Search tips
Search criteria 


Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
PLoS One. 2010; 5(8): e12094.
Published online 2010 August 10. doi:  10.1371/journal.pone.0012094
PMCID: PMC2930871

Symbiotic Legume Nodules Employ Both Rhizobial Exo- and Endo-Hydrogenases to Recycle Hydrogen Produced by Nitrogen Fixation

Malcolm James Horsburgh, Editor



In symbiotic legume nodules, endosymbiotic rhizobia (bacteroids) fix atmospheric N2, an ATP-dependent catalytic process yielding stoichiometric ammonium and hydrogen gas (H2). While in most legume nodules this H2 is quantitatively evolved, which loss drains metabolic energy, certain bacteroid strains employ uptake hydrogenase activity and thus evolve little or no H2. Rather, endogenous H2 is efficiently respired at the expense of O2, driving oxidative phosphorylation, recouping ATP used for H2 production, and increasing the efficiency of symbiotic nodule N2 fixation. In many ensuing investigations since its discovery as a physiological process, bacteroid uptake hydrogenase activity has been presumed a single entity.

Methodology/Principal Findings

Azorhizobium caulinodans, the nodule endosymbiont of Sesbania rostrata stems and roots, possesses both orthodox respiratory (exo-)hydrogenase and novel (endo-)hydrogenase activities. These two respiratory hydrogenases are structurally quite distinct and encoded by disparate, unlinked gene-sets. As shown here, in S. rostrata symbiotic nodules, haploid A. caulinodans bacteroids carrying single knockout alleles in either exo- or-endo-hydrogenase structural genes, like the wild-type parent, evolve no detectable H2 and thus are fully competent for endogenous H2 recycling. Whereas, nodules formed with A. caulinodans exo-, endo-hydrogenase double-mutants evolve endogenous H2 quantitatively and thus suffer complete loss of H2 recycling capability. More generally, from bioinformatic analyses, diazotrophic microaerophiles, including rhizobia, which respire H2 may carry both exo- and endo-hydrogenase gene-sets.


In symbiotic S. rostrata nodules, A. caulinodans bacteroids can use either respiratory hydrogenase to recycle endogenous H2 produced by N2 fixation. Thus, H2 recycling by symbiotic legume nodules may involve multiple respiratory hydrogenases.


Legume root and stem nodules fix atmospheric dinitrogen (N2) yielding anabolic-N, which augments growth and reproduction of host plants. In these nodules, the biochemical conversion of N2 to ammonium is owed to endosymbiotic rhizobia (bacteroids) who carry the N2 fixation genes encoding the dinitrogenase complex. Whether N2 fixation occurs in legume nodules [1] or in pure cultures of diazotrophic (able to use N2 as sole N-source) bacteria [2], hydrogen gas (H2) is then co-produced. From subsequent mechanistic studies of dinitrogenase activity, H2 co-production is both stoichiometric and requires 2 ATP per H2 formed [3][4]. Yet in agronomic surveys, many legume nodules typically evolve H2 at high levels, and such H2 evolution rates correlate with N2 fixation rates [5]. However, in certain symbiotic legume nodules, bacteroids avidly fix N2 yet reproducibly evolve little or no H2 [1]. As this endogenous H2 production consumes metabolic energy, H2 recycling, which recoups that energy, allows increased efficiency of N2 fixation and, in principle, increased plant biomass yields [6][7]. This symbiotic nodule H2 recycling capability correlates with specific bacteroid strains, although host legume cultivars also contribute to H2 recycling and yield [8], [9]. Indeed, in biochemical assays, bacteroids isolated from H2 recycling (non-evolving) nodules show high levels of respiratory uptake hydrogenase activity [10], [11].

H2 recycling during N2 fixation was first observed with the aerobe Azotobacter vinelandii, a diazotroph but not a legume symbiont. In pure culture, A. vinelandii induces a particulate (respiratory) hydrogenase activity which oxidizes H2 at the expense of and tolerant of O2 [2]. In following studies with legume nodule bacteroids, such uptake hydrogenase activity was also affirmed [10]. In the ensuing forty years, hydrogenases, extensively studied, have proven both biochemically diverse and broadly distributed across bacteria and archaea [12]. Among aerobes and microaerophiles able to use H2 as energy source, uptake hydrogenase activities are typically classified as group I: heterodimeric, globular, hydrophilic proteins carrying a heteronuclear Ni,Fe-catalytic center; group I hydrogenases are generally O2 tolerant [12]. In cellular terms, the group I, Ni,Fe uptake hydrogenases are tightly associated with respiratory membranes via integral diheme b-type cytochromes, required for physiological activity [13], [14] [Bernhard]. As the group I cell membrane-peripheral complexes face the periplasm, or cell exterior [15], they may be termed exo-hydrogenases.

Azorhizobium caulinodans, a microaerophilic α-proteobacterium originally isolated as nodule endosymbiont of the host legume Sesbania rostrata, is capable of N2 fixation both in planta and in pure diazotrophic culture [16]. Recently, we discovered in A. caulinodans a second, novel respiratory hydrogenase encoded by the seven-gene hyq operon [17]. The inferred Hyq hydrogenase includes six different structural proteins, including a heterodimeric Ni,Fe-catalytic center hydrogenase conserved with group I enzymes. From bioinformatic analyses, the remaining four Hyq proteins are all membrane-integral [17]. Because all six Hyq hydrogenase subunits are NADH:quinone oxidoreductase (respiratory complex I) homologs [18], the Hyq complex is classified with the reversible group IV hydrogenases [12]. Given structural and functional homology to respiratory complex I [18], the Ni,Fe-catalytic center heterodimers of group IV complexes associated with respiratory membranes presumably face the cell-interior and thus may be termed endo-hydrogenases.


In symbiotic legume nodules both exo- and endo-hydrogenases recycle H2 produced by N2 fixation

To assess physiological roles for both bacteroid Hup exo- and Hyq endo-hydrogenases in symbiotic legume nodules fixing N2 and recycling H2, A. caulinodans haploid derivatives carrying precise (to the nucleotide pair) in-frame deletions of hup and hyq structural genes encoding the conserved catalytic subunits of, respectively, exo- and endo-hydrogenases were constructed and verified by nucleotide sequencing of mutant loci [17]. Specifically, A. caulinodans exo-hydrogenase null mutants carried in-frame, precise, complete hupSL deletions; endo-hydrogenase null mutants comprised precise, complete hyqRBCEFGI operon deletions. As well, haploid recombinant double-mutants carrying both exo- and endo-hydrogenase null alleles were also constructed (Methods). Pure A. caulinodans cultures were used to inoculate both stems and roots of S. rostrata seedlings aseptically germinated and individually cultivated under N-limitation (Methods). In S. rostrata, symbiotic nodules are developmentally determinate, not meristematic. While both stem- and root-nodules subsequently developed on inoculated plants only, as they are invariably absent on uninoculated plants, stem nodules were chosen for further study. Three week-old and five week-old determinate stem nodules were excised from inoculated plants and individually tested for N2 fixation activity, assaying acetylene-dependent ethylene production by gas chromatography with flame ionization detection (Methods). Excised stem nodules all showed similar (±15%) high levels of acetylene reduction activity when normalized per fresh nodule biomass (Table 1). Accordingly, all A. caulinodans strains tested were assigned both nodulation-competent (Nod+) and N2 fixation-competent (Fix+) phenotypes.

Table 1
N2 fixation and H2 recycling in S. rostrata–A. caulinodans stem nodules.

Additional excised nodules from these S. rostrata plants were simultaneously tested under air for H2 evolution activity using gas chromatography coupled to a reducing-compound photometric detector (Methods). In kinetic studies with excised nodules elicited by A. caulinodans strains 61305R (parental), 66081 (exo-hydrogenase mutant) and 66132 (endo-hydrogenase mutant), H2 evolution was nonexistent (Figs. 1a,b). Whereas, nodules elicited by double (exo- and endo-hydrogenase) mutant 66204 evolved H2 at very high rates (Fig. 1a) comparable to those measured for acetylene reduction (Table 1). Thus, H2 evolution by double-mutant 66204-elicited nodules was quantitatively owed to N2 fixation (dinitrogenase) activity. Results with five week-old determinate nodules from additional S. rostrata plants entirely corroborated results with three week-old nodules (data not presented). Pure bacterial cultures were reestablished from aseptically crushed nodules and strain identities verified by nucleotide sequencing of hup and hyq loci (Methods). In conclusion, A. caulinodans bacteroids in S. rostrata nodules employ both exo- and endo-hydrogenases to recycle endogenous H2 produced by N2 fixation. Moreover, H2 recycling is quantitative, entirely accounting for N2 fixation activities. Yet as measured by H2 evolution rates, bacteroid exo- and endo-hydrogenase are interchangeable and individually are fully competent to handle endogenous H2 recycling in symbiotic S. rostrata nodules.

Figure 1
Hydrogen (H2) evolution by excised S. rostrata stem nodules elicited by indicated A. caulinodans strains as endosymbiont.

N2 fixing, microaerophilic α-proteobacteria able to recycle H2 carry exo- and endo-hydrogenase gene-sets

From bioinformatic analyses (Table 2), orthologous hyq+ operons encoding endo-hydrogenase are generally present in N2 fixing microaerophiles able to recycle endogenous H2. These strains include both free-living diazotrophs as well as certain rhizobia, such as B. japonicum, the endosymbiont of Glycine max (soy). In Rhizobium leguminosarum, a metastable species with several descendant biovars each with genomes comprised of variable multipartite replicons, H2 recycling capability in symbiotic legume nodules varies among strains. As well, both the hup+/hyp+ (exo-hydrogenase) and the hyq+ (endo-hydrogenase) gene-sets are also variables [9], [19], [20]. Yet other diverse rhizobia (e.g. Sinorhizobium meliloti 1021; Mesorhizobium loti MAFF303099; Rhizobium etli CFN42; Rhizobium sp. NGR234) all incapable of H2 recycling in symbiotic legume nodules, completely lack both hup+/hyp+ and hyq+ gene-sets (Table 2). As the hyq+ operon is also absent from anaerobic (fermentative) diazotrophs, fully aerobic diazotrophs (e.g. Azotobacter spp.), and non-diazotrophs generally, Hyq endo-hydrogenase seems co-selected with N2 fixation in microaerophilic (non-fermentative) α-proteobacteria. Nevertheless, in every N2 fixing microaerophile with both exo- and endo-hydrogenases, these gene-sets, as well the nif genes encoding N2 fixation activities are all unlinked (Table 2). Moreover, A. caulinodans haploid strains carrying complete (20-gene) hup+/hyp+ (including hupSL+) operon deletions entirely lacking exo-hydrogenase and ancillary activities, nevertheless retain full H2 recycling activity both in pure cultures and in S. rostrata stem nodules. As well, Rhodocista centenaria (aka Rhodospirillum centenum) SW, which possesses the hyq+ operon but not the hup+/hyp+ operon (Table 2), completely recycles H2 in diazotrophic culture (data not presented). Accordingly, these exo- and endo-hydrogenase gene-sets seem fully autonomous.

Table 2
N2-fixing microaerophilic α-proteobacteria carrying orthologous hup+/hyp+ (exo-hydrogenase) and hyq+ (endo-hydrogenase) genes.


Among legume-Rhizobium symbioses, H2 recycling was first reported in Pisum sativum (garden pea) nodules elicited by specific Rhizobium leguminosarum bv. viciae strains [2]. Genetic studies were subsequently undertaken with [Brady]Rhizobium japonicum strains able to recycle H2 in Glycine max (soy) nodules [21], [22]. Many subsequent studies with H2 recycling legume nodules all presumed uptake hydrogenase activity a single entity. These studies include combined genetic and physiological analyses which might have challenged this assertion. For the case of A. caulinodans, single mutants W58, U58 as well as hupSL impaired strain ORS571.2 all were reported to suffer substantial to complete loss of uptake hydrogenase activity [23], [24], [25]. Such conclusions are incompatible with the present finding: A. caulinodans employs two structurally and functionally distinct, genetically-independent, respiratory hydrogenases to recycle endogenous H2 produced by N2 fixation.

Whereas, early on the investigative timeline, B. japonicum single mutants unable to be cultured autotrophically on exogenous H2 yet still able to recycle endogenous H2 in soy nodules were identified [26]. As these strains showed induction of uptake hydrogenase activity in cultures shifted to O2 limitation (≤11 µM DOT), they were perhaps understandably considered transcriptional control mutants hypersensitive to O2. With the benefit of hindsight, this phenotype is precisely that expected of true loss-of-function point mutants affecting hup operon structural genes encoding Hup exo-hydrogenase activity, were the observed limiting-DOT uptake hydrogenase activity in fact owed to Hyq endo-hydrogenase. In A. caulinodans, hyq operon expression requires NifA as transactivator [17], and the pnifA+ promoter is in turn strongly transactivated by Fnr, which process requires physiological O2 limitation in diazotrophic culture [27]. In principle, both exo- and endo-hydrogenase gene-sets, despite being encoded at disparate loci in all organisms identified, might nevertheless share a common genetic predisposition, allowing strategic single mutations to convey dual loss-of-function. However, as strains carrying complete hyq operon deletions still possess wild-type Hup exo-hydrogenase activity, and vice versa, evidence for any genetic, post-transcriptional interaction or interdependence between the two gene-sets is entirely lacking.

As shown previously, in pure diazotrophic (N2 as sole N-source) cultures, A. caulinodans exo-hydrogenase knockout mutants grow as wild-type, whereas endo-hydrogenase knockout mutants exhibit slow growth [17]. Are exo- and endo-hydrogenase H2 recycling efficiencies in pure culture and in legume nodules then demonstrably different? Or, do diazotrophic phenotypes imply additional endo-hydrogenase function(s), e.g. chemiosmotic work associated with membrane ion translocation [28] not undertaken by exo-hydrogenase? Obviously, effective exo- and endo-hydrogenase cellular concentrations and/or distributions might be dissimilar in legume nodules and in pure diazotrophic cultures, even though both hup+/hyp+ (exo-hydrogenase) and hyq+ (endo-hydrogenase) gene-sets are then strongly transcribed [17], [25], [29]. Because hup mutants suffer loss of chemoautotrophy with exogenous H2 as energy substrate [17], [26], exo-hydrogenase kinetic behavior may constitute simple diffusion control. Because hyq mutants do not adversely impact chemoautotrophy with exogenous H2, endo-hydrogenase kinetic behavior might not constitute simple diffusion control. A critical test of this hypothesis is still lacking. Diazotrophic liquid batch cultures typically employ constant sparging with relatively high gas-phase exhaust rates (0.5 min−1), complicating kinetic behavior and analysis of cellular processes with gaseous substrate(s) subject to simple diffusion control. In such pure liquid diazotrophic batch cultures bacterial densities typically reach 108 cc−1, whereas in determinate S. rostrata nodules, bacteroid densities approach 1011 cc−1, the latter obviously more conducive to endogenous H2 recycling under simple diffusion control. Notwithstanding, given their apparent co-selection in N2 fixing micoaerophilic α-proteobacteria capable of H2 recycling, exo- and endo-hydrogenases likely possess additional, distinctive functionalities yet to be elucidated.


Bacterial strains and culture conditions

Azorhizobium caulinodans ORS571 wild-type (strain 57100), originally isolated from Sesbania rostrata stem-nodules [16], was cultured as previously described [30]. As 57100 wild-type is a pyridine nucleotide auxotroph, to serve as ‘virtual’ wild-type, all experiments reported here employ A. caulinodans 61305R, a 57100 derivative carrying an IS50R insertion in the (catabolic) nicotinate hydroxylase structural gene. Precise, in-frame deletion mutants were constructed by a ‘crossover PCR’ method [31]. Haploid exo-hydrogenase knockout mutants each carry a hupΔSL2 allele in which the (upstream) hupS translation initiation codon is fused in-frame to a synthetic 21np linker sequence fused in-frame to the (downstream) hupL termination codon. Similarly, haploid endo-hydrogenase mutants each carry a hyqΔRI7 allele, in which the hyqRBCEFGI operon has been replaced by a deletion allele comprising the hyqR initiation codon fused in-frame to the 21np linker sequence fused in-frame to the hyqI termination codon. After gene replacement, haploid strains carrying deletion alleles were verified by PCR and DNA sequencing analyses [17].

Sesbania rostrata nodulation tests

S. rostrata plants were germinated, cultivated aseptically, and stem inoculated with pure A. caulinodans strain cultures as described [16]. Either three or five weeks post-inoculation, stem nodules were detached, weighed, individually placed in septated vials. Dinitrogenase activity was assayed kinetically by acetylene reduction [32] and product ethylene was measured by gas chromatography with flame-ionization detection. H2 evolution was assayed kinetically and measured by gas chromatography with reducing compound photometer detection (RCP1; Peak Laboratories LLC, Mountain View, CA.), both at atmospheric pressure and 29°C [33]. Enzymatic activities are expressed per gram nodule fresh-biomass at 29°C.


The authors thank Henk Vreman (Stanford University), Steve Hartman and Alex Lowe (Peak Laboratories LLC) for technical assistance with gas chromatographic H2 analyses.


Competing Interests: The authors have declared that no competing interests exist.

Funding: This work was supported by the California Energy Commission. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


1. Phelps AS, Wilson PW. Occurrence of hydrogenase in nitrogen-fixing organisms. Proc Soc Exp Biol. 1941;47:473–476.
2. Hyndman LA, Burris RH, Wilson PW. Properties of hydrogenase from Azotobacter vinelandii. J Bacteriol. 1953;65:522–531. [PMC free article] [PubMed]
3. Thorneley RNF, Lowe DJ. in Molybdenum Enzymes; In: Spiro TG, editor. New York: Wiley-Interscience; 1985. pp. 221–284.
4. Burgess BK, Lowe DJ. Mechanism of molybdenum nitrogenase. Chem Rev. 1996;96:2983–3011. [PubMed]
5. Schubert KR, Evans HJ. Hydrogen evolution: A major factor affecting the efficiency of nitrogen fixation in nodulated symbionts. Proc Natl Acad Sci USA. 1976;73:1207–1211. [PubMed]
6. Albrecht SL, Maier R J, Hanus FJ, Russell SA, Emerich DW, et al. Hydrogenase in Rhizobium japonicum increases nitrogen fixation by nodulated soybeans. Science. 1979;203:1255–1257. [PubMed]
7. Eisbrenner G, Evans HJ. Aspects of hydrogen metabolism in nitrogen-fixing legumes and other plant-microbe associations. Ann Rev Plant Physiol. 1983;34:105–136.
8. Bedmar EJ, Phillips DA. Pisum sativum cultivar effects on hydrogen metabolism in Rhizobium. Can J Bot. 1984;62:1682–1686.
9. López M, Carbonero V, Cabrera E, Ruiz-Argüeso T. Effects of host on the expression of the H2-uptake hydrogenase of Rhizobium in legume nodules. Plant Sci Lett. 1983;9:191–199.
10. Dixon ROD. Hydrogenase in legume root nodule bacteroids: occurrence and properties. Arch Mikrobiol. 1972;85:193–201. [PubMed]
11. Dixon ROD. Hydrogenase in pea root nodule bacteroids. Arch Mikrobiol. 1968;62:272–283. [PubMed]
12. Vignais PM, Billoud B, Meyer J. Classification and phylogeny of hydrogenases. FEMS Microbiol Rev. 2001;25:455–501. [PubMed]
13. Cauvin B, Colbeau C, Vignais PM. The hydrogenase structural operon in Rhodobacter capsulatus contains a third gene, hupM, necessary for the formation of a physiologically competent hydrogenase. Mol Microbiol. 1991;5:2519–2527. [PubMed]
14. Bernhard M, Benelli B, Hochkoeppler A, Zannoni D, Friedrich B. Functional and structural role of the cytochrome b subunit of the membrane-bound hydrogenase complex of Alcaligenes eutrophus. Eur J Biochem. 2004;16:179–186. [PubMed]
15. Burgdorf T, Lenz O, Buhrke T, van der Linden E, Jones AK, et al. [NiFe]-hydrogenases of Ralstonia eutropha H16: modular enzymes for oxygen-tolerant biological hydrogen oxidation. J Mol Microbiol Biotechnol. 2005;10:181–196. [PubMed]
16. Dreyfus BL, Dommergues YR. Nitrogen fixing nodules induced by Rhizobium on strains of the tropical legume Sesbania rostrata. FEMS Microbiol Lett. 1981;10:313–317.
17. Ng G, Tom CGS, Park AS, Zenad L, Ludwig RA. A novel endo-hydrogenase activity recycles hydrogen produced by nitrogen fixation. PLoS ONE. 2009;4(3):e4695. doi: 101371/journalpone0004695. [PMC free article] [PubMed]
18. Sazanov LA, Hinchcliffe P. Structure of the hydrophilic domain of respiratory complex I from Thermus thermophilus. Science. 2006;311:1430–1436 (2006). [PubMed]
19. Ruiz-Argüeso T, Hanus J, Evans HJ. Hydrogen production and uptake by pea nodules as affected by strains of Rhizobium leguminosarum. Arch Microbiol. 1978;116:113–118.
20. Nelson LM, Grosskopf E, Tichy HV, Lotz W. Characterization of hup-specific DNA in Rhizobium leguminosarum strains of different origin. FEMS Microbiol Lett. 1985;30:53–58.
21. Lepo JE, Hickok RE, Cantrell MA, Russell SA, Evans HJ. Revertible hydrogen uptake deficient mutants of Rhizobium japonicum. J Bacteriol. 1981;146:614–620. [PMC free article] [PubMed]
22. Drevon JJ, Frazier L, Russell SA, Evans HJ. Respiratory and nitrogenase activities of soybean nodules formed by hydrogen uptake negative (Hup) mutant and revertant strains of [Brady]Rhizobium japonicum characterized by protein patterns. Plant Physiol. 1982;70:1341–1346. [PubMed]
23. De Vries W, Ras J, Stam H, van Vlerken MMA, Hilgert U, et al. Isolation and characterization of hydrogenase-negative mutants of Azorhizobium caulinodans ORS571. Arch Microbiol. 1988;150:595–599.
24. Boogerd FC, Ferdinandy-Van Vlerken M, Mawadza C, Pronk AF, et al. Nitrogen fixation and hydrogen metabolism in relation to the dissolved oxygen tension in chemostat cultures of the wild type and a hydrogenase-negative mutant of Azorhizobium caulinodans. App Env Microbiol. 1994;60:1859–1864. [PMC free article] [PubMed]
25. Baginsky C, Palacios JM, Imperial J, Ruiz-Argüeso T, Brito B. Molecular and functional characterization of the Azorhizobium caulinodans ORS571 hydrogenase gene cluster. FEMS Microbiol Lett. 2004;237:399–405. [PubMed]
26. Maier RJ, Merberg DM. [Brady]Rhizobium japonicum mutants that are hypersensitive to repression of H2 uptake by oxygen. J Bacteriol. 1982;150:161–167. [PMC free article] [PubMed]
27. Loroch AI, Nguyen B, Ludwig RA. FixLJK and NtrBC signals interactively regulate Azorhizobium nifA transcription via overlapping promoters. J Bacteriol. 1995;177:7210–7221. [PMC free article] [PubMed]
28. Ohnishi T, Salerno JC. Conformation-driven and semiquinone-gated proton-pump mechanism in the NADH-ubiquinone oxidoreductase (complex I). FEBS Lett. 2005;579:4555–4561. [PubMed]
29. Brito B, Martinez M, Fernandez M, Rey L, Babrera E, et al. Hydrogenase genes from Rhizobium leguminosarum bv viciae are controlled by the nitrogen fixation regulatory protein NifA. Proc Natl Acad Sci USA. 1997;94:6019–6024. [PubMed]
30. Donald RGK, Nees D, Raymond CK, Loroch AI, Ludwig RA. Three genomic loci encode [Azo]Rhizobium sp ORS571 N2 fixation genes. J Bacteriol. 1986;165:72–81. [PMC free article] [PubMed]
31. Link AJ, Phillips D, Church GM. Methods for generating precise deletions and insertions in the genome of wild-type Escherichia coli: application to open reading frame characterization. J Bacteriol. 1997;179:6228–6237. [PMC free article] [PubMed]
32. Burris RH, Arp DJ, Benson DR, Emerich DW, Hageman RV, et al. The biochemistry of nitrogenase, In: Stewart WDP, Gallon JR, editors. London: Academic Press; 1980. pp. 37–54. Nitrogen Fixation,
33. Vreman HJ, Mahoney JJ, Van Kessel AL, Stevenson DK. Carboxyhemoglobin as measured by gas chromatography and with the IL282 and 482CO-oximeters. Clin Chem. 1988;34:2562–2566. [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science