Introduction

Volumetric absorptive microsampling devices (VAMS) allow patients to collect blood via fingerprick sampling at home. This approach enables collection of frequent longitudinal samples that can be used for health surveillance of disease, including cancer, for early diagnostics, ongoing monitoring, and recurrence. We have demonstrated that VAMS devices aid sample preparation for proteomics, facilitating the depletion of high abundance proteins via washing, prior to digestion. Reducing the dynamic range allows detection of more than 3000 proteins using a mid-throughput mass spectrometry (MS) method running 18 samples per day (SPD). Using a clinical cohort of non-small cell lung cancer (NSCLC), we compared the use of both mid- and high-throughput methods to identify relevant biomarkers and differentiate healthy from disease patients.

Methods

Blood from 16 patients with NSCLC and 18 age and sex matched controls were loaded onto VAMS devices and dried at room temperature for 24 hours. VAMS devices were washed overnight in LiCl wash buffer, followed by two subsequent shorter washes. The proteins remaining in the tip were extracted using lysis buffer, reduced, alkylated and digested with 1 µg overnight at 37 °C. Samples were desalted using solid phase extraction, quantified and 500ng used for MS injection. Two MS methods were employed, first a mid-throughput method using 18 SPD method on a QEHFX (Thermo) and the second a 60 SPD method on a 7600 Zeno-TOF (SCIEX) equipped with an Evosep LC.

Preliminary data

Due to the use of a longer gradient and LC column for the mid-throughput method, significantly more proteins were identified in comparison to the high-throughput method (mean and SD of 3370 ± 117 and 1483 ± 137, respectively). Per minute of run time however, the high-throughput produced 1.5-fold more IDs. The high-throughput method had greater sparsity with less than 50 % of the data quantitated in all samples compared to more than two thirds for the mid-throughput method. Overall, 36 (high-throughput) and 455 (mid-throughput) differentially expressed proteins were discovered for each method. There were 31/36 of these identified proteins thar overlapped with the mid-throughput results, with several of the overlapping proteins previously identified as biomarkers of NSCLC. Using a machine-learning package, ProMor, a fourteen-marker prediction model was identified for the high-throughput method which gave an area under the ROC Curve (AUC) of 72.2% using k-nearest neighbours (knn) compared with 94.4% using a nine-maker panel for mid-throughput. There was only one protein common between the mid- and high-throughput models, a serine protease, Myeloblastin (PRTN3), which has been linked to KRAS mutations in lung cancer patients. There were 5 pathways identified that were statistically significant and were common to each method. All these pathways are closely related, with neutrophil degranulation, antimicrobial peptides, and ROS and RNS production in phagocytes being pathways under the innate immune system, indicating a strong immune system response in the disease cohort that is detected by both methods.

Novel aspect

Combination of dried blood spots and high-throughput mass spectrometry to identify cancer biomarkers.