Major Facilitator Superfamily (MFS) evolved without 3-transmembrane segment unit rearrangements.

Based on alleged functional residue correspondences, a recent study proposed a model of 3-transmembrane segment (TMS) repeat unit rearrangements in Major Facilitator Superfamily (MFS) carriers (1). A rebuttal of “Evolutionary mix-and-match with MFS transporters” (1) is currently in press in the Journal of Molecular Microbiology and Biotechnology (2). In their follow-up paper, “Evolutionary mix-and-match with MFS transporters II” (3), Madej and Kaback extend their postulate (1) to xylose (XylE), peptide (PepT) (4), and phosphate (PiPT) (5) porters by suggesting additional 3-TMS unit rearrangements (3). In this letter, we suggest that these transporters are all homologous throughout their lengths, having evolved from a common ancestor without 3-TMS unit rearrangements. First, MFS transporters are well known to contain conserved motifs in cytoplasmic loops (CLs), CL23 and CL89, conforming to the sequence D[RK]-X2–4-[RK]2, where X = any residue (2). In CL23 and CL89, the same motifs could be found in the expected positions corresponding to [DN][RK][IL]GLK [KN] (LacY) and DK[ILV]GR[KF]K (PiPT). In L-fucose permease (FucP) and PepT, CL23 contained [DK]K-X3-KR and DR-X4-RR, respectively. In XylE, CL89 contained DR-X2RR. Thus, 7 of 10 CL23 and CL89 loops in these five proteins contained this motif, an observation expected only in the absence of rearrangements (2). Second, using MAFFT 7.023 (E-INS-i) to align members of the lactose permease (LacY), FucP, XylE, PepT, and PiPT families (see the Transporter Classification Database, TCDB; www.tcdb.org), and using a Perl script to visualize TMHMM predictions in Jalview, we extracted the 3-TMS units of each family and trained hidden Markov models (HMMs) on each triplet using HHmake 2.0.15. For each family comparison, we recorded the average probability of homology using HH-suite 2.0 according to both the mix-and-match (M&M) hypotheses (1, 3) and the null hypothesis (no rearrangements) (2). The null hypothesis scored higher (Table 1). Third, the sequences of the five transporters were subdivided into their 3-TMS units, and the average percent identity for each comparison was derived according to the two models. In 8 of 10 comparisons, the null hypothesis gave higher scores than the M&M model (Table 1). Fourth, we calculated the number of conserved basic residues (R and K) in each tail and loop of the alignments for the five families. All clearly followed the “positive inside rule.” If the 3-TMS units had rearranged, the loops would assume opposite orientation in the membrane. Fifth, there is a shared and characteristic “kink” in the interdomain linker, just before TMS7 in LacY, FucP, and PiPT. This common feature would not be expected from the M&M model. Finally, although Madej and Kaback (3) show superimpositions of allegedly corresponding triplets, they do not systematically compare structures in the same conformation. By superimposing structures available in the same conformational state, four of five comparisons favored the null model. These observations lead to the suggestion that functionally equivalent residues involved in dissimilar substrate binding and catalysis have evolved independently from a single structural scaffold without repeat unit rearrangements.