1. Introduction

Chitinases (EC 3.2.1.14) are classified as glycosyl hydrolases and catalyze the degradation of chitin, an insoluble linear b-1,4-linked polymer of N-acetyl-D-glucosamine (GlcNAc). Chitin is a major component of the exoskeleton of insects, of crustacean shells and of the cell wall of many fungi. According to the glycosyl hydrolase classification system that is based on amino acid sequence similarity of the catalytic domains, chitinases have been placed in families 18 and 19 (Henrissat, 1991). Family 18 chitinases are found in bacteria, fungi, yeast, viruses, plants and animals whereas family 19 members are almost exclusively present in plants. A single family 19 chitinase was identified in Streptomyces griseus (Ohno et al., 1996; Watanabe et al., 1999). Chitinases of both families do not share sequence similarity and have a different 3D-structure, suggesting that they have arisen from a different ancestor (Hamel et al., 1997). They also differ in several of their biochemical properties. For instance, family 18 chitinases use a retention mechanism, keeping the catalysis product in the same configuration as the substrate (i.e. b-anomeric form) whereas family 19 members use an inversion mechanism turning the product into the a-anomeric form (Brameld and Goddard, 1998; Iseli et al., 1996). In addition, family 18 members hydrolyze GlcNAc-GlcNAc or GlcNAc-GlcN linkages whereas family 19 chitinases do so with GlcNAc-GlcNAc or GlcN-GlcNAc linkages (Ohno et al., 1996). Finally, family 18 chitinases are likely to function according to a substrate-assisted catalysis model (Brameld et al., 1998), whereas family 19 chitinases probably use a general acid-and-base mechanism (Garcia-Casado et al., 1998; Hart et al., 1995).

In all plants analyzed to date, chitinases of both families are present (Graham and Sticklen, 1994). They are organized in five different classes numbered from I to V, according to their sequences and structure (Neuhaus et al., 1996) and chitinases from classes I, II and IV belong to the family 19 whereas classes III and V chitinases are made of family 18 chitinases. Chitinases are often considered as pathogenesis-related (PR) proteins, since their activity can be induced by fungal, bacterial and viral infections, but also by more general sources of stress such as wounding, salicylic acid, ethylene, auxins and cytokinins, heavy metal salts or elicitors such as fungal and plant cell wall components (reviewed in Graham and Sticklen, 1994). Plants do not contain chitin in their cell walls, whereas major agricultural pests such as most fungi (i.e. Ascomycetes, Basidiomycetes and Deuteromycetes; Collinge et al., 1993) and insects do, leading to the obvious and often quoted hypothesis that chitinases act as a defense mechanism against pathogens. Evidence has been reported that chitinases can indeed degrade fungal cell walls and inhibit fungal growth in vitro, especially in combination with b-1,3-glucanases (Arlorio et al., 1992; Mauch et al., 1988; Schlumbaum et al., 1986). The expression of a number of chitinase genes appeared to be induced upon fungal infection (Majeau et al., 1990; Roby et al., 1990) and they were shown to accumulate around hyphal walls at infection sites in planta (Wubben et al., 1992). Several transgenic studies showed that by increasing the expression level of some chitinases the susceptibility of transformed plants to certain pathogens was significantly reduced (Broglie et al., 1991; Jach et al., 1995), providing an excellent tool for improving pest control. However, other studies were less conclusive. A 120-fold increase in expression of a tobacco class I chitinase did not result in any change in resistance to fungal infection (Neuhaus et al., 1991a). Similarly, down-regulation of the Arabidopsis ATHCHIA class III chitinase by antisense suppression did not increase susceptibility to fungi either (Samac et al., 1994). Therefore it remains an open question whether the primary role of chitinases is plant defense or whether they have other functions.

There are several reports of developmentally-regulated chitinase expression, with specific isoforms being present only in certain organs and at specific stages, e.g. in flowers from tobacco (Neale et al., 1990; Trudel and Asselin, 1989), Arabidopsis (class IV AtEP3/AtchitIV; Passarinho et al. 2001 and class III ATHCHIA; Samac et al., 1990), potato (SK2; Ficker et al., 1997), parsley (class II PcCHI1; Ponath et al., 2000) or rice (class I OsChia1; Takakura et al., 2000); in ripening banana fruit (Clendennen and May, 1997) or grape berries (class IV, VvChi4; Robinson et al., 1997); in roots from rice (class I RC24; Xu et al., 1996) or Sesbania rostrata (class III Srchi13; Goormachtig et al., 1998); in seeds of barley (class III Chi26; Leah et al., 1994), carrot (class IV EP3; van Hengel et al., 1998), pea (Chn; Petruzzelli et al., 1999), soybean (classIII; Yeboah et al., 1998) or in embryogenic cultures of carrot (class IV EP3; van Hengel et al., 1998), chicories (Helleboid et al., 2000), pine tree (Domon et al., 2000), spruce (Dong and Dunstan, 1997; Egertsdotter, 1996). The specificity of expression of some chitinase genes suggests that they could also play a role in developmental processes such as pollination, senescence, root and root nodule development, seed germination and somatic embryogenesis. It was shown that chitinases could rescue the carrot somatic embryo mutant ts11 (Baldan et al., 1997; de Jong et al., 1992; de Jong et al., 1993; Kragh et al., 1996) and could therefore play a crucial role in somatic embryo development. The study of Patil and Widholm (1997) also suggested the active participation of chitinases in development by over-expression of the maize Ch2 chitinase in tobacco that resulted in taller and stronger plants. Furthermore, the role of plant chitinases in Nod factor degradation during the formation of root nodules in the Rhizobium-legume symbiosis was shown in pea (Ovtsyna et al., 2000). Chitinase-mediated Nod factor degradation was already hypothesized several times and is especially interesting in line with the work of de Jong et al. (1993) showing that Nod factor-like molecules may exist in plants since rhizobial nodulation factors are also able to rescue the same carrot embryo mutant ts11.

In conclusion, chitinases are probably involved in a broad range of processes ranging from plant defense to development and there might be different functions associated with the different types of chitinases (reviewed in Graham and Sticklen, 1994). So far, attention has been mainly focused on agronomically important crops based on the preconceived idea that the natural role of plant chitinases is indeed in defense against pathogens. Very few studies were carried out in Arabidopsis thaliana and dealt with three different chitinases only (de A. Gerhardt et al., 1997; Passarinho et al., 2001; Samac et al., 1990; Verburg and Huynh, 1991). We have performed a survey of all putative chitinase genes in Arabidopsis and present here a detailed overview of their characteristics in relation with other plant chitinases. Based on these characteristics we discuss some of their possible functions and propose a modified annotation for some of the sequences, since in the release of the complete Arabidopsis genome sequence (The Arabidopsis Genome Initiative, 2000), most chitinases were annotated as “pathogen-induced or defense-related proteins”. In another database plant chitinases are annotated as being involved in the “biogenesis of cell wall”, based on homology with yeast chitinases. Moreover the AtEP3 endochitinase (Passarinho et al., 2001) is classified as a protein involved in “cell rescue, defense, cell death and ageing – biogenesis of cell wall”; for sure a highly versatile protein.

2. Arabidopsis chitinase genes and their genomic distribution.

Using the word chitinase, we performed a keyword-based search on several Arabidopsis annotation databases (MATDB (MIPS (Munich Information Center for Protein Sequences) Arabidopsis thaliana DataBase); Mewes et al., 2000;  http://mips.gsf.de/proj/thal/db/index.html), TIGR (The Institute for Genomic Research;  http://www.tigr.org/tdb/e2k1/ath1/ath1.shtml) and DAtA (Database of Arabidopsis thaliana Annotation;  http://luggagefast.stanford.edu/group/arabprotein/index.html). Each search gave a slightly different result, mostly due to differences in clone names and annotations. We compared all returned accessions for redundancy and finally came to a total of 24 DNA sequences that, based on their annotation, encode putative chitinases (Table 1). The corresponding loci are distributed on all five chromosomes of the Arabidopsis genome (Figure 1), with a remarkable degree of clustering at the bottom of chromosome II where 6 putative genes are organized in tandem and in the middle of chromosome IV where 9 genes are organized in two clusters with 2 unrelated genes in between (Figure 1). It has now become obvious from several studies (Blanc et al., 2000; Vision et al., 2000) that the Arabidopsis genome contains large segmental duplications, suggesting that Arabidopsis could have originated from an ancient tetraploid ancestor (Blanc et al., 2000). It is likely that some of the duplicated genes have acquired a certain degree of specialization and are now expressed in different conditions. As found during systematic gene knockout in yeast (Ross-MacDonald et al., 1999), many insertion mutants in Arabidopsis do not show an obvious phenotype (Bouche and Bouchez, 2001; Pereira, 2000). This can be the result of gene redundancy or may point to a failure to detect subtle phenotypes perhaps only seen at the level of genome-wide gene expression as found in yeast (Beh et al., 2001).

Expressed Sequenced Tags (ESTs) were found for 16 of these sequences (Table 1) indicating that the corresponding genes are transcribed and most likely encode a functional protein, whereas the others are putative genes. This must be taken into consideration when drawing conclusions from their sequence, since they may be pseudogenes or are only expressed in conditions that were not studied in the various EST projects (Blanc et al., 2000).

3. Classification and structure of the Arabidopsis chitinase sequences.

The deduced amino acid sequences of all 24 accessions revealed that they all have a length of around 300 amino acids and a molecular weight of 25–35 kDa, which is typical for chitinases in general (Graham and Sticklen, 1994). The predicted proteins they encode belong to different groups according to the classification proposed for plant chitinases (Neuhaus et al., 1996). Based on their amino acid sequence all plant chitinases are endochitinases (EC 3.2.1.14) and have been organized in five different classes (Figure 2). Class I chitinases have a highly conserved N-terminal cysteine-rich region of approximately 40 amino acid residues that is involved in chitin-binding (Iseli et al., 1993). It is separated from the catalytic domain by a short proline-rich variable hinge region and the catalytic domain is often followed by a C-terminal extension that is involved in vacuolar targeting (Class Ia; Neuhaus et al., 1991b).

Class II chitinases lack both the N-terminal cystein-rich region and the C-terminal extension, but have a catalytic domain with a high sequence and structural similarity to that of class I chitinases. Class IV chitinases resemble class I chitinases with a very similar main structure, but they are significantly smaller due to four deletions distributed along the chitin-binding domain and the catalytic region. Class III chitinasesare more similar to fungal and bacterial chitinases than to other plant chitinases (Graham and Sticklen, 1994), except for class V chitinases, that also belong to the family 18 of glycosyl hydrolases whereas all other classes belong to family 19. In addition, class V chitinases have a C-terminal extension for vacuolar targeting and may contain a chitin-binding domain as well (Heitz et al., 1994; Ponstein et al., 1994). Finally, cass III and class V chitinases display an additional lysozymal activity (Heitz et al., 1994; Majeau et al., 1990).

As in all plants analyzed to date (Graham and Sticklen, 1994), members of all five classes are present in the Arabidopsis genome. It is also remarkable that classes I and III are poorly represented with only one member each (Figure 2), whereas the other classes are more abundant, especially classes IV and V with 9 members each. It is also noteworthy that the class I chitinase contains a C-terminal extension, hence belongs to subclass Ia, and none of he class V members possesses a chitin-binding domain.

Figure 3 shows the phylogenetic tree generated with the 24 sequences by using the CLUSTALW Multiple Sequence Alignment program at the GenomeNet WWW server ( http://clustalw.genome.ad.jp/). The different classes are nicely clustered and it is clear that class V has diverged from the other classes very early during evolution. It also seems that the very similar classes I and IV may have arisen from class II in which they are imbedded. Araki and Torikata (1995) have indeed suggested that class I chitinases arose from class II chitinases by insertion of the chitin-binding domain. This probably occurred in the case of class IV chitinases as well, considering their degree of similarity with class I members, including the presence of the chitin-binding domain.

4. Sequence characteristics of the Arabidopsis chitinases.

Based on the classes obtained from the phylogenetic tree, the deduced amino acid sequences of all chitinase genes were compared to each other by multiple sequence alignment and the presence of elements essential for chitinase activity was analyzed for each sequence.

Figure 4 shows the sequences of class I and class III chitinases, both of which represent actual genes that were isolated by Samac et al. (1990). The class I chitinase sequence contains all characteristics of class I chitinases including the C-terminal extension, specific of subclass Ia, indicating that it is targeted to the vacuole. All residues shown to be involved in substrate binding and catalytic activity are also present (Garcia-Casado et al., 1998) and indicate that it is most likely an active chitinase and one of that is actively transcribed (Samac et al., 1990). The same holds true for the class III chitinase, of which the catalytic domain possesses all essential residues known to date (Watanabe et al., 1993).

Figure 5 shows the multiple alignment of the class II chitinase sequences and one can see that they share a relatively high degree of similarity, especially in the catalytic domain. However it also appears that two of these sequences do not possess all conserved residues essential for chitinase activity. As a matter of fact, only the sequences of the two underlined accessions fulfill all requirements described by Garcia-Casado et al. (1998). For example, the H-E-T-T motif including the essential glutamic acid residue shown in bold is absent from the two other sequences. The same holds true for the first cysteine in the Chitinase 19_1 conserved domain as well as for most of the residues in bold that are essential for catalytic activity and the boxed residues involved in substrate binding. Nevertheless these residues were only shown to play a specific role in a class I chitinase (Garcia-Casado et al., 1998) and there are no reports so far of a similar study with class II chitinases. Therefore it could still be that especially the residues involved in substrate binding (boxed) are different in this class. We can eliminate the last 2 sequences (At1g05870 and At3g16920) as non-active chitinases based on the absence of the H-E-T-T motif and of some of the other residues essential for catalytic activity. Furthermore, At1g05870 and At3g16920 were also put together at the bottom of the phylogenetic tree (Figure 3) indicating that although they are similar to each other they also diverge considerably from the other class II members. Interestingly the sequences At1g02360 and At4g01700 considered as encoding active chitinases are also paired in the dendrogram shown in Figure 3 and are located on chromosomal regions that were shown to be duplicated (i.e the top of chromosome I and the top of chromosome IV; Blanc et al., 2000) and are therefore likely to represent a duplication of the same gene.

Figure 6 shows the same comparison for class IV chitinases to which the only other Arabidopsis chitinase studied, AtEP3/AtChitIV (At3g54420; de A. Gerhardt et al., 1997; Passarinho et al., 2001) belongs. In this class the degree of conservation is very high and all elements specific for class IV chitinases are present, except for accession At3g47540 that lacks the chitin-binding domain as well as the accompanying hinge region. Nevertheless it was put in class IV, since its shorter catalytic domain is more closely related to that of this class than to that of class II chitinases. It is also shorter than the other class IV chitinase genes in the second half of the catalytic domain where it also lacks some of the important amino acid residues (i.e. glutamate-170 and serine-172, as seen in the At2g43590 sequence). Furthermore, there was no EST found for At3g47540, so it could very well be that it represents a pseudogene. There were three other sequences for which no EST was found (marked by the asterisk) and those also appear to lack some essential amino acids in the second half of the catalytic domain, especially At2g43600 that lacks the essential glutamic acid residue at position 140 and is therefore probably not active as a chitinase. It is also remarkable that in this class some of the residues shown to be involved in substrate binding in class I chitinases are here consistently different (Garcia-Casado et al., 1998). For example the H-E-T-T motif seems to be replaced by H-E-[TS]-G, and the tryptophan residue that should have been at position 153 (see the At2g43590 sequence) is replaced by a tyrosine. The same holds true for the glutamine-212 and the lysine-214 of the same sequence that are replaced by a valine. These differences most likely reflect a class-related difference in substrate specificity, which is also illustrated by the tyrosine (shown by the arrow) that was shown to be essential for substrate binding, but not for catalysis in the class I chitinase (Verburg et al., 1993) and is replaced by a phenylalanine, especially in sequence At3g54420 (i.e. AtEP3/AtChitIV), of which we know that it is an active chitinase (Passarinho et al., 2001). As for class II chitinases, based on the missing essential amino acid residues and the failure to find ESTs we can conclude that the accessions At1g56680, At2g43580, At2g43600 and At3g47540 are not very likely to encode active chitinases. It is also noteworthy that the majority of class IV chitinases is clustered at the bottom of chromosome II and is also found on the lower arm of chromosome III (Figure 1) that also seems to be an area duplicated on chromosome II (Blanc et al., 2000).

Figure 7 presents the multiple alignment of class V chitinases. The chitinases of this class are longer than the members of the other classes. They also seem to possess additional motifs, which were not found in other classes and of which we do not know the functional relevance. Little is known about class V chitinases and we can therefore only base our analysis on what is known for the glycosyl hydrolase family 18 (Watanabe et al., 1993), of which the conserved characteristic motif represents a small segment of the whole protein. In this small conserved region we can already see that two members of this class (At4g19720 and At4g19820) deviate from the others since a lysine residue (arrow) replaces the proposed essential glutamic acid. This resembles the situation of concanavalin B present in seeds of Canavalia ensiformis (Hennig et al., 1995), where the glutamic acid residue is replaced by a glutamine. As a consequence, concanavalin B, a close relative of family 18 chitinases, lost its enzymatic activity, but retained its carbohydrate-binding function (Hennig et al., 1995).

Concanavilin B is biochemically and structurally similar to narbonin that is a storage protein found in seeds of Vicia narbonensis (Hennig et al., 1992; Nong et al., 1995) and could be involved in “trapping” carbohydrate molecules necessary for the seed. A similar function could be proposed here for At4g19820 and At4g19720. The other sequences, including those for which no EST was found, all have an intact catalytic site and should therefore be active class V chitinases. As seen for class IV chitinases they are also clustered on a particular chromosomal location, on the lower arm of chromosome IV (Figure 1), but this region does not seem to have been duplicated elsewhere in the genome.

5. Putative function and reannotation of the Arabidopsis chitinase sequences.

In order to obtain additional clues with respect to the putative function of all chitinases, each sequence was also analyzed for the presence of additional specific motifs by using the InterPro domain search ( http://www.ebi.ac.uk/interpro/; Apweiler et al., 2001) and for the presence of targeting sequences using the PSORT ( http://psort.nibb.ac.jp/) and targetP ( http://www.cbs.dtu.dk/services/TargetP/; Emanuelsson et al., 2000) servers. A PSI-BLAST search ( http://www.ncbi.nlm.nih.gov/BLAST/; Altschul et al., 1997) was also performed in order to obtain more functional data on similar chitinases. The results of this analysis are detailed in Table 2.

6. Conclusions.

Sequencing and systematic automated annotation of the Arabidopsis genome has led to the classification of 24 sequences as putative chitinase-encoding genes. A more detailed analysis of the individual sequences reveals one of the limitations of large-scale automated genome annotation. Sequence details that are functionally important can be missed because at present it is difficult to incorporate an integrated view of all data available on protein families into the annotation software. Indeed, out of the 24 chitinase sequences, 8 are not likely to be transcribed while 3 others do not contain amino acid residues that are essential for catalytic activity. Consequently, they probably have a function different from the hydrolysis of chitin-derived molecules. This is also true for most of the sequences for which no ESTs were found.

The genomic distribution of the chitinase-encoding genes shows a remarkable degree of clustering per class (class IV on chromosome II and class V on chromosome IV; Figure 8). Similar genes are indeed repeated in tandem but also duplicated on other chromosomal regions like At1g02360 and At3g16920. This reflects one of the characteristics of the Arabidopsis genome, that is largely made up of duplicated chromosomal regions (Blanc et al., 2000; Vision et al., 2000). Chitinase genes belong to relatively large families (Graham and Sticklen, 1994) that are probably the result of such duplication events.

Chitinases are grouped into five different classes that differ in sequence, 3D structure and biochemical properties (Neuhaus et al., 1996). In Arabidopsis, as in all other plants studied so far, chitinases of each class are present. These are rather equally represented, if one removes all sequences that are most likely not transcribed (Figure 8), and it is reasonable to assume that they have developed class-specific functions, especially between chitinases of family 18 and 19. Furthermore, the analysis we performed here reveals that there are also differences between related classes such as class I and class IV as well as within classes, like in classes II and V. This is probably indicative of different substrate specificities and thereby suggest a rather high degree of specialization. It is also clear that most chitinases, independently from their class, are probably involved in several functions.

Some chitinases (e.g. Arabidopsis classes I and III (Samac et al., 1991; Verburg and Huynh, 1991) and some isoforms of class II, e.g. in parsley (Ponath et al., 2000) and peanut (Kellmann et al., 1996)) are only activated upon infection with specific strains of fungi, implying a role in a highly specialized defense response. Others (e.g. bean class IV (Margis-Pinheiro et al., 1993) and some isoforms of class II, e.g. in parsley (Ponath et al., 2000) and peanut (Kellmann et al., 1996)) seem to be involved in more general stress responses that do not require a very specific interaction with a pathogen. Furthermore, their range of action in response to pathogen infection also seems to be different. Classes III and V chitinases that belong to the glycosyl hydrolase family 18, seem to be involved in a short-range response that suggests a direct action on the invading pathogen.

The Arabidopsis class III chitinase ATHCHIA that is induced by very specific strains of pathogens and does not seem to require any other form of signaling (e.g. ethylene) for activation, is a typical example. This is supported by its activation directly at the infection site (Samac and Shah, 1991). Furthermore, the inactive chitinases of the concanavalin B-type found in class V suggest a putative role in the perception and recruitment of chitin-derived molecules (Hennig et al., 1995; Kim et al., 2000). This may strengthen the idea of a direct interaction with the invading pathogen. And last, the additional lyzosymal activity that is characteristic of these two classes combined with the putative localization of some isoforms in the peroxisomes could also indicate an activity involved in direct degradation of the pathogen. Genes of the other classes are more likely to be activated indirectly via a signaling cascade triggered upon identification of a specific pathogen by, for example, a class V chitinase of the concanavalin B-type. This is probably the case for the Arabidopsis class I chitinase ATHCHIB and for some specific isoforms of class II (Kellmann et al., 1996; Ponath et al., 2000). Other isoforms of class II as well as class IV chitinases are probably activated by more general forms of stress that eventually may lead to the same general response. Plant hormones, such as ethylene, may be the mediators of these signaling events.

The role ethylene plays in development also brings us to the developmental regulation of chitinase genes. This seems to be valid for all classes and their exact function at this level is probably determined by the part of the plant in which they are localized and on the available substrates. These substrates can be of a symbiotic origin (rhizobial Nod factors) that upon perception and processing by chitinases are able to trigger a cascade of specific events leading to the formation of a root nodule (Ovtsyna et al., 2000). Alternatively, substrates must be of plant origin, implying the existence of plant endogenous GlcNAc-containing molecules. Recent work has demonstrated that these molecules could be AGPs (van Hengel et al., 2001). This is in line with the large distribution of AGPs in different plant tissues (Knox, 1999) and their great plasticity in carbohydrate composition. Thus, GlcNAc- or GlcN-containing AGPs could exist in many plant organs and provide highly specific substrates to matching specific chitinases.

In conclusion, it is clear that the function of plant chitinases is still poorly understood. Chitinases seem to be involved in many different aspects of the plant life cycle, and it will be difficult to dissect such aspects in great detail. Understanding the role of plant chitinases will require the generation of mutant plants that lack one or several specific chitinases, to create a background with different combinations of chitinases and circumvent problems of gene redundancy but also to understand the specific interrelations between the different classes. It will also imply the combined study of the role of AGPs following similar approaches and most certainly detailed immunocytological and biochemical studies to unravel the complex chitinase-AGP combinations in association with very specific processes.