Life with no sugars?

Whether hydrolytic or synthetic, glycosyl transfer is, simply in terms of quantity, one of the most important biological reactions on earth. The formation and cleavage of glycosidic bonds are catalysed by glycosyltransferases and glycoside hydrolases which are crucial for a number of biological pathways (biosynthesis and degradation of structural and storage polysaccharides, cellular signalling, host-pathogen interactions, etc.). As complete genomes are becoming available, comparative genomics are emerging and allow complete metabolic pathways to be searched and evaluated (Tatusov et al., 1997). Glycoside hydrolases and glycosyltransferases have been classified in families of related proteins for several years (Henrissat, 1991; Henrissat and Bairoch, 1993; Henrissat and Bairoch, 1996; Campbell et al., 1997). These families can be conveniently accessed through the CAZY server at URL http://afmb.cnrs-mrs.fr/~pedro/CAZY/db.html. This classification system has several advantages, one of which is that one can search complete genomes for these families. We have compiled a library containing these enzymes and related proteins (>4,700 entries). Any genome can be searched for putative glycoside hydrolases and glycosyltransferases by BLASTing (Altschul et al., 1997) its putative coding regions against this library. Glycoside hydrolases and glycosyltransferases frequently display a modular structure with a catalytic domain linked to one or several non-catalytic modules (Gilkes et al., 1991). A given non-catalytic module can be found attached to catalytic modules of different families but also to proteins which are neither glycoside hydrolases nor glycosyltransferases. To avoid « false positive hits » with the non-catalytic modules, we have excised the non-catalytic modules from our sequences. We have analyzed the five complete archeal genomes publicly available: Aeropyrum pernix (Kawarabayasi et al., 1999), Archaeoglobus fulgidus (Klenk et al., 1997), Methanobacterium thermoautotrophicum (Smith et al., 1997), Methanococcus jannaschii (Bult et al., 1996) and Pyrococcus horikoshii (Kawarabayasi et al., 1998). Surprisingly, even using permissive thresholds, no similarity with any known glycoside hydrolase was found in three out of these genomes (A. fulgidus, A. pernix and M. thermoautotrophicum) while the remaining two (M. jannaschii and P. horikoshii) contained only two and seven respectively (Table 1). On the other hand, the five archae all appear to contain putative glycosyltransferases (Table 1). For comparison, the genomes of two hyperthermophilic bacteria, Aquifex aeolicus (Deckert et al., 1998) and Thermotoga maritima (Nelson et al., 1999), and of a mesophilic bacterium, Escherichia coli (Blattner et al., 1997), were also searched for glycoside hydrolases and glycosyltransferases (Table 1). While A. aeolicus appears to contain only four putative glycoside hydrolases, T. maritima and E. coli have respectively 39 and 35. There are two possibilities to explain the apparent total lack of glycoside hydrolase in three out of five archaeal genomes: (i) these organisms have glycoside hydrolases completely different (and to be discovered) from those already known, a puzzling hypothesis given the evidence of extensive horizontal transfers between archae and bacteria (Nelson et al., 1999); (ii) these organisms indeed do not require any glycoside hydrolase, perhaps in line with observations that certain sulfur-dependent heterotrophic archae do not grow on carbohydrates (Grote et al., 1999). The presence of putative glycosyltransferases in genomes apparently devoid of glycoside hydrolase constitutes another surprise as one would expect an organism to be able to degrade the glycosidic bonds it builds.