In Arabidopsis, a homology-based search of EST databases yielded two CCR genes [14]. Their identity was confirmed by the catalytic properties of the corresponding recombinant enzymes. Expression analyses suggested the involvement of AtCCR1 in lignification of vascular tissue. Independently, Jones et al. [15] described the irx4 mutation, which affected AtCCR1 and resulted in collapsed xylem elements and reduced lignin levels. The role of AtCCR1 in constitutive lignification was further ascertained more recently through an antisense strategy [16].

The other gene, AtCCR2, was shown to be responsive to pathogen attack [14]. Lignins play an important role in plant defence and indeed, one of the characteristic responses of plants to injury and disease is the reinforcement of cell walls in the vicinity of the wound/infection site with lignin, as well as with suberin [17–19]. The role of AtCCR2 in stress response could be twofold: synthesis of defence lignins that help prevent the progression of bacteria, and/or synthesis of other defence-related compounds. It has already been reported that coniferaldehyde and coniferyl alcohol may function as phytoalexins in the incompatible interaction between flax and Melampsora lini [20].

More recently, Raes et al. [21] demonstrated by RT-PCR the expression of AtCCR2 in stems at late developmental stages, seedlings, and roots. Moreover, the expression of AtCCR2, but not AtCCR1, was enhanced in cellulose-deficient Arabidopsis seedlings exhibiting ectopic lignification [22]. Hence, AtCCR2 may take part in the synthesis of monolignols in some tissues and in response to developmental cues. It would be of interest to study an AtCCR2 mutant to investigate the involvement of this isoform in defence-related and/or developmental lignification.

After completion of Arabidopsis genome sequencing [23], 11 sequences have been annotated as ‘putative CCR’, although their identity with AtCCR1 can be as low as 32.8%. As underlined by Costa et al. [24], their catalytic properties have not been assessed biochemically, with the exception of AtCCR1 and AtCCR2. Hence we propose that, based on current evidence, there are only two bona fide CCR genes in Arabidopsis. It is clear that conclusions drawn only from sequence homology can lead to misannotation of genes, since enzymes with distinct biochemical functions can share a common ancestor and hence exhibit a strong degree of homology, making it difficult to delimit gene families appropriately in silico. A recent analysis of Arabidopsis genes involved in the phenylpropanoid pathway [24] revealed that for most enzyme families considered, gene identification awaits experimental evidence such as the properties of recombinant proteins or the study of mutants.

New CCR genes have been isolated on the basis of homology to known CCR clones from a few agronomically important species including maize [25], poplar [26], and tomato (Benoît van der Rest, unpublished data). So far, the number of CCR isoforms identified per species has never exceeded two. In maize, as in Arabidopsis, the two isoforms show differential expression patterns, with one predominantly associated with lignifying cells. So, it appears that CCR polymorphism within species is very limited. Paralogues exhibit distinct expression profiles, which probably relate to separate functions of the corresponding genes. There would be one CCR dedicated to constitutive lignification, and in some cases another one potentially involved in defence mechanisms.

The first transgenics with downregulated CCR activity were obtained in tobacco [27]. Lignin content was decreased down to 50% of the wild-type level for transgenic lines with a dramatic reduction of CCR activity. This represents the strongest reduction of lignin content that can be achieved by downregulation of a single enzyme in the pathway.

In Arabidopsis irx4 mutants, total lignin content estimated by thioglycolic acid assays was decreased by more than 50% in inflorescence stems, and phloroglucinol or Maüle staining failed to reveal the presence of ‘classical’ lignin in the xylem (these staining methods indicate the presence of cinnamaldehydes and S units respectively). These mutants most likely lack a functional AtCCR1 [15], meaning that their level of residual CCR activity is presumably very low. Unfortunately, no details of CCR activity or lignin composition in these plants have been published yet. Goujon et al. [16] described a set of AtCCR1 antisense lines with residual CCR activities down to about 20% of the wild type. Again, the lignin content was severely reduced in these transformants. Detailed analysis of lignin composition by thioacidolysis showed no consistent change in S/G ratio, while ferulic acid accumulated in lignins to 2–4 fold the level measured in the wild type. The general conclusion we can draw from these studies is that CCR downregulation affects lignin content by blocking the pathway to cinnamaldehydes, which results in the incorporation into lignin of earlier precursors like ferulic acid [28]. The S/G ratio being essentially unaffected, we can hypothesize that CCR does not play an important role in the control of this parameter.

Another important observation coming from antisense studies is that it takes a strong downregulation (a threshold was observed at around <30% activity) to bring about a decrease of lignin content [16,27]. Therefore, although the reaction catalysed by CCR is essential for monolignol synthesis, it is unlikely the rate limiting step in the pathway, since the activity is far in excess at least in mature organs.

CAD activity was first isolated from Picea abies and Glycine max in 1981 [9]. Several years later, purification of the corresponding enzyme was undertaken from Eucalyptus xylem. This work led to the identification of two major peaks of activity on coniferyl alcohol. The corresponding enzymes, named CAD1 and CAD2, were then purified to apparent homogeneity and their catalytic properties were further analysed. While CAD2 can reduce all three cinnamyl aldehydes into the corresponding alcohols, CAD1 only uses coumaryl and coniferyl aldehydes, and exhibits a lower affinity for these substrates than CAD2. Cloning and sequencing of CAD1 and CAD2 genes revealed that their amino acid sequences did not share any homology. Rather, CAD1 is related to CCR and dihydroflavonol reductase (DFR) sequences (48% identity/70% similarity and 43% identity/62% similarity, respectively).

A search for CAD1 homologues in EST databases allowed us to identify closely related proteins in many species. CAD1 seems to be highly conserved in tracheophytes: Eucalyptus gunnii CAD1 has 80% identical homologues in angiosperms and 74% identical homologues in some gymnosperms. In non-vascular land plants (e.g., Physcomitrella patens), expressed sequences homologous to CAD1 and CCR could represent a putative common ancestor for these genes. Based on available sequences, the Physcomitrella protein would be 78% and 67% similar to E. gunnii CAD1 and CCR, respectively.

CAD1 and CAD2 belong to two different families of alcohol dehydrogenases: short chain dehydrogenases and zinc-binding dehydrogenases, respectively. The active form of CAD2 is a dimer while CAD1 is active as a monomer.

Because CAD2 efficiently catalyses the synthesis of all three monolignols, it was considered to be the more relevant enzyme for lignification. The first cDNA-encoding CAD2 was cloned from tobacco [29], and the first transformants carrying antisense CAD2 constructs were also obtained in this species [30]. The analysis of these plants as well as transformants and mutants in other species established the role of CAD2 in lignification in both gymnosperms and angiosperms.

Subsequent search for CAD genes from other species was always homology-based (using CAD2 as a probe), and thus led only to the identification of CAD2 orthologues. Therefore, genes annotated as CAD1 in various species (including Arabidopsis and alfalfa) do not correspond to CAD1 but to CAD2 orthologues.

Tobacco CAD1 was further characterised in our lab [31] and was shown to use a wider range of substrates, including benzaldehyde derivatives. Besides the obvious participation of CAD2 in lignin synthesis, the involvement of CAD1 in lignification cannot be ruled out for the following reasons:

• – CAD1 was originally extracted from xylem tissue;
• – CAD1 is not able to synthesize S units, but in certain tissues or developmental stages lignins mostly comprise G units (e.g. primary xylem).

In this respect, it is interesting to note that the ‘red xylem’ phenotype characteristic of CAD2-downregulated plants is confined to secondary xylem: primary xylem cells, which are also lignified, appear normal [32]. Also, lignins enriched in H and G units can be found in the outermost part of the cell wall (middle lamella and cell corners), this region corresponding to the initiation sites for the assembly of lignin polymers.

The synthesis of syringyl units of lignin represents an example of pathway diversification in angiosperms since gymnosperms are devoid of such units in their lignins. Li et al. [33] have suggested that the reduction of coniferyl and sinapyl aldehydes was achieved by distinct enzymes. The authors postulated that the previously identified CADs were guaiacyl-specific, and identified a new syringyl-specific isoform, termed SAD for sinapyl alcohol dehydrogenase. Interestingly, they went about searching for SAD by probing an aspen cDNA library with a CAD2 clone. Two cDNAs were isolated: one with very high homology with CAD (PtCAD), and the other with only 50% homology (PtSAD). Recombinant enzymes were able to produce all three monolignols, but with a strong preference for coniferyl and for sinapyl alcohol, respectively. The distinct CAD and SAD functions are linked spatiotemporally to the differential biosynthesis of guaiacyl and syringyl lignins in different cell types (vessels enriched in guaiacyl units/phloem fibres enriched in syringyl units). The authors conclude that the last step of the synthesis of G and S units occurs via distinct enzymes in angiosperms. However, sinapaldehyde is abundant in lignins of CAD-deficient poplars [34], which suggests that SAD may not play a substantial role in constitutive lignification in poplar. Here again, studies in Arabidopsis can shed light on the question of CAD polymorphism, since (a) the full genome sequence and a great number of ESTs are available, (b) recombinant proteins have been obtained for all CAD2 homologues and their catalytic properties analysed, and (c) mutants for the main CAD2 isoforms have been described [35]. Among nine genes encoding putative cinnamyl alcohol dehydrogenases [24,36], only two (At3g19450 and At4g34230) possess a significant activity on hydroxycinnamaldehydes, based on the analysis of the corresponding recombinant proteins [37]. Another four (At4g37990, At4g37980, At2g21890, At2g21730) show comparatively little activity, while the last three (At4g39330, At4g37970, At1g72680) did not exhibit any detectable CAD activity. None of them was found to behave like a SAD (i.e. exhibited a preference for sinapaldehyde). Among these genes, three seem to be expressed at a very low level, or not at all, as suggested by the absence of the corresponding sequences in EST databases [36]: At2g21890, At2g21730 and At4g37970.

These data suggest that there is no specific SAD in Arabidopsis, and that conversion of sinapaldehyde to sinapyl alcohol is achieved by the two main CAD isoforms (At3g19450 and At4g34230). This is confirmed by the analysis of the corresponding mutants [35]: the S/G ratio decreased in both AtCAD-C and AtCAD-D mutants, showing that both isoforms participate in the synthesis of S units.

Looking at the molecular data available for other angiosperms, it is difficult to tell how widely SAD is represented, especially because large EST collections or genomic sequences are only available for very few plants. Many species possess several genes homologous to CAD2. Strong homologies with SAD (in the range of 65 to 75% identity) can be found in several species, notably woody species (e.g., Citrus sinensis, Vitis vinifera, Prunus persica). In many other taxa, the EST that comes closest to SAD is only around 50% identical, and most likely represents another type of CAD variant.

Consistent with these observations, a biochemical approach based on isotope tracer experiments performed on Robinia, Oleander, Magnolia and Arabidopsis revealed that the synthesis of S units is achieved via different pathways in these species [38]. Robinia and Oleander have a pathway that produces sinapyl alcohol from sinapic acid via sinapoyl CoA, while Magnolia and Arabidopsis have a sinapyl alcohol biosynthetic pathway that does not require sinapic acid as an intermediate. Similarly, previous data from Chen et al. [39] obtained on Magnolia kobus showed the incorporation of deuterium-labelled coniferyl alcohol into syringyl units. Hydroxylation and methylation steps would then occur at the level of the monolignols themselves. Together, these data strongly suggest that the S lignin biosynthetic pathway varies among Angiosperms.

An in-depth biochemical study of Eucalyptus gunnii CAD2 iso-enzymes purified from periderm tissue [40] introduced the idea that the polymorphism of CAD catalytic properties could result in part from distinct combinations of subunits. In that study, it was shown that CAD2 isoforms were dimers of two different subunits. The two CAD subunits might represent allelic differences as previously reported in loblolly pine [41] or the products of two different genes as reported for soybean aliphatic alcohol dehydrogenase [42]. They were differentially combined to give the heterodimer but also two homodimers, resulting in three different combinations associated with distinct enzymatic properties. The high molecular weight homoCAD works most efficiently with sinapaldehyde, whereas heteroCAD works equally efficiently with both sinapaldehyde and coniferaldehyde. These data suggest that beyond the differences in protein sequence, combinatory mechanisms can generate a polymorphism of functional significance.

As in the case of CCR, there seems to exist a sub-family of CAD-related enzymes devoted to defence. For example, two lucerne CAD cDNAs (MsaCAD1 and MsaCAD2) are differentially regulated during development and plant defence [43]. A significant amount of data [19,44,45] suggests that the chemical composition of defence lignin is different to that of developmental lignin. In Eucalyptus plantlets, for example, defence lignins produced after wounding are poor in syringyl units unlike the usual G/S mixed type. Preliminary approaches indicated that two different types of cinnamyl alcohol dehydrogenase activity are induced and apparently regulated differentially, in response to wounding in Eucalyptus trees [19].

In addition to ‘true’ CAD cDNAs that have been isolated from a number of species (for a review, see [46]), several other cDNAs showing a homology of approximately 50% with the typical CAD cDNAs have also been characterized. Interestingly, these other cDNAs were all initially isolated as pathogen-related defence genes. For example, the ELI-3 (for elicitor-induced 3, [47]) transcript was identified in a cDNA library of parsley cells challenged with a fungal elicitor. Later, two ELI3 homologues were identified in Arabidopsis. Sequence analysis first suggested that ELI3 was a mannitol dehydrogenase. Further studies involving ELI3 recombinant protein proved this assumption wrong and revealed that the highest activity of ELI3 was against benzaldehydes. Although ELI3 shares significant homology with CAD2 (60% similarity of the amino acid sequences), it is only weakly active on cinnamaldehydes [48]. It is not clear at the moment whether the proteins encoded by these genes participate in the synthesis of unusual lignin monomers and thus in lignification or in the production of specific phenolics with a putative role in defence.

Hence, it appears that CAD genes have evolved into larger families than CCR (particularly in angiosperms), which resulted in a diversification of biochemical function: some members of the family could be specialized in the synthesis of lignin monomers, while others (like ELI-3) are involved in the production of defence-related compounds outside the scope of lignin synthesis.