You are here

PhD Project - Nodulation gene phylogenetics

This page is a quick conversion to text format, for figures, tables, and references, see the pdf file.

Nodulation gene phylogenetics


Rhizobia--legume symbiosis

Rhizobia live in a mutualistic symbiotic relationship with legumes---a relationship that has existed and co-evolved for tens of millions of years Sprent94. The nodulation process includes a complex array of signalling molecules, molecular recognition, and regulation. Legumes secrete secondary metabolites known as flavonoids into the soil; rhizobia, which are motile, are attracted to these flavonoids and attach to the root surface (rhizoplane). The flavonoids also induce the bacteria to secrete specific signal molecules, known as Nod factors Werner04.

Nod factors bind to a receptor in the root hair cell, and cause root hair curling, and eventual penetration of the bacterium into the root hair cell. Hence, Nod factors are a critical molecules for nodule formation. After entering the root hair, bacteria travel down an infection thread---a plant structure made specifically for this purpose Gage00.

The growing infection thread branches as it reaches the developing nodule primordium, formed by dividing cortical cells. This growth is also initiated by Nod factors, which reactivate the cell cycle Patriarca04. In most cases rhizobia then differentiate morphologically to form bacteroids, which are usually larger than the free-living bacteria and have altered cell walls; bacteroids are released from the infection thread and form symbiosomes in nodule cells Oke99. Bacteroids are the nitrogen-fixing cells, and are incapable of cell division and further reproduction Perret00.

A compatible Nod factor is not the only requirement for effective nodulation. Bacterial cell surface components such as lipopolysaccharides (LPS), cyclic--glucans, exopolysaccharides (EPS), capsular proteins, and K-antigens are also recognised by the plant, and help determine host specificity Spaink00,Fraysse03,Mathis05. If these components are not recognised by the host, then the process is disturbed to various degrees. For example, if infection threads fail to form, non-fixing empty nodules (Nod Fix) may result Perret00.

nod genes and Nod factors

The Nod factor is produced by, and is under the control of, nod genes. Typically, nod genes are present in the bacterial cell on a transmissible genetic element, such as a Sym plasmid Martinez87,Sharma05, or symbiosis island Sullivan98. In Bradyrhizobium, nod genes are integrated into the chromosome in a putative symbiosis island Kaneko02. Thus, because they are transmissible, the evolutionary lineage of the symbiosis genes may be different from the housekeeping genes. As an example, rhizobia nodulating common bean (Phaseolus vulgaris) have been placed in several rhizobial genera, based largely on 16S rRNA gene sequence analysis. In contrast, when nodC (nodulation) and nifH (fixation) genes were analysed, the isolates were found to be more closely related Laguerre01.


The structure of the Nod factor of Ensifer meliloti was first described by Lerouge90. The molecule is a `lipo-chito-oligosaccharide' (LCO), consisting of a chitin-like backbone, with a fatty acyl side chain. The `core symbiosis genes' nodA, nodB, and nodC, are required to synthesise the Nod factor. NodC, a -glucosaminyl transferase, links UDP-N-acetyl glucosamine monomers into the chitin-like backbone. NodB removes an acetyl group from the terminal residue of the chitin oligomer. NodA then catalyses the transfer of a fatty acyl chain onto the resulting free amino group Hirsch01. Although NodB and NodC have homologies to known proteins, NodA is a unique acyl transferase because, in contrast to all other fatty acylated polysaccharides which have acylated sugars added during elongation, NodA adds a fatty acyl chain to a preformed polysaccharide Hirsch01. Additionally, nodABC genes have a lower G+C content than other Rhizobiaceae genes, and a different codon usage pattern Galibert01, suggesting an ancestral horizontal transfer from a presently unknown source.

After synthesis, the Nod factor is modified by the addition of various substituents (acetate, sulphate, carbamoyl groups, or sugars such as arabinose, mannose, or fucose), under the control of other nod genes, and it is these modifications (`decorations') that confer most of the specificity Laeremans98.

Previous work on New Zealand nod genes

Some research has been done on characterising the nod genes in New Zealand rhizobia. Although the type strain for Mesorhizobium loti was isolated in New Zealand, and significant molecular analysis has been done on the locally isolated R7A strain, the symbiosis region of these strains were derived from an exotic Mesorhizobium spp.specific to Lotus spp.Sullivan98,Sullivan02. McCallum96 used a nod gene probe to determine that nod genes were mostly carried on plasmids in Mesorhizobium spp.isolated from native legumes, although they were located on the chromosome in some strains.

Since New Zealand legumes have been reproductively isolated from relatives overseas, it may be reasonable to assume that host-specific symbioses have established, and this may have led to unique nodulation genes.



The objectives of this part of the research were:

itemize To determine the evolutionary history and origins of putatively novel nodulation genes specific to rhizobia that nodulate New Zealand legumes. To determine if strains of Rhizobium leguminosarum nodulating native legumes acquired symbiotic genes from nodulating Mesorhizobium species. itemize



The nodA gene was chosen for analysis as it is one of three that construct the critical Nod factor molecule, and many sequences were available on GenBank for comparison. nodA was PCR amplified and sequenced for each strain where possible, using conditions and cycle parameters detailed in the methods chapter. The sequence data were aligned with nodA sequences from other strains of rhizobia available in GenBank, and the DNA and protein sequences analysed with Maximum Likelihood and Bayesian inference to build phylogenetic trees. Protein sequences were used, as the protein is the unit of selection, and to counteract the effect of substitution bias at the third codon position (`wobble base') on phylogenies. Saturation of this base has been shown to contribute to phylogenetic misinformation Mindell96. Selection of an outgroup was difficult, as nodA has no homologue outside of rhizobia, hence trees were rooted with Azorhizobium strain SD02, due to its divergence from all other sequences.


Amplification, alignment, and analysis

Amplification of the nodA gene was successful for almost all isolates attempted after extensive optimisation of primer choice and PCR cycle parameters, although no product was able to be amplified from strain 12635 (Carmichaelia petriei).

figure [width=14cm]nodA-MB [nodA Bayesian inference phylogenetic tree]Bayesian inference phylogenetic tree showing the relationship of nodA sequences of rhizobia isolated from New Zealand legumes compared with global sequences. Letters in bold indicate genomic grouping as defined by the 16S rRNA phylogeny in Chapter 3. The model of evolution used was GTR+I+ and was run for 1010 generations. Scale bar indicates number of expected changes per site. Clade posterior probability is indicated above the node. p-nodA-MB figure

figure [width=14cm]nodA-protein-MB [NodA protein Bayesian inference phylogenetic tree]Bayesian inference phylogenetic tree of NodA protein sequences of rhizobia isolated from New Zealand legumes compared with global sequences. Letters in bold indicate genomic grouping as defined by the 16S rRNA phylogeny in Chapter 3. The model of evolution used was was Jukes+ run for 210 generations. Scale bar indicates number of expected changes per site. Clade posterior probability is indicated above the node. p-nodA-protein-MB figure

The alignment of the nodA DNA sequence had 62 taxa and was 576-bp long. A total of 21 sequences were obtained from native legumes, and 10 from introduced woody legume weeds (Table t-GenBank-nodA), these were compared with the nodA gene from 31 diverse strains of rhizobia (Table t-GenBank-nodA-type). All Bradyrhizobium strains, Methylobacterium, and Burkholderia, have a three base pair indel, and type 8 nodA genes have another three base pair indel (GAC) that is shared with Mesorhizobium loti strains. There is also a six base pair insertion at a different position in the Mesorhizobium septentrionale sequence.

The ML tree is not shown, as it was identical to the Bayesian inference DNA tree. The data were analysed under Bayesian analysis using the GTR+I+ model for ten million generations. The consensus tree is shown in Figure p-nodA-MB, numbers above nodes are the marginal posterior probabilities of the clade being correct. The gene sequence was translated to a protein sequence of 191 aa, and the ML model of protein evolution selected was JTT+. The same model was used for the Bayesian analysis of the protein data (called Jones+ in MrBayes), and run for two million generations (Fig.p-atpD-protein-MB). The ML protein tree is not shown as it is nearly identical to the Bayesian tree, with the exception of some deep branching (see corresponding low clade probability in the Bayesian tree).

Grouping of nodA types

The topology of the nodA tree was quite different from the housekeeping gene trees described in Chapter 3. In this case there appears to be a host-specific grouping of nodA genotypes. New Zealand native legumes were nodulated by rhizobia carrying five different nodA genes; introduced weed legumes were nodulated by rhizobia carrying two nodA types.

Type 1 -- `Carmichaelinae 1' The type 1 clade has the largest number of members (ten) and the sequences were nearly identical, apart from a single amino acid change from glutamic acid to aspartic acid. These sequences are very divergent from all other currently known nodA genes, as indicated by the branch to the root of the tree. Because of this divergence, the phylogenetic position of the type 1 clade is unknown. It appears to be joined to the Azorhizobium sequence; but this is almost certainly Long Branch Attraction (LBA), where two divergent clades appear artificially close to each other [reviewed in][]Bergsten05. A BLAST of these nucleotide sequences to others on GenBank reveal that they share little similarity with any other sequence. The only similarity is 50--100-bp of the sequence to Bradyrhizobium nodA sequences from Australia (82 identity over that region), possibly indicating a conserved functional region. In a BLAST of protein sequences however, there was 61 identity (75 similarity) of the entire sequence (179 aa) to M.ciceri strain UPM-Ca7 isolated from Cicer arietinum. The relative conservation of the protein sequence compared to the DNA sequence indicates that some of the differences this gene has accumulated are silent (in the third codon position). All strains containing the type 1 nodA gene were Mesorhizobium species isolated from Clianthus, Carmichaelia and Montigena plants that are classified in the `Carmichaelinae' legume subtribe Wagstaff99.

Type 2 -- `Carmichaelinae 2' The type 2 (`Carmichaelinae 2') clade also has sequences from Clianthus, Carmichaelia, and Montigena, but only has three members. The sequence from strain 11541 has four base pair changes resulting in four amino acid changes. These sequences are grouped in a larger clade containing other Mesorhizobium species: M.ciceri (Cicer), M.mediterraneum (Cicer), M.temperatum (Astragalus), and M.tianshanense (Glycyrrhiza).

Type 3 -- `Sophora' The type 3 (`Sophora') clade's three members are solely from Group A Mesorhizobium strains isolated from Sophora. Two sequences are identical, but the sequence from strain 11736 has two base pair changes resulting in two amino acid changes. The phylogenetic position of the clade is uncertain as in the DNA tree this clade groups with type 6 (`Phaseoli'), but in the protein tree has it has no close relation and the branch goes to the root of the tree, although in nucleotide BLAST searches the closest matches are M.tianshanense, and M.temperatum (82 identity). In the DNA tree the clade probability is 0.71, but in the protein tree it is only 0.56, (low support---the tree is 50 consensus majority rule).


Types 4,5,6 -- leguminosarum biovars The type 4 (`Trifolii') clade is made up of five R.leguminosarum strains (and the NLNP isolate, see section s-pristine-soil). All strains appear to be of the trifolii biovar. Three Genomic group E strains isolated from native legumes are in this clade. The type 5 (`Viciae') clade is made up of three R.leguminosarum strains including one isolate from Sophora. No New Zealand isolates were found in the `Phaseoli' clade (type 6), but this was included for a comparison study described in the next chapter.

Types 7,8 -- Introduced weeds The type 7 (`Genisteae') clade contains five Bradyrhizobium isolates from Ulex and Cytisus, as well as three comparison strains from Genista, Lupinus, and Cytisus. The type 8 (`Acacia') clade contains five Bradyrhizobium isolates from Acacia and Albizia, as well as three comparison strains from Acacia and Gompholobium.


figure [width=12cm]isolates-nodA [Geographical distribution of nodA types]Map of New Zealand showing geographical distribution of nodA gene types. The genus of the rhizobial isolate is indicated by the shape of marker, the number inside indicates gene type. p-nodA-map figure

It appears that there is no geographical localisation of nodA genotypes (Fig. p-nodA-map), and each genotype was found throughout New Zealand.



Tree topology

The topology of the phylogenetic trees was generally consistent between different methods of analysis, and between the DNA and protein data. In fact, the Maximum Likelihood and Bayesian DNA trees were nearly identical; and each had excellent support in the tree-island profile, and measures of convergence, respectively.

Nevertheless, there were a few differences between DNA and protein trees, exclusively in the deep branching of the tree, where clade posterior probability was low. The type 3 (`Sophora') clade and type 6 (`Phaseoli') appear to be linked in the DNA tree, but are separated in the protein tree. Another deviation between the trees is Bradyrhizobium japonicum USDA110, which in the DNA trees is an outgroup to all other Bradyrhizobium sequences, but in the protein tree it is also an outgroup to Methylobacterium and Burkholderia. The clade support of closely related clades was significantly greater, with many clades having a posterior probability of 1.00.

It is likely that the deep branching differences can be explained by long branch attraction, which could be rectified by the addition of taxa that have a phylogenetic position between that of the existing clades Bergsten05. However, adding taxa similar to the affected clades is difficult, as some nodA sequences are novel and distinct and do not have close relations in available databases. Further investigations of non crop-and-forage legumes from other countries, particularly those related to the New Zealand legumes such as the Australian Swainsona and Sophora species, would help to elucidate these relationships further.

Horizontal transfer of nodA genes within rhizobial genera

Inferred nodA phylogenies from New Zealand rhizobia has revealed an evolutionary history distinct from that of the housekeeping genes and from which genomic groups were assigned. Horizontal gene transfer is the most plausible hypothesis to explain this phylogenetic incongruence Martinez96,Young96a.

With the exception of nodA groups 2 and 3, there was little correspondence of nodA type to genomic group determined in Chapter 3. Type 1 nodA sequences were isolated from genomic groups A, B, C, and D. nodA types 2 and 3 were smaller groups, that did show a relation of to genomic group; type 2 sequences came from Group D strains, and type 3 sequences from Group A strains. It is possible that in this case the congruence of nodA and genomic groups may represent a nodA--strain specificity or may be due to small sample size bias.

There was also no clear correspondence of nodA type to genomic group in Bradyrhizobium spp., but a clear correlation to the host legume. This may provide more evidence for the transmissibility of symbiotic elements in Bradyrhizobium that has been suggested in the literature Kaneko02,Moulin04, but not yet verified.

nodA sequences from Mesorhizobium strains formed three clades (1, 2, and 3), two of which appear to be novel genotypes, and the other (type 2) grouping with known Mesorhizobium nodA sequences. nodA genes of R.leguminosarum strains, isolated from New Zealand legumes, grouped with typical known R.leguminosarum nodA genes, predominantly of the trifolii biovar. All Bradyrhizobium nodA sequences clustered together, in two related clades. These data suggest that nodA genes (and by extension transmissible genetic elements) have not transferred between rhizobial genera, although they may transfer within a genus. This pattern has been noted before in several genera using nodB and nodC Wernegreen99.

The cause of the incongruence in inferred phylogenies of housekeeping and nodulation genes is almost certainly horizontal transfer of nod genes mediated by either symbiosis plasmids or symbiosis islands. This hypothesis provides an explanation for the presence of multiple genomic groups of rhizobia capable of nodulating each native legumes species, seen in Chapter 3. Such conservation of nod genes, despite genotypic diversity has been noted several times before, such as in Astragalus sinicus rhizobia Zhang00, and Rhizobium galegae Suominen01.

Although nodulation genes have been shown to transfer to different genera before, even phylogenetically distant ones Moulin01, this apparently has not occurred in the rhizobia of New Zealand. A possible explanation for this may lie in the physiological differences of rhizobia strains. It is possible that rhizobia may not have the correct mechanisms for integration and propagation of `foreign' transmissible elements. Nevertheless, a bacterium not related to the Rhizobiaceae, vis. Sphingobacterium multivorum (in the phylum Bacteroidetes) was shown to carry and express nodulation genes from plasmids, although the nodulation was ineffective Fenton94, indicating that nodulation genes could be expressed in diverse organisms.

Alternatively, it may be possible that plasmids or symbiosis islands can exist in different genera, and produce specific Nod factors, yet successful nodulation of a plant may be restricted by bacterial cell determinants of host-specificity such as lipopolysaccharides, cyclic--glucans, exopolysaccharides, and capsular proteins. However, this latter explanation would not account for R.leguminosarum strains capable of nodulating native legumes.



Specificity of nodA to host legume

The original hypothesis (see Section EOS) was that nod genes would be genus-specific, or there would be a single broad host-range transmissible element carrying nodulation genes. These data indicate that nodA is transmissible within rhizobial genera, and seems specific to its original host legume to the genus or subtribe. Host specificity of the nodA gene has been reported elsewhere for rhizobia that nodulate Vicia, Medicago, Trifolium, Pisum, and Galega Ritsema96,Roche96,Suominen01

All the sequences from introduced woody legume weeds were placed in either of two clades. Sequences from Acacia and the related Albizia were found in the type 8 clade, along with sequences from Acacia found elsewhere in the world. Sequences from broom (Cytisus) and gorse (Ulex) were found in the type 7 clade along with the related species of Genista tinctoria, Cytisus sp., and Lupinus albus. The close relation of these nodA types to their overseas counterparts may indicate recent transfer of effective strains to New Zealand. This pattern was seen with rhizobia nodulating introduced lupins in Australia, where it was concluded that lupin nodulating strains were not native to Australia, but were instead of European origin Stepkowski05. The sequences of the lupin clade (Clade II) in the Stepkowski study match those of the type 7 (`Genisteae' -- gorse and broom) clade of this analysis (Fig. p-nodA-MB). This may indicate that lupins in New Zealand are nodulated by the same rhizobia as gorse and broom, although this was not tested in this thesis.

The specificity of nodA type to isolated host legume was also seen with Mesorhizobium spp. Type 3 sequences were solely isolated from Mesorhizobium spp.nodulating Sophora species. Carmichaelia, Clianthus and Montigena were nodulated by strains with nodA types 1 and 2. Species or genera specificity was not seen, but the three genera are related to each other in the Carmichaelinae sub-tribe Wagstaff99, which may indicate specificity to the sub-tribal level. An Australian genus, Swainsona, is the only other member of the Carmichaelinae. Greenwood69 demonstrated that strains isolated from New Zealand native legumes were able to nodulate Swainsona, providing more evidence for the hypothesis of sub-tribal host specificity. The ability of native legume rhizobial strains to nodulate exotic legumes is described in the next chapter.

nodA is only one component that contributes to the host-range limits of rhizobia. However it is located in close proximity to other symbiosis genes on the bacterial chromosome (or plasmid), and it is likely that all of the nodulation genes transfer as a single unit; hence the examination of a single gene may indicate the host-range abilities of an entire symbiosis region. Nevertheless, the sequencing of other nodulation genes would allow for a more complete picture of the structure of Nod factors, including specific modifications.

Rhizobium leguminosarum nodA types

In the previous chapter, it was established that most strains nodulating the native legumes are members of Mesorhizobium, the exception to this was four isolates identified as Rhizobium leguminosarum. One explanation for their nodulation of native legumes was that R.leguminosarum strains acquired specific nodulating genes from native Mesorhizobium spp. However, the data presented here shows no evidence for transfer of Mesorhizobium nodulation genes into R.leguminosarum. It appears that R.leguminosarum strains isolated from native legumes have typical nodA genes for that species, and segregate well with the known biovars.

The ability of these putatively introduced strains to nodulate native legumes raises fundamental questions about the relationship of nodA to specificity. The core nodulation genes (nodABC) are often referred to as the `common' nodulation genes, as nodulation ability can be restored to strains carrying non-functional mutants of these genes by complementation of the genes from other strains Coplin89. However, this original work was done with nod genes from closely related rhizobial strains. Later research showed that NodA from a Bradyrhizobium sp.was unable to attach a fatty acyl chain from R.leguminosarum bv.viciae on to the chitin backbone of the Bradyrhizobium Nod factor Ritsema96. Other research showed that allelic variation of the nodABC genes plays an important role in signalling variation and in the control of host range Roche96,Debelle96. Hence the most recent literature supports the notion that nodA type does correlate to host-range nodulation ability.

This raises an interesting question in this research of how `leguminosarum' Nod factors (bv.trifolii and bv.viciae) were able to nodulate native legumes. Assuming that there was no error made in the accession of these strains, there are two possibilities, notably either legume or rhizobial promiscuity.

The first possibility is that the New Zealand native legumes are promiscuous and allow nodulation by rhizobia producing many different kinds of Nod factors. Some evidence for this is indicated by the two distinct types (1 and 2) that nodulate the Carmichaelinae species. However, the absence of nodulation by other strains (Rhizobium spp., Ensifer spp., and Bradyrhizobium spp.) argues against the possibility of legume promiscuity.

The second possibility is that the R.leguminosarum used in this study are promiscuous (i.e. have a broad host-range). This is not uncommon, as Ensifer fredii strain NGR234 is known to nodulate 232 species in 112 genera Hirsch01,Pueppke99,Saldana03, much of this ability to nodulate stems from the more than 80 different Nod factors that it secretes Berck99.

In order to answer these questions, the host range of these strains was thoroughly investigated and described in Section s-R.leg of the next chapter.

Chromosomal organisation of nod genes

The chromosomal arrangement of nodulation genes varies between rhizobial genera, species, and strains. In most species the core nod genes (nodABC) are part of a single operon, but in other species the arrangement can vary VanRhijn95. An investigation of these arrangements may give insight into the evolutionary history of this region of the transmissible symbiosis element.

The arrangement of genes can be determined by the binding of PCR primers. Primers nodA1 and nodA2 correspond to residues 14--37 of nodA and 65--43 of nodB of the E.meliloti 1021 sequence (GenBank: M112684) Haukka98. These primers amplified most R.leguminosarum nodA sequences (types 4 and 5), and Mesorhizobium type 3 sequences, but not types 1 or 2. When primer nodA2 (binding to nodB) was substituted with nodA3 (binding to nodA), products were amplified for types 1 and 2, but not for R.leguminosarum sequences. This implies that the nodA and nodB genes are separated in type 1 and 2 strains, and adjacent with type 3, 4, and 5 strains.

This is consistent with M.loti strains MAFF303099 and R7A, where nodB is about 10 kb downstream of nodA, Scott96,Sullivan02,Kaneko00 and with Mesorhizobium spp.isolated from Astragalus sinicus Zhang00, where nodB is 22 kb upstream. nodB is also separated from nodA in some fast growing strains from native Australian legumes Watkin05. In other Mesorhizobium species; (M.ciceri, M.mediterraneum, M.plurifarium, and M.tianshanense) nodB is adjacent Zhang00.

In Bradyrhizobium sp.USDA110, nodA and nodB are adjacent, in the order nodD1YAB. All Bradyrhizobium spp.from New Zealand amplified with the primers TSnodD1-1a and TSnodB1 (binding to nodD1 and nodB genes), which indicates that they have the same arrangement of nod genes as other Bradyrhizobium spp., providing further evidence for the close relationship of the Bradyrhizobium symbiosis regions in New Zealand strains to those found elsewhere in the world.

Novel nodA genotypes in Mesorhizobium spp.

Nodulation genes differ from housekeeping genes, in that they are under different evolutionary pressures, by producing a molecule that interfaces with another organism; this selective pressure could potentially lead to novel nod types in different rhizobia--legume symbioses. In addition, unlike the house-keeping genes, only a tiny fraction of existing nodA genes have been sequenced, implying that poorly researched symbioses (typically those associated with non crop-and-forage legumes) may have novel genotypes, as yet uncharacterised.

Sequences from type 1 (`Carmichaelinae 1') and type 3 (`Sophora') do not have close matches in the GenBank database, and are distinct in the phylogenetic trees. These sequences are therefore considered to represent novel nodA genotypes. Although these sequences are substantially different from previously known sequences, they are almost certainly nodA genes, as they are exactly the same length as other nodA sequences, have many homologous regions, are amplified by nodA primers, and flanked by other nod genes.

The relationship of these novel nodA types to other genotypes is difficult to establish due to the long branch lengths and would require more sequence data from the rhizobia of related legumes. The nodA types of native legumes, and possibly by extension the entire transmissible symbiosis region, may represent separate dispersal events to New Zealand, possibly from different geographical sources.

An interesting property of the novel sequences is that they are highly conserved within a clade, it is remarkable that these sequences have diverged so much from other known sequences, and yet have little within-group variation. This is probably indicative of isolated symbiotic co-evolution with the native legumes of New Zealand, rather than random genetic drift.

Co-evolution and novel symbiosis genotypes

The novel nodA genotypes seen in Mesorhizobium isolates from native legumes may have arisen through co-evolution of rhizobia and recently dispersed legumes.

Evolution is driven by natural selection. Specifically this means an environmental pressure, acting on natural variation, providing a competitive advantage to an individual (conferring greater reproductive success) Dawkins86. A relevant well-understood example is bacterial plant--pathogen interactions, where co-evolution leads to an `arms-race' where each side develops better mechanisms of attack (pathogen), and better mechanisms of defence (plant). Successful attack or defence leads to greater reproductive success of the individual Dawkins79,Frank92. Co-evolution in rhizobia--legume interactions is somewhat more complicated, as it is generally a mutualistic symbiosis, but has elements of parasitism in the form of ineffective nodulation.

Effective nodulation is beneficial for both bacteria and plant for reasons already covered in Section s-rhizobia-intro. In the effective nodulation of a plant by an established strain there should be no significant selective pressure for change of nodulation ability. Although random mutations would arise in either the Nod factor or receptor, these would be typically be disadvantageous to the established relationship, thus this process would lead to relatively stable gene sequences for nodulation genes, and the corresponding Nod factor receptor of plants. There may also be selective pressure acting on nitrogen fixation genes, and regulatory genes, caused by selection of the best nitrogen fixing strains by the legume; this however cannot be determined by examining nodA gene sequences.

In contrast, ineffective nodulation (not the empty nodule kindEmpty-nodule ineffective nodulation is where rhizobial Nod factors cause nodule formation, but bacteria do not exist in the nodule, perhaps because of an aborted infection thread Perret00.) can be compared to parasitism with the rhizobia gaining all the benefits of symbiosis at no cost. It is beneficial for legumes to prevent ineffective nodulation, as a significant proportion of photosynthate goes to nodule production and upkeep Provorov98b. The bacterial partner of the symbiosis benefits from the physical shelter/protection and access to plant-derived nutrients. The cost is that cells direct energy into N fixation and become unable to divide, ultimately resulting in death. This would therefore seem to be an evolutionary driver for nodulation without fixation.

It is also in the best interests of rhizobia to nodulate as widely as possible, and this is seen with host ranges of plants, where often the ineffective nodulation range is broader than the effective range Crow81. Whilst it may seem that this scenario would lead to exclusive parasitism, plants can reduce the reproductive success of individual ineffective nodules (for example by restricting the oxygen supply Denison04b).

These concepts, notably that novel nod genotypes may arise in response to plant mechanisms to prohibit ineffective nodulation, may apply to the evolution of novel nod genes in New Zealand rhizobia. Legumes have two pre-nodulation mechanisms to prohibit nodulation; one is to change the structure of the Nod factor receptor, such that it no longer recognises the Nod factors of ineffective strains. Such a change would drive co-evolution in the Nod factors of both its preferential effective symbionts and ineffective parasitic strains. Another mechanism to prohibit nodulation is by hydrolysing Nod factors in the rhizosphere. Legumes produce at least six different classes of chitinases which can cleave the backbone of a Nod factor destroying its function Perret00. As the specificity of susceptibly to chitinases is determined by Nod factor structure Schultze98, this may also drive evolution in the nod genes to produce a molecule less susceptible to chitinase action.

It is proposed that parasitism by ineffective strains, at some point in the history of legumes in New Zealand, drove the co-evolution of effective strains and the host legume to prohibit the ineffective nodulation.

The evolutionary process probably does not act directly on the NodA protein, but from downstream effects. NodA catalyses the addition of a fatty acyl chain onto a chitin backbone, it would be changes in the acyl chain and backbone that would confer different specificity, but NodA would likewise have to alter in order to recognise its substrates (altered acyl chain and backbone).

There are other mechanisms that could explain novel nodA genotypes. One is genetic drift, and this certainly seems to have played a role, as the type 1 sequence was more similar to other nodA genes in the protein rather than DNA sequence. However, even the protein sequence was very divergent from all other sequences. The main problem with a genetic drift hypothesis is that these data show little internal variation in nodA gene types. All ten type 1 sequences are nearly identical (excepting one residue change). Had genetic drift played a large role then one would expect these sequences to be more divergent.



The nodA gene was sequenced from rhizobial strains nodulating both New Zealand native legumes and introduced woody weeds. An inferred phylogenetic tree showed a topology distinct from those of the housekeeping genes, that correlated to the host legumes of the strain. Horizontal transfer of nodA genes within (but not between) rhizobial genera on transmissible genetic elements was proposed as a mechanism explaining this pattern.

Horizontal transfer does not explain the ability of R.leguminosarum strains to nodulate the native legumes, which were found to possess typical nodA genes for that species.

A mechanism of co-evolution of effective strains and legumes in response to ineffective nodulation by parasitic strains was proposed for the presence of novel nodA genotypes. In the next chapter the host-range of these rhizobial strains is determined by inoculation of legumes under controlled conditions. This allowed examination of the phenotypic effect of different nod genes, and a determination of effective or ineffective nitrogen-fixing ability.