Overview of proteomic landscape of trace FFPE samples in duodenum cancer and gastric cancer
We performed proteomic profiling of trace 40 samples collected from 4 duodenum and 4 gastric cancer patients who had not experienced prior chemotherapy or radiotherapy. The amounts of peptides were diluted from each patient samples as 2.0 µg (D1/G1), 1.0 µg (D2/G2), 0.5 µg (D3/G3), 0.25 µg (D4/G4), and 0.125 µg (D5/G5), respectively (Fig. 1 and Supplementary Fig. S1a). With the dilution of the AOPs, the number of identified proteins slightly decreased from ~ 4,000 in the D1/G1 to ~ 2,000 in the D5/G5 (Fig. 2a).
Proteomic analysis was performed using a label-free quantification strategy [7, 11]. Protein abundance of all samples was firstly calculated by intensity-based absolute quantification (iBAQ) [12, 13] and then normalized as a fraction of the total (FOT). The nearly same number of identified proteins in samples with the same AOPs indicated the stability and the highly quality control of MS platform. As well as in gastric cancer, we found the top proteins extensively expressed in all samples in duodenum cancer, including HBA2, HBB, AGR2, HIST2H4B, HIST2H2BF, etc. (Fig. 2a, Supplementary Table S1a). However, the lower proteins detected in the samples with 5 different AOPs were distinctive. For example, the low abundance proteins of NIPBL, GTF3C1, HEATR5A, MDN1, and CLASP2 were observed in the D1, and of SAMD9, NBAS, C6, ABCC1, and PRRC2A were disclosed in D2 (Fig. 2a). Likewise, the top proteins identified all gradients samples in gastric cancer, such as ALB, HBB, KRT8, HIS2H4B, HIS2H2BF, etc. (Supplementary Table S1b). The differential low expressed proteins were identified in 5 AOPs samples groups. For example, ITPR2, CAPN15, and ABCB1 were detected in G1, and UBR2, VAV2, and WDR7 were observed in the G2 (Supplementary Fig. S1b). Taken together, we built a proteomic landscape of trace FFPE sample in duodenum cancer and gastric cancer for the first time, and disclosed the different features of high/low proteins in duodenum and gastric cancers with different AOPs.
Proteomic characterization of trace FFPE samples with 5 different AOPs
To explore the coverage in different gradients, we applied Venn diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/) to the D1 to the D5, in which the D1 was regarded as the basal. As well as in the gastric cancer samples, we found the overlapped proteins were decreased with the dilution of peptides in duodenum cancer samples ranging from 3,228 in overlap 1 (D-3,228, D1 & D2) to 1,638 in overlap 4 (D-1,638, D1&D2&D3&D4&D5) (Fig. 2b and Supplementary Fig. S1c). In addition, when the input of AOPs of the trace samples was no less than 0.5 µg, the coverage of identified proteins was over 60% in duodenum cancer and 70% in gastric cancer (Fig. 2b and Supplementary Fig. S1c).
To characterize the proteomic profiles of the trace samples with different AOPs, we performed principal component analysis (PCA) and found the clearly discrimination among trace samples when the input of AOPs was no less than 0.5 µg in duodenum cancer and gastric cancer (Fig. 3a and Supplementary Fig. S2a). Consensus clustering identified 4 clusters in duodenum cancer and 3 clusters in gastric cancers, which further revealed the gathering of trace samples with the same input of AOPs (Fig. 3b and Supplementary Fig. S2b). Especially, the cluster 1 in gastric cancer contained the G1, G2, and G3, indicating the similarity of the trace samples (Supplementary Fig. S2b). As well as in gastric cancer, Spearman’s correlation analysis illustrated that the higher correlation coefficients were detected in the trace samples with the same input of AOPs in the duodenum cancer (Fig. 3c and Supplementary Fig. S2c). These findings implied the key role of the same input in cancer cohort researches, and revealed the minimum standard of the input of AOPs (0.5 µg) in trace samples for proteomic profiles, providing a new insight for early-stage cancers.
Biological pathways in 5 AOPs trace samples
Previously, we showed that Wnt signaling and extracellular matrix organization were the driver pathways in gastric cancer . Indeed, we also detected Wnt signaling (p < 0.05, FDR < 0.1) and extracellular matrix organization (p < 0.05, FDR < 0.1) in all samples of gastric cancer (Fig. 4a). In addition, visualization of the Gene Ontology (GO) disclosed that the overlapped proteins participated in the metabolic machinery (e.g., tricarboxylic acid cycle and glycolysis) (p < 0.05, FDR < 0.1), immune response (e.g., antigen processing and presentation, NF-kappaB signaling) (p < 0.05, FDR < 0.1), and canonical cancer-related pathways (e.g., cell cycle and Mapk signaling) (p < 0.05, FDR < 0.1), suggesting the potential carcinogenesis in gastric cancer (Fig. 4a, Supplementary Table S1c). These biological pathways (p < 0.05, FDR < 0.1) were also overrepresented in duodenum cancer, indicating their potential cancer driven effects in duodenum cancer.
To investigate the loss functions with the dilutions of AOPs, we integrated the differential expressed proteins (DFP, fold change ≤ 0.5), in which the D1 was the basal. We then performed GO analysis of the DEPs and found antigen processing and presentation was decreased when the input of AOPs were no more than 0.5 µg in duodenum cancer, which was also observed in gastric cancer when the input of AOPs of trace samples was no more than 0.25 µg. The loss functions of cell cycle, Wnt signaling, and NF-kappaB signaling were detected when the input of AOPs were no more than 0.25 µg in duodenum cancer (Fig. 4b, Supplementary Table S1d). Anti-EGFR strategy is prevalent for gastrointestinal cancer, for example panitumumab to anti-EGFR in colorectal cancer [14, 15], and was employed in treating duodenal adenomatous [16, 17]. However, compared with the D1, EGFR signaling (p = 2.93E-4, FDR = 0.022), was notably attenuated in the D5 (Fig. 4b). In gastric cancer, we found the reduction of complement cascade when the input of AOPs was no more than 0.5 µg, and the loss function of NF-kappaB signaling (p = 1.15E-7, FDR = 4.22E-7), Wnt signaling (p = 1.24E-5, FDR = 1.27E-3), and cell cycle (p = 3.91E-6, FDR = 4.91E-4), indicating the useless when the input was in significantly light amounts. Together, we revealed the driver pathways for the carcinogenesis of duodenum cancer and gastric cancer, disclosed the loss functions with the dilutions of AOPs, and indicated the significantly light amounts was useless for scientific researches.