Phylogenomics of phosphoinositide lipid kinases: perspectives on the evolution of second messenger signaling and drug discovery

Abstract

Background: Phosphoinositide lipid kinases (PIKs) generate specific phosphorylated variants of phosatidylinositols (PtdIns) that are critical for second messenger signaling and cellular membrane remodeling. Mammals have 19 PIK isoforms spread across three major families: the PtIns 3-kinases (PI3Ks), PtdIns 4-kinases (PI4Ks), and PtdIns-P (PIP) kinases (PIPKs). Other eukaryotes have fewer yet varying PIK complements. PIKs are also an important, emerging class of drug targets for many therapeutic areas including cancer, inflammatory and metabolic diseases and host-pathogen interactions. Here, we report the genomic occurrences and evolutionary relationships or phylogenomics of all three PIK families across major eukaryotic groups and suggest potential ramifications for drug discovery.

Results: Our analyses reveal four core eukaryotic PIKs which are type III PIK4A and PIK4B, and at least one homolog each from PI3K (possibly PIK3C3 as the ancestor) and PIP5K families. We also applied evolutionary analyses to PIK disease ontology and drug discovery. Mutated PIK3CA are known to be oncogenic and several inhibitors are in anti-cancer clinical trials. We found conservation of activating mutations of PIK3CA in paralogous isoforms suggesting specific functional constraints on these residues. By mapping published compound inhibition data (IC50s) onto a phylogeny of PI3Ks, type II PI4Ks and distantly related, MTOR, ATM, ATR and PRKDC kinases, we also show that compound polypharmacology corresponds to kinase evolutionary relationships. Finally, we extended the rationale for drugs targeting PIKs of malarial Plasmodium falciparum, and the parasites, Leishmania sp. and Trypanosoma sp. by identifying those PIKs highly divergent from human homologs.

Conclusion: Our phylogenomic analysis of PIKs provides new insights into the evolution of second messenger signaling. We postulate two waves of PIK diversification, the first in metazoans with a subsequent expansion in cold-blooded vertebrates that was post-emergence of Deutrostomia\Chordata but prior to the appearance of mammals. Reconstruction of the evolutionary relationships among these lipid kinases also adds to our understanding of their roles in various diseases and assists in their development as potential drug targets.

Figure 1

General pathway for phosphatidylinositide (PI) synthesis. Major PI types with phosphorylation sites labeled (3,4,5 in red) are shown along with the phophorylation and dephosphorylation reactions catalyzed by different phosphoinositide kinase (PIK) types and phosphatases, respectively. Figure partially adapted from Figure 1 of Weernink et al. [6].

Figure 2

Phylogenetic tree of phosphoinositide 3-kinases (PI3K) and Type III phosphoinositide 4-kinases (PI4K). PI3K proteins are labeled according to HUGO human gene name with classes identified by commonly accepted nomenclature [2]. Main organism taxonomic groups are color coded. For simplicity, Protosomatia (insects/arthropods), Pseudocoelomata (nematode worms) are considered collectively as Invertebrates. The tree was reconstructed by neighbor-joining (NJ) method using protein distance matrices of core conserved amino acids (see Methods). Asterisks ("*") indicate those nodes supported 70% or greater in 1000 bootstrap replicate NJ trees and 0.95 Bayesian posterior probability. Scale bar represents 0.1 expected amino acid residue substitutions per site.

Figure 3

Phylogenetic tree of Type II phosphoinositide 4-kinasess kinases (PI4K2). Nomenclature and phylogenetic reconstruction methods are as described for Figure 2.

Figure 4

Phylogenetic tree of phosphatidylinositol-4-phosphate 5-kinases (PIP5K). Nomenclature and phylogenetic reconstruction methods are as described for Figure 2. Numbers at major nodes show the percent occurrence of nodes in greater than 50% of 1000 bootstrap replicate neighbor joining trees followed by Bayesian posterior probability values. Dashed blue line shows the placement of Anopheles gambiae and Drosophila melanogaster in consensus bootstrap neighbor-joining and Bayesian phylogenetic trees which depict invertebrates as monophyletic and ancestral to the Chordata/Deutrostomia/Vertebrata clade.

Figure 5

Phyletic distribution of phosphoinositide kinases. Shown is a summary of PIK orthologs and paralogs (relative to mammals) across major eukaryotic groups. Cladogram at the top represents relative evolutionary relationships among these groups according to the Tree of Life web project [46].

Figure 6

Occurrence of missense cancer mutations in PIK3CA gene relative to orthologous and paralogous PI3K kinases. PIK3CA sequences are from human (hs_PIK3CA), mouse (mus_PIK3CA), and dog (dog_PIK3CA) and cow (cow_PIK3CA) as well as the human paralogs PIK3CB (hs_PIK3CB), PIK3CD (hs_PIK3CD), and PIK3CG (hs_PIK3CG). Also included is a composite cancer mutant human PI3KCA (hs_PI3KCAm in red) with amino acid substitutions (mutations) mapped as reported by Samuels et al. [8] and the Sanger COSMIC database [27]. Regions of the alignments are shown where a cancer missense mutation is identical to an amino acid occurring in normal human paralogs. Numbers indicate coordinates in normal human PI3KCA. Arrows at the bottom of the alignment point to those specific changes across paralogues. Note for H1047, three different amino acid substitutions have been observed and font size of label indicates the relative high (large font) to low (small font) oncogenic potency of each type [28]. Structural domains were taken from the alignment of PI3K kinases to the PI3K C-γ structure reported by Walker et al. [47] and are not drawn to scale.

Figure 7

Phylogenetic tree of PIK3/4III and related protein kinases with IC₅₀ values from tested inhibitors. Compound names in column are as reported by Knight et al. [31] (bold font) and Apsel et al. [32] (italics). Kinase assay are aligned with their branching order in the phylogenetic tree. IC₅₀ values (μM) are shaded according to potency with smaller values being more effective inhibitors of kinase activity. The tree was constructed using the neighbor-joining method as described for Figure 2 with numbers at nodes showing percent support in 1000 bootstrap replications.

Figure 8

Malarial and human PIK3/4III kinase domains. Alignment of kinase domains of PI3K and PI4KIII kinases from Homo sapiens (Hs), Plasmodium falciparum (Plfa) and P. vivax (Plvi). Plasmodium species names and residues implicated in the kinase ATP-binding pocket are highlighted in red. Domains are from the PI3K C-γ (PIK3CG) structure reported by Walker et al. [47].