Overview of proteomic landscape of trace FFPE samples in EESCC and EDAC

To characterize the comprehensive proteomic landscape of trace FFPE samples, we collected proteomics and phosphoproteomics data from 3 early-stage ESCC (EESCC) patients and 3 early-stage DC (EDAC) patients who had not experienced prior chemotherapy or radiotherapy. A schematic of the experimental design is shown in Fig. 1a and Supplementary Fig. S1a. Proteomic analysis was performed on the basis of mass spectrometry (MS)-based label-free quantification strategy [11, 20]. Protein abundance of all samples was firstly calculated by intensity-based absolute quantification (iBAQ) [21, 22] and then normalized as a fraction of the total (FOT), allowing for comparisons between experiments.

Fig. 1

Proteomic landscape of trace FFPE samples in EESCC and EDAC. a Overview of the experimental design. A total of 12 samples were collected from EESCC (3 cases) and EDAC (3 cases). b Proteomic landscape of EESCC. c Proteomic landscape of EDAC. d The number of identified proteins in EESCC (purple) and EDAC (red). e Phosphoproteomic profiles of EESCC. f Phosphoproteomic profiles of EDAC. The high abundance proteins/phosphoproteins were shown in the box. g The number of identified phosphosites in EESCC and EDAC (t-test)

At the protein level, identification of ~ 6000 proteins were observed in single trace FFPE sample. A total of 4667 and 6450 were identified in the normal and tumor tissues of EESCC, respectively (Supplementary Fig. S1b – d). In EDAC, 6562 and 6789 proteins were detected in the normal and tumor tissues of EDAC, respectively. Interestingly, the high abundance identifies were consistent both in the tumor and normal tissues in EESCC and EDAC at the protein level, including HBB, HBA1, HBD, ACTB, ALB, etc. (Fig. 1b – d; Supplementary Table S1). Comparably, those high abundance proteins were also expressed in the human stomach cancer [23] and lung adenocarcinoma [24] and their corresponding normal tissues. These findings implied that those proteins might be highly expressed in all human tissues. Additionally, ESCC biomarkers identified in previous study, such as ACTA2, ANXA1 [25], HSPA9, THBS1 [26] etc. were also detected in EESCC, indicating the key events in advanced-stage cancer happened as earlier as in the early-stage cancer.

At the phosphoprotein level, identification of over 10,000 phosphosites was observed in the single trace FFPE sample. A total of 18,072 phosphosites corresponding to 4173 phosphoproteins, and 12,200 phosphosites corresponding to 3390 phosphoproteins were identified in tumor tissues of EESCC and EDAC, respectively (Fig. 1e – g). In addition, the highly phosphorylation (e.g., HSPB1 S82, KRT14, KRT13 S427, etc.) were both detected in tumor tissues and their paired normal tissues in EESCC (Fig. 1e; Supplementary Table S2). Whereas, the highly phosphorylation in EDAC were STK24 S4, THRAP3 S682, TMPO T74, etc. at the phosphoprotein level (Fig. 1f). Observation of significantly more phosphosites were detected in tumor tissues compared with paired normal tissues in EESCC (t test, p = 0.013), which was not notable in EDAC (Fig. 1g). These findings indicated the similarly high abundance identifies in EESCC and EDAC at the protein level, whereas the highly expressed phosphoproteins were different. Overall, we established a comprehensive landscape of trace FFPE samples in early-stage gastrointestinal cancer (EESCC and EDAC) at multi-omics levels, and identified ~ 6000 proteins and > 10,000 phosphosites in single FFPE samples.

Proteomic characterization of tumor tissues compared with paired normal tissues

To investigate the differential proteomic features of tumor tissues compared with paired normal tissues, we performed principal component analysis (PCA) at the protein and phosphoprotein levels. To an end, we found obvious separation between normal tissues and tumor tissues of EESCC at the protein level and phosphoprotein levels, as well as in EDAC (Fig. 2a, d, and Supplementary Fig. S2a, d). We then integrated the differential expressed proteins and phosphoproteins (DEPs) between the tumor tissues and paired normal tissues in EESCC and EDAC, respectively. In EESCC, we found the normal tissues highly expressed proteins/phosphoproteins participated in the primary physiological functions of the esophagus, such as keratinization (e.g., TGM1, KLK13, etc.) and metabolism (e.g., PGD, PLCD1, etc.), etc. (Fig. 2b, c, g, and Supplementary Fig. S2b, c). Over-representation of glycolysis and cell cycle were observed in the tumor tissues of EESSCC.

Fig. 2

The proteomic features of tumor tissues compared with paired normal tissues in in EESCC and EDAC. a PCA analysis showing distinct separation between the tumor tissues and paired normal tissues in EESCC at the protein level. b Volcano analysis depicted the differential expressed proteins of the tumor tissues and paired normal tissues in EESCC at the protein level. c Bar chart presenting the functional pathways in up-regulated and down-regulated pathways in EESCC at the protein level. d PCA analysis showing distinct separation between the tumor tissues and paired normal tissues in EDAC at the protein level. e Volcano analysis depicted the differential expressed proteins of the tumor tissues and paired normal tissues in EDAC at the protein level. f Bar chart presenting the functional pathways in up-regulated and down-regulated pathways in EDAC at the protein level. g A brief summary descripting the differential proteomic features of tumor tissues and paired normal tissues in EECC (left) and EDAC (right)

In EDAC, immune response (e.g., PTK2B, TRADD, etc.) was prominent in the normal tissues, including interleukins signaling, antigen processing and presentation (Fig. 2e, f, g, and Supplementary Fig. S2e, f). AMPK and TNF receptor signaling (e.g., MTOR, MAP4K1) were overrepresented in the tumor tissues of EDAC at the protein and phosphoprotein levels. As well as in EESCC, cell cycle (e.g., PCM1, LIG1, etc.) and glycolysis (e.g., PGK2, ENO3, etc.) were predominant in the tumor tissues of EDAC. Clinical Proteomic Tumor Analysis Consortium (CPTAC) show that glycolysis can be a potential target to overcome the resistance of microsatellite instability-high (MSI-H) tumors to immune checkpoint blockade [10]. These findings suggested that MSI-H might be prevalent in the gastrointestinal cancer, and glycolysis could be a potential target in EESCC and EDAC. Taken together, we disclosed the proteomic features of tumors compared with paired normal tissues in EESCC and EDAC.

Proteomic characterization of EESCC and EDAC

To investigate the correlations between the proteome and phosphoproteome of EESCC and EDAC, we performed integration analysis and found significant association between proteome (n = 6450) and phosphoproteome (n = 4173) in EESCC (Pearson correlation, R = 0.25, p < 2.2E-16) (Supplementary Fig. S3a). As well, notably association between proteome (n = 6789) and phosphoproteome (n = 3390) in EDAC (Pearson correlation, R = 0.19, p = 8.1E-15) (Supplementary Fig. S3b), which allowed us to further investigating the characterizations of early-stage cancer.

Spearman’s correlation showed lower coefficients (mean = 0.62) between EESCC and EDAC than the same cancer type (mean = 0.74 (ESCC) and 0.78 (EDAC)) indicated the difference in gastrointestinal cancer (Fig. 3a). To explore the difference between EESCC and EDAC, we performed PCA at the protein and phosphoprotein levels. Visualization of PCA differentiated the proteome profiles between EESCC and EDAC, as well as at the phosphoprotein level (Fig. 3b). Interestingly, notable separation was also observed in the normal tissues between EESCC and EDAC at the protein and phosphoprotein levels (Supplementary Fig. S3c). The distinct separation suggested the fundamental difference between EESCC and EDAC.

Fig. 3

Proteomic characterization of EESCC and EDAC. a Spearman’s correlation coefficients among 6 gastrointestinal cancer samples. b PCA analysis showing distinct separation between EESCC and EDAC at the protein (left) and phosphoprotein (right) levels. c Comparative analysis the dominant pathways of normal tissues (top) and tumor tissues (bottom) in EESCC and EDAC. d Proteins in functional pathways that were differentially expressed in EESCC and EDAC at protein and phosphoprotein levels. e A brief of the differential proteins and functional pathways in EESCC (top) and EDAC (bottom). f Boxplot showing MTOR was highly expressed in EDAC at the protein (left) and phosphoprotein (right) levels (t-test). g Boxplot showing MTOR (S1261) was highly expressed in the tumor tissues compared with paired normal tissues (t-test). h Pearson’s correlation coefficients indicated significantly positive association between MTOR proteome and phosphoproteome (S1261)

To access whether the differential features of the tumor tissues were derived from the origins of their normal tissues, we compared the proteomes of normal tissues of EESCC and EDAC. Notably, the dominant pathways of normal tissues in the EESCC were enriched in keratinization and cell adhesion, which were also over-represented in the tumor tissues of EESCC (Fig. 3c). In the EDAC, AMPK signaling, canonical glycolysis, ATP biosynthetic process, tricarboxylic acid (TCA) cycle, and fatty acid beta oxidation, were dominant in the normal tissues and tumor tissues. These findings suggesting the diversity in the EESCC and EDAC was derived from their origin normal tissues.

The significance analysis of microarray (SAM) [27] was performed to investigate the characteristics of EESCC and EDAC at the protein level, which identified 791 differentially expressed proteins (DEPs) between EESCC and EDAC (t-test, p < 0.05, fold change (FC) (EDAC/EESCC) ≥ 2 or ≤ 0.5), including 678 elevated (EDAC-proteins) and 113 descend proteins (EESCC-proteins) (Supplementary Fig. S3d). GO-enrichment analysis presented that the EESCC-proteins were related to the primary functions of esophagus tissues, including keratinization (p = 2.0E-4) (e.g., CDH1, KRT2, etc.), cell division (p = 1.2E-3) (e.g., NCAPD2, MCMBP, etc.), and epidermis development (p = 1.7E-3) (e.g., TCHH, CETN2, etc.) (Fig. 3d, e). The EDAC-proteins participated in Mapk signaling (p = 7.7E-3) (e.g., MAPK1, MAPK3, etc.) and metabolic processes (p = 2.0E-4) (e.g., ACO2, PAK1, etc.), covering tricarboxylic acid cycle (TCA) (p = 2.6E-9) (e.g., SDHA, IDH1, etc.), fatty acid-beta-oxidation (p = 3.6E-8) (e.g., ACOX2, CPT2, etc.), ATP biosynthesis process (p = 6.5E-6) (e.g., ATP5B, ATP5S, etc.), canonical glycolysis (p = 2.8E-4) (e.g., PFKL, TPI1, etc.).

Observation of the difference between EESCC and EDAC was also identified at the phosphoprotein level. For example, the phosphorylation of keratinization−/epidermis development−/cell division- related phosphoproteins were detected in EESCC, such as KRT2 S26, KRT5 S571, CDH3 T694, TGM3 S471, NCAPD2 S13, MCMBP S298, etc. In addition, the phosphorylation of MAPK T185/Y187, MAPK3 T202, ACO2 S79, ATP5B T140, PFKL S763, etc. was elevated in EDAC (Supplementary Fig. 3e, f). Notably, the high phosphorylation was also observed in the tumor tissues compared paired normal tissues of EESCC and EDAC. Furthermore, we found the EESCC-phosphoproteins were associated in cell cycle (p = 1.9E-3) (e.g., RB1 T373, RBL2 S662, etc.) and signal transduction (p = 1.3E-3) (e.g., EGFR T693, ROCK1 S1108, etc.). The mutation of RB1 was prevalent in ESCC [28], and the mutation of EGFR was gefitinib-sensitizing mutation in ESCC [29]. In our datasets, we found the significant association between RB1 T373 and EGFR T693 (Pearson correlation, R = 0.97, p = 1.7E-3) at the phosphoprotein level, indicating the co-functions in RB1 and EGFR in esophageal carcinogenesis (Supplementary Fig. S3g). The EDAC-phosphoproteins were involved in regulation of apoptosis (p = 9.9E-4) (e.g., APC S2093, ATM T1885, etc.). The mutation of MTOR was the driven mutation for small-bowel carcinoma, and is regarded as the potential targets in the clinic [30, 31]. In this study, we found the higher expression of MTOR (t-test, FC (EDAC/EESCC) = 4.1, p = 0.19) and MTOR S1261 (t-test, FC (EDAC/EESCC) = 4.4, p = 0.04) in EDAC (Fig. 3f). Specifically, MTOR S1261 was also significantly expressed in the tumor tissues in EDAC (t-test, FC (tumor/normal) = 5.3, p = 0.019), which showed significantly positive association (Pearson correlation, R = 0.96, p = 2.4E-3) (Fig. 3g, h). Collectively, this study presented comprehensive proteomic characterization of EESCC and EDAC at multi-omics level, and revealed the functional pathways of cell cycle in EESCC, the positive impacts of apoptosis, and metabolic processes in EDAC, and further demonstrated the potential co-functions of RB1 and EGFR in ESCC, and the key role of MTOR at the protein and phosphoprotein levels in EDAC.

Immune-based features of EESCC and EDAC

Recent studies have well-established the connection between inflammatory and tumorigenesis, and have considered the inflammatory is an important risk factor for gastrointestinal cancer [32]. To gain insight into features of immune infiltration of early-stage gastrointestinal cancer, we analyzed the proteomic profiles of EESCC and EDAC, and deconvoluted immune, stromal, and microenvironment cell signature using xCell (https://xcell.ucsf.edu) [33]. As a results, we found the cellular characteristics of dendritic cells (DCs) was dominant in EESCC, evidenced by the highly expressed biomarkers at the protein level, such as cluster of differentiation 14 (CD14), CD276, and CD36 (Fig. 4a, b; Supplementary Table S3). Furthermore, the cell signatures of endothelial cells, epithelial cells, and keratinocytes were prominent in EESCC, implying the specific characteristics of the esophagus tissues. Specifically, higher immune score (t-test, FC (EDAC/EESCC) = 1.39, p = 0.04) and microenvironment score (t-test, FC (EDAC/EESCC) = 1.35, p = 0.03) were observed in EDAC, evidenced by the overrepresentation of cell signatures, such as CD8+ Tem, B cells, monocytes, neurons, and platelets (Fig. 4a). In addition, the cell markers of B cells (e.g., CD200 and CD38) and T cells (e.g., CD226 and CD81) were also overrepresented in EDAC (Fig. 4b). In EDAC, the molecules of major histocompatibility complex class I and II (MHC-I/II), were highly expressed, including HLA-B, HLA-C, HLA-E, HLA-DQA1, HLA-DQB1, HLA_DRA, HLA-DRB1, etc. (Fig. 4c).

Fig. 4

Immune-based features of EESCC and EDAC. a Heatmap showing the immune infiltration of EESCC and EDAC at the protein level. Top: heatmap showing the immune and microenvironment score. Bottom: heatmap showing the cell signatures. b The proteome-level expression of the biomarkers of DCs, B cells, and T cells in EESCC and EDAC. c Expression of MHC-I (top) and MHC-II (bottom) in EESCC and EDAC at the protein level. d Pie chart showing eight major classifications of cytokines identified in early-stage gastrointestinal cancer. e Bar charts illustrating the proportion of eight major classifications of cytokines in ESSCC (purple) and EDAC (red). f Heatmap showing the expression of the proteins of the major cytokines in EESCC and EDAC at the protein level. g Heatmap showing the expression of the proteins of the major cytokines in EESCC and EDAC at the phosphoprotein level

The increased pro-inflammatory cytokines permit immune activity in disease [34]. To gain insight of the difference of early-stage gastrointestinal cancer, we performed cytokines-classification analysis of all identified proteins (n = 7870) in early-stage gastrointestinal cancer. As a result, we identified 153 cytokines, which were classified into 8 major cytokines types as following: chemokine (2.6%, n = 4), interferon (13.7%, n = 21), interleukin (8.5%, n = 13), growth factor (23.5%, n = 36), integrin (20.9%, n = 32), matrix metalloproteinase (MMP) (3.3%, n = 5), tumor necrosis factor (5.9%, n = 9), and CD (21.6%, n = 33) (Fig. 4d). Comparative analysis elucidated proteins of chemokines, tumor necrosis factor, and CD were prevalent in EDAC, while the proteins of interferon, interleukin, growth factor, integrin, and MMP were prominent in EESCC (Fig. 4e and Supplementary Fig. S4). Specifically, the expression of IFI16 (S106 and S568), IFIT3 (S478), IRF2BP1 (S453), IRF2BPL (S69 and S547), IL1RAP (S556 and S557), ILF3 (S482 and T592), EGFR (T693 and S991), IGF2R (S2409), ICAM3 (S532), ITGA5 (S123 and S126), TRAF4 (S426), etc. was overrepresented in EESCC at the protein and phosphoprotein levels (Fig. 4f, g). In addition, we found those proteins/phosphoproteins were overrepresented in the tumor tissues of ESCC compared with paired normal tissues. In EDAC, the overrepresentation of the proteome expression and phosphorylation of CD44 (T720), CD300A (S260), and CD2AP (T87 and S458), was detected, which was also notably observed in the tumor tissues compared with paired normal tissues of EDAC. Taken together, we elucidated the higher immune infiltration in EDAC, disclosed the cytokines classification, and revealed the specific immune features of EESCC and EDAC.

Proteomic kinases profiles of EESCC and EDAC

Human protein kinases mediate the majority of signal transduction pathways in biological processes, including cell metabolism, cell cycle, apoptosis, etc. [35], and the novel targets and inhibitors are applied to clinic strategy [36, 37]. To access the specific kinases of early-stage gastrointestinal cancer, the identified protein kinases (n = 166) were mapped to the human protein kinases. As a result, we depicted a Kinome Tree (Supplementary Fig. S5a), and found the more kinases of casein kinase 1 (CK1) group (57.1%), cyclin-dependent/mitogen-activated protein kinase (CMGC) group (58.3%), tyrosine-protein kinase (/receptor) (TK) group (55.3%), and Serine/Threonine-protein kinase (/receptor) (TKL) (54.5%) were detected in EESCC (Fig. 5a and Supplementary Fig. S5b). In addition, the kinases of cAMP/cGMP/calcium/phospholipid-dependent kinase (AGC) (73.9%), serine/tyrosine/threonine protein kinase (STE) group (73.1%), and Ca2+/CaM-dependent protein kinase (CAMK) group (67.9%) were overrepresented in EDAC (Fig. 5b).

Fig. 5

Proteomic kinases profiles in EESCC and EDAC. a Distribution and the number of the seven major kinases types in EESCC (purple) and EDAC (red). b Bar charts showing the differential proportion of major kinases types in EESCC (purple) and EDAC (red). c Heatmap representing the expression of the proteins of major kinases types in EESCC and EDAC at the protein level. d Heatmap representing the expression of the proteins of major kinases types in EESCC and EDAC at the phosphoprotein level. e Enrichment of the kinases and the downstream substates showing the dominant pathways in EESCC (purple) and EDAC (red). P value < 0.05 and FDR q value < 0.3 were considered as significant enrichment. f Heatmap showing the expression of the inhibitors (FDA drug) to kinases (top) (proteome) and the kinases regulated substates (phosphoproteome) in EESCC and EDAC. g Pathways based on the selected phospho-substates and kinases, with relevant drugs shown by targets in EESCC (left) and EDAC (right)

Specifically, CSNK1G3, ICK, CDK1, CDK2, EGFR, FGFR1, PDGFRB, EPHB3, MAP3K7, ZAK, etc. were prominent in EESCC at the protein level. Consistently, the phosphorylation of CSNK1G3 T374, CDK1 T14, ICK S584, EGFR T693 and S991, LCK Y394, PDGFRB S712, EPHB3 T613, MAP3K7 S389 and ZAK S637, were overrepresented in EESCC at the phosphoprotein level (Fig. 5c, d). Comparably, those specific proteins/phosphoproteins were also highly expressed in the tumor tissues compared with paired normal tissues of EESCC. In EDAC, higher expressions of the proteome and phosphoproteome of PDPK1 (S241), PRKCD (T507), RPS6KA1 (S221), PRKACA (S339), MARK1 (S219), DAPK2 (S229), CAMK2D (T306), MAP3K2 (S153), PAK1 (S165), etc. were identified in EDAC. Consistently, those higher proteome−/phosphoproteome levels were also detected in the tumor tissues than paired normal tissues of EDAC.

To explore the functions of those kinase, we performed kinases-substates relationship analysis on the basis of kinases-substrates database [14, 16], and integrated the substrates in EESCC (n = 249) and EDAC (n = 297) (Supplementary Table S4). GO-enrichment analysis of those substates revealed the elevation of cell cycle (p = 2.9E-10), p53 signaling pathways (p = 5.2E-5), base excision repair (p = 4.6E-4), nucleotide excision repair (p = 2.4E-3), and Ras signaling pathways (p = 1.9E-3) in EESCC, and disclosed the overrepresentation of mTOR signaling (p = 3.3E-5), PI3K-AKT signaling pathways (p = 1.4E-5), Mapk signaling pathways (p = 2.6E-6), ErbB signaling pathways (p = 2.1E-13) in EDAC (Fig. 5e).

Kinases were applicated to the clinic strategy, such as anti-EGFR (abemaciclib) in breast cancer [38], and anti-MAP2K1 (trametinib) in melanoma [39]. We then accessed the drug targets approved by the US Food and Drug Administration database (FDA) (https://www.fda.gov) [40]. To an end, we found 6 kinases in our study were recorded in FDA datasets, in which EGFR, WEE1, CDK4, and PDGFRB were prominent in EESCC, and MAP2K1 was prevalent in EDAC (Fig. 5f). Phosphorylation impacts multiple cellular processes, with site occupancy tightly regulated by the activity of kinases and phosphatases [41]. We then performed integrative analysis of the differential kinases-substates (site), and proposed the functions of drugs approved by FDA in EESCC and EDAC. In EESCC, anti-EGFR with abemaciclib decreased the expression of EGFR (T693) and the downstream phosphorylation of GAB1 (S547), and anti-PDGFRB with imatinib in-activated the phosphorylation of PDGFRB (S712) and ABL (S210), which participated in cell cycle. Additionally, the inhibitor of ribociclib to WEE1 down-regulated the CDK1 at the protein and phosphoprotein levels, and the anti-CDK4 with afatinib decreased the phosphorylation of RBL2 (S662) and RB1 (T373), resulting in the stability of cell cycle-checkpoint which was the final safeguard of genomic fidelity (Fig. 5g). In EDAC, the inhibitor of trametinib to MAP2K1 was negative associated with the phosphorylation of MAPK1 (T185) and MAPK3 (T202), which down-regulated Mapk signaling and inhibited cell proliferation in EDAC (Fig. 5g). Collectively, we revealed EESCC-specific and EDAC-specific kinases, elucidated the functional kinases-substates relationship network, and proposed the potential clinic strategy in EESCC and EDAC, providing a novel insight for gastrointestinal cancer in the clinic.