Comparing Symbiotic Efﬁciency between Swollen versus Nonswollen Rhizobial Bacteroids 1[C][W][OA]

Symbiotic rhizobia differentiate physiologically and morphologically into nitrogen-ﬁxing bacteroids inside legume host nodules. The differentiation is apparently terminal in some legume species, such as peas ( Pisum sativum ) and peanuts ( Arachis hypogaea ), likely due to extreme cell swelling induced by the host. In other legume species, such as beans ( Phaseolus vulgaris ) and cowpeas ( Vigna unguiculata ), differentiation into bacteroids, which are similar in size and shape to free-living rhizobia, is reversible. Bacteroid modiﬁcation by plants may affect the effectiveness of the symbiosis. Here, we compare symbiotic efﬁciency of rhizobia in two different hosts where the rhizobia differentiate into swollen nonreproductive bacteroids in one host and remain nonswollen and reproductive in the other. Two such dual-host strains were tested: Rhizobium leguminosarum A34 in peas and beans and Bradyrhizobium sp . 32H1 in peanuts and cowpeas. In both comparisons, swollen bacteroids conferred more net host beneﬁt by two measures: return on nodule construction cost (plant growth per gram nodule growth) and nitrogen ﬁxation efﬁciency (H 2 production by nitrogenase per CO 2 respired). Terminal bacteroid differentiation among legume species has evolved independently multiple times, perhaps due to the increased host ﬁtness beneﬁts observed in this study.

Legume-rhizobia interactions vary widely across a diverse paraphyletic group of soil bacteria known for symbiotic nitrogen fixation inside root nodules of over 18,000 species of legumes throughout the world (Lewis et al., 2005). In several legume species, rhizobial cells are induced to swell during their differentiation into nitrogen-fixing bacteroids (Oono et al., 2010). These legume species belong to five different major papilionoid clades (inverted repeat-lacking clade, genistoids, dalbergioids, mirbelioids, and millettioids), a pattern suggestive of convergent evolution. Swelling apparently leads to terminal differentiation; swollen bacteroids no longer divide normally (Zhou et al., 1985). In other legume host species, bacteroid differentiation is less extreme, leading to nonswollen bacteroids. Nonswollen bacteroids are similar in shape and size to free-living rhizobia and divide normally once outside of their nodules. The proximate mechanisms for host-imposed bacteroid swelling have been investigated ( Van de Velde et al., 2010), but what drove the repeated evolution of this trait? The multiple independent origins of host traits causing bacteroids to swell suggest that swollen bacteroids may provide more net benefit to legumes. Could the swelling of bacteroids improve nitrogen fixation efficiency (e.g. nitrogen fixed relative to carbon cost)? In this study, we compare symbiotic efficiencies of rhizobia in legume hosts that are evolutionarily diverged but share a common effective rhizobial strain, whose bacteroids are swollen in one host and nonswollen in the other.
Variations among host species in benefits and costs of symbiosis with rhizobia are not commonly explored (Thrall et al., 2000) because legume species typically nodulate with only one group of rhizobia (e.g. Sinorhizobium sp. in Medicago), although some legumes and some rhizobia are more promiscuous. Rhizobium sp. NGR234 has the largest known host range but does not fix nitrogen effectively with any legume species currently recognized to induce swelling of rhizobial bacteroids (Pueppke and Broughton, 1999). Some Sinorhizobium fredii strains apparently fix nitrogen in certain cultivars of soybean (Glycine max; hosting nonswollen bacteroids) and alfalfa (Medicago sativa; hosting swollen bacteroids; Hashem et al., 1997), but our efforts to replicate these results did not lead to successful nodulation. Therefore, we studied two strains, a transgenic strain that nodulates beans (Phaseolus vulgaris) and peas (Pisum sativum) and a second wild strain harvested from cowpeas (Vigna unguiculata) that also nodulates peanuts (Arachis hypogaea). Beans and cowpeas are both within the Phaseolid group and do not induce terminal differentiation of rhizobial bacteroids. Peas and peanuts both host ter-1 This work was supported by the National Science Foundation (grant no. 0918986). * Corresponding author; e-mail oonox001@umn.edu. The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Ryoko Oono (oonox001@umn.edu).
[C] Some figures in this article are displayed in color online but in black and white in the print edition.
[W] The online version of this article contains Web-only data. [OA] Open Access articles can be viewed online without a subscription.
www.plantphysiol.org/cgi/doi/10.1104/pp.110.163436 minally differentiated bacteroids but are in distant clades and likely have different genetic origins for traits that induce terminal differentiation (Oono et al., 2010). Also, the swollen bacteroids in peas are branched while those in peanuts are spherical. Differences in symbiotic qualities between swollen and nonswollen bacteroids have been previously explored in peanuts and cowpeas by Sen and Weaver (1980, who also hypothesized that swollen bacteroids are more beneficial to the host plant than nonswollen ones. They found 1.5 to 3 times greater acetylene reduction by nitrogenase (as well as plant nitrogen) per nodule mass in peanuts than in cowpeas at multiple nodule ages (Sen and Weaver, 1980). Acetylene reduction per bacteroid was also greater in peanuts than in cowpeas when measuring whole nodules, but this difference disappeared when isolated bacteroids were assayed (Sen and Weaver, 1984). They concluded that swelling of peanut bacteroids per se was not responsible for the higher rate of nitrogen fixation per bacteroid. They suggested that in cowpea nodules, with greater numbers of smaller bacteroids per nodule volume, availability of oxygen to each bacteroid might be restricted such that the rate of oxidative phosphorylation, necessary for nitrogen fixation, is reduced. Fixation rates per bacteroid may be different between hosts due to nodule gas permeability or bacteroid crowding within nodules. However, fixation efficiency (nitrogen fixed per carbon respired) would not necessarily be affected by these and may be more important for the host than the rate of fixation.
Rhizobial performances are often compared by measuring the symbiotic benefits, e.g. rates of acetylene reduction or plant growth (Sen and Weaver, 1984;Hashem et al., 1997;Lodwig et al., 2005), but rarely by measuring the symbiotic costs, e.g. carbon consumed or respired. Up to 25% of a legume's net photosynthate may be required for nitrogen fixation by rhizobia (Minchin et al., 1981). Faster fixation rates (mol nitrogen per s) can be beneficial for hosts, but carbon costs can also be important. Rhizobia that fix more nitrogen per carbon respired could free more carbon for other functions, including the option of supporting more nodules with the same amount of photosynthate. If legumes are sometimes carbon limited, then improved carbon-use efficiency could enhance plant fitness. Measuring both benefits and costs is therefore key to an accurate understanding of the symbiotic performance of a rhizobial strain.
While we recognize the many physiological differences between peas and beans or peanuts and cowpeas, the fact that terminal differentiation induced by host legumes evolved multiple times independently (Oono et al., 2010) suggests there may be some consistent host symbiotic benefit, such as improved fixation efficiency. Here, we measured the efficiency of each of two strains as swollen bacteroids in one host and nonswollen bacteroids in another. We measured nitrogenase activity as hydrogen (H 2 ) production in an N 2 -free atmosphere (Layzell et al., 1984;Witty and Minchin, 1998), and compared it to carbon dioxide (CO 2 ) respiration to estimate return on nodule operation cost. We also compared host biomass growth per total nodule mass growth to estimate return on nodule construction cost. To further assess carbon allocation to the different types of bacteroids, we also measured the average amounts per bacteroid of polyhydroxybutyrate (PHB), an energy storage compound that can comprise up to 50% of bacteroid dry weight (Trainer and Charles, 2006). A greater PHB accumulation per bacteroid may require a decreased allocation of carbon for nitrogenase activity within the bacteroids, and hence, less plant growth per carbon invested in bacteroids. We demonstrate that peas and peanuts that host swollen bacteroids have higher fixation efficiency as well as greater plant return on nodule construction than beans and cowpeas, respectively, nodulated with the same rhizobial strains. PHB was not consistently correlated with plant:nodule growth efficiency with the tested strains. These findings show that swollen bacteroids can indeed provide greater benefits to their legume hosts.

Plant Return on Nodule Construction Cost in Pea versus Bean
Peas and beans nodulated with Rhizobium leguminosarum A34 were harvested periodically to obtain a range of plant sizes and their dry plant weights and nodule weights were measured. Since A34 is a transgenic strain transformed from a rhizobial strain that only nodulates beans, we compared its performance on each host with the performance of a natural strain. The relationship between nodule fresh weight and shoot dry weight was not significantly different between pea hosts nodulated by A34 versus R. leguminosarum 3841, a wild strain that only nodulates peas (Table I; Supplemental Fig. S1). The relationship was also not significantly different between bean hosts nodulated by A34 versus R. leguminosarum 4292, a wild strain that only nodulates beans (Table I; Supplemental Fig. S2). However, the relationship between nodule fresh weight and shoot dry weight was significantly different between pea and bean hosts, controlling for a constant linear effect of plant age, when each was nodulated by A34 (Table I). Dry weights of shoots and roots (g) per nodule dry weight (g) were about 5 times greater for pea than for bean (10.68 versus 2.64, Fig. 1A). During the time period in which the host plants grew the fastest, pea plants grew 1.27 g dry weight per day per g of nodule dry weight while bean plants grew only 0.27 g dry weight per day per g of nodule dry weight with A34 in each host. Since shoot nitrogen concentration did not differ significantly between peas and beans (2.7% and 2.5%, respectively, P = 0.18, n = 50), greater plant biomass indicates greater fixed nitrogen. Both pea and bean biomasses were lower than expected under field conditions since they were grown in plastic pouches in the growth chamber.

Plant Return on Nodule Construction Cost in Peanut and Cowpea
To compare plant growth per gram of nodule growth between peanuts (hosting swollen spherical bacteroids) and cowpeas (hosting nonswollen bacteroids), peanuts and cowpeas were nodulated with Bradyrhizobium sp. 32H1 and harvested intermittently across 50 d. The relationship between nodule dry weight and host dry weight was significantly different between peanuts and cowpeas, controlling for a constant linear effect of plant age (Table II). Peanuts grew about 3 times more than cowpeas in shoot and root dry weight (g) per nodule dry weight (g; 39.16 versus 12.62, Fig. 1B). This 3-fold difference between peanuts and cowpeas is consistent with results reported for multiple rhizobial strains by Sen and Weaver (1981;5.15 versus 1.5, Fig. 1B inset, t = 5.83, degrees of freedom = 7, P , 0.001, using t test on a linear model of plant weight on nodule weight with host species as a factor). During the harvest period in which the host plants grew the fastest, peanuts grew about 3.74 g per day per g of nodule dry weight while cowpeas grew about 2.01 g per day per g of nodule dry weight.
H 2 evolution is a by-product of the nitrogenase reaction, with at least 25% of nitrogenase activity going Figure 1. Peas (white circles) and peanuts (white squares) grew more per nodule mass growth than beans (black circles) and cowpeas (black squares), respectively, when nodulated with common rhizobial strain. A, Relationship between plant dry weight (including shoots and roots) and total nodule dry weight in peas and beans (P , 0.0001). B, Relationship between plant dry weight (including shoots and roots) and total nodule dry weight in peanuts and cowpeas (P , 0.0001). B (inset), Study by Sen and Weaver (1981) shows that peanut plant mass grows about 3 times greater than cowpeas with other rhizobia strains (P , 0.001). to H 2 production in ambient air (Schubert and Evans, 1976). For R. leguminosarum A34, there were detectable levels of H 2 produced in air (N 2 :O 2 ), which increased 3-fold when switched to argon:O 2 , indicating that, in air, approximately 33% of nitrogenase activity was used in H 2 production instead of nitrogen fixation. However, when sampling H 2 evolution from Bradyrhizobium sp. 32H1 in either cowpea or peanut nodules, there were no detectable levels of H 2 until the atmosphere surrounding nodules was switched to argon:O 2 . Hence, 32H1 may contain hydrogen uptake (hup) enzymes, commonly found among cowpea rhizobia (Martins et al., 1997). If 32H1 expresses hup genes, our measurements of hydrogen production may not be equally proportional to nitrogen fixation in cowpeas and peanuts because differences in nodule permeability, for example, could affect nodule-interior H 2 and therefore H 2 uptake. However, if H 2 uptake was saturated in both species by the greater H 2 production in argon:O 2 , then presence of hydrogenase would not affect marginal increase in H 2 efflux with respiration (i.e. the slope of H 2 versus CO 2 line), which we used to calculate efficiency (Fig. 2C). Differences between host nodule physiology, such as the fraction of a nodule that contains bacteroids, could also affect total fixation/ respiration, but again, such differences would only affect the baseline respiration rather than the slope of the efficiency line.

PHB Accumulation per Bacteroid
R. leguminosarum A34 bacteroids lacked PHB inside pea nodules (Fig. 4), which is a similar result seen in natural pea rhizobia, such as 3184. However, A34 accumulated high levels of PHB in beans (Fig. 4), as did 4292 (Lodwig et al., 2005) and many other natural bean rhizobia. PHB in swollen bacteroids was analyzed separately from PHB in the undifferentiated cells by distinguishing large and small cells by forward scatter in flow cytometry. Bradyrhizobium 32H1 bacteroids had low levels of PHB in both cowpeas (0.045 pg/cell 6 0.03 SD, n = 20 nodules) and peanuts (0.03 pg/cell 6 0.02 SD, n = 14 nodules) compared to beans (0.25 pg/cell 6 0.13 SD, n = 20 nodules). PHB in cowpea bacteroids was significantly lower than normally found in closely related legume species with nonswollen bacteroids, e.g. bean bacteroids can accumulate up to 0.72 pg/cell, siratro (Macroptilium atropurpureum) bacteroids accumulate on average 0.35 pg/cell (W.C. Ratcliff, unpublished data).

Peas and Peanuts with Swollen Bacteroids Can Invest Less in Nodule Construction Than Beans and Cowpeas with Nonswollen Bacteroids
Peas grew about 5 times more per nodule mass than beans when each was nodulated by R. leguminosarum A34. Peanuts grew about three times more per nodule mass than cowpeas when each was nodulated by Bradyrhizobium sp. 32H1. The greater efficiency of 32H1 in peanuts compared to cowpeas can be generalized to other strains that nodulate both species as seen by Sen and Weaver (1981;Fig. 1B, inset), who compared four different effective strains' host to nodule mass ratio, on peanuts, cowpeas, and siratro (data not shown). The study of Sen and Weaver (1981) shows that the superiority of peanut symbiosis over cowpeas was not unique to the 32H1 strain.

Peas and Peanuts Have Greater Fixation Efficiency Than Beans and Cowpeas, Respectively, with the Same Rhizobial Strain
Peas and peanuts, both hosting swollen rhizobial bacteroids, had higher nitrogen fixation efficiency (H 2 production per CO 2 respiration) than beans and cowpeas, respectively. The possible presence of hup genes in Bradyrhizobium sp. 32H1 would lead to underestimates of nitrogen fixation rates via H 2 evolution, but it would not affect efficiency estimates based on marginal rates, so long as uptake was saturated over the range used for calculating efficiency.
Similar methods have not shown consistent effects of bacteroid differentiation in the past. Our efficiency calculations are based on the approach of the late John Witty (Witty et al., 1983), who used acetylene reduction rather than hydrogen evolution to estimate nitrogenase activity. They measured fixation efficiency in 12 different legume genera and found no consistent difference between those species with swollen bacteroids and those with nonswollen ones. For example, peas ranged from 2.25 to 4.52 CO 2 :C 2 H 4 moles whereas beans ranged from 2.65 to 3.29 CO 2 :C 2 H 4 moles, depending on host cultivar and rhizobial strain. They also found no clear difference between cowpeas and peanuts (1.97 CO 2 :C 2 H 4 moles in cowpeas and 2.08 CO 2 :C 2 H 4 moles in peanuts) that were nodulated by the same strain of rhizobia RCR 3824. These comparisons by Witty et al. (1983), among others (Hunt et al., 1989), gave similar values to our H 2 :CO 2 ratios assuming 1 mol of H 2 for 1 mol of C 2 H 4 conversion. Further comparisons of single strains nodulating both hosts Figure 3. Nitrogen fixation efficiency measured as marginal increase in ratio of nitrogenase activity (mmol H 2 g 21 h 21 ) with increasing respiration (CO 2 mmol g 21 h 21 ) in pea and bean nodules nodulated by R. leguminosarum A34 (A) and in peanut and cowpea nodules nodulated by Bradyrhizobium sp. 32H1 (B). Error bars indicate one SD, with n = 3; all differences significant at * P , 0.05, ** P , 0.001.  that do or do not impose bacteroid swelling would be informative. Unfortunately, such dual-host strains are rare. Even R. leguminosarum A34 did not effectively nodulate a wide array of bean cultivars, so the experiment could not be extended even within bean. Interestingly, A34 could effectively nodulate other pea cultivars, including Green Arrow, albeit with delayed nodulation with even higher efficiency (0.70 H 2 / CO 2 6 0.18 SD, n = 3) than in our first tested cultivar, Maestro.
Greater nodule operation efficiency (N 2 fixed per CO 2 respired) often correlates with greater return on nodule construction cost (gram shoot per gram nodule). Higher fixation efficiency often correlates with the production of more plant mass relative to nodule mass. Legumes will typically continue to form nodules until they have attained an adequate nitrogen source, i.e. legumes will form many more nodules with a single less-effective strain than with a more-effective one. Hence, the per-plant construction cost of nodules is much greater for legumes when they do not find effective strains. Peas had fewer nodules per plant (157 nodules per g of plant mass) than beans (414 nodules per g of plant mass) while the average nodule for the two hosts weighed about the same (1.6 mg). Peanuts had slightly more nodules than cowpea per g plant (68 versus 53), but their much smaller size (0.3 versus 1.4 mg) still resulted in less nodule construction cost per gram shoot.

PHB in Swollen and Nonswollen Bacteroids
We hypothesized that if bacteroid PHB accumulation is less in swollen bacteroids than in nonswollen ones, this could affect differences we saw in plant: nodule growth efficiencies. PHB tends to be absent from bacteroids of peas, and other vicioid legumes with swollen bacteroids, but abundant in nonswollen bean bacteroids. But we did not see significantly less PHB in swollen peanut bacteroids compared to nonswollen cowpea bacteroids. Plant interference with bacteroid PHB synthesis, if it exists at all, is apparently neither universal nor essential for increased nitrogen fixation efficiency.

Why Do at Least Some Bacteroids Have Higher Symbiotic Efficiency?
We do not have any direct evidence for possible mechanisms whereby swollen and terminally differentiated bacteroids would have increased efficiency, but there are some possibilities that might merit further research. When bacteroids are reproductive, there are multiple bacteroids per peribacteroid unit, which may lead to some bacteroids not having any contact with the peribacteroid membrane that separates them from host cytoplasm. This might decrease nutrient exchange with the host. Swollen Y-shaped bacteroids may have polar localization allowing partitioning of metabolic functions in different areas of the cytoplasm (Young 2006), which might increase fixation efficiency. Terminally differentiated bacteroids are also known to have genomic endoreduplication and their genomic size can be commonly observed as 4C (Oono et al., 2010;Van de Velde et al., 2010). This could increase the potential rate of nitrogen fixation per bacteroid. A faster rate would not necessarily increase bacteroidlevel efficiency (nitrogen fixed per carbon respired by bacteroids). But nodule-level efficiency also depends on respiration by plant mitochondria. If endoreduplication leads to more active bacteroids and their respiration is therefore a larger fraction of total nodule respiration, then nodules containing swollen bacteroids could perhaps have more of their respiration contributing directly to nitrogen fixation, increasing nodule-level efficiency. Just as Sen and Weaver (1984) found that isolated swollen and nonswollen bacteroids did not differ in the fixation rate per bacteroid, isolated bacteroids might not differ in their fixation:respiration efficiency.
The terminal differentiation of bacteroids has been shown to be mediated by plant factors known as nodule-specific Cys-rich peptides (Van de Velde et al., 2010). These peptides interfere with normal rhizobial cell division (cytokinensis) once inside the peribacteroid membrane, which leads to a single bacteroid per peribacteroid unit. This can lead to increased rhizobial genome copies per cell if DNA synthesis was already occurring. The cells typically become larger because daughter cells cannot split off from each other during cytokinesis. This tightly links the three characteristics of (1) genomic endoreduplication, (2) single bacteroid per host cell, and (3) swelling, making it difficult to assess which characteristic might be most beneficial for the host. Furthermore, other unknown peptide effects could be the cause of a higher symbiotic efficiency.

CONCLUSION
Only a fraction of legume species host terminally differentiated rhizobial bacteroids (Oono et al., 2010). This may be because of certain trade-offs depending on environmental conditions, much like C4 photosynthesis has a greater carbon fixation efficiency than the dominant C3 systems mainly at high temperatures. Investigating bacteroid differentiation may reveal how swollen bacteroids could be more symbiotically efficient than nonswollen ones, and allow us to modify other host species to have higher nitrogen fixation efficiency as well.

Plant/Rhizobia Culture and Experimental Conditions
Two rhizobial strains were compared for their symbiotic efficiencies: (1) Rhizobium leguminosarum A34, a transgenic strain, previously studied by Lodwig et al. (2005) and Mergaert et al. (2006) among others, that nodulates beans (Phaseolus vulgaris) and peas (Pisum sativum) and (2) a wild strain, Bradyrhizobium sp. 32H1 (=USDA3384), that nodulates cowpeas (Vigna unguiculata) and peanuts (Arachis hypogaea; Sen and Weaver, 1980). Seeds of peas (cv Maestro) and beans (cv Royal Burgundy) were surface sterilized with 0.09% hypochlorite (3% commercial bleach) for 5 min, rinsed in deionized water, and inoculated with 1 mL (approximately 10 9 cells) of stationary phase R. leguminosarum 4292 , A34 (Gotz et al., 1985), or 3841 (Johnston and Beringer, 1975), which were grown in tryptone yeast media (Somasegaran and Hoben, 1994). A34 and 4292 are both derived from Rhizobium phaseoli 8002 (Lamb et al., 1982) but contain different plasmids allowing nodulation in peas (PRL1J1) or beans (PRL2J1), respectively. A34 retained ability to nodulate beans (albeit with delayed nodulation compared to 4292) and was used as the common strain to compare pea and bean host effects. 4292 only nodulates beans and 3841 only nodulates peas. Hence, these strains were used to measure natural (control) host effects on symbiotic efficiency to compare with that of the common (A34) strain. Peas and beans grew in plastic growth pouches with nitrogen-free Fahraeus nutrient media (Fahraeus, 1957) using growth conditions previously described (Ratcliff et al., 2008). Pea and bean plants were sampled at random, but their position in the growth chamber was not randomized by species. However, differences between species unrelated to bacteroid morphology were unavoidable and presumably larger than placement effects.
Seeds of peanuts (cv Starr) and cowpeas (cv California Blackeye) were surface sterilized and planted in 15-inch deep cones with sterile vermiculite: perlite (1:1) mixture. Growth chambers were set at 29°C light 16 h and 22°C dark 8 h. Peanuts and cowpeas were inoculated with Bradyrhizobium sp. 32H1, grown in modified arabinose Glu media (Somasegaran and Hoben, 1994). Since peanuts and cowpeas commonly share many strains, we did not include control strains to assess natural host effects on symbiotic efficiency. These plants were also watered with nitrogen-free Fahraeus media (Fahraeus, 1957) and mixed haphazardly throughout the growth chamber. Plant individuals were randomly chosen for harvest between days 50 and 100.

Harvesting Nodules
Peas and beans were grown for 74 d, and four to nine plants each were harvested at five different time intervals. Only plants from the first four time intervals were included in the data analysis because as the plants grew older, some leaves were lost, which underestimated the plant dry weights. Cowpeas and peanuts were harvested intermittently between day 50 and 100. All nodules were harvested from each plant and their total fresh weights were recorded. The host shoots were dried in an oven overnight and weighed. Pea and bean shoots were further processed for nitrogen content using elemental combustion analysis. Pea and bean root weights were estimated from typical shoot:root ratios estimated from a separate set of peas and beans nodulated with A34 (pea root = 0.25 3 shoot, r 2 = 0.47, n = 7, bean root = 0.64 3 shoot, r 2 = 0.77, n = 11). For peanuts and cowpeas, actual root weights were used. Nodule fresh weight was measured and dry weights were estimated by regression based on a separate experiment (pea dry nodule weight = 0.17 3 wet nodule weight, r 2 = 0.99, bean dry nodule = 0.18 3 wet, r 2 = 0.99, peanut dry nodule = 0.21 3 wet, r 2 = 0.99, cowpea dry nodule = 0.25, r 2 = 0.99).

Nodules for H 2 :CO 2 Efficiency Measurements
Healthy, mature, pink nodules were harvested and used immediately for measuring H 2 production and CO 2 respiration in nitrogen-free air, using a method adapted from Witty and Minchin (1998)'s open-flow-through system. Witty and Minchin (1998) showed respiratory cost (mol CO 2 /mol ethylene from acetylene reduction by nitrogenase) remained relatively constant with plant age for detached nodules, so nodules of various ages were pooled as long as they looked healthy. These nodules were detached from plants but still connected to some root fragments to minimize wounding or introduction of ambient oxygen into nodule interior. Detached nodules have lower fixation and respiration rates but the relationship between them apparently does not change (Witty and Minchin, 1998). Detached nodules were pooled from two or more individual host plants to obtain detectable levels of H 2 production. Peanut and cowpea plants were acclimated to a cooler growth chamber of 20°C for 24 h before harvesting at room temperature (20°C) to reduce the temperature shock for the nodules.
Productions of H 2 and of CO 2 by nodules (total fresh weight between 0.5 g and 1 g) were assayed in a flow-through chamber ( Fig. 2A). Argon:O 2 flowed through the nodules from below at a controlled rate (100 mL/min). Due to the absence of N 2 , 100% of nitrogenase activity went to H 2 production. A subsample of gas from above the nodules was pulled through an H 2 analyzer (Witty and Minchin, 1998) and infrared gas analyzer for CO 2 (Qubit Systems) at a lower flow rate than the supply rate, with excess gas vented ( Fig. 2A). After equilibrating under 21% O 2 in argon, oxygen percentage was increased in steps to 33%. Increments of oxygen increases varied from 1% (v:v, i.e. 1 kPa) every 4 min to 3% every 10 min depending on nodule sensitivity to O 2 (Fig.  2B). When external O 2 partial pressure is increased gradually, internal O 2 partial pressure will rise enough to reduce O 2 limitation of nodule interior respiration (increasing nitrogenase activity as well as respiration) but not enough to damage nitrogenase irreversibly (Denison et al., 1992;Witty and Minchin, 1998). Efficiency was calculated from linear regression of nitrogenase activity (H 2 production) on respiration (CO 2 production; Fig. 2C; Witty et al., 1983). If nitrogenase was damaged at high external oxygen concentrations, then when external oxygen was returned back to 21% at the end of the experiment from 33%, there would be a definitive drop in H 2 evolution from the initial 21% reading (Fig. 2C). Hence, CO 2 and H 2 values at some high oxygen percentages were not included in the regression (Witty and Minchin, 1998) unless they remained linear. The H 2 sensor was calibrated at 0.5 mL L 21 (ambient air) and at 50 mL L 21 (H 2 standard).

Comparing PHB in Bacteroids
We also measured the amount of a carbon storage polymer, PHB, per bacteroid in each host since PHB is known to compete with nitrogen fixation for photosynthates (Trainer and Charles, 2006, and refs. therein). We tested whether there is less PHB in the swollen bacteroids than in the unswollen bacteroids, which might help explain any differences in symbiotic efficiency between hosts, if they are present.
Five to eight random nodules from each host plant were weighed individually and saved for PHB analysis via flow cytometry. Procedures for PHB analysis are described by Ratcliff et al. (2008).

Statistics
To test if the effect of nodule weight on plant weight was significantly different between the two strains on the same host species or between two host species with the same strain, controlling for a constant effect of plant age, we compared linear models using a t test in R. Formula: plant weight = b + m 3 (nodule weight) + I (strain:host species combination) + plant age + I (strain or host species) 3 m 3 (nodule weight), where I designates indicators for the model, b is the intercept of the model, and m is the coefficient of the nodule weight variable or the slope of the linear regression that was estimated. Slopes indicate plant growth per nodule growth, and allow discounting differences in initial seed weights. Fixation efficiencies were compared with unpaired equal variance two-tailed t test.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Pea plants nodulated by R. leguminosarum A34 and 3841 across four harvest dates (four plants each for 3841 and four to nine plants each for A34).
Supplemental Figure S2. Bean plants nodulated by R. leguminosarum A34 and 4292 across four harvest dates (four to six plants each).