The circadian clock regulated RNA-binding protein AtGRP7 autoregulates its expression by influencing alternative splicing of its own pre-mRNA

Authors

  • Dorothee Staiger,

    Corresponding author
    1. Institut für Pflanzenwissenschaften, ETH, CH-8092 Zürich, Switzerland, and
    2. Molekulare Zellphysiologie, Universität Bielefeld, D-33615 Bielefeld, Germany
      For correspondence (fax +49 521 106 6410; e-mail type="email">dorothee.staiger@uni-bielefeld.de).
    Search for more papers by this author
  • Laura Zecca,

    1. Institut für Pflanzenwissenschaften, ETH, CH-8092 Zürich, Switzerland, and
    Search for more papers by this author
  • Dominika A. Wieczorek Kirk,

    1. Institut für Pflanzenwissenschaften, ETH, CH-8092 Zürich, Switzerland, and
    Search for more papers by this author
  • Klaus Apel,

    1. Institut für Pflanzenwissenschaften, ETH, CH-8092 Zürich, Switzerland, and
    Search for more papers by this author
  • Luca Eckstein

    1. Institut für Pflanzenwissenschaften, ETH, CH-8092 Zürich, Switzerland, and
    Search for more papers by this author
    • Present address: Max-Planck-Institut für Züchtungsforschung, D-50829 Köln, Germany.


For correspondence (fax +49 521 106 6410; e-mail type="email">dorothee.staiger@uni-bielefeld.de).

Summary

The clock-regulated RNA-binding protein AtGRP7 is part of a negative feedback circuit through which the protein influences circadian oscillations of its own transcript. Constitutive overexpression of AtGRP7 in transgenic plants leads to the appearance of a low amount of an alternatively spliced Atgrp7 transcript with a premature stop codon. It is generated by the use of a 5′ cryptic splice site in the middle of the intron at the expense of the fully spliced mRNA, indicating a role for AtGRP7 in splice site selection. Accelerated decay of this transcript accounts for its low steady state abundance. This implicates a mechanism for the AtGRP7 feedback loop: Atgrp7 expression is downregulated, as AtGRP7 protein accumulates over the circadian cycle, partly by the generation of an alternate transcript that due to its instability does not accumulate to high levels and does not produce a functional protein. Recombinant AtGRP7 protein specifically interacts with the 3′ untranslated region and the intron of its transcript, suggesting that the shift in splice site selection and downregulation involves binding of AtGRP7 to its pre-mRNA. AtGRP7 also influences the choice of splice sites in the Atgrp8 transcript encoding a related RNA-binding protein, favoring the production of an alternatively spliced, unstable Atgrp8 transcript. This conservation points to the importance of this regulatory mechanism to control the level of the clock-regulated glycine-rich RNA-binding proteins and shows how AtGRP7 can control abundance of target transcripts.

Introduction

Endogenous rhythms in plants with a period of a day, so-called circadian rhythms, manifest themselves in diverse physiological processes such as leaf movement, photosynthetic activity, stomata opening and closure, photoperiodic flower induction as well as gene expression (Barak et al., 2000; McClung, 2001; Somers, 1999; Staiger, 2002). They are controlled by an internal timekeeper, the ‘circadian clock’.

Based on work in several model organisms including Drosophila, Neurospora and humans, a general picture of the molecular clock mechanism has emerged (Dunlap, 1999; Young and Kay, 2001). The oscillators responsible for the generation of circadian rhythms consist of clock proteins whose abundance rises and falls within a day. This is caused by rhythmic transcriptional activation of clock genes and subsequent interference of the clock proteins with activators of their own genes. Superimposed on the regular pattern of transcriptional activation, followed by negative autoregulation, are several layers of post-transcriptional control (Edery, 1999; Suri et al., 1999). Computer modeling has predicted that the half-life of the Drosophila clock transcript period changes over the circadian cycle, with the mRNA being more stable during the rising phase of the mRNA curve and less stable during the declining phase (So and Rosbash, 1997). This implicates trans-acting protein factors in the regulation of period mRNA degradation. It is thought that post-transcriptional control may serve to fine-tune the transcription-based negative feedback loops to a cycle of 24 h.

Although the clock proteins identified in Drosophila, mammals and Neurospora do not have counterparts in higher plants, the basic principle of negative autoregulatory circuits appears to be conserved. Two MYB-type transcription factors, CCA1 (circadian clock associated) and LHY (late elongated hypocotyl), that regulate each other by negative feedback have been implicated in the generation of circadian rhythms (Schaffer et al., 1998; Wang and Tobin, 1998). Their constitutive overexpression causes arhythmicity of transcript oscillations and leaf movement. TOC1 (timing of CAB expression) is a nuclear protein that was identified in a mutant with a short period phenotype (Millar et al., 1995; Strayer et al., 2000). CCA1/LHY and TOC1 interact reciprocally: TOC1 activates LHY and CCA1 expression and LHY/CCA1 in turn repress TOC1. This interdependent loop contributes to circadian clock function in Arabidopsis (Alabadi et al., 2001).

Transcripts encoding putative RNA-binding proteins that undergo circadian oscillations in Arabidopsis thaliana and Sinapis alba and reach their highest level at the end of the day were previously identified (Carpenter et al., 1994; Heintzen et al., 1994; van Nocker and Vierstra, 1993). These AtGRP (A. thaliana glycine-rich RNA-binding protein) and SaGRP proteins feature an N-terminal RNA-binding domain of the RNA recognition motif (RRM) type with the two ribonucleoprotein consensus sequences, RNP-1 and RNP-2, and a glycine-rich C-terminal domain. The AtGRP7 protein level also undergoes circadian oscillations with a slight delay of the protein peak relative to the RNA peak (Heintzen et al., 1997). Constitutive overexpression of the AtGRP7 protein in transgenic plants depresses the oscillation of the endogenous Atgrp7 transcript, placing both the transcript and the protein into a negative feedback circuit (Heintzen et al., 1997). Analysis of transgenic plants carrying Atgrp7 promoter-β-glucuronidase (gus) fusions indicated that the promoter is under transcriptional control by the endogenous clock (Staiger and Apel, 1999). Analysis of crosses between the promoter-gus line and the AtGRP7 overexpressors showed that oscillations of the chimeric Atgrp7 promoter-gus transcript continue in the presence of an elevated AtGRP7 protein level, indicating that the promoter by itself does not mediate the negative feedback of the AtGRP7 protein. Instead, Atgrp7 transcript levels are elevated through rhythmic transcriptional activation by the endogenous clock during the day. When AtGRP7 protein has accumulated after a lag phase, Atgrp7 mRNA accumulation is limited at least partly through the AtGRP7 protein itself at the post-transcriptional level, contributing to a stable high amplitude circadian oscillation of both the transcript and the protein (Staiger and Apel, 1999). Since AtGRP7 is predicted to be an RNA-binding protein, this may involve direct interaction of the protein with its own RNA. This is in contrast to the mechanism of known circadianly regulated negative feedback loops in which clock proteins act mainly through interfering with transcription of their own genes with a 24-h rhythm (Dunlap, 1999; Young and Kay, 2001).

Additionally, AtGRP7 overexpression depresses the oscillations of Atgrp8 encoding a related RNA-binding protein, indicating that AtGRP7 also regulates other clock-controlled transcripts apart from its own (Heintzen et al., 1997). The Sagrp, Atgrp7 and Atgrp8 genes contain introns of approximately 300 nucleotides between RNP-2 and RNP-1 within the RMM. Cloning of cDNAs that retain the upstream half of the intron suggests that the pre-mRNAs undergo alternative splicing at a cryptic 5′ splice site in the middle of the intron (Carpenter et al., 1994; Heintzen et al., 1994; D. S., unpublished).

Here, we show that overexpression of the AtGRP7 protein promotes a change in splice site within the endogenous Atgrp7 pre-mRNA. The use of a cryptic 5′ splice site in the middle of the intron generates an alternatively spliced transcript with a premature stop codon at the expense of the mature mRNA. Steady state abundance of this transcript form is greatly reduced due to a decreased half-life. Recombinant AtGRP7 protein specifically binds to two elements within its own pre-mRNA suggesting that the negative autoregulation of the endogenous transcript by an elevated level of AtGRP7 is brought about by direct contacts between AtGPR7 and its own pre-mRNA.

Results

Specific interaction of AtGRP7 with its own transcript

Constitutive overexpression of the AtGRP7 protein in transgenic plants, designated AtGRP7-ox, depresses the oscillations of the endogenous Atgrp7 transcript by negative feedback. This downregulation is not mediated by the promoter, indicating that AtGRP7 affects abundance of its own transcript at the post-transcriptional level (Staiger and Apel, 1999).

Two prerequisites need to be fulfilled for post-transcriptional autoregulation, firstly a specific, direct or indirect recognition of the transcript by the encoded protein which, secondly, should trigger the change in transcript abundance, presumably by promoting RNA decay. The presence of an RNA-binding domain in AtGRP7 suggests that the autoregulation may be caused by a direct interaction. Therefore, we first tested the ability of AtGRP7 to bind its transcript in vitro.

The Atgrp7 transcribed region is schematically shown in Figure 1(a). 32P-labeled in vitro transcripts, corresponding to the 5′ and 3′ untranslated regions (UTR) as well as to the first and second halves of the intron (Figure 1b), were incubated with AtGRP7-glutathione-S-transferase fusion protein. On native polyacrylamide gels, complex formation is observed with the second half of the intron (Figure 2a) as well as the 3′ UTR (Figure 2b). The intensity of the retarded complexes decreases upon inclusion in the binding reaction of an increasing amount of the unlabeled homologous transcript but not of an unrelated transcript, indicating specific interactions.

Figure 1.

Organization of the Atgrp7-transcribed region and RNA-binding substrates.

(a) Scheme of the Atgrp7 pre-mRNA (top), an alternatively spliced transcript generated by the use of a cryptic splice site at position 337 (middle) and the mature mRNA (bottom). Numbering is relative to the transcription start site (Staiger and Apel, 1993). Hatched: 5′ UTR; dark grey: first half of intron; light grey: second half of intron; black: 3′ UTR.

(b) Fragments derived from the 5′ UTR, the first half of the intron (including 10 bp of exonI and 15 bp of exonII), the second half of the intron (including 15 bp of exonII) and the 3′ UTR that were individually subcloned into pBSK(–) to generate RNA binding substrates by in vitro transcription.

Figure 2.

Electrophoretic mobility shift assay.

32P-labeled in vitro transcript comprising the second half of the Atgrp7 intron (a), the 3′ UTR (b), the 5′ UTR (c) and the first half of the intron (d) were incubated with 500 ng AtGRP7–GST fusion protein in the presence of 30 µg yeast tRNA and resolved on polyacrylamide gels. For competition, unlabeled homologous transcripts or an unrelated transcript derived from the tapetum-specific Satap44 gene (Staiger and Apel, 1999) were added in the amount indicated.

A retarded complex observed with the labeled 5′ UTR is not specifically competed (Figure 2c). Incubation of the first half of the intron with AtGRP7–GST does not result in the formation of a retarded complex (Figure 2d). None of the fragments interacts with purified GST carrier protein lacking the AtGRP7 moiety (not shown).

To localize the sequence motifs involved in AtGRP7 binding to the 3′ UTR, band shift experiments were performed with transcripts successively truncated from their 3′ ends. A transcript terminating at position 1003 retains binding that is competed more efficiently by homologous unlabeled RNA than by an unrelated RNA (Figure 3a). The fragment terminating at position 990 interacts with AtGRP7 only at a reduced concentration of tRNA competitor, but still shows a somewhat preferential competition with homologous RNA (Figure 3b). Further truncation to position 976 abolishes specific complex formation (Figure 3c), implicating the nucleotides between positions 976 and 1003 as target sites.

Figure 3.

Delineation of the AtGRP7 binding site within the 3′ UTR.

32P-labeled in vitro transcripts extending to positions 1003 (a), 990 (b) and 977(c) were incubated with AtGRP7–GST fusion protein in the presence of 30 µg tRNA (a) or 2.5 µg tRNA (b, c). For competition, unlabeled homologous transcripts or the unrelated Satap44 transcript (Staiger and Apel, 1993) were added in the amount indicated. The first lanes contain the free fragments in the absence of AtGRP7–GST.

In a similar way, band shift experiments with the second part of the intron successively truncated from its 3′ end reveals that the intron deleted to position 469 and thus lacking the 3′ splice site (Figure 1) retains binding to AtGRP7 (supplementary Figure S1). Deletion to position 456 strongly reduces the interaction and further deletion to 443 abolishes binding at higher concentrations of tRNA competitor. Thus, the nucleotides between 443 and 469 are part of the binding site(s).

To assess nucleotide specificity, UV-crosslinking in the presence of homoribopolymer competitors was performed (Figure 4). Poly(G) competes most efficiently for binding to both the 3′ UTR and the second half of the intron. Poly(U) also competes, although to a lesser extent. Poly(A) and poly(C) do not compete even at a 500-fold molar excess. These data implicate that the G/U-rich stretches within the delineated binding sites may be involved in contacts with AtGRP7 (supplementary Figure S2).

Figure 4.

UV crosslinking/homoribopolymer competition assay.

20 ng of AtGRP7–GST fusion protein were incubated with 2 fmol 32P-labeled in vitro transcript comprising the second half of intron (top) and 3′ UTR (bottom), respectively, in the absence (–) or in the presence of a 5-fold, 50-fold and 500-fold excess of unlabeled homoribopolymers as indicated.

AtGRP7 influences differential usage of alternative splice sites within its own pre-mRNA

Binding of recombinant AtGRP7 to two elements within its transcript in vitro suggests that a direct interaction of AtGRP7 with its pre-mRNA leads to the changes of the endogenous transcript in AtGRP7-ox plants.

In wild-type plants, two forms of the Atgrp7 transcript are detectable at the time of the circadian maximum (zeitgeber time (zt) 12, 12 h after lights on): the fully spliced transcript containing the uninterrupted AtGRP7 reading frame predominates over a small amount of the unspliced pre-mRNA (Figure 5a,b, lanes 1, 2). Arabidopsis plants expressing a transgenic copy of the Atgrp7 cDNA under the control of the constitutive cauliflower mosaic virus (CaMV) 35S promoter show constitutive high total (endogenous and transgenic) Atgrp7 transcript levels at zt0 and zt12, as shown by hybridization with the Atgrp7 cDNA (Figure 5a, lanes 5, 6). However, hybridization with a gene-specific probe derived from the 5′ UTR shows that the endogenous Atgrp7 transcript level is strongly decreased and the pattern of Atgrp7 transcripts is altered: an intermediate length transcript accumulates in AtGRP7-ox plants (Figure 5b, lane 6). This transcript is not detected in control plants that constitutively overexpress a mutated cDNA with a frameshift introduced upstream of the RRM, which leads to early translation termination so that no functional protein is made (AtGRP7-mut). In these plants, the mutated transcript accumulates to a high level (Figure 5a, lanes 3, 4), but the AtGRP7 protein level is not elevated (Figure 5c) and the endogenous Atgrp7 transcript is not influenced (Figure 5b, lanes 3, 4). This suggests that an elevated level of functional AtGRP7 protein is required for the appearance of the alternate Atgrp7 transcript form.

Figure 5.

Alteration of Atgrp7 transcript pattern upon AtGRP7 overexpression requires functional AtGRP7 protein.

Transgenic plants overexpressing the authentic Atgrp7 cDNA (AtGRP7-ox) or a mutated Atgrp7 cDNA with a premature stop codon (AtGRP7-mut) were grown along with wild-type plants in 16-h light/8-h dark cycles and harvested at zt0 and zt12.

(a) The RNA gel blot was hybridized with the Atgrp7 cDNA to measure total transcript.

(b) The RNA gel blot was hybridized with a gene-specific Atgrp7 probe derived from the 5′ UTR that specifically recognizes the endogenous transcript. u, unspliced pre-mRNA; a, alternatively spliced RNA; s, spliced mRNA; zt, zeitgeber time, hours after lights on.

(c) Immunoblot analysis of AtGRP protein. Protein extracts were prepared from AtGRP7-ox, AtGRP7-mut and wild-type plants harvested at zt12. The immunoblot was sequentially probed with serum against SaGRP1 (top) and β-ATPase (bottom) to confirm equal loading.

(d) Analysis of Atgrp7 transcript forms. RT-PCR was performed on poly(A)-containing RNA from AtGRP7-ox or wild-type plants (left). PCR was performed on subcloned cDNAs corresponding to the mature mRNA (Atgrp7a), the alternatively spliced mRNA (Atgrp7b) or the unspliced pre-mRNA (Atgrp7c) (right) as templates. The DNA gel blot was hybridized with the Atgrp7 cDNA. The sizes of the amplification products as calculated from the genomic sequence are indicated. The RT-PCR product corresponding to the pre-mRNA in wild-type plants was only visible after a much higher number of cycles (not shown).

Previously, three cDNA variants have been isolated corresponding to the fully spliced Atgrp7 mRNA, the pre-mRNA comprising the unspliced 300-bp intron, and an alternatively spliced form retaining only the first 195 bp of the intron, respectively, as has been done for the S. alba homologues (Carpenter et al., 1994; Heintzen et al., 1994; D. S., unpublished). To determine the identity of the transcript form in AtGRP7-ox plants, RT-PCR was performed with plant RNA using primers derived from exonI and the 3′ UTR, respectively (see Experimental procedures; Figure 5d). PCR of cloned cDNAs served as a control. In wild-type plants, a band similar in size to the spliced mRNA is amplified whereas in the overexpressors a band corresponding to the alternatively spliced mRNA accumulates. Sequence analysis of the RT-PCR product confirmed the identity of the alternatively spliced band. Thus, an elevated level of AtGRP7 protein leads to preferential use of a cryptic 5′ splice site, resulting in removal of only the downstream half of the intron.

Accelerated decay of the alternatively spliced Atgrp7 transcript in AtGRP7-ox plants

The low Atgrp7 transcript abundance in AtGRP7-ox plants may be due to an increased turnover. To follow the fate of the transcripts, wild-type and AtGRP7-ox plants were transferred to a solution containing 150 µg ml−1 cordycepin to inhibit RNA synthesis (Holtorf et al., 1999). At different time points plants were withdrawn for RNA analysis.

Mature Atgrp7 RNA, the predominant form in wild-type plants, is reduced to about 50% over the 4-h time course (Figure 6a) while the level of an actin RNA remains constant (Figure 6b). In contrast, the alternatively spliced RNA in the overexpressors practically disappears within the first hour (Figure 6a). Assuming a first order decay, the half-life was calculated to be 0.5 h. Because the mature mRNA is below the detection limit in AtGRP7-ox plants, it was not possible to measure its half-life. Again, the actin control RNA remains more or less constant (Figure 6b). Thus, the alternatively spliced transcript prevailing in AtGRP7-ox plants has indeed a very short half-life.

Figure 6.

Determination of half-life of Atgrp7 transcript forms.

Wild type and AtGRP7-ox plants were raised in 16-h light/8-h dark cycles and transferred to liquid medium supplemented with 150 µg ml−1 cordycepin at zt10. Aliquots were removed at 1-h intervals. The RNA gel blot was hybridized with the gene-specific Atgrp7 probe (a) and an actin probe (b). Hybridization signals were quantified using a Phosphorimager and ImageQuant software. Data representative of four repetitions and two independent transgenic lines are shown.

Ongoing translation is required for the decay of the alternatively spliced Atgrp7 transcript

To test whether the decay of the alternatively spliced transcript depends on ongoing translation, AtGRP7-ox plants were subjected to cycloheximide (CHX) treatment. Plants raised in light–dark cycles were transferred to a solution containing CHX at zt8, during the rising phase of the Atgrp7 oscillation (Figure 7a). After 2 h of CHX treatment, at zt10, the level of the alternatively spliced transcript is significantly higher than in untreated AtGRP7-ox control plants (Figure 7b, compare lanes 2 and 7). This difference is maintained for at least 5 h (zt11 to zt13, compare lanes 3–5 and 8–10). In contrast, the actin transcript level is not augmented by CHX treatment but rather slightly declines (Figure 7b, compare lanes 2–5 and 7–10).

Figure 7.

Influence of inhibition of protein synthesis on the abundance of the Atgrp7 transcripts.

(a) Experimental protocol for the experiments shown in (b) and (c). AtGRP7-ox plants were raised in 16-h light/8-h dark cycles and at zt8 transferred to liquid medium with or without CHX. Cordycepin was added at zt10.

(b) CHX treatment increases steady-state abundance of the alternatively spliced Atgrp7 transcript. AtGRP7-ox plants were transferred to liquid medium without (lanes 1–5) or with CHX (lanes 6–10). Plants were removed at the indicated time points for RNA isolation. The RNA gel blot was hybridized with the gene-specific Atgrp7 probe (top) and an actin probe (bottom).

(c) CHX pre-treatment inhibits the decay of the alternatively spliced Atgrp7 transcript. AtGRP7-ox plants were transferred to liquid medium without (lanes 1–5) or with CHX (lanes 6–10) at zt8. Cordycepin was added at zt10 and plants were removed at the indicated time points. The RNA gel blot was hybridized with the gene-specific Atgrp7 probe (top) and an actin probe (bottom). a, alternatively spliced RNA.

(d) CHX treatment moderately affects abundance of the fully spliced Atgrp7 transcript. Wild-type plants (lanes 1–4) and AtGRP7-ox plants (lanes 5–8) were transferred to medium without (lanes 1–2 and 5–6) or with CHX (lanes 3–4 and 7–8) at zt9. Plants were removed 2 h, at zt11 (lanes 1, 3, 5, 7), and 4 h, at zt13 (lanes 2, 4, 6 and 8) after beginning of the CHX treatment. The RNA gel blot was hybridized with the gene-specific Atgrp7 probe. a, alternatively spliced RNA; s, spliced mRNA.

Moreover, a 2-h incubation with CHX precludes the decay of the alternatively spliced transcript as evidenced by subsequent addition of cordycepin (Figure 7a). At zt10, cordycepin was added to plants pre-treated with CHX for 2 h and to untreated control plants. Whereas in control plants the alternatively spliced transcript largely disappears within 1 h of cordycepin treatment (Figure 7c, compare lanes 2–5 to lanes 2–5 in Figure 7b), in plants pre-treated with CHX, the Atgrp7 level did not change upon cordycepin treatment (compare lanes 7–10 of Figure 7c to lanes 7–10 in Figure 7b). The actin mRNA is not substantially affected by cordycepin treatment, indicating a higher intrinsic stability (Figure 7c, lanes 1–5), and CHX pre-treatment did not augment the level of actin mRNA (Figure 7c, lanes 6–10).

The mature mRNA that prevails in wild-type plants is only moderately affected by CHX treatment (Figure 7d). After 2 h, the Atgrp7 level is comparable to untreated control plants (compare lanes 1 and 3) and after 4 h, CHX-treated wild-type plants show an elevated level of the fully spliced transcript (compare lanes 2 and 4). In contrast, the alternatively spliced transcript in AtGRP7-ox control plants is already strongly elevated 2 h after the beginning of the treatment (compare lanes 5 and 7) as well as 4 h after addition of CHX (compare lanes 6 and 8).

Taken together, these data suggest that the decay of the unstable alternatively spliced and, to a lesser extent, of the mature Atgrp7 transcript, requires ongoing translation. CHX may prevent the translation of labile protein factors critical for the degradation process. Since already within 2 h of CHX treatment, the steady state level of the alternatively spliced transcript has increased, a factor involved in the degradation of this transcript would have to have a very short half-life. Alternatively, stabilization of the transcripts upon addition of CHX may point to a requirement for a component associated with ribosomes.

AtGRP7 influences alternative splicing of the heterologous Atgrp8 pre-mRNA

To test whether AtGRP7 overexpression affects splicing of other transcripts, we first investigated in AtGRP7-ox plants the Atgrp8 transcript encoding a closely related RNA-binding protein. Circadian oscillations of this transcript previously have been shown to be under negative control by AtGRP7 (Heintzen et al., 1997). Atgrp8 contains an intron of similar size to Atgrp7 with the potential cryptic 5′ splice site. RT-PCR analysis showed that in wild-type plants the mature mRNA, as well as the pre-mRNA, predominate, while an intermediate size band is present in only a low amount. In contrast, in the AtGRP7-ox plants this intermediate band is strongly increased in abundance relative to the two others (Figure 8a). Sequence analysis confirms that it corresponds to an alternatively spliced Atgrp8 transcript, suggesting again a switch from the regular to the cryptic 5′ splice mode. As observed in the case of Atgrp7, the alternatively spliced Atgrp8 transcript is stabilized upon CHX treatment of AtGRP7-ox plants, and in plants pre-treated with CHX the Atgrp8 level does not decline upon cordycepin treatment (not shown).

Figure 8.

Influence of AtGRP7 overexpression on heterologous alternatively spliced transcripts.

(a) Analysis of Atgrp8 transcripts. RT-PCR was performed on poly(A)-containing RNA from AtGRP7-ox or wild-type plants (left). The DNA gel blot was hybridized with the Atgrp8 cDNA. The sizes of the PCR products as calculated from the genomic sequence are indicated.

(b) Analysis of U1-70K transcripts. Wild type and AtGRP7-ox plants grown in light–dark cycles were harvested at 4-h intervals. The RNA gel blot was hybridized with a U1-70K probe derived from the expressed sequence tag clone no. 156J20T7 that was kindly provided by the ABRC.

Thus, alternative splicing is conserved among the Atgrp7 and Atgrp8 transcripts, and AtGRP7 overexpression promotes the usage of the cryptic 5′ splice site that gives rise to an unstable Atgrp8 transcript form.

To determine how widespread the effect of AtGRP7 is, an unrelated transcript known to undergo alternative splicing was investigated. The Arabidopsis U1 small nuclear ribonucleoprotein particle 70K gene codes for the U1-70K RNA-binding protein that features an N-terminal RMM, a glycine-rich region and an arginine/serine-rich region. Two transcripts of 2.8 and 1.7 kb, respectively, are produced by alternative splicing, resulting in inclusion or exclusion of a 910-bp intron (Golovkin and Reddy, 1996). Inclusion of the intron leads to the production of an mRNA with a premature stop codon, coding for a truncated protein. In AtGRP7-ox plants the intensity and the ratio of the two bands remain unchanged compared to the pattern in wild-type plants throughout the day (Figure 8b). AtGRP7 thus does not affect splice site selection in the U1-70K pre-mRNA.

Discussion

The predicted RNA-binding protein AtGRP7 is part of a clock-regulated negative feedback loop through which it influences oscillations of its transcript by a post-transcriptional mechanism (Heintzen et al., 1997; Staiger and Apel, 1999). Here, we show that recombinant AtGRP7 protein interacts with the 3′ UTR and part of the intron of its own transcript in vitro, suggesting that the autoregulation may be a consequence of direct interaction of AtGRP7 with its pre-mRNA in vivo.

In transgenic plants constitutively overexpressing AtGRP7, the endogenous Atgrp7 transcript is strongly downregulated. Moreover, an alternate Atgrp7 transcript appears at the expense of the mature mRNA. It is generated by the use of a cryptic 5′ splice site in the middle of the intron and retains the upstream half of the intron. The alternatively spliced transcript does not accumulate in transgenic plants overexpressing a mutated Atgrp7 cDNA that prevents AtGRP7 synthesis (AtGRP7-mut), suggesting that an elevated level of functional AtGRP7 protein is required to activate the cryptic 5′ splice site.

The preferential use of the alternative splice site cannot be caused by titration of a limiting splicing factor in AtGRP7-ox plants because the construct used for overexpression lacks the intron. Furthermore, AtGRP7 is not likely to generally regulate alternative splicing, as evidenced by the pattern of transcripts encoding the Arabidopsis U1-70K protein, another RNA-binding protein with an RMM and a glycine stretch, that is not affected in AtGRP7-ox plants. Instead, alternative splicing of glycine-rich RNA-binding proteins may represent a regulatory event because the cryptic splice site is conserved between the Arabidopsis Atgrp7 and Atgrp8 genes as well as among genes encoding glycine-rich RNA-binding proteins in different plant species (Heintzen et al., 1994; Hirose et al., 1993). In favor of this, cDNAs corresponding to the alternate forms have been identified in S. alba and Nicotiana sylvestris (Heintzen et al., 1994; Hirose et al., 1993). Moreover, in AtGRP7-ox plants selection of the splice site within the related Atgrp8 transcript is also biased towards the internal site.

Steady-state abundance of the alternatively spliced Atgrp7 and Atgrp8 transcripts in AtGRP7-ox plants is strongly reduced as compared to the Atgrp7 and Atgrp8 mRNAs in wild-type plants due to an accelerated decay. It is tempting to project the findings of the overexpressors on the changes in AtGRP7 levels over the circadian cycle in wild-type plants. Transcriptional activation of Atgrp7 during the day leads to accumulation of the transcript and, somewhat delayed, of the protein. Above a certain threshold AtGRP7 prevents further accumulation of its transcript at the post-transcriptional level (Heintzen et al., 1997; Staiger and Apel, 1999). This presumably occurs at least partly by a shift to the alternatively spliced transcript that due to its instability does not accumulate to significant levels. In fact, it is barely detectable in RT-PCR experiments (not shown). Whether the decay of the mature mRNA is also accelerated in the presence of an elevated AtGRP7 level remains unclear, because the very low amount of mature endogenous mRNA in AtGRP7-ox plants precluded determination of its half-life. Compatible with a potential dual role of AtGRP7 on splicing and RNA stability, an AtGRP7-green fluorescent protein fusion is found in the nucleus and the cytoplasm in transiently transfected tobacco BY-2 protoplasts, onion epidermal cells and stably transformed Arabidopsis plants (A. Ziemienowicz, D. Haasen, C. Lippuner, D.S., and T. Merkle, unpublished). A dual and independent role in alternative splicing and stability has been uncovered for mammalian ASF/SF2 that determines, in a dose-dependent manner, the choice of alternative splice sites and recently has been found to influence the stability of a target transcript by binding to a purine-rich sequence element within the 3′ UTR (Lemaire et al., 2002).

To change the choice in splice site, AtGRP7 either represses the regular distal 5′ splice site or activates the internal 5′ splice site. Although we cannot exclude an indirect effect on splicing factors recognizing the respective splice sites, the specific interaction between recombinant AtGRP7 and its own pre-mRNA strongly suggests that the influence on pre-mRNA processing is accomplished by a direct interaction in vivo. A detailed investigation of the consequences of engineered mutations on binding and splicing is hampered by the lack of plant-derived in vitro splicing extracts. However, it seems plausible that the effect on the splice site selection is mediated by AtGRP7 binding to G/U-rich sequences within the second half of the intron. The preference of AtGRP7 for U and G residues correlates with previous observations that the mustard, rapeseed and tobacco homologues of AtGRP7 preferentially bind G/U-rich sequences. The mustard SaGRP protein and GR–RNP from rapeseed additionally bind to poly(C) in vitro (Dunn et al., 1996; Hirose et al., 1993; C. Heintzen and D. S., unpublished).

Feedback regulation at the post-transcriptional level has emerged as a mechanism to control levels of splicing factors that in turn may modulate spatial or temporal expression of specific target genes. The Drosophila factor TRA-2 and the murine splicing factor SR20 autoregulate their own pre-mRNAs. Increasing protein levels activate an alternative splicing event that leads to the expression of a truncated, non-functional protein (Jumaa and Nielsen, 1997; Mattox and Baker, 1991). In transgenic Arabidopsis plants overexpressing the atSRP30 protein, the splicing pattern of several transcripts is altered including atSRP30 itself (Lopato et al., 1999). AtSRP30 belongs to the SR protein family whose distinguishing feature is a C-terminal region enriched for alternating arginine (R) and serine (S) residues downstream of at least one RRM. SR proteins influence splice site selection in a dose-dependent manner in metazoa (Hastings and Krainer, 2001). AtGRP7 is the first example of a plant protein shown to influence utilization of splice sites not belonging to the general class of SR proteins.

Decisions of alternative splicing events involve antagonistic effects of general splicing factors or the participation of tissue-specific and developmentally regulated factors (Brown and Simpson, 1998; Lorkovic et al., 2000; Reddy, 2001). One could envisage AtGRP7 binding to its own pre-mRNA and acting as a gene-specific factor in co-operation with proteins of the basal splicing machinery to control its expression. Resolving such interactions will greatly contribute to our understanding of the basic mechanism of splice site selection in higher plants.

In addition to autoregulating its own expression, AtGRP7 influences splice site selection within the Atgrp8 transcript, promoting the production of an alternatively spliced, unstable transcript. This is a possibility of how the circadianly regulated AtGRP7 feedback loop may exert temporal control on specific target genes: The AtGRP7 autoregulatory loop receives rhythmic input from other oscillators such as LHY/CCA1 and TOC1 (Alabadi et al., 2001; Staiger and Apel, 1999; Staiger and Heintzen, 1999) and in turn regulates a subset of oscillating transcripts. Such temporal control could involve changes in the splicing pattern of potential target genes with concomitant alterations in transcript half-life.

Regulation of alternative splicing may well turn out to be a more widespread factor in the regulation of circadian transcript oscillations. For example, alternative splicing of the transcript encoding the Drosophila clock protein PERIOD (PER) within the 3′ UTR gives rise to two transcripts. At lower temperatures, removal of the intron is favored. This correlates with an earlier rise in period transcript during the circadian cycle and PER protein within the PER feedback loop. Such alterations in the appearance of the PER clock protein during the day has been suggested to influence behavioral rhythms such as fly activity during cold, short days in winter (Majercak et al., 1999). Consistent with post-transcriptional regulation within the circadian system, RNA-binding proteins have been shown to undergo circadian oscillations in Drosophila and Gonyaulax, although their mode of action has not been defined (Mittag et al., 1994; Newby and Jackson, 1996).

The protein synthesis inhibitor cycloheximide prevents the decay of the alternatively spliced Atgrp7 and Atgrp8 transcripts (Figure 7c). CHX may prevent the translation of a labile protein factor critical for the degradation of the alternatively spliced transcript. Alternatively, inhibition by CHX could indicate a requirement for a ribosomal component. In metazoa, transcripts containing premature termination codons are selectively depleted from the cell by so-called nonsense-mediated decay. As the decrease in aberrant mRNAs is presumably due to recognition of the stop codon by ribosomes, protein synthesis inhibitors reverse this mechanism (Carter et al., 1995). Several findings suggest that nonsense-mediated decay does exist in plants although the components have not been identified (Jofuku et al., 1989; Petracek et al., 2000; van Hoof and Green, 1996). The ORF of the alternatively spliced Atgrp7 transcript terminates at a premature stop codon within the retained part of the intron. Its translation would result in a truncated polypeptide lacking the RNP-1 consensus sequence and the glycine-rich C-terminal domain. Immunoblot analysis and immunoprecipitation so far failed to detect this 48 amino acid polypeptide. If the degradation of the alternatively spliced Atgrp7 transcript would involve nonsense-mediated decay, the alternatively spliced transcript that is present in trace amount in wild-type plants should be stabilized and accumulate to a high steady state abundance upon CHX treatment which has not been found in our experimental system (Figure 6d).

Regardless of the decay mechanism, the shift to the unproductively spliced transcript upon increasing AtGRP7 levels limits the production of functional protein and thus may contribute to the daily oscillations of AtGRP7 steady state concentration. Precedents for such a temporal regulation include the developmental control of the Drosophila regulatory gene suppressor-of-white-apricot (Zachar et al., 1987; Zachar et al., 1994). In pre-cellular blastoderm embryos, the primary transcript is processed to the mature 3.5-kb mRNA encoding functional suppressor-of-white-apricot protein. During subsequent development, accumulation of the protein inhibits splicing of the first and/or second intron, giving rise to two transcripts of 4.4 and 5.2 kb that are not translated and apparently non-functional by-products.

In conclusion, specific binding of AtGRP7 to parts of its transcript and the appearance of an alternative unstable splice form in AtGRP7-ox plants provide a possible mechanism for the post-transcriptional circadian autoregulatory loop: in response to a rising AtGRP7 level during the circadian cycle it may bind to its own pre-mRNA, causing a shift to the production of an unstable transcript form that is rapidly depleted and, moreover, does not produce a functional protein.

Experimental procedures

Recombinant AtGRP7–glutathione-S-transferase fusion

The Atgrp7 coding region was amplified with flanking EcoRI and NotI sites and inserted into the vector pGEX-5X-1 (Pharmacia). Extracts from E. coli BL21 harboring this construct were purified on glutathione sepharose according to the manufacturer's recommendations.

RNA–protein interaction

To prepare binding substrates, parts of the Atgrp7 transcribed region (Figure 1) were subcloned into pBSK(–)(Stratagene). The 5′ UTR including the ATG spans positions −5 to +60 relative to the transcription start (Staiger and Apel, 1999). The first half of the intron was amplified from a cDNA corresponding to alternatively spliced Atgrp7 and contains the regions between +162 to +337 and +481 to +495 relative to the transcription start site including 10 bp of exonI and 15 bp of exonII. The second half of the intron spans positions +328 to +495, i.e. comprises 10-bp upstream of the cryptic splice site and 15 bp of exonII. Nested deletions extend from +328 to +469, +456, +443, +428, +409, +389 and +349, respectively. The 3′ untranslated region spans positions +898 to +1118. Nested deletions extend from +898 to +1056, +1016, +1003, +990, +976, +956 and +936, respectively.

Subclones were linearized and transcribed with T7 RNA polymerase (Promega). Heterologous competitor RNA was derived from the tapetum-specific transcript Satap44 (Staiger and Apel, 1993). Quantification on ethidium bromide-stained agarose gels was done relative to RNA standards of known concentration. Radioactively labeled RNAs were made in the presence of 0.4 mm each ATP, CTP, GTP and 2 µm UTP supplemented with 20 µCi 32P UTP. Full-length transcripts were purified on 6% polyacrylamide urea gels.

Transcripts were briefly denatured at 80°C and allowed to re-nature by cooling down to room temperature. Binding assays were performed at room temperature in a total volume of 15 µl of 20 mm HEPES, pH 7.5, 0.1 m NaCl, 1 mm MgCl2, 0.01% NP-40, supplemented with 10 U RNAsin, 500 ng AtGRP7–GST fusion protein, RNA substrates (5000 cpm), and 30 µg of yeast tRNA. Free and bound fragments were resolved on 5% polyacrylamide gels in 40 mm Tris-acetate, 1.8 mm EDTA.

UV RNA–protein crosslinking

Recombinant AtGRP7–GST fusion protein (0.5 pmol corresponding to 20 ng) and denatured 32P-labeled in vitro transcripts (2 fmol) were incubated with or without competitor in a total volume of 10 µl crosslinking buffer (10 mm HEPES-KOH, pH 7.9, 2 mm MgCl2, 50 mm KCl, 0.025% Nonidet P-40, 1 mm DTT, 10% glycerol, 5 U RNAsin (Promega)). UV crosslinking was performed by irradiating the samples on ice in the Stratalinker (Stratagene) for 20 min at 2.3 J cm−2. Samples were subsequently digested with 10 µg of RNAse A for 30 min at 37°C and analysed on 15% SDS-polyacrylamide gels. For competition, poly(A), poly(C), poly(U) and poly(G) homoribopolymers were used in 5, 50 and 500-fold excess (calculated in moles of nucleotides over 32P-labeled RNA).

Chimeric gene construct

To overexpress a mutated Atgrp7 transcript containing a premature termination codon, the protein-coding region was amplified by PCR from the cDNA with the upstream primer 5′-GGCCATGGCGCGGTGATGTTGAGTATCG-3′ covering the translation start (bold) and introducing a frameshift that results in a stop codon (italics) six amino acids downstream of the ATG start codon, and the downstream primer 5′-GGGATCCTTACCATCCTCCACC-3′ covering the translation stop codon (bold). The primers comprise engineered NcoI and BamHI sites, respectively (underlined). The gel-purified 540 bp amplification product was inserted between the CaMV promoter with the duplicated enhancer fused to the tobacco mosaic virus omega element (Gallie et al., 1987) and the CaMV polyadenylation signal, as done for overexpression of the authentic cDNA (Heintzen et al., 1997). The subcloned PCR product was verified by sequencing and the entire cassette was inserted into pBin19.

Plant transformation

The chimeric gene was transformed into Agrobacterium tumefaciens and transferred into A. thaliana Columbia by vacuum infiltration (Bechtold et al., 1993). Kanamycin-resistant seedlings were selected and grown to flowering in soil.

RNA gel-blot analysis

Seeds were germinated on one-half strength MS plates (Murashige and Skoog, 1962) containing 1% sucrose and 50 µg ml−1 kanamycin and grown in 16-h light/8-h dark cycles at a constant temperature of 20°C. After 2 weeks, resistant plants were transferred to one-half MS plates. Plants were harvested at the indicated time points. RNA isolation and hybridization were performed as described (Heintzen et al., 1994; Staiger and Apel, 1999).

RT-PCR analysis of Atgrp7 and Atgrp8 isoforms

Total RNA was reverse transcribed with SuperscriptII reverse transcriptase (Invitrogen) using an oligo dT12-18 primer. cDNAs encoding alternatively spliced isoforms were amplified using primers derived from the first exon, 5′-CTCTTGAGCTGCCTTCG-3′, and the 3′ untranslated region immediately downstream of the translation stop codon, 5′-AGAACATTCATTGGTAATCCC-3′, respectively. Control fragments were amplified from the cloned cDNA forms either without the intron (Atgrp7a) or with the entire intron (Atgrp7c) or the alternatively spliced intron (Atgrp7b) (Carpenter et al., 1994; Heintzen et al., 1997). The DNA gel blot was hybridized with the Atgrp7 cDNA. For sequence determination, the PCR product derived from the overexpressors was eluted from a 5% polyacrylamide gel.

Primers for RT-PCR analysis of Atgrp8 are positioned at the translational start site, 5′-ATGTCTGAAGTTGAGTAC-3′, and downstream of the RMM, 5′-TCTCGACTGAGCCTCGTTC-3′.

Treatment with cordycepin and cycloheximide

Wild type and AtGRP7-ox plants were raised in 16-h light/8-h dark cycles. After 3 weeks, they were transferred to a solution containing 1 mm PIPES-NaOH, pH 6.3, 1 mm KCl, 1 mm sodium citrate, 15 mm sucrose (Holtorf et al., 1999). Cordycepin and cycloheximide were added to a final concentration of 150 µg ml−1 (Holtorf et al., 1999) and 100 µg ml−1, respectively, at the times indicated in the figure legends. Plants were subjected to vacuum infiltration for 4 min and incubated on a shaker at room temperature. Plants were withdrawn at defined time points and quickly frozen in liquid nitrogen for RNA isolation. The half-life of the mRNAs was calculated from PhosphorImager scans using the Origin Program (Microcal Software, Northampton, MA,USA).

Immunoblot analysis

Protein extraction and incubation of protein gel blots with antiserum raised against mustard SaGRP1 followed by chemiluminescence detection were done as described (Heintzen et al., 1997).

Acknowledgements

Thanks are due to Martin Neuenschwander for generating deletion clones of the binding substrates and to Christian Ochsenbein for an Arabidopsis actin probe. We thank the Arabidopsis Biological Resource Center at Ohio State University for providing expressed sequence tags, Stephan Lange for critical comments on the manuscript and Dr Dieter Rubli for help with the figures. This work was supported by grants from the Swiss National Foundation and the ETH Research Commission to D.S.

Supplementary Material

The following material is available from http://www.blackwellpublishing.com/products/journals/suppmat/tpj/tpj1629/tpj1629sm.htm

Ancillary