Phylogenetic Characterization of Transport Protein Superfamilies: Superiority of SuperfamilyTree Programs over Those Based on Multiple Alignments

Abstract

Transport proteins function in the translocation of ions, solutes and macromolecules across cellular and organellar membranes. These integral membrane proteins fall into >600 families as tabulated in the Transporter Classification Database (www.tcdb.org). Recent studies, some of which are reported here, define distant phylogenetic relationships between families with the creation of superfamilies. Several of these are analyzed using a novel set of programs designed to allow reliable prediction of phylogenetic trees when sequence divergence is too great to allow the use of multiple alignments. These new programs, called SuperfamilyTree1 and 2 (SFT1 and 2), allow display of protein and family relationships, respectively, based on thousands of comparative BLAST scores rather than multiple alignments. Superfamilies analyzed include: (1) Aerolysins, (2) RTX Toxins, (3) Defensins, (4) Ion Transporters, (5) Bile/Arsenite/Riboflavin Transporters, (6) Cation:Proton Antiporters, and (7) the Glucose/Fructose/Lactose superfamily within the prokaryotic phosphoenol pyruvate-dependent Phosphotransferase System. In addition to defining the phylogenetic relationships of the proteins and families within these seven superfamilies, evidence is provided showing that the SFT programs outperform programs that are based on multiple alignments whenever sequence divergence of superfamily members is extensive. The SFT programs should be applicable to virtually any superfamily of proteins or nucleic acids.

The Amino Acid-Polyamine-Organocation Superfamily

“…2 a). By contrast, our newly developed SFT programs [Chen et al, 2011;Yen et al, 2009;Yen et al, 2010] proved reliable, with members of each family clustering coherently together. These SFT programs base phylogenetic position on tens of thousands of BLAST bit scores, giving much greater degrees of reliability.…”

“…The resulting database files were then used to generate a phylogenetic tree using the SFT1 program by generating tens of thousands of comparative BLAST bit score matrices of the superfamily through 100 repeat shuffles [Chen et al, 2011;Yen et al, 2009;Yen et al, 2010]. The programs, Fitch and Consense [Fitch and Margoliash, 1967; http://evolution.genetics.washington.edu/ phylip/doc/protpars.html], utilized the matrix information to generate 100 phylogenetic trees and consolidate those trees into a single consensus tree.…”

“…However, branch lengths in trees derived using the SFT programs are not proportional to phylogenetic distance. In previous communications, the reliability of these methods has been evaluated [Chen et al, 2011;Yen et al, 2009;Yen et al, 2010].…”

“…Members of the CCC, AGCS, and HAAAP families also segregated according to family, and these occur in the upper central part of the tree, sandwiched between the APC and AAAP families. It is interesting to note that all of these families, which had previously been known [Chen et al, 2011;Yen et al, 2009;Yen et al, 2010].…”

Establishing homology between mitochondrial calcium uniporters, prokaryotic magnesium channels and chlamydial IncA proteins

“…The SFT1 program, which is based on BLAST BIT scores for large numbers (tens of thousands) of binary comparisons, has been shown in six publications (Chen et al, 2011;Wong et al, 2012;Yen et al, 2009Yen et al, , 2010 to be superior to other programs for depicting deep phylogenies within superfamilies of distantly related proteins. This is in contrast to programs that construct phylogenetic trees that are based on multiple alignments such as neighbour-joining, Bayesian integration over alignments and maximum-parsimony.…”