Published in: Journal of Clinical Oncology, Volume 37, no. 15_suppl., 26 May 2019

Authors

Vincenzo Lagani, Ioannis Tsamardinos

Olav Toai Duc Nguyen, Maria Markaki, Christina Chatzipantsiou, Animesh Sharma, Oluf D. Røe

Abstract

Background:
The high incidence and high mortality rate of small-cell lung cancer (SCLC) calls for identification of methods for early diagnosis. Searching of cancer-related proteins and proteins signature in biofluids is an emerging approach in early diagnostic of malignancies. In the present study we have used proteomics-based profiling of serum collected 2 months to 5 years before SCLC diagnosis to search for early diagnostic biomarkers.

Methods:
All serum samples in this study were obtained from the Nord-Trøndelag Health Study (HUNT) Research Centre’s Biobank. Discovery sample set (Cohort I): Serum samples from 12 individuals that subsequently developed SCLC and 12 matched controls were obtained. Validation sample set (Cohort II): Serum samples from 5 future SCLC patients and 5 matched controls were obtained. The serum samples in both cohorts were collected in a time frame of 2 months to 5 years before diagnosis. All controls included in the study were matched to the cases for smoking status (pack years and quit time), gender and age, and were cancer-free at least 5 years before blood sampling. All subjects were smokers or ex-smokers. Twenty (20) µl of serum was depleted of its high-abundant proteins and the cleaned-up peptides were analysed by LC- MS with an Orbitrap Elite mass spectrometer. Data were processed/analysed using MaxQuant software. Using Cohort I as training set, proteins most related to diagnosis were identified at first with limma univariate analysis. Reference and equivalent signatures were identified using JADbio with SES as feature selection algorithm. The JADbio matched pipeline with 50 repetitions for performance assessment (AUC with 95% CI) was used.

Results:
In each serum sample the analysis detected 435 proteins in Cohort I and 609 proteins in Cohort II. Reference signature for 12 SCLC future cases vs 12 matched controls had a conservative AUC of 0.686 (95% CI: 0.557, 0.765) in the discovery cohort. Three proteins in the reference and equivalent signatures were also found to be differentially expressed in validation Cohort II; all three proteins discriminated small-cell lung cancer patients from their respective controls (AUC 0.757-0.826 in the discovery Cohort I and AUC 0.68-0.84 in the validation Cohort II).

Conclusions:
The results indicate that differential levels of a few proteins in serum may help detecting SCLC 2 months to 5 years prior to clinical diagnosis. This is one of the first large-scale proteomics screening studies of pre-diagnostic serum of future SCLC patients. Further validation studies are in progress.