Introduction

Drought and salinity are generally considered by agronomists to be a problem for nodulated legume crops, and yet there are many wild legumes that grow in deserts, some of which are also saline and/or of high pH. With the current levels of desertification and climate change, such areas are extending rapidly, and are now estimated to occupy up to 45% of the earth’s land surface (Schimel 2010). Studies on nodulated legumes have been conducted in a number of arid and semi-arid areas of Africa, including Egypt, Ethiopia, Kenya, Senegal, Sudan and Tunisia (Zahran 2001; Zakhia et al. 2004; Wolde-meskel et al. 2004; Odee et al. 1997; Nick et al. 1999). Very drought and saline tolerant species of Prosopis have been studied in the Americas (Sprent 2009), as have Mimosa spp. in the semi-arid Caatinga biome of Brazil (dos Reis Junior et al. 2010). The plants studied in Africa were mainly either pasture species, or trees, especially African acacias. Some shrubby legumes such as Caragana spp have been studied in high altitude deserts in China (Hou et al. 2009). However, there are many other potentially nodulated legumes native to arid and semi-arid areas. Some of these have been cited in a review of nine major arid and semi-arid areas around the world (Sprent and Gehlot 2010). Diverse rhizobial genera have been isolated from 19 wild legume species growing in the Tunisian infra-arid zone in which rainfall does not exceed 180 mm (Zakhia et al. 2004). Native (endemic) legumes are widely distributed throughout India (Sanjappa 1992), including the arid and semi-arid regions of the Indian Thar Desert (Bhandari 1990). This desert, the 18th largest in the world, is 200,000 km2 in area and extends from India into Pakistan. It contains a number of legume genera not previously studied for their in situ nodulating ability or the identity of the bacteria nodulating them. The present study aims firstly to survey nodulation in these legumes, secondly to characterise the nodules formed, and thirdly to identify the rhizobia that nodulate them.

Material and methods

Studies were carried out in several districts of the state of Rajasthan. Figure 1 gives a general map of the area, together with a more detailed map of the Jodhpur district and surroundings. The current work was undertaken at the marked sites (Table 1 gives details). Overall the area is alkaline, being based on limestone. The soil in arid zone regions of Western Rajasthan has been classified as ‘desert soil’ (Ray Choudhury et al. 1963); this comes under aridisol in the USDA soil classification system (Dhir 1977). Compared with the more arid and sometimes acid desert areas in the Southern hemisphere, the soils are relatively rich in plant nutrients, including potassium and phosphorus, although much of the latter is not in a plant available form. Minimum monthly average temperatures vary from 10 to 29°C and maxima from 26 to 43°C. Rainfall varies greatly from year to year, in dry years being less than 300 mm and in wet years over 400 mm. Most of it falls in the monsoon season, between July and August. In the dry season, relative humidity is very low (around 40%), but rises to over 80% in the monsoon period. Light winds (5–10 km h−1) occur in the summer.

Fig. 1
figure 1

Map of the Rajasthan area of the Thar desert showing sampling sites. The latitudes and longitudes of Barmer and Sikar are 23° 45′ 0″N; 71° 23′ 0″E and 27° 37′ 0″N; 75° 9′ 0″E, respectively

Full size image
Table 1 Characteristics of soils at different locations in Jodhpur, Nagaur and Pali Districts of Rajasthan (Data for Sirohi site are not available)
Full size table

Determination of soil properties

Soil properties were estimated by standard methods as follows: pH (aqueous extract) and cation exchange capacity (CEC) according to Jackson (1973); electrical conductivity (EC) according to Richards (1954), organic carbon by wet digestion (Walkley and Black 1934), and available phosphorus and potassium by the methods of Olsen et al. (1954) and Pratt (1982), respectively.

Sampling methods

Field studies were carried out during the monsoon and post monsoon season (July–August) in years 2007, 2008 and 2009. Native legumes were identified using Flora of the Indian Desert (Bhandari 1990). The general classification of legumes follows Lewis et al. (2005) and the taxonomy of the Indian species is from Bhandari (1990) and Kumar and Sane (2003). Whole plants of the smaller species were excavated. Nodulated roots of 10–20 plants of each species were kept moist and brought to the laboratory for counting of nodules, making photographic records, preparation for microscopy and isolation of rhizobia. For mature tree species it was not possible to see nodules in the field because of their deep root systems. Nodulation for these was studied in seedlings growing in the field as well as in trap seedlings grown in pots with soil taken from sites where adult plants were growing. Seeds of all species were grown in pots in order to record stages of nodule development. After thorough washing, nodules were photographed under low magnification. A selection of nodules was halved and examined for pink internal colouration, indicative of leghaemoglobin and presumed nitrogenase activity (dos Reis Junior et al. 2010). Healthy nodules were halved and prepared for microscopy according to James et al. (2001). Herbarium specimens of all sampled plants were lodged in the Department of Botany, J.N. Vyas University, Jodhpur, and samples of seeds were deposited at the regional station of the National Bureau of Plant Genetic Resources, India at the Central Arid Zone Research Institute, Jodhpur. Nodulation records were checked against the database http://www.ars-grin.gov/~sbmljw/cgi-bin/taxnodul.pl and against our own records, updated from Sprent (2009).

Isolation of bacteria and cultural conditions

Bacterial strains were isolated from root nodules after surface sterilization by immersion in 0.1% (w/v) sodium hypochlorite for 5–7 min followed by immersion in 70% ethanol for one minute. After washing six times in sterile water the nodules were then crushed in one drop of sterile water and the suspension was streaked on yeast extract mannitol agar (YEMA). Bacterial colonies were grown on YEMA and on tryptone yeast (TY) agar at 28°C for 3–8 days according to the methods of Vincent (1970) and Somasegaran and Hoben (1994). They were routinely checked for their purity by repetitive streaking and Gram staining.

Authentication on original host

Isolates were tested for nodulation on their original hosts in glasshouse pot experiments as described by Yates et al. (2004). After six weeks the host plants were harvested to record nodulation. Rhizobia were reisolated from nodules and confirmed to be the inoculant strain by PCR fingerprinting using RPO1 primers (Richardson et al. 1995).

Molecular fingerprinting

Bacterial cells were used as a source of DNA for templates in all PCR reactions. Cells from a freshly grown culture in 1/2 Lupin Agar (or LA, which is described as Medium 1 in Howieson et al. 1988 and Yates et al. 2007) were concentrated to an OD600 of 6.0 in 0.89% (w/v) saline for molecular fingerprinting. PCR reaction volumes of 20 μL contained 1 μL of concentrated cells and 2.5 Units of TAQ-1 DNA polymerase (Biotech International, Perth, Australia), 1 μM of RPO1 primer, 1.5 mM MgCl2, 5X PCR Polymerisation buffer (PB-1, Biotech International, Perth, Australia).

All the primers used in this study are listed in Table 2. The reaction mixture for amplification with the RPO1 primer (Richardson et al. 1995) was held at 94°C for 5 min followed by 5 cycles at 94°C for 30 s, 50°C for 20 s, 72°C for 90 s and then 30 cycles at 94°C for 30 s, 55°C for 20 s and 72°C for 90 s and a final extension at 72°C for 5 min. A Bio-Rad Sub-Cell GT Agarose gel electrophoresis system was used to analyse the band pattern of the amplified products. Electrophoresis using a 1.5% (w/v) agarose gel prepared in TAE buffer (40 mM Tris acetate, 2 mM EDTA, pH 8.0) was carried out at 80 V for 3 h. The molecular weight marker used was 100 bp DNA Ladder (Cat. No. G5711, Promega). Gels were pre-stained by incorporating Sybr Safe (Invitrogen) (8 μL/100 ml) in molten agarose. The gels were visualised using a UV transilluminator and a digital image of the gel was captured using GEL-DOC 2000 (BioRad, USA).

Table 2 Primers used for molecular fingerprinting, amplification and sequencing
Full size table

16S rRNA gene sequencing and bacterial identification

Cells from a freshly grown culture in ½LA were concentrated to an OD600 of 2.0 in 0.89% (w/v) saline solutions. The intragenic fragments of 1475 bp from the 16S rRNA gene were amplified using the primers 18 F and 1492R. The PCR reaction mixture was held at 94°C for 5 min followed by 35 cycles of 94°C for 30 s, 53°C for 30 s, 72°C for 60 s and a final extension at 72°C for 5 min. The amplified gene products were purified using the QIAquick™ PCR purification kit (QIAGEN, Melbourne, Australia). The primers 18F, 800F, 820R and 1492R were used for sequencing of the amplified DNA. 4 μL of purified amplified template was added to 4 μL of Big-Dye (version 3.1) with the relevant primer. All DNA sequencing was carried out as described by the manufacturer (Applied Biosystems, Melbourne, Australia) by using a dye terminator cycle sequencing ready reaction kit and automated sequencer (ABI Model 377A).

Sequence contigs were assembled using the Geneious software (www.geneious.com). Databases at the National Centre for Biotechnology Information (NCBI) were searched using BLASTN (Altschul et al. 1990) for similar DNA sequences. All primers were synthesized by GeneWorks (Adelaide, Australia).

atpD gene amplification and sequencing

The housekeeping atpD gene which codes for ATP synthase subunit β of size 500 bp was amplified using the primers atpD-294 F and atpD-771 described by Gaunt et al. (2001). 5μL of lysed cell suspension was added to a 50 μL PCR mixture (1.5 mM MgCl2, 10 pmol each primer, 1 U Taq DNA polymerase and 5X PCR Polymerization buffer) and incubated at 94°C for 5 min followed by 30 rounds of thermal cycling (94°C for 45 s, 55°C for 60 s and 72°C for 90 s) and final extension at 72°C for 5 min. The amplified product was confirmed by electrophoresis using 4 μL products on a 1% agarose gel at 80 V for 2 h. 100 bp DNA Ladder (Promega) was used to estimate the length of the amplification products. The purified DNA was sequenced as described above in for the 16S rRNA gene.

Phylogenetical analysis

Sequences of type strains and/or NCBI reference sequences (NR) were included in the analyses. The GenBank accession numbers are listed in parentheses for the 16S rRNA and atpD genes. To perform molecular phylogenetic analyses, reference sequences required for comparison were downloaded from the NCBI . All the sequences of 16S rRNA and atpD were aligned using the multiple sequence alignment program CLUSTAL W (Thompson et al. 1997) and were saved as molecular evolutionary genetics analysis (MEGA) format in software MEGA v5.1 (Tamura et al. 2011). The pair wise evolutionary distances were computed using the Kimura 2-parameter model (Kimura 1980). Phylogenetic trees were inferred using the neighbour-joining method (Saitou and Nei 1987). To obtain the confidence values, the original data set was resampled 1000 times using the bootstrap analysis method. The MEGA 5.1.0 package (Tamura et al. 2011) was used for construction of phylogenetic trees, inferring distances and percent similarity.

Results

Nodulation

Nodulation of 35 native species recorded at the different sites is given in Table 3. Of the four caesalpinioid legumes found, nodules were only seen on Chamaecrista pumila. Nodules were absent in all investigated species of Cassia or Senna. All mimosoid species found are from genera known to nodulate and all were nodulated in our study sites. Nodulation of Acacia jacquemontii is a new report. Similarly, all papilionoid species found are from known nodulating genera and all were found to nodulate in the present work. Two species of Rhynchosia, were found, one of which, R. aurea, is a new report of nodulation. Seven species of the large pan-tropical genus Indigofera were recorded, as well as the related Cyamopsis tetragonoloba, a very important food legume also known as clusterbean. Tephrosia is another large pan-tropical genus well represented here, and T. falciformis is a new report of nodulation.

Table 3 Occurrence and nodulation of native legumes in arid and semi arid regions of Rajasthan. Taxonomy as in Kumar and Sane (2003). A = Annual; P = perennial
Full size table

The type genus of tribe Crotalarieae contains a large number of species, many of which have xerophytic characters, as in the two species found here. The woodiest papilionoid species studied is the valuable timber tree Dalbergia sissoo. This species is not particularly well adapted to dry conditions and may be at the limit of its habitat range, although it is often listed as drought tolerant. Two species of tribe Trifolieae, Melilotus indicus and Trigonella corniculata (synonym T. balansae Boiss. & Reut.), are also on the borderlines of their normal habitat. Taverniera is a small genus in tribe Hedysareae that is largely restricted to drier areas and deserts. Table 4 lists the major morphological features associated with drought tolerance found in the present study and Fig. 2 illustrates some of these.

Table 4 Characteristics of some native legumes of the Indian Thar Desert growing in open field, gravel and sandy soil
Full size table
Fig. 2
figure 2

Plants from the Thar desert, illustrating xerophytic characters; (a) Acacia jacquemontii, a small drought-tolerant tree which has large spines and sheds its leaves in the dry season (b) Alysicarpus vaginalis, a drought-escaping annual which has a prostrate habit and a short life cycle (2–3 months); (c) Crotalaria burhia, a drought-tolerant shrub which has thick multibranched photosynthetic stems, and small leaves (arrows) that become thick and scaly under dry conditions; (d) Indigofera sessiliflora, a drought-escaping annual with hairy leaves (arrows) and a short life cycle (2–3 months); (e) Tephrosia villosa, a drought-tolerant shrub with hairy subcoriaceous leaves, and profusely hairy pods (arrow). See Table 3 for further details about these species and other nodulated legumes native to the Thar Desert

Full size image

The number of nodules per plant varied in the field depending on growing conditions, the sample site and type of plant species (herbaceous annual, perennial or trees). However, during authentication experiments in pots, host plants inoculated with their respective isolated microsymbionts had between 20 and 35 nodules per plant, depending on host species and inoculant strains (data not shown).

Nodule morphology and anatomy

Nodule morphology is shown in Fig. 3. Nodules with determinate growth and surface lenticels were found in Alysicarpus (not shown) and Rhynchosia (Fig. 3a). Except for Dalbergia sissoo, which has aeschynomenoid nodules (not shown, but see Sprent 2009), a type with a form of determinate growth that is associated with emergence of lateral roots, all other species had nodules that showed indeterminate growth (Fig. 3b–f), although they varied greatly in the degree of branching. Nodules on species of Acacia are usually branched (Sprent 2001), unlike the example shown here. Indigofera nodules are particularly interesting as they appeared classically determinate at early stages and then became elongated, but retained the lenticels typical of other determinate nodules (Fig. 3f). This is the first report illustrating this type of nodule, although it has also been found in South African species of Indigofera (I. Law, pers. comm.; Sprent, unpublished data). The surfaces of all nodules were dark in colour, due to drying soil. This has been shown in other nodules from dry areas to arise by production of bark from a cortical phellogen (Fig. 3a & d), which also produces aerating tissue in the wet season (see for example Fig. 3.5 in Sprent 2001).

Fig. 3
figure 3

Morphology of root nodules of (a) Rhynchosia aurea; (b) Chamaecrista pumila; (c) Crotalaria burhia; (d) Tephrosia leptostachya; (e) Acacia jacquemontii; (e); (f) Indigofera linifolia. All the nodules are indeterminate, except for those on R. aurea (a). Note the bark (*) on (a) and (d) that has arisen from a cortical phellogen, and the pronounced lenticels (arrow) on (f). Bars = 5 mm

Full size image

All nodules collected were red in internal colour and therefore assumed to be active in fixing nitrogen. This was confirmed by immuno-labelling of NifH protein in bacteroids (data not shown here, but similar to that shown in other work from our laboratory e.g. dos Reis Junior et al. 2010). The general structure of nodules for species of Chamaecrista, Acacia, Rhynchosia and Mimosa was similar to previous descriptions given by Naisbitt et al. (1992), Räsänen et al. (2001), Garau et al. (2009) and dos Reis Junior et al. (2010), respectively. Here we illustrate the nodule structure of three genera, in particular the species Crotalaria burhia, Indigofera linnaei and Tephrosia wallichii (Fig. 4). All have a distinct apical meristem, an inner cortex containing vascular bundles and an outer cortex with some indications of phellogen activity (Fig. 4b, d, f). Where they differ is in the region behind the apex, where cells become infected with rhizobia and in the arrangement of the cells in the active nitrogen fixing tissues. In C. burhia (Fig. 4a) no uninfected (interstitial) cells are present in the infected tissue, whereas in I. linnaei (Fig. 4c) and T. wallichii (Fig. 4e) the latter consists of a mixture of infected and uninfected cells. The absence of interstitial cells appears to be a characteristic of nodules from tribe Crotalarieae (Sprent 2007). Thus, the nodules of C. burhia and Ch. pumila that may appear similar morphologically (Fig. 3a, b) are internally quite different, as the Chamaecrista nodules have interstitial cells (not shown, but see Naisbitt et al. 1992).

Fig. 4
figure 4

Light microscopy of longitudinal sections of nodules of Crotalaria burhia (a, b); Indigofera linnaei (c, d) and Tephrosia wallichii (e, f); (a, c, e) illustrate for each species details of the nodule meristem (m) and the nitrogen-fixing infected zone (iz), whereas (b, d, e) are higher magnification views of the cortex, including vascular bundles (arrows). Asterisks mark uninfected cells in the infected zone in (c – f). Bars = 100 μm (a, c, e) and 30 μm (b, d, f)

Full size image

Bacteria isolated from nodules and their identification on the basis of 16S rRNA sequences

A collection of 70 rhizobial isolates was obtained from the five genera and 8 nodulated species of legumes. Of these we focussed on isolates from Mimosa hamata, as the genus Mimosa is known to have the betaproteobacterial genus Burkholderia as its major symbiont in Brazil (dos Reis Junior et al. 2010), on Prosopis cineraria, because of the world wide occurrence of this genus in arid areas (Sprent 2009), on species of Tephrosia, as they are the first to resume growth after monsoonal rains, and on Rhynchosia aurea and Acacia jacquemontii, as both are new reports of nodulation. In addition, R. aurea is of interest, as it is from a genus that is also known to nodulate with Burkholderia (Garau et al. 2009). Selected isolates from the aforementioned plants were sequenced for their 16S rRNA genes based on groups obtained by RPO1 fingerprints (RAPDs patterns). The selected isolates were initially identified through BLASTN searches using their partial (>500 bp) 16S rRNA gene sequences, and then a nearly full length fragment (>1300 bp) of the 16S rRNA gene was amplified and sequenced for sixteen selected isolates (Table 5). All nodulating isolates were from the Alphaproteobacteria, and they included species of Sinorhizobium, Bradyrhizobium and Rhizobium, with Sinorhizobium spp. being particularly common (Table 5). In the case of strains identified as Bradyrhizobium on the basis of their 16S rRNA sequences, another housekeeping gene, atpD, was also sequenced in order to more precisely determine their phylogenetic relationships (Table S1). All rhizobial isolates were authenticated on their original hosts.

Table 5 Range of nodulating bacteria isolated from legumes native to the Thar desert (all strains were Nod+ Fix+on their respective hosts)
Full size table

Sequence analysis and phylogeny

The sequences of the 16 selected strains were aligned and compared with 16S rRNA and atpD gene sequences of rhizobial type strains available in the GenBank database. Dendrograms showing the phylogenetic relationships of the unclassified Thar desert rhizobia with those of previously designated species of rhizobia are shown in Figs. 5a, b and S1. On the basis of 16S rRNA gene sequences our results indicate that strains from Mimosa hamata (JNVU MH3a & JNVU MH8) and Acacia jacquemontii (JNVU AJ10) were 99.8% similar to Sinorhizobium sp. ORS 1085 (AJ295078) from nodules of Acacia tortilis ssp. raddiana growing in Sahara, Africa (Ba et al. 2002). These three strains showed a similarity to the type strain of S. saheli LMG 7837 (NR_026096) between 99.7 to 99.8%, but formed a separate group from the latter. Strains from species of Tephrosia, such as T. wallichii (JNVU TW8 & JNVU TW10), T. falciformis (JNVU TF7) and T. villosa (JNVU TV3), as well as those from Prosopis cineraria (JNVU PC2) and Rhynchosia aurea (JNVU RA9) were very close (with 99.7% similarity) to Sinorhizobium sp. ORS 3178 (AY875975), which was isolated from Acacia nilotica in Senegal. Although these strains had 99.2 to 99.6% similarity with sequences of the type strains of S. terangae LMG 7834 (NR_044842) and S. saheli LMG 7837 (NR_026096), respectively, they formed two lineages in a subclade together with S. terangae strain LMG 7834 and S. saheli strain LMG 7837 (Fig. 5a). On the other hand, strains from T. purpurea (JNVU TP13 & JNVU TP18) formed separate lineages from the other Tephrosia sinorhizobia, and had 99.7% similarity with S. terangae PHABA-2 (GU784793) isolated from Phaseolus vulgaris in Cuba, and 99.3 and 99.6% similarity with the S. terangae (NR_044842) and S. saheli (NR_026096) type strains, respectively. In addition, isolate JNVU TV1 from T. villosa showed 99.7% similarity with Sinorhizobium sp. ORS 3180 (AY875976) isolated from Acacia nilotica in Senegal and with the type strain of S. terangae (NR_044842), forming yet another separate lineage in the same (S. terangae) clade. In summary, nine of the 12 Sinorhizobium strains isolated and described during this study formed four lineages in a subclade together with the S. terangae type strain, whereas the remaining three strains (all from M. hamata and A. jacquemontii) formed another lineage in a subclade of S. saheli (Fig. 5a).

Fig. 5
figure 5

Neighbour-joining phylogenies for 16S rRNA gene sequences of rhizobia isolated from legumes in the Thar Desert together with those of reference strains. The trees were generated using MEGA version 5.0 with default parameters, K2P distance model and the Neighbour-Joining algorithm. Bootstrap values calculated for 1,000 replications are indicated. The scale bars indicate 1% substitutions per site. Accession numbers from GenBank are in parenthesis. B., Bradyrhizobium; R., Rhizobium; S., Sinorhizobium; E., Ensifer; M., Mesorhizobium; T, type strain and (NR), NCBI Reference sequence. (a) Sinorhizobium strains; (b) Rhizobium and Bradyrhizobium strains. The tree is rooted using the sequence from Rhodobacter capsulatus

Full size image

Of the four non-sinorhizobial symbionts isolated during this study (Table 5), isolate JNVU TP3 from T. purpurea was identified as a Rhizobium strain with 99.2% similarity to Rhizobium sp. ORS 3441 (EU584258) isolated from Acacia senegal in Senegal (Fall et al. 2008), and 99.1% similarity with the type strain R. mesosinicum CCBAU 25010 (NR_043548) forming a separate subclade (Fig. 5b). The remaining three strains, all from Tephrosia spp., were identified as bradyrhizobia. Strains JNVU TF17 from T. falciformis and JNVU TV4 from T. villosa showed 99.7 and 99.2% similarity respectively with the B. yuanmingense type strain B071 (NR_028768). Strains JNVU TF17 and JNVU TV4 showed 99.9 and 99.5% similarity respectively with B. yuanmingense strain M11 (AB601666) isolated from Vigna radiata in Nepal, but had separate lineages within this clade (Fig. 5b). Isolate JNVU TV11 was also identified as a species of Bradyrhizobium with 99.5% similarity to Bradyrhizobium sp. strain M16 (AB601671), which was isolated from V. radiata in Nepal, and 99.2% similarity to the B. iriomotense type strain EK05 (AB300992), although it formed a separate subclade within B. iriomotense (Fig. 5b). Two of the bradyrhizobial isolates from Tephrosia (JNVU TV4 & JNVU TV11) were selected on the basis of their locations within separate 16S rRNA groups (Fig. 5b) for further sequencing of their atpD housekeeping gene in order to more clearly define their phylogenetic positions within the genus Bradyrhizobium (Table S1). Phylogenetic analysis of the atpD gene showed that isolate JNVU TV4 had 99.3% similarity with the type strain of B. yuanmingense LMG 21827 (FM253140) isolated from Lespedeza cuneata in China, whereas the other isolate, JNVU TV11, had 95.2% similarity to B. yuanmingense CCBAU 33112 (FJ418711) isolated from V. angularis in China, and 95.6% similarity to the type strain of B. yuanmingense (LMG 21827) (Fig. S1).

Discussion

Nodulation of native legumes of the Thar desert

Plants have a number of ways in which they adapt to dry conditions, basically by tolerating or avoiding drought. Legumes are no different in this respect (Fig. 2, Table 4). Comparison of our data with the known occurrence of potentially nodulating genera in other alkaline desert soils shows that the same genera are often found. For example in dry areas of Madagascar (Du Puy et al. 2002) and Soqotra, an island off the Yemeni archipelago (Miller and Morris 2004), the genera Acacia (s.l.), Crotalaria, Indigofera and Tephrosia are all well represented. They are also common in arid areas of Southern Africa, in soils with a range of pH values (Jürgens et al. 2010). It is now thought that legumes evolved in seasonally dry areas bounding the Tethys Sea about 55 Mya (Schrire et al. 2005). From here they extended into other biomes, including arid ones, and evidence suggests that these processes are continuing (Crisp et al. 2004). Thus it is not surprising that legumes are found in deserts today. What is surprising is that their nodulation characters (rather than their taxonomic ones) have frequently been ignored. We found all known types of nodule morphology, plus an additional intermediate determinate/indeterminate form (in Indigofera). Nodule structure also showed the known range of variation, and new observations on Tephrosia, Crotalaria and Indigofera confirm the proposed taxonomic and evolutionary significance of these characters (Sprent 2007). There appears to be no particular advantage of a specific type of nodule in desert environments, even though indeterminate nodules have the capacity to re-grow after adverse conditions, whereas determinate nodules normally have to be replaced (Sprent 2001).

On the other hand, the nodules in our study did show some morphological and structural adaptations to their semi-arid environment, such as a covering of bark on the determinate nodules of R. aurea and the indeterminate nodules of T. leptostachya; this arises from a corky phellogen. Indeed, in their study of endemic Mimosa spp. in the seasonally dry Cerrado and Caatinga regions of Brazil, dos Reis Junior et al. (2010) also observed a thickened cortex in the nodules on woody perennial species that were subjected to prolonged periods of drought, and hypothesised that such a “corky” layer would help the nodules to retain moisture. Interestingly, some of the nodules, such as those on Indigofera linifolia also displayed adaptations to inundation, such as lenticels (James et al. 2001). This highlights the highly seasonal nature of precipitation in such environments, which necessitates adaptations to both drought and flooding.

Thar desert legumes are predominantly nodulated by Sinorhizobium (Ensifer) strains

As with other recent studies from natural arid and semi-arid environments (Nick et al. 1999; Odee et al. 1997; Wolde-meskel et al. 2004; Zahran 2001; Zakhia et al. 2004, 2006; Romdhane et al. 2006), we found a wide range of bacteria to be present inside nodules, but with a particularly high incidence of Sinorhizobium (Ensifer) strains. Interestingly, the type of rhizobia associated with some of the native legumes of the Indian Thar desert are similar to those found in leguminous trees in seasonally dry regions of Sudan and Kenya (Nick et al. 1999). On the basis of 16S rRNA gene sequencing, Nick et al. (1999) reported that 25 Sudanese and 5 Kenyan strains isolated from root nodules of Acacia senegal and Prosopis chilensis showed similarity with Sinorhizobium saheli and Sinorhizobium terangae, but were sufficiently different from these established species to be included in two new species, Sinorhizobium arboris and S. kostiense. Similarly, Benata et al. (2008) isolated a diverse range of bacteria from root nodules of Prosopis juliflora growing in two arid soils in eastern Morocco that again were dominated by sinorhizobia, including strains of S. kummerowiae and S. fredii (although strains of R. multihospitium, R. tropici and R. giardinii were also isolated). Furthermore, Deng et al. (2011) have reported diverse rhizobia belonging to eight putative species in the genera Mesorhizobium, Rhizobium and Sinorhizobium to be nodulating a single legume species, Sphaerophysa salsula, from two regions of the semi-arid Loess Plateau of China.

On the other hand, the aforementioned studies, as well as the present one, contrast with an extensive study of legumes from dry areas of Senegal by Doignon-Bourcier et al. (2000) which found that species in many of the genera that we have studied in India, such as Chamaecrista, Crotalaria, Indigofera, Rhynchosia and Tephrosia, were exclusively nodulated by strains of Bradyrhizobium. In addition although the Rhynchosia spp. in our study were nodulated by Sinorhizobium spp., a recent study on a South African species of Rhynchosia has shown that it is nodulated by Burkholderia (Garau et al. 2009). This difference might be explained by the fact that unlike those in our work, the species (R. ferulifolia) examined by Garau et al. (2009) was native to acid soils which could favour the growth of nodulating Burkholderia (dos Reis Junior et al. 2010). This is also likely be the case with the native Indian Mimosa species, M. hamata, which is nodulated by Sinorhizobium rather than by Burkholderia strains. Mimosas in their native environments are commonly nodulated by Burkholderia, as was recently shown in a study of 70 Mimosa species growing in acidic soils in central Brazil (dos Reis Junior et al. 2010). Thus it is likely that local soil factors play an important role in defining the bacteria nodulating indigenous legumes.

Concluding remarks

From the results of the current investigation it appears that the native rhizobia of arid regions of the Indian Thar desert belong to both fast- and slow-growing rhizobial species, but also that the legumes may have a particular preference for Sinorhizobium. Indeed, the fast growing sinorhizobia characterized in this study are phylogenetically close to those reported from semi-arid regions of African countries (Sudan, Kenya, Tunisia, Morroco, Senegal). Interestingly, as with the African sinorhizobia from semi-arid regions, the Thar desert sinorhizobia reported here appear to have evolved to be promiscuous symbionts, as they can nodulate several native legumes (6 species) as well as crop legumes, such as Vigna radiata, V. aconitifolia and V. unguiculata (data not presented here). Our data further suggest that our strains may constitute two to three new species of Sinorhizobium, and possibly also a new species each of Rhizobium and Bradyrhizobium. However a polyphasic approach, including DNA-DNA hybridization and GC% analyses, is needed to confirm this.