R. leguminosarum Rlv13 and Rlv21 were isolated from V. sativa root nodules at Zaidín Estacion Experimental (Granada, Spain). M. loti U261 (=NZP2037, Irisarri et al., 1996) was obtained from the Department of Scientific and Industrial Research, Palmerston South, New Zealand. Rhizobium and Mesorhizobium strains were grown at 28°C in tryptone-yeast extract (TY), yeast-manitol broth (YMB), or Rhizobium minimal media (Rmin) (Bascones et al., 2000). Escherichia coli strains were grown at 37°C in Luria-Bertani medium supplemented with the corresponding antibiotics. Minitransposon TnHB100, carrying the entire R. leguminosarum hup cluster was transferred into (Meso)Rhizobium strains by conjugation as previously described (Bascones et al., 2000). For bacterial conjugations, fresh cell cultures of E. coli S17.1 λpir(pTnHB100) donor (Simon et al., 1983) and rhizobial recipient strains were mixed on TY plates and incubated overnight at 28°C. Transconjugants were selected on Rmin plates supplemented with 200 μg/ml of spectinomycin.

Seeds of L. corniculatus cv. La Estanzuela, L. tenuis, L. subbiflorus, and L. uliginosus were surface sterilized with 10% sodium hypochlorite and pregerminated at 28°C for 2 days in the dark. V. sativa cv. Filón seeds were sterilized with ethanol/sodium hypochlorite treatment and pregerminated as previously described (Brito et al., 2000). Three-day old seedlings were sown in pots containing vermiculite supplied with a nitrogen-free nutrient solution (Handberg and Stougaard, 1992; Brito et al., 1994) and inoculated with liquid cultures of rhizobial strains. Plants were maintained in a growth chamber under controlled conditions (16/8 h light/dark; Monza et al., 1992; Brito et al., 2000). Bacteroid supensions were prepared from freshly detached nodules as previously described (Brito et al., 2008).

Plasmid DNA preparations, restriction enzyme digestions, DNA cloning, transformation of DNA into E. coli, agarose gel electrophoresis, Southern hybridizations, and PCR amplifications were performed by standard methods (Sambrook and Russell, 2002). Total DNA from rhizobial strains was isolated as described (Bascones et al., 2000). For the determination of nucleotide sequence flanking the TnHB100 insertions, the region was first cloned as a PstI restriction fragment in a pBluescript vector plasmid (Agilent Technologies, La Jolla, CA), and the DNA sequence was determined using a primer complementary to the 3′ end of the spectinomycin resistance cassette (5′-GCTGGCTTTTTCTTGTTATCG-3′). Transposon insertions in Rlv13H9 and Rlv21H19 were located in genes encoding potential galactonate dehydratase/epimerases highly similar to protein RLV_6446 from R. leguminosarum bv viciae (Sanchez-Canizares et al., 2018). These proteins are moderately expressed in pea and lentil bacteroids induced by Rlv UPM791 (Duran et al., 2021), and might function in carbon metabolism pathway, so a potential effect of the gene inactivation on bacterial performance cannot be excluded. In contrast, insertion in Rlv13H6 was located in an intergenic region, with no similarity to described genes, whereas Rlv21H10 carried TnHB100 insertion in a hypothetical protein not expressed in bacteroids as deduced from a recent proteomic analysis (Duran et al., 2021). In the case of Ml U261H4, the site of insertion was located in an intergenic region downstream of a potassium uptake gene (kup) located at positions 2366770-2368683 of the genome (Kelly et al., 2014).

The removal of spectinomycin cassette from Rlv13 and Rlv21 TnHB100-containing derivatives was carried out as described (Bascones et al., 2000). Briefly, DNA regions flanking the resistance marker on each strain were obtained and fused to hupS upstream region by restriction cloning in pK18mobSac suicide vector (Schäfer et al., 1994). Then, the plasmid was conjugated into the strains carrying the TnHB100 insertion, and a double recombination event excising Spc^R marker was obtained based on the selection provided by sacB gene.

Nitrogenase activity was determined by the acetylene reduction assay, and hydrogenase activity was estimated by measuring bacteroid hydrogen uptake amperometrically, or hydrogen evolution from nodules by gas chromatography (Brito et al., 2000, 2008) using detopped nodulated roots from 35 day-old plants. Plant dry weight was measured after maintaining shoots in an oven at 80°C for 48 h. Nitrogen content of Lotus plants was determined from detached leaves collected after 40, 60 and 70 days growing period and analyzed for total nitrogen with a mass spectrometer (IRMS Micromass Isochrom) at SIDI- Laboratorio de Isótopos Estables (Montevideo, Uruguay). Nitrogen content was calculated according with IAEA (IAEA, 2001). Nitrogen content of Vicia shoots was determined by a combustion method using a nitrogen analyzer (FP-528-LECO). Experiments were established in a completely random design with at least three replicates. Data were analyzed using ANOVA and means were compared using Fisher's least significant difference (LSD) test at 5% probability level.

In order to analyze the potential effect of hydrogen recycling in symbiotic performance of forage legumes, the R. leguminosarum hydrogenase gene cluster was introduced into strains effective in Vicia sativa (R. leguminosarum bv viciae Rlv13 and Rlv21) and in L. corniculatus (Mesorhizobium loti U261, a broad-range Lotus inoculant strain). Stable insertion of hup gene package into the recipient strains was carried using transposon TnHB100. Following conjugation with E. coli S17.1λpir(pTnHB100), spectinomycin-resistant derivatives arose at frequencies of ca. 10⁻⁴/recipient (Rlv strains) or 10⁻⁸ /recipient (Ml U261). The presence of hup genes in spectinomycin-resistant transconjugants was confirmed through identification of hupS-hybridizing bands in Southern experiments (Supplementary Figure 1) or through PCR amplification of specific bands corresponding to hypB gene using adequate primers (data not shown).

The ability of transposants to effectively recycle hydrogen evolved by nitrogenase was assessed in several derivatives per each recipient strain. In the case of strains effective in V. sativa, four derivatives obtained from each of Rlv13 and Rlv21 strains were analyzed for hydrogenase and nitrogenase activities in vetch nodules (Table 1). With one exception, all transconjugants from Rlv strains showed high levels of nitrogenase activity. The exception was the Rlv13 derivative Rlv13H24, which showed a poor symbiotic performance, attributed to a potential deleterious effect of the insertion. Although there were differences in the levels of hydrogenase activity induced in each case, in both sets of strains we found derivatives with levels of hydrogenase activity high enough as to fully recycle the hydrogen evolved by nitrogenase, thus resulting in nodules with no detectable hydrogen evolution: Rlv13H6 and H9, and Rlv21H10 and H19 (Table 1). Insertion sites were determined for these four derivatives (see Materials and Methods), and derivatives Rlv 13H6 and Rlv21H10, bearing insertions with no predicted effect on bacteroid physiology were selected for the analysis of the effect of hydrogen recycling on vetch productivity.

Transposon TnHB100 carries a Spc^R determinant whose presence in bacteria designed to be used as inoculant in the field is not desirable. In order to generate marker-free derivatives more acceptable for field release, strains Rlv13H6 and Rlv21H10 were further engineered for the removal of Spc^R determinant by a double recombination event carried out as described in Materials and Methods. The absence of the spectinomycin resistance cassette in the resulting strains was confirmed by Southern hybridization of genomic DNA using a probe internal to the Spc^R gene (Supplementary Figure 1). Both strains, designated as Rlv13H6DS and Rlv21H10DS, induced nodules with levels of nitrogenase activity similar to those in the strains bearing the Spc^R determinant. Bacteroids from these strains showed high levels of hydrogenase activity that fully recycled the nitrogenase-derived hydrogen (data not shown).

In the case of M. loti U261 derivatives, 6 transconjugants were selected to assess the expression of the introduced hup genes in symbiosis with L. corniculatus as the legume host. Each of the transconjugants was used as inocula for L. corniculatus plants, and the levels of nitrogenase activity and hydrogen evolution from nodules were determined (Table 1). Levels of nitrogenase activity in the transconjugant strains did not show significant differences with that from the wild type. Nodules induced by the hup-containing derivatives showed significant reductions of the level of hydrogen evolution as compared to nodules induced by the wild-type strain (Table 1). The complete suppression of hydrogen evolution, indicating full recycling of nitrogenase-produced hydrogen, was detected in one of the transconjugants (Ml U261H4). Inspection of the corresponding genomic region (see Materials and Methods section) revealed that this insertion is likely in an intergenic region not affecting the expression of any other gene.

In order to determine the effect of the hydrogen recycling on symbiotic performance of the vetch/R. leguminosarum bv viciae system, V. sativa plants were inoculated with the wild type strains Rlv13 and Rlv21 and with the hup-engineered derivatives obtained from them and described above. Inoculated plants were maintained in a growth chamber, and shoot dry matter and nitrogen accumulation were determined 30 days after inoculation. In this analysis, hydrogen-recycling strains derived from Rlv13 showed significant increases (p < 0.05) in N fixed per plant (Table 2). Values of shoot dry weight also showed marked increases (24.1% and 21.7 %, respectively) although in this case the higher variability of the data led to a lower significance of the differences (p < 0.1). In the case of strain Rlv21, derivatives incorporating the ability of full hydrogen recycling did not show significant variations of any of the two productivity parameters analyzed. These data suggest that the rhizobial background is relevant for the expression of the benefits associated to hydrogen recycling.

Symbiotic performance of M. loti U261 and its derivative strain U261H4 were compared in L. corniculatus plants grown under controlled conditions. Plant shoot dry weight, and N content of the shoots were determined at 40, 60, and 70 days after inoculation. 60- and 70-day old plants inoculated with the engineered strain showed levels of nitrogen accumulation significantly higher than those inoculated with the wild-type strain. This result was consistently observed in three independent experiments carried out under controlled chamber conditions. The increase in nitrogen was accompanied with an increase in dry weight at 70 days (Table 3). At 40 days plants showed higher values of N accumulation, but the variability of the data made these differences not statistically significant.

Since, M. loti U261 is effective in other Lotus species of agronomical interest (Gonnet and Díaz, 2000), we also estimated the potential benefits of hydrogen recycling carried out by M. loti U261H4 when in symbiosis with other 3 hosts of the same genus, namely L. tenuis, L. subbiflorus, and L. uliginosus. Similarly to what was observed in L. corniculatus, nodules induced by M. loti U261H4 engineered strain showed undetectable levels of hydrogen evolution in symbiosis with these 3 hosts, whereas nodules induced by the wild type strain Ml U261 evolved large amounts of hydrogen in all cases (data not shown). These results indicate that R. leguminosarum hup system is expressed under symbiotic conditions in all host partners at levels high enough as to fully recycle hydrogen evolved by nitrogenase. Despite full hydrogen recycling, there were no significant differences between the wild type and U261H4 engineered strain neither in shoot dry weight nor in N content in symbiosis with L. tenuis, L. subbiflorus, L. uliginosus at 60 days of cultivation (Table 4), indicating a relevant role for the host on the effect of hydrogen-recycling on nitrogen fixation.

Forage legumes constitute a relevant input of protein for the sustainable feeding of livestock, a crucial issue in a context of increased demand of livestock products (FAO, 2011; Michalk et al., 2019). The productivity of pastures is dependent on an adequate supply of nitrogen (Phelan et al., 2015), therefore ensuring optimal functioning of symbiotic nitrogen fixation is a relevant aspect for pasture management. This optimization includes the use of highly efficient rhizobial strains. Even the most efficient rhizobial strains still dedicate a minimum of 25% of electrons going through nitrogenase to reduce protons to hydrogen. In this work we have generated rhizobial strains effective in two forage legumes (Lotus and Vicia) and able to fully recycle the hydrogen evolved from nitrogenase activity in the nodules. Engineered strains were associated to levels of nitrogen fixation higher than the respective wild-type strains under controlled conditions. It has to be noted, however, that for L. tenuis and L. subbiflorus the data showed a higher level of variability which, along with the limited number of replicates, did not allow to adscribe statistical significance to the observed variation of the mean values.

Previous evidence had shown that R. leguminosarum hydrogenase system is expressed in nodules when the hydrogenase gene cluster is introduced into strains effective on Vicia and Lotus (Monza et al., 1997; Brito et al., 2000). In contrast, introduction of the same system into Sinorhizobium meliloti strains resulted in only residual levels of hydrogenase activity in alfalfa nodules. It was concluded that several factors, including nickel provision and effectiveness of heterologous regulators, limit the expression in different backgrounds (Brito et al., 2000). The original transposon contains a spectinomycin resistance cassette to help on the selection of transposants (Bascones et al., 2000). We have also generated derivative strains in which such marker has been deleted, thus resulting in engineered strains in which only rhizobial DNA was present. As expected, removal of Spc^R tag did not change the symbiotic performance of the strains. These strains should be more amenable to its use as inoculants.

The results presented in this work indicate that the introduced ability of hydrogen recycling has the potential to increase nitrogen fixation in forage legumes, confirming the notion of hydrogen evolution as a source of inefficiency of nitrogen fixation by the Rhizobium-legume symbiosis (Evans et al., 1988). There was scarce previous information on the effect of hydrogen recycling on forage legumes. The presence of hydrogenase genes in rhizobia is highly variable depending of the system (Ruiz-Argüeso et al., 2000), and in fact there are no reports on the presence of functional hup gene clusters in rhizobia isolated from forage legumes such as clover or alfalfa. Even in R. leguminosarum bv viciae the presence of hydrogenase system is highly uncommon (Fernandez et al., 2005), and so the actual evolutionary advantage of possessing such system is a matter of controversy. Now we demonstrate for the first time that the advantages of hydrogen recycling can be exploited also in forage legumes, at least under controlled conditions. The results obtained also indicate that both the host species and the rhizobial strain condition the existence of significant increases of fixed nitrogen associated to hydrogenase activity. Factors proposed to account for the beneficial effect of hydrogen recycling are the protection of nitrogenase against oxygen and against the detrimental effect of hydrogen, and the availability of additional energy for nitrogen fixation (Palacios et al., 2005). Additional work is required to elucidate the reasons for the observed effect of both the rhizobial background and host legume. It has to be noted that nickel provision and effectiveness of heterologous regulators for expression of hydrogenase genes have been cited before as potential limiting factors (Brito et al., 2000).

The role of hydrogen evolution and hydrogen recycling in plant productivity is not a simple question. Results presented here refer to the levels of nitrogen fixed by the inoculated legume, and have been obtained in an artificial substrate without the soil microbiome. From a more integrated point of view, the positive role of hydrogen recycling in the productivity of legumes must be considered along with the potential effect of hydrogen evolved from nodules in the fertility of surrounding soil. It has been shown that the presence of hydrogen evolving nodules in legume plants inoculated with hydrogenase-negative rhizobia results in an increase of the ratio of H₂-oxidizing bacteria in soil surrounding nodules, as could be expected from the presence of an additional energy substrate in the environment. In fact, a “hydrogen fertilization effect” has been proposed as a factor contributing to the observed benefit of legumes on the following crop in rotation practices (Golding and Dong, 2010). The direct addition of hydrogen to soils resulted in increases in tiller/plant and plant dry weight in barley (Dong et al., 2003). There is also one report indicating that barley crop following soybean inoculated with Hup⁻, hydrogen evolving B. japonicum strain resulted in higher grain yield than in the case of isogenic, hydrogen recycling strain (Dean et al., 2006). In that work an increase of hydrogen oxidation activity in soils surrounding Hup⁻ nodules was also detected. Changes in soil microbiome around nodules producing hydrogen are expected, and it has been recently shown that exposure of soil to hydrogen levels similar to those on the soil-nodule interface indeed causes changes in soil microbiome structure and function, including the stimulation of carbon turnover (Khdhiri et al., 2017). All these data indicate that the potential role of hydrogen metabolism in plant productivity should be analyzed from an integrated perspective. The hydrogen-recycling strains effective in forage legumes developed in this work can be used to further study the effect of hydrogen metabolism on mixed grass-legume pastures in which the potential role of hydrogen in stimulating other crops can be analyzed in the same season as legume symbiosis occurs, so a more integrated study of the effect of nitrogen fixation on mixed pasture production can be carried out.