One of the groups of genes potentially involved in the senescence process are the nuclease I enzymes. Together with other hydrolytic enzymes, nucleases can provide RNA and DNA degradation products to be used in other parts of the plant as part of the mechanism of nutrient salvage that occurs in the plant following cell death (Bleecker, 1998). Nevertheless, and in spite of extensive studies on the gene expression that occurs during plant senescence (Lohman et al., 1994; Oh et al., 1996; Buchanan-Wollaston, 1997; Gan and Amasino, 1997; Weaver et al., 1998), the genes that encode senescence-induced nucleases have not been identified. To better understand plant senescence, it is important to isolate and study the genes responsible for degradation of both bulk RNA and DNA, i.e. genes for nuclease I enzymes.

All living organisms contain enzymes responsible for the degradation of single-stranded nucleic acids (Gite and Shankar, 1995), including nuclease I proteins (EC 3.1.30.1). Nuclease I enzymes are extracellular heat-stable glycoproteins that degrade both RNA and single-stranded DNA endonucleolytically. They have a preference for bonds adjacent to adenine and produce 5′-phosphoryl-oligo and mononucleotides. Several biochemical characteristics define this family of enzymes: they have a molecular mass range between 31 and 42 kD, are highly sensitive to EDTA, have acidic pH optima, and require Zn²⁺ for activation and for stability (Fraser and Low, 1993). Two fungal nuclease I proteins have been extensively characterized, and their sequences have been determined. These are nuclease P1 from Penicillium citrum (Maekawa et al., 1991) and nuclease S1 from Aspergillus oryzae (Iwamatsu et al., 1991).

Plant nuclease I and other single-strand-specific nucleases are induced during plant growth and developmental processes such as germination, xylem differentiation, the hypersensitive response, stress responses, and senescence (for review, see Bariola and Green, 1997). Many examples of such activities have been reported. They include proteins with nuclease I characteristics from mung bean (Laskowski, 1980), tobacco cell suspension cultures (Oleson et al., 1982), tobacco pollen (Matousek and Tupy, 1984), barley (Brown and Ho, 1986, 1987), zinnia (Zinnia elegans) (Thelen and Northcote, 1989), rye (Siwecka et al., 1989; el Adlouni et al., 1993), and Lentinula edodes (Kobayashi et al., 1995). In addition, a number of activities with some similarities to the tobacco pollen nuclease I (Matousek and Tupy, 1984) have been identified in pollen from various other plants (Matousek and Tupy, 1985). Finally, single-strand-specific nucleases have also been found in a variety of species, including spinach (Strickland et al., 1991; Yupsanis et al., 1996; Christou et al., 1998), scallion (Uchida et al., 1993), wheat chloroplasts (Kuligowska et al., 1988; Monko et al., 1994), pea seeds (Naseem and Hadi, 1987), and pea chloroplasts (Kumar et al., 1995).

Proteins from Arabidopsis with some of the properties of nuclease I enzymes have been identified using activity gels (Yen and Green, 1991). Specifically, a doublet of about 33 kD appears in both RNase and DNase activity gels. More recently, analysis of altered RNase profile (arp) mutants confirmed that nuclease I enzymes exist in Arabidopsis. Several arp mutants lack or overproduce one or both of the 33-kD activity bands in RNase and DNase activity gels, with RNase patterns mirroring the DNase patterns in each mutant (M.L. Abler and P.J. Green, unpublished data). These results confirm the presence of bifunctional nuclease activities at 33 kD.

Until recently, sequence information was available only for several fungal nuclease I genes and short N-terminal regions of proteins with properties of nuclease I enzyme from barley and zinnia. In barley, it was found that aleurone layers secrete a nuclease into the endosperm in response to gibberellic acid (Brown and Ho, 1986). The first 17 amino acids corresponding to the NH₂-terminal sequence of the secreted protein were determined (Brown and Ho, 1987). During xylogenesis in cultured cells, zinnia secretes a single-strand-specific nuclease (Thelen and Northcote, 1989). The NH₂-terminal amino acid sequence of this mature protein was also determined and shown to be similar to that of the barley enzyme (Thelen and Northcote, 1989). Only very recently were the cDNA sequences corresponding to the barley and zinnia proteins described above reported (Aoyagi et al., 1998).

Previously, we characterized three Arabidopsis S-like RNase genes, RNS1, RNS2, and RNS3, that are each induced to different extents during leaf senescence. RNS1 is induced only slightly (Bariola et al., 1994), whereas RNS2 and RNS3 are more strongly induced (Taylor et al., 1993; Bariola et al., 1994). Isolation and characterization of Arabidopsis genes encoding nuclease I enzymes, especially those induced during senescence, will enable us to better understand the role of nucleases in senescence.

We report the identification of an Arabidopsis nuclease I cDNA that is induced specifically during leaf and stem senescence. We also identified two zinnia cDNAs that are similar to the Arabidopsis clone. The Arabidopsis gene, designated BFN1, encodes a bifunctional nuclease I enzyme, a protein with both RNase and DNase activities. The expression characteristics of BFN1 suggest a role in nucleic acid degradation to facilitate nucleotide and phosphate recovery during senescence.

Zinnia (Zinnia elegans cv Envy) leaves were collected from 2-month-old plants grown in growth chambers in the same conditions as for Arabidopsis plants.

All samples were frozen in liquid N₂ immediately after harvesting and stored at −70°C until analysis.

Plasmid p1626 and unmodified pBI121 as a control were introduced into Agrobacterium tumefaciens GV3101 C58C1 Rif^r (pMP90) (Koncz and Schell, 1986) by electroporation using a Gene-Pulse apparatus (Bio-Rad Laboratories, Hercules, CA) according to manufacturer's instructions. Arabidopsis plants were transformed with T-DNA by the vacuum infiltration method of Bechtold et al. (1993) with the modifications described in van Hoof and Green (1996), and at http://www.bch.msu.edu/pamgreen/vac.htm. T₁ seeds from these plants were plated on solid Arabidopsis growing medium (containing 0.8% [w/v] phytagar), which contained 50 μg mL⁻¹ kanamycin for selection of transformants and 500 μg mL⁻¹ vancomycin to limit the growth of A. tumefaciens. One kanamycin-resistant plant from seeds derived from each originally infiltrated plant was transferred to soil. T₁ plants were grown to maturity and T₂ seeds were collected. T₂ seeds were sterilized and plated on Arabidopsis growing medium with 50 μg mL⁻¹ kanamycin. After 2 weeks, seedlings were harvested and analyzed for expression of the BFN1 transgene by RNA blot analysis. Selected lines were also analyzed by RNase and DNase activity gels (Yen and Green, 1991).

A zinnia cDNA library from tracheary elements differentiated in vitro (Ye and Varner, 1993) was screened by plaque hybridization (Sambrook et al., 1989) using a ³²P-labeled BFN1 probe. Positive plaques were purified, and the corresponding cDNAs were sequenced. ZEN2 and ZEN3 cDNA sequences were deposited into the EMBL, GenBank, and DDBJ databases as NucZe1 accession number U90265 and NucZe2 accession number U90266.

The Arabidopsis recombinant inbred lines between ecotypes Columbia and Landsberg erecta were used to map the BFN1 gene (Lister and Dean, 1993). DNA samples from 30 recombinant inbred lines were digested with HincII and analyzed by DNA gel blot as described above, using BFN1 cDNA from p1504 as a probe. Results were scored with RFLP markers using MAPMAKER (Lander et al., 1987) at the Nottingham Arabidopsis Stock Center (http://nasc.nott.ac.uk/new_ri_map.html).

RNase and DNase activities were assayed using activity gels basically as described previously (Yen and Green, 1991). After electrophoresis and before incubation, gels were washed in 100 mm Tris-HCl, pH 7.0, containing 2 μm ZnCl₂ for 50 min to restore the Zn²⁺ needed for nuclease activity. With this method, nuclease activity appears as a clear band on a dark background.

Figure 1 A, Known peptide sequences used to clone BFN1. N-terminal sequences determined via peptide microsequencing of barley and zinnia nuclease I proteins are shown (Brown and Ho, 1986, 1987; Thelen and Northcote, 1989). Sequences were aligned and a consensus (more ...)

The BFN1 cDNA is 1,161 nucleotides long, with the longest ORF from position 53 to 967, encoding a protein of 34.9 kD. Figure 1B shows the first 60 amino acid residues and the corresponding nucleotide sequence of this cDNA. The hydrophobicity profile of the deduced protein identifies a highly hydrophobic potential signal sequence at the N terminus, with a predicted cleavage site (Nielsen et al., 1997) between residues 28 and 29 (highlighted in Fig. 1B). The predicted molecular mass for the mature protein after the cleavage of the signal is 31.9 kD. In addition, three putative N-glycosylation sites are present, at positions 122, 140, and 214 (94, 112, and 186 of the mature protein), based on the presence of the consensus sequence Asn-Xaa-Ser/Thr (Marshall, 1972). These results suggest that BFN1 is a glycosylated protein that is targeted to the secretory pathway.

Figure 2 Alignment of the deduced amino acid sequences of nuclease I enzymes. A, Nuclease BFN1 from Arabidopsis (accession no. U90264) and ZEN2 and ZEN3 (accession nos. U90265, and U90266, respectively) from zinnia, are compared with the deduced amino acid sequences (more ...)

Similarity among these proteins is dispersed throughout the entire sequence. BFN1 and SA6 are the most similar, sharing 74% of identical amino acids. Compared with the zinnia nucleases, BFN1 is more similar to ZEN1 than to ZEN2 or ZEN3, even though the latter two nucleases were identified using BFN1 cDNA as a probe (70%, 52%, and 48% identity, respectively). Among plant nucleases, BFN1 is most distantly related to BEN1, with 46% identity. The PROTDIST/NEIGHBOR programs of the Phylogeny Inference Package were used to generate a gene genealogy with these nuclease I genes (Fig. 2B). According to the consensus tree generated from 2,000 bootstrapped data sets, plant nucleases can be divided in two groups. BFN1 and SA6, along with ZEN1, form one of them. The other group is formed by ZEN2 and ZEN3, together with the monocot BEN1. As expected, fungal nucleases fall in a separate cluster.

BFN1, ZEN2, and ZEN3 contain the major sequence features characteristic of the fungal nuclease I proteins, as highlighted in Figure 2A. Two regions of high similarity include residues surrounding the active sites for both RNase and DNase activities in nucleases P1 and S1 (His residues at positions 60 and 132, respectively, in the BFN1 mature protein) (Maekawa et al., 1991). Nucleases P1 and S1 contain four Cys residues that form two disulfide bonds that are responsible for the tertiary structure of the protein (Iwamatsu et al., 1991; Maekawa et al., 1991). These residues are conserved in the Arabidopsis and zinnia nucleases (+ symbol in Fig. 2A). Similarly, several Trp, His, and Asp residues, which have been implicated in the binding of zinc atoms in nucleases P1 and S1, are also conserved (asterisks in Fig. 2A) (Maekawa et al., 1991; Volbeda et al., 1991; Gite and Shankar, 1992; Gite et al., 1992). Finally, two of the Asn residues glycosylated in the P1 nuclease (Maekawa et al., 1991) are conserved in BFN1 and zinnia nucleases (# symbol in Fig. 2A).

Figure 3 Genomic DNA gel-blot analysis of the BFN1 gene. Genomic DNA (20 μg per lane) from Arabidopsis was digested with BamHI, EcoRI, HincII, or XbaI, electrophoresed in agarose gels, blotted, and probe with 32P-labeled BFN1 cDNA. DNA marker sizes are (more ...)

The BFN1 gene was mapped using 30 Columbia/Landsberg recombinant inbred lines digested with HincII, which generates a RFLP. BFN1 was found to be located close to the top of chromosome 1 (−9.86 cM), between markers mi443 (−9.31 cM) and ATTS0477 (−10.44 cM). While this manuscript was under review, a genomic sequence corresponding to BFN1 was released (BAC T28P6 from position 65,333–67,938). Restriction analysis of the genomic sequence fully confirmed data in Figure 3.

Figure 4 Overexpression of the BFN1 cDNA in Arabidopsis. A, BFN1 expression construct and GUS control construct in plant transformation vectors p1626 and pBI121, respectively. 35S, 35S promoter from cauliflower mosaic virus; NOS 3′, 3′-UTR from (more ...)

Protein extracts from transgenic pBI121 and p1626 plants were assayed for RNase and DNase activity in activity gels as described in “Materials and Methods” (Fig. 4C) (Yen and Green, 1991). A highly intense band of RNase activity of approximately 38 kD was detected in transgenic p1626 plants (Fig. 4C, top). At the same position, no activity was detected in transgenic pBI121 plants. Similarly, a unique band of DNase activity of approximately 38 kD present in transgenic p1626 plants did not appear in extracts from transgenic pBI121 plants (Fig. 4C, bottom). This result demonstrates that BFN1 encodes a bifunctional nuclease from Arabidopsis capable of degrading RNA and DNA.

Overexpression of BFN1 cDNA does not result in any obvious visible phenotype. When plants from transgenic Arabidopsis lines p1626 and pBI121 were grown to maturity in parallel under the same growth conditions, no evident differences in morphological characteristics were detected. In addition, neither timing nor the onset of senescence was altered (data not shown), indicating that overexpression of BFN1 in Arabidopsis does not affect normal plant growth and development.

Figure 5 shows BFN1 mRNA levels in leaves and stems at different developmental stages. A unique mRNA of about 1.2 kb was detected using BFN1 cDNA as a probe. The abundance of the BFN1 mRNA was extremely low in young or mature leaves (lanes YL and ML) but was induced to a high level during senescence (lane SL). Induction was also observed in senescing stems, albeit to a lower extent than in leaves (compare lanes YB, YS, MS, and SS). Relative to eIF4A mRNA, which was used as an internal standard (Taylor et al., 1993; Bariola et al., 1994), BFN1 mRNA levels increased 10- and 2-fold during leaf and stem senescence, respectively.

Figure 5 RNA gel-blot analysis of BFN1 expression during leaf and stem growth and senescence. Lanes contain 10 μg of total RNA extracted from young leaves (YL), mature green leaves (ML), senescent leaves (SL), young bolts (stems 1–3 cm long) (YB), (more ...)

Figure 6 shows RNase and DNase activity gels of Arabidopsis tissues during leaf and stem development and senescence. No RNase activity of the size of BFN1 was detected in young or mature leaves (Fig. 6A, lanes YL and ML) or during stem growth and development (Fig. 6A, lanes YB, YS, and MS). However, such an RNase activity did appear during leaf and stem senescence (Fig. 6A, lanes SL and SS). To correlate this activity with that of BFN1, protein extracts from transgenic p1626 plants overexpressing BFN1 cDNA were electrophoresed in the same gel. BFN1 RNase activity co-migrated with the RNase activity induced during leaf and stem senescence. During stem and especially leaf senescence, additional RNases of approximately 40, 33, and 26 to 23 kD were strongly induced. This made it difficult to observe BFN1 RNase activity in senescent leaves (Fig. 6A). To circumvent this problem, less protein (40 μg compared with 100 μg per lane) from senescent leaves (lane SL) or from plants transgenic for p1626 or pBI121 was applied to the RNase activity gel (Fig. 6B). In this gel, BFN1 RNase activity could be resolved from the major 40-kD activity.

Figure 6 Examination of RNase and DNase activities during leaf and stem growth and senescence. Protein extracts from leaves and stems of wild-type plants at the same stages as in Figure 5 are compared. Protein extracts from transgenic T2 plants transformed with (more ...)

Results from DNase activity gels were very similar, as shown in Figure 6C. DNase activity of approximately 38 kD was not detected in young or mature leaves, but was induced during senescence. A less-intense DNase activity of the same size detected during bolting and stem growth and development was strongly induced during stem senescence. This DNase activity co-migrated with BFN1 DNase activity overexpressed in plants transgenic for p1626. It is not clear why levels of BFN1 mRNA and nuclease activities were not parallel in senescent leaves and stems. Senescing stems have more BFN1 activity but less mRNA than senescing leaves. Perhaps there is some impact of translational or post-translational control on BFN1 in one or both organs. Nevertheless, these data indicate that BFN1 RNase and DNase activities are induced during leaf and stem senescence in both the mRNA and nuclease activity levels in Arabidopsis.

In contrast to senescence, phosphate starvation did not lead to induction of BFN1. When seeds were germinated on complete medium with phosphate for 2 d and then transferred to fresh plates with or without phosphate for 14 d, BFN1 mRNA was not detected in either sample (data not shown). Also, BFN1 mRNA was not detected during germination or during early seedling growth (data not shown). The earliest time point analyzed was 2 d after plating. Under our growth conditions, the seed coat was broken and the radicle had already emerged at this stage. These results provide strong evidence that BFN1 expression is highly specific for leaf and stem senescence and is not detected during phosphate starvation or germination.

We also determined the expression of ZEN2 and ZEN3 expression during zinnia leaf senescence. Interestingly, ZEN3 and, to a lesser extent, ZEN2 mRNA levels were also elevated during leaf senescence (data not shown). This indicates that induction during senescence is a common feature of a number of plant nuclease I enzymes.

Figure 7 BFN1 expression in organs of Arabidopsis. Total RNA (10 μg) from roots (R), stems (S), leaves (L), flowers (Fl), and green siliques (Sl) of wild-type plants and from seedlings transgenic for pBI121 or p1626 were subjected to RNA gel-blot analysis. (more ...)

Bifunctional nucleases in the nuclease I class that degrade both RNA and DNA, have been known to exist in plants for many years, but their molecular analysis began only recently. BFN1, described here, is the first example to our knowledge of a senescence-associated gene encoding a nuclease I enzyme, and is also the first nuclease I cloned and characterized from Arabidopsis. The properties of this gene indicate that BFN1 will be a useful tool in the study of senescence and the degradation of nucleic acid that occurs during this process.

Our observations indicate that BFN1 helps fulfill this role in Arabidopis. We observed a 10-fold increase in BFN1 mRNA levels during leaf senescence and a 2-fold induction in senescing stems. Concomitant with mRNA accumulation, an induction of BFN1 activity was observed in the corresponding senescing tissue. In Arabidopsis, senescence of leaves and stems occurs when the plant is flowering and producing fruits, a time at which the released nutrients likely contribute to completion of fruit development and seed maturation (Nooden, 1988). Thus, it seems likely that BFN1 participates in this process.

A number of other plant nucleases characterized at the protein level have been implicated in senescence. Wheat produces several single-strand-specific nucleases during leaf senescence (Blank and McKeon, 1989). Induction of these nuclease activities can be detected by RNase activity gels at the onset of senescence, just when chlorophyll loss is initiated (Blank and McKeon, 1989). In addition, at least three RNase activities of 20 to 27 kD are induced during wheat senescence (Blank and McKeon, 1991). In Arabidopsis, RNS2 and RNS3, which encode S-like RNases (Taylor and Green, 1991), increase in abundance during senescence (Taylor et al., 1993; Bariola et al., 1994). Another S-like RNase induced during senescence is RNase LX of tomato (Lers et al., 1998). All of these S-like RNases are expressed in non-senescing tissues and therefore their induction does not appear to be senescence specific (Taylor et al., 1993; Bariola et al., 1994; Lers et al., 1998).

Similar to BFN1, the two zinnia nuclease I genes described in this report, ZEN2 and ZEN3, are also induced during senescence at the RNA level. It is not known whether ZEN1, the xylogenesis-associated (Thelen and Northcote, 1989) nuclease I gene isolated previously (Aoyagi et al., 1998), is induced during senescence. According to the notation listed with its GenBank entry, a daylily nuclease I gene, SA6 (accession no. AF082031), may exhibit induction during petal senescence, but no characterization of this gene or its expression has been published. Nevertheless, at least a subset of nuclease I enzymes are senescence associated if not senescence specific.

In addition to senescence, there are several other processes or conditions, including germination, xylogenesis, and phosphate starvation, during which it would be advantageous for the plant to induce nucleic-acid-degrading activities. The bifunctional nucleases, RNases, and DNases, presumably work together with phosphatases and phosphodiesterases to release phosphate from DNA and RNA for remobilization (Glund and Goldstein, 1993). Some enzymes or their genes are known to be induced by more than one of these conditions, such as RNS2 of Arabidopsis, which responds to both senescence and phosphate starvation.

In contrast, our data are consistent with a senescence-specific role for BFN1 in vegetative tissues. BFN1 mRNA was not detected in seedlings grown in phosphate-depleted medium. This is consistent with a previous study that did not detect a RNases of 38 kD in activity gels following phosphate starvation (Bariola et al., 1994). Our data further indicate that, unlike a barley nuclease I (Brown and Ho, 1986, 1987), BFN1 expression is not induced during germination. We did observe BFN1 expression in flowers that cannot solely be explained by the presence of senescing tissues in those preparations. Still other explanations are possible, especially in light of RNS1, which is barely induced by senescence in leaves but markedly expressed in flowers. Another role suggested for nucleic-acid-degrading activities in flowers is to protect that organ, more specifically the style, from invasion by pathogens (Bariola et al., 1994).

Sequences encoding nuclease I enzymes have been conserved throughout evolution. The sequence of Arabidopsis nuclease BFN1, as well as nucleases ZEN2 and ZEN3 from zinnia, are highly similar to the two other plant nucleases that have been cloned (Aoyagi et al., 1998) and to other nucleases from fungi described previously (Iwamatsu et al., 1991; Maekawa et al., 1991). Even though the plant and fungal nucleases differ in length, they contain several conserved regions, including the catalytic sites for the porcine pancreatic DNase I (Paudel and Liao, 1986) and pancreatic RNase A (Blackburn and Moore, 1982; Cuchillo et al., 1997) mentioned above. In addition, all nuclease I enzymes reported to date enter the secretory pathway and are known to be extracellular. The deduced amino acid sequence of the BFN1, ZEN2, and ZEN3 proteins start with a typical signal peptide, indicating that these proteins also enter the secretory pathway. They also lack a KDEL-like sequence for retention in the endoplasmic reticulum or an obvious C- or N-terminal vacuolar targeting signal (Bar-Peled et al., 1996), so they too may be extracellular. It is not yet clear whether the amino acid sequences of senescence-associated nuclease I enzymes have any distinct features, but this issue should be resolved once more genes are characterized.

The deduced BFN1 protein sequence predicts a mature protein of 32 kD after cleavage of the signal peptide. However, expression of the BFN1 cDNA in Arabidopsis results in the production of a unique 38-kD protein with both RNase and DNase activity. It is likely that the size discrepancy occurs because BFN1 is glycosylated at Asn residues during its transit through the secretory pathway. Predicted glycosylated residues based on the consensus sequence Asn-Xaa-Ser/Thr, where Xaa indicates any amino acid residue (Marshall, 1972), are located at positions 94, 112, and 186 of the mature BFN1 protein. At least two of these align with glycosylated Asn residues in nuclease P1 (Maekawa et al., 1991). Another Asn residue in the BFN1 sequence that is not predicted to be an N-glycosylation site (position 142) is conserved in nuclease P1 protein as an Asn residue that is N-glycosylated. Thus, it is highly likely that BFN1 is modified post-translationally by the addition of two or three carbohydrate moieties. A number of the other nucleases, including nuclease S1 (Iwamatsu et al., 1991) and nucleases from mung bean (Laskowski, 1980), pea seed (Naseem et al., 1987), barley seed (Brown and Ho, 1986, 1987), rye germ ribosome (Siwecka et al., 1989), and spinach (Strickland et al., 1991), are glycoproteins, with carbohydrate contents that account for 17% to 29% of the final relative M_r (for review, see Gite and Shankar, 1995).

Now that the BFN1 gene has been isolated and its activity identified, the role of its potential glycosylation sites and its location within the secretory system can be investigated. Further, the strong and specific response of BFN1 to senescence indicates that it should be an excellent tool with which to study the mechanisms of senescence induction, as well as the role of the enzyme in senescence using reverse genetic approaches and other methodologies in Arabidopsis.

We thank Dr. Z.H. Ye for providing the zinnia cDNA library. We also thank Linda Danhof for excellent technical assistance.