Matrix association region/scaffold attachment region: the crucial player in defining the positions of chromosome breaks mediated by bile acid-induced apoptosis in nasopharyngeal epithelial cells

Genetic alterations [1], epigenetic changes [2], and environmental factors [3] are thought to be involved in the development of nasopharyngeal carcinoma (NPC). Several environmental risk factors that contribute to NPC have been identified. These include Epstein-Barr virus (EBV) infection [4, 5], dietary exposure to nitrosamines [6] as well as occupational exposure to smokes, wood dust, formaldehyde and intense industrial heat [6–8]. In addition, prior history of chronic nose and ear diseases (such as chronic rhinitis, sinusitis and otitis media) has also long been recognised as a risk factor for developing NPC [9–15]. Individuals with chronic rhinosinusitis (CRS), the inflammation of the nose and paranasal sinuses, have been shown to have a significantly higher risk of developing NPC as compared with the control individuals without CRS [15]. Although chronic inflammation of nose or ear has long been recognised as a risk factor for NPC, the underlying mechanisms by which this risk factor may contribute to NPC pathogenesis remain elusive.

Gastro-oesophageal reflux disease (GORD) is one of the major aetiological factors of chronic inflammation of sinonasal tract or ear [16–20]. GORD is caused by the flowing back of gastric duodenal contents into the oesophagus. It has been reported that the gastric duodenal refluxate may flow beyond the oesophagus. In turn, the gastric duodenal contents may affect the tracheobronchopulmonary tree, larynx, pharynx, sinonasal tract and middle ear [18, 21, 22]. The typical GORD symptoms such as heartburn and acid regurgitation may not be present in half of these patients [19]. Thus, these atypical manifestations of GORD are not only referred as extraoesophageal reflux (EOR) or laryngopharyngeal reflux [18, 23] but also as ‘silent reflux’ [19].

GORD is related to various inflammatory disorders. These inflammatory disorders include gastritis [24, 25], oesophagitis [26–28], laryngitis [29–31], pharyngitis [32, 33], post-nasal drip [34], otitis media [35–38] and asthma [39–41]. Moreover, the relation between CRS and GORD has increasingly received much concern [33, 42, 43]. It has been reported that individual with GORD has a significantly higher risk of developing CRS [44]. The prevalence of acid pharyngeal reflux in patients with CRS has been found to be higher than that in the normal controls (64% vs 18%) [42]. Seventy-eight percent of patients with CRS has been observed to have GOR [45]. Nasopharyngeal reflux has been demonstrated in both paediatric [46–49] and adult groups [34, 42, 43, 50].

Besides, GORD has also been related to various cancers. These cancers include stomach cancer [51, 52], oesophageal adenocarcinoma [53, 54], laryngeal cancer [55], pharyngeal cancer [56] and lung cancer [57]. Bile acid (BA), the major component of acid refluxate has been identified as a carcinogen in human malignancies (reviewed in [58]). It has been found that the levels of total pepsin and BA in the saliva of patients with laryngopharyngeal reflux were approximately three-fold higher than those of the normal volunteers [59]. It has also been reported that the levels of total pepsin and BA in the saliva of early laryngeal carcinoma patients were about four-fold higher than those of the normal controls [60]. In addition, BA has also been shown to have the carcinogenic effect in human hypopharyngeal squamous carcinoma FaDu cells through epithelial-mesenchymal transition (EMT) [61]. EMT is a major pathway related to cancer invasion and metastasis [62]. These observations suggested a potential role for biliary reflux in the pathogenesis of laryngeal and pharyngeal cancers.

There are strong associations among oxidative stress, inflammation and cancer [63–65]. Oxidative stress may activate nuclear factor-kappa B (NF-kappa B) [66] which plays a vital role in inflammatory response [67]. The activation of this transcription factor leads to the expression of genes involved in inflammation [66]. On the other hand, the inflammatory condition generates excessive reactive oxygen species (ROS) in cells. The free radicals may interact directly with DNA or interfere with DNA repair system. These, in turn, elevate the mutation rate in the inflammatory cells. Therefore, chronic inflammation predisposes the cells to neoplastic transformation. Cytokines have been found to be the important mediators that relate inflammation to cancer through oxidative stress [68]. It has been demonstrated that the combination of BA and acid triggered NF-kappa B activation in human hypopharyngeal epithelial cells. This, in turn, leads to overexpression of genes associated with antiapoptosis and oncogenic properties [69]. NF-kappa B pathway is well known to be a proinflammatory signalling pathway. This pathway is mainly activated by proinflammatory cytokines such as interleukin 1 (IL-1) and tumour necrosis factor-alpha (TNF-alpha) [70]. ROS are known to act as the messengers in NF-kappa B activation. It has been found that the anti-inflammatory cytokine IL-10 was able to inhibit the NF-kappa B activation in the stimulated macrophages via reduction of ROS [71].

It has recently been reported that the level of BA in the serum of NPC patients was significantly higher than that of the normal controls. The level of BA in the serum of NPC patients significantly inhibited the secretion of the IL-10 protein in CD4+ CD5- T cells [72]. IL-10 is suggested to have an anti-inflammatory role through reduction of oxidative stress induced by the proinflammatory factors. Treatment of Caco-2 intestinal epithelial cells with proinflammatory factors such as TNF-alpha, serotonin, adenosine and melatonin has been shown to induce oxidative damage in proteins and lipids. IL-10 was found to be able to reverse the oxidative damage by restoring the activities of antioxidant enzymes such as catalase, superoxide dismutase and glutathione peroxidise [73]. It has also been demonstrated that IL-10 inhibited hydrogen peroxide (H₂O₂) generation triggered by interferon (IFN)-gamma or TNF-alpha-activated macrophages [74]. Our previous study provided clear evidence that BA triggered oxidative stress in normal nasopharyngeal epithelial and NPC cells. The effect of BA in the induction of oxidative stress was enhanced by the acid [75]. These findings unravelled a possibility that oxidative stress provoked by the acidic gastric duodenal content may be a vital factor leading to the inflammation-induced carcinogenesis in the nasopharyngeal epithelium. It will be intriguing to investigate the relation between BA and proinflammatory or anti-inflammatory factors in the context of direct exposure of the nasopharyngeal epithelial cells to the refluxate.

In addition, BA-induced apoptosis has been suggested to be a possible mechanism underlying the pathogenesis of Barrett’s oesophagus, oesophagus adenocarcinoma and colon cancer [76–78]. Chromosomal cleavage is a hallmark of apoptosis. Initially, the chromosomal DNA is being cleaved and detached from their binding sites on the nuclear scaffold. The release of rosettes and loops of chromatin produces the high-molecular-weight (HMW) DNA of 200 to 300 and 30 to 50 kbp, respectively [79–81]. In the later stage of apoptosis, the HMW DNA is further degraded into internucleosomal DNA fragments of 180 to 200 bp [82, 83]. In our previous study, we demonstrated that BA was able to induce apoptosis in normal nasopharyngeal epithelial and NPC cells. We further demonstrated that BA-induced apoptosis resulted in chromosome breakages within the AF9 gene. These chromosome breakages were abolished by the caspase-3 inhibitor. Given that caspase-3 is the primary activator of caspase-activated DNase (CAD), our findings suggested that CAD may play an important role in mediating the chromosomal cleavages during BA-induced apoptosis [75].

It has been observed that the apoptotic nuclease CAD is closely associated with the nuclear matrix in the cells undergoing apoptosis [84]. Chromosomal DNA binds to the nuclear matrix through matrix association region/scaffold attachment region (MAR/SAR) [85]. It is plausible that when CAD cleaves the chromosomal DNA, it potentially cleaves at MAR/SAR. Thus, we hypothesised that BA-induced apoptosis may cause DNA breakages preferentially at MAR/SAR sites leading to chromosome rearrangement in NPC. Our study focuses on the AF9 gene which is located at 9p22 because 9p22 is one of the deletion hotspots in NPC [86]. In the present study, we performed in silico prediction of MAR/SAR within the AF9 gene. We demonstrated that the AF9 gene cleavage frequency within the SAR region was significantly higher in BA-treated cells as compared with the untreated control. By contrast, there was no significant difference in the AF9 gene cleavage frequency within the non-SAR region between BA-treated and untreated control cells. Our results suggest a role for MAR/SAR in defining the positions of chromosomal breakages mediated by BA-induced apoptosis.

Lately, the association between chronic inflammation of sinonasal tract and NPC has increasingly received much attention [15]. One of the major risk factors for the development of CRS is GORD [42, 43, 102]. It has been demonstrated that gastric duodenal refluxate can reach the larynx, pharynx, oral cavity, nasopharynx, nose, sinus, eustachian tube and middle ear. Repeated exposure to gastric duodenal content may result in localised inflammation of these regions [18, 20, 43, 103–106]. More recently, BA has been shown to cause cell injury and inflammation in the airway epithelium. Treatment of the immortalised human bronchial epithelial cells (BEAS-2B) with BA resulted in increased activity of proinflammatory cytokines (interleukin-8 and interleukin-6) [107]. The airways do not have intrinsic protective mechanisms as found in the oesophagus. Therefore, the tissues of airways are more vulnerable to acid-peptic injury as compared with the oesophagus. Due to this reason, it is conceivable that when the tissues of airways are repeatedly exposed to the gastric duodenal refluxate, the genotoxicity and mutagenicity of gastric duodenal content may also contribute to carcinogenesis in the airways [108].

By using flow cytometric analyses of phosphatidylserine (PS) externalisation and mitochondrial membrane potential (MMP) disruption, we have previously demonstrated that BA induced apoptosis in normal nasopharyngeal epithelial cells (NP69) and NPC cells (TWO4) [75]. We further demonstrated that BA-induced apoptosis triggered oxidative stress and caspase activity. These events, in turn, resulted in cleavages within the AF9 BCR. These cleavages were abolished by the caspase inhibitor, suggesting that these cleavages were mediated CAD. Our findings suggested that one of the potential mechanisms contributing to chromosome rearrangement in NPC could be BA-induced apoptosis, where CAD may be involved [75]. In the present report, we intended to investigate the relation between the positions of BA-induced chromosomal cleavages and the chromatin structure.

It has been known that the BCRs of the AF9 and MLL genes share similar structural elements. These structural elements include MAR/SAR, topo II cleavage site and DNase I hypersensitive site. The similarity in the structural features in the BCRs of the AF9 and MLL genes has been suggested to serve as recombination hot spots leading to the formation of the MLL-AF9 fusion gene in leukaemogenesis [88]. MAR/SARs are DNA sequences that are responsible for chromosomal loop attachment [109]. Topo II cleavage site and DNase I hypersensitive site frequently co-localise with MAR/SAR [109–111]. Therefore, we attempted to study the role of MAR/SAR in determining the positions of chromosomal cleavages in BA-induced apoptosis.

The targeted gene in this study was the AF9 gene located on the short arm of chromosome 9 at position 9p22, a common deletion region in NPC. The AF9 gene frequently translocates with the MLL gene at 11q23 resulting in the reciprocal translocation t(9;11)(p22;q23) in leukaemia [88]. The fusion of these two genes has been found to occur predominantly among the patients with de novo acute myelogenous leukaemia (AML). The MLL-AF9 fusion gene was less commonly observed in patients with acute lymphocytic leukaemia (ALL), with myelodysplastic syndrome (MDS) and with therapy-related AML (t-AML) [88, 112].

In the present study, possible MAR/SAR sites in the AF9 gene were predicted by using MRS. MRS is a bipartite sequence element which has been strongly associated with MAR/SARs. MRS is composed of two individual sequence elements which are found together within a distance of approximately 200 bp. However, when the DNA is wrapped around the histones, these two sequence elements exist at a location near the dyad axis of the nucleosome. Hence, they are found parallel to each other in MAR/SAR when the nucleosomes are positioned. This allows them to generate a protein binding site in MAR/SAR. van Drunen and co-workers have analysed more than 300 kb of DNA sequences from several eukaryotic organisms by using MRS. Their studies reported that all the MRS predictions were associated with the experimentally determined MAR/SARs [87]. MRS has been widely used in other studies and proved to be reliable [113–115].

The MRS predictions obtained in the present study were compared with the location of experimentally determined MAR/SARs reported in the previous studies [88, 89]. Strissel and co-workers have analysed exons 4 to 10 of the AF9 gene for MAR/SAR. In this region of 61 kb in length, two biochemically extracted MAR/SARs have been reported. These two MAR/SARs were indicated as SAR1 and SAR2. SAR1 is a 6.2 kb MAR/SAR found in intron 4. SAR2 is a 4.6 kb MAR/SAR spans through parts of introns 5 to 7 [88]. To the best of our knowledge, there is no previous report on the experimentally determined MAR/SAR for the AF9 region from exon 1 to intron 3. Four MRS predictions (MAR/SARs 24–1 to 24–4 in Fig. 1) are associated with SAR1. One of these four MRSs was located in a region < 1 kb centromeric to SAR1 (MAR/SAR 24–1 in Fig. 1), whereas the other three MRSs were found within SAR1 (MAR/SARs 24–2 to 24–4 in Fig. 1). One MRS prediction (MAR 27 in Fig. 1) correlates with SAR2. This MRS was found in a region < 1.5 kb telomeric to SAR2. Interestingly, all the MRS-predicted MAR/SARs were found in the introns of the AF9 gene. These results are consistent with those of the other studies which found that MAR/SARs occur more frequently in introns than in exons. This has been previously confirmed by both the experimental extraction [116, 117] and computational prediction [91].

Based on the in silico prediction and the previous studies which reported the experimentally determined MAR/SARs [88], the SAR and non-SAR regions were determined (Fig. 1). A study by van Drunen and colleagues (1999) showed that they never found a MRS which did not correlate with an experimentally verified MAR/SAR. However, their studies also revealed that not all biochemically defined MAR/SARs contain a MRS. Their findings suggested that there is at least a distinct type of MAR/SAR which does not contain a MRS [87]. Thus, in order to investigate if the region which was considered as a non-SAR region contains another type of MAR/SAR which was not predicted by MRS, we further analysed the AF9 sequence by predicting the presence of MAR/SAR using MAR-Finder and SMARTest.

The MAR/SAR analysis rules utilised in these two programs are different from the criteria used in the MAR/SAR prediction by MRS. MAR-Finder utilises statistical inference to predict the occurrence of MAR/SARs. MAR-Finder was developed by using the formulation of a set of biological rules based on the correlation of MAR/SAR with various DNA sequence motifs. These motifs include the TG-rich sequences, origin of replication (ORI), kinked DNA, curved DNA, AT-rich sequences and topo II sites. MAR-Finder has been shown to successfully identify MAR/SAR sites which correlate with those experimentally verified in the human beta-globin gene, PRM1-PRM2-TNP2 domains and human apolipoprotein B locus [92]. By contrast, SMARTest predicts MAR/SARs based on a density analysis of MAR/SAR-associated features described by a weight matrix library [91]. This MAR/SAR matrix library was mainly derived from the following MAR/SAR-associated patterns. Firstly, MAR/SARs have a minimum sequence length of 200–300 bp [118]. Secondly, MAR/SAR sequences are AT-rich [117]. Thirdly, MAR/SARs are associated with a few motifs. These motifs include ATTA, ATTTA, AATATT, ATATTT and AATATATTT [85, 118–121].

In order to evaluate the capability of SMARTest, Frisch and co-workers analysed six genomic sequences (three human sequences and three plant sequences) with a total of 310 kb in length by using SMARTest. These six genomic sequences contain a total of 37 experimentally determined MAR/SARs. Their studies showed that 19 of 28 SMARTest predictions were true positives (specificity = 68%). These 19 true positives only correlate with 14 of 37 biochemically extracted MAR/SARs in these genomic sequences (sensitivity = 38%), as some of the MAR/SARs contain more than one SMARTest predictions. For comparison, the authors analysed the same six sequences by using MAR-Finder. Twenty of 25 MAR-Finder predictions were true positives (specificity = 80%). These 20 true positives only correlate with 12 of 37 biochemically extracted MAR/SARs in these sequences, as some of the MAR/SARs contain more than one MAR-Finder predictions (sensitivity = 32%) [91]. Given that the MAR/SAR matrix library utilised by SMARTest was derived from AT-rich MAR/SAR, other MAR/SAR classes distinct from AT-rich class were not predicted by SMARTest. However, these MAR/SAR classes distinct from AT-rich class were found by MAR-Finder. Frisch and co-workers further found that some of the experimentally determined MAR/SARs which were not identified by MAR-Finder were detected by SMARTest. Hence, SMARTest and MAR-Finder were suggested to mutually complete each other in MAR/SAR prediction [91].

Due to a lack of report on the experimentally determined MAR/SARs for the AF9 region from exon 1 to intron 3 (approximately 220 kb in length), the sensitivity and specificity of MRS, SMARTest and MAR-Finder were unable to be compared in this study. Nevertheless, the comparison of accuracy among these prediction tools was not the aim of this study. The main purpose of using these different MAR/SAR prediction tools was to predict the MAR/SARs of different classes. By using MRS, SMARTest and MAR-Finder for MAR/SAR prediction, our findings suggested that the non-SAR region does not contain any MAR/SAR (Fig. 1).

Given that chromosomal cleavage is an initial event in both apoptosis and chromosome rearrangements, we employed nested IPCR to identify the chromosome breaks mediated by BA-induced apoptosis. Our findings showed that, for the SAR region, the cleavage frequencies in BA-treated cells were significantly higher than those in the untreated control. On the contrary, for the non-SAR region, there was no significant difference in the cleavage frequencies between the BA-treated cells and untreated control cells. These observations were true for both NP69 and TWO cell lines. However, in both untreated NP69 and TWO4 cells, the cleavage frequencies of the non-SAR region were significantly higher than those of the SAR region. By using CENSOR program, we found that the overall content of repeat elements in the non-SAR region is 3.0-fold higher (41.37% vs 13.81%) than that in the SAR region. Considering that no significant difference in the cleavage frequencies of the non-SAR region was found between the untreated and BA-treated cells, it seems not unlikely that the chromosome breaks in the non-SAR region were not mediated by BA-induced apoptosis. Rather, the chromosome breaks detected in this region were mostly spontaneous breaks due to DNA fragility contributed by these repeat elements. It is likely that repeat elements make the chromosome to be more prone to cleavage. These results are consistent with those of the other studies which reported a high proportion of repeat elements in common fragile sites, including FRA3B, FRA7G, FRA7H, FRA16D and FRAXB. These repeat elements include interspersed repeat elements, long terminal repeats (LTR), transposable elements, Mirs, L1 elements, L2 elements and Alu elements [122, 123]. Therefore, we concluded that MAR/SAR may play an essential role in mediating the gene cleavages in BA-induced apoptosis in NP69 and TWO4 cells at both neutral and acidic pH.

Knowing that topo II was involved in mediating illegitimate recombination [124, 125], we further analysed the SAR and non-SAR regions for topo II consensus sites. The topo II consensus sites were predicted by using an 18 bp consensus sequence [98, 99]. Our findings showed that the proportion of topo II sites in the SAR region was approximately 3-fold higher than that in the non-SAR region (1.41% vs 0.43%). These results seemed to reaffirm the findings of those studies which unravelled that MAR/SARs are the dominant sites for topo II binding and cleavage [109, 126].

In an in vitro system, topo II has been demonstrated to play a critical role in mediating DNA cleavages at acidic pH. Topo II has also been shown to be involved in mediating mutation and cytotoxicity induced by acidic pH in tissue culture models. These findings suggested that topo II-mediated DNA damage may lead to the development of cancers associated with gastro-oesophageal acid reflux [127]. In addition, previous studies have demonstrated that topo II is responsible for chromosomal loops excision in the early stage of apoptosis induced by oxidative stress. This initial event was followed by activation of nucleases leading to degradation of chromosomal DNA into nucleosomal DNA [128].

Our previous study has shown that BA and/or acidic pH induced apoptosis via oxidative stress in nasopharyngeal epithelial cells. In BA-induced apoptosis, we demonstrated that DNA cleavages within the SAR region occurred in a caspase-3-dependent manner, suggesting that CAD is responsible for these DNA cleavages [75]. Besides, we have also previously demonstrated that CAD cleaves the DNA preferentially at MAR/SAR sites during oxidative stress [97]. It has been observed that CAD was closely associated with the nuclear matrix of apoptotic cells [84]. It is conceivable that when CAD binds to the nuclear matrix during apoptosis, CAD preferentially cleaves the DNA at MAR/SAR sequences. It is possible that in BA and/or acidic pH-induced apoptosis, which involves oxidative stress, both topo II and CAD do play a role in mediating the DNA cleavages. The former may take part in mediating the cleavage of loop-sized DNA into HMW fragments whereas the latter may involve in mediating the degradation of chromosomal DNA into nucleosomal DNA. Therefore, our current findings which revealed that BA-induced apoptosis resulted in DNA cleavages within the SAR region may be explained by the close relation among topo II, CAD and MAR/SAR.

Sequencing of IPCR bands detected in the SAR region showed the positions of chromosome breaks within the AF9 BCR1 mediated by BA-induced apoptosis. The AF9 BCR1 is bordered by SAR1 and SAR2 [88, 89]. It is noteworthy that the positions of the chromosome breaks identified in the present study were highly similar to those previously reported in leukaemia patients. A few chromosome breaks were mapped within the AF9 region that was previously reported to translocate with the MLL gene in an ALL patient. This reciprocal translocation t(9;11)(p22;q23) resulted in the formation of MLL-AF9 fusion gene in the ALL patient [GenBank:AM050804]. Additionally, we identified a breakpoint which is identical with that identified in the ALL patient [GenBank:AM050804].

In summary, our current results reaffirm our previous findings that BA-induced apoptosis may cause chromosomal breakages in nasopharyngeal epithelial cells. In addition, our findings further implicate that MAR/SAR, which has a close association with topo II and CAD, plays a critical role in determining the positions of BA-induced chromosomal breakages. The positions of these BA-induced chromosome breaks shared high similarity with those identified in patients with leukaemia. Given that chromosomal breakage is an initial event of chromosome rearrangement and that cells may survive apoptosis upon compromised DNA repair, repeated exposure of nasopharyngeal epithelial cells to acid refluxate may contribute to genomic instability. The elevated mutation rate may in turn lead to the development of NPC. In order to clarify the relation between GOR and NPC, there are a few questions remain to be answered by further investigations: (i) Whether GOR may directly contribute to the pathogenesis of NPC through the cytotoxicity and genotoxicity driven by acid refluxate? (ii) Whether GOR may indirectly contribute to the pathogenesis of NPC through chronic inflammation of sinonasal tract (such as CRS) that has been recognised as a precursor of NPC? (iii) Whether both chronic inflammation of sinonasal tract and NPC share a similar underlying mechanism contributed by GOR? Nevertheless, our findings have unfolded the potential role of refluxed gastro-duodenal contents in contributing to NPC chromosome rearrangements.