Endogenous isoflavones are essential for the establishment of symbiosis between soybean and Bradyrhizobium japonicum


*(fax +1 314 587 1541; e-mail oyu@danforthcenter.org).


Legume iso/flavonoids have been implicated in the nodulation process, but questions remain as to their specific role(s), and no unequivocal evidence exists showing that these compounds are essential for nodulation. Two hypotheses suggest that the primary role of iso/flavonoids is their ability to induce rhizobial nod gene expression and/or their ability to modulate internal root auxin concentrations. The present work provides direct, genetic evidence that isoflavones are essential for nodulation of soybean roots because of their ability to induce the nodulation genes of Bradyrhizobium japonicum. Expression of isoflavone synthase (IFS), a key enzyme in the biosynthesis of isoflavones, is specifically induced by B. japonicum. When IFS was silenced using RNA interference in soybean hairy root composite plants, these plants had severely reduced nodulation. Surprisingly, pre-treatment of B. japonicum or exogenous application to the root system of either of the major soybean isoflavones, daidzein or genistein, failed to restore normal nodulation. Silencing of chalcone reductase led to very low levels of daidzein and increased levels of genistein, but did not affect nodulation, suggesting that the endogenous production of genistein was sufficient to support nodulation. Consistent with a role for isoflavones as endogenous regulators of auxin transport in soybean roots, silencing of IFS resulted in altered auxin-inducible gene expression and auxin transport. However, use of a genistein-hypersensitive B. japonicum strain or purified B. japonicum Nod signals rescued normal nodulation in IFS-silenced roots, indicating that the ability of isoflavones to modulate auxin transport is not essential to nodulation.


Legume nodulation is a highly host-specific interaction mediated by the chemical structure of signals exchanged between the two symbiotic partners. Plant roots secrete particular flavonoid compounds that are recognized by the compatible rhizobial symbiont, leading to the induction of the nodulation (nod) genes. The products of the nod genes are enzymes that synthesize a unique lipo-chitin signal molecule (Nod signal) that is recognized by the plant host, activating a cascade of events culminating in the formation of a functional nodule organ (Geurts and Bisseling, 2002; van de Sande and Bisseling, 1997).

Extensive evidence has accumulated over the years stressing the role of iso/flavonoids in nodulation. Indeed, research in a number of laboratories (Banfalvi et al., 1988; Peters et al., 1986; Recourt et al., 1992) identified the nod gene inducing iso/flavonoids from a variety of legume species using root exudates, extracts or seed-coat preparations (Begum et al., 2001; Hungria et al., 1992; Smit et al., 1992). In soybean, isoflavones (e.g. genistein and daidzein), a group of flavonoid compounds that are found almost exclusively in legumes, were shown to be the primary inducers of Bradyrhizobium japonicum nod gene expression (Banfalvi et al., 1988; Kosslak et al., 1990; Loh et al., 1994; Smit et al., 1992). In addition to isoflavones, a few other flavonoid compounds, including isoliquiritigenin, the chalcone precursor of daidzein, were also shown to induce nod gene expression in B. japonicum (Kape et al., 1992). In some rhizobial species, non-flavonoid compounds such as betaines (Phillips et al., 1994) or xanthones (Yuen et al., 1995) were also able to induce nod gene expression. Although iso/flavonoid and non-flavonoid nod gene inducers are clearly made by legumes, no unequivocal evidence exists to indicate that any of these specific compounds are essential for nodule formation. Isoflavones, in particular, have a variety of effects, including inhibition of nod gene expression in certain rhizobial species (Pankhurst and Biggs, 1980). They also act as precursors to defense compounds, termed phytoalexins, to inhibit the growth of various microbes (Graham and Graham, 1996). In soybean, isoflavones were also shown to enhance the resistance of B. japonicum to phytoalexin defense compounds (Parniske et al., 1991).

Biochemical analysis of soybean roots showed that isoflavone levels increased in response to B. japonicum treatment (Cho and Harper, 1991a). This increase was suggested to be regulated by the Nod signaling pathway, because higher isoflavone content was detected in hypernodulating soybean mutants (Cho and Harper, 1991b). We earlier showed that expression of soybean isoflavone synthase (IFS), a key enzyme in the biosynthesis of isoflavones (Figure 1), increased in response to B. japonicum treatment in a tissue-specific manner (Subramanian et al., 2004). When an IFS1 gene promoter: β-glucuronidase (IFS1:GUS) fusion was transformed into soybean, only root hairs and xylem showed significant, rhizobium-induced IFS1 promoter activity. The IFS expression in root hairs may contribute to the isoflavones in root exudates. However, expression in deeper cell layers raised the possibility that, in addition to serving as external Nod signal inducers, isoflavones may have other functions in the roots. There is evidence supporting a role for isoflavones in modulation of auxin transport. The isoflavone genistein was reported to inhibit binding of the auxin transport inhibitor naphthyl pthalamic acid (NPA) in zucchini hypocotyl segments (Jacobs and Rubery, 1988). In Arabidopsis, there is strong evidence to suggest that flavonoids are endogenous auxin efflux inhibitors (Brown et al., 2001). Flavonoid-deficient mutants exhibited elevated auxin transport in seedlings, and this elevation could be reduced by adding the flavanone naringenin externally (Murphy et al., 2000). In a transparent testa mutant (tt4), the expression and sub-cellular localization of auxin efflux carrier PIN proteins are related to local flavonoid concentrations (Peer et al., 2004). In legumes, flavonoid compounds were found to accumulate in cells near the site of nodule initiation. It was suggested that this leads to localized auxin accumulation, resulting in nodule organogenesis (Mathesius et al., 1998, 2000b). Therefore it is possible that iso/flavonoid accumulation preceding nodule initiation might lead to localized increase in auxin levels. Another possible role for isoflavones is the induction of bacterial nod genes inside the roots. It was reported recently that plant Nod signal receptors may be expressed in deeper layers of the root, suggesting that sustained expression of the rhizobial nod genes (encoding Nod signal synthesis) may be required inside the root (Madsen et al., 2003; Parniske and Downie, 2003). This would necessitate the presence of nod gene inducers (isoflavones in this case) in deeper cell layers of the root.

Figure 1.

 Partial diagram of the phenylpropanoid pathway, showing intermediates and enzymes involved in isoflavone biosynthesis, as well as some branch pathways.
Legume-specific enzymes are underlined; legume-specific compounds are shown in bold type. *Glyceollins are the major phytoalexin compounds in soybean; other legumes produce different downstream metabolites from daidzein and/or genistein.

The current work focused on providing evidence that isoflavones are essential for soybean nodulation, and seeks to identify the role of isoflavones in the root. In soybean the major isoflavonoids, daidzein and genistein, are produced via two branches of the isoflavone biosynthetic pathway (Figure 1). The chalcone precursor of daidzein, isoliquiritigenin, is produced by the enzyme chalcone reductase (CHR) and subsequently converted to daidzein by chalcone isomerase (CHI) and IFS. Genistein is produced from naringenin chalcone by the action of CHI and IFS. So CHR is essential for the production of daidzein; and IFS is essential for the production of both daidzein and genistein (Figure 1). RNA interference (RNAi)-induced silencing of IFS greatly reduced isoflavone levels and led to severe reduction of nodulation in transgenic hairy roots of composite soybean plants. Surprisingly, exogenous addition of isoflavones did not significantly increase nodulation in IFS-silenced roots, suggesting a role for isoflavones beyond initial signaling between the plant host and symbiont. Indeed, reduced isoflavone levels increased polar auxin transport in the transgenic roots, suggesting that isoflavones may play a role as auxin transport regulators inside the root. To differentiate between the roles of isoflavones as either nod gene inducers or auxin-transport regulators in soybean, we used a genistein-hypersensitive B. japonicum strain (Bj30053) that produces the lipo-chitin Nod signal at very low levels of inducers (Ip et al., 2001). This strain was able to rescue nodulation to normal levels, even in IFS-silenced roots, and purified Nod signals were able to induce pseudo-nodules on the isoflavone-null lines. These data indicate that the role of isoflavones as nod gene inducers inside the root is essential for nodulation, while the effects of isoflavones on auxin transport are not.


Tissue-specific localization of IFS in response to B. japonicum treatment

In a previous report using an IFS1:GUS transcriptional fusion (Subramanian et al., 2004), a unique xylem-specific induction of the IFS1 promoter by B. japonicum was detected in stably transformed soybean lines. To validate these earlier observations, in situ hybridization was performed to localize IFS transcripts in cross-sections of control and B. japonicum-treated soybean roots. The probe used was designed to detect the highly identical IFS1 and IFS2 transcripts (Jung et al., 2000). In control root sections (approximately 1 cm from the tip) we observed low levels of IFS transcripts primarily in the epidermal tissue, and to some extent in the vasculature (Figure 2a). In B. japonicum-treated roots, abundant expression was seen both in the epidermis and in the vascular tissue (Figure 2b), confirming our earlier observations. Hybridizations with the sense IFS transcript as probe did not show a significant signal (Figure 2c). This observation confirms the xylem-specific induction of isoflavone biosynthesis by B. japonicum inoculation, and suggests a possible vascular role for isoflavones.

Figure 2.

IFS expression was induced in the vasculature in response to Bradyrhizobium japonicum treatment (8 h) in soybean roots.
(a–c) In situ hybridization of soybean (cv. Bragg) root cross-sections (approximately 1 cm from the tip) probed with digoxigenin (DIG)-labelled antisense (a,b) or sense (c) IFS1 probe that can detect both IFS1 and IFS2 transcripts. Cross-sections of untreated control roots (a) and B. japonicum-treated roots (b) showed IFS expression. The sense probe did not show any significant hybridization (c). All three panels are of the same magnification (10×).

RNAi-mediated suppression of IFS and CHR in soybean hairy roots

As discussed above, although isoflavones can induce B. japonicum nod gene expression, an essential role for isoflavones in soybean nodulation has not been documented unequivocally. To examine this issue directly, an RNAi-based approach was used to silence IFS and CHR genes in hairy roots of soybean cv. Bragg. We previously reported that there are probably only two functional IFS genes in the soybean genome, and they are highly similar to each other (Jung et al., 2000). An IFS RNAi binary vector was constructed to silence both IFS genes by targeting the region of most identity in the coding sequence (Subramanian et al., 2005). We used a similar approach to construct a CHR RNAi binary vector. A search of the TIGR soybean gene index (http://www.tigr.org/tdb/gmgi) revealed four putative CHR ESTs homologous to the previously published CHR sequence (Welle et al., 1991). We named them CHR1 (TC219205, identical to previously published sequence); CHR2 (TC203399); CHR3 (TC227002); and CHR4 (TC229251). An RNAi vector that targeted a region of most identity among these four CHR genes was constructed to inactivate all CHR transcripts simultaneously. The IFS-RNAi (IFSi) and CHR-RNAi (CHRi) binary vector constructs contained a detectable GFP marker and were used to generate hairy root composite soybean plants by transformation with Agrobacterium rhizogenes K599 (Collier et al., 2005). These plants have transgenic as well as non-transgenic roots, and each root is an independent transformation event. On average, each composite plant had approximately 25% GFP transgenic roots. All results presented below compare data obtained from transgenic roots only, comparing vector control transgenic roots with RNAi transgenic roots except where indicated.

The resulting IFSi and CHRi transgenic roots were analysed for the levels of IFS1, IFS2, CHR1, CHR2, CHR3 and CHR4 transcripts by quantitative RT-PCR. As expected, the IFSi-transformed roots showed a significant reduction in IFS1 and IFS2 expression (Figure 3a). While we observed more than 40-fold reduction of IFS1 and IFS2 transcripts in IFSi roots, we did not observe any reduction of these transcripts in CHRi roots. Quantitative RT-PCR assays for the four putative CHR transcripts in CHRi roots indicated varying levels of gene silencing for each gene. While CHR1 and CHR4 were significantly silenced, CHR2 and CHR3 were not (Figure 3a). In IFSi roots, there was no reduction of CHR1, CHR2 and CHR3 transcript levels. However, we did observe an apparent decrease in transcript levels of CHR4 in IFSi roots (Figure 3a). It is possible that this is due to a feedback repression of CHR4 expression by the accumulation of liquiritigenin in these roots (see below).

Figure 3.

 RNAi-mediated silencing of IFS and CHR gene expression in soybean hairy roots.
(a) Transcript levels of IFS1, IFS2, CHR1, CHR2, CHR3 and CHR4 in hairy roots transformed with control vector (white bars); IFSi construct (black bars); or CHRi construct (hatched bars). Transcript levels were estimated using the relative Ct-value method and normalized to that of soybean Ubi3. Error bars indicate the range of possible values based on SD of replicate Ct values. Data presented are representative of at least two independent experiments.
(b) Isoflavone contents of soybean hairy roots transformed with control vector (white bars); IFSi construct (black bars); or CHRi construct (hatched bars). Compounds were extracted and quantified by HPLC analysis as described in Experimental procedures. Each data point is the average of at least 12 roots from five different composite plants. Error bars, SEM. Dz, daidzein; Gn, genistein.

We further assayed the levels of isoflavonoid compounds in these roots. The majority of the soybean root isoflavonoids are present in their glycosylated form (Figure S1). We present total isoflavonoid levels below. Approximately 98% of the IFSi transgenic roots (data not shown) had <2 and <8% of the wild-type levels of the major isoflavones daidzein and genistein, respectively (Figure 3b). As expected, a slight but significant increase in liquiritigenin, the precursor of daidzein, was observed (Figure 3b). Naringenin, the precursor of genistein and several other flavonoid compounds, did not accumulate in either transgenic or non-transgenic roots (data not shown). In over 96% of CHRi transgenic hairy roots tested, little or no daidzein was detectable by HPLC, suggesting successful silencing of all CHR activity in the roots (Figure 3b). However, suppression of CHR activity led to an eightfold increase in genistein levels (Figure 3b), thereby keeping the total isoflavone levels unaffected between control roots (99.7 ± 21.8 nmol per 100 mg root) and CHRi roots (129.0 ± 16.6 nmol per 100 mg root). There was no difference in the levels of isoflavones between transgenic or non-transgenic control roots in composite plants generated by transformation with A. rhizogenes containing the empty vector (S. Subramanian, M. Matsuno and O. Yu, unpubl. results). Although CHR2 and CHR3 were not effectively silenced, reduced daidzein levels in CHRi roots indicated that all functional CHR activity was silenced in the roots. Therefore CHR2 and CHR3 may not encode proteins with CHR activity, or they may not have significant CHR activity in the roots. These results clearly indicate that we were successfully able to silence the functional IFS and CHR genes in soybean roots using RNAi constructs.

Suppression of IFS activity, but not CHR activity, leads to severely reduced nodulation

We performed control experiments to test the suitability of soybean hairy root composite plants to study nodulation by B. japonicum. When inoculated with B. japonicum, the transgenic hairy roots nodulated as efficiently as the non-transgenic adventitious roots in these plants (Figure S2). In addition, we observed fourfold higher nodule numbers on the roots of a hypernodulating soybean mutant (NTS382; Carroll et al., 1985), indicating that hairy root composite plants retain the hypernodulating phenotype (Figure S2). This suggested that the hairy root composite plants are suitable for studying nodulation. We tested the effect of IFS silencing on nodulation of soybean roots by B. japonicum. In Bragg IFSi roots, the average number of nodules per root was severely reduced compared with empty vector control roots (Figure 4a). While vector control transgenic roots had 14 ± 4 nodules, IFSi roots had only 2 ± 1 nodules per root. This reduction was statistically significant at P < 0.005 based on Poisson's distribution analysis (see Experimental procedures). On the other hand, the CHRi-silenced roots that had altered composition of isoflavones, but unaltered total isoflavone levels, nodulated normally when inoculated with B. japonicum. There was no significant difference in nodule numbers between CHRi roots and vector control roots (Figure 4a). In summary, a reduction in isoflavone levels led to significantly reduced nodulation (as in IFSi roots), whereas unaltered isoflavone levels (as in CHRi roots) did not affect the extent of nodulation. These results indicate that isoflavones are indeed essential for efficient nodulation in soybean.

Figure 4.

 RNAi-mediated suppression of isoflavone levels in soybean hairy roots leads to severe reduction in nodulation by Bradyrhizobium japonicum.
(a) Average nodule numbers per root in vector control (white bars); IFSi (black bars); and CHRi (hatched bars) transgenic hairy roots. Each data point is the average of at least 52 roots from 24 composite plants generated from three independent experiments. Error bars, SEM. aSignificantly different from the respective vector control based on Poisson's distribution analysis (P < 0.005).
(b) Relationship between the amounts of daidzein (squares) or genistein (circles), and the number of nodules in IFSi roots. Roots with very low total isoflavone levels nodulated very poorly; those with near wild-type isoflavone levels nodulated normally.

Given the partial penetration of the hairy root RNAi phenotype, it is likely that the small number of nodules observed in IFSi roots is caused by incomplete silencing of IFS in a few roots. In fact, the levels of isoflavones assayed in individual transgenic hairy roots showed varying degrees of reduction, suggesting that each individual transgenic root probably had a varying degree of gene silencing. We attempted to associate the extent of nodulation in selected transgenic roots with the amounts of remaining endogenous isoflavones (Figure 4b). Roots with very low isoflavone levels nodulated very poorly, and those with near-wild-type isoflavone levels nodulated normally. These data suggest that levels of isoflavones in the root are an important factor that determines the extent of nodulation. Thus the few remaining nodules formed on the IFSi roots are probably the result of incomplete silencing. The above results clearly suggest that isoflavones are essential for proper nodulation of soybean by B. japonicum.

Exogenous isoflavones do not restore nodule numbers

The isoflavones daidzein and genistein were previously shown to be the primary inducers of nod gene expression in B. japonicum (Banfalvi et al., 1988). Therefore exogenous addition of different concentrations of daidzein or genistein was used in an attempt to rescue the loss of nodulation in the IFSi roots. Bradyrhizobium japonicum cells pre-treated with 5 μm genistein were used to inoculate soybean cv. Bragg vector control, IFSi- or CHRi-transformed roots (see Experimental procedures). When observed 3 days post-inoculation, we observed root hair-curling responses in all three genotypes (Table 1), suggesting that the plants were recognizing the Nod signal induced in these B. japonicum cells by the added genistein. However, 1 week after inoculation, the number of nodule primordia in the IFSi roots were reduced significantly compared with control roots. CHRi roots were not affected in the number of nodule primordia compared with control roots (Table 1). Similar results were obtained when B. japonicum cells pre-treated with daidzein were used (data not shown). This suggested that the IFSi roots were deficient in inducing nodule primordia, despite initiating the root hair-deformation response to B. japonicum. As Nod signals were present (as indicated by the root hair-curling response), these data clearly suggest a role for isoflavones in nodule initiation beyond that of simply inducing nod gene expression in the rhizosphere.

Table 1.   Different root hair deformation response and nodule primordia initiation in IFSi and CHRi roots
GenotypeRoot hair deformationNumber of nodule primordia per root
  1. Root hair deformation was scored in a 5-cm region of the root closest to the tip 3 days after inoculation with 5 μm genistein-pretreated Bradyrhizobium japonicum cells in 10 roots for each genotype. Observation of a curling response or corkscrew-like response was considered positive. Nodule primordia were counted in cleared roots (see Experimental procedures) 1 week after inoculation with B. japonicum. Each data point represents the average ± SEM.

Vector+12.17 ± 1.98
IFSi+4.42 ± 1.52
CHRi+10.91 ± 1.84

We also attempted pre-treatment of B. japonicum cells with different concentrations of genistein or the exogenous addition of genistein to the root system before B. japonicum inoculation. However, there was no rescue of nodule numbers to wild-type levels by any of these genistein treatments (Figure S3). Similar results were obtained when daidzein was used (data not shown).

Thus an increase in endogenous isoflavone levels was able to rescue nodulation (as in CHRi roots), while pre-treatment of B. japonicum cells with, or the addition to the root system of, exogenous isoflavones failed to rescue the nodulation phenotype in IFSi roots (Figure S3). A chief difference between exogenously applied isoflavones and endogenous isoflavones could be their tissue-specific distribution. Indeed, we observed tissue-specific induction of IFS expression in response to B. japonicum (Figure 2). Therefore endogenous isoflavones, perhaps because of their tissue-specific distribution, are critical to nodulation. As discussed above, the essential role of endogenous isoflavones in nodulation could be modulation of auxin transport and/or the induction of nod genes in the root.

Auxin-inducible gene expression and auxin transport are altered in isoflavone-silenced roots

Given the known role of auxin in initiation of nodule primordia and the postulated role of isoflavones in controlling auxin movement in roots, we examined IFSi roots for altered auxin homeostasis. Soybean composite plants expressing the auxin marker gene DR5:GUS (Ulmasov et al., 1997), alone or in combination with the IFSi construct, were generated. Soybean hairy roots transformed with DR5:GUS alone showed GUS expression primarily in the lateral root initials (Figure 5a). Mock-inoculation (arrowheads in Figure 5a) or spot-inoculation of daidzein-pretreated B. japonicum cells (arrowheads in Figure 5b) did not result in any obvious alteration in DR5:GUS expression (Figure 5b). Surprisingly, hairy roots transformed with the construct containing both IFSi and DR5:GUS had abundant DR5:GUS expression, noticable all over the root (Figure 5c). The expression pattern was not altered by mock-inoculation (arrowheads in Figure 5c) or spot inoculation with daidzein pre-treated B. japonicum cells (arrowheads in Figure 5d).

Figure 5.

 IFSi roots have increased auxin-inducible gene expression and auxin transport.
(a–d) Histochemical staining for DR5:GUS expression in roots transformed with DR5:GUS alone (a,b) or DR5:GUS in combination with IFSi (c,d). Arrowheads indicate the sites of mock-inoculation (a,c) or spot-inoculation with 5 μm daidzein-pretreated Bradyrhizobium japonicum cells (b,d). All four panels are of the same magnification (60×).
(e) Transcript levels of an endogenous auxin-inducible gene (GH3) measured by quantitative RT-PCR in vector control (white bars); IFSi (black bars); and CHRi (hatched bars) transgenic hairy roots. Transcript levels were estimated using the relative Ct-value method normalized to that of soybean Ubi3 (see Experimental procedures). Error bars indicate the range of possible values based on SD of replicate Ct values. Data presented are representative of two independent experiments.
(f) Acropetal auxin transport assayed using 3H-IAA as tracer in vector control (white bars); IFSi (black bars); and CHRi (hatched bars) transgenic hairy roots. Data are the average of 12 roots per sample; error bars, SEM. aSignificantly different from the respective control based on Student's t-test (P < 0.05).

We also tested the expression of an endogenous auxin-inducible marker gene, GH3, in these roots by RT-PCR assays. In IFSi roots there was a 10-fold increase in GH3 levels compared with vector control roots (Figure 5e). Interestingly, CHRi roots (which had altered isoflavone composition, but unaltered isoflavone levels) did not show any significant difference in GH3 expression levels compared with vector control. These two observations suggest that silencing of IFS in the roots caused a dramatic increase in auxin-inducible gene expression.

This increased auxin-inducible expression in IFSi roots was probably caused by increased auxin transport, given the role of iso/flavonoid compounds as auxin transport regulators. To test this directly, acropetal auxin transport was measured in vector and IFSi root segments using 3H-IAA as tracer. The results shown in Figure 5(f) document that the IFSi roots had significantly elevated auxin transport compared with the empty vector-transformed roots. However, CHRi roots did not show a significant increase in auxin transport, but rather a small decrease in auxin transport. As an additional test, we measured auxin transport in intact hairy root composite plants and wild-type soybean seedlings. These results (Figure S4) were consistent with those obtained using root segments. These observations, together with the increased auxin-regulated gene expression, suggested that the IFSi-silenced roots accumulated auxin because of elevated acropetal transport. Therefore, during nodulation, induction of isoflavone synthesis by B. japonicum inoculation may cause localized inhibition of auxin transport in soybean roots, leading to accumulation of auxin. However, no increase in DR5:GUS expression was observed at the site of rhizobia inoculation. This is probably due to the narrow range of auxin sensitivity of the DR5:GUS marker. We also measured isoflavone levels in roots and mature nodules for comparison. Total isoflavone levels in mature nodules (502 ± 56 μg g−1 FW) were only slightly higher than in the roots (422 ± 33 μg g−1 FW).

Isoflavone-mediated inhibition of auxin transport may not be critical to soybean nodulation

The roles of isoflavones as auxin transport inhibitor and nod gene inducer can be differentiated by the use of genistein-hypersensitive B. japonicum mutant strains. These mutant cells produce three- to fivefold higher Nod signals, even in the presence of suboptimal concentrations of genistein (Ip et al., 2001). If the essential role of isoflavones is to modulate auxin transport and cause initiation of nodule primordia, use of these mutant strains would not restore nodulation in IFSi roots. On the other hand, if isoflavones are necessary for nod gene induction in the roots, these strains should be able to nodulate IFSi roots equally well as vector control roots. Vector control roots inoculated with the wild-type strain of B. japonicum (USDA110) or the genistein-hypersensitive B. japonicum mutant (Bj30053) nodulated normally (Figure 6a). IFSi roots inoculated with the wild-type B. japonicum strain had significantly reduced nodulation, as observed before. In contrast, there was no significant difference in the number of nodules between vector transgenic and IFSi roots when inoculated with Bj30053 (Figure 6a). The other genistein-hypersensitive strain tested, Bj30055, nodulated vector transgenic hairy roots in a less efficient manner. Nevertheless, there was no significant reduction in the number of nodules between vector transgenic and IFSi roots inoculated with Bj30055 (Figure 6a). The restoration of nodulation in IFSi roots by genistein-hypersensitive B. japonicum strains suggests that the essential role of isoflavones during nodulation is nod gene induction, and perhaps not modulation of auxin transport.

Figure 6.

 Genistein-hypersensitive Bradyrhizobium japonicum strains are not affected in their ability to nodulate IFSi roots.
(a) Average nodule numbers per root in vector control (white bars) and IFSi (black bars) transgenic hairy roots inoculated with the wild-type strain Bj110, or the genistein-hypersensitive mutant strains Bj30053, or Bj30055. aSignificantly different from the respective vector control based on Poisson's distribution analysis (P < 0.005).
(b) Poor association between the number of nodules produced by Bj30053 and the levels of daidzein (squares) or genistein (circles). Even roots with undetectable levels of isoflavones nodulated efficiently.

We also measured isoflavone levels in these roots and attempted to associate the number of nodules to the isoflavone levels (or the level of silencing), both normalized to unit weight of the root. As before, we observed that roots with lower isoflavone levels nodulated poorly, whereas roots with near wild-type isoflavone levels nodulated normally (data not shown). However, we did not see the same association (Figure 6b) between the number of nodules produced by Bj30053 and the levels of isoflavones. Note that several roots with undetectable levels of isoflavones produced a very large number of nodules (clustering of data points on the y-axis in Figure 6b). This is in contrast to the very few such roots seen when IFSi roots were inoculated with Bj110 (Figure 4b). This is consistent with the notion that the expression of bacterial nod genes in the absence of detectable levels of isoflavones (as in the case of strain Bj30053) is sufficient to allow efficient nodulation.

We further tested the ability of IFS-silenced roots to produce pseudo-nodules when spot-treated with purified B. japonicum Nod signals. We spot inoculated approximately 200–300 nl 10−8 m Nod signal preparation on to emerging root hairs of vector, IFSi and CHRi transgenic hairy roots of Glycine soja. This has been the species of choice for root-hair deformation or nodulation assays using purified Nod signals, because of its small size that facilitates microscope examination (Stokkermans et al., 1995). We confirmed by HPLC analyses that the expression of our silencing constructs in this species led to inhibition of isoflavone biosynthesis as expected (not shown). All three genotypes (vector control, IFSi and CHRi) produced comparable numbers of pseudo-nodules (Table 2), suggesting that the presence of the Nod signals is sufficient to initiate nodule primordia, even in the absence of isoflavones. This is in contrast to our results obtained using genistein pre-treated B. japonicum cells (Table 1), where there was perhaps no sustained nod gene induction in IFSi roots resulting in reduced nodule primordial initiation. These results confirm that the role of endogenous isoflavones as nod gene inducers inside the root is more critical to nodulation than their role as auxin transport regulators.

Table 2.   Isoflavone synthase-silenced roots are able to respond to purified Nod signals as efficiently as control roots
GenotypeExperiment I (number of pseudo-nodules from 25 spot treatments)Experiment II (number of pseudo-nodules from 50 spot treatments)
  1. Emerging root hairs were spot-treated with approximately 200–300 nl 10−8 m purified Nod signals from Bradyrhizobium japonicum (dissolved in 18% acetonitrile) 1 week after planting in sand (see Experimental procedures). Three weeks after inoculation, the sites of treatment were examined for the presence of pseudo-nodules. None of the 20 control spot treatments with 18% acetonitrile on vector control roots resulted in any nodules.



Recent advances in identification and characterization of key isoflavone biosynthetic enzymes (Jung et al., 2000; Steele et al., 1999) allowed us to genetically engineer the level and composition of isoflavones in soybean roots. Silencing of IFS expression in soybean hairy roots resulted in a significant reduction in nodule formation on inoculation with B. japonicum. The IFS-silenced roots had increased levels of liquiritigenin (Figure 1), one of the substrates of IFS, compared with control roots (Figure 3b). The other IFS substrate, naringenin (Figure 1), was not detected, probably because it is shared by a few other enzymes in the flavonoid pathway. Both naringenin and liquiritigenin were previously shown to induce B. japonicum nod gene expression (Kape et al., 1992). These results suggest that, under the conditions assayed, these alternative nod gene inducers are not sufficient to restore nodulation in the absence of isoflavones. The partial penetration of the RNAi-silencing phenotype probably accounts for the level of nodulation in the IFS-silenced roots. Indeed, we observed that roots with low isoflavone levels nodulated poorly, whereas roots with near wild-type isoflavone levels nodulated normally. Therefore isoflavones are indeed a necessity for nodulation.

RNAi-mediated suppression of CHR gene expression led to a reduction in daidzein levels, as expected, but also to an increase in genistein levels, thereby keeping the total isoflavone content unaffected. Daidzein is the precursor for several downstream metabolites that are used by plants mostly to protect against pathogenic microbes. However, genistein is not known to be converted to any downstream metabolite, except for glycosylation for storage in the vacuole (Graham and Graham, 1996). The observation that elevated genistein levels were able to restore nodulation even in the absence (or reduced levels) of the major root isoflavone daidzein clearly indicates that isoflavones, and not their downstream compounds, play a crucial role in nodulation.

Given the fact that isoflavones were shown to induce nod gene expression in B. japonicum (Banfalvi et al., 1988), the obvious explanation for the reduced nodulation of IFS-silenced roots is that these plants lack sufficient levels of inducer. However, pre-treatment of B. japonicum cells with isoflavones failed sufficiently to rescue the nodulation phenotype of the IFSi roots. Similar results were found when roots were treated with isoflavones before inoculation, or increasing levels of isoflavones were added externally. Indeed, addition of exogenous isoflavones resulted in higher levels of isoflavones inside the root, suggesting that the level of isoflavone alone is not an indicator of nodulation proficiency. Together, the data obtained using the IFSi roots clearly show that endogenous isoflavones are essential for nodulation. These results are consistent with an important role of isoflavones as inducers of B. japonicum nod gene expression in the rhizosphere, but leave open the possibility that endogenous isoflavones play additional roles in nodulation inside the root.

There are different possible roles for isoflavones inside the root during nodulation. Recent results suggest that bacterial nod gene induction inside the root may be critical. It was reported that putative Nod signal receptors are expressed in deeper layers of the root, providing a second layer of specificity for Nod signal recognition (Madsen et al., 2003; Parniske and Downie, 2003). Therefore the isoflavone nod gene inducers may be required inside the plant to maintain Nod signal synthesis. However, the ability of isoflavones to act as auxin transport inhibitors may also be critical to nodule formation. Data obtained using Medicago truncatula and clover showed that isoflavones and other flavonoid compounds influence auxin transport and/or degradation (Mathesius, 2001; Mathesius et al., 1998, 2000a). Accumulation of flavonoid compounds at the sites of nodule initiation was suggested to modulate auxin transport in order to initiate primordial cell division (Mathesius et al., 1998). It is possible that isoflavones also influence nodule primordial cell division in soybean, by modulating auxin transport. This would necessitate the presence of isoflavones inside the root in a tissue-specific manner. The ability of genistein to act as auxin transport inhibitor was demonstrated several years ago (Jacobs and Rubery, 1988).

Expression of the auxin-inducible reporter gene DR5:GUS (Ulmasov et al., 1997) was significantly increased in IFSi roots, suggesting that auxin levels were increased. This phenotype did not result in any reduction in lateral root number or root mass between vector control and IFS-silenced roots by visual observation. It was found that the IFSi roots had a significantly higher level of auxin transport. Therefore the increased DR5:GUS expression in IFSi roots is probably the result of auxin accumulation because of increased polar transport. It was previously reported that the isoflavone genistein can inhibit NPA binding in zucchini hypocotyls. The inhibition of NPA binding by genistein (approximately 40%) was much lower than that of kaempferol (approximately 60%) or quercetin (approximately 80%) (Jacobs and Rubery, 1988). It is well established in Arabidopsis that the flavonoid compounds kaempferol and quercetin act as auxin-transport regulators. We were not able to detect measurable levels of kaempferol or quercetin in soybean roots (not shown). Therefore the increased auxin transport in IFSi roots is probably due to the reduced levels of isoflavones.

The above-mentioned roles of isoflavones could be differentiated clearly by the use of the genistein-hypersensitive B. japonicum mutant Bj30053, which produces Nod signals in the presence of very low amounts (0.1 μm) of the inducer compounds genistein or daidzein. If isoflavones were critical to nodulation because of their role as auxin transport inhibitors, then one would expect that nodulation would be significantly impaired, even when roots were inoculated with strain Bj30053. However, the results clearly showed that Bj30053 nodulated IFSi roots similarly to wild-type roots. A complementary experiment was performed where the ability of purified B. japonicum Nod signals to produce pseudo-nodules on IFS-silenced roots was tested. If the auxin transport-modulator role of isoflavones is critical to nodulation, purified Nod signals should not be able to produce nodules in IFSi roots. However, purified Nod signals were able to produce pseudo-nodules on IFSi roots with the same efficiency as on vector control roots. We conclude from these results that it is the role of isoflavones as nod gene inducers that is essential for nodulation. If isoflavones are also critical for modulating nodule primordium formation through effects on auxin transport, then (unidentified) redundant plant functions may compensate for the loss of this function in IFSi roots. There appears to be no redundancy for the role of isoflavones as nod gene inducers.

Legumes develop one of two types of nodule, determinate and indeterminate (Kijne, 1992). Determinate nodules (for example, formed on roots of soybean and Lotus japonicus) arise from sub-epidermal cortical cells, have limited cell division, and grow primarily by cell expansion. In contrast, indeterminate nodules (for example, formed on roots of pea and M. truncatula) arise from deeper cortical cell layers and involve continuous cell division at an apical meristem. The ontogeny of both nodule types is known to coincide with localized changes in auxin distribution that precedes cortical cell division. It is possible that the conclusions reached from the present work are specific for soybean or, perhaps, determinate nodulating plants. As mentioned above, previous experiments in M. truncatula and white clover using an auxin-responsive GH3–GUS fusion clearly showed that polar auxin transport was impaired below the site of rhizobial spot inoculation (Mathesius et al., 1998). These authors suggested that cessation of auxin transport resulted in a localized increase in auxin levels triggering nodule primordium formation. In accordance with this, RNAi-mediated suppression of flavonoid biosynthesis in M. truncatula hairy roots led to the abolition of auxin transport inhibition at the site of rhizobial inoculation, leading to reduced nodulation (Wasson et al., 2006). Therefore, during indeterminate nodulation, modulation of local auxin levels by inhibition of polar auxin transport may be a crucial step in nodule formation. However, in the case of both L. japonicus (Pacios-Bras et al., 2003) and soybean (this report), the pattern of reduced auxin-responsive promoter–GUS fusion expression below the site of rhizobial inoculation, as reported for M. truncatula, was not seen. Does the mechanism of nodule primordium formation in determinate nodule-forming plants differ significantly with regard to the role of auxin? Recent evidence suggests so. Grafting experiments by Lohar and VandenBosch (2005) suggest that determinate and indeterminate nodulating legumes may use different signals to initiate nodulation. It was reported recently that in M. truncatula high auxin levels are necessary to sustain nodulation (van Noorden et al., 2006), in contrast to the ‘auxin-burst hypothesis’ proposed for soybean (Gresshoff, 1993). Indeterminate nodules expand primarily by continuous cell division at a terminal meristem that may require highly localized auxin accumulation. In contrast, after initial cell divisions, determinate nodules expand by radial cell expansion that may be less dependent on modulation of auxin levels.

Experimental procedures

Seeds, DNA vectors and plant growth conditions

Seeds of the soybean cultivar Bragg and the corresponding hypernodulating mutant NTS382 were kindly provided by Dr P. Gresshoff (University of Queensland, Brisbane, Australia). Soybean growth conditions and generation of hairy root composite plants were as described previously (Collier et al., 2005), except that the nutrient solution used was free of nitrogen. Briefly, 7-day-old soybean shoot tips were infected with A. rhizogenes K599 in a FibrGro rockwool plug (Hummert International, Earth City, MO, USA). The infected stem segments produced non-transgenic adventitious roots, transgenic hairy roots without the binary vector, and transgenic hairy roots expressing the gene of interest. On average approximately 25–30% of the roots produced were useful for analysis. Each transgenic root results from an independent transgenic event. These ‘composite’ plants were transferred to sand approximately 3 weeks after infection, and allowed to grow for the required period of time. Hairy root composite plants for nodulation assays were watered with N-free nutrient solution on alternate days. See Supplementary material for details of the construction of plant transformation vectors.

In situ hybridization

A 150-bp fragment of the IFS1 transcript, including the entire 3′ UTR, was amplified from cDNA and cloned in to pGEM Teasy vector (Promega, Madison, WI, USA) and the orientation of the insert was verified by PCR assays. DIG-labelled sense or antisense probes were generated using SP6 or T7 RNA polymerases, respectively, using a DIG RNA-labeling kit according to the manufacturer's recommendations (Roche Diagnostics, Indianapolis, IN, USA). Surface-sterilized soybean seedlings were germinated on N-free nutrient agar for 4 days, then treated with a B. japonicum cell suspension (OD600 = 0.08) by flooding the roots. After 8 h, seedlings were rinsed to get rid of excess B. japonicum cells on the root surface, and roots were excised from the plant and fixed in 3.7% formaldehyde, 5% acetic acid, 50% ethanol. Preparation of tissue sections and in situ hybridization and detection were performed based on a protocol modified from Jackson (1991) (kindly provided by Dr Mark Running, Donald Danforth Plant Science Center).

RT-PCR assays and analysis of root isoflavones

RNA isolation, cDNA synthesis and RT-PCR assays were performed as described (Subramanian et al., 2004). Relative transcript levels were determined using the comparative threshold cycle (Ct) value method, as described previously (Bovy et al., 2002) (ABI Prism 7700 sequence detection system user manual, December 1997, Perkin-Elmer Applied Biosystems, Boston, MA, USA) using ubiquitin as control. Primer concentrations of each gene were adjusted to obtain comparable PCR efficiencies with ubiquitin. However, there was slight variations in PCR efficiencies among the different genes assayed. Therefore the expression levels of the same gene under different conditions or genotypes are comparable, whereas the expression levels of two different genes under the same or different conditions are not directly comparable. See Supplementary material for the sequences of primers used in quantitative PCR assays.

Roots for flavonoid assays were immediately frozen in liquid N after microscopic analysis for GFP expression. Flavonoids were then extracted by grinding roots under liquid N2 and extracting with 1 ml 80% methanol per 100 mg tissue. Filtered extracts were directly subjected to HPLC analysis. Identities of compounds were verified by comparison of retention times and UV absorbance spectra to authentic standards. HPLC separation and data analysis were as published (Yu et al., 2003). For the experiment to associate nodule numbers and isoflavone content, roots were frozen in liquid N after counting the nodules.

Bradyrhizobium japonicum treatments and nodulation assays

Growth conditions for B. japonicum strain USDA 110 and inoculation of plants were as described previously (Subramanian et al., 2004). For pre-treatment, B. japonicum cells were diluted to an OD600 of 0.2 and the appropriate amount of isoflavone stock solution was added. The cells were incubated at 28°C for 16 h with shaking. Before inoculation, cells were harvested by centrifugation and resuspended in N-free nutrient solution (OD600 = 0.08). For exogenous application of isoflavones to soil, the appropriate amount of isoflavone stock solution was added to the N-free nutrient solution, which was applied through bottom irrigation.

For nodule counts, hairy root composite plants were allowed to grow on sand for 1 week. A 20-mL suspension of B. japonicum (OD600 = 0.08) was applied to each composite plant. Two weeks after inoculation, roots were harvested, genotyped based on GFP epifluorescence, and the nodules counted. Nodule counts were analysed for randomness in time and space using Poisson distribution and χ2 statistical analyses (http://helios.bto.ed.ac.uk/bto/statistics/tress1.htm).

Genistein-hypersensitive B. japonicum strains Bj30053 and Bj30055 were kindly provided by Dr T. Charles (University of Waterloo, Waterloo, ON, Canada). They were cultured and grown similarly to B. japonicum USDA 110. For nodulation assays involving Bj30053 and Bj 30055, all three strains were allowed to nodulate for 3 weeks after inoculation.

For assaying root hair deformation, 5-cm root sections proximal to the tip were excised, mounted in 30% glycerol and observed under the microscope. For counting nodule primordia, roots were fixed in ethanol: acetic acid (9:1) for 2 h, followed by rinsing with 90% ethanol then briefly with 70% ethanol. The roots were then mounted in chloralhydrate:glycerol:water (9:1:2) and observed under a dissection microscope to count the nodule primordia.

DR5:GUS assays and auxin-transport assays

Transgenic roots in composite plants expressing DR5:GUS or IFSi-DR5:GUS constructs were spot-treated with a suspension of B. japonicum (OD600 = 0.08). Spot treatments were performed using a thin Pasteur pipette. Each spot was approximately 250–500 nl. The site of treatment was marked with a glass bead on the opposite side of the root. Roots were excised after 8 h treatment and stained for histochemical GUS activity (Jefferson et al., 1987).

Acropetal auxin transport in soybean root segments was assayed as described previously (Pacios-Bras et al., 2003). Soybean root-tip segments 4 cm long without any lateral roots were excised from wild-type seedlings or hairy root composite plants, and immediately placed upside down in a culture tube containing 0.5 ml of a mixture of 9.9 μm indole-3-acetic acid (IAA) and 100 nm3H-IAA in 1% agarose. The tubes were sealed with Parafilm and placed in the dark at room temperature. After 24 h, 5 mm of the root end that was directly in contact with the 3H-IAA-agarose was excised, the rest of the root crushed in scintillation liquid, and the amount of 3H-IAA assayed in a liquid scintillation counter. Auxin transport was normalized to the root FW, because of the variation in root diameter.

Nod signal spot treatments

Glycine soja composite plants (seeds kindly provided by Dr Eliot Herman, USDA, Donald Danforth Plant Science Center) were also generated using the same procedure as Glycine max except that seedlings for transformation were 12 days old at the time of infection. Nod signal spot treatments were performed based on the method of Carlson et al. (1993). Composite G. soja plants were allowed to grow in sand for 1 week. Pots were then slit on one side and the roots viewed under a dissection microscope for GFP epifluorescence. These roots were then viewed under normal light at ×40 magnification to view emerging root hairs. This zone was spot-treated with a drop (approximately 200–300 nl) of 10−8 mB. japonicum Nod signal (dissolved in 18% acetonitrile) using a thin Pasteur pipette. Inoculated spots were marked with a small piece of tape, allowed to dry for about 10 min, then sealed again. Three weeks later the pots were opened and examined for pseudo-nodules at the treated sites. Mock inoculations with 18% acetonitrile did not result in pseudo-nodule formation.


We thank Drs S. Clough, R. Nelson and P. Gresshoff for soybean mutant seeds, Dr E. Herman for Glycine soja seeds, Dr T. Charles for genistein-hypersensitive B. japonicum strains, Dr T. Guilfoyle for the DR5:GUS construct, Drs C. Taylor and T. Graham for various cloning vectors, the soybean composite plant protocol and modified hairy root protocol, Dr M. Matsuno for help with HPLC analysis, Dr M. Preuss for critically reading the manuscript and C. Menne, L. Walker and J. Zhong for technical assistance. We also thank Drs Q. Zeng and M. Running for excellent advice on the in situ hybridization protocol. This research was supported by grants from the Illinois–Missouri Biotechnology Alliance 34346-13070 (to O.Y.), Missouri Soybean Merchandising Council 03-242 (to O.Y.), and the National Science Foundation, Plant genome program DBI-0421620 (to G.S.).