Proteomic and transcriptomic analyses identify apo-transcobalamin-II as a biomarker of overall survival in osteosarcoma

Abstract

Background: The large-scale proteomic platform known as the SomaScan® assay is capable of simultaneously measuring thousands of proteins in patient specimens through next-generation aptamer-based multiplexed technology. While previous studies have utilized patient peripheral blood to suggest serum biomarkers of prognostic or diagnostic value in osteosarcoma (OSA), the most common primary pediatric bone cancer, they have ultimately been limited in the robustness of their analyses. We propose utilizing this aptamer-based technology to describe the systemic proteomic milieu in patients diagnosed with this disease.

Methods: To determine novel biomarkers associated with overall survival in OSA, we deployed the SomaLogic SomaScan® 7k assay to investigate the plasma proteomic profile of naive primary, recurrent, and metastatic OSA patients. Following identification of differentially expressed proteins (DEPs) between 2-year deceased and survivor cohorts, publicly available databases including Survival Genie, TIGER, and KM Plotter Immunotherapy, among others, were utilized to investigate the significance of our proteomic findings.

Results: Apo-transcobalamin-II (APO-TCN2) was identified as the most DEP between 2-year deceased and survivor cohorts (Log2 fold change = 6.8, P-value = 0.0017). Survival analysis using the Survival Genie web-based platform indicated that increased intratumoral TCN2 expression was associated with better overall survival in both OSA (TARGET-OS) and sarcoma (TCGA-SARC) datasets. Cell-cell communication analysis using the TIGER database suggested that TCN2+ Myeloid cells likely interact with marginal zone and immunoglobin-producing B lymphocytes expressing the TCN2 receptor (CD320) to promote their proliferation and survival in both non-small cell lung cancer and melanoma tumors. Analysis of publicly available OSA scRNA-sequencing datasets identified similar populations in naive primary tumors. Furthermore, circulating APO-TCN2 levels in OSA were then associated with a plasma proteomic profile likely necessary for robust B lymphocyte proliferation, infiltration, and formation of intratumoral tertiary lymphoid structures for improved anti-tumor immunity.

Conclusions: Overall, APO-TCN2, a circulatory protein previously described in various lymphoproliferative disorders, was associated with 2-year survival status in patients diagnosed with OSA. The relevance of this protein and apparent immunological function (anti-tumor B lymphocyte responses) was suggested using publicly available solid tumor RNA-sequencing datasets. Further studies characterizing the biological function of APO-TCN2 and its relevance in these diseases is warranted.

Keywords: immunotherapy; osteosarcoma; plasma; proteomics; transcobalamin-II.

Figure 1

Overview of experimental design. (A) Proteomic investigation of OSA patient plasma samples using the SomaLogic SomaScan® 7k assay. Step-by-step process by which the SomaScan platform measures circulatory proteins (provided by SomaLogic Inc.). (B) Downstream analyses including cell identification using scRNA-seq analysis, overall survival analysis in publicly available datasets, and correlation analyses to support identified protein’s role or association in solid tumors. Figure generated using BioRender.com. (C) Tree diagram depicting those OSA patient plasma samples selected for downstream analyses in this study.

Figure 2

Proteomic analysis of naive OSA patients, stratified by 2-year survival status. (A) Volcano plot depicting those differentially expressed proteins (DEPs – blue) with a Log2FC > 0.585 or < -0.585 and P-value < 0.05. (B) Reactome analysis identifies those pathways associated with 2-year survival in OSA. Significant pathways include those with multiple entities and FDR < 0.05. Box plot depicting APO-TCN2 [seq.15560.52] (C) and HOLO-TCN2 [seq.5584.21] (D) levels in all naive OSA patients (n = 14) stratified by 2-year survival status, with n = 3 deceased and n = 11 survivors. Box plot depicting APO-TCN2 [seq.15560.52] (E) and HOLO-TCN2 [seq.5584.21] (F) levels in naive OSA patients with advanced disease (n = 5) stratified by 2-year survival status, with n = 2 deceased and n = 3 survivors. Each dot of the box plot is representative of an individual OSA patient. Y-axis represents the relative fluorescence units (RFUs) in thousands. Significant P-values (t-test) are presented for each comparison. Median values from assessment of plasma samples in a normal, healthy patient population (provided by SomaLogic, Inc.) are indicated by orange bar.

Figure 3

TARGET-OS and TCGA-SARC survival analysis supports TCN2 as a marker of better overall survival. Analysis of TCN2 in TARGET-OS (A) and TCGA-SARC (B) databases using Survival Genie web-based platform. Box plots representing TCN2 FPKM normalized expression in primary tumors with stratification into low and high expressing groups. The relative fraction of TILs was estimated using the tumor-infiltrating immune cell type matrix LM22 gene signature and CIBERSORTx deconvolution. Pearson correlation matrix of deconvoluted immune cell RNA-seq gene expression data and TCN2, with shape (square or circle) denoting significance and color denoting positive (red) or negative (blue) correlation with TCN2. Kaplan-Meier (KM) survival curves with high (red) and low (blue) group stratification are compared using the log-rank test, with log-rank P-value < 0.05 considered statistically significant. Forest plot details the association between the high and low groups [stratified by cut point (CP)] based on the Cox Proportional Hazards regression model. nLow and nHigh represent the number of patients in low and high expressing groups, respectively. Hazard ratio (HR) with 95% confidence interval as well as the associated Wald-test (HR) and log-rank (LR) P-values are reported.

Figure 4

TIGER database confirms TCN2 expression on various myeloid and B cellular clusters of NSCLC datasets. (A) Cell Type Marker (Log2FC) diagram represents differential expression of TCN2 on various clusters and subclusters in each scRNA-seq tumor dataset of the TIGER database, with differential expression represented by circle size and distance from center. (B, C) UMAPs of all cellular clusters of the NSCLC dataset highlight expression of TCN2 on myeloid and endothelial cells. (D, E) UMAPs of the Myeloid subclusters of the NSCLC dataset highlighting expression of TCN2. (F, G) UMAPs of all cellular clusters of the NSCLC6 dataset highlights expression of TCN2 on endothelial and plasma cells. (H, I) UMAPs of the B cell sub-clusters of the NSCLC6 dataset highlighting expression of TCN2. UMAPs generated using the TIGER portal.

Figure 5

TIGER database confirms CD320 expression on various myeloid and B cellular clusters of NSCLC datasets. (A) Cell Type Marker (Log2FC) diagram represents differential expression of CD320 on various clusters and subclusters in each scRNA-seq tumor dataset of the TIGER database, with differential expression represented by circle size and distance from center. (B, C) UMAPs of all cellular clusters of the NSCLC5 dataset highlights expression of CD320 on endothelial, malignant, and fibroblast cellular clusters. (D, E) UMAPs of the B cell subclusters of the NSCLC5 dataset highlighting expression of CD320. (F, G) UMAPs of B cell subclusters of the NSCLC dataset highlighting expression of CD320. (H, I) UMAPs of B cell subclusters of the NSCLC1 dataset highlighting expression of CD320. UMAPs generated using the TIGER portal.

Figure 6

scRNA-seq analysis of naive primary OSA samples reveals CD320 and TCN2 expression. All clustering representative of patient OS-6 (A) All cellular clusters identified by canonical marker expression. (B) Representative log-normalized CD320 expression across various cellular clusters. (C) Representative log-normalized TCN2 expression across various cellular cluster of patient OS-6. (D) Representative clustering of Plasmocytes with log-normalized expression of IGHG1 and MZB1 using the “Feature Min” function. (E) Representative clustering of B cells with log-normalized expression of MS4A1 and CD79A using the “Feature Min” function. (F) Representative clustering of Myeloid cells with log-normalized expression of LYZ and CD68 using the “Feature Min” function. (G) Representative clustering of CD320+ Plasmocytes with log-normalized expression of CD320, IGHG1, and MZB1 using the “Feature Min” function. (H) Representative clustering of CD320+ B cells with log-normalized expression of CD320, MS4A1, and CD79A using the “Feature Min” function. (I) Representative clustering of TCN2+ Myeloid cells with log-normalized expression of TCN2, LYZ, and CD68 using the “Feature Min” function. (J) Quantification of total cells for the CD320+ Plasmocytes, CD320+ B cells, and TCN2+ Myeloid cells populations in all naive OSA patient tumors (n = 6). (K) Quantification of the percent (%) of total cells for CD320+ Plasmocytes, CD320+ B cells, and TCN2+ Myeloid cells population in all naive OSA patient tumors (n = 6). (L) Quantification of the percent (%) of TCN2+ Myeloid cells, relative to all Myeloid cells in all naive OSA patient tumors (n = 6).

Figure 7

Correlation between APO-TCN2 and markers of plasma cell maintenance, B cell activation, and TLS formation. Spearman correlation analysis between measured APO-TCN2 (x-axis, RFUs) and markers of plasma cell maintenance including IL-6 [seq.2573.20, seq.4673.13] (A, B), IL-6Ra [seq.15602.43, seq.4139.71, seq.8092.29] (C–E), IL-6Rb [seq.2620.4] (F), CXCL12 [seq.2330.2, seq.3516.60, seq.9278.9] (G–I), and TNF [seq.5692.79, seq.5936.53] (J, K), markers of B cell activation including IL2RA [seq.3151.6] (L), TNFSF13B [seq.3059.50] (M), TNFRSF13B [seq.2704.74] (N), and TNFRSF17 [seq.2665.26] (O), as well as markers of TLS formation including LTa1b2 [seq.3505.6] (P), FCRL5 [seq.6103.70] (Q), SELL [seq.4831.4] (R), and TNFSF14 [seq.5355.69, seq.5988.49] (S, T) (y-axis, RFUs). Spearman correlation coefficient (R) and P-value (P) presented on each scatter plot. Each chemokine identified by multiple SOMAmers [example: TNFSF14 (1) and (2), representing seq.5355.69 and seq.5988.49, respectively] is presented with multiple correlation plots.

Figure 8

Correlation analysis between APO-TCN2 and measured chemokines of the 12-CK signature. Spearman correlation analysis between measured APO-TCN2 (x-axis, RFUs) and CCL2 [seq.2578.67] (A), CCL3 [seq.3040.59] (B), CCL5 [seq.2523.31, seq.5480.49] (C, D), CCL8 [seq.13748.4, seq.2785.15] (E, F), CCL18 [seq.3044.3] (G), CCL19 [seq.4922.13] (H), CCL21 [seq.2516.57] (I), CXCL9 [seq.11593.21, seq.9188.119] (J, K), CXCL10 [seq.4141.79] (L), CXCL11 [seq.18171.25, seq.3038.9] (M, N), CXCL13 [seq.13701.2, seq.3487.32] (O, P) (y-axis, RFUs) in naive OSA patient plasma samples (n = 14). Spearman correlation coefficient (R) and P-value (P) presented on each scatter plot. Each chemokine identified by multiple SOMAmers [example: CCL5 (1) and (2), representing seq.2523.31 and seq.5480.49, respectively] is presented with multiple correlation plots.

Figure 9

KM Plotter Immunotherapy survival analysis supports TCN2 and CD320 as a markers of immunotherapy response. Overall (A) and progression free (B) survival analysis of TCN2. Overall (C) and progression free (D) survival analysis of CD320. Each analysis included specimens from all solid tumors, collected pre-treatment, irrespective of immunotherapy target (anti-PD-1, anti-PD-L1, and anti-CTLA-4). Default KM Plotter settings were utilized with auto selection of best cutoff into high (red) and low (black) gene expression groups based on the calculation of all upper and lower quartiles with selection of the best performing threshold. KM survival curves with reported HR and log-rank P-value were constructed, with y-axis representing probability of survival and the x-axis representing time (months). The total number of patients at risk for each time point is reported. Survival curves generated using KM Plotter Immunotherapy.