Neuroblastoma tyrosine kinase signaling networks involve FYN and LYN in endosomes and lipid rafts

Abstract

Protein phosphorylation plays a central role in creating a highly dynamic network of interacting proteins that reads and responds to signals from growth factors in the cellular microenvironment. Cells of the neural crest employ multiple signaling mechanisms to control migration and differentiation during development. It is known that defects in these mechanisms cause neuroblastoma, but how multiple signaling pathways interact to govern cell behavior is unknown. In a phosphoproteomic study of neuroblastoma cell lines and cell fractions, including endosomes and detergent-resistant membranes, 1622 phosphorylated proteins were detected, including more than half of the receptor tyrosine kinases in the human genome. Data were analyzed using a combination of graph theory and pattern recognition techniques that resolve data structure into networks that incorporate statistical relationships and protein-protein interaction data. Clusters of proteins in these networks are indicative of functional signaling pathways. The analysis indicates that receptor tyrosine kinases are functionally compartmentalized into distinct collaborative groups distinguished by activation and intracellular localization of SRC-family kinases, especially FYN and LYN. Changes in intracellular localization of activated FYN and LYN were observed in response to stimulation of the receptor tyrosine kinases, ALK and KIT. The results suggest a mechanism to distinguish signaling responses to activation of different receptors, or combinations of receptors, that govern the behavior of the neural crest, which gives rise to neuroblastoma.

Fig 1. Proteins that cluster with RTKs.

Proteins with tyrosine kinase, tyrosine phosphatase, SH2, and/or SH3 domains (PNCPs) that co-clustered with ALK (A), IGF1R (B), and EGFR (C) by t-SNE using both Spearman and Euclidean dissimilarity representations (“hard” filter), graphed as networks with PPI edges (left) and heat maps (right). The size and color of nodes is scaled to graph total phosphopeptides detected for each protein; blue represents phosphorylation on inhibitory sites (e.g., SFKs labelled “SFK_i”), yellow, all other sites. Heat maps (right) display the relative total phosphopeptide amounts for each protein on a blue-yellow scale (black represents NA; key, bottom right), sorted most to least left to right and top to bottom for samples and proteins, respectively, in each cluster.

Fig 2. The neuroblastoma cluster-filtered network (CFN).

Edges from the neuroblastoma network shown in S1 Fig were filtered to show only edges among proteins that co-clustered based on Spearman, Euclidean, or hybrid Spearman-Euclidean dissimilarity. Nodes are graphed as in Fig 1 using an edge-weighted, spring-embedded layout. Small isolated groups with limited interactions are at the bottom of the figure. Proteins that have no interaction edges within clusters are not shown. PNCPs in aggregated, co-clustering collaborative groups, indicated by shaded regions, are displayed in Fig 3. (The dark blue node next to the region shaded in green is CDK1 phosphorylated on its inhibitory site.)

Fig 3. Collaborative groups of RTKs.

Interaction networks show proteins that co-cluster and have Spearman correlation > 0.5 (yellow edges) or are known to interact from PPI databases (grey edges) with ALK, PDGFRA, FGFR1, and IGFR1 (A); EGFR, DDR2, EPHA2, EPHB3, and PDGFRB (B). Only edges linked to RTKs are shown. Proteins are grouped by the number of interactions with RTKs (e.g. the group in the center containing PAG1 and BCAR1 in A interacts with all four RTKs). Nodes were filtered to exclude those with lowest representation in the phosphoproteomic data. Node size and color indicates total phosphorylation as in Fig 2. Proteins that are both known to interact and have a positive correlation have two edges (e.g. ALK edges connecting FYN, SHC1, and IRS1). PAG1, FYN, and PIK3R2, the only nodes in common to both networks, are highlighted in green.

Fig 4. Fold change in response to RTK stimulation or inhibition.

Shown are changes of more than twofold from representative experiments where peak intensity was measured for treatment and control conditions in the same experiment with cell lines and treatments indicated on column labels (e.g., “NGF to C” means NGF-treated compared to control). (A) Total phosphorylation changes in tyrosine kinases. SFKs phosphorylated on their C-terminal inhibitory site were tracked separately (SFK_i). In addition to results summarized in the text, the ALK inhibitor, TAE684, inhibited RET and IGF1R phosphorylation about threefold in SH-SY5Y cells. NGF stimulated phosphorylation of IGF1R and PDGFRA, and BDNF treatment increased phosphorylation of FGFR1, in SMS-KCN cells. EGFR and EPHA2 were affected in opposite ways in LAN-6 and SH-SY5Y cells. AXL, PDGFRB, EPHA7, and EPHB1 phosphorylation were decreased by NGF in LAN6 cells. Individual phosphorylation site changes are shown for SFKs (B). Activating (SFK Y411-426) and inhibitory (SFK Y508-531) sites on FYN, LYN, YES1, and SRC were affected differently by different treatments. Phosphorylation sites represent the sum of all peptides surrounding that site; peptides whose conserved sequence is present in several proteins are indicated with multiple names, e.g., “FYN 420; LCK 394; SRC 419; YES1 426.” Fold changes are graphed on a blue-yellow color scale with blue representing a decrease, and yellow, an increase, compared to control (key). Data are sorted from most to least for each total row (protein or phosphorylation site) and column (treatment) from left to right and top to bottom, respectively.

Fig 5. PNCP enrichment in endosomes and DRM fractions.

(A) The most highly phosphorylated PNCPs that were present in endosome fractions from two or more cell lines, graphed as in Fig 1 except node size and color intensity represents total phosphorylation in endosome fractions. (B) Enrichment of proteins in endosome and DRM fractions was calculated as the ratio of amounts in endosomes or DRMs vs. the average in all other fractions and samples from that cell line, graphed as a heat map. (C) SFK and PAG1 phosphorylation site enrichment in LAN-6 cells (left) and SK-N-BE(2) cells (right).

Fig 6. Enrichment of RTKs in endosomes and DRM fractions.

Enrichment was graphed in PPI networks as big yellow nodes for positive enrichment and small blue nodes for de-enrichment. Green nodes of intermediate size indicate equal amounts in all fractions. PNCPs from LAN-6 endosomes (A) and DRMs (B); SK-N-BE(2) expressing TrkA (NTRK1) endosomes (C) and DRMs (D).

Fig 7. Intracellular localization of FYN and LYN changed in response to PTN and SCF.

(A-D) Velocity gradient fractionation of intracellular organelles after serum starvation (control; squares) or 60 min stimulation of LAN-6 cells with PTN (A, C) or SCF (B, D). Data were quantified from western blots using antibodies against FYN (A, B) and LYN (C, D) and expressed as the percent of each protein in each gradient fraction after quantifying amounts in all other cell fractions (percent in whole cell). Shown are means from 2–4 experiments for each condition; error bars are SEM. (E) Organelle fractions, defined as pools of velocity gradient fractions lys, E1, E2, E3, and cyt as shown in (C), were subjected to flotation equilibrium centrifugation [32]. Western blots show these fractions and detergent-resistant (DRM) and-soluble (P1M) fractions (see S10 Fig) after no treatment (C = control) or treatment with PTN, probed with antibodies to FYN and LYN (indicated). Both floating (F) and non-floating (NF, defined as material in higher density fractions at the bottom of flotation equilibrium gradients) membranes were analyzed. That SFKs associated with floating (F) fractions indicates that they were robustly bound to membranes. (F) Phospho-SFK (left) and non-phospho-SFK antibodies (right) were used to immunoprecipitate proteins from endosome E1 fractions under unstimulated or stimulated conditions as in (A). Western blots were probed with FYN- and LYN-specific antibodies (indicated). (G) Box plot shows amounts of FYN (red) and LYN (blue) in detergent-resistant (DRM) and-soluble (P1M) fractions under control, ALK- or KIT-stimulated conditions as in A-D. (H) Bar plots show fold change (treatment/control if positive;-(treatment/control)^-1 if negative) in all cell fractions under unstimulated or stimulated conditions as in A-D. (G, H) Amounts of FYN and LYN in all cell fractions were quantified from 3–7 experiments; boxes show quartiles and whiskers show ranges in G, error bars are SEM in H.

Fig 8. Phosphorylation site clusters displayed as a co-clustered correlation networks (CCCNs).

RTK, SFK and PAG1 phosphorylation sites were selected from the co-cluster correlation network shown in S11 Fig, where edges represent Spearman correlations greater than 0.5 from phosphorylation sites that co-clustered from t-SNE embeddings. Edge line thickness is proportional to correlation and the number of cases in which phosphorylation sites co-clustered from different embeddings. Two separate groups containing the most highly phosphorylated RTK and SFK sites are shown. Heat maps (right) show the primary data for phosphorylation on specific sites. (A) The network containing phosphorylation sites ALK 1507 (the most highly phosphorylated site on ALK), FYN 420, PDGFRA 762 and LYN 508 was extended to include other PNCP phosphorylation sites that co-clustered with positive Spearman correlation; no PAG1 phosphorylation sites co-clustered with this group. (B) Five highly phosphorylated PAG1 sites co-clustered with phosphorylation sites on IGF1R, DDR2, EGFR, and FGFR1. Note that the conserved peptide sequence for FYN 531; SRC 530; YES1 537 was inclusively summed to both FYN 531; YES1 537 and SRC 530 in the phosphorylation site network, but exclusively summed to FYN_i in the total protein phosphorylation networks based on the presence of other FYN phosphopeptides in the same samples. By the exclusive criteria, FYN 420; LCK 394; SRC 419; YES1 426 in (A) most likely represents FYN activation; FYN 531; YES1 537 most likely represents FYN inhibitory phosphorylation; and the inclusive summing method likely over-represents the amounts of SRC inhibitory phosphorylation in (B).