We used 32 human proteomic microarrays to profile serum samples from 16 healthy persons (HC) and 16 renal cancer patients (RC), and anti-human antibodies with different fluorescence are used to mark the IgG or IgM which can bind to the recombinant proteins of the microarray, where goat anti-human IgG (labeled by Cy3) antibody in green, the donkey anti-human IgM (labeled by Cy5) antibody is in red, and the workflow is shown in Fig. 1a. Comparing RC and HC, for IgG or IgM, there are 3 possible results on the microarray, 1) Unchange, autoantibodies of RC and HC have the same response to the majority of target proteins, 2) Up, autoantibodies from RC are more responsive than HC to some target proteins, 3) Down, autoantibodies from RC serum are less responsive than HC to the remaining proteins of the microarray. According to these three situations, we found the corresponding target proteins in the experimental results (Fig. 1b).
We performed a statistical analysis of the signal of each protein on the microarray and set criteria for screening autoantibody and their target proteins in the "Methods" section. The volcano plot shows all Up, Down and Unchange autoantibodies of IgG and IgM in RC (Fig. 2a, 2b). Among them, 420 autoantibodies have changed significantly, and the positive rate, FC and p-value of all target proteins are shown in Table S1. Compare RC to HC, Up autoantibodies included 139 types of IgG and 43 types of IgM and Down autoantibodies included 20 types of IgG and 218 types of IgM. Among IgG autoantibodies, Up autoantibodies are significantly more than Down autoantibodies, while in IgM autoantibodies, the statistical results are just the opposite. Some up autoantibodies can be applied as candidate biomarkers, for example, anti-BCAS4 IgG and anti-RCN1 IgM have high positive rates, 43.75% (7/16) and 31.25% (5/16), suggesting that they may be RC-related autoantibody target proteins (Fig. 2d, 2f). Besides, down autoantibodies are hardly be used as candidate biomarkers, but their targeted proteins are closely related to renal cancer. For example, anti-GAP43-IgG and anti-CCT8-IgM in RC are significantly lower than HC (Fig. 2c, 2e). Among the 420 differential target proteins that were screened out, 10 were present in both IgG and IgM. After removing duplicate counts, the remaining 410 differential target proteins were analyzed by bioinformatics.
To understand the biological relevance of the target proteins, we applied the online protein classification tool, PANTHER, to the 410 screened target proteins to identify enrichment for specific biological process, cellular component and molecular function. We found that the target proteins could be classified into 12 biological process groups (Fig. 3a), the largest of which was the cellular processes group (GO: 0009987), followed by metabolic processes (GO: 0008152). In terms of cellular component, the candidates could be classified into 11 groups. Finally, the candidates could be classified into 8 protein class groups, the top 3 of which were binding (GO: 0005488), catalytic activity (GO: 0003824), and molecular function regulator (GO: 0098772).
To gain insight into possible functional roles of the identified targeted proteins, the enrichment of ontology terms and components of molecular pathways of the candidates were analyzed using DAVID in comparison with their occurrence. The candidate proteins were examined for enrichment, the GO terms of P < 0.05 were shown in Fig. 3b. There is a significant number of the candidates localizing in Mitogen-activated protein kinase (P = 0.001180). Besides zona pellucida receptor complex (P = 0.013122), other significantly enriched terms were IGF-1 Signaling Pathway (P = 0.003730) and DNA damage response (P = 0.035909).
To create significance out of the targeted proteins, biological protein-protein interaction networks of these proteins were constructed. The candidate proteins were imported into STRING to build the network (Fig. 4a). Three pathways of ErbB, Antigen processing: ubiquitination & proteasome degradation and Cargo recognition for clathrin-mediated endocytosis were also identified (Fig. 4b-d). 4 proteins were overlapped with the MAPK pathway of papillary renal-cell carcinoma (Fig. 4e)