Sequence similarity in structurally dissimilar proteins

It is a central tenet of protein evolution that the threedimensional structure of a protein family is better conserved than the sequences themselves. So structural similarities between related proteins should be detectable over longer evolutionary distances and be most useful for the functional prediction of proteins without closely related homologs. Comparative sequence and structural analysis of prokaryotic phospholipase A2 (PLA2) from Streptomyces violaceoruber reveals a clear violation of this dogma. While the enzymatic properties of this enzyme match closely those of eukaryotic secretory PLA2s, the corresponding structures appear very different, and no appreciable structural similarity is detected by a number of comparison methods. Nevertheless, contemporary sequence analysis methods demonstrate a highly significant relationship between the sequences of both families and provide a structurally correct alignment of the active site residues. The recent literature documents many cases of unexpected structural conservation in the complete absence of detectable sequence relatedness. Although in many of those cases the purported lack of sequence similarity does not stand closer scrutiny with sophisticated sequence comparison methods, there clearly are some impressive examples of structure-based functional predictions where sequence comparisons fail [1]. Here, we describe an unusual case where this situation appears to be reversed: related proteins with an unambiguously detectable sequence similarity but whose structures have diverged beyond recognition. Secretory (s)PLA2s are a class of small enzymes that hydrolyze the 2-acyl ester bond of 1,2diacylglycero-3-phospholipids [2]. Most eukaryotic cells produce multiple sPLA2 isoenzymes, and several poisonous animals, including bees and snakes, use members of this enzyme class as a major component of their toxins. All secreted PLA2 forms contain Ca(II) ions, which are directly involved in the catalytic reaction [2]. Recently, the first prokaryotic PLA2 has been isolated and cloned from the bacterium S. violaceoruber, and subsequently was structurally characterized [3,4]. The enzymatic properties of bacterial PLA2 resemble closely that of the eukaryotic sPLA2 forms, including the strict requirement for Ca(II). But the lack of visible sequence similarity and the fundamentally different structural fold have been interpreted as indicating a distinct evolutionary origin [3]. As indicated in the original report on cloning Streptomyces PLA2, sequence database searches with standard methods like BLAST [5] reveal significant similarities to a number of uncharacterized proteins from other bacteria, but fail to show a relationship to established PLA2 forms or to other hydrolytic enzymes. To search for more distant sequence relatives, we constructed generalized profiles [6] from a multiple alignment of the bacterial PLA2 and its reliable BLAST matches (Figure 1, upper part). Unexpectedly, the result of the profile search clearly demonstrates a significant relationship to a number of established PLA2s: The best match (p<0.01) was the conodipine-M α chain from cone snails, a well-characterized PLA2 toxin [7]. Among the next seven high-scoring sequences, six corresponded to known eukaryotic PLA2s, including the group XIV enzyme from Drosophila and the mammalian group XIII enzymes. The seventh sequence was an uncharacterized Pseudomonas protein, another likely PLA2. The best-scoring non-PLA2 was the MAP kinase ERK4, which reached an insignificant p value of only 0.6. As expected for profile searches, the significance values of other eukaryotic PLA2 sequences further improved after incorporating conodipine-M in a subsequent cycle of iterative profile refinement. As shown in Figure 1, two classes of residues are nearly invariant between prokaryotic and eukaryotic PLA2 isoenzymes: two cysteine residues that form a structurally important disulfide bridge, and the polar residues required for catalysis and the coordination of one Ca(II) ion. The eukaryotic PLA2 enzymes typically Figure 1. Alignment of prokaryotic and eukaryotic sPLA2 families. Multiple alignment of the prokaryotic PLA2 family and representative members of the eukaryotic sPLA2 families for comparison. Sequence names are in the leftmost column: PLA2 from S. violaceoruber and S. coelicolor; SCO01048 from S. coelicolor; Pst1 from Tuber borchii; RHSA from S. clavuligerus; p15 from Helicosporium sp; CGL2546 from Corynebacterium glutamicum; ORF from Neurospora crassa; and PLA2s from cone snail, cobra, bee, and various human isoenzymes. Only the most conserved part of the catalytic domain is shown. The position of the fragment relative to the complete sequence is indicated by numbers. Residues invariant or conservatively replaced in at least 50% of the sequences are printed on black or grey background, respectively. The catalytically important residues conserved between all PLA2 subtypes are printed on red background. Strep viol 64 AYEFDWSTDLCTQAPDNP.........FGFPFNTACARHDFGYRNYKA.........AGSFDANKSRIDSAFYEDMKRVCTGYT Strep 1048 63 AYGFDWTTDYCSSSPDNP.........FGFPFNTSCARHDFGYRNYKD.........AGTFSANKSRLDSAFYEDLKRVCAGYG Tuber borc 118 PGNLDWSDDGCSKSPDRP.........AGFNFLDSCKRHDFGYRNYKK........QHRFTEANRKRIDDNFKKDLYNECAKYS Strep RHSA 600 RGDLVWSDDGCSAPWYSHIVIGPSVGYYSGQFYWPCARHDFGYRNYRK........QNRRTRANKDKIDNRFRYDMKKRICEPK Helicospor 59 PSTLDWSSDGCSSSPDDP.........FGFDFLSSCHRHDFGYRNYKK........QNRFTAPNKARIDTNFKTDMYNQCNTES Corynebact 447 DPDAYGRHDYCTLSPDSY.....GPLGKKAEFSGACARHDLCMDAVDA............NGTGYAPCHPAFYTWMSTVCTTNY Neurospora 65 PATLDWSSDSCSYSPDNP.........LGFPFSPACNRHDFGYRNYKA........QSRFTDNNKLKIDGNFKTDLYYQCDTHG