Autoantibody signatures defined by serological proteome analysis in sera from patients with cholangiocarcinoma

Cholangiocarcinoma (CC) is a primary liver tumour which results from the malignant transformation of epithelial cells in any portion of the bile ducts. The incidence of CC has increased to 18 % of all liver cancers during the past 30–40 years [1]. Its rate of incidence differs as a function of geographical region [2]. The prognosis for CC is poor and treatment options are very limited. This is partly due to its late diagnosis and onset of symptoms; because of this, the early diagnosis of CC using specific biomarkers remains an important challenge. Antigens such as carbohydrate antigens (CA) and carcinoembryonic antigen are released from digestive tract tumour cells, and their detection in the blood would be a valuable tool to diagnose the cancer and monitor its treatment. However, only a few of them appear to be specific biomarkers for CC [3, 4]. Moreover, standard techniques lack sufficient sensitivity, particularly during the early stages of the tumour process, thus hampering their use in routine practice.

Differences in protein expression between normal and tumour tissues have been studied extensively, including in patients with CC [5]. Studies have also targeted secreted proteins that are directly accessible in biological fluids in the context of CC [6, 7].

Other molecules directly accessible in biological fluids are autoantibodies (AAbs), and their presence has been reported in sera from patients suffering from a variety of malignancies [8]. The origin of the immune response in this setting is largely unknown, although it involves mutation, incorrect protein folding, over-expression and also post translational modifications which cause the neo-antigen to be presented to the immune system [8, 9]. Because circulating AAbs are produced in large quantities despite the presence of small amounts of the corresponding antigen, and because of their persistence and stability in sera, they may be of considerable diagnostic value. In the case of primary liver tumours, and more specifically CC, very few studies on AAbs to tumour associated antigen (TAA) have been reported.

We therefore focused in this study on identifying autoantibodies as biomarkers for the diagnosis of CC, using serological proteome analysis (SERPA) which integrates 2D electrophoresis, western blotting and mass spectrometry. A similar approach was recently reported by Rucksaken et al. [10]. By using antigenic preparations from other cholangiocarcinoma cell lines (CCLP1 and CCSW1), from the liver of healthy humans and from tumour and adjacent non-tumour liver tissues from patients with cholangiocarcinoma, we were able to highlight the heterogeneity of autoantigenic patterns reflecting the diversity of the immune response. The classification of autoantibody targets according to gene ontology analysis enabled the definition of two main patterns which depended on the antigenic extract used, i.e., antigens with structural activity if a tumour specimen was used for AAbs screening, and antigens with catalytic activity if non-tumour specimens were used. In this way, it was also possible to identify the most appropriate substrate producing the best sensitivity.

The design of the study is summarized in Fig. 1. Briefly, sera from a pool of ten normal subjects, and from eight patients with CC, were tested on two-dimensional immunoblots performed using two CC cell lines and human liver from healthy subjects. In addition, five CC sera were tested on these three substrates and also on immunoblots performed using tumour and non-tumour tissue from the same CC patients. Spots of interest were identified by mass spectrometry (MS), and autoantigens categorised according to the Gene Ontology project before further classification.

This study highlights the heterogeneity of autoantigen patterns that reflect the diversity of the immune response as a function of the serum tested, the different fractions used; thus, as previously, it has underlined the specific nature of the immune response in the setting of cancer [5]. Our data are in accordance with the recent report by Rucksaken [10] which demonstrated that CC sera contain AAbs reacting with enolase and heterogeneous nuclear ribonucleoprotein L. However, we were not able to detect AAbs to ribonuclease/angiotensin inhibitor 1 or to heat shock protein 70, as reported by the same authors. It was nevertheless not surprising to observe different immunoreactive patterns being displayed by CC sera on the different antigenic extracts used, probably due to the specific nature of the cancer cells involved or the technique employed. It is postulated that autoantibodies in cancer are induced by a breakdown in self-tolerance resulting from over-expression, mutations, changes to post-translational modifications or the truncation of proteins in a cancer cell [7]. One hallmark of cancer is genome instability, which can differ from one cell to another and of course from normal cells to cancer cells [8]. In terms of protein expression and modification, choliangocarcinoma cell lines differed from the five tumour extracts, which in turn differed from each other and also from non-tumour specimens. It follows that a protein may be abnormally autoantigenic regarding induction of the autoimmune process, and not be present in the substrate used for AAb screening.

Secondly, the hepatocytes and cholangiocytes present in the liver and cholangiocarcinoma cell lines have different metabolic activities or cellular specializations with differencies on the level of expression of antigenic targets.

Thirdly, the 2D electrophoresis technique used during our study involved a whole homogenate, implying a bias towards abundant proteins.

Added to the previous considerations, some of the proteins resolved were found in sufficient quantities to be immunoreactive when transferred to a nitrocellulose membrane, whereas others were not; for example, nuclear proteins other than histones.

Taken together, these considerations may explain the variability of the immunoblotting patterns we noted and which differed somewhat from the results obtained by Rucksaken [10].

The Gene Ontology classification of autoantigenic targets as a function of their origin revealed two patterns of molecular function: catalytic activity or structural activity (Table 2).

Identified proteins displaying catalytic activity were mainly found when liver was used as antigens, including normal liver or non-tumour tissues adjacent to the CC, and to a lesser extent to tumour-affected part of the liver. Hepatocytes are the principal site for carbohydrate metabolism, and three of the main targets we found (identified by at least 40 % of CC sera with the most appropriate antigenic extracts) were enzymes implicated more specifically in glycolysis and fructose-galactose metabolism: i.e., alpha-enolase, fructose biphosphate aldolase and glyceraldehyde3-phosphate dehydrogenase. Interestingly, these three enzymes were found in the secretome of a cholangiocarcinoma cell line HuCCA1 [6]. Furthermore, enolase and glyceraldehyde3-phosphate dehydrogenase were reported as being over-expressed through study of the HuCCA1 proteome [5]. Even more interesting, and mentioned above, was the report of enolase as an antigenic target for CC sera, using substrates other than those we used [10]. More generally, by probing a protein array with numerous sera from patients with a variety of cancers, increased reactivity to glycolytic enzymes has been reported [13].

However, these three targets were not all specific to CC sera and have been reported (sometimes at a high frequency) in liver diseases, including hepatocellular carcinoma [14, 15], as well as glyceraldehyde3-phosphate dehydrogenase in melanoma [16], enolase in some autoimmune diseases and infections [17‐19] and fructose biphosphate aldolase in a case of drug hepatotoxicity [20].

Concerning other immunoreactive targets identified by fewer CC sera and having catalytic activity, i.e., liver arginase and ATP synthase sub-unit β, they have been reported in a variety of settings including autoimmune diseases [21], Alzheimer’s disease [22], coeliac disease [23], idiopathic nephrotic syndrome [24] and also in 11 % of hepatocellular carcinoma patients [25]. However, to our knowledge, ATP synthase sub-unit β has never previously been reported as an auto-antigen. Interestingly, ATP synthase is also located at the cell surface and may contribute to the development of an acidic micro-environment in tumour tissues [26]. Its surface location allows it to gain access to the immune system, and it has been reported as the target of a subset of T-gamma-delta lymphocytes [27].

In terms of antigenic targets with structural activity, they were essentially found when CC cell lines or tumour liver specimens were used as antigens. Three of them were identified by more than 50 % of CC sera on the most appropriate substrate, i.e., vimentin, prelamine A/C and actin. Both actin and members of the intermediate filament family have been demonstrated to be strongly implicated in tumorogenesis [28‐30]. But AAbs to these proteins lack specificity as potential biomarkers, because they may be detected in various autoimmune disorders [31], in liver diseases [32], and in hepatocellular or digestive cancer [32, 33]. However, to our knowledge, AAbs to actin have never been reported in the context of liver carcinoma. The presence of AAbs to vimentin and actin was confirmed using a fluorescence technique that enabled access to a conformational epitope, whereas serological proteome analysis generates denatured antigens. Interestingly, actin has been reported to be found in the secretome of HuCCA CC cell lines [6], and its presence in a bodily fluid may cause a loss of immune tolerance. In this study of Srisomsap [6], annexin was also identified in the secretome. We also found various isoforms of annexin, A1, A2, A4 and A5 as the principal antigenic targets for CC sera if they were tested on a tumour substrate (Table 1). Annexin is involved in many cellular processes [34], and is implicated in the genesis of numerous diseases. Annexin A1 was recently reported to be highly expressed in CC, but not in hepatocellular carcinoma [35]. The over-expression or post-translational modification of annexin A2 has been reported in various cancers, such as colorectal, oral and lung cancers [36‐40]. However, AAbs to annexin A2 have also been reported in the context of anti-phospholipid syndrome, sometimes in association with cancer [41, 42].

One final and interesting observation was the presence of proteins categorised as transfer/carrier proteins and accounting for 20 % of autoantigenic targets in the tumour tissues, compared to 6.3 % in the non-tumour tissues adjacent to the CC (Additional file 3: Table S3). These proteins included serotranferrin, which was recognised by 100 % of CC sera. Serotansferrin carries iron from its absorption sites and delivers the metal to cells [43], and may also contribute to stimulating cell proliferation [44]. It was very exciting to note that once again, serotransferrin was identified in the secretome of CC cell lines [6]. Until now, anti-serotransferrin auto-antibodies had been found in 30 % of sera from patients with hepatocellular carcinoma and at a lower rate of 5 % in the context of liver cirrhosis and chronic hepatitis [15].

In order to be used as biomarkers, AAbs need to be both highly sensitive and highly specific. However, most of the AAbs detected during the present study had previously been reported not only in cancers, but also in the context of autoimmune disorders. We therefore cannot conclude that AAb alone could be considered as a biomarker, in agreement with a previous report [10]. Nevertheless, a combination of several AAbs tested on a panel of a significant number of patients, and using the most appropriate substrate defined during this study, might be able to identify the best biomarkers for CC.