Response to Shields: Molecular evolution of CXC chemokines and receptors

Abstract

The letter by Denis Shields [1] comments on our article on the molecular evolution of CXC chemokines and their receptors [2], in which we hypothesized that the chemokines originate from the central nervous system (CNS). Shield’s comments hinged mainly on the evolutionary analyses. Clearly, inferences drawn from phylogenetic analyses do have their limitations. In the initial stages of the analyses, we constructed phylogenetic trees using an outgroup for both receptor and ligands. Although in general outgroups are chosen arbitrarily, we constructed a CXC receptor (CXCR) tree with CC chemokine receptors (CCRs) as an outgroup. The topology did not change from what was presented, except for CXCR6, which was located in the CCR cluster, close to the starting point of cluster formation but with low bootstrap values. In general, the tips of a phylogenetic tree are well resolved, whereas the deeper nodes are poorly resolved. ‘Supertrees’, which in theory could combine different types of information, and not merely DNA or protein sequence information, might be an alternative for studying long evolutionary histories [3]. We adopted the philosophy of a ‘supertree’ approach by substantiating our hypothesis using not only phylogenetic analyses but also other information, such as protein function, several CXC ligands (CXCL) sharing the same receptor, the presence of the ELR (glutamic acid, leucine, arginine) motif, and chromosomal localizations. Shields also argues that the absence of CXCR4 in Ciona favours a primordial immunological function of the CXC family. However, in this respect, the presence of CXCR4 in jawless fish is more important. Kuroda et al. [4] sequenced 9312 cDNA clones from a lamprey (Petromyzon marinus). The serendipitous finding of only CXCR4 could be a testimony to its ancestral status. The lamprey CXCR4 receptors cluster, with a high bootstrap value, with all other CXCR4 found in jawed fish, amphibia, birds and mammals. In addition, the architecture and sophistication of the CNS of Ciona differ markedly compared with that found in other more advanced chordates [5]. Although using a small number of amino acids to construct a phylogeny can lead to incorrect topologies, this can be overcome by the bootstrap test [6], which we used extensively to identify statistically reliable orthologues. The use of complete sequences is preferred to detect true orthologues, although when only incomplete sequences are available, the phylogenetic analyses can be performed in two ways, using pair-wise deletion or complete deletion of missing information. We performed both, and the results were essentially identical topologies with only minor deviations, notably the position of the Xenopus CXCR3 and rainbow trout CXCR. Phylogenetic trees are a hypothetical model for the evolutionary history of a gene family. A gene conversion event might result in two paralogous genes clustering closer together in trees than in reality. Gene conversions occur relatively frequently; Shields [7] observed that 45% of a control group of clustered genes showed evidence of gene conversion. However, this should not preclude phylogenetic analyses; we should merely be aware of the possibility of gene conversion imposing a bias on the analyses. The gene conversion events between the clustered CXCR1 and CXCR2 sequences analysed by Shields [7] using human, rabbit and rat sequences probably occurred between ,90 and 110 million years ago [8]. The evolutionary scale on which these gene conversions could bias the phylogenetic tree is therefore limited to the clustering of CXCR1 and CXCR2 in mammals. We believe that putative gene conversion events do not interfere with our conclusions and hypotheses derived from phylogenetic analyses on a larger, vertebrate-wide scale that comprises more CXCRs. Shields also touches upon the controversy surrounding genome duplication. Some studies favour whole-genome duplications, whereas others point to accumulated duplication of blocks or single chromosomes. Most data supporting whole genome duplications are derived from analyses of either the major histocompatibility complex (MHC) [9] or the Hox [10] gene clusters. This issue is still unresolved. Following either mode of duplication, it is apparent that extensive reshuffling and loss of genes or gene regions has occurred. This obscures the reconstruction of events occurring during the ‘birth’ of genomes [11]. Our hypotheses do not refute either the whole-genome duplication or the block duplication hypothesis. Regardless of the pitfalls associated with phylogenetic analyses, this approach has the potential to unravel the evolutionary history of the CXCL/CXCR multigene families. Our opinion, based on a combination of these analyses and additional data, is an attempt to set these gene families in a compelling evolutionary perspective. Corresponding author: B.M. Lidy Verburg-van Kemenade (lidy.vankemenade@ wur.nl). Update TRENDS in Immunology Vol.24 No.7 July 2003 356