Molecular phylogeny of the actinorhizal Hamamelidae and relationships with host promiscuity towards Frankia

Several of the most studied actinorhizal symbioses involve associations between host plants in the subclass Hamamelidae of the dicots and actinomycetes of the genus Frankia. These actinorhizal plants comprise eight genera distributed among three families of ‘higher’ Hamamelidae, the Betulaceae, Myricaceae, and Casuarinaceae. Contrasting promiscuity towards Frankia is encountered among the different actinorhizal members of these families, and a better assessment of the evolutionary history of these actinorhizal taxa could help to understand the observed contrasts and their implications for the ecology and evolution of the actinorhizal symbiosis. Complete DNA sequences of the chloroplast gene coding for the large subunit of ribulose 1,5‐bisphosphate carboxylase (rbcL) were obtained from taxa representative of these families and the Fagaceae. The phylogenetic relationships among and within these families were estimated using parsimony and distance‐matrix approaches. All families appeared monophyletic. The Myricaceae appeared to derive first before the Betulaceae and the Casuarinaceae. In the Casuarinaceae, the genus Gymnostoma derived before the genera Casuarina and Allocasuarina, which were found closely related. The analysis of character‐state changes in promiscuity along the consensus tree topology suggested a strong relationship between the evolutionary history of host plants and their promiscuity toward Frankia. Indeed, the actinorhizal taxa that diverged more recently in this group of plants were shown to be susceptible to a narrower spectrum of Frankia strains. The results also suggest that the ancestor of this group of plant was highly promiscuous, and that evolution has proceeded toward narrower promiscuity and greater specialization. These results imply that a tight relationship between the phytogenies of both symbiotic partners should not be expected, and that host promiscuity is likely to be a key determinant in the establishment of an effective symbiosis.


Introduction
The actinorhizal symbiosis derives from an association between the filamentous soil bacteria Frankiu (Actinomycetales) and an array of host woody dicots plants belonging to more than 20 genera in eight families ). The fixation of nitrogen resulting from this symbiosis is central to the dynamics of several ecosystems, many of the host plants being main components of early-successional communities established on poor and marginal sites (Benson & Silvester 1993). When considered all together, there is no close taxonomic affinity among the different actinorhizal plant families ). However, actinorhizal genera of the families Casuarinaceae, Myricaceae, and the single actinorhizal genus Alnus of the Betulaceae all involve hyphal penetration of deformed root hairs by Frankia followed by prenodule development and nodule lobe formation (Callaham eta!. 1979, Benson & Silvester 1993. These genera are also classified in the same coherent group of 'higher' Hamamelidae (Cronquist 1981(Cronquist , 1988. They are presumably closely linked to each other, together with the non-actinorhizal family Fagaceae (Takhtajan 1980;Cronquist 1988).
These various actinorhizal Hamamelidae taxa have been shown to differ drastically in their susceptibility to Frankia. The term 'promiscuity' has been coined to describe the plant's tolerance level to a range of genetically diverse Frankia strains (Baker 1987;Torrey & Racette 1989). For instance, within the Casuarinaceae, Gymnostoma has been shown to be susceptible to a wider array of Frankia strains (hence being more promiscuous) than Allocasuarina or Casuarina (Torrey & Racette 1989). Myrica, the largest genus of Myricaceae, has also been shown to be more promiscuous than either Alnus or Casuarina taxa (Baker 1987;Torrey & Racette 1989). Without a sound phylogenetic framework, it remains difficult to understand the ecological and evolutionary implications of these differences, which are central to a proper understanding of this symbiosis and its optimal utilization in field.
Recently, the chloroplast gene coding for the large subunit of the enzyme ribulose 15-bisphosphate carboxylase (rbcL) has been used to estimate the phylogeny of the family Betulaceae, and the results obtained were in complete agreement with the phylogeny estimated from morphological data (Bouquet et al. 1992b) and ribosomal DNA internal transcribed spacer sequences (Savard et al. 1993). rbcL gene sequences have also been used fruitfully to estimate phylogenetic relationships in various other goups of dicots (Soltis et al. 1990Giannasi etal. 1992;Chase etal. 1993). Using rbcL gene sequences, the purpose of this study was to clarify the phylogenetic relationships among actiorhizal families of 'higher' Hamamelidae in order to help understand the ecological and evolutionary implications of promiscuity differences observed among these taxa.

Plant materials
In addition to rbcL gene sequences already published (Table l), the complete nucleotide sequence of the rbcL ORF was determined for five taxa: Allocasuarina vwticillata (EMBL no. X69527), Casuarina cunninghamiana (EMBL no. X69528), Gymnostoma webbianum (EMBL no. X69531), Myrica gale (EMBL no. X69530), and Comptonia peregrina (EMBL no. X69529). Seeds from A. verticillata, C. cunninghamiana, and G. webbianun were germinated. After one month, a single branchlet from one individual of each species was used for DNA extraction. Lyophilized leaf tissues from M . gale and fresh leaves from C. peregrina were also used for DNA extraction. In all cases, DNA extractions were performed using a CTAB procedure (Bouquet et al. 1990).
DNA amplification and sequencing of rbcL rbcL was amplified symmetrically following previously published procedures (Bousquet ef af. 1992b). The primers used, including those upstream and downstream of rbcL, are described elsewhere (Frascaria et al. 1993). For each of the two DNA strands, a second asymmetrical amplification was conducted with a primer ratio of 1:50 where the primer in excess was at a concentration of 50 ~.LM and the limited primer at 1 p~. The asymmetrically amplified fragments were purified by ultrafiltration with Centricon-30 (Amicon). Direct sequencing of the two DNA strains was performed by the dideoxy-nucleotide chain-termination procedure using the Sequenase Version 2.0 Kit (USB). Sequencing was conducted using 7% polyaaylamide gels according to the manufacturer's recommendations (LR). X-ray films were exposed for 24 h to several days. Pairwise synonymous (K,) and nonsynonymous (K,) numbers of nucleotide substitutions corrected for multiple hits, and their standard errors, were calculated according to the two-parameter method of Li et al. (1985) modified by Li (1993). This method takes into account transition and transversion rates. Overall numbers of substitutions (KO) were calculated as a weighted average of K, and K,. Substitution rates were also estimated using the one-parameter method of Jukes and Cantor (1969).
One-parameter and KO pairwise substitution rates were analysed with the neighbour-joining method of phylogenetic tree construction (Saitou & Nei 1987). In addition, parsimony analysis of nucleotide sequences were conducted using the Branch and Bound algorithm of PAUP 3.1 (Swofford 1993

Evolutiona y study of promiscuity diflerences
Information regarding the promiscuity of actinorhizal Hamamelidae genera towards Frankia were regrouped in Table 2.' Height groups of Frunkiu strains were delineated on the basis of their host genus of origin. Only the Frunkiu strains shown to reinfect their host of origin were considered. A host genus was considered nodulated by a particular Frunkiu strain if one nodulation was induced on at least one species within the genus. The genus Comptoniu of Myricaceae was not considered because of insufficient data available on promiscuity. From Table 2, a cladistic character-state matrix was constructed where for each actinorhizal genus analysed, promiscuity towards each group of Frankia strains delineated was scored as positive (coded 1, all strains tested led to infection), or negative (coded 0, no infection), or polymorphic (coded 1/0, some ' Table 2 is shown overleaf. strains tested led to infection). For each group of Frankia strains (the cladistic characters), character-state changes were mapped on the consensus phylogenetic tree derived from the analysis of rbcL sequences of actinorhizal Hamamelidae taxa, using MACCLADE 3.1 (Maddisson & Maddisson 1992). Changes were assumed irreversible and two different analyses were performed, whether promiscuity toward each group of Frankiu strains was assumed derived or ancestral. Total number of steps and consistency index were monitored for each scenario.

rbcL sequences
The rbcL ORF sequence determined here for the three species of Casuarinaceae and the two species of Myricaceae was 1428 bp long. The degree of DNA homology was greater than 98Y0 within families of 'higher' Hamamelidae (Table 3). Between the families, the DNA homology was about 97%, whether the families belonged to the same order or to different orders. The synonymous rate was 10-20 times larger than the nonsynonymous rate, except for the Myricaceae (Table 3). Within 'higher' Hamamelidae, Jukes and Cantor's one-parameter rates were almost identical to the overall two-parameter rates (KO) calculated as weighted averages of K, and K, (Table 3).

Estimated phylogenies
Using two outgroups, Magnolia mucrophylla and Liquidambar styracpua, standard parsimony led to one most parsimonious tree with a total length of 273 steps (Fig. 1A). Of these, 153 were accounted for by the internal network linking the 'higher' Hamamelidae taxa. The consistency index (CI) excluding uninformative characters was 0.778 when considering the most parsimonious network linking the 'higher' Hamamelidae taxa. The topology obtained from neighbour-joining analysis of one-parameter substitution rates ( Fig. 18) was identical to the one obtained using two-parameter overall numbers of Abbreviations used: n, number of pairwise sequence comparisons involved; K , K , and K , synonymous, nonsynonymous, and overall numbers of substitutions per site, respectively; J.C., Jukes and Cantor's number Lf s;bstitut$ns per site.  DDB16060820  IMB16092115  IAE16248302  ULQ00231058  ULQ132500106  SIB13010118  IAE13310107  IMB13312061  IMB13271510  SIB1 3320273  SIB13320373  1AE13131211  IAE13360085  SIB13250131  IAE14010016  IAE14010021  lAE14010034  IAE14010037  SIB14010104  ULQ130100144  ULI13270210  ULI13270237  ULI13270250  5 Catalog numbers as defined by Lechevalier (1985).

100
Compronia strap value higher than 50% (Fig. 1B) (the alternative bootstrap for the Fagales using this method was only So/ , ). Standard parsimony analysis led to a regrouping of the families Casuarinaceae and Fagaceae with a bootstrap value (43%) higher than that obtained for the Fagales (15%). Overall, the divergence between the Fagaceae, the Betulaceae, and the Casuarinaceae appeared essentially as a trichotomy. Neighbour-joining and parsimony analyses both indicated the Myricaceae to have diverged first among the families of 'higher' Hamamelidae analysed (Fig. 1A,B). Any other topologies placing the Fagaceae, Betulaceae, or Casuarinaceae as first group to diverge among the 'higher' Hamamelidae received bootstraps lower than 50%. In the family Casuarinaceae, both neighbour-joining and parsimony analyses indicated a subclade which contained Allocnsuarina and Casunrinn (the Casuarineae) (Fig.  1A,B). As previously reported, the Betulaceae were divided into two subclades, the tribe Betuleae containing the genera Alms and Betula and the tribe Coryleae containing the remaining genera (Bousquet rt al. 1992b).

Mognobo
The promiscuity of the each actinorhizal genus of the 'higher' lid^^ was scored successively for group of Frnnkin strains (Fig. Z), based on available data presented in Table 2. Character-state changes, that is, the promiscuity differences among host genera relative to the Fig. 1 (a) Tree obtained from parsimony analysis of rbcL nucleotide sequences; (b) tree obtained from neighbor-joining analysis of rhcL sequences using the one-parameter method. Numbers on nodes indicate bootstrap cstirnates from 1000 replications. substitutions (KO), and highly congruent with results from parsimony analysis.
In both parsimony and neighbour-joining trees, the different families were supported by high bootstraps (Fig. 1A,B). The order Fagales, which contains the families Betulaceae and Fagaceae, did not appear to form a coherent group from the different phylogenetic analyses conducted. The neighbour-joining analysis of substitution rates led to a regrouping of the families Casuarinaceae and Betulaceae, which was supported by a boot-various groups of Frnnkin strains, were then monitored over the rbcL consensus tree of actinorhizal plant taxa.
Compfunin could not be included in the analysis because of insufficient data on promiscuity (see Table 2). A strong relationship could be observed between the phylogeny of actinorhizal Hamamelidae and their spectrum of promiscuity towards the various groups of Frankin strains (Fig.2). Myricn, assumed to diverge first, is equally or more promiscuous than the other genera. Within the Casuarinaceae, Cymnostornn, assumed to diverge first, is more promiscuous than the other genera of the family.  Table 2) and analysis of rhararter-state changes in promiscuity of host-genera along the rbcL consensus tree of actinorhizal genera. The host genus Conrptotzio was not included in the consensus tree because of insufficient data regarding its promiscuity. Names of host genera a t the top of columns of the character-state matrix correspond to those at the tip of branches of the consensu tree. On the right, numbers of steps corresponds to the numbers of reversals in promiscuity along the branches of the tree, as produced by each group of Fronkia strains on the left and the associated array of promiscuity differences.  Moreover, the Casuarinae group (Allocasuarina and Casuarina) appears to be restricted in its symbiotic association to a group of Frankia strains not uninfective on Ahus roots, while Alms can only be infected by strains not infective on Casuarinae roots ( Table 2 and Fig. 2

).
When promiscuity was assumed ancestral, the number of steps necessary to explain the observed character-state variation among actinorhizal Hamamelidae was much smaller than when promiscuity was assumed derived (Fig. 2).

Molecular phylogeny of acfinorhizal Hamamelidae taxa
The different families analysed appeared as natural entities through the different phylogenetic analyses conducted, but the order Fagales was not observed as a coherent group. This observation is congruent with those of Nixon (1989), who suggested a paraphyletic or polyphyletic origin of the order Fagales, based on morphological characters. However, the closer relationship we observed between the Fagales (Betulaceae and Fagaceae) and the Casuarinales (Casuarinaceae), than with the Myricales (Myricaceae), is in agreement with views proposed by Takhtajan (1980) andConquist (1988). This is in contrast to other views considering the Casuarinaceae more primitive and classifying the family as evolutionary intermediate between the 'lower' Hamamelidae and other more specialized 'higher' Hamamelidae such as the Betulaceae, Fagaceae, and Myricaceae (Barabe et at. 1982). Part of the disagreement could be explained by the interpretation of morphological variation. As indicated by Cronquist (1988), Casuarinaceae flowers are reduced rather than primitively simple, and such a reduced aspect in morphological characters should not necessarily be l i k e d to a primitive state, because of potential adaptation to environmental conditions. The likely divergence of the Myricaceae before the Betulaceae (Alnus) and Casuarinaceae is in agreement with the fossil record where the family Myricaceae appeared in the Cenomanian stage of the Upper Cretaceous (Gladkova 1962;MacDonald 1977), before the Betulaceae and Casuarinaceae, which appeared later in the Coniacian-Santonian stage of the Upper Cretaceous (Cronquist 1988;Crage 1989) and at the limit between the Cretaceous and the Tertiary (Johnson & Wilson 1989), respectively. At lower taxonomic level, the topology of the Casuarinaceae obtained from rbcL sequences is in complete agreement with the treatment of the family by Johnson & Wilson (1989), which was based on morphological and chromosomal characters. Gymnostoma would have first diverged, followed by Casuarina and Allocasuarina which, in our analyses, were regrouped (Casuarineae clade).
This topology is also confirmed by the fossil record, where Gymnostoma is much postdated by the other genera of the family (Barlow 1983).

The evolution of promiscuity
Our survey of infectivity tests showed that Myrica and Gymnostoma can be infected by all host groups of Frankia strains tested, unlike other actinorhizal genera of 'higher' Hamamelidae which were found to be more specific in their Frankia symbiotic association. Because both molecular and traditional approaches indicate an earlier divergence of the family Myricaceae relative to the Betulaceae and Casuarinaceae, and an earlier divergence of Gymnostoma relative to Casuarina and Allocasuarina within the Casuarinaceae, it appears that evolution has proceeded in the direction of narrower promiscuity. This is also supported by the smaller number of state changes in promiscuity along the consensus topology of actinorhizal taxa, when promiscuity was assumed to be ancestral. Therefore, a narrower promiscuity could be interpreted as a more specialized feature, with possible evolution towards total resistance to Frankia infection, such as observed quite frequently for Allocasuarina under field conditions aohnson & Wilson 1989).
This evolutionary trend is also observed in the Betulaceae, even though it contains only a single actinorhizal genus, Alnus. The Betuleae clade, which contains the actinorhizal genus Alnus and the non-actinorhizal genus Betula, has been shown to be less morphologically advanced that the Corylear clade (Crane 1989;Bousquet et al. 1992b), which contains only nonactinorhizal genera. This suggests the actinorhizal state typical of Alnus to be less advanced than the nonactinorhizal state. Furthermore, the most ancient Betulaceae fossils were pollen grains typical of the actinorhizal A h u s (Crane 1989), with pollen typical of other genera, particularly from the Coryleae clade, appearing later in the fossil record. Moreover, the non-actinorhizal genus Betula is found more closely related to Alnus than to other non-actinorhizal genera of the Coryleae clade (Bousquet et al. 1992b;Savard et al. 1993), and this is paralleled by evidence that Betula is more dependent on Frankia that taxa from the morphologically advanced Coryleae clade, again suggesting the association with Frankia to be less advanced. Indeed, higher densities of Frankia of more than one order of magnitude have been observed in soils under stands of different Betula species, as compared to stands of spruce, oak (a Fagaceae), or even Corylus (Coryleae clade) (Van Dijk 1984;Smolander 1990;Paschke & Dawson 1992). Under such Betula stands, it has further been shown that Frankia could survive as an associative nitrogen fixer in a loose, unspecific but beneficial relationship with the nonhost plant Befula roots (Ronkko et al. 1993).

A C T I N O R H I Z A L H A M A M E L I D A E A N D Frniikia 465
Ecological and evolutionary implications for the actinorhizal symbiosis This evolutionary trend toward a narrower promiscuity of the host plants, suggests that the host plant root system plays a key role in the specific relationships established with Frankia. Not only the array of Frankiu strains capable of infecting a host root system seems to be controlled by the plant, but also other types of characters involved in the symbiosis such as vesicle formation and morphology (Lalonde 1979;Tjepkema et al. 1980;Benson & Silvester 1993), or the mechanism of Frankia penetration into host roots: in this case, Frankia strains able to infect both Elaeagnaceae (subclass Rosidae of the dicots) and some actinorhizal Hamamelidae species show differential type of root penetration depending of the host, either by root hair infection for the Hamamelidae taxa, or by intracellular penetration for the Elaeagnaceae (Miller & Baker 1986;Racette & Torrey 1989b).
Furthermore, variation in the efficiency of the symbiosis has been shown to be much more controlled by plant clonal effects than Frankia strain effects (Simon, et al. 1985;Mackay et al. 1987;Prat 1989;'Sougoufara et al. 1992). This might simply derives from differences in plant genotypes with respect to promiscuity towards compatible Frankia strains, which would result in efficiency differences. Such larger control of the symbiotic efficiency by the host plant has also been reported in the legumes (see Galiana et af. 1991) and in free nitrogen-fixation associations involving various bacteria and nonhost plants (Ronkko et al. 1993). Therefore, for the practical use of the actinorhizal symbiosis in the field where an efficient system is required, selection of the optimal host genotype-Frankiu combination should exploit knowledge of the promiscuity of the plant genotypes used. This would imply a larger emphasis on the number of host genotypes tested rather than the number of Frankia strains involved.
It is now generally agreed that taxonomic grouping of Frunkia strains from root nodules is much more dictated by the host plant from which the strains were isolated, rather than by the spectrum of infectivity of the strains (Lalonde et al. 1988;Fernandez et ul. 1989;Beyazova & Lechevalier 1992). While this observation should be considered as a direct effect of host promiscuity, it is also likely that Frunkia strains naturally found on promiscdow plant taxa or genera would exhibit more genetic or taxonomic diversity than strains naturally found on less promiscuous plant taxa or genera. Indeed, for Frankia strains naturally found on actinorhizal Hamamelidae taxa and which could reinfect their host of origin, this relationship between host promiscuity and strain diversity seems to hold. It is supported by the apparently narrower genetic diversity of Frankia strains isolated from Casuarina and Allocasuarina, as compared to strains isolated from the more promiscuous Ahus (Femandez et al. 1989;Nazaret et ul. 1991). It is also supported by the apparently much larger biochemical and genetic diversity of Frankia strains isolated from the largely promiscuous genus Myricu, which usually fail to form coherent taxonomic groups (Gardes et al. 1987;Bloom et al. 1989;Simon rt al. 1989).
As a consequence of these promiscuity differences among actinorhizal plants, the stringency of coevolutionary relationships between Frankia and the plant host should vary extensively, and a tight relationship between the phylogenies of both partners involved in the symbiosis should not be expected. Such a scenario is also observed in the legume-Rhizobium symbiosis, where little correspondence is observed between the phylogenies of the symbiont and the host (Young &Johnston 1989). This is in contrast with other mutualistic or parasitic systems where much tighter relationships are often observed between the phylogenies of both partners (Mitter et ul. 1991; Moran & Baumann 1994). As for the 1egmeRhizobium symbiosis, differences in promiscuity among actinorhizal Hamamelidae and total resistance of closely related taxa to root invasion by Frunkia should reflect various levels of long-term mutualistic and antagonistic interactions between the two partners (Young & Johnston 1989). Because Frankia is not an obligate symbiont, being able like Rhizobium to survive as a free-living organism in the soil, the lack of coevolutionary relationships should even be more apparent with rhizospheric Frankin, where the plant host effects on Frankiu taxonomic and genetic diversity would be much relaxed, if not negligible.