Nodule Parenchyma–Specific Expression of the Sesbania rostrata Early Nodulin Gene SrEnod2 Is Mediated by Its 3' Untranslated Region

Correspondence to: Frans J. de Bruijn, at Michigan State University Department of Energy Plant Research Laboratory, East Lansing, MI 48824-1312., debruijn{at}pilot.msu.edu (E-mail), 517-353-9168 (fax).

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

The early nodulin Enod2 gene encodes a putative hydroxyproline-rich cell wall protein and is expressed exclusively in the nodule parenchyma cell layer. The latter finding suggests that the Enod2 protein may contribute to the special morphological features of the nodule parenchyma and to the creation of an oxygen diffusion barrier. The Enod2 gene of the stem-nodulating legume Sesbania rostrata (SrEnod2) is induced specifically in roots by the plant hormone cytokinin, and this induction occurs at a post-transcriptional level. Here, we characterize the cis determinant(s) in the SrEnod2 locus responsible for nodule parenchyma–specific expression and show that the 3' untranslated region (UTR) of the SrEnod2 gene is both required and sufficient for directing chimeric reporter gene expression in the nodule parenchyma of transgenic Lotus corniculatus plants. Moreover, we show that the SrEnod2 3' UTR does not act as a tissue-specific enhancer element. By conducting a detailed deletion analysis of the 5' and 3' SrEnod2 regions, we delimited the minimal promoter of the SrEnod2 gene, and it appears that the 5' flanking sequences are not essential for nodule parenchyma–specific expression. This finding is in contrast with the report that the 5' upstream region of the soybean Enod2 gene directs nodule parenchyma–specific expression, indicating that different mechanisms may be involved in regulating the expression of these two genes. We definitively demonstrate that the cis element(s) for tissue-specific expression is located within the 3' UTR of a plant nuclear gene.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Symbiotic interactions between legume plants and soil bacteria of the genera Rhizobium, Bradyrhizobium, and Azorhizobium result in the formation of new plant organs, namely, root and stem nodules. Nodule formation is initiated by the exchange and recognition of signal molecules between the symbiotic partners (Long 1989 ; Fisher and Long 1992 ). Flavonoids and isoflavonoids secreted by roots of legume plants serve as chemoattractants as well as inducers of the bacterial nodulation (nod) genes. Expression of the nod genes, in turn, results in both the synthesis of lipochitooligosaccharides, known as Nod factors, which are capable of initiating the process of nodule formation (Lerouge et al. 1990 ; Truchet et al. 1991 ; Vijn et al. 1993 ; Yang et al. 1994 ), and the signal transduction pathway responsible for activating nodulin gene expression (Pingret et al. 1998 ). When infected with rhizobia, roots of legume plants undergo specific morphological changes, such as root hair deformation and curling, cortical cell division, formation of infection threads, and the release of bacteria via endocytosis, leading ultimately to the development of mature nodules (reviewed in Verma and Delauney 1988 ; Gloudemans and Bisseling 1989 ; Nap and Bisseling 1990 ; Kijne 1992 ; Mylona et al. 1995 ).

Concomitant with the observed morphological changes in the infected plant roots, the expression of specific plant genes is induced or enhanced. These genes are referred to as early- or late-nodulin genes (van Kammen 1984 ). Early-nodulin genes are activated at early stages of nodule development, well before the onset of nitrogen fixation. Their gene products are involved in the infection process and nodule ontogeny. Late-nodulin genes are first expressed around the onset of nitrogen fixation. Their products are generally responsible for nodule functioning (Nap and Bisseling 1990 ; de Bruijn and Schell 1992 ).

The Enod2 gene is an early-nodulin gene that was first identified by differential screening of a soybean root nodule cDNA library (Franssen et al. 1987 ). Isolation of Enod2 homologs from other legumes and subsequent DNA sequence comparisons have shown that the Enod2 gene is highly conserved among leguminous plants (Franssen et al. 1987 ; Dickstein et al. 1988 ; Govers et al. 1990 ; Szczyglowski and Legocki 1990 ; Van de Wiel et al. 1990a ; Dehio and de Bruijn 1992 ). The amino acid sequence derived from the soybean Enod2 cDNA clone revealed that the Enod2 protein is very proline rich and is composed mainly of two repeating pentapeptides, Pro-Pro-His-Glu-Lys and Pro-Pro-Glu-Tyr-Gln (Franssen et al. 1987 ; Van de Wiel et al. 1990a ). The Enod2 proteins of other legumes appear to be composed of the same or similar repeating pentapeptides (Dickstein et al. 1988 ; Szczyglowski and Legocki 1990 ; Van de Wiel et al. 1990a ; Dehio and de Bruijn 1992 ). A putative signal peptide is present at the N terminus of all of the Enod2 proteins identified thus far, and the amino acid sequences of both pea and soybean Enod2 proteins highly resemble a hydroxyproline-rich cell wall protein identified in soybean seed cell walls (Averyhart-Fullard et al. 1988 ; Van de Wiel et al. 1990a ). Therefore, it has been proposed that the Enod2 protein is a cell wall protein (Van de Wiel et al. 1990a ).

The Enod2 gene is expressed during early stages of nodule organogenesis (Franssen et al. 1987 ; Van de Wiel et al. 1990a ; Dehio and de Bruijn 1992 ). It has been shown that this early-nodulin gene is expressed also in empty nodules induced by certain Rhizobium strains on the roots of soybean and alfalfa (Finan et al. 1985 ; Franssen et al. 1987 ; Dickstein et al. 1988 ). These empty nodules contain neither infection threads nor intracellular bacteroids. Expression of the Enod2 gene also has been demonstrated in "pseudonodules" induced by auxin transport inhibitors (Hirsch et al. 1989 ; Van de Wiel et al. 1990b ) or by Rhizobium nodulation-deficient (Nod^-) mutants secreting the cytokinin trans-zeatin (Cooper and Long 1994 ). The expression of the Enod2 gene in these empty/pseudonodules strongly suggests that the Enod2 protein is not involved in the infection process per se but may play a role in nodule ontogeny.

The finding that the Enod2 gene is expressed in alfalfa pseudonodules induced by auxin transport inhibitors (Hirsch et al. 1989 ) extends the observations made several decades ago suggesting that plant hormones are involved in the induction of nodules on plant roots (Thimann 1936 ; Arora et al. 1959 ; Libbenga et al. 1973 ; reviewed in Hirsch 1992 ; Verma 1992 ). In fact, plant hormones clearly appear to be involved in the expression of the Enod2 gene; Dehio and de Bruijn 1992 first demonstrated that the expression of the Enod2 gene of Sesbania rostrata is specifically upregulated in roots after the application of cytokinin. Similar results have been reported for the alfalfa Enod2 gene (Hirsch et al. 1993 ), although the size of the cytokinin-enhanced Enod2 transcript in roots is larger than that observed in alfalfa nodules.

In situ hybridization studies have demonstrated that the soybean Enod2 gene is expressed in the nodule parenchyma and in the nodule tissue surrounding the vascular bundle that connects the nodule to the root central cylinder (Van de Wiel et al. 1990a ). Low levels of the Enod2 transcript also have been found in the nodule endodermis and the nodule outer cortical cells directly adjacent to the endodermis (Van de Wiel et al. 1990a ). The nodule parenchyma is composed of several layers of cells, with fewer and smaller intercellular spaces than most other cortical cells (Van de Wiel et al. 1990a ). The compact structure of these cell layers has suggested to various investigators that they may be involved in forming an oxygen diffusion barrier (Tjepkema and Yocum 1974 ; Witty et al. 1986 ; Nap and Bisseling 1990 ; Van de Wiel et al. 1990a ). Because the Enod2 protein is localized in the same cell layers, it has been hypothesized to contribute to the special morphology of the nodule parenchyma and therefore to the formation of the oxygen barrier (Nap and Bisseling 1990 ; Van de Wiel et al. 1990a ). However, no experimental evidence supporting this hypothesis exists.

Recently, considerable efforts have been made to understand how Enod2 gene expression is controlled in developing soybean and S. rostrata nodules and in roots of S. rostrata plants after cytokinin treatment (Lauridsen et al. 1993 ; de Bruijn et al. 1994 ; Silver et al. 1996 ). We have shown that the cytokinin-mediated upregulation of the SrEnod2 gene in uninfected roots of S. rostrata occurs post-transcriptionally and mainly in the cytoplasm, requires protein synthesis, and appears to involve protein phosphorylation (Silver et al. 1996 ). We also have reported that a 3-kb 5' upstream region of the SrEnod2 gene failed to consistently direct the expression of the uidA reporter gene (gus; Jefferson et al. 1987 ) in the nodule parenchyma of transgenic Lotus corniculatus plants (de Bruijn et al. 1994 ). In contrast, a 3-kb 5' region of the soybean Enod2(B) gene has been shown to direct reporter gene expression in the nodule parenchyma of transgenic L. corniculatus plants (Lauridsen et al. 1993 ). Here, we provide a detailed analysis of DNA sequence determinants in the SrEnod2 locus responsible for tissue-specific expression in nodules. We conclude from this analysis that although the 5' proximal region of the SrEnod2 gene is required for gene expression, the 3' untranslated region (3' UTR) of the SrEnod2 gene contains DNA sequence determinants responsible and sufficient for nodule parenchyma–specific induction.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Localization of the SrEnod2 Transcript in the Parenchyma of S. rostrata Stem and Root Nodules
The tropical legume S. rostrata forms nodules on both roots and stems after infection with A. caulinodans ORS571 (Dreyfus and Dommergues 1981 ; de Bruijn 1989 ). Because the primitive crack entry–mediated infection process of both stems and roots makes the symbiotic interaction between S. rostrata and A. caulinodans ORS571 unique (Tsien et al. 1983 ; de Bruijn 1989 ; Ndoye et al. 1994 ), we studied the tissue-specific expression of the SrEnod2 gene in S. rostrata stem and root nodules. The results presented in Figure 1 show that the SrEnod2 gene is expressed in the nodule parenchyma cells of both stem and root nodules (Figure 1A and Figure 1B). In both types of nodules, hybridization signals were also observed in cells immediately adjacent to the nodule vascular tissue (Figure 1C and Figure 1D).

View larger version (119K): [in this window] [in a new window]   Figure 1. In Situ Localization of SrEnod2 Gene Expression. Sections (7 µm) of S. rostrata root and stem nodules harvested 6 days after inoculation with A. caulinodans ORS571 were hybridized with a digoxigenin-labeled antisense RNA probe derived from the SrEnod2 coding region. Pink and black colors indicate hybridization signals. No signals were observed in neighboring sections hybridized with the sense probe (data not shown). (A) Section of a root nodule. (B) Section of a stem nodule. (C) Magnification of the section shown in (A). (D) Magnification of the section shown in (B). The arrowhead indicates expression in the nodule endodermis. CT, central tissue; NP, nodule parenchyma; VB, vascular bundle.

Moreover, in stem nodules, hybridization signals also were detected in cell layers corresponding to the nodule endodermis (Figure 1D, arrowhead). Expression of the SrEnod2 gene was not observed in other nodule cells. Therefore, the expression pattern of the SrEnod2 gene in the parenchyma cells of S. rostrata stem and root nodules appears to be similar to that of the Enod2 genes of other legumes (Van de Wiel et al. 1990a ).

The 3' Flanking Region of the SrEnod2 Gene Is Required for Gene Expression in the Nodule Parenchyma
To identify DNA sequence determinants responsible for nodule parenchyma–specific expression of the SrEnod2 gene, we constructed a chimeric reporter gene fusion (gus; Jefferson 1987 ; Jefferson et al. 1987 ) in which the 5' and 3' flanking regions of the SrEnod2 gene were connected to the coding region of the gus reporter gene (5' SrEnod2–gus–3' SrEnod2; Figure 2A, construct 1). A construct containing the 3' terminator region of the nopaline synthase gene (3' nos) was used as a control (5' SrEnod2–gus–3' nos; Figure 2A, construct 2). These two constructs were introduced into L. corniculatus plants by Agrobacterium rhizogenes–mediated plant transformation (Szabados et al. 1990 ), and their presence in transgenic plants was verified by DNA gel blotting (data not shown). Expression of the gus reporter gene was determined histochemically and quantitatively in tissues of transgenic plants 21 days after inoculation with rhizobia.

View larger version (32K): [in this window] [in a new window]   Figure 2. Structure of gus Fusion Constructs and Determination of GUS Activity in Transgenic L. corniculatus Plants. (A) Structure of gus fusion constructs. The construct numbers are listed at left. The boxes in black and white represent gus, SrEnod2, and nos gene sequences, as indicated. The vertical lines and wavy arrows represent the transcriptional start site of the SrEnod2 gene, as determined by Dehio and de Bruijn 1992 . The short vertical bars attached to the boxes immediately downstream of the transcriptional start sites represent the translational start of the corresponding genes. The thin open boxes immediately downstream of the SrEnod2 translational start site represent a 75-bp putative signal peptide sequence. (+) and (-) designate positive and negative staining results, respectively. (B) Quantification of GUS activity in transgenic L. corniculatus plants. The data shown represent mean values in units of picomoles of 4-methylumbelliferone (MU) per minute per milligram of protein. The error bars represent the standard deviation of the mean. n indicates the number of transgenic plants tested. Construct numbers are listed at left.

Histochemical staining revealed that gus reporter gene expression was visible as a ringlike pattern in cross-sections of mature nodules expressing the 5' SrEnod2–gus–3' SrEnod2 construct (Figure 3B). No gus reporter gene expression could be detected in leaves, stems, or uninfected roots of the transgenic plants (Figure 2A). However, weak reporter gene expression could be detected in regions near the root tips of some transgenic plants, but only after infection with rhizobia (Figure 3B). When the 3' nos region was present instead of the SrEnod2 3' region, no reporter gene expression could be detected in any tissues of the transgenic plants (Figure 2A and Figure 3A).

View larger version (109K): [in this window] [in a new window]   Figure 3. Microscopic Analysis of the Expression of Reporter Gene Constructs in Transgenic L. corniculatus Plants. (A) Hand-cut sections of transgenic nodules harboring the 5' SrEnod2–gus–3' nos construct (Figure 2A, construct 2). The grayish blue hue of the nodule parenchyma area in these sections does not indicate GUS activity but is an artifact of the microscopy conditions used (reflection of light). (B) Hand-cut sections of transgenic nodules harboring the 5' SrEnod2–gus–3' SrEnod2 construct (Figure 2A, construct 1). GUS activity (blue color) can be seen in the nodule parenchyma and in regions near the root tip of some infected roots. (C) Dark-field microscopy of a 10-µm section of one of the transgenic nodules shown in (A). (D) Dark-field microscopy of a 10-µm section of one of the transgenic nodules shown in (B). GUS activity (brick-red color) can be seen in the nodule parenchyma, vascular, and outer cortical tissues. CT, central tissue; E, endodermis; NP, nodule parenchyma; OC, outer cortical tissue; VB, vascular bundle. Bars in (C) and (D) = 100 µm.

Dark-field microscopy revealed that the expression of the 5' SrEnod2–gus–3' SrEnod2 construct was localized in the nodule parenchyma and vascular tissues (Figure 3D). A low level of GUS activity was detected also in the outer cortex of the nodule. The histochemical staining pattern of the reporter gene constructs correlates well with the in situ Enod2 mRNA hybridization pattern found in S. rostrata nodules (Figure 1) and those observed for other legumes (Van de Wiel et al. 1990a ). The only exception to this correlation is that our reporter gene construct also was expressed in the nodule vascular tissue. Dark-field microscopy confirmed that no GUS activity was detectable in nodules of transgenic plants harboring the 5' SrEnod2–gus–3' nos construct (Figure 3C).

The histochemical staining analysis revealed limited variations in the expression of the reporter gene. Twelve of 17 transgenic plants harboring 5' SrEnod2–gus–3' SrEnod2 (Figure 2A, construct 1) had significant reporter gene expression in nodules, five plants were inactive, and one plant had a high level of reporter gene expression in both nodules and infected roots (data not shown). A variation in reporter gene expression pattern was observed also in one of 18 transgenic plants harboring the 5' SrEnod2–gus–3' nos construct, which had a low level of gus expression in the nodule vascular tissue (data not shown). It is likely that these variations are due to "position effects" (Odell et al. 1985 ; Sanders et al. 1987 ; Benfey et al. 1989 ); therefore, these transgenic plants were excluded from further analysis.

GUS activity in nodules and infected roots of transgenic plants was quantified using a fluorometric assay. As shown in Figure 2B, GUS activity was detected only in nodules of transgenic plants harboring the 5' SrEnod2–gus–3' SrEnod2 construct. No GUS activity was detected in nodules of transgenic plants harboring the 5' SrEnod2–gus–3' nos construct, confirming the results of the histochemical staining analysis. These results strongly suggest that the 3' region of the SrEnod2 gene is required for tissue-specific reporter gene expression in transgenic plants.

In previous studies examining the interaction of DNA binding proteins (trans-acting factors) with the SrEnod2 locus, a specific protein–DNA interaction with sequences encoding the putative SrEnod2 leader peptide was identified (A. Goel and F.J. de Bruijn, unpublished data). Therefore, we also examined the role of the SrEnod2 putative signal peptide sequence in modulating gus reporter gene expression. A 75-bp nucleotide sequence encoding the N-terminal putative signal peptide of 25 amino acids (Dehio and de Bruijn 1992 ) was transcriptionally fused to the truncated SrEnod2 5' flanking region previously used (Figure 2A, constructs 1 and 2). The transcriptional fusions (Figure 2A, constructs 3 and 4) displayed a pattern and level of reporter gene expression similar to the original truncated constructs lacking the leader sequence (Figure 2B). Therefore, the putative signal peptide sequence does not appear to influence the expression of the reporter gene when fused to it in a transcriptional configuration.

gus Reporter Gene Activities Correlate with Steady State gus mRNA Levels in Transgenic Nodules
RNA gel blot analysis was performed to determine whether the gus reporter gene activities observed correlate with the levels of the gus transcript accumulation. As shown in Figure 4, nodule poly(A)⁺ RNA isolated from a pool of three independent transgenic plants harboring constructs containing the SrEnod2 3' region strongly hybridized with the riboprobe prepared from the gus coding region (Figure 4, lanes 1 and 3). The transcript detected was ~2.3 kb in length, as expected. In contrast, nodule poly(A)⁺ RNA isolated from a pool of three independent transgenic plants harboring constructs containing the 3' nos terminator did not hybridize with the gus probe (Figure 4, lanes 2 and 4). The same blot was stripped and reprobed with an eIF4A gene probe (Taylor et al. 1993 ) to standardize loading. The signal peptide sequence of the SrEnod2 gene was not found to influence the accumulation of the reporter gene transcript when fused in a transcriptional configuration to the gus coding region (Figure 4, lanes 1 and 3).

View larger version (31K): [in this window] [in a new window]   Figure 4. RNA Gel Blot Analysis of gus Reporter Gene Expression in Transgenic Nodules Harboring the Reporter Gene Constructs 1 to 4 (Shown in Figure 2A). The lane numbers correspond to the construct numbers in Figure 2A. Two micrograms of poly(A)+ RNA was loaded in each lane and hybridized with a 32P-labeled riboprobe derived from the gus coding region. The same blot was stripped and rehybridized with a 32P-labeled eIF4A probe (Taylor et al. 1993 ) as a loading control.

These data clearly show that the accumulation of the reporter gene transcript correlates well with the gus reporter activity detected in transgenic plants. In addition, they reveal that the reporter gene transcript accumulates in transgenic plants harboring the constructs containing the SrEnod2 3' region but not the 3' nos terminator.

The 3' Region of the SrEnod2 Gene Contains Sequence Determinants Sufficient for Expression in the Nodule Parenchyma
To examine further the role of the SrEnod2 3' region in directing nodule parenchyma–specific expression, we constructed additional chimeric reporter gene fusions containing heterologous promoters, for example, the leghemoglobin (Srglb3) promoter from S. rostrata (Szabados et al. 1990 ; Szczyglowski et al. 1994 ), the nos minimal promoter (-150 to +1 bp; Depicker et al. 1982 ; Ebert et al. 1987 ), and the cauliflower mosaic virus (CaMV) 35S minimal promoter (-90 to +6 bp, domain A; Franck et al. 1980 ; Pietrzak et al. 1986 ; Benfey et al. 1989 ). The resulting constructs (see Figure 5A) were introduced into L. corniculatus, and the expression of the gus gene was determined histochemically and quantitatively.

View larger version (44K): [in this window] [in a new window]   Figure 5. Structure of gus Fusion Constructs and Determination of GUS Activity in Transgenic L. corniculatus Plants. (A) Structure of gus fusion constructs. The construct numbers are listed at left. The open boxes with diagonal lines represent the S. rostrata leghemoglobin gene promoter. The open boxes with horizontal lines represent the CaMV 35S minimal promoter (-90 to +6 bp). The open boxes with vertical lines represent the nos minimal promoter (Pnos; -150 to +1 bp). Other symbols are as described in the legend to Figure 2A. (B) Quantification of GUS activity (picomoles of 4-methylumbelliferone [MU] per minute per milligram of protein) in transgenic L. corniculatus plants. The data represent the mean values. The error bars represent the standard deviation of the mean. Construct numbers are listed at the bottom.

It was shown previously that a 5' Srglb3–gus–3' nos construct (Figure 5A, construct 5) is expressed exclusively in the infected cells of transgenic nodules (Szabados et al. 1990 ; Szczyglowski et al. 1994 ; see also Figure 6A and Figure 6C). However, when the 3' nos terminator was replaced by the SrEnod2 3' region (Figure 5A, construct 6), GUS activity was detected not only in the infected cells of transgenic nodules but also in the nodule parenchyma cells, the vascular tissue, and the outer cortex (Figure 6B and Figure 6D). Dark-field microscopy revealed that GUS expression directed by the 5' Srglb3–gus–3' SrEnod2 construct also could be detected in the central uninfected (interstitial) cells of transgenic nodules (Figure 6D) but could not be detected in transgenic stem, leaf, and root tissues (data summarized in Figure 5A). This expression pattern suggests an additive effect of the leghemoglobin promoter and the SrEnod2 3' region. In addition, the average expression level of the 5' Srglb3–gus–3' SrEnod2 construct was consistently higher than that of the 5' Srglb3–gus–3' nos construct (Figure 5B).

View larger version (132K): [in this window] [in a new window]   Figure 6. Microscopic Analysis of the Expression of gus Fusion Constructs in Transgenic L. corniculatus Plants. Staining for GUS activity was conducted overnight ([C] to [J]) or for 1 hr ([A], [B], [K], and [L]). (A) Hand-cut section of a transgenic nodule harboring the 5' Srglb3–gus–3' nos construct (Figure 5A, construct 5). GUS activity (blue color) can be seen in the nodule central tissue. (B) Hand-cut section of a transgenic nodule harboring the 5' Srglb3–gus–3' SrEnod2 construct (Figure 5A, construct 6). GUS activity (blue color) can be seen in the nodule central tissue as well as in surrounding tissues. (C) Dark-field microscopy of the nodule section shown in (A). GUS activity (blue color) can be seen in the infected cells of the nodule central tissue. (D) Dark-field microscopy of the nodule section shown in (B). GUS activity (blue and brick-red colors) can be seen in the nodule central tissue as well as in the nodule parenchyma, vascular bundles, and outer cortical tissue. (E) Dark-field microscopy of a transgenic nodule section harboring the 5' CaMV 35S(-90/+6 bp)–gus–3' nos construct (Figure 5A, construct 7). (F) and (G) Dark-field microscopy of a proximal and a distal view, respectively, of a transgenic nodule section harboring the 5' CaMV 35S(-90/+6 bp)–gus–3' SrEnod2 construct (Figure 5A, construct 8). GUS activity can be seen in the nodule parenchyma, vascular tissue bundles, and the root central cylinder. No GUS activity was observed in the root cortex. (H) and (I) Dark-field microscopy of a proximal and a distal view, respectively, of a transgenic nodule section harboring the 5' Pnos(-150/+1 bp)–gus–3' SrEnod2 construct (Figure 5A, construct 10). GUS activity can be seen in the nodule parenchyma, vascular bundles, and outer cortical tissue as well as in the root central cylinder and in the root cortex. (J) Dark-field microscopy of a transgenic nodule section harboring the CaMV 35S(-150/-90 bp)–5' SrEnod2(-115/+23 bp)–gus–3' SrEnod2 construct (Figure 8, construct 21). GUS activity can be seen in the nodule parenchyma, the nodule vascular bundles, and the nodule central tissue. (K) and (L) Dark-field microscopy of nodule and root sections, respectively, from plants harboring the CaMV 35S(-150/-90 bp)–5' SrEnod2(-50/+23)–gus–3' SrEnod2 construct (see Figure 8, construct 22). GUS activity can be seen in the nodule parenchyma, vascular bundles, and central tissue as well as in the root central cylinder and cortex. CC, central cylinder; CT, central tissue; E, endodermis; NP, nodule parenchyma; OC, outer cortical tissue; RC, root cortex; VB, vascular bundle. Bars in (C) to (L) = 100 µm.

The 5' CaMV 35S(-90/+6)–gus–3' nos construct (Figure 5A, construct 7) did not direct detectable levels of gus expression in either nodules or roots of transgenic plants (Figure 5A and Figure 6E). However, when the 3' nos terminator was replaced by the 3' SrEnod2 region (Figure 5A, construct 8), gus expression was detected in the nodule parenchyma, vascular bundles, and tissues surrounding the vascular bundle connecting the nodule to the root central cylinder (Figure 6F and Figure 6G). gus expression also was detected in root vascular tissues but not in stems or leaves of transgenic plants (data not shown). Fluorometric measurements confirmed this observation (Figure 5B).

The Pnos(-150/+1 bp)–gus–3' nos construct (where the P in Pnos stands for promoter; Figure 5A, construct 9) did not direct detectable levels of gus expression in nodules but was weakly expressed in the root tissues of transgenic plants (Figure 5B). Again, when the 3' nos terminator was replaced by the 3' SrEnod2 region (Figure 5A, construct 10), gus expression could be detected readily in the nodule parenchyma, vascular bundles, and outer cortex as well as in root vascular tissues and cortical cells (Figure 6H and Figure 6I). No staining was found in stems or leaves (data not shown). GUS activity measurement confirmed these observations (Figure 5B).

These data reveal that the 3' region of the SrEnod2 gene can extend the expression of the gus reporter gene to the parenchyma and vascular tissues of transgenic plants when fused to heterologous promoters, indicating that the SrEnod2 3' region contains sequence determinants necessary and sufficient for directing reporter gene expression in the nodule parenchyma and vascular tissues.

Delimitation of Sequences in the 5' Proximal Region of the SrEnod2 Gene Required for Detectable Levels of Gene Expression
A series of 5' to 3' unidirectional deletions in the 5' SrEnod2 region was constructed and fused to the gus coding region as well as to the 3' region of the SrEnod2 gene (Figure 7A). The resulting constructs were introduced into L. corniculatus plants, and reporter gene expression was analyzed histochemically and quantitatively. GUS activity measurement revealed that constructs with deletions up to position -191, relative to the transcriptional start site mapped by Dehio and de Bruijn 1992 , directed reporter gene expression in the same fashion as did the full-length 5' SrEnod2–gus–3' SrEnod2 construct (Figure 2A, construct 1), namely, to the nodule parenchyma and vascular tissues (Figure 7A and Figure 7B). Only the level of GUS activity varied slightly (Figure 7B). These data suggest that sequences up to position -191 (Figure 7A, construct 18) are not required for nodule parenchyma–specific reporter gene expression. A deletion down to position -115 (Figure 7A, construct 19) was found to lead to a significant reduction of reporter gene expression in nodules (Figure 7B), although a limited degree of nodule-enhanced expression was maintained. Histochemical GUS staining revealed that this construct retained gus gene expression in the nodule parenchyma and vascular tissues (data summarized in Figure 7A). Therefore, we suggest that the sequences upstream of position -115 may not be absolutely necessary for nodule parenchyma–specific expression.

View larger version (35K): [in this window] [in a new window]   Figure 7. Structure of 5' Deletion Constructs and GUS Activity in Transgenic L. corniculatus Plants. (A) Structure of gus fusion constructs. The construct numbers are listed at left. The numbers in base pairs indicate the end point of the 5' deletion relative to the start of transcription. The short vertical bars attached to the boxes represent the translational start of the corresponding genes. The interrupted lines downstream of the gus reporter gene represent the 2.4-kb 3' flanking sequences of the SrEnod2 gene. For other designations, see the legend to Figure 2A. (B) Quantification of GUS activity (picomoles of 4-methylumbelliferone [MU] per minute per milligram of protein) in transgenic L. corniculatus plants. The data represent the mean values. The error bars represent the standard deviation of the mean. n indicates the number of transgenic plants tested. Construct numbers are listed at left.

A construct with a deletion up to position -50 in the 5' SrEnod2 region was found to be inactive in transgenic plants (Figure 8, construct 20). To test whether the lack of reporter gene expression was due to a deletion of a tissue-specific cis element(s) or to a deletion of a proximal promoter element(s), we fused the transcriptional enhancer from the CaMV 35S promoter (-150 to -90 bp, part of domain B; Benfey et al. 1989 ) in cis to the -50-bp fragment (Figure 8, construct 22). GUS activity measurement in transgenic plants harboring this construct revealed that the reporter gene was expressed strongly in nodule, root, and stem tissues (Figure 8). Within nodules, the reporter gene was highly expressed in the nodule parenchyma and vascular tissues and at a lower level in the infected tissue (Figure 6K). Within roots, the gus gene was expressed in the central cylinder and cortical tissues (Figure 6L). When the same transcriptional enhancer was fused to the -115-bp deletion fragment of the 5' SrEnod2 gene (Figure 8, construct 21), a GUS activity staining pattern identical to that of construct 22 (Figure 6K, Figure 6L, and Figure 8) was observed, although GUS activity was much lower than in plants harboring construct 22 (Figure 6J; data summarized in Figure 8). These data suggest that sequences between positions -115 and -50 in the 5' SrEnod2 region are necessary for reporter gene expression. However, they may not contain any tissue-specific cis elements.

View larger version (17K): [in this window] [in a new window]   Figure 8. Structure of 5' Deletion Constructs/Enhancer Fusions and GUS Activity in Transgenic L. corniculatus Plants. The construct numbers are listed at left. The numbers in base pairs indicate the end point of the 5' deletions relative to the start of transcription. The interrupted open boxes represent the 2.4-kb 3' flanking region of the SrEnod2 gene. The open boxes represent the transcriptional enhancer of the CaMV 35S promoter (-150 to -90 bp). For other designations, see the legend to Figure 2A.

The 3' UTR of the SrEnod2 Gene Constitutes the Minimum Determinant for Nodule Parenchyma–Specific Gene Expression
To delineate the sequence determinants in the 3' region of the SrEnod2 gene responsible for nodule parenchyma–specific expression, we constructed a series of 3' to 5' unidirectional deletions in the 3' region of the SrEnod2 gene (Figure 9A). The 3' deletion derivatives were fused to the gus coding region flanked by the full-length 5' SrEnod2 region. The resulting constructs were introduced into L. corniculatus, and reporter gene expression was determined. As shown in Figure 9B, deletions in the 3' region up to position +400 bp relative to the stop codon (Figure 9A, construct 29) did not affect tissue-specific expression of the reporter gene. The level of gene expression was reduced greatly when the 3' flanking sequences were further limited to position +241 bp, although the nodule parenchyma–specific expression pattern appeared to be maintained (Figure 9A, construct 30; data not shown). Therefore, the 241-bp fragment immediately downstream of the stop codon of the SrEnod2 gene appears to be sufficient to direct gene expression to the nodule parenchyma and vascular tissues. This 241-bp fragment contains most of the SrEnod2 3' untranslated sequence because the poly(A) addition site is located between positions 256 and 258 (Dehio and de Bruijn 1992 ; Chen 1996 ).

View larger version (36K): [in this window] [in a new window]   Figure 9. Structure of 3' Deletion Constructs and Determination of GUS Activity in Transgenic L. corniculatus Plants. (A) Structure of gus fusion constructs. The construct numbers are listed at left. The numbers in base pairs indicate the end points of the 3' deletions relative to the stop codon of the SrEnod2 gene. The interrupted lines at the 5' end represent the 1.9-kb 5' SrEnod2 region. The short vertical bars attached to the boxes represent the translational start of the corresponding genes. The vertical arrows indicate the position of the polyadenylation site(s) of the SrEnod2 gene (Dehio and de Bruijn 1992 ; Chen 1996 ). Other designations are as given in the legend to Figure 2A. (B) Quantification of GUS activity (picomoles of 4-methylumbelliferone [MU] per minute per milligram of protein) in transgenic plants. The data represent the mean values. The error bars represent the standard deviation of the mean. n indicates the number of transgenic plants tested. Construct numbers are listed at left.

The 3' UTR of the SrEnod2 Gene Does Not Function as a Transcriptional Enhancer Element
To investigate the potential mode of action of the SrEnod2 3' UTR element, we inserted a fragment carrying 400 bp of 3' UTR sequences (see Figure 9A, construct 29) in both orientations in front of a 684-bp 5' fragment (see Figure 7, construct 14) fused to the gus reporter gene and the 3' nos terminator (Figure 10, constructs 33 and 34). As controls, constructs were made containing the same 684-bp 5' fragment fused to the gus reporter gene and the 400-bp SrEnod2 3' UTR or the 3' nos terminator (Figure 10, constructs 31 and 32, respectively). These constructs were introduced into L. corniculatus plants, and GUS activity was analyzed in different tissues of the transgenic plants by GUS staining. The results are summarized in Figure 10. Although the positive control construct 31 was found to display the typical nodule parenchyma– and vascular bundle–specific expression pattern, neither of the chimeric constructs containing the SrEnod2 3' UTR fused to the SrEnod2 5' (promoter) region and the nos terminator displayed any detectable GUS activity (Figure 10, constructs 33 and 34). Therefore, the SrEnod2 3' UTR does not act as a transcriptional enhancer.

View larger version (16K): [in this window] [in a new window]   Figure 10. Structure of Chimeric Constructs Containing the SrEnod2 3' UTR Fused to the SrEnod2 Promoter Region in cis and GUS Activity in Transgenic L. corniculatus Plants. The construct numbers are listed at left, and the number of transgenic plants analyzed is in parentheses. The arrows in the 3' SrEnod2 boxes indicate the direction of the fusion. For other designations, see the legend to Figure 2A. For construction details, see the text.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

The 3' UTR of SrEnod2 Confers Nodule Parenchyma–Specific Expression
The Enod2 genes of several legumes have been shown to be expressed in the nodule parenchyma cell layer (Van de Wiel et al. 1990a , Van de Wiel et al. 1990b ; Allen et al. 1991 ; Figure 1). Here, we show that in the case of the tropical legume S. rostrata, the primary determinant of this cell-specific expression pattern is located in the 3' UTR of the SrEnod2 gene. The following lines of evidence support this conclusion: (1) Chimeric gus reporter constructs flanked by both 5' and 3' regions of the SrEnod2 gene are expressed in the parenchyma of transgenic L. corniculatus nodules, whereas constructs containing a heterologous 3' region (e.g., 3' nos) are not (Figure 2 and Figure 3). (2) The presence of the SrEnod2 3' region is necessary for the accumulation of the gus mRNA in transgenic nodules (Figure 4). (3) The addition of the SrEnod2 3' region to chimeric reporter gene constructs containing different heterologous promoters (e.g., Srglb3 promoter, CaMV 35S minimal promoter, and nos minimal promoter) extends the expression pattern normally observed with these promoters to the nodule parenchyma in a similar fashion, regardless of the promoter used (Figure 5).

One interesting observation during our analysis of gus reporter gene expression was that all of the gus constructs containing the SrEnod2 3' region directed gus expression to the vascular bundles of transgenic L. corniculatus nodules as well. This observation is in contrast with our in situ hybridization results (Figure 1), which did not reveal SrEnod2 mRNA accumulation in the vascular bundles. It also is in contrast with previous results from soybean Enod2 gene/promoter expression analyses, which did not reveal reporter gene expression in the vascular tissues either (Lauridsen et al. 1993 ). Although we cannot formally rule out the possibility that the observed gus expression in the nodule vascular bundles is due to an artifact, we have observed that positive GUS staining first occurred in the nodule vascular tissues and subsequently in the parenchyma cells rather than in the opposite order. This observation argues against artifacts caused by the diffusion of the GUS enzyme or its colored product and suggests instead that the GUS staining pattern reported here does reflect true gene expression. Consistent with the hypothesis that the S. rostrata Enod2 gene differs from the soybean homolog, the S. rostrata Enod2 gene is inducible by cytokinin (Dehio and de Bruijn 1992 ; Silver et al. 1996 ), whereas the soybean Enod2 gene is not. Moreover, in the case of the soybean Enod2 gene, nodule parenchyma–specific expression has been reported to be controlled by elements far upstream in the 5' region of the gene (Lauridsen et al. 1993 ; also see below).

Another interesting observation was made with the chimeric 5' Srglb3–gus–3' SrEnod2 construct. We found that the SrEnod2 3' region and the Srglb3 promoter have additive effects on cell specificity, because the gus reporter gene was expressed in both the nodule parenchyma/vascular bundles and the infected cells of the central tissue (Figure 6D). In addition, the gus gene also was expressed in the central uninfected cells. This unexpected expression pattern may be due to an unusual interaction of the cis-acting elements in these 5' and 3' regions. Alternatively, because the results shown in Figure 6 suggest that the SrEnod2 3' region appears to confer both a tissue-specific as well as a quantitative function(s), it may reflect a stimulation of Srglb3 promoter activity in the uninfected cells to a previously undetectable level (Szczyglowski et al. 1994 ).

Our SrEnod2 3' deletion analysis revealed that the DNA sequences downstream of position +400 bp, relative to the SrEnod2 stop codon, can be omitted without affecting the expression pattern of chimeric reporter gene constructs (Figure 9). In fact, even a deletion construct lacking DNA sequences 3' to position +241 directs expression to the nodule parenchyma, whereas expression in roots, stems, and leaves is virtually undetectable (Figure 9). This suggests that sequences between the SrEnod2 stop codon and position +241, upstream of the poly(A) addition site (positions +256 to +258; Dehio and de Bruijn 1992 ; Chen 1996 ), are capable of directing cell-specific expression in transgenic nodules, albeit at a low level.

DNA elements in the 3' region of plant genes have been implicated previously in regulation of gene expression. Examples include the potato proteinase inhibitor II gene (Thornburg et al. 1987 ; An et al. 1989 ), the petunia ribulose bisphosphate carboxylase small subunit gene (Dean et al. 1989 ), the oilseed rape AX92 gene (Dietrich et al. 1992 ), the Arabidopsis GLABROUS1 gene (Larkin et al. 1993 ), and the potato sucrose synthase genes (Fu et al. 1995a , Fu et al. 1995b ). Multiple examples have also been reported in mammalian systems. Examples include the chicken adult ß-globin gene (Choi and Engel 1986 ), the chicken histone H5 gene (Trainor et al. 1987 ), the human A globin gene (Bodine and Ley 1987 ), the mammalian ribonucleotide reductase R1 gene (Chen et al. 1993 ), and the human transferrin receptor gene (Owen and Kuhn 1987 ; Binder et al. 1994 ). In these cases, the 3' sequences act either as transcriptional enhancers or as sequence determinants modulating gene expression post-transcriptionally. The cis element(s) acting as a transcriptional enhancer generally is located downstream of the poly(A) addition site(s) and often acts as a tissue-specific and/or developmental regulatory element(s) (e.g., the Arabidopsis GLABROUS1 gene, chicken adult ß-globin gene, chicken histone H5 gene, and human A globin gene). In those cases in which post-transcriptional regulation is involved, the 3' sequences modulate mRNA stability (petunia small subunit ribulose bisphosphate carboxylase gene and human transferrin receptor gene) or efficacy of mRNA 3' end formation and processing (potato proteinase inhibitor II gene). In addition, different 3' end regions also have been found to influence the level of gene expression in plant cells (Ingelbrecht et al. 1989 ).

However, our finding that the cis element(s) responsible for cell-specific expression is located within the 3' UTR appears to be a unique example among plant nuclear genes. The mechanism by which the SrEnod2 3' UTR mediates nodule parenchyma–specific expression remains to be elucidated. In a separate but related study, we have shown that cytokinin-mediated enhancement of SrEnod2 gene expression in S. rostrata roots, in the absence of infection with rhizobia, appears to occur post-transcriptionally (Silver et al. 1996 ). This aspect of SrEnod2 regulation also involves the 3' UTR of the gene (Silver 1996 ). However, we postulate that a post-transcriptional mechanism may be involved in tissue-specific gene expression mediated by the SrEnod2 3' UTR as well, because the SrEnod2 3' UTR element failed to act as a transcriptional enhancer when fused to the SrEnod2 promoter in cis (Figure 10).

Whether nodule parenchyma–specific expression and cytokinin upregulation of the SrEnod2 gene share common regulatory elements currently is not known, and several potential mechanisms, including interaction of the 3' UTR with promoter elements to enhance transcription in nodule parenchyma cells, can be postulated. Cell-specific mRNA stability elements have been identified in the 3' UTR of the 2 globin gene (Weiss and Liebhaber 1995 ). It is possible that the SrEnod2 3' UTR directs nodule parenchyma–specific expression in a manner similar to the 3' UTR of the 2 globin gene.

The 5' Region of SrEnod2 Does Not Appear to Harbor Significant Tissue-Specific cis Elements
Our deletion analysis of the SrEnod2 5' region revealed that sequences upstream of position -191, relative to the transcriptional start site, are not required for nodule parenchyma–specific expression. A deletion up to the -50-bp position abolishes reporter gene expression in nodules, but fusing a transcriptional enhancer (CaMV 35S -150 to -90 bp; see Szabados et al. 1990 ) to the -50-bp fragment restores reporter gene expression to the nodule parenchyma/vascular tissue, to stem and root tissues, and to a limited extent to the nodule central tissue (Figure 6K). This pattern of reporter gene expression was found to be similar to the pattern directed by a fusion of the same transcriptional enhancer to the -115-bp fragment of the SrEnod2 5' region (Figure 6J). These data suggest that sequences between positions -115 and -50 are required for a significant level of reporter gene expression in nodules. However, these sequences are not likely to contain a nodule tissue–specific cis element(s).

The latter conclusion is very intriguing in light of the fact that in the case of the soybean Enod2B gene, Lauridsen et al. 1993 have reported that the 3-kb promoter region directed reporter gene expression to the parenchyma of L. corniculatus and Trifolium repens transgenic hairy root nodules. The 3' nos terminator was used in their study. Two critical tissue-specific cis elements were identified in regions located between positions -1792 and -1582 and between -380 and -53 that were relative to the transcriptional start site of the soybean Enod2B gene. When the upstream tissue-specific cis element was fused to the promoter region of a leghemoglobin gene (lb), the resulting gene fusion was found to direct a low level of reporter gene expression in a scattered pattern in the nodule central area. These data appear to contradict the data presented in this study and/or to suggest further that there are profound differences in the mechanisms regulating SrEnod2 and soybean Enod2B gene expression (also see the discussion above).

Additional support for this concept is the finding that DNA sequences in the promoter regions of the SrEnod2 and the soybean Enod2B genes share remarkably little homology (de Bruijn et al. 1994 ), in contrast to the late-nodulin lb promoters, which are highly similar (Stougaard et al. 1987 ; de Bruijn et al. 1990 ; Szabados et al. 1990 ; de Bruijn and Schell 1992 ; Ramlov et al. 1993 ; Szczyglowski et al. 1994 ). Perhaps this discrepancy can be explained once the nucleotide sequences of the cis element(s) responsible for the nodule parenchyma–specific gene expression are identified and compared directly. Whatever the case may be, the apparent involvement of the 3' region of the SrEnod2 gene, rather than its promoter (5') region, in nodule parenchyma–specific expression is unique in a variety of ways and opens up a whole new area of investigations into the mechanisms of plant gene expression in response to rhizobial infection.

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Plant Growth and Nodulation
Sesbania rostrata seeds were surface-sterilized and germinated on Petri dishes in the dark at 28°C for 1 day, as described by Pawlowski et al. 1991 . Seedlings were grown for 3 weeks in autoclaved soil mix (MetroMix [Scotts-Sierra Horticultural Products, Marysville, OH]/sand; 2:1) in growth chambers with an 18-hr-light (28°C) and 6-hr-dark (22°C) cycle. Plants were inoculated with a 2-day-old culture of Azorhizobium caulinodans ORS571 (Dreyfus and Dommergues 1981 ) either in pots for root nodulation or on stems for stem nodulation. Six-day-old root nodules and stem nodules were collected and processed (see below).

Transgenic Lotus corniculatus plants were grown for 1 week in autoclaved soil mix (vermiculite/sand/MetroMix; 10:10:1) in growth chambers with a 16-hr-light (24°C) and 8-hr-dark (18°C) cycle. Plants were inoculated with a 3-day-old culture of Rhizobium loti NZP2037 (Pankhurst et al. 1983 ) and grown for 21 days. Plant tissues (nodules, infected roots, and leaves plus stems) were collected and stained for ß-glucuronidase (GUS) activity or stored in liquid nitrogen for GUS activity quantification.

In Situ RNA Localization
In situ RNA localization studies were conducted essentially as described by Cox and Goldberg 1988 and Van de Wiel et al. 1990a , except that a nonradioactive approach was used according to De Block and DeBrouwer 1993 . Stem and root nodules were fixed in 10 mM sodium phosphate buffer, pH 6.8, containing 4% paraformaldehyde, 0.25% glutaraldehyde, and 100 mM NaCl at room temperature, dehydrated in a graded ethanol and xylene series, and embedded in Paraplast X-tra (Fisher Scientific, Pittsburgh, PA). Nodule sections (7 µm thick) were cut and attached to Superfrost/plus microscopic slides (Fisher Scientific). After removal of the paraplast with xylene and rehydration through an ethanol series, sections were treated with proteinase K (5 mg/mL) in 100 mM Tris-HCl, pH 7.5, and 50 mM EDTA for 30 min at 37°C and with 0.25% acetic anhydride in 100 mM triethanolamine, pH 8.0, for 10 min at room temperature. Slides were subsequently dehydrated through an ethanol series and dried and hybridized with digoxigenin-labeled sense and antisense RNA probes generated by transcribing the coding region of the S. rostrata early nodulin SrEnod2 gene (Dehio and de Bruijn 1992 ) with T3 and T7 RNA polymerase, according to the manufacturer's instructions (Boehringer Mannheim).

Hybridization and washing were essentially performed according to Cox and Goldberg 1988 and De Block and DeBrouwer 1993 . Sections were dehydrated via an ethanol series and mounted with Polymount (Polysciences Inc., Warrington, PA). Sections were photographed using an Axiophot microscope (Zeiss, Oberkochen, Germany).

Isolation of Total RNA and Poly(A)⁺ RNA
Total RNA was isolated according to the hot phenol method of Verwoerd et al. 1989 , with the following modifications. Briefly, plant material (up to 1 gm) was ground in liquid nitrogen to a fine powder and then transferred into a 15-mL tube containing 4 mL of preheated (65°C) RNA isolation buffer (Hall et al. 1978 ) and 4 mL of preheated (65°C) phenol. Samples were vortexed vigorously for 5 min, 4 mL of chloroform was added, and the mixture was vortexed again for 15 min and centrifuged at 10,000 rpm for 20 min. The supernatant was extracted once with an equal volume of chloroform and centrifuged again. RNA in the supernatant was precipitated at 10,000 rpm for 30 min after adding one-quarter volume of 8 M LiCl and one-tenth volume of 1.6 M KCl and after an overnight incubation on ice. The RNA pellet was resuspended in 400 µL of diethyl pyrocarbonate (DEPC)–treated water and extracted with an equal volume of chloroform until a clear interface was formed. RNA was then precipitated and resuspended in DEPC–H₂O. The RNA concentration was determined photometrically.

Poly(A)⁺ RNA was prepared by passing total RNA through oligo(dT) columns, according to the manufacturer's instructions (5' to 3', Boulder, CO). RNA gel blot analysis was performed according to standard procedures (Maniatis et al. 1989 ). Briefly, Poly(A)⁺ RNA was loaded on a 1% formaldehyde agarose gel and separated at 4°C. The gel was soaked in 10 x SSC (1 x SSC contains 0.15 M NaCl and 0.015 M sodium citrate) for 30 min to remove formaldehyde and subsequently blotted to a 0.22-µm nitrocellulose membrane (NitroPlus; Micron Separation Inc., Westborough, MA). The membrane was rinsed briefly in DEPC-treated water and baked at 80°C for 2 hr under vacuum. Prehybridization was conducted at 51°C overnight in a solution containing 5 x SSC, 10 x Denhardt's solution (Maniatis et al. 1989 ), 0.1% SDS, 0.1 M KPO₄, pH 6.8, and 100 µg/mL salmon sperm DNA. Hybridization reactions were performed at 65°C for 18 hr in a solution containing 1 x SSC, 10 x Denhardt's solution, 0.1 M KPO₄, pH 6.8, 100 µg/mL salmon sperm DNA, 10% dextran sulfate, and 50% formamide. After hybridization, the membrane was washed at 65°C in 2 x SSC and 1 x SSC for 20 min each and finally in 0.2 x SSC for 40 min. Blots were exposed to x-ray film at -75°C.

Chimeric Reporter Gene Construction
A 5.3-kb EcoRI fragment, which contains a 1.9-kb 5' region, a 1-kb coding region, and a 2.4-kb 3' downstream region of the SrEnod2 gene, was subcloned from the genomic clone CD1 (Dehio and de Bruijn 1992 ) into the pBluescript KS- vector (Stratagene, La Jolla, CA). Restriction sites (BamHI and SalI) were introduced at the translational initiation codon and at the 3' end of the putative signal peptide sequence of the SrEnod2 gene by using site-directed mutagenesis, as described by Kunkel 1985 . The oligonucleotide primers used for introducing the BamHI and SalI sites were 5'-GAGTAGTGTAGA-GAGGATCCT TCTATCTAT-3' and 5'-GGCTCATAGTAAT TAGTCGAC-ACTGGAG-3', respectively. The underlined nucleotide sequences correspond to the restriction endonuclease recognition sites, and the boldface nucleotides correspond to the mutations introduced.

A 1.9-kb EcoRI-BamHI fragment, which contains the entire 5' untranslated region (UTR) and upstream region of the SrEnod2 gene, was inserted in front of the gus coding region carried on plasmid pBI101.1 (Clontech Laboratories, Palo Alto, CA). The EcoRI end of the fragment was filled in with the Klenow fragment of DNA polymerase I before cloning. The resulting binary vector was used to generate constructs 1 and 2 (Figure 2A).

To construct a transcriptional fusion between the 1.98-kb 5' sequences of the SrEnod2 gene and the gus coding region, we performed second-round mutagenesis with the 5.3-kb EcoRI fragment containing the introduced SalI site with an oligonucleotide carrying a mutation in the translational initiation codon (ATG to GGA). The resulting 1.98-kb EcoRI-(BamHI)-SalI fragment was excised and inserted in front of the gus coding region carried on plasmid pBI101.1. This vector was used to generate constructs 3 and 4 (Figure 2A).

A restriction site (SacI; underlined and boldface) was introduced immediately downstream of the translational stop codon (TAA) of the SrEnod2 gene by using the oligonucleotide primer 5'-GTAGTG-GTAGTGGT TGGAGCTCT TAAT T T T T T T TGG-3'. A 2.4-kb SacI-EcoRI fragment, containing the entire 3' untranslated and downstream region of the SrEnod2 gene, was inserted into pBI101.1 to generate constructs 1 and 3 (Figure 2A).

Constructs 5 and 6 were generated by inserting a 3.4-kb HindIII-SacI fragment of plasmid pLP14 (Szczyglowski et al. 1994 ), containing a 1.4-kb 5' sequence of the Srglb3 gene and the gus coding region, into pBI101.1 and pBI101-3' SrEnod2, respectively. Constructs 7, 8, 9, and 10 were generated in a similar fashion by inserting into the BamHI-SacI sites of pBI101.1 and pBI101-3' SrEnod2 a 2.1-kb BamHI-SacI fragment of pLP70, containing the 96-bp minimal cauliflower mosaic virus (CaMV) 35S promoter (-90 to +6 bp) and the gus coding region, and a 2.15-kb BamHI-SacI fragment of pLP62, containing the 150-bp minimal nopaline synthase (nos) promoter (-150 to +1 bp) and the gus coding region, respectively.

Several 5' deletion fragments were generated by restriction enzyme digestion of the 1.9-kb EcoRI-BamHI fragment. XbaI, Sau3A, RsaI, EcoRV, DraI, MspI, HinfI, AluI, SspI, and Tsp were used to generate deletions ending at positions -1628, -1180, -854, -684, -464, -384, -306, -191, -115, and -50 bp, respectively, relative to the transcriptional start point. Several 3' deletion fragments were generated by nested deletions, using the Erase-a-Base kit (Promega). A XhoI linker was inserted into the EcoRI site at the end of the 3' sequences of the SrEnod2 gene in construct 1. The 3' deletion fragments were then inserted into the SacI-XhoI sites of the modified construct 1.

In construct 31, the gus reporter gene was fused to a 684-bp 5' SrEnod2 (promoter) fragment and a 400-bp 3' SrEnod2 3' UTR fragment, respectively (Figure 10). In construct 32, the gus reporter gene was fused to the same 684-bp 5' SrEnod2 fragment but was flanked with the 3' nos terminator fragment. Constructs 33 and 34 were directly derived from construct 32 by inserting the SrEnod2 400-bp 3' UTR fragment in front of the SrEnod2 promoter region in both orientations (Figure 10).

Regeneration of Transgenic L. corniculatus Plants
Transgenic plants (L. corniculatus cv Rodeo) were generated according to Szabados et al. 1990 . Binary vectors were conjugally transferred into Agrobacterium rhizogenes A4 (Tempe and Casse-Delbart 1989 ). Regenerated plants were transferred to a growth chamber (Conviron, Asheville, NC) and nodulated as described previously (Szabados et al. 1990 ).

Quantification of GUS Activity
GUS activity was quantified according to Jefferson et al. 1987 . The units of GUS activity used were picomoles of 4-methylumbelliferone produced per minute per milligram of protein. The concentration of protein in the extract was determined by the Bradford assay (Bradford 1976 ). Fluorometric assays were performed using a fluorescence spectrophotometer (model F-2000; Hitachi, Tokyo, Japan).

Histochemical Staining of GUS Activity
GUS activity was analyzed histochemically according to the procedure described by Jefferson 1987 , with the following modifications. Hand-cut nodule sections were incubated at 37°C in a solution containing 50 mM Na₂PO₄ and NaHPO₄, pH 7, 1 mM ferricyanide and ferrocyanide, 20% methanol, and 1 mM 5-bromo-4-chloro-3-indolyl ß-D-glucuronide (Gold Biotechnology Inc., St. Louis, MO). Stained sections were then embedded in historesin (Reichert-Jung; Cambridge Instruments, Heidelberg, Germany), as described by De Block and DeBrouwer 1992 , or in Paraplast X-Tra, as described by Cox and Goldberg 1988 . Sections (10 µm thick) were generated and examined by dark- and light-field microscopy by using an Axiophot microscope.

	FOOTNOTES

¹ Current address: Laboratory of Genetics, University of Wisconsin, Madison, WI 53706.
² Current address: Department of Medicine, Columbia University, New York, NY 10027.

<