Alteration of protein expression and spliceosome pathway activity during Barrett’s carcinogenesis

In Barrett’s esophagus (BE), the normal squamous lining of the lower esophagus is replaced by gastric type columnar epithelium [1]. This condition is considered a consequence of chronic gastroesophageal reflux disease (GERD). Because BE is asymptomatic, it is most commonly diagnosed by endoscopy in patients with GERD symptoms [2]. It is, therefore, difficult to assess the prevalence for the general population, and a biased group of patients undergo endoscopy because of symptoms that are not necessarily related to BE [3]. Dependent on the scope and population of a study, the reported average prevalence of histologically confirmed BE is around 1.5% (0.1–9.0%) [3‐5]. BE is considered a premalignant precursor for esophageal adenocarcinoma (EAC), which might progress continuously through the sequence of low-grade dysplasia (LGD), high-grade dysplasia (HGD) and ultimately adenocarcinoma. It follows that both non-dysplastic BE and dysplastic BE are important risk factors for EAC [6]. The prognosis of EAC is poor; the 5-year survival rate is low, at 10–18% dependent on sex and ethnicity [7, 8]. EAC occurs predominately in males, with the highest rates in Western and Central Asia regions [9, 10], and is currently the sixth most frequent cancer, with the highest increase of incidence rate in the past 3 decades [11].

BE is diagnosed by the presence of endoscopically visible and histopathologically confirmed metaplasia [12]. The grade of dysplasia is strongly related to the risk of carcinogenesis [13] and defines the intensity of the required surveillance and treatment [14]. However, distinguishing between different grades of dysplasia is challenging and in the past resulted in low inter-observer agreement and variation in the assessment of risk of progression between studies [15]. As a consequence, the risk prediction of EAC solely on basis of the dysplastic grade is of limited reliability, potentially may lead to overtreatment [16]. The pathological progression from BE into EAC is associated with biological processes such as proliferation, tumor suppression, cell adhesion and inflammation. Molecules involved in these pathways might predict the development of EAC. A wide range of molecular markers have been studied, such as genomic alterations, epigenetic markers and proteins expression[17‐19]: DNA copy number variations and aneuploidy have been found to be altered in EAC [20], and regions of loss of heterozygosity have been identified as promising predictive markers for EAC [21]. Gains of chromosomes 7 and 17 determined by FISH have been found correlated with the grade of oncogenic progression; the detection rate of dysplasia improved when the assessment of these gains was added to cytology [22]. EAC is characterized by a high mutational burden due to genomic micro-satellite instability compared to other cancers [23]. The predictive power of mutational load is limited for EAC, because in non-dysplastic BE (NDBE), the mutational load is also already elevated [24]. Correspondingly, gene expression studies have shown that the transcription profile of BE is more similar to that of EAC than that of normal esophagus [25]. Alterations of driver genes and frequency of genetic events have been found associated with EAC development [26]. Overexpression of p53, determined by immunohistochemistry (IHC), to predict development of EAC has been intensively studied [27]. Loss of heterozygosity in chromosome 17p is linked to inactivation of the p53 tumor suppressor gene. This inactivation of p53 was found in a higher frequency in HGD patients compared to NDBE patients, and is associated with a higher risk of progression to EAC [28]. Consequentially, p53 immunostaining has been suggested as an adjunct molecule marker for the diagnosis of dysplasia in BE [29].

Far fewer studies comparing the proteomes of BE and EAC tissue have been conducted. Zhao and co-workers compared premalignant Barrett metaplasia tissues with esophageal adenocarcinoma tissues taken from the same six patients. 2D liquid chromatography protein separation and time-of-flight mass spectrometry (MS) identified 38 differentially expressed proteins, of which 20 correlated with mRNA expression levels; and validated by IHC (3 of 3 proteins positive) [30]. Elsner and co-workers used imaging MS to determine m/z profiles of metaplastic and carcinogen tissue areas in a set of fresh-frozen samples taken from 38 Barrett’s adenocarcinoma patients. They found 22 m/z species that were differentially expressed and identified six of these as proteins potentially involved in tumor development and metastasis [31]. Through an LC–MS analysis of NDBE, HGD, and EAC epithelium, Zaidi and co-workers determined a diagnostic 4-protein biomarker panel that was successfully evaluated in serum by an ELISA assay on an independent cohort to discriminate between GERD and EAC patients with an accuracy of 87% [32]. O’Neill and co-workers acquired by MS-based proteomics a set of more than 6000 proteins from EAC, normal esophagus and gastric tissue samples of seven patients. Around half of the proteins quantified in tumor samples were differentially expressed, and quantification was successfully validated by IHC staining of seven proteins [33]. Despite these efforts, so far, none of the potential markers has been further developed for application in clinical practice.

In this study, we conducted an analysis specifically on the epithelial cell compartment and the surrounding stroma to determine proteomic alterations related to Barrett’s carcinogenesis. Because the proportion of epithelial cells relative to all cells of a specimen as well as the proportion of dysplastic/EAC epithelial cells relative to all epithelial cells vary widely, laser capture microdissection (LCM) was chosen as an appropriate method to collect samples that are, from a microscopical perspective, sufficiently uniform in tissue volume and stage of disease [34]. Proteins were identified and quantified by label-free bottom-up proteomics using high-resolution LC–MS. Results were validated by IHC for two selected proteins. Knowledge about these proteins and the underlying functions and pathways might add another puzzle piece to the molecular mechanisms of Barrett’s carcinogenesis. This addition could ultimately help to accurately predict the risk of carcinogenic progression, and thus decide on the most effective treatment and disease management.

Tissue samples with different pathologic grades of the progression sequence from non-dysplastic Barrett’s epithelium, dysplasia and ultimately Barrett’s adenocarcinoma are characterized by high heterogeneity, and apparently non-dysplastic cells can be found in close spatial proximity to dysplastic and cancerous cells. In an earlier study, we used biopsies taken from non-dysplastic tissue to successfully analyze the proteomes of epithelial and surrounding stromal cells [34]. In the present study, however, using biopsies from dysplastic/EAC tissue did not prove to be a viable option for this analysis, because in the majority of cases not enough cells of the selected type and disease stage could be found. Moreover, making a clear histopathological diagnosis on the basis of a single fresh-frozen biopsy was often not possible. Therefore, for the sake of pathohistological confidence, we used fresh-frozen EMR specimen obtained by ER-cap resection as sample type to obtain biological specimen with high biological and clinical fidelity. As a consequence, sample sizes were small, but on the other hand, fewer samples had to be excluded for the reason of insufficient tissue with the targeted cell type or uncertain stage of dysplasia. Nevertheless, we still found different stages of dysplasia/EAC between different specimens of an individual but also within the same specimens, and staging coincided just partially between samples. This heterogeneity of dysplastic tissue in samples that are spatially very close to each other is probably a substantial factor for the common disagreement on grading of dysplasia reported in literature [49‐52]. For the above reasons, during the statistical analysis, we were constrained to categorize solely NDBE and dysplastic/EAC tissue, without further differentiating between the grades of dysplasia.

When comparing the protein abundances between non-dysplastic and dysplastic/EAC samples, we found more different protein identifications and a higher total protein abundance in dysplastic/EAC epithelial tissue; interestingly, this observation did not hold for stromal tissue. Because same volumes of epithelial tissue were collected for both non-dysplastic and dysplastic/EAC cells, higher protein abundance is likely to be a result of the higher density of the dysplastic epithelial compartment, probably due to the absence of goblet cells and abnormal cellular organization such as atypic and crowded nuclei, and irregular compacted growth. Because for statistically analyses normalized abundances were used that corrected for variation due to overall differences, the higher number of differentially up-regulated proteins could not be exclusively explained by an overall difference of total protein abundance. The increased number and abundance of proteins are therefore probably mainly related to the higher heterogeneity of dysplastic/EAC tissue. An overall up-regulation in number of proteins was also shown in earlier studies that used LCM sampling to investigate malignant epithelial cells in breast cancer [53] and cervix carcinoma [54]. The up-regulation in these two studies as well might be explained by the possibility of a higher density of tumor cells.

In the present study, up-regulation of proteins in dysplastic samples did not apply uniformly to all types of proteins. Secreted mucins and mucin-associated trefoil factors had lower abundance in dysplastic/EAC tissue compared to non-dysplastic tissue. This group of proteins has characteristic expression patterns that in general decline during the progression from BE into EAC. The group includes MUC2, characteristically secreted by goblet cells; gastric MUC5AC, expressed at the surface epithelium and the submucosal glands; and MUC6 and MUC5B, found inside the glands [55‐57]. Associated to mucins, too, is the family of trefoil factors, including the gastric tumor suppressors TFF1 and TTF2, which are co-localized with MUC5AC and MUC6, respectively; and TFF3, which is typically not secreted by gastric mucosa but, like MUC2, by goblet cells. Trefoil factors are essential in mucosal protection and repair, and decreased expression is associated with increased risk of dysplastic progression [58‐60]. The presence of TFF3 in samples taken by Cytosponges [61] further allows specific and sensitive diagnosis of BE [62, 63]. Our results are consistent with the expected expression profile, because we found MUC6, TFF1 and TFF2 significantly down-regulated in dysplastic samples. In contrast, significant abundance differences between dysplastic and non-dysplastic samples were not found for MUC2, MUC5AC, MUC5B and TFF3. The probable reason for this latter finding is that LGD samples, from the group of dysplastic samples, had abundancies comparable to those of the non-dysplastic group. These abundance patterns of mucins and TFFs reflect the origin of secretion [59]. TFF2 and MUC6, secreted from mucous neck cells of fundic glands, show reduced expression already in LGD tissue. MUC5AC and TFF1, which are expressed in gastric mucosa cells, were similarly expressed in LGD and non-dysplastic tissue. The extent of expressions of MUC2 and TFF3, which are expressed by goblet cells, in LGD samples was between that of non-dysplastic and dysplastic/EAC tissue. However, the detection of mucins and TFFs in dysplastic/EAC samples indicates moreover that the samples collected were heterogeneous in terms of dysplasia and EAC. Collagens as well had in general a lower expression in dysplastic/EAC epithelium. Still, two of these, the endostatin precursor and angiogenesis inhibitor COL18A1 [64] and COL16A1, were significantly lower expressed in dysplastic/EAC tissue. This latter finding may be related to the proportional reduction of extra-cellular matrix surrounding epithelial cells in response to the dysplastic growth of the epithelial compartment.

A group of up-regulated proteins—PAXX, TOP2A, TOP2B, and MSH6—are involved in the stimulation of cellular response to DNA damage. PAXX is executing ligation in damage repair as response to double-strand breakage (DSB), mediated by TOP2A [65, 66]. There is further evidence that mismatch repair (MMR) genes are involved in DSB repair [66, 67] and that MSH6 regulates NHEJ activity by interaction with Ku70 [68]. Up-regulation of MSH6 and other MMR genes, such as MSH2 and MLH, have also been reported in various types of cancer [67, 69]. Mutations of MMR genes causes micro-satellite instability, which in turn leads to increased mutation rates that can ultimately lead to cancer. Still, microsatellite instability is less common in BE-associated EAC [70, 71]. In the present study, the elevated levels of MSH6 determined in the discovery experiment were confirmed by significantly increased MSH6 IHC staining in dysplasia/EAC.

Exportin-5 (XPO5) transports micro-RNAs and proteins from the nucleus to the cytoplasm [72]. MicroRNAs are small non-coding RNAs that regulate gene expression by binding to mRNA during translation, which process is frequently dysregulated in cancer [73]. In normal, dysplastic and cancerous Barrett’s epithelium, discriminating micro-RNA signatures have been found for different stages of dysplasia of esophageal tissue [74]. In colorectal cancer, elevated expression of XPO5 is primarily found in the nucleus and correlates with advanced disease stage and poor prognosis. XPO5 has an oncogenic role because its down-regulation reduces the invasive capacities and cell proliferation [75]. In prostate carcinoma, a DNA micro-array analysis revealed that XPO5 was 1.6-fold up-regulated [76]. Another exportin, XPO1, showed distinct nucleic and cytoplasmic staining patterns associated to the Gleason score [77]. In the present study, we validated the elevated expression of XPO5 in dysplastic/EAC tissue by IHC staining. This revealed a significant increase in cytoplasmic and nuclear XPO5 staining in dysplastic/EAC tissue.

The most significantly up-regulated protein in our study, PRKCI, is a known genetic driver and genomic EAC biomarker [26, 79]. PRKCI is an oncogene that shows increased copy numbers in invasive tumors and has a locus in a commonly amplified region due to 9p loss of heterozygosity during progression of BE to EAC [78, 79]. CSNK2A and CSNK2B, which both were found significantly up-regulated, are subunits of the protein kinase CK2 (CSNK2), which has been associated with various cancer types, such as breast, lung, colon, and prostate cancer. CSNK2 is an emerging candidate for targeted therapy [80]. Deregulation of the regulatory subunits is suggested to promote various cancer types, and are considered potential biomarkers and therapeutic targets [81]. In a recent study, Xiao and co-workers found that CSNK2B attenuates the inhibition of NF-κB in hepatocellular carcinoma [82].

By gene set enrichment analysis, we examined the data to identify pathways potentially underlying the set of differentially expressed gene products. On the basis of the significant up-regulation of 19 proteins, the most significantly enriched pathways identified were the Processing of Capped Intron-Containing Pre-mRNA pathway and the sub-pathway mRNA Major Splicing, which both are part of the RNA metabolism super-pathway. These pathways stood out for their highly significant enrichment (FDR = 9e−9) and differed distinctly in this respect from other pathways (e.g., 4th pathway, T cell receptor signaling: FDR = 4e−6). Strong enrichment for spliceosome components was earlier found in a study by Francavilla and co-workers in epithelial ovarian cancer by using a mass-spectrometry-based proteomics approach [83]. A meta-analysis comparing four publicly available BE- and EAC-associated micro-array datasets, published by Nangraj et al. revealed that RNA metabolism and spliceosome are critical in the formation and development of EAC [84]. The manifold and complex associations of spliceosome and cancer have been reviewed in detail by Srebrow et al. and El Marabti et al. [85, 86] What it basically comes down to is that due to mutations and alterations of expression levels of the splicing factors, the cancer is able to affect splicing and, thus, is potentially able to promote the selection of certain splicing variants. This observation is relevant for the pathology of cancer because, on the one hand, the functions of a protein are often related to the splicing form and, on the other hand, alternative splicing affects more than 90% of human genes [87]. Thus, changes of splicing factors can affect the splicing isoform selection—and hence processes related to cancer [85]. Dysregulation of splicing has in multiple studies been linked to cancer development, involving both oncogene and tumor suppressor activities [88]. Jiménez-Vacas et al. showed that the up-regulation of splicing factor SF3B1 is associated with the expressions of oncogenic splicing variants and the progression of prostate cancer [89]. Highly relevant in the development of EAC is the expression of different p53 protein isoforms as a result of different TP53 splicing forms [90]. Equally relevant is that MYC, an EAC driver gene [26], regulates the splicing of selected genes via the activation of alternative splicing factors or components of the core spliceosome [91, 92]. Spliceosome core components have been suggested as potential therapeutic target in various types of cancer, such as lung, breast, ovarian and prostate cancer [89, 93]. In the present study, we found the highest up-regulation within the spliceosome for the cleavage and polyadenylation complex, due to elevated expression of CPSF1, CSTF1 and SYMPK, potentially leading to deregulation of alternative polyadenylation (APA), which further yields mRNA 3′ untranslated region (UTRs) isoforms with modified characteristics including oncogenic activities [94]. For example, an SNP in the 3′ UTR of TP53 that is transcribed as a consequence of APA (lengthening) forms a risk factor for different types of cancer, such as prostate cancer, glioma, and colorectal adenoma [94, 95]. The regulation of proliferation marker Ki-67, which we found up-regulated, is as well mediated by APA in breast cancer [96].

IHC staining of nuclear MSH6 and nuclear and cytoplasmic XPO5 was observed in all samples. This observation partly reflects the results of the MS-based discovery study, in which XPO5 and MSH6 were found almost exclusively in dysplastic samples—and not in non-dysplastic samples. The discrepancy between absence in the discovery experiment and presence in the validation experiment might be primarily due to limited sensitivity of mass spectrometric detection, but could also be related to limited selectivity of the antibodies used. Nevertheless, the latter consideration seems unlikely, especially in the case of MSH6, because we used clinically validated antibodies. Hence, the relatively high fold-change ratios of the discovery study relative to those found during IHC validation are likely a result of zero imputation during statistical analysis. It must be noted, however, that the significance determined was not affected by zero imputation because a non-parametric test was applied. Furthermore, a linear quantitative relationship need not necessarily be assumed between the IHC score and the actual protein concentration of the tissue or the protein concentration determined in the discovery phase by mass spectrometry. In summary, both proteins had significantly higher IHC scores in dysplastic/EAC tissue, and thus positively validated the result of the discovery experiment.

In summary, this study provides insights in the alteration of epithelial proteomes during progression from NDBE into EAC. We determined a set of differentially expressed proteins that overall are up-regulated in dysplasia/EAC and point to increased activity of DNA mismatch repair, micro-RNA transport and RNA splicing. We showed increased immunostaining of MSH6 and XPO5 to confirm these findings. The proteomic finding of associations of spliceosome and polyadenylation activity with dysplastic progression of BE confirms recent novel findings, and extends the current knowledge of Barrett’s carcinogenesis.