The diagnosis of sarcoidosis is still a significant challenge in China because of the need to exclude lung diseases that are especially common in China (ie. TB, COPD, lung cancer of which in patients with similar clinical and/or radiographic findings). This is the first study demonstrating potential effectiveness of protein profiling in serum with ClinProt as a diagnostic tool for sarcoidosis in ethnic Han Chinese. We confirm that the SAA serum level and tissue expression of sarcoidosis are higher than in other lung diseases.
Over the last decade, accumulated evidences in the research of protein biomarkers of sarcoidosis have shown that it is possible to find sarcoidosis-specific or associated proteins through analysis of body fluids. Researchers used to compare protein profiles from sarcoidosis patients with healthy controls in specimens of bronchoalveolar lavage fluid (BALF)
[21–26]. However, no further analysis for the positive identification of the sarcoidosis-associated peak was performed. Collection of blood is much less invasive than collection of BALF, and not subject to variabilities of BALF collection such as dilution and contamination by oropharyngeal secretions. Several studies have reported novel serum biomarkers in sarcoidosis. Bons JA and his colleagues from The Netherlands first applied surface-enhanced laser desorption ioni-zation time-of-flight mass spectrometry (SELDI-TOF-MS) to compare protein profiles from 35 sarcoidosis patients to 35 healthy controls in 2007
. The differential protein peak M/Z 17,377 was identified as the alpha-2 chain of haptoglobin, although the total haptoglobin level in the serum detected by ELISA was not statistically significant between the disease group and the controls. In the same study, there were two other unidentified serum proteins that were also up-regulated in sarcoidosis patients. Recently, Bargagli et al, using proteomic tools found that serum concentrations of SAA were significantly higher in sarcoidosis patients than in healthy controls. They thought that SAA is likely one of the unidentified biomarkers in Bons’s proteomic study
This study is the first effort using the ClinProt system of serum protein profiling in Chinese subjects to discover new diagnostic biomarkers in sarcoidosis. The ClinProt system is considered as more advanced technologies than SELDI-TOF-MS
. We are charged with the need to differentiate sarcoidosis from other similar diseases with enrolling 37 patients with sarcoidosis, 20 healthy volunteers and otherwise non-sardoidosis diseases in lung including tuberculosis, COPD, interstitial lung disease and lung cancer as control. In our pilot study, among the multiple differential protein peaks detected by ClinProt-MALDI-TOF, the expression levels of M/Z 3,210 in the sarcoidosis group was significantly elevated in peak height than those in non-sarcoidosis groups (p < 0.05. Thus M/Z 3,210 was the potential candidate biomarker, which was further analyzed by LC-MS/MS. The investigations confirmed a 29 amino acid peptide sequence of “SLADQAANEWGRSGKDPNHFRPAGLPEKY”, which corresponds to the N-terminal fragment of SAA.
SAA is an acute phase protein, which belongs to a group of heterogeneous proteins of the apolipoprotein family. It is possible that SAA could activate the NF-kB signaling pathway through interactions with Toll-like receptor 2 (TLR2), leading to the development of chronic granuloma as has been recently reported by Chen et al
. Although the underlying mechanism of the increase in the secretion of SAA or the chemotatic tissue deposition due to changes in the internal environment remain unclear. Nonetheless these findings suggest that SAA could be a potential biomarker for the diseases.
A literature survey indicates that besides sarcoidosis, SAA is also over-expressed in several diseases such as COPD
, lung cancer
, and other interstitial lung diseases
. Our study showed that expression of SAA was significantly higher in sarcoidosis than in the other non-sarcoidosis disease groups (Figure
4). Furthermore, we found that sarcoidosis patients also had increased tissue levels of SAA, and it was deposited in the granulomas of patients (Figure
The optimal cutoff of the serum SAA concentration obtained by the ROC analysis was 101.98 ng/ml yielding a sensitivity of 96.3% and specificity of 52.5% for diagnosis, respectively. By including not only healthy controls but by expanding the comparator groups to include patients with other inflammatory lung disease, our results demonstrated that SAA might be an important player in the development of sarcoidosis. However, although SAA is sensitive, it was less specific for the diseases tested in the study. This means that it is not for confirmative or for ruling- in a diagnosis of sarcoidosis.
A follow up question is whether the SAA is a marker of disease activity or it was a sequel of the treatments? As early as 1989, Rubinstein et al had shown that SAA concentration in 25 sarcoidosis patients (13 cases with “active sarcoidosis” and 12 cases with “inactive sarcoidosis”) and 94 healthy volunteers. He showed that SAA concentrations in both types of sarcoidosis were significantly higher than that of healthy controls. However, SAA was of limited usefulness as a marker for disease activity
. A report from Italy demonstrated that concentration of SAA in serum is significantly higher in patients with sub-acute onset requiring prolonged and multiple steroid treatments than in patients with sub acute onset but not requiring therapy (p < 0.001). In this study, we found that SAA is similarly increased in both “treated” and “untreated” sarcoidosis, which implies that SAA is not affected by the corticosteroid treatment.
Since the usefulness of multiple markers for diagnosis is now widely recognized
. ClinProt software could be used in combination with three different algorithms (Supervised Neural Network (SNN); Genetic Algorithm (GA) and QuickClassifier (QC) to create testable diagnostic models for classifying potential biomarkers with respect to whether they are associated with the sarcoidosis. In this study, the SNN method has achieved the optimal classification of sarcoidosis and non-sarcoidosis with efficiency of identification of 100%, and predictive capability of 78.62%; 93.15%, and 69.24% of efficiency of identification and predictive capability, respectively, between the sarcoidosis and the healthy controls; and 100% and 81.62% of efficiency of identification and predictive capability, respectively, between the sarcoidosis and the diseased comparator groups. The machine learning algorithms greatly improved the efficiency in identification of potential serum biomarkers for the diagnosis of sarcoidosis when compared to the previous study. GA mimics evolution in nature, together with the SNN and the QC are univariate sorting algorithms, which use the p-values at certain peak positions for classification. Therefore, these ClinProt system-based serological classification models can be applied to the clinical diagnosis of diseases, providing additional validation methods. However, the current classification models based on the ClinProt platform would be expensive and complicated to operate in routine clinical settings. Thus they are limited in wider clinical applications.