Phylogenetic Analyses of Rhizobia Isolated from Nodules of Lupinus angustifolius in Northern Tunisia Reveal Devosia sp. as a New Microsymbiont of Lupin Species

: Thirty-two bacterial isolates were obtained from root nodules of Lupinus angustifolius growing in Northern Tunisia. Phylogenetic analyses based on recA and gyrB partial gene sequences grouped the strains into six clusters: four clusters belonged to the genus Bradyrhizobium (22 isolates), one to Microvirga (8 isolates) and one to Devosia (2 isolates), a genus that has not been previously reported to nodulate lupin. Representative strains of each group were further characterized. MultiLocus Sequence Analysis (MLSA) based on recA and glnII gene sequences separated the strains within the genus Bradyrhizobium into four divergent clusters related to B. canariense , B. liaoningense , B. lupini , and B. algeriense , respectively. The latter might constitute a new Bradyrhizobium species. The strains in the Microvirga cluster showed high identity with M. tunisiensis . The Devosia isolates might also represent a new species within this genus. An additional phylogenetic analysis based on the symbiotic gene nodC afﬁliated the strains to symbiovars genistearum , mediterranense , and to a possibly new symbiovar. These results altogether contributed to the existing knowledge on the genetic diversity of lupin-nodulating microsymbionts and revealed a likely new, fast-growing, salt-tolerant rhizobial species within the genus Devosia as a potentially useful inoculant in agricultural practices or landscape restoration.


Introduction
The Leguminosae or Fabaceae family includes more than 19,700 species, many of which are agronomically important [1]. The Genisteae tribe is presently considered to include 618 species in 25 genera [2]. Lupinus is considered the largest Genisteae genus in terms of species number with around 270 species [3,4]. This genus may be divided into two groups, the first of which is centered in the Mediterranean Basin and the Canary Islands [5,6] and comprises 13 Old World species with smooth and rough seeds [3,7]. The second group includes the New World species [8]. The New World species number ranges from 200 to 400 located in the Americas. Lupins are able to colonize very different environments and diverse ecosystems [9], including marginal impoverished or contaminated soils [10]. L. angustifolius, commonly named narrow leaf lupin and blue lupin, grows wild in agricultural lands as well as in the wetlands of coastal plains in the Northern region of Tunisia.
Lupins, partly due to their effective nitrogen-fixing symbiosis with different soil bacteria (collectively called rhizobia), have been grown since antiquity as an important pulse crop and as a green manure. Lupinus species have been studied from various perspectives as indicators of phosphorus and nitrogen in soils poor in nutrients, because of their dual efficacy in fixing atmospheric nitrogen and solubilizing soil phosphates [11], phytoremediation of polluted soils [12,13] and as medicinal plants [14].The high protein content of lupin seeds is similar to that of soybean, and lupin cultivation has been suggested as an alternative to soybean in Europe [15], as new sources are needed to meet the increasing demand for plant protein [16].
The genus Devosia was defined is a result of the reclassification of Pseudomonas riboflavin as Devosia riboflavina [31]. Devosia bacteria are aerobic, Gram-negative, rod-shaped or oval and they do not form spores. Twenty-one validated species have been identified within the genus (http://www.bacterio.net/devosia.html, accessed on 16 July 2020). Some Devosia species have been reported to nodulate legumes [32][33][34]. We have previously reported that L. angustifolius from Tunisia was nodulated by new lineages within the Bradyrhizobium and Microvirga genera [35], but, to the best of our knowledge, there are no reports on Devosia symbiotic nitrogen-fixing bacteria that nodulate Lupinus species.
The main objective of the present paper was to further investigate and characterize the diversity of the rhizobia that nodulate L. angustifolius growing in different sites in Northern Tunisia.

Sampling Sites, Isolation of Bacterial Strains and Culture Conditions
Effective nodules were collected from L. angustifolius roots growing wild or in soil collected from six different locations in Northern Tunisia: Mraissa, Takelsa, Borj Hfaied, Hammamet, Sejnane and Tabarka. The ecological and climatic characterization of the sampling sites is indicated in Table 1.
Nodules were surface-sterilized with 95% ethanol for 1 min, 25% sodium hypochlorite for 3 min, then rinsed several with sterile distilled water; they were individually crushed on sterile plates, and a loopful of nodule material was streaked onto yeast-mannitol agar (YMA) [36]. Plates were then incubated aerobically at 28 • C and colonies could be visualized after 4 to 10 days.
Single colonies were streaked on fresh YMA plates. Pure isolates were maintained on YMA plates at 4 • C for temporary preservation, or stored in 20% (w/v) glycerol suspensions at −80 • C. The isolates used in this study are listed in Table 1.

Nodulation and Cross-Inoculation Test
The ability of all purified isolates to nodulate their original host was assessed. The isolates were also tested for nodulation on L. luteus, L. micranthus, L. mariae-josephae, Retama sphaerocarpa, Vigna unguiculata and Glycine max. Seed germination and plant inoculation were carried out as previously described [37]. Plants were grown under bacteriologically controlled conditions in a greenhouse for 3 to 8 weeks, depending on the legume host, with a 16.0/8.0 h light/dark photoperiod at 25/23 • C, and they were watered with sterile Jensen's liquid medium once a week. Three uninoculated Leonard jars per legume were included as negative controls. Nodules were examined for number and inside color. Three replicates of each strain were used in the nodulation tests.

DNA Isolation, PCR Amplifications and Sequencing
Bacterial DNA was obtained by the alkaline lysis method [38]. PCR was performed in a 25 µL volume containing DNA (5-10 ng), 2.5 µL of 10× PCR buffer containing magnesium chloride (Roche Applied Science), 10 µM of each primer, 10 mM of each dNTP, 1 µL DMSO and 1 U of Taq DNA polymerase (Roche Applied Science, Mannheim, Germany).
glnII appeared to be absent in the genomes of Microvirga spp., as it could not be retrieved from data banks or from any sequenced Microvirga genomes. Hence, dnaK was used, and the phylogenetic tree for this group of strains was assembled using gyrB and dnaK concatenated sequences.
Amplification products were purified from the unincorporated primers and (deoxynucleosidetriphosphates (dNTPs) with the NucleoSpin ® Extract II (Macherey-Nagel, Dueren, Germany) or by gel electrophoresis followed by band purification with the same kit, if necessary. Sequencing was performed externally at Secugen (Madrid, Spain) and STAB Vida (Lisbon, Portugal).
Phylogenetic and molecular evolutionary analyses were conducted using the MEGA7.0 software [40]. The maximum likelihood (ML) and neighbor-joining (NJ) statistical methods and the Kimura two-parameter model were used for gene and for multilocus sequence analysis (MLSA) (combined sequences of glnII and recA; gyrB and dnaK). ML and NJ analyses provided similar results. NJ phylogenetic trees are shown in Figures. Phylogenetic trees were bootstrapped with 1000 bootstrap replications. Accession numbers are shown in the tree figures and were obtained after deposit in the Genbank sequence database.

Isolation and Symbiotic Efficiency of L. angustifolius-Nodulating Bacteria from Northern Tunisia
A total of 32 rhizobial bacteria were isolated from L. angustifolius nodules growing naturally in six different areas in Northern Tunisia (Table 1). These isolates were tested for re-nodulation and efficient nitrogen-fixing activity by re-inoculating L. angustifolius seedlings and growing the plants under bacteriologically axenic conditions. All isolates were able to nodulate their original host, as corroborated by the formation of nodules that were pink inside, which is indicative of active nodules. The nodule number ranged from 3 to 10 per plant. Indirect nitrogen fixation effectiveness could be assessed by the dark green color of the leaves compared to uninoculated yellowish control plant leaves, and by visual observation of pink-red color due to the leghemoglobin presence inside the nodules.

Phylogenetic Analysis Based on recA and gyrB Genes
The recA gene was amplified and sequenced for all isolates. The sequences were compared with those available in GenBank. As shown in the recA phylogenetic tree topology (Figure 1a), L. angustifolius strains are grouped in six different clades, four within the Bradyrhizobium genus (22 strains) and one with Devosia (2 strains). A sixth group, based on gyrB gene phylogeny, formed by Microvirga (8 strains) is shown in Figure 1b. The bradyrhizobia subgroups contained 17 strains affiliated to B. lupini, 2 strains to B. canariense, 1 to B. liaoningense (86%) and 2 strains to B. algeriense. At least one representative isolate of each pattern was selected for further phylogenetic analyses.
In order to define a more phylogenetically robust position of the strains, an analysis based on the concatenation of relevant housekeeping genes was performed. recA and glnII were used for Bradyrhizobium strains as they have been proven to be appropriate in phylogenetic studies of this genus [23,37]. glnII could not identified in any Microvirga strains draft genome sequences. Thus, dnaK was used instead, and the phylogenetic tree for this group of strains was made using gyrB and dnaK concatenated sequences. The trees again distributed L. agustifolius nodulating strains in six distinct groups, four (I, II, III and IV) contained all the bradyrhizobia, one (V) grouped the Devosia strain and one (VI) comprised the Microvirga strains (Figure 2a  In order to define a more phylogenetically robust position of the strains, an analysis based on the concatenation of relevant housekeeping genes was performed. recA and glnII were used for Bradyrhizobium strains as they have been proven to be appropriate in phylogenetic studies of this genus [23,37]. glnII could not identified in any Microvirga strains draft genome sequences. Thus, dnaK was used instead, and the phylogenetic tree for this group of strains was made using gyrB and dnaK concatenated sequences. The trees again distributed L. agustifolius nodulating strains in six distinct groups, four (I, II, III and IV) contained all the bradyrhizobia, one (V) grouped the Devosia strain and one (VI) comprised the Microvirga strains (Figure 2a,b).
Strains LanT1, LanTb3, LanM5 and LanTb18 clustered with B. lupini USDA3051.LanS2 and LanTb17 grouped with B. canariense BTA-1 and B. liaoningense LMG18230, respectively. LanH1 occupied a position between B. valentinum LmjM3 and B. algeriense RST89. LanTb5 was located in a taxonomic position separate from the Bradyrhizobium spp., which suggested that it might constitute a new species under a different genus. The phylogenetic tree based on the concatenated gyrB and dnaK sequences ( Figure  2b) revealed that strains in Group VI (LanM6, LanTb9 and LanH5) grouped with M. tunisiensis LmiM8 T with almost 100% identity.

16S rRNA Gene Phylogeny
Phylogenetic analysis of the nearly full length sequences of 16S rRNA gene of 11 representative strains showed five large groups ( Figure 3). Groups I and II of the analysis based on housekeeping genes form a single group when analyzed based on the rrs gene. This group contained five strains (LanTb18, LanM5, LanS2, LanT1, and LanTb3) grouped with B. canariense BTA-1 and B. lupini USDA 3051. The rest of the groups were similarly distributed. LanTb17 grouped with B. liaoningense LMG18230. Strain LanH1 occupied a separate position in the extra-slow bradyrhizobial clade formed by B. valentinum LmjM3, B. pachyrhizi PAC48 and B. lablabi CCBAU 23086, and might constitute a new species in the Bradyrhizobium genus. Three fast growing strains (LanM6, LanH5 and LanTb9) were identical to Microvirga tunisiensis LmiM8,also isolated in Northern Tunisia from L. micranthus [37]. The 16S rRNA sequence of LanTb5 shared 99% identity with Devosia sp. Dc2a-9 isolated from algal-bacterial consortia (https://www.ncbi.nlm.nih.gov/nuccore/300068687, accessed on 10 March 2021). Altogether, the phylogenetic analyses revealed several Bradyrhizobium, Microvirga and Devosia species, which had not been previously described, as microsymbionts of L. angustifolius.

16S rRNA Gene Phylogeny
Phylogenetic analysis of the nearly full length sequences of 16S rRNA gene of 11 representative strains showed five large groups ( Figure 3). Groups I and II of the analysis based on housekeeping genes form a single group when analyzed based on the rrs gene. This group contained five strains (LanTb18, LanM5, LanS2, LanT1, and LanTb3) grouped with B. canariense BTA-1 and B. lupini USDA 3051. The rest of the groups were similarly distributed. LanTb17 grouped with B. liaoningense LMG18230. Strain LanH1 occupied a separate position in the extra-slow bradyrhizobial clade formed by B. valentinum LmjM3, B. pachyrhizi PAC48 and B. lablabi CCBAU 23086, and might constitute a new species in the Bradyrhizobium genus. Three fast growing strains (LanM6, LanH5 and LanTb9) were identical to Microvirga tunisiensis LmiM8,also isolated in Northern Tunisia from L. micranthus [37]. The 16S rRNA sequence of LanTb5 shared 99% identity with Devosia sp. Dc2a-9 isolated from algal-bacterial consortia (https://www.ncbi.nlm.nih.gov/nuccore/300068687, accessed on 10 March 2021). Altogether, the phylogenetic analyses revealed several Bradyrhizobium, Microvirga and Devosia species, which had not been previously described, as microsymbionts of L. angustifolius.

NodC Gene Phylogeny
Partial nodC sequences were obtained from only six bradyrhizobial strains. Five were in groups I and II defined in Figure 1 and lined up with symbiovar genistearum formed by strains able to nodulate legumes in the Genisteae tribe, such as B. lupini, B. rifense, B. cytisi and B. canariense type strains ( Figure 4). LanH1, in group IV of Figure 1, might correspond to a new symbiovar designated new sv. A, which is distant from other Bradyrhizobium symbiovars and close to symbiovar retamae, defined with strains nodulating Retama spp., and L. mariae-josephae [41]. nodC sequences from Microvirga strains (LanM6, LanTb9 LanH5) were identical to that of Microvirga tunisiensis LmiM8 belonging to the new mediterraneense symbiovar (Figure 4) [42]. The LmiM8 strain was an effective endosymbiont isolated from L. micranthus.  Amplification of nodC and nodA could not be obtained for strains LanTb17 and LanTb5. nifH gene sequencing affiliated LanTb17 (MZ343378) to Bradyrhizobium japonicum CCBAU43129 with 97% identity. LanTb5 (MZ369158) nifH gene was aligned to Rhizobium tropici CIAT 899 with 78% identity. These results might indicate that these novel L. angustifolius microsymbionts might have acquired their nif genes from phylogenetically distant sources through lateral transfer.

Phenotypic Analyses of Narrow-Leaf Lupin-Nodulating Strains
The rhizobial isolates presented a broad biodiversity in their growth characteristics. Ten isolates could be classified as fast-growing (Devosia and Microvirga) and twenty-two of them as slow-growing rhizobia (Bradyrhizobium). The phenotypic traits of the 11representative strains are summarized in Table 3. All strains were able to grow in optimal growth conditions (28 • C, pH 6.8 and <0.01% NaCl). Cells were rod shaped, Gram negative and aerobic. The Bradyrhizobium isolates were slow growers with mean generation times >15 h in YMB. Colonies produced in YMA were translucent, gummy and circular. They presented convex smooth margins, and their sizes were 2-3 mm in diameter after 6-7 days in YMA medium at 28 • C. LmiH1 was an extra-slow grower (generation time in YMB > 30 h) that produced punctiform, non-mucoid colonies (<1 mm) after 10 to 11 days in YMA plates. Fast growers showed mean generation times of 4-7 h in TY. They also grew well on YMA plates. After 8 days at 28 • C, Microvirga colonies were 2-3 mm in diameter, convex, smooth, mucilaginous pink, and circular. The effect of temperature, salinity, pH and some antibiotics, used to quantify and qualify stress tolerance, showed a degree of heterogeneity (Table 3). All the Microvirga strains grew well between 20 and 37 • C, at pH 4-12, and on media containing 0.1; 0.5 and 1% salt. These strains were very sensitive to all antibiotics used (kanamycin (50 µg/mL), ampicillin (100 µg/mL), gentamicin (30 µg/mL), spectinomycin (50 µg/mL), tetracycline (5 µg/mL), streptomycin (10 µg/mL). The Bradyrhizobium strains grew at temperatures between 20 and 28 • C, at pH range 4 to 11 and on media containing 0.1% salt, with only strains LanS2, LanTb17 and LanH1 growing at temperatures between 20 and 37 • C. These strains were sensitive to kanamycin, gentamicin, spectinomycin, and streptomycin and were less resistant to ampicillin and tetracycline, except for strain LanTb17, which showed a good resistance to kanamycin, ampicillin and chloramphenicol. The Devosia LanTb5 strain grew well between 20 and 28 • C, at pH 4-11, and tolerated 1% salt added to the medium. This strain was sensitive to ampicillin, gentamicin, tetracycline, and chloramphenicol. LanTb5 could grow in presence of spectinomycin, kanamycin and streptomycin.

Discussion
In this study, the isolation, identification and phenotypic characterization of rhizobial bacteria isolated from effective root nodules of L. angustifolius plants growing in Northern Tunisia is reported. All isolated strains were effective endosymbionts of their original host, as evidenced by the formation of red-pink nodules and the leaves' dark green color. Nodule numbers ranged from 3 to 10 per plant Phylogenetic analyses revealed high genetic diversity levels among the L. angustifolius nodule isolates. Previous reports also described considerable genetic diversity among rhizobia that nodulate Lupinus by using diverse methodological approaches, such as M13 random amplified polymorphic DNA-PCR (RAPD-PCR) [20,21,23]. High diversity was shown by phylogenetic, symbiotic and phenotypic analyses. Some rhizobia assigned to the Bradyrhizobium genus were slow-growing bacteria, while some isolates were fastgrowing and were assigned to the Microvirga and, unexpectedly, for the first time to our knowledge, to the Devosia genus. It is not unusual that legumes nodulated by slow-growing Bradyrhizobium species are also efficiently nodulated by different fast-growing rhizobial species [43,44].
The housekeeping genes glnII/recA and gyrB/dnaK were used because of their proven performance as molecular markers, either individually or in combination, for assessing the evolutionary genetics of Bradyrhizobium and Microvirga species, as described in previous works [28,45]. Phylogenetic analyses based on the housekeeping genes glnII/recA and gyrB/dnaK congruently placed some tested isolates in a new strongly supported clade within the Bradyrhizobium (Group III) and Devosia (Group V) genera. The concatenated tree of the representative strain showed that four strains affiliated with B. lupini USDA3051, one strain grouped with B. canariense BTA-1, one strain clustered together with B. liaoningense LMG18230, one strain occupied a separate position between B. valentinum LmjM3 and B. algeriense RST89, and could constitute a new species within the genus Bradyrhizobium, Three strains were included in the Microvirga genus and, surprisingly, one strain was related to the genus Devosia, never described before as a lupin microsymbiont. The concatenated phylogenetic trees constructed confirmed the results obtained by the single sequence of glnII, recA, gyrB and dnaK genes. The distribution of some strains separated from the existing type species might be indicative of novel species. Nonetheless, further polyphasic studies to precisely characterize and classify them are required.
The genus Devosia has been reported to establish nitrogen-fixing symbiosis with the aquatic legume Neptunia natans growing in India [34,46]. To our knowledge, no Devosia strain has previously been described as a lupin symbiont. In this study, two strains (LanTb4 and LanTb5) closely clustered with Devosia species. These isolates were clearly separated from other Devosia spp. and might constitute a new species under this genus. In previous works, Lupinus spp. has been reported to be nodulated by fast-growing bacteria affiliated to Mesorhizobium, Allorhizobium, Rhizobium and Sinorhizobium [24,27,30,[47][48][49]. However, in general, annual lupins are nodulated by slow-growing Bradyrhizobium sp. rhizobia [21,[50][51][52]. This is in agreement with our results, as approximately 60% of our isolates belonged to different Bradyrhizobium species.
The phylogeny of the important symbiotic gene nodC is well correlated with legume host range but independent of the rhizobial taxonomic position [53][54][55][56]. The nodC gene has been used as a molecular marker related to host specificity and range, and used to define Bradyrhizobium symbiovars [41,45,57]. The nodC tree revealed that the majority Bradyrhizobium strains clearly grouped with sequences withinthe genistearum symbiovar, which has been previously described for strains that nodulate species in the Genisteae tribe, including rhizobia nodulating Cytisus, Lupinus and Retama in Europe, America and Africa [41,45,58]. The sequences of Microvirga strains LanM6 and LanH5 grouped with sequences characteristic of the mediterraneense symbiovar. This type of nodC sequence was previously described for M. tunisiensis LmiM8 nodulating L. micranthus [42]. The remaining Bradyrhizobium-type representative isolate, LanH1, might correspond to a new symbiovar, as it was placed distant from other Bradyrhizobium symbiovars defined to date. Sanger chromatograms for nifH sequences of LanTb5 strain displayed sequence ambiguities. This may be explained by the presence of more than one copy [59].
Phenotypic tests have been used to study the diversity of rhizobial bacteria and to assess their tolerance under stress conditions [60,61]. The first macroscopic test showed that the morphology of the bradyrizobial strains in the present work was similar to those previously reported [23,60]. The colonies formed by the Microvirga strains were pinkish in color, similar to those of M. lupini [28] and M. vigna [39]. pH, salinity and extreme temperatures are abiotic stresses for nodulation of agricultural legumes in Mediterranean areas. The isolates exhibited differences in their response to abiotic stress (Table 3). All isolates grew at pH 4, in agreement with reports of other strains isolated from Lupinus spp. showing resistance to acid soils [62]. Most strains were also tolerant to alkaline conditions (pH 11), as previously reported for rhizobia that nodulate Prosopis juliflora growing in alkaline soil [63]. All strains were sensitive to salinity, as they were not able to grow when NaCl concentration was higher than 0.5%. Conversely, Microvirga strains were able to grow in with the presence of 1% sodium chloride. These results are also corroborated by those reported for Microvirga vignae [39]. Slow-growing rhizobia in the Bradyrhizobium genus have been reported to be more sensitive to salt stress than fast-growing rhizobia isolated from the same soils [61,64]. All strains in the present study grew at the optimum temperature of 28 • C, in contrast with a previous report that stated that the optimum temperature for M. lupini growth was 39 • C [28]. The effect of temperature, salinity and pH, used to qualify and quantify stresses tolerance, showed a degree of heterogeneity even when they were isolated from the same site (Table 3).
In conclusion, the isolates in this work showed a wide phenotypic and genotypic diversity. Significant heterogeneity among lupin endosymbionts was unraveled based on phylogenetic analysis using housekeeping and symbiotic gene sequences, and on their legume-host ranges. The study revealed the presence of rhizobia belonging to the three divergent genera Bradyrhizobium, Microvirga and Devosia. To our knowledge, this is the first report on the Alphaproteobacterial genus Devosia as a nodule-forming species in Lupinus spp. A practical aspect of the ability of lupin species to trap diverse rhizobial genotypes is that they might display important differences in their capacities to fix nitrogen depending on edaphoclimatic conditions [65,66]. The potentially new Bradyrhizobium and Devosia species reported here require further taxonomic characterization for confirmation and formal description. Finally, this work provides an array of native rhizobia that might have practical application as biofertilizers and/or inoculants with native legumes, including shrubs, in the restoration or revegetation of arid and semiarid Mediterranean areas.