Serum autoantibody profiling using a natural glycoprotein microarray for the prognosis of early melanoma

Abstract

The poor prognosis of melanoma and the high cost of lymph node biopsy for melanoma patients have led to an urgent need for the discovery of convenient and accurate prognostic indicators. Here, we have developed a natural glycoprotein microarray to discover serum autoantibodies to distinguish between patients with node negative melanoma and node positive melanoma. Dual-lectin affinity chromatography was used to extract glycoproteins from a melanoma cell line. Liquid-based reverse phase separation and microarray platforms were then applied to separate and spot these natural proteins on nitrocellulose slides. The serum autoantibodies were investigated by exposing these proteins to sera from 43 patients that have already been diagnosed to have different stages of early melanoma. The combination of 9 fractions provides a 55% sensitivity with 100% specificity for the detection of node positive against node negative and a 62% sensitivity with 100% specificity for the detection of node negative against node positive. Recombinant proteins were used to confirm the results using a sample set with 79 patients with diagnosed melanoma. The response of sera against recombinant 94 kD glucose-regulated protein (GRP94), acid ceramidase (ASAH1), cathepsin D (CTSD), and lactate dehydrogenase B (LDHB) shared a similar pattern to the fractions where they were identified. The glycoarray platform provides a convenient and highly reproducible method to profile autoantibodies that could be used as serum biomarkers for prognosis of melanoma.

Figure 1. Humoral response experimental overview

Proteins were first extracted from melanoma cell lines then dual-lectin (ConA&WGA) affinity chromatography was applied to enrich the N-linked glycoproteins and proteins interacting with these glycoproteins. These proteins were separated by a reverse phase HPLC then the separated fractions were spotted on nitrocellulose slides. Antibody-antigen response of each spot was detected using anti-human IgG conjugated to a fluorophore after probing with sera from patients with early node negative melanoma (node negative) and patients with early node positive melanoma (node positive). Following statistical analysis, fractions which exhibited a different humoral response between the node negative and node positive were selected, and then proteins from these fractions were identified by ESI-MS/MS and MALDI-QIT. We further investigated the performance of these proteins on the prognosis of different stages of melanoma, using recombinant proteins and a confirmation set of melanoma patients.

Figure 2. Heat map with dendrograms of hierarchical clustering for the natural glycoprotein microarray

The heat map was based on the intensities of response from 9 fractions against sera from node negative and node positive patients. Each column indicates a fraction which showed significantly different response between node negative and node positive (p<0.05), and each row indicates a serum from patient with node negative or node positive. The intensity of response increases from red color to blue color. The 9 fractions allow us to distinguish between most of the patients with ENN (red rectangle) and the patients with ENP(blue rectangle). The heat map was generated with R language.

Figure 3. Performance of Selected fractions and protein identification

Forty-seven fractions collected from RP-HPLC fractionation were used as bait to explore the different levels of serum autoantibodies between patients with early node negative melanoma (node negative) and early node positive melanoma (node positive). Serum from patients with node negative showed higher humoral response to 3 fractions while serum from patients with node positive showed higher humoral response to the other 6 fractions. (a) The combined Received Operating Characteristic (ROC) Curve for the 3 fractions which showed higher reactivity to serum of node positive. (b) The combined Receiver Operating Characteristic (ROC) Curve for the 6 fractions which showed higher reactivity to serum of node negative.

Figure 4. The reactivity of selected fractions against serum and protein identification

(a) The scatter plot of the reactivity of fraction 43, 44 and 45 to the serum from different melanoma patients. Each spot represents the normalized intensity of serum autoantibodies from one patient against one of the selected fractions. Patients with node negative showed significantly higher response to fraction 43, 44 and 45, compared to the node positive group (*p<0.05; ** p<0.01). (b) MS/MS sequencing data of 4 peptides from cathepsin D (CTSD) identified from the fractions 43, 44 and 45.

Figure 5. The scatter plot of the reactivity of recombinant GRP94, ASAH1, CTSD and LDHB to serum from patients with node negative and patients with node positive

Each spot represents the normalized intensity of serum autoantibodies from one patient against one of the recombinant proteins. These recombinant proteins showed the similar pattern of humoral response to the fractions from which they were identified.

Figure 6. Performance of recombinant GRP94, ASAH1, CTSD and LDHB

(a) Performance of the 4 recombinant proteins on distinguishing between patients with node negative and patients with node positive.(b) The combined ROC Curve for the 4 recombinant proteins.