ATM activation in 1 mM cFP-treated INT-407 cells using anti-phospho ATR (S428) and anti-phospho ATM (S1981) antibodies respectively. Western blot analysis revealed that phosphorylation level of ATR (S428) did not change by cFP treatment as compared to untreated control (CTR) or linear Phe-Pro dipeptides (FP) treatment ( Fig. 2A). However, we observed increased phosphorylation of ATM (S1981) and CHK2 (T68), indicating that cFPinduced DNA damage response is occurred through the activation of ATM-CHK2 signalling pathway. In addition, we did not observe phosphorylation of ATM and CHK2 in FP dipeptides-treated cells ( Fig. 2A). We further confirmed the activation of ATM by co-localization of phos-  phorylated ATM (S1981) with cH2AX foci in 1 mM cFP-treated INT-407 cells (Fig. 2B). We also observed co-localization of the 53BP1 checkpoint protein with cH2AX foci in 1 mM cFP-treated cells (Fig. 2C).

Requirement of ATM in cFP-induced DNA damage response
Next, we tested whether cFP-induced DNA damage is dependent on ATM kinase, which functions as a DNA damage sensor and phosphorylates H2AX (S139). We first depleted expression of ATM by siRNA in INT-407 cells, and cells were treated with 1 mM cFP (Fig. 3A). Using ATM-depleted INT-407 cells, the level of DNA damage was determined with cH2AX foci formation after 1 mM cFP treatment for 48 hrs. We found decreased cH2AX foci formation in 1 mM cFP-treated cells as compared to control siRNA-transfected cells (Fig. 3B). We further confirmed a decreased level of H2AX (S139) phosphorylation by Western blot analysis (Fig. 3C). Given that kinase activity of ATM is critical for the DNA damage sensor and H2AX (S139) phosphorylation, we treated INT-407 cells with 1 mM cFP in conjunction with 10 lM ATM inhibitor (KU-55933) or DMSO solvent. As with ATM depletion by siRNA, we found decreased cH2AX foci formation by immunocytochemistry (Fig. 3D) and decreased H2AX phosphorylation (S139) by Western blot analysis (Fig. 3E). Collectively, these results suggest that ATM is required for cFP-induced DNA damage response such as H2AX (S139) phosphorylation.

Treatment of cFP alters the gene expression related to ROS production and scavenging
To gain insight into the molecular mechanism underlying cFP-induced DNA damage response in INT-407 cells, RNA-seq was performed with total RNA from 1 mM cFP-treated or -untreated INT-407 cells. A relatively small number of genes (38) were regulated in cFP-treated INT-407 cells (8 up-regulated and 30 down-regulated) as compared to untreated cells. Table 1 shows the top 8 up-regulated and 10 downregulated genes. When 38 regulated genes were analysed using GO (gene ontology), we found that some of the genes participated in GO including response to hypoxia and oxygen level (Table 1). Interestingly, some of the down-regulated genes such as MT1X, MT2A, ADM, ANGPTL4, CTSS, CYP1A1, PFKFB4, FABP3 and DKK1 are involved in regulation of ROS scavenging and production [32][33][34][35][36][37][38][39][40][41]. We confirmed the result from RNA-seq by real-time PCR (Fig. 4). These results imply that cFP may induce DNA damage by suppression of genes related to regulation of ROS scavenging and production in mammalian cells.

ROS-dependent DNA damage in cFP-treated INT-407 cells
Because cFP treatment suppresses a set of genes related to regulation of ROS scavenging and production in INT-407 cells, we determined ROS level in cFP-treated cells by flow cytometry. After INT-407 cells were treated with 1 mM cFP for 6, 12, 24, 48, 72 hrs, cells were further incubated with DCFDA for 90 min. and flow cytometry was performed. While we did not detect significant ROS production in untreated or FP dipeptides-treated cells, 1 mM cFP-induced ROS production in an incubation time-dependent manner ( Fig. 5A and Fig. S3A, compare the black, blue and red lines). We also found that ROS is produced before DNA damage is occurred, which is demonstrated by phosphorylation of CHK2 (T68), ATM (S1981) and H2AX (S139) (Fig. 5B and Fig. S3B). When cells were co-treated with 1 mM cFP and 0.5 mM N-acetylcysteine (NAC) ROS scavenger, increased ROS production by cFP treatment was abolished (Fig. 5A, compare the red and pink lines). Since cFP-induced ROS production and DNA damage response, we investigated whether NAC treatment attenuates DNA damage response by Western blot analysis. We found decreased DNA damage response, which is demonstrated by decreased phosphorylation of CHK2 (T68) and ATM (S1981; Fig. 5B). Immunocytochemical results further confirmed that NAC treatment results in decreased cH2AX foci formation (Fig. 5C). Finally, a neutral comet analysis confirmed that cFP-induced physical DSB formation is attenuated significantly by removal of ROS (Fig. 5D).
Since the mitochondrial respiratory chain is the main source of ROS production and superoxide is predominant ROS in mitochondria [42][43][44][45], we used MnTBAP, superoxide dismutase (SOD) mimetic, to reduce level of superoxide in cFP-treated cells [46,47]. To detect superoxide, MiotSox Red reagent also was used for flow cytometric anlysis and cell imaging. MnTBAP treatment efficiently reduced cFPinduced superoxide from mitochondria (Fig. S4A). In addition, MnTBAP attenuated cFP-induced DNA dmage response, which is assessed by immunocytochemistry, Western blotting, and neutral comet assay (Fig. S4B-D). By measuring cytochrome C release and  membrane potential, we further tested whether cFP impairs the mitochondria membrane. As expected, cFP treatment induced increased cytochrome C release (Fig. 5E) and disturbance of mitochondria potential (Fig. 5F). These results indicate that cFP may also induce DNA damage via ROS, especially superoxide, production in the mitochondria.  We next tested whether DNA damage repair is restored after washout of cFP in a cell culture. After INT-407 cells were treated with 1 mM cFP for 48 hrs, cell culture media was changed with new media without cFP and cultured for the indicated times. We found that washout of cFP decreases ROS production in INT-407 cells (Fig. 6A, compare the red and pink lines). In addition, immunocytochemistry clearly showed a decreased number of cH2AX foci formation at 24 hrs after washout of cFP (Fig. 6B). Consistently, Western blot analysis indicated decreased phosphorylation of H2AX (S139) after washout of cFP (Fig. 6C). We further examined whether DNA damage repair would restore after washout of cFP by a neutral comet assay. Parallel with the above results, DNA damage was significantly decreased (Fig. 6D). Interestingly, INT-407 cells treated with 1 mM cFP for 120 hrs retained the DNA damage repair capacity (Fig. S5). However, more than 8 mM cFP treatment for 48 hrs failed to restore DNA damage repair, showing decreased survived cells (Fig. S6A). Consistently, flow cytometric analysis demonstrated that 8 mM cFP treatment increases cell necrosis (Fig. S6B).
Collectively, we demonstrated that 1 mM cFP induces DNA damage such as DSB through elevation of ROS by suppression of a subset of genes related to ROS metabolism as well as perturbation of the mitochondrial membrane in mammalian cells.

Discussion
Naturally abundant cFP has various biological functions such as antibacterial activity, anti-fungal activity, quorum sensing, bacterial virulence, biofilm formation and plant auxin production [5][6][7][8][9][10][11]. However, a few studies have investigated the biological effects of cFP on mammalian cells. For example, cFP treatment induces cell growth arrest and apoptosis through caspase-3 activation and PARP cleavage in HT-29 colon cancer cells [19,20]. In addition, cFP induces cell differentiation in HT-29 cells [48]. Since cFP inhibits DNA Topoisomerase I activity based on an in vitro DNA relaxation assay [14], it would be interesting whether cFP induces DNA damage response in mammalian cells. In this study, we found that 1 mM cFP induces activation of DNA damage response such as CHK2/ATM activation. Furthermore, it induced DSB as determined by H2AX (S139) phosphorylation and a neutral comet assay. Although concentration of cFP tested in this study was relatively high, millimolar cFP seems to be the physiological concentration. In fact, some bacteria such as V. vulnificus produces approximately 1 mM cFP in culture [6]. In addition, millimolar concentration of cFP regulates diverse biological events such as quorum sensing, virulence factor production, auxin signalling, cell growth and differentiation in bacteria, plant and mammalian cells [6,7,15,[18][19][20]48]. However, more than 8 mM cFP treatment induced cell death. We also found that cFP induces DNA damage in various mammalian cells including INT-407, Huh7 and U2OS cells with different response kinetics. We believe that this variation may be caused by different cellular origins and contextures. Similarly, it has been reported that doxorubicin treatment induces DNA damage in different cell lines with different incubation time and concentration [49][50][51].
Our study further suggests that DSB was not formed immediately but slowly accumulated within 48 hrs after 1 mM cFP treatment, indi-cating that cFP may alter cell metabolism rather than inducing DSB directly. In addition, RNA-seq. analysis demonstrated that cFP treatment may induce DNA damage via elevated ROS production and/or decreased ROS scavenging. Most of the down-regulated genes such as MT1X, MT2A, ADM, CTSS, PFKFB4, FABP3 and DKK1 are involved in response to hypoxia and oxygen level and DNA damage. Among them, it has been reported that metallothioneins such as MT1X and MT2A protect against ROS toxicity by ROS scavenging activity in plant and animal cells [52][53][54][55][56]. Adrenomedullin (ADM) also has antioxidant activity in angiotensin II-induced ROS generation in vascular smooth muscle cells [57]. This effect may be because of the increased activities of glutathione peroxidase and reductase and maintained total and active reduced thioredoxin levels [34]. In addition, when Cathepsin S (CTSS) was inhibited, autophagy and intracellular ROS were induced in HONE1 nasopharyngeal carcinoma cells [36]. Depletion of 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (PFKFB4) resulted in a significant decrease in the levels of Dihydronicotinamide-adenine dinucleotide phosphate (NADPH) that maintain cellular stores of reduced glutathione antioxidant, leading to increased accumulation of ROS in prostate cancer cells [38]. Furthermore, depletion of Fatty acid binding protein 3 (FABP3) enhanced ROS production in P19 embryonic carcinoma cells [39]. Depletion of Dickkopf WNT signaling pathway inhibitor 1 (DKK1), an antagonist of Wnt/b-catenin signalling, induced intracellular ROS because of upregulation of ROMO1 (ROS modulator 1) expression in A549 cells [40]. Although decrease in some genes such as CYP1A1, ANGPTL4 and CCN1 resulted in the reduction of ROS production in specific situations [35,58,59], we have suggested that 1 mM cFP treatment may increase intracellular ROS production by dysregulation of genes related to ROS scavenging and production.
Consistent with RNA-seq results, cFP treatment indeed resulted in increased ROS production. In addition, elevated ROS was observed before DNA damage is occurred by cFP treatment (Fig. 5A and Fig. S3). Accumulating evidence indicates that ROS induces oxidative DNA damage by formation of oxidative DNA base adducts including 8-oxoG, leading to DSB formation [60][61][62]. Although cFP down-regulated genes related to regulation of ROS scavenging and production in mammalian cells, we further found that cFP treatment triggers mitochondria perturbation, which results in ROS production. In fact, it has been known that mitochondrial respiratory chain is the main source of ROS production and superoxides are predominant ROS in mitochondria [42][43][44][45]. Consistently, we found that MnTBAP, SOD mimetic, treatment abolishes cFP-induced DNA damage, indicating that superoxide may be the main ROS induced by cFP treatment (Fig. S4). However, we cannot rule out the possible existence of ROS-independent DSB formation by cFP treatment. For example, cFP inactivates the activity of DNA topoisomerase I that relaxes DNA supercoiling during replication and transcription [14]. In the case of the camptothecin DNA topoisomerase I inhibitor, it forms a cleavage complex and prevents the DNA re-ligation step, leading to DNA single strand breaks (SSBs) [63,64]. This cleavage complex collides with the replication fork during the S phase, resulting in conversion of SSB into DSB [65].  Given that millimolar cFP is produced in many pathogenic bacteria including V. vulnificus, and the fact that it plays an important role in regulation of virulence and biofilm formation, which may affect bacteria-host interaction [6,7,66], it is tempting to speculate that cFP itself functions as a virulence factor. Several reports have demonstrated that bacterial infections or pathogens induce DSB in host mammalian cells. For example, infection of pathogenic Escherichia coli of phylogenetic group B2 expressing a putative hybrid peptidepolyketide colibactin induces DSB and activation of DNA damage response, resulting in cell cycle arrest and eventually host cell death [67]. Similarly, Mycoplasma pneumoniae infection induces changes of protein expression related to oxidative stress and ROS production, leading to DSB in A549 cells [68]. In addition, Chlamydia trachomatis infection results in ROS production, ERK-dependent DSB formation and malignant transformation [69]. In contrast, Helicobacter pylori infection induces ROS-independent and ATM-dependent DSB formation, leading to genomic instability and gastric carcinogenesis, although no virulence factor or DSB inducing factor has been identified [70].
Collectively, we demonstrate that cFP induces elevation of ROS, especially superoxide by perturbation of mitochondria membrane as well as dysregulation of a subset of genes related to ROS scavenging and production. In turn, this event results in DNA damage such as DSB in mammalian cells (Fig. 6E). Our findings may explain the bacterial infection-induced activation of DNA damage response or carcinogenesis in host mammalian cells. However, it is necessary to identify the mechanism underlying cFP uptake, signalling pathway and interacting partners for the clarification of the biological function of cFP. In addition, it is important to develop bacteria that cannot synthesize cFP by destroying responsible genes such as nonribosomal peptide synthetases or tRNA-dependent cyclodipeptide synthases to test our hypothesis.

Supporting information
Additional Supporting Information may be found in the online version of this article: