Introduction
Colorectal cancer (CRC) accounts for 608,700 deaths per year worldwide [
1] which makes it one of the most common causes of cancer related deaths. Randomized clinical trials have demonstrated the value of population-based screening to reduce CRC-related mortality. In part, this can be ascribed to the detection of early-stage CRC with provision of timely treatment [
2‐
4]. Therefore, there is a strong interest in the identification and clinical validation of new CRC biomarkers to be used for early detection of this disease.
Modalities as the Faecal Occult Blood Test (FOBT) and stool DNA (sDNA) tests are presently the only approved non-invasive screening tests available for detection of CRC in asymptomatic individuals. The performances of these tests have varied [
4‐
6] and there is an immense problem with compliance [
7,
8]. In the results of a Danish study it was demonstrated that ≥60% compliance is a prerequisite in order to obtain successful effect of the screening [
9]. A similar compliance problem in symptomatic individuals is not observed when using serological tests where the compliance is over 90% [
10]. One potential serological CRC screening test is the Septin 9 (SEPT9) methylated DNA test which has demonstrated good test performance in a prospective screening study including nearly 8,000 asymptomatic individuals. However, the SEPT9 test still leaves about 33% of cancer patients undetected while the false-positive rate is 12% [
11]. Furthermore, the technical platform could be improved as the SEPT9 test requires a large volume of plasma per test.
We addressed the clinical needs for a blood-based test by initiating a protein biomarker study evaluating 74 different protein biomarkers in plasma samples from case–control patient material consisting of symptomatic individuals represented by CRC patients, adenoma patients, patients with non-neoplastic large bowel diseases and healthy individuals. Presently, patients both with adenoma and CRC need further examination using endoscopy in order to evaluate the pathology of the neoplasm. Hence, it is not as important to discriminate between these two groups if the aim is to develop a pre-colonoscopy screening test. However, it is a technical challenge to measure proteins in plasma due to their biological complexity and a wide range in protein concentrations [
12,
13], a problem that is reflected in the general absence of serological CRC screening protein biomarkers being implemented in clinical settings [
14,
15]. Furthermore, in order to increase the success rate of a potential blood-based test, we found inspiration in a number of studies which demonstrated that higher discrimination power could be obtained by combining biomarkers [
16‐
18]. However, aiming for a panel of biomarkers for a final test requires an assay with multiplexing capability, but without loss of technical sensitivity and specificity [
19]. Moreover, low sample consumption and good assay performance in general is needed in order to facilitate high quality biomarker studies. We addressed these technical issues by applying the novel proximity extension assay (PEA) which is an improved version of a biomarker discovery tool with assay performance superior to the related Proximity Ligation Assay (PLA) in plasma samples [
19].
To investigate the reason for the additive effect of our biomarker panel we evaluated the independence among the potential biomarkers as well as exploring their biological associations and interactions by performing network and pathway analysis.
In this study, we applied the novel PEA technique to identify five serological proteins that could discriminate between patients with colorectal neoplasias and control groups of healthy individuals and patients with other diseases. Furthermore, by focusing on the early stages of CRC in the statistical evaluation, we identified three proteins that are potential candidate biomarkers for early detection of CRC.
Discussion
The present study demonstrates that the novel multiplex PEA assay is suitable to identify potential plasma biomarkers for detection of CRC. Forming a biomarker panel consisting of plasma CEA, TFRC, MIF, OPN/SPP1 and CA242 could represent a biomarker test for detection of CRC in symptomatic individuals. Of particular interest is the observation that the plasma levels of CEA, TFRC and CA242 were identifiers of early-stage CRC, suggesting that the panel holds potential as an early detection method of CRC. However, future studies including asymptomatic individuals as well as patients with other diseases, including other cancer types, are needed to test this hypothesis. We also investigated plasma obtained from adenoma patients in order to search for a precancerous biomarker, but unfortunately no adenoma-specific biomarkers with sufficient statistical significance could be identified.
Among symptomatic individuals the five CRC biomarkers identified were shown to detect CRC patients with a sensitivity of 56% at a specificity of 90% (Figure
1). Furthermore, functional in silico analyses showed that the four biomarker proteins are involved in different pathways, which supports their lack of association and their individual contribution to the additive effect of the panel. These four biomarkers did not have any direct interactions, yet all had interactions with TGFβ1, indicating some common functional character. The most significant associated diseases and disorders associated with the four biomarkers were inflammatory response (CEA, MIF, OPN/SPP1), cancer (all four), gastrointestinal disease (all four), cardiovascular disease (MIF, OPN/SPP1, TFRC) and infectious disease (MIF, OPN/SPP1, TFRC), supporting that these biomarkers have a biological role in CRC. The three early-stage CRC biomarkers were shown to identify stage I and II CRC patients with a sensitivity of 53% at 90% specificity (Figure
2). In asymptomatic individuals, the sensitivities of the FOBT range from 40-90% at specificities ranging from 85-90% [
8], but the sensitivity is highly affected by the low compliance for faeces tests [
7,
30]. Nevertheless, it has been shown that CRC screening by the FOBT reduces mortality from CRC by 15-21% [
6,
31,
32] and that CRC screening in general increases the number of patients detected with early-stage disease [
33]. Measurements of SEPT9 methylated DNA in serum of asymptomatic individuals have recently been shown to identify CRC patients with a sensitivity of 67% at 89% specificity [
11]. The performance of our biomarker panel is thus within the ranges of the FOBT and almost within the ranges of the SEPT9 test, although a direct comparison between asymptomatic and symptomatic individuals may be inaccurate. Furthermore, like the SEPT9 test, our biomarker panel has the advantage of being a simple blood test and thus the compliance will most probably be high. It should be noted, however, that our results are obtained from analysis of symptomatic patients prior to endoscopy. Initiating a validation study in an asymptomatic population-based cohort might result in altered sensitivity and specificity performance as well as an altered compliance.
To obtain further information about the value of our biomarker panel as identifier of early-stage CRC we compared the analysis of CRC TNM stage I-II with their matching control groups (healthy individuals, other diseases and adenoma patients). We found three positive identifiers; TFRC, CEA and CA242 with known biological roles and previous associations to cancer. TFRC has an important role in the inflammation process and it has also been described as being specific for tumor cell proliferation as it provides high iron uptake required for cell division [
34,
35]. CEA is a glycoprotein and a well-known CRC monitoring marker [
19,
36] suggested to mediate cell-cell adhesion, maintain the bacterial environment of the intestine and protect the colon from infectious microorganisms [
36]. Lastly, CA242 is a blood-group antigen defined by the monoclonal antibody C242 and it has also previously been described as a potential CRC biomarker [
37]. The adenoma patients were included in the study as it would be an advantage to include adenoma biomarkers in a potential screening panel. However, only NSE was a significant identifier of adenoma, but it did not demonstrate any impact in the ROC analysis. The low success rate of adenoma biomarkers could be a consequence of our initial literature search from which the 74 proteins were initially chosen as the search was focused on CRC biomarkers.
A challenge of both adenoma and early-cancer serological markers is the concentration of each analyte which needs to be of a sufficient level in the blood before it is theoretically possible to measure it [
38]. From the literature we know that CEA is increased in tumor tissue [
39], that the
MIF (Hs.407995) is up-regulated in colorectal carcinomas [
40], that OPN is significantly higher in CRC tumor tissue compared to normal tissue [
41] and that TFRC is higher in tumor tissue compared to normal mucosa [
42]. The tumor tissue level of CA242 is not well established. Searching for detection markers in the early CRC stages when the tumor is minor or in the adenomateous lesions it has been argued that the production of protein markers is not sufficient to make an imprint in the systemic circulation [
42]. This could be a potential pitfall for discovery of biomarkers, especially for the early stage CRC and the adenomateous lesions, as one would expect less activity from these. However, to take one example, the TIMP-1 plasma level has been described by several authors to be elevated in CRC patients. In most of these studies it was found that stage I, II and III CRC patients had the same degree of TIMP-1 elevation as compared to non-neoplastic individuals and thus the plasma level was not related to tumor burden [
43]. In the present study we were able to statistically point out discriminators of both CRC stage I-II and all stage CRC in plasma suggesting that it is possible to develop serological screening markers for CRC. The origin of these markers is unknown and could be derived from the tumor microenvironment rather than the tumor itself. This question needs yet to be answered and it would be interesting to compare the tumor tissue level with the serological level of each of the identified CRC biomarkers.
Conclusions
Survival after CRC is currently being improved by screening, but improvement of detection methods for early stage CRC is needed. Such methods should be easy to perform, have high sensitivity, high specificity and high compliance and be inexpensive. Since CRC is a heterogeneous disease it could be an option, as indicated in the present study, to combine different screening tests in order to obtain a high test performance. Consequently, implementation of a sensitive and specific serological screening test could be of great importance as an add-on to the FOBT or SEPT9. In that regard it could be an advantage that the test platforms are run on the same specimen in order to simplify sample collection and maintain high compliance. Interestingly, proximity assays may be applied for analysis of stool samples [
44] and saliva (unpublished observations) which could add further to the usability of this assay in CRC biomarker research.
In summary, five plasma protein biomarkers were identified by the novel multiplex PEA assay as potential CRC discriminators and three of these were additionally found to be discriminators of early-stage CRC. The performance of our biomarker panel in symptomatic individuals was within the range of sensitivity and specificity seen for asymptomatic individuals applying the FOBT and the performance of the SEPT9 test. PEA assays of the five identified protein biomarkers have been thoroughly validated. However, analyses in a new independent and larger sample cohort of asymptomatic individuals are needed in order to further validate the performances of our CRC biomarker panel.
Acknowledgements
This work was funded by the EU FP7-Health-2007-B under the PROACTIVE project, VINNOVA within the Forska&Väx programme, The Kathrine and Vigo Skovgaard Foundation, Danish Cancer Society, The Kornerup Foundation, The Aage and Johanne Louis-Hansen Foundation, The Aase and Ejnar Danielsen Foundation, The Henrik Henriksen Foundation, The Walter and Kristiane Christensen Foundation, The Oda and Hans Svenningsen Foundation, The “Midtjyske Bladfond”, The Beckett-Foundation, The Einar Willumsen Foundation, The Hede-Nielsen Family Foundation, The Bechgaard Foundation, The Grønbech-Olsen Foundation, The Sophus and Astrid Jacobsen Foundation, The KID Foundation, The Obel Family Foundation, The Sven and Ina Hansen Foundation, The Torben and Astrid Frimodt Foundation, and The Willy and Ingeborg Reinhard Foundation. Furthermore, The Danish Endoscopy Study Group is acknowledged for providing the plasma sample material.
Competing interests
M.L., E.A., A.V. and S.F. are employees at Olink Bioscience AB owner of intellectual property on the proximity ligation assay.
Authors’ contributions
EA, JS, SF, AV, HJN and NB conceived the study and coordinated activities. HJN collected all samples and information on the patients. ML, SBT, SLCT, BSN, SF and EA established, optimized, validated and performed the proximity assays. MK and NG optimized the conjugation step in the probe preparation process. IJC performed all statistical analyses. SBT, NB, and JS drafted the manuscript. All authors read and approved the final manuscript.