Nodulation and nitrogen fixation by Mimosa spp. in the Cerrado and Caatinga biomes of Brazil


Author for correspondence:
E. K. James
Tel: +44 1382 560017


  • An extensive survey of nodulation in the legume genus Mimosa was undertaken in two major biomes in Brazil, the Cerrado and the Caatinga, in both of which there are high degrees of endemicity of the genus.
  • Nodules were collected from 67 of the 70 Mimosa spp. found. Thirteen of the species were newly reported as nodulating. Nodules were examined by light and electron microscopy, and all except for M. gatesiae had a structure typical of effective Mimosa nodules. The endosymbiotic bacteria in nodules from all of the Mimosa spp. were identified as Burkholderia via immunolabelling with an antibody against Burkholderia phymatum STM815.
  • Twenty of the 23 Mimosa nodules tested were shown to contain nitrogenase by immunolabelling with an antibody to the nitrogenase Fe- (nifH) protein, and using the δ15N (15N natural abundance) technique, contributions by biological N2 fixation of up to 60% of total plant N were calculated for Caatinga Mimosa spp.
  • It is concluded that nodulation in Mimosa is a generic character, and that the preferred symbionts of Brazilian species are Burkholderia. This is the first study to demonstrate N2 fixation by beta-rhizobial symbioses in the field.


Until recently, ‘rhizobia’ were considered to consist of a limited number of genera in the family Rhizobiales in the Alpha-Proteobacterieaceae (Graham, 2008), but evidence has accumulated to show that legumes, particularly those in the genus Mimosa, but also other mimosoids and some papilionoids, such as Cyclopia (Elliott et al., 2007a) and Rhynchosia (Garau et al., 2009), may also form effective nodules with beta-proteobacteria, the so-called ‘beta-rhizobia’ (Elliott et al., 2007b). Although beta-rhizobia have been isolated in many tropical regions from invasive Mimosa species, such as M. diplotricha (often cited as M. invisa), M. pigra and M. pudica (Chen et al., 2001, 2005b; Verma et al., 2004; Barrett & Parker, 2005, 2006; Parker et al., 2007), and selected strains of beta-rhizobia, such as Burkholderia phymatum STM815, Burkholderia mimosarum PAS44 and Cupriavidus taiwanensis (syn. Ralstonia taiwanensis) LMG19424, have shown a particular ability to nodulate Mimosa spp. (Elliott et al., 2007b, 2009), little is known about the symbionts of Mimosa species in their native environments.

Mimosa is a large and complex genus with over 500 species and is mainly native to the New World (Barneby, 1991; Simon & Proença, 2000). Species vary in habit from tall trees and shrubs to vines and herbs, and they are found in a wide variety of habitats from wet to dry, growing on many different soils. Mimosa is considered by Barneby (1991) to have ‘differentiated profusely in tropical and warm temperate savanna habitats’, but is particularly abundant and diverse in the Cerrado and Caatinga biomes of Brazil (Lewis, 1987; Barneby, 1991; Simon & Proença, 2000; Mendonça et al., 2008; de Queiroz, 2009). The Cerrado has c. 200 Mimosa species, most of them endemic to either the Cerrado biome as a whole or to very specific localities within it (Simon & Proença, 2000). Similarly, the semi-arid Caatinga biome in the northeast of Brazil is home to 38 species of Mimosa (de Queiroz, 2009), plus 15 species that grow in the highlands of the Chapada Diamantina (Lewis, 1987; de Queiroz, 2009). As with the Cerrado, many of these species are endemic.

The present study aimed to:

  • determine the extent of nodulation of Mimosa species in the Cerrado and Caatinga. Additional collections were made in the Pantanal, a wetland biome neighbouring the Cerrado, in which nodulated Mimosa spp. have previously been recorded (James et al., 2001);
  • identify the microsymbionts ‘in situ’ using antibodies specific either to the genus Burkholderia or to the species C. taiwanensis;
  • determine whether nodulated species fix nitrogen in the field by immunolabelling nodules with an antibody against the nitrogenase nifH (Fe-) protein, and by applying the δ15N (15N natural abundance) technique to Mimosa spp. using neighbouring nonnodulated plants as references.

Materials and Methods

Cerrado and Caatinga: vegetation types, soils and precipitation

The collection sites are indicated in Fig. 1 and Supporting Information, Table S1. The Cerrado biome covers an area of 2 × 106 km2 (Ratter et al., 1997). Its soils are dystrophic, with low cation exchange capacity, and acidic with pH values of c. 4.0–5.5 (Adámoli et al., 1986), and they support a mainly savannah vegetation, dominated by grasses, which is frequently subjected to natural burning (Felfili et al., 2004; Simon et al., 2009). The Caatinga biome, which has an area of 850 000 km2, is typified by Fe-rich soils and xeric scrub vegetation (de Queiroz, 2009). Both biomes contain areas of campo rupestre, which is a type of vegetation (mainly herbaceous with shrubs and small trees) that occurs above 900 m in litholic soils that are acidic and low in fertility (Ribeiro & Walter, 2008). Precipitation in the Cerrado is highly seasonal, but generally higher than 1000 mm yr−1, whereas that in the Caatinga is < 800 mm yr−1 (classified as semi-arid) and is concentrated in only a very short and sometimes unpredictable wet season of 2–3 months’ duration. Precipitation in the high-altitude campo rupestre is generally higher than in the surrounding Cerrado or Caatinga, but the rocky soils within them have low water retention capacity (Ribeiro & Walter, 2008).

Figure 1.

 The locations of the Mimosa nodule-collecting sites in the various biomes of central Brazil. Two-letter codes are Brazilian states: BA, Bahia; GO, Goiás; MT, Mato Grosso; .

Collection of plant material

Species of Mimosa were collected from various locations (mainly highland regions) in the Cerrado in September 2005 (dry season) and in the Cerrado, Caatinga and Pantanal in March 2006 (wet season). Additional collections were made in the Cerrado in Minas Gerais and in the Caatinga in Bahia during the wet season of 2009 (Fig. 1, Table S1). Voucher herbarium specimens were taken for all species and deposited in UB, HUEFS, BHCB and FHO (herbarium abbreviations follow Holmgren et al., 1990). In both expeditions, nodules (if present) were collected and preserved in silica gel for later bacterial isolation. Some nodules (three to four per plant) were also cut in half to determine if they were potentially active and effective by the appearance of a pink coloration resulting from the presence of leghaemoglobin (Lb), and these were then placed into vials containing 2.5% glutaraldehyde in 50 mM phosphate buffer (pH 7.5) for microscopic analysis. Since many of the plants in the September 2005 dry season expedition only had dried nodules on their roots, these were used solely for bacterial isolation (see Bontemps et al., 2010); nodules sampled in the March 2006 wet season were generally sufficiently fresh for both bacterial isolation and microscopy. Mimosa seeds were collected whenever they were found, but this was generally restricted to the dry season. Leaf material was collected in the wet season for δ15N analysis (see the section on δ15N analysis for details), and soil samples were collected in the Caatinga for pH analysis.

Microscopy and in situ detection of microsymbionts

All pink nodules collected in the field were prepared and sectioned for light microscopy to determine general nodule structure, and then were further analysed by in situ immunogold labelling (plus silver enhancement) using antibodies raised against B. phymatum STM815T and C. taiwanensis LMG19424T according to Chen et al. (2005b) and Elliott et al. (2007b). The cross-reactions of the antibodies against a wide range of bacteria, including both alpha- and beta-proteobacteria (Table S2), were first tested with the enzyme-linked immunosorbent assay (ELISA) according to Gyaneshwar et al. (2001). Two preliminary experiments established the ability of the antibodies to detect defined Burkholderia and Cupriavidus strains within Mimosa nodules.

Seedlings of various Mimosa spp. collected in the Cerrado and Caatinga were inoculated with either B. phymatum STM815 or C. taiwanensis LMG19424 according to Elliott et al. (2007b).

Seedlings of M. setosissima and M. decorticans, both Cerrado endemics, and M. blanchetii, a campo rupestre endemic from Chap. Diamantina, were inoculated, respectively, with rhizobial strains JPY164, JPY297 and JPY578, which had originally been isolated from dry nodules of these hosts, and have been identified via their 16S rDNA and recA sequences as belonging to the genus Burkholderia (Bontemps et al., 2010).

In both experiments, the plants were harvested at 6 wk after inoculation, tested for nitrogenase activity via acetylene reduction assays (ARAs) according to Chen et al. (2003), and any nodules formed were sampled for microscopy and immunogold labelling as for the field samples.

Pre-immune sera were used as negative controls in all of the immunogold assays. The reactions of the test samples were compared visually with the corresponding negative and positive controls to determine which of the antibodies had reacted with each nodule, and these were then scored as being positive for either B. phymatum (‘Bp’) or C. taiwanensis (‘Ct’), or giving no reaction with either antibody (‘NR’).

Western blots and in situ detection of nitrogenase nifH protein

A polyclonal antibody raised against the Fe-protein (nifH) of the nitrogenase enzyme complex from Rhodospirillum rubrum (Norén et al., 1997) was used to detect the presence of nitrogenase in some of the nodules collected from the Cerrado and the Caatinga according to Chen et al. (2003). The specificity of the nifH antibody was tested by Western blots of nodule extracts of M. pudica infected with either B. phymatum STM815T or C. taiwanensis LMG19424T (Elliott et al., 2007b) according to Rubio et al. (2009). The nifH antibody was used at a dilution of 1 : 5000, and extracts of nodules from alfalfa (Medicago sativa) infected with Sinorhizobium meliloti 102F78 were used as positive controls. The ability of the nifH antibody to detect the nitrogenase enzyme via immunogold silver enhancement in Mimosa nodules collected in the field was first tested on sections of nodules of Cerrado species inoculated with B. phymatum STM815T or C. taiwanensis LMG19424T that were positive or negative for nitrogenase (acetylene reduction) activity, as well as on nodules of M. setosissima (+ Burkholderia strain JPY164) and M. blanchetii (+ Burkholderia strain JPY578) from actively N2-fixing plants (see previous section). In addition to immunogold silver enhancement for light microscopy, more detailed analyses using transmission electron microscopy (TEM) with both the B. phymatum and nifH antibodies were performed on nodules from selected field-collected Mimosa spp. according to Elliott et al. (2007b) and Chen et al. (2005a). Nonimmune serum was used as a negative control in all immunogold assays.

δ15N analysis

Leaf samples were taken from Mimosa spp. for analysis of their δ15N content in the 2006 wet season expedition (in which nodules were more likely to be actively fixing N2) according to Sprent et al. (1996). Samples were only taken from sites that were apparently undisturbed (e.g. pristine Cerrado, Caatinga or campo rupestre), and not from agricultural land or roadsides where the δ15N signal can be greatly altered by human activities (Unkovich et al., 2008). Sites were effectively delineated by the presence of at least three individuals of each endemic or biome-restricted Mimosa sp. within 5 m of each other, and leaves were taken from three separate nodulated plants at each site. Where available (i.e. within a 1 m radius of the test plant, but not immediately adjacent to it), samples were also taken from three replicate plants of three nonnodulated dicotyledonous reference species (preferably nonnodulating legumes, if present), as recommended by Boddey et al. (2000). Samples of test and reference plants were dried, ground to a fine powder, and their δ15N enrichment was determined using the methodology of Teixeira et al. (2006). A ‘B-value’ for a defined Mimosa symbiosis was obtained according to Unkovich et al. (2008) by growing seedlings of M. caesalpiniifolia inoculated with B. sabiae strain Br3407 (Chen et al., 2008) in a mixture of vermiculite/perlite without any added N. The plants were harvested at 6 months and were assessed for the presence of root nodules and nitrogenase activity using ARA. The aerial parts of those plants with active N2-fixing nodules were dried and ground and their δ15N content determined. Aerial parts had a mean (± SEM, = 6) δ15N enrichment of −1.24 ± 0.22‰, which was within the range of reported B-values for other (albeit nonwoody) legumes (Unkovich et al., 2008), and thus it was used to determine the proportion of N derived from air (%Ndfa) according to the equations given by Unkovich et al. (2008).


Nodulation of Mimosa species in the Cerrado and Caatinga

Seventy species of Mimosa were examined for nodulation in the Cerrado and Caatinga (Tables S3, S4). Of these, 59 were shrubs or small trees (2–3 m in height), such as M. decorticans (Fig. S1a), M. splendida (Fig. S1b) and M. crumenarioides (Fig. S1c), five were herbs (M. debilis, M. lewisii, M. pudica, M. skinneri, M. ursina; Fig. 2d), and six were trees (M. acutistipula, M. caesalpiniifolia, M. dominarum, M. hexandra, M. ophthalmocentra, M. tenuiflora). The species were categorized according to Simon & Proença (2000) as either ‘endemic’ (E), ‘biome-restricted’ (R) or ‘widespread’ (W) (i.e. occurring in more than one biome). The majority of the endemic and biome-restricted species (29 in the Cerrado, nine in the Caatinga) were found at high altitude (> 800 m) mainly in pristine vegetation (often growing in campo rupestre on rocky outcrops at elevations above 1000 m; Fig. S1a–c), whereas the widespread species were generally found at lower altitudes alongside roads or on disturbed ground close to roads (Fig. S1d), and sometimes in urban areas (Tables S3, S4). In total, nodules were found on 67 separate species (84 taxa); 44 of these had not previously been reported as nodulating before these expeditions to the Cerrado and the Caatinga, but as effective rhizobia were isolated from 31 of them in our parallel study from the same expeditions (Bontemps et al., 2010; see Tables S3, S4), the present study has completely new reports on nodulation by the following 13 species: M. calodendron, M. campicola, M. dominarum, M. gatesiae, M. irrigua, M. pseudosepiaria, M. pyrenea, M. regina, M. setuligera, M. skinneri, M. speciosissima, M. ulbrichiana and M. verecunda.

Figure 2.

 (a) Light microscopy of a section of a N2-fixing nodule formed on the Cerrado endemic, Mimosa setosissima, after inoculation with a Burkholderia strain (JPY164) which was originally isolated from this species. Note that it has a structure typical of an effective nodule, with an apical meristem and a large infected zone which contains the N2-fixing symbionts. (b–d) Light microscopy of sections of nodules collected in the Cerrado and the Caatinga biomes. Nodules from Mimosa callithrix (b), endemic to Chapada dos Guimarães in the western Cerrado (Mato Grosso); Mimosa cordistipula (c), an endemic of campo rupestre in Chapada Diamantina; and Mimosa pseudosepiaria (d), an endemic of the Caatinga. Arrows indicate thick cortex in M. setosissima (a) and M. pseudosepiaria (d). *, N2-fixing infected zone; m, meristem. Bars, 200 μm.

Soil samples taken from the rhizosphere of the Mimosa spp. confirmed the characteristic low pH of the Cerrado soils; for example, those near Brasilia had a pH of 4.9, whereas those from the western Cerrado (Chap. dos Guimarães) had an average pH of 5.1. The pH of the Mimosa rhizosphere soils in Chap. Diamantina (Caatinga) were similarly very low, with an average pH of 4.7.

Nodule structure and in situ identification of microsymbionts

As only fresh nodules could be analysed using microscopy, most of the nodules examined were those collected in the 2006 and 2009 wet seasons. The only exceptions from the dry season expedition were the nodules on M. decorticans, and those on some widespread species in Chap. dos Guimarães (Table S3). The majority of the nodules on all the species examined from both biomes (49 species in total) were found to be effective in appearance when examined under the light microscope (Tables S3, S4, Fig. 2). To ascertain whether nodules collected in the field could be regarded as likely to be ‘effective’ in terms of N2 fixation they were compared anatomically with actively N2-fixing (i.e. ARA positive) nodules on Cerrado/Caatinga species that had been grown under controlled (N-free) conditions after inoculation of seedlings with strains of Burkholderia isolated from nodules collected in the field (Bontemps et al., 2010). The example shown in Fig. 2(a) is from an effective nodule on M. setosissima, a species endemic to Serra dos Pirineus in Goiás that was inoculated with the Burkholderia strain JPY164. This can be compared with nodules on field-collected nodules on the endemics M. callithrix (Fig. 2b), M. cordistipula (Fig. 2c) and M. pseudosepiaria (Fig. 2d). In all cases the field-collected nodules have the typical appearance of effective Mimosa nodules with an apical meristem, and an elongated infected zone containing enlarged host cells packed with N2-fixing symbiotic bacteroids (cf. Chen et al., 2005a; Elliott et al., 2007b). In some cases (but not all, e.g. Fig. 2b,c), such as M. setosissima (Fig. 2a) and M. pseudosepiaria (Fig. 2d), the nodules had thick cortices with layers of tannin-containing ‘corky’ cells external to the endodermis which may help the nodules to retain moisture in a semi-arid or seasonally dry environment (Brown & Walsh, 1994; Gross et al., 2002). Mimosa gatesiae from the Cerrado was the only species examined by microscopy in which all the nodules examined appeared to be senescent, although ineffective/senescent nodules were also observed on some plants of M. adenocarpa and M. lewisii (Table S4).

An attempt was then made to identify the symbionts in all of the nodules prepared for microscopy. This involved immunolabelling them with antibodies raised against two common Mimosa symbionts, B. phymatum STM815T and C. taiwanensis LMG19424T (Chen et al., 2005b; Elliott et al., 2007b). It was first necessary, however, to check the specificity of the antibodies, and an ELISA confirmed that the B. phymatum antibody strongly recognized B. phymatum STM815T, as well as all the other Burkholderia strains that were tested, including legume-nodulating strains, plant-associated diazotrophs and environmental isolates (Table S1), but it did not react with other bacteria, including Cupriavidus, Ralstonia and Rhizobium strains. The C. taiwanensis antibody was more specific, reacting only with C. taiwanensis strains, and giving no cross-reaction even with other Cupriavidus or Ralstonia species. Preliminary immunolabelling tests showed that the antibodies could detect B. phymatum and C. taiwanensis within Cerrado and Caatinga Mimosa hosts (Table 1), but also that they did not cross-react with nodules containing the ‘alternative’ strain (see Elliott et al., 2007b for more details). In a second preliminary experiment, the B. phymatum antibody could detect Burkholderia within symbiotic nodules from Cerrado and Caatinga endemic Mimosa species inoculated with strains originally isolated from them (Bontemps et al., 2010), for example, M. blanchetii inoculated with JPY578 (Fig. 3a), but also M. decorticans inoculated with JPY297 and M. setosissima inoculated with JPY164 (not shown). Sections of nodules treated with pre-immune serum gave no reaction (Fig. 3b). Finally, when the field-collected nodules were examined, it was shown that almost all of them (i.e. 118 of the 126 Mimosa nodules from 49 species) reacted positively with the B. phymatum antibody, with the antibody reaction coinciding with the infected, N2-fixing zone in each case (Tables S3, S4, Fig. 3c,d). There was no reaction with the pre-immune serum negative controls that were run concurrently with each sample. These results were confirmed under the TEM for selected samples, in which it was shown that the B. phymatum antibody mainly recognized antigens on the bacteroid surfaces (Fig. 3e,f). No nodules reacted with the antibody against C. taiwanensis, and the eight samples that did not react with either antibody included one nodule from M. misera (Caatinga), one from M. verecunda (Chap. dos Veadeiros), and six nodules from three separate plants of the widespread species M. xanthocentra (Chap. dos Guimarães). It should be noted, however, that all the other samples of M. cordistipula, M. verecunda and M. xanthocentra that were tested reacted positively with the B. phymatum antibody (Table S4).

Table 1.   Acetylene reduction activity and nodule structure of Mimosa species from the Cerrado and the Caatinga inoculated with Burkholderia phymatum STM815 or Cupriavidus taiwanensis LMG19424
 B. phymatum STM815C. taiwanensis LMG19424
ARA (nmol C2H4 per plant h−1)Nodule structureaARA (nmol C2H4 per plant h−1)Nodule structure
  1. Nodules from all species/strain combinations were confirmed to contain the strain with which they were inoculated using antibodies against B. phymatum STM815 or C. taiwanensis LMG19424 (Elliott et al., 2007b). E, endemic to the Cerrado; E*, endemic to the Caatinga; R, restricted to the Cerrado; W, widespread (Simon & Proença, 2000).

  2. aEff, effective in appearance; In, ineffective in appearance.

M. albolanata (R)0–34Eff0In
M. callithrix (E)0In0In
M. claussenii (R)0In0In
M. cordistipula (E*)10Eff0In
M. debilis (W)1562–44640Eff0–10Eff
M. decorticans (E)0In0In
M. densa (R)0In0In
M. foliolosa (R)0–106Eff0In
M. melanocarpa (R)0In0In
M. setosa (W)551–5554Eff0In
M. setosissima (E)0No nodules0No nodules
M. ursina (W)250–1399Eff0In
M. velloziana (W)271–1232Eff0In
M. xanthocentra (W)94–1489Eff0–41Eff
Figure 3.

 Light microscopy (a–d) and transmission electron microscopy (TEM) (e, f) of sections of Mimosa nodules immunogold-labelled either with an antibody against Burkholderia phymatum STM815 (a, c–f) or with pre-immune serum (b). (a, b) Nodules from M. blanchetii inoculated with Burkholderia sp. JPY578. Note the strong reaction with the STM815 (a, c–f) or with pre-immune serum (b). (c, d) Light microscopy of nodules from the Cerrado endemic M. manidea (c) and the widespread species M. xanthocentra var. subsericea (d). (e, f) TEM of the Cerrado endemic M. regina (e) and the Caatinga endemic M. cordistipula (f). Note that the surfaces of the bacteroids in both nodules are strongly labelled (arrows), but not the bacteroid cytoplasm. *, N2-fixing infected zone; b, bacteroids; m, meristem; s, senescent zone. Bars, 200 μm (a–d); 500 nm (e, f).

Nodulation of Cerrado/Caatinga species with common beta-rhizobial strains

The ‘standard’Mimosa-nodulating beta-rhizobial strains, B. phymatum STM815T and C. taiwanensis LMG19424T, nodulated the five widespread species (M. debilis, M. setosa, M. ursina, M. velloziana, M. xanthocentra) highly effectively, but nodulated Cerrado-restricted species either much less effectively (M. albolanata, M. foliolosa) or ineffectively (M. claussenii, M. densa, M. melanocarpa), and the Cerrado endemic species either ineffectively (M. callithrix, M. decorticans) or not at all (M. setosissima) (Table 1). By contrast, C. taiwanensis LMG19424T could only form slightly effective nodules on M. debilis and M. xanthocentra, and ineffective nodules on all the other species, except for M. setosissima, which it could not nodulate. No Caatinga endemics were tested, but Burkholderia phymatum STM815T, but not C. taiwanensis LMG19424T, was also capable of forming slightly effective nodules on M. cordistipula, the only Caatinga endemic tested (Table 1).

In situ immunolocalization of nifH protein in field-collected nodules

Nodules from 23 species were examined for nitrogenase expression by immunogold labelling with an antibody raised against the Fe-protein (nifH protein) of nitrogenase (Table 2). When possible, the plants from which the nodules were sampled were the same as those from which leaves were collected for δ15N analysis (Table 3). The specificity of the nifH protein antibody was demonstrated in Western blots of extracts of M. pudica nodules that were infected with either of the ‘standard’ beta-rhizobial strains, B. phymatum STM815T and C. taiwanensis LMG19424T (Fig. 4). The symbioses formed by M. pudica with either of these strains are known to be highly effective in terms of N2 fixation (Elliott et al., 2007b). The nifH antibody recognized a strong single immunoreactive band of an approximate molecular mass of 36 kDa in both kinds of nodules and also a single band of 37 kDa in alfalfa nodules used as positive controls. In preliminary tests on sections of nodules from plants grown under controlled-environment conditions and which had tested positive for nitrogenase activity, nodules from M. debilis and M. ursina inoculated with B. phymatum STM815T (Table 1), M. blanchetii plants inoculated with Burkholderia sp. JPY578 (Table 2) and M. setosissima plants inoculated with Burkholderia sp. JPY164 (Table 2) all gave a positive immunogold reaction with the antibody (Fig. 5a), whereas nodules from nonN2-fixing associations (e.g. M. himalayana inoculated with C. taiwanensis LMG19424T; Elliott et al., 2007b) gave no signal (not shown, but similar to Fig. 3b), thus demonstrating that the antibody could be used to probe nodules collected in the field. Fifteen of the 17 Cerrado species and five of the six Caatinga species gave a positive signal with the nifH protein antibody (Table 2), and the reaction coincided only with regions of the infected zone that were nonsenescent (Fig. 5b,c). The specificity of the antibody only to bacteroids was confirmed under the TEM for M. caesalpiniifolia (Fig. 5d,e) and M. regina (both from the Cerrado), and for M. cordistipula (from campo rupestre in Chap. Diamantina) (Fig. 5f) and M. ulbrichiana (from the Caatinga) (not shown). Interestingly, the three samples that were negative (M. adenocarpa, M. gatesiae and M. lewisii) had previously been identified as having nodules that were senescent and/or ineffective in appearance (Table S4).

Table 2. In situ immunogold reaction of nodules from Cerrado and Caatinga Mimosa species with an antibody against the nifH protein (Fe-protein) of the nitrogenase enzyme complex
SpeciesnifH antibody reaction
  1. aAll nodules examined had previously been identified as being effective in anatomical appearance with the exception of those marked as ineffective (i).

  2. bSamples from plants showing significant contributions by biological N2 fixation using the δ15N assay (Table 3).

  3. cNodules from actively N2-fixing plants inoculated with homologous Burkholderia strains giving mean acetylene reduction assay (ARA) values ± SEM (= 4) of 1152 ± 422 nmol C2H4 per plant h−1 (M. setosissima + JPY164) and 420 ± 77 nmol C2H4 per plant h−1 (M. blanchetii + JPY578).

M. albolanata+
M. adenocarpa (i)a
M. caesalpiniifolia+
M. callithrix+
M. claussenii+, +
M. cyclophylla+
M. densa (i)
M. foliolosa+
M. laniceps+
M. manidea+
M. melanocarpa+
M. radula+
M. regina+
M. speciosissima+
M. splendida+
M. ulei+
M. venatorum+
M. setosissima+
Burkholderia sp. JPY164c 
M. blanchetiib+
M. cordistipula+
M. gemmulatab+
M. lewisii (i)
M. miserab+
M. ulbrichianab+
M. blanchetii+
Burkholderia sp. JPY578c 
Table 3.   δ15N (‰ ± SEM) values of nodulated Mimosa spp. in the Caatinga biome together with those of the immediately neighbouring nonlegume reference plants
Test plantδ15NReference plantsδ15N%Ndfa
  1. Proportion of N derived from air (%Ndfa) was calculated using a ‘B-value’ of −1.24 δ15N for those plants that were significantly different from their neighbouring reference plants.

  2. *Significantly different at the 5% level from value obtained from neighbouring Mimosa sp. using a two-tailed t-test; ns, not significantly different; nd, not determined.

M. blanchetii5.16 ± 0.72Sebastiana sp. (Euphorbiaceae)9.45 ± 0.12*38
Tibouchina sp. (Melastomataceae)8.76 ± 0.68*
M. gemmulata5.58 ± 0.78Microtea maypurensis (Phytolaccaceae)13.29 ± 1.05*51
Unknown dicot.11.85 ± 0.29*
M. hypoglauca0.59 ± 0.28Byrsonima sp. (Malpighiaceae)1.0 ± 0.25nsnd
Croton sp. (Euphorbiaceae)0.56 ± 0.16ns
M. lewisii1.24 ± 0.33Eriope latifolia (Lamiaceae)2.76 ± 0.44nsnd
Sebastiania corniculata (Euphorbiaceae)2.51 ± 0.46ns
M. misera8.51 ± 0.96Croton campestris (Euphorbiaceae)11.65 ± 0.42nsnd
Sida cordifolia (Malvaceae)10.88 ± 0.29ns
M. polydidyma3.83 ± 1.98Eugenia sp. (Myrtaceae)4.99 ± 0.33nsnd
Stigmaphyllon paralias (Malpighiaceae)3.38 ± 0.74ns
M. ulbrichiana3.00 ± 1.22Glischrothamnus ulei (Molluginaceae)11.03 ± 1.06*60
Simaba ferruginea (Simaroubaceae)8.83 ± 0.52*
Copaifera coriacea (Caesalpinioideae)8.57 ± 0.14*
Figure 4.

 Western blot of extracts from Mimosa pudica nodulated with Cupriavidus taiwanensis LMG19424 (Ct) or with Burkholderia phymatum STM815 (Bp) probed with an antibody against the iron protein (nifH protein) of nitrogenase. A third lane contains extracts from alfalfa nodules infected with Sinorhizobium meliloti 102F78.

Figure 5.

 Light microscopy (a, c, e) and TEM (b, d, f) of sections of Mimosa nodules immunogold labeled either with an antibody against the nifH protein (a, c - f) or with pre-immune serum (b). (a) Serial section to Figs 4 a, b of a nodule from an actively N2-fixing M. blanchetii plant after inoculation with Burkholderia sp. JPY578 showing a positive signal with the nifH protein antibody. Nodules from the Cerrado endemic M. regina (c) and M. ulbrichiana (e) both also show positive reactions, but note that there is no signal in the senescent part of the latter nodule. TEM of bacteroids from nodules on M. caesalpiniifolia (d) and M. cordistipula (f). Panel (b) was treated with pre-immune serum and shows that there is no non-specific labeling of the bacteroids, whereas the cytoplasm of the bacteroids in (e) and (f) are strongly labelled with the nifH protein antibody. Note that the labelling of the bacteroids is largely concentrated within electron dense regions of their cytoplasm (arrows in f). Bars = 200 μm (a, c, e), 500 nm (b, d, f).

δ15N content of Mimosa spp.

Leaf samples taken from vegetation growing in undisturbed areas of the Cerrado had significantly lower δ15N values than those from corresponding Caatinga vegetation. This is well illustrated by the Mimosa spp. (Fig. 6), but is also clearly evident in a whole range of dicotyledonous species from both biomes. The mean δ15N signals of the nonlegumes, nonnodulated legumes (all Bauhinia spp.) and Mimosa spp. in the Cerrado were similar to those reported previously by Sprent et al. (1996) and Bustamante et al. (2004), and the significantly lower values of the Mimosa spp. compared with the nonnodulated plants (Fig. 6) suggest that the mimosas were probably fixing N2 (Unkovich et al., 2008). However, on a site-to-site basis, the highly variable and rather low δ15N signals in the Cerrado made it impossible to apply the δ15N technique to estimate the amounts of N fixed by the nodulated Mimosa spp., as the differences between the ‘fixing’ plants and the immediately neighbouring ‘nonfixing’ plants were not significant (data not shown). On the other hand, in the Caatinga, although the δ15N signals of the nodulated Mimosa spp. were much higher than those from the Cerrado, on average the neighbouring nonlegume reference plants had δ15N values that were > 3‰ greater than the corresponding nodulated legumes (Fig. 6), and in many cases the difference between them was even greater (up to 5‰ for M. gemmulata and M. ulbrichiana; Table 3), and so the technique could be applied to those species that were significantly different from their nearest nonnodulated neighbours (Table 3). Thus, using a B-value obtained from M. caesalpiniifolia plants inoculated with B. sabiae Br3407, %Ndfa values of 38% for M. blanchetii, 51% for M. gemmulata, and 60% for M. ulbrichiana were obtained. However, for others, such as M. hypoglauca, M. lewisii, M. misera and M. polydidyma, there were no significant differences between the Mimosa spp. and their neighbouring nonlegumes (Table 3), and so their %Ndfa was assumed to be negligible.

Figure 6.

 Mean δ15N signatures (± SEM) of leaves from Mimosa and neighbouring dicotyledonous nonlegumes and nonnodulated legumes (mainly Bauhinia spp.) collected in the Cerrado (open bars) and the Caatinga (tinted bars) in March 2006 (wet season). Means differ statistically among plant types for both Cerrado and Caatinga (ANOVA, = 9.2, = 0.0002; t-test, = −3.7, = 0.0004; respectively), and a multiple comparison test found that Mimosa and nonlegumes are statistically different in the Cerrado (post hoc Tukey HSD test, < 0.05).


Nodulation is a generic characteristic of Mimosa

This study, together with the parallel study of Bontemps et al. (2010) which examined the molecular phylogeny of the bacteria isolated from many (but not all) of the species in the present study, has shown that 67 species of Mimosa native to Brazil are nodulated and that 44 of these are new reports of nodulation. Sprent (2001) listed 54 species as being nodulated, and so the present study, together with those of Elliott et al. (2007b) and Lammel et al. (2007), brings the total number of Mimosa spp. reported to nodulate to 109, which is > 20% of this large genus of > 500 species (Barneby, 1991; Simon & Proença, 2000; Du Puy et al., 2002). In the majority of cases (47 species), nodulation was confirmed in the present study not just by visual inspection of roots for nodule-like outgrowths, which when used alone has the potential to generate erroneous reports of nodulation (Sprent, 2001), but also by detailed anatomical analyses. These have shown that all of the samples (except for those from M. gatesiae) were typical of effective, N2-fixing Mimosa nodules, being very similar in longitudinal profile to nodules reported in studies in which N2 fixation was also measured and/or the nitrogenase enzyme was localized (compare Fig. 3 with Chen et al., 2005a,b; Elliott et al., 2007b). For the other 20 species on which only dry nodules were sampled, their symbiotic ability was confirmed in several cases via isolation of effective Burkholderia symbionts (Bontemps et al., 2010) (e.g. M. setosissima; Fig. 3a). No nodules could be found on three species in the present study (i.e. M. echinocaula, M. polydidyma, M. modesta), but given the apparently generic nodulation trait in Mimosa that this study and others have shown (Elliott et al., 2007b), it is unlikely that they are nonnodulating, and their nodulation was probably inhibited by the dry conditions in which they were found (Sprent, 2009).

Burkholderia are the predominant symbionts of Mimosa in the Cerrado and the Caatinga

In situ immunolocalization of 128 separate field-collected Mimosa nodules with an antibody against B. phymatum STM815T (Chen et al., 2005b; Elliott et al., 2007b) has demonstrated that the symbionts of 47 species of Mimosa in important centres of diversity of this major legume genus are Burkholderia. This is also supported by sequences of the 16S rRNA and recA genes of 148 strains that were isolated from nodules of 47 Mimosa species (Bontemps et al., 2010), many of which were from among the same species examined by microscopy in the present study. Indeed, by using two techniques of symbiont identification from nodules collected in the field, the present study and that of Bontemps et al. (2010) have brought the total number of South American Mimosa spp. known to have Burkholderia as symbionts to > 60 (including those reported by Chen et al., 2005a).

As with nodulation, the degree of endemism apparently had no effect on the general choice of symbiont by the Mimosa spp., but it should be noted that the B. phymatum antibody used by the in situ immunolocalization technique does not allow for detailed identification of the Burkholderia symbionts beyond genus level, and so differences between Mimosa spp. in their selection of particular Burkholderia species/strains cannot be excluded by using this technique alone. Indeed, Bontemps et al. (2010), after examining Burkholderia strains that were isolated from nodules collected during the same expeditions, have shown that one particular species complex (‘SC5’) predominated at altitudes above 800 m. This suggests that although endemism per se is not a factor governing symbiont selection, it is possible that endemics, most of which have evolved at higher altitudes, such as M. decorticans and M. setosissima (Simon & Proença, 2000; Simon & Hay, 2003), have selected symbionts of a particular type that live predominantly in the soils at these altitudes. Another indicator that high-altitude endemics are selective in their choice of symbionts was the fact that the promiscuous Mimosa symbiont, B. phymatum STM815T (Elliott et al., 2007b), could not nodulate any of the Cerrado endemics effectively, and, in the case of M. setosissima, not at all.

Other symbionts of Mimosa in the Cerrado and Caatinga

Given that it is very frequently found as a symbiont of the (nonnative) invasive Mimosa spp. (M. diplotricha and M. pudica) in Asian countries, such as India (Verma et al., 2004), Papua New Guinea (Elliott et al., 2009) and Taiwan (Chen et al., 2001, 2005b), as well as in native Mimosa spp. in Costa Rica and Texas (Barrett & Parker, 2006; Andam et al., 2007), one of the most striking results of the present study was that no nodules reacted with the antibody against C. taiwanensis LMG19424. Bontemps et al. (2010) also failed to isolate Cupriavidus from any of the Mimosa spp. that they examined, and it has been previously shown in comparative studies of nodulation with B. phymatum STM815T and C. taiwanensis LMG19424T that Cerrado and Caatinga Mimosa spp. prefer Burkholderia as a symbiont (Table 3; Elliott et al., 2007b). One possible explanation for its apparent absence is that Burkholderia is so dominant as a symbiont of Mimosa spp. in these biomes that it may out-compete other potential symbionts for nodulation even of species that have a strong affinity for nodulation by them, such as M. pudica and M. diplotricha (Chen et al., 2003). Indeed, this would be especially true in the very low N-containing soils of the Cerrado and Caatinga, which would give Burkholderia a competitive advantage over other potential Mimosa-nodulating bacteria (Elliott et al., 2009). The other thing that would favour Burkholderia is the low pH of the soils in both the Cerrado (Adámoli et al., 1986) and the Caatinga (e.g. in Chap. Diamantina) (this study; Reis et al., 2004; Garau et al., 2009). On the other hand, there was some evidence that Rhizobium was occasionally a symbiont of M. xanthocentra, a very widespread species in South America (Simon & Proença, 2000). The six samples from M. xanthocentra collected in the Chap. dos Guimarães that failed to react with either the B. phymatum or C. taiwanensis antibodies did react faintly with an antibody raised against Rhizobium etli bv. mimosae (EK James & W-M Chen, unpublished), a symbiont commonly found in nodules from Mimosa species in Mexico (Wang et al., 1999). In addition, although most were identified as Burkholderia, some of the bacteria isolated from the M. xanthocentra nodules collected in this location were identified as Rhizobium sp. (Bontemps et al., 2010), and hence M. xanthocentra was the only species that was shown by both in situ immunogold labelling and via identification of bacterial isolates to have symbionts other than Burkholderia. Interestingly, it was also the only Cerrado/Caatinga species that tested positively for (slightly) effective nodulation by C. taiwanensis LMG19424 (Table 3).

Burkholderia fixes N2 within Mimosa nodules in the Cerrado and the Caatinga

Although it is now known that many Burkholderia spp. are diazotrophic (Estrada de Los Santos et al., 2001; Martínez-Aguilar et al., 2008), including nodulating strains (Chen et al., 2005b; Elliott et al., 2007b), this is the first study to examine a component of the nitrogenase enzyme complex in Burkholderia. Unsurprisingly for such a highly conserved protein, the iron protein (nifH protein or dinitrogenase reductase) in both B. phymatum and C. taiwanensis bacteroids had a similar molecular weight to that of other diazotrophs, such as Sinorhizobium meliloti. Moreover, the nifH protein in B. phymatum and C. taiwanensis is probably a dimer, as is also the case with S. meliloti (Petrova et al., 2002). Sections of nodules from 23 Cerrado and Caatinga Mimosa species immunogold-labelled using the nifH antibody confirmed that the Burkholderia endophytes of 20 of them expressed nitrogenase protein and hence were likely to be symbiotically effective. The high specificity of the antibody to nitrogenase proteins was further illustrated using TEM, which showed that the enzyme was highly localized on electron-dense regions within the bacteroids of M. caesalpiniifolia, M. cordistipula, M. regina and M. ulbrichiana.

Further evidence for biological N2 fixation (BNF) by the MimosaBurkholderia symbioses comes from δ15N analysis of leaf material collected from Mimosa spp., especially those collected from Chap. Diamantina in the Caatinga of Bahia state. In this biome, although the δ15N values were significantly higher than those in the Cerrado (this study; Sprent et al., 1996; Bustamante et al., 2004), and are similar to those reported in the Caatinga from the more northerly states of Paraiba and Pernambuco (Freitas et al., 2010), the difference between the mean δ15N signal of the nodulated legumes (Mimosa spp.) and the neighbouring nonlegumes was sufficiently large to apply the δ15N assay, and a range of %Ndfa values were obtained suggesting very substantial contributions of BNF to the N nutrition of Mimosa spp. In the cases of M. blanchetii, M. gemmulata and M. ulbrichiana, evidence for BNF was actually obtained from both techniques (immunogold labelling of nifH protein and δ15N), whereas a Mimosa sp. that showed no BNF contribution via its δ15N signal was not nodulated (M. polydidyma). Similarly, two other species which showed no BNF contribution via their δ15N signals had only either dry nodules (M. hypoglauca) or senescent ones (M. lewisii). In the case of the Cerrado, however, although the mean δ15N signal of the nonlegumes and the nonnodulating legumes (Bauhinia app.) was significantly higher than the nodulated Mimosa spp., thus suggesting that the latter were fixing N2 (Sprent et al., 1996), the high variability of the δ15N signal at all sites did not allow for the calculation of their %Ndfa in this biome (Gehring & Vlek, 2004; Unkovich et al., 2008).

Concluding remarks

This study has shown that nodulation is a generic character of the genus Mimosa, and that Burkholderia strains are the predominant symbionts in two of the major centres of diversity of this genus, that is, the Cerrado and the Caatinga. It has also shown that nodulated mimosas, including rare and endangered ones, such as M. splendida (Simon & Amaral, 2003), M. setosissima and M. decorticans (Simon & Hay, 2003), can fix N2 within their native environments, and therefore that they may make a valuable contribution to the N-cycle of the fragile ecosystems of the Cerrado and the Caatinga. This study is also the first demonstration of BNF by beta-rhizobial symbioses under field conditions.


This work was funded by the Natural Environment Research Council, UK (NE/B505038/1). We thank Pete Rowell for use of his GC, and Altiberto Baêta and Roberto G. de Souza for the N and 15N analyses. We thank Murray Unkovich for useful discussions. The collection and transfer of nodules were authorized by permit no. 007/2005 from the Ministério do Meio Ambiente, Brazil (IBAMA).