GS52 ecto-apyrase plays a critical role during soybean nodulation.

Apyrases are non-energy-coupled nucleotide phosphohydrolases that hydrolyze nucleoside triphosphates and nucleoside diphosphates to nucleoside monophosphates and orthophosphates. GS52, a soybean (Glycine soja) ecto-apyrase, was previously shown to be induced very early in response to inoculation with the symbiotic bacterium Bradyrhizobium japonicum. Overexpression of the GS52 ecto-apyrase in Lotus japonicus increased the level of rhizobial infection and enhanced nodulation. These data suggest a critical role for the GS52 ecto-apyrase during nodulation. To further investigate the role of GS52 during nodulation, we used RNA interference to silence GS52 expression in soybean (Glycine max) roots using Agrobacterium rhizogenes-mediated root transformation. Transcript levels of GS52 were significantly reduced in GS52 silenced roots and these roots exhibited reduced numbers of mature nodules. Development of the nodule primordium and subsequent nodule maturation was significantly suppressed in GS52 silenced roots. Transmission electron micrographs of GS52 silenced root nodules showed that early senescence and infected cortical cells were devoid of symbiosome-containing bacteroids. Application of exogenous adenosine diphosphate to silenced GS52 roots restored nodule development. Restored nodules contained bacteroids, thus indicating that extracellular adenosine diphosphate is important during nodulation. These results clearly suggest that GS52 ecto-apyrase catalytic activity is critical for the early B. japonicum infection process, initiation of nodule primordium development, and subsequent nodule organogenesis in soybean.

There are two major categories of apyrases: ectoapyrases that typically have an extracellular catalytic domain (Plesner, 1995) and endo-apyrases with an intracellular catalytic domain on the inside face of the cell membrane (Komoszynski and Wojtczak, 1996). For example, in yeast, two endo-apyrases are required to regulate the glycosylation of N-and O-linked oligosaccharides in the Golgi lumen (Abeijon et al., 1993;Gao et al., 1999). These endo-apyrases, encoded by the gda1 and ynd1 genes, control the turnover of GDP (released by hydrolysis of GTP sugar) to GMP (Abeijon et al., 1993;Gao et al., 1999). In animals, ecto-apyrases have several important physiological roles such as involvement in neuron signaling (Sarkis and Salto, 1991;Plesner, 1995;Komoszynski and Wojtczak, 1996), blood platelet aggregation (Marcus and Safier, 1993), and ATP-mediated immunoresponses (Virgilio, 1998). For instance, animal ecto-apyrases play a critical role at the synaptic junction of nerve cells where degrada-tion of extracellular ATP to AMP occurs (Sarkis and Salto, 1991;Komoszynski and Wojtczak, 1996). The AMP activates 5#-nucleotidase, an enzyme abundant in the synaptic space, releasing adenosine, which reenters the cell and restores the cellular ATP pool (Komoszynski and Wojtczak, 1996). Therefore, ectoapyrase plays a key role during synaptic junction activity.
In plants, endo-apyrases have been characterized in soybean (Day et al., 2000), potato (Solanum tuberosum; Kettlun et al., 2005), and pea (Shibata et al., 2002). For instance, in potato, an endo-apyrase was suggested to be involved in the regulation of several key steps during starch synthesis (Handa and Guidotti, 1996). In pea, a nuclear-localized apyrase was reported to be stimulated by red light (Chen and Roux, 1986) and mediated by calmodulin in a calcium-dependent manner (Chen et al., 1987) or casein kinase II (Hsieh et al., 2000). This apyrase was expressed both in light-and dark-grown pea roots, but more strongly expressed in dark-grown plumules and stems (Hsieh et al., 1996).
Plant ecto-apyrases have been proposed to play several roles, including phosphate transport and mobilization (Thomas et al., 1999), toxin resistance (Thomas et al., 2000), and cytoskeleton-based cellular metabolism (Shibata et al., 1999). Early reports have suggested that this cytoskeleton-associated apyrase may be involved during signal transduction on the cytoskeleton (Komoszynski and Wojtczak, 1996) or transported to other locations through the cytoskeleton (Davies et al., 1996). Thomas et al. (1999) showed that transgenic Arabidopsis plants expressing the psNTP9 ecto-apyrase exhibited enhanced growth in comparison to wild-type plants when supplied with exogenous ATP as an inorganic phosphate source, suggesting that this apyrase may play an important role in the uptake of inorganic phosphate from the extracellular matrix. A role for ecto-apyrases in nodulation has previously been proposed. For example, in D. biflorus, an ecto-apyrase originally isolated from roots as a unique lectin (DB46) was named a lectinnucleotide phosphohydrolase (Db-LNP; Etzler et al., 1999). Although showing no significant sequence similarity to classical lectins, this Db-LNP/apyrase was shown to bind to the lipo-chitin Nod factor, produced by Rhizobium and essential for nodulation. Using antibodies, Db-LNP was localized to the epidermal cell surface of young roots, predominantly on the root hair surface, the primary site of rhizobial infection. Pretreatment of roots with anti-LNP serum inhibited root hair deformation and nodulation (Kalsi and Etzler, 2000). These data argued strongly for a role for Db-LNP during the early nodulation response . In M. truncatula, four putative apyrase genes were identified (Mtapy1-4); two of which (Mtapy1 and Mtapy4) showed evidence of increased mRNA levels (within 3-6 h) when roots were inoculated with Sinorhizobium meliloti, while mRNA levels of Mtapy2 and Mtapy3 remained unaffected by rhizobial inoculation (Cohn et al., 2001). In soybean, two apyrases, GS50, an endo-apyrase localized in the Golgi, and GS52, an ecto-apyrase localized to the plasma membrane, were identified and partially characterized (Day et al., 2000). Semiquantitative reverse transcription (RT)-PCR showed that GS52, but not GS50, was induced within 6 h of inoculation with Bradyrhizobium japonicum, suggesting a possible role for this ectoapyrase during early nodulation. Roots treated with antibody directed against GS52 or GS50 blocked soybean nodulation only with the former (Day et al., 2000). A phylogenetic analysis also showed that the GS52 apyrase, the Db-LNP, and other legume orthologs belonged to an apparent legume-specific clade, suggesting a unique function for these enzymes in legumes Cannon et al., 2003). More recently, McAlvin and Stacey (2005) reported that overexpression of soybean GS52 ecto-apyrase in L. japonicus enhanced infection by Mesorhizobium loti and increased nodulation, again suggesting an important role during nodulation. Despite several reports indicating apyrase function in legumes, the specific role for GS52 ecto-apyrase during nodulation in soybean remains unknown. We therefore investigated the role of GS52 ecto-apyrase using a RNA interference (RNAi) approach during nodulation in soybean (Glycine max).
RNA silencing or posttranscriptional gene silencing is a powerful tool for the analysis of gene function in plants (Waterhouse and Helliwell, 2003). RNAi using Agrobacterium rhizogenes-mediated root transformation is an efficient method to silence genes in legume roots (e.g. M. truncatula; Limpens and Bisseling, 2003;and L. japonicus;Kumagai and Kouchi, 2003). In this study, we show that RNAi-mediated silencing of the GS52 ecto-apyrase results in severe suppression of nodule primordium development and maturation. Addition of exogenous ADP (100 mM) to the GS52 silenced roots restores mature nodule formation. These data suggest that the catalytic activity of the GS52 ecto-apyrase, which would result in ADP release, likely plays a beneficial role during nodulation. Our data clearly indicate that GS52 ecto-apyrase is important for the early B. japonicum infection process, nodule primordium development, and subsequent maturation revealing a novel role for GS52 during nodulation in soybean.

Differential Expression of Two Apyrase Genes in Soybean Tissues
Soybean contains two distinct apyrase genes, GS50 and GS52, with 73% sequence identity (Day et al., 2000). As a result, we used quantitative real-time-PCR (qRT-PCR) analysis, with gene-specific primers, to measure the expression of GS50 and GS52 mRNA in different soybean tissues. The qRT-PCR was performed with total RNA isolated from 10 different soybean tissue types: root tips, stripped roots (roots devoid of root hairs), root hair cells, mature whole roots, mature stems, leaves, apical meristems, flowers, young pods, and green pods (Fig. 1). The expression levels for GS50 and GS52 genes were normalized with a constitutively expressed reference gene cons6, an F-box protein (Libault et al., 2008). GS50 mRNA levels were higher than those of GS52 in all tissues tested except in the root tips, where the expression levels were similar. GS50 mRNA was weakly expressed in the root tips, but strongly expressed in mature whole roots, root hair cells, mature stems, leaves, apical meristems, flowers, young pods, and green pods and very strongly expressed in stripped roots (Fig. 1). GS52 mRNA was strongly expressed in 18-d-old roots. Lower levels of GS52 expression were observed in root hair cells, root tip, and stripped root samples. In contrast to GS50, GS52 mRNA was detected only in the stem and not in the other aerial tissues of the soybean plant.

GS52 Ecto-Apyrase Is Induced during Nodulation
To determine the role of apyrase during nodulation in soybean, we used qRT-PCR analysis to assess GS50 and GS52 mRNA levels during the later stages of infection by B. japonicum. The qRT-PCR analysis was performed with total RNA isolated from uninoculated roots and roots inoculated with B. japonicum at 0, 4, 8, 16, 24, and 32 d postinoculation (dpi; Fig. 2). The expression of ENOD40 mRNA, an early nodulin gene, was also measured as a positive control because it shows rapid accumulation in response to B. japonicum inoculation (Kouchi and Hata, 1993;Yang et al., 1993;Asad et al., 1994;Papadopoulou et al., 1996). The inoculated and uninoculated soybean roots may be in different metabolic states, with the inoculated roots nodulating and fixing nitrogen, while the uninoculated roots are in a nitrogen-deprived state. Therefore, soybean encoding cons6 was used as a reference gene (Libault et al., 2008) to normalize the expression levels of GS50, GS52, and ENOD40 in both inoculated and uninoculated roots. The normalized values from inoculated roots were compared against uninoculated roots to give the ratio of expression levels for GS52, GS50, and ENOD40 (Fig. 2). The GS52 and ENOD40 mRNA levels were strongly induced during nodulation. Our experiments showed that the GS52 mRNA levels were not induced at 3, 6, 12, 24, and 48 h postinoculation (data not shown). However, GS52 expression levels increased at 8, 16, and 24 dpi and continued to remain high 32 dpi. GS50 mRNA levels increased only during the later stages of nodulation Figure 1. Expression profile of apyrase GS52 and GS50 mRNA levels in different soybean tissues. Expression levels of GS52 and GS50 mRNA were measured by real-time PCR analysis from apical meristem at 14 d; trifoliate leaves, stem, and roots at 18 d; flowers, seeds, and pods at R6 stage; and root hair cells, stripped roots, and root tips at 3 d postgermination in soybean plants. Soybean encoding Cons6 was used as a reference gene (expression level = 1) to normalize the expression of GS52 (black bars) and GS50 (gray bars). The error bars represent the SEs of three independent biological replicates and two technical repeats. Figure 2. Ratio of GS52, GS50, and ENOD40 expression levels in soybean roots after inoculation with B. japonicum. Using real-time PCR analysis, the expression levels for GS52, GS50, and ENOD40 were measured in inoculated roots and uninoculated roots at 4, 8, 16, 24, and 32 dpi. The uninoculated roots were mock inoculated with water. ENOD40 was used as an internal control. Soybean encoding Cons6 was used as a reference gene to normalize the expression levels of GS52 (black bars), GS50 (gray bars), and ENOD40 (white bars) in inoculated roots and in uninoculated roots. Each bar represents the SEs of three independent biological replicates and two technical repeats. The expression levels for GS52, GS50, and ENOD40 in inoculated roots were normalized and compared to uninoculated roots and were significantly different (* = P , 0.05; ** = P , 0.01; *** = P , 0.001 using Student's t test). and were not as highly expressed as the GS52 gene at 24 and 32 dpi. These results suggest GS52, previously reported as an early nodulin (Day et al., 2000), may also play a major role during later stages of nodule development.

Reduction in GS52 Gene Expression Affects Nodule Development
To enhance our understanding of the role of GS52 during nodulation, we chose to silence GS52 mRNA expression using RNAi in roots transformed with A. rhizogenes. The empty vector control and RNAi constructs (GUS control and GS52 ecto-apyrase) driven by the constitutive figwort mosaic virus (FMV) promoter were expressed in composite soybean plants (Collier et al., 2005;Govindarajulu et al., 2008). Transgenic roots were identified using a scorable GFP marker driven by the superubiquitin promoter (Collier et al., 2005;Govindarajulu et al., 2008). The expression of GFP indicated that 50% to 60% of the roots produced from the inoculated stem were transformed. Phenotypic characterization of composite plants expressing the empty vector control, RNAi GUS control, and RNAi GS52 gene showed no obvious differences (i.e. no statistically significant differences in the number of transgenic GFP roots formed per shoot, the length or width of the transgenic roots produced, or the number of lateral roots per transgenic root [data not shown]). However, nodule number per transgenic root varied according to construct tested. Roots transformed with the empty vector control produced 6.9 6 0.7 pink, fully formed nodules per transgenic root (Fig. 3, A and D). The RNAi GUS control exhibited 6.6 6 0.7 fully formed nodules per transgenic root (Fig. 3, B and E).
Roots expressing the RNAi GS52 gene exhibited a dramatic decrease in the formation of fully formed pink nodules (0.1 6 0.01 fully formed nodules per transgenic root; Fig. 3F). Along with numerous nodule primordia, we observed the formation of small bumps emerging on the root surface from the root cortex. These emerging nodules were translucent and devoid of bacteria. Both nodule primordia and emerging but empty nodules were grouped together under the small empty nodule category. These small empty nodules in the GS52 silenced transgenic roots failed to develop into mature pink nodules (14.4 6 0.9 small empty nodules formed per transgenic root; Fig. 3, C and F) upon extended incubation. It is likely that this ob- Figure 3. Nodulation profile in transgenic soybean GS52 knock-down roots after inoculation with B. japonicum. A to C, Fourweek-old GFP-expressing root nodule phenotypes for empty vector (A; control); RNAi GUS (B; control); and RNAi GS52 apyrase on soybean hairy roots (C). Inset, The RNAi GS52 roots exhibited very poor nodulation (as small bumps or empty nodules), whereas the controls show fully matured nodules as indicated by white arrowheads. SE, Small empty nodule. D to F, Phenotypic nodulation analysis in empty vector (control), RNAi GUS (control), and RNAi GS52 apyrase on soybean hairy roots. Small empty nodules (gray bar) were counted as small bumps just emerging on the root surface from the root cortex and as clearly visible small translucent nodules. The data represent average of 24 individual plants (per biological replicate) containing transgenic nodulated roots. SE bars are shown for total number of mature nodules and small empty nodules of three independent biological replicates. served decrease in fully formed nodules is primarily due to the inhibition of nodule development after nodule primordia are formed. There was no apparent alterations in the morphology of leaves, stems, or roots on the RNAi GS52 composite plants other than the aberrant nodule morphology.
Addition of ADP to RNAi GS52 Roots Rescues Mature Nodule Formation Due to the predicted catalytic activity of GS52, we investigated the effect of exogenous application of nucleotide phosphates on transgenic silenced GS52 soybean roots during nodulation. Figure 4, A and B, shows the effect of the addition of various concentrations of exogenous ADP and ATP to nodulation on wild-type soybean roots, whereas control wild-type roots were mock inoculated with autoclaved water. Exogenous addition of 100 mM ADP to wild-type roots showed the highest number of mature nodules formed (18.2 6 0.3) in comparison to the control roots (7.6 6 0.14), whereas addition of 5, 10, 25, 50, and 250 mM ADP to wild-type roots gave a similar number of root nodules (7.2 6 0.14, 7 6 0.14, 7 6 0.13, 8.4 6 0.1, and 8 6 0.16 nodules per root, respectively; Fig. 4A) to that found on control roots (7.6 6 0.14 nodules per root). However, exogenous addition of 125, 150, 175, and 200 mM ADP to wild-type roots had a small effect on nodulation (17.9 6 0.3, 16.8 6 0.4, 15.2 6 0.4, and 11.6 6 0.3 nodules per root, respectively; Fig. 4A) in comparison the control roots (7.6 6 0.14 nodules per root). In similar experiments, exogenous addition of 100 mM ADP to the wild-type roots of L. japonicus Gifu and M. truncatula A17 wild-type roots resulted in higher nodule numbers (Supplemental Fig. S1). Addition of exogenous 10, 50, 100, 150, and 200 mM ATP to wild-type roots gave a similar number of mature root nodules (7.3 6 0.11, 6.2 6 0.1, 7.7 6 2.3, 7.2 6 0.13, and 5.9 6 0.07 nodules per root, respectively; Fig. 4B) to that found on control roots (7.6 6 0.14 nodules per root). Therefore, exogenous 100 mM ADP was chosen to study its effect on nodulation in silenced GS52 roots. Addition of 100 mM ADP to either the empty vector control or RNAi GUS control transgenic roots 48 h after inoculation with B. japonicum did not change the nodule number or nodule size (5.9 6 0.4 and 6.4 6 0.3 mature nodules per transgenic root, respectively; Fig.  5A). However, addition of exogenous 100 mM ADP to RNAi GS52 transgenic roots partially rescued mature nodule formation (3.6 6 0.4 mature nodules per transgenic root; Fig. 5A). Addition of exogenous 100 mM AMP to either the wild-type, empty vector control or RNAi GUS control transgenic roots showed no effect on nodule size or number, while application of AMP to RNAi GS52 roots did not have any effect on the aberrant nodule phenotype observed in the silenced GS52 (data not shown). Additionally, exogenous application of 100 mM ATP to empty vector control or RNAi GUS control transgenic roots did not produce any change in the nodule number or size (5.8 6 0.3 and 6.6 6 0.3 nodules per transgenic root, respectively; Fig.  5B), while application of ATP to RNAi GS52 roots had no effect on the aberrant nodule phenotype observed with the silenced GS52 (9.2 6 0.7 small empty nodules per transgenic root; Fig. 5B). These results suggest that ADP is a critical component for nodule development in soybean roots.

GS52-Infected Roots
To verify silencing of GS52, we analyzed GS52 expression levels in the transgenic roots. In addition, to ensure sequence-specific silencing of GS52 and not GS50, we quantified GS50 mRNA levels. Figure 6, A to C, shows the transcript levels of GS52, GS50, and ENOD40 measured by qRT-PCR from total RNA isolated from empty vector control, RNAi GUS control, and RNAi-GS52 roots after inoculation with B. japonicum. Soybean encoding cons6 was used as a reference gene (Libault et al., 2008) to normalize the expression levels of GS50, GS52, and ENOD40 in empty vector control, RNAi GUS control, and RNAi GS52 nodulated roots. As expected, the transcript levels of GS52 were significantly reduced in the RNAi GS52 roots, whereas GS52 expression in RNAi GUS control roots remained unaffected when compared to empty vector control (Fig. 6A). However, GS50 expression was also altered in RNAi GS52 roots and in RNAi GUS roots in comparison to GS50 transcript levels in empty vector controls (Fig. 6B). Although the RNAi construct was designed for GS52 sequence, the levels of GS52 and GS50 were both significantly reduced in the RNAi GS52 transgenic roots. It is possible that the high sequence similarity between GS52 and GS50 (i.e. identity in the coding region sequence is 73%) resulted in silencing of GS50. However, it is also possible that, similar to the results for ENOD40 (Fig. 6C), the physiological effects of GS52 silencing may also have led to lower GS50 expression. This may also explain why GS50 expression was slightly reduced in RNAi GUS roots. Given that GS50 was slightly down-regulated in roots transgenic for RNAi GUS (Fig. 6B) and yet these same roots were unaffected in nodule formation (Fig.  3E), we conclude that down-regulation of GS50 alone has no effect on nodulation. In RNAi GS52 nodulated roots, the expression level of ENOD40 was reduced when compared with the empty vector control nodulated roots (Fig. 6C). These results likely reflect the poor nodulation response on the GS52 silenced roots.
The transcript levels of GS52, GS50, and ENOD40 were also measured in RNAi GS52 transgenic roots treated with exogenous 100 mM ADP. Interestingly, ENOD40 mRNA levels in silenced roots were restored by exogenous 100 mM ADP application (Fig. 6C), consistent with the restoration of the nodulation response. However, the reduction of expression of GS50 and GS52 was not restored by the addition of ADP (Fig. 6, A and B). The GS52 Ecto-Apyrase Is Essential for Mature Nodule Formation To determine whether GS52 is required for nodule development, nodules formed on empty vector con-trol, RNAi GUS control, and RNAi GS52 roots were analyzed using light microscopy and transmission electron microscopy (TEM). Longitudinal sections of the mature pink nodules from empty vector control, RNAi GUS control, and the small translucent nodules from RNAi GS52 roots were cytologically analyzed. In the RNAi GUS and empty vector nodules, the infected host cells were populated with mature bacteroids within a symbiosome (Fig. 7, A and B, black arrowheads). However, the RNAi GS52 nodule sections revealed that all the cells in the infected zone of the nodular tissue were devoid of bacteroids, containing only infection thread stubs (Fig. 7C). TEM studies also showed that, within the infected region of the mature root nodules (empty vector and RNAi GUS controls), several symbiosomes enclosing mature bacteroids (two to four) that contained poly-b-hydroxybutyrate crystals were clearly visible (Fig. 8, A and B). In the RNAi GS52 nodule sections, TEM studies confirmed the lack of functional bacteroids within a symbiosome and a few released bacteria remained inside the dead plant cell (Fig. 8C). Because GS50 expression was reduced in the RNAi GUS plants (Fig. 6B), but nodule numbers and nodule ultrastructure were not affected (Figs. 3E and 8B), these data argue that reduced GS50 transcript levels do not adversely affect nodulation. Interestingly, light microscopy and TEM analyses of RNAi GS52 nodules after the addition of exogenous ADP also displayed well-developed bacteroids (two to four) surrounded by a symbiosome (Figs. 7D and 8D), which was very similar to that seen in the empty vector control and RNAi GUS control nodules (Fig. 8,  A and B). These data clearly indicate that reduction in GS52 expression levels in the roots interfered with bacteroid development within the nodule and, therefore, demonstrate that GS52 is essential for normal nodule development.

DISCUSSION
The previous study by Day et al. (2000) identified and partially characterized two distinct soybean apyrase genes, GS50 and GS52. Using semiquantitative RT-PCR, they reported that GS52, but not GS50, was transcriptionally induced within 6 h of inoculation in soybean roots, thus identifying GS52 as a possible early nodulin. In addition, McAlvin and Stacey (2005) demonstrated that transgenic L. japonicus constitutively expressing soybean apyrase GS52 not only doubled nodule number, but also increased root infections by M. loti. To expand upon these earlier studies, we silenced GS52 expression in soybean roots by means of transient expression of double-stranded RNA using hairy root transformation mediated by A. rhizogenes. GS52 silencing resulted in a significant reduction in both GS52 and GS50 expression, likely due to the high level of sequence identity between these two genes. However, the data argue that it is silencing of GS52 expression that was responsible for the nodule phe- Figure 6. Expression levels of GS52, GS50, and ENOD40 using genespecific primers in soybean silenced hairy roots 4 weeks after inoculation with B. japonicum. A to C, GS52 (A), GS50 (B), and ENOD40 (C) mRNA levels in nodulated roots for RNAi GUS (control), RNAi GS52 apyrase, and RNAi GS52 apyrase after exogenous ADP (100 mM) treatment (as represented in the x axis). The y axis represents the fold change (log 2 ratio) in comparison to the empty vector (control). Soybean encoding Cons6 was used as a reference gene to normalize the expression levels of GS52, GS50, and ENOD40 measured using real-time PCR analysis. Each bar represents the SEs of three experimental repeats each evaluating 25 independent nodulated roots. Asterisks (*) indicate statistically significant differences compared to empty vector control (Student's t test; * = P , 0.05).
notypes seen. Also, previous studies (Day et al., 2000;McAlvin and Stacey, 2005) clearly implicated GS52 and not GS50 as playing an important role both in soybean and L. japonicus nodulation. For example, as shown in Figure 2 as well as earlier studies by Day et al. (2000), only GS52 mRNA levels increased significantly upon inoculation with B. japonicum. Day et al. (2000) also demonstrated that the addition of anti-GS52 antibodies to soybean roots, but not anti-GS50 antibodies, blocked nodulation. Moreover, reduction of GS50 expression in RNAi GUS control plants (Fig.  6B) did not result in an alteration in the nodulation phenotype (Fig. 3, B and E), which would be expected if GS52 played an important role in nodulation.
With respect to nodulation phenotypes, our data demonstrated a significant decrease in the formation  of mature nodules in GS52 knock-down roots. Control roots showed fully formed, mature nodules. RNAi knock-down of GS52 by hairy root transformation led to the formation of numerous small empty nodules instead of fully developed pink nodules. McAlvin and Stacey (2005) showed that transgenic L. japonicus plants overexpressing apyrase GS52 had significantly more infections, thus suggesting that GS52 plays a role during the early infection events. Complete nodule formation in the silenced GS52 roots was strongly inhibited during the nodule primordia stage resulting in small empty nodules. Such nodule phenotypes have been reported in a number of legume symbiotic mutants disrupted in genes involved in early nodulation events (Kumagai et al., 2006). For example, Kumagai et al. (2006) demonstrated that silencing of ENOD40, an early nodulin, significantly suppressed nodule primordium initiation and subsequent nodule development in the transgenic L. japonicus roots. In nodulationdeficient mutant lines of M. truncatula, apyrase expression was severely reduced, consistent with defects in bacterial infection and the formation of nodule primordia (Cohn et al., 2001). Taken together, our results suggest that apyrase GS52 plays a key role in the early nodulation response. The lack of efficient infection in the RNAi GS52 roots likely results in the subsequent defects seen in nodule primordium development and the deviant nodule ultrastructure observed. Our microscopy analysis also showed GS52 silencing resulted in aberrant nodules with impaired bacteroid development. Such aberrant nodule phenotypes have also been observed in other early nodulindeficient mutant lines (Wan et al., 2007).
The predicted protein structure for GS52 and its localization to the soybean plasma membrane predicts that GS52 is an ecto-apyrase, with its ATPase catalytic domain positioned to be extracellular matrix (Day et al., 2000). Kim et al. (2006) clearly demonstrate the presence of extracellular ATP at the root tips of M. truncatula using a novel cell wall-bound luciferase. Demidchik et al. (2003) demonstrate that the addition of exogenous ATP can increase intracellular calcium levels in root hairs. It is well known that an increase in cytoplasmic calcium is an essential component during the early stages of rhizobial infection of the root hair (Cohn et al., 1997). Our results suggest that the GS52 ecto-apyrase may act to hydrolyze extracellular ATP at the soybean root hair surface. The ADP formed by this activity thus appears to be beneficial during nodulation in soybean roots by B. japonicum. Indeed, addition of exogenous ADP enhanced nodulation in control roots and largely reversed the effects of RNAi GS52 silencing. The specific mechanism of ADP action remains to be elucidated. However, extracellular ATP clearly exists in plants and can have profound effects on growth and metabolism. For example, Kim et al. (2006) showed that addition of potato apyrase to M. truncatula root hairs inhibited root hair growth, which is essential for rhizobial infection. It may be that fine control of ATP and ADP at the root hair surface is essential to allow for the proper balance between root hair growth and rhizobial infection.
In summary, this study highlights a critical role of the GS52 ecto-apyrase catalytic activity during the early events in soybean nodulation leading to normal nodule development. The data argue that the key step is the hydrolysis of extracellular ATP, leading to the release of ADP, which appears beneficial, at the appropriate concentration, for efficient nodulation. These results provide strong evidence that soybean GS52 ecto-apyrase is essential for the initiation of nodule primordia and nodule development.

RNAi Plasmid Construction
The RNAi gene constructs were made according to Collier et al. (2005). In short, a 343-bp gene fragment for the RNAi construct was amplified by PCR from soybean genomic DNA using apyrase GS52 specific primers as given in Supplemental Table S1. A control RNAi gene fragment in plants was created using a fragment of GUS (uidA; Jefferson et al., 1987). The gene fragments for the RNAi GUS construct was amplified via PCR from the GUS expression plasmid pCGT 1427 using GUS-specific primers (Supplemental Table S1) and blunt-end cloned into pCR-BLUNT (Invitrogen). The amplified GS52 and cloned GUS fragments were cloned into the FMV-driven RNAi shuttle vector, pCGT 2255 (XhoI-KpnI sites for the sense orientation and XbaI-HindIII sites for the antisense orientation; Collier et al., 2005). The entire RNAi shuttle cassettes (promoter/double-stranded RNA/terminator) was excised using flanking Sse8387I restriction endonuclease sites and cloned into the T-DNA of the binary vector, pAKK 1467B (the T-DNA of pAKK 1467B contains GFP and BAR expression modules along with a unique Sse8387I restriction site; Collier et al., 2005). RNAi binary vectors created included the RNAi apyrase GS52 plasmid (pCGT 6288) and the RNAi GUS plasmid (pCGT 5200). An empty vector control (pCGT 6419A) was also generated by cloning the FMV RNAi shuttle (without gene fragments) using the flanking Sse8387I sites into pAKK 1467B. The fidelity of the clones was verified by sequencing and electroporated into Agrobacterium rhizogenes strain K599 (McCormac et al., 1998).

Production of Hairy Roots Using the Composite Plant System
Soybean composite plants were generated according to Govindarajulu et al. (2008). In short, soybean seeds (cv Williams 82) were surface sterilized using chlorine gas. Sterilized seeds were sown in germinating mix (4 M; Hummert International) and grown in the greenhouse for 2 weeks. Hairy root transformation was carried out with A. rhizogenes strain K599 transformed with the RNAi apyrase GS52 and RNAi GUS constructs along with the empty vector control, pCGT 6419A. A. rhizogenes cultures were grown in Luria-Bertani broth with kanamycin (50 mg mL 21 ) in a flask at 225 rpm at 28°C overnight. Bacterial cells were spun down at 3,000g for 10 min at 23°C, and resuspended in sterile water to an OD 600 nm of 0.3. Sterilized FibrGro cubes (1 cm 3 ; Hummert International) were inoculated with A. rhizogenes bacterial suspension. Apical stem sections were excised from greenhouse-grown soybean plants and inserted into the inoculated cubes. These plants were incubated in the growth chamber (temperature 22°C; light 15 mmol m 22 s 21 ; humidity 30%, 16-h day/8-h night photoperiod) for a week. FibrGro cubes were removed and each composite soybean plant was placed in individual 4-inch pot containing sterile vermiculite: perlite (3:1) wetted with nitrogenfree plant nutrient solution (Lullien et al., 1987). The pots were incubated in the growth chamber (temperature 26°C; light 200 mmol m 22 s 21 ; humidity 60%; 16-h-day/8-h-night photoperiod) for 2 weeks. Plants were watered every other day alternating water with nitrogen-free plant nutrient solution. Production of transgenic hairy roots on soybean stems was monitored by observing the formation of GFP-expressing roots.

Inoculation of B. japonicum and Phenotypic Nodulation Analysis
Soybean composite plants were inoculated with B. japonicum USDA110 strain for nodulation. Cultures were grown in HEPES-MES liquid medium (Cole and Elkan, 1973) and 20 mg mL 21 chloramphenicol at 30°C for 2 d with agitation to an OD 600 nm of 0.5. The cells were centrifuged at 7,000g for 15 min at 10°C, pellet washed with sterile water, and then resuspended in nitrogenfree plant nutrient solution to a final OD 600 of 0.08 (approximately 10 8 cells mL 21 ). After 2 weeks of plant growth in vermiculite/perlite, 10 mL of the B. japonicum bacterial suspension were inoculated onto each plant. Control plants (noninoculated) were mock inoculated with 10 mL of nitrogen-free plant nutrient solution. Roots used in this analysis were verified as transformed by their GFP epifluorescence using a Zeiss Stemi SV11 microscope outfitted with a 480-nm excitation/515 emission fluorescein isothiocyanate filter. Twenty-four individual plants were scored for GFP roots containing nodules for RNAi construct. Twenty-five root tissues with nodules were frozen in liquid nitrogen after nodule counting and stored at 280°C for RNA extractions. For each RNAi construct, experiments were performed three times and analyzed.

Treatment of Soybean Roots with ATP or ADP
Soybean seeds were surface sterilized and composite plants were generated as described above, the difference being sterilized FibrGro cubes (1 cm 3 ; Hummert International) were inoculated with autoclaved water to produce wild-type roots. Two days after inoculation with B. japonicum (OD 600 = 0.2), each wild-type root produced from the composite plant was treated with 10 mL of different concentrations of ADP (5, 10, 25, 50, 75, 100, 125, 150, 175, 200, and 250 mM, respectively) or ATP (10, 50, 100, 150, and 200 mM respectively). Control plants were mock inoculated with 10 mL of water. After 4 weeks, 24 individual plants were scored and roots were washed (n . 25 per plant) with water and mature root nodules counted. For complementation assays, individual transgenic plants containing either empty vector, RNAi GUS control or RNAi apyrase GS52 constructs were treated with 10 mL of 100 mM ATP or ADP, 48 h after inoculating plants with B. japonicum (OD 600 = 0.08). Control plants were mock inoculated with 10 mL of autoclaved water. Four weeks after B. japonicum inoculation, GFP roots were carefully excised, rinsed with water, and the number of nodules counted manually. Twenty-four individual plants were scored for GFP roots containing nodules for each RNAi construct. Twenty-five root tissues with nodules were frozen in liquid nitrogen after nodule counting and stored at 280°C for RNA extractions. For each RNAi construct, experiments were performed three times and analyzed.

Isolation of RNA and Real-Time PCR
To quantify GS50, GS52, and ENOD40 gene expression, total RNA was extracted from tissues (as described above) using TRIzol reagent (Invitrogen) and purified using chloroform extraction. cDNA synthesis was performed as described by Libault et al. (2008). Gene-specific primers used to quantify and normalize gene expression are described in Supplemental Table S1. qRT-PCR using SYBR Green PCR master mix (Applied Biosystems) was performed with a 7900 HT sequence detection system and a 7500 real-time PCR system (Applied Biosystems). The following PCR parameters were used: 50°C for 2 min, 95°C for 10 min, 40 cycles at 95°C for 15 s, and 60°C for 1 min. The data were analyzed using SDS 2.2.1 software and the 7500 System, version 1.3.0 (Applied Biosystems), respectively, when the 384-well and the 96-well plate qRT-PCR machines were used. PCR efficiencies (E) were calculated according to a linear regression analysis using LinRegPCR software (R 2 value . 0.995; Ramakers et al., 2003). Absolute gene expression levels relative to the housekeeping gene Cons6 were calculated for each cDNA sample using the following equation: relative ratio gene/Cons6 = (E gene 2(Ctgene) )/(E Cons6

2(CtCons6)
). The values of three replicates were used in a Student's t test to calculate probabilities of distinct induction or repression, and the average ratio of these values was used to determine the fold change in transcript level in treatment samples compared with control.

Preparation of Nodule Samples for Light Microscopy and TEM
Freshly harvested mature nodules from empty vector control transgenic roots, RNAi GUS control transgenic roots, and small empty nodules from RNAi apyrase GS52 transgenic roots were dissected to fit the specimen planchettes for high-pressure freezing, packed in 0.15 M Suc in 50 mM PIPES buffer (pH 6.8) and frozen in a Bal-Tec high-pressure freezer (Danforth Plant Science Center). Frozen nodule samples were freeze substituted in acetone containing 2% osmium tetroxide and 0.1% uranyl acetate for 5 d at 285°C, 24 h at 220°C, and 1 h at 0°C (on ice). Then, the samples were warmed gradually to room temperature for 1 h, rinsed in acetone, and infiltrated with Epon/ Araldite resin slowly for 7 d (Hess, 2007). Thin sections were stained in uranyl and lead salts and observed using a Leo 912 energy filter TEM; images were digitally captured. For light micrographs, 0.5-mm sections from resin-embedded material was stained with toluidine blue and observed using a phase contrast light microscope.
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers AF207687 (GS50) and AF207688 (GS52).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Effect of nodulation using different concentrations of nucleotides applied to wild-type legume roots.
Supplemental Table S1. Nucleotide sequences of primer sets used in this study.