Uncovering the promiscuous activity of IL-6 proteins: A multi-dimensional analysis of phylogeny, classification and residue conservation

Abstract

The IL-6 family of cytokines, known for their pleiotropic behavior, share binding to the gp130 receptor for signal transduction with the necessity to bind other receptors. Leukemia inhibitory factor receptor is triggered by the IL-6 family proteins: leukemia inhibitory factor (LIF), oncostatin-M (OSM), cardiotrophin-1 (CT-1), ciliary neurotrophic factor (CNTF), and cardiotrophin-like cytokine factor 1 (CLCF1). Besides the conserved binding sites to the receptor, not much is known in terms of the diversity and characteristics of these proteins in different organisms. Herein, we describe the sequence analysis of LIF, OSM, and CT-1 from several organisms, and m17, a LIF ortholog found in fishes, regarding its phylogenetics, intrinsic properties, and the impact of conserved residues on structural features. Sequences were identified in seven classes of vertebrates, showing high conservation values in binding site III, but protein-dependent results on binding site II. GRAVY, isoelectric point, and molecular weight parameters were relevant to differentiate classes in each protein and to enable, for the first time and with high fidelity, the prediction of both organism class and protein type just using machine learning approaches. OSM sequences from primates showed an increased BC loop when compared to the remaining mammals, which could influence binding to OSM receptor and tune signaling pathways. Overall, this study highlights the potential of sequence diversity analysis to understand IL-6 cytokine family evolution, showing the conservation of function-related motifs and evolution of class and protein-dependent characteristics. Our results could impact future medical treatment of disorders associated with imbalances in these cytokines.

Keywords: IL-6 cytokine family; leukemia inhibitory factor; machine learning; protein evolution; sequence diversity analysis.

FIGURE 1

Variability and variety of LIF, m17, OSM, and CT‐1 proteins. (a) Maximum‐likelihood cladogram showing clade distribution of LIF, m17, OSM, and CT‐1 proteins in different species. For terminal label information, please see supplementary information for a complete, full resolution cladogram. The cladogram was bootstrapped 1,000 times and nodes with a bootstrap value under 70 were disregarded. (b) Percentage identity matrix (PIM) of the MUSCLE alignment for all sequences and categorized in terms of total sequence; alignment of ±20 amino acids (aa) from FXXK sequence of human LIF; alignment of ±3 aa from FXXK sequence of human LIF; and alignment of ±3 aa from 121 to 127 sequence of human LIF of putative site II. Colour codes from red (lower) to green (higher) denote the similarity between sequences. *–LIF sequences misplaced near OSM sequences

FIGURE 2

Maximum‐likelihood cladograms with artificial classification of organisms for Leukemia inhibitory factor (a) and Oncostatin‐M (b). Terminal labels show species name (see supplementary information for readable terminal labels). The cladogram was bootstrapped 1,000 times and nodes with a bootstrap value under 70 were deleted. The inner label corresponds to organism class and the outer label to the (order, clade, family) classification of animals

FIGURE 3

Maximum‐likelihood cladograms with artificial classification of organisms for m17 (a) and Cardiotrophin‐1 (b). Terminal labels show species name (see supplementary information for readable terminal labels). The cladogram was bootstrapped 1,000 times and nodes with a bootstrap value under 70 were deleted. The inner label corresponds to organism class and the outer label to the (order, clade, family) classification of animals

FIGURE 4

Protein parameters and influence on classification. FreeViz plots of multivariate projections using the protein characteristics molecular weight (KDa, MW), average pI, charge at pH 7.4 and grand average of hydropathy (GRAVY) on sequences from LIF (a), OSM (b), CT‐1 (c), and all obtained sequences divided by organism class (d) and by protein family (e). The direction and size of each vector indicates the relative prevalence of that feature to explain group stratification. Background color is an intensity gradient related to cluster positioning and size

FIGURE 5

Modelling analysis for protein classification. (a) Confusion Matrix indicating the protein family classification of the 806 sequences predicted via Random Forest model (AUC = 0.990); (b) Confusion Matrix indicating the organism class of the 806 protein sequences predicted via Random Forest model (AUC = 0.961); Scatter plots of protein family (c) and organism class (d) classifications relatively to most important characteristics. Filled circles: correct classifications; non‐filled circles: incorrect classifications

FIGURE 6

Residue conservation in LIF (a), OSM (b), m17 (c) and CT‐1 (d) after analysis by the ConSurf server. 3D model structures of human LIF (PDB reference 1EMR), human OSM (PDB reference 1EVS), CT‐1 (SWISS‐model Q16619), and m17 (HHpred structure from the Consurf platform using MODELLER) were used as base models. The conservation score has a color code and numbered score (0–9). Site III of binding to LIFR is highlighted in blue

FIGURE 7

Residue conservation in LIF (a), OSM (b), m17 (c) and CT‐1 (d) associated with gp130 binding. 3D model structures of human LIF (PDB reference 1EMR), human OSM (PDB reference 1EVS), CT‐1 (SWISS‐model Q16619), and m17 (HHpred structure from the Consurf platform using MODELLER) were used as base models. The conservation score has a color code and numbered score (0–9)

FIGURE 8

Sequence differences in OSM from primates when compared to the remaining mammals. (a) 3D structure of hOSM (PDB reference 1EVS) highlighting the FXXK motif (green) of site III and the region near the BC only present in primates (red); (b) sequence alignment of all OSM mammal sequences denoting the extra region near the BC loop (dark blue arrow) found for primates and their placement in terms of helix distribution