The late steps of plant nonsense-mediated mRNA decay




Nonsense-mediated mRNA decay (NMD) is a eukaryotic quality control system that identifies and degrades mRNAs containing premature termination codons (PTCs). If translation terminates at a PTC, the UPF1 NMD factor binds the terminating ribosome and recruits UPF2 and UPF3 to form a functional NMD complex, which triggers the rapid decay of the PTC-containing transcript. Although NMD deficiency is seedling lethal in plants, the mechanism of plant NMD remains poorly understood. To understand how the formation of the NMD complex leads to transcript decay we functionally mapped the UPF1 and SMG7 plant NMD factors, the putative key players of NMD target degradation. Our data indicate that the cysteine–histidine-rich (CH) and helicase domains of UPF1 are only essential for the early steps of NMD, whereas the heavily phosphorylated N- and C–terminal regions play a redundant but essential role in the target transcript degradation steps of NMD. We also show that both the N- and the C–terminal regions of SMG7 are essential for NMD. The N terminus contains a phosphoserine-binding domain that is required for the early steps of NMD, whereas the C terminus is required to trigger the degradation of NMD target transcripts. Moreover, SMG7 is a P–body component that can also remobilize UPF1 from the cytoplasm into processing bodies (P bodies). We propose that the N- and C–terminal phosphorylated regions of UPF1 recruit SMG7 to the functional NMD complex, and then SMG7 transports the PTC-containing transcripts into P bodies for degradation.


Nonsense-mediated mRNA decay (NMD) is a eukaryotic quality control mechanism that degrades mRNAs containing in-frame premature termination codons (PTCs), thereby preventing the accumulation of potentially harmful truncated proteins (Stalder and Muhlemann, 2008; Brogna and Wen, 2009). In addition, NMD controls the expression of several wild-type genes: 1% of the protein coding and approximately 20% of the mRNA-like non-protein-coding Arabidopsis RNAs are regulated by NMD (Kurihara et al., 2009; Rayson et al., 2012). NMD deficiency is lethal in plants, presumably because impaired NMD leads to a constitutive pathogen response (Jeong et al., 2011; Rayson et al., 2012; Riehs-Kearnan et al., 2012; Shi et al., 2012).

Mechanistically, NMD can be separated into two phases. The early steps of NMD include the translation termination-coupled identification of the PTC and the subsequent formation of an NMD complex. During the late phase, the NMD complex triggers the rapid decay of target mRNA.

The PTC identification step is postulated to be conserved within eukaryotes. For efficient translation termination, the eukaryotic release factor 3 (eRF3) component of the terminating ribosome has to bind the poly(A)-binding protein (PABP; Amrani et al., 2004; Silva et al., 2008). However, if NMD-inducing cis elements, such as an unusually long 3′ untranslated region (UTR; mainly in yeasts and invertebrates, less frequently in vertebrates) or a 3′–UTR-located intron (predominantly in vertebrates), are present in the mRNA, eRF3 fails to interact with PABP and instead binds the UPF1 NMD factor (Le Hir et al., 2001; Gatfield et al., 2003; Muhlemann et al., 2008). The irreversible step of NMD is the formation of the functional NMD complex. In yeast, the functional NMD complex is formed when (eRF3-bound) UPF1 recruits the UPF2 and UPF3 NMD factors to the target transcript, either co-translationally or after translocation of the UPF1-bound transcript into P bodies (Sheth and Parker, 2006; Hu et al., 2010). P bodies are cytoplasmic compartments enriched in translationally repressed mRNAs and RNA degradation factors, including decapping proteins (DCPs) and the XRN1 5′ → 3′ exonuclease (Kulkarni et al., 2010). In mammals, the formation of the UPF1-2-3 NMD complex triggers SMG1 kinase-mediated phosphorylation of UPF1 at specific S/TQ motifs (threonine 28 and serine 1096), which leads to translational repression and the decay of the NMD target mRNA (Isken et al., 2008; Okada-Katsuhata et al., 2012).

It has been suggested that the NMD target degradation steps are not conserved among eukaryotes (Nicholson and Muhlemann, 2010). In yeast, NMD complex formation leads to target transcript decay by triggering either deadenylation-independent decapping, followed by Xrn1p-mediated 5′ → 3′ mRNA decay (DCP-XRN pathway) or accelerated deadenylation followed by 5′ → 3′ exonucleolytic degradation (Cao and Parker, 2003; Mitchell and Tollervey, 2003; Hu et al., 2010). In mammals, three non-redundant 14-3-3-like phosphoserine binding domain-containing proteins (SMG5, SMG6 and SMG7) connect NMD complex formation with mRNA degradation. The SMG–SMG7 heterodimer binds a C–terminal phosphoserine (S1096) of UPF1, followed by binding SMG6 to an N–terminal phosphothreonine (T28) (Okada-Katsuhata et al., 2012). NMD target transcript degradation can be initiated by SMG6 endonuclease-mediated cleavage close to the PTC, or by an SMG5–SMG7-activated exonucleolytic pathway (Huntzinger et al., 2008; Eberle et al., 2009). SMG7 is the key factor in the latter pathway. Artificial binding (tethering) of SMG7 to any part of a mammalian mRNA leads to quick target transcript decay. Moreover, SMG7 is localized to P bodies, and relocalizes the mainly cytoplasmic UPF1 to P bodies (Unterholzner and Izaurralde, 2004). It has been suggested that the 14-3-3-like domain of SMG7 binds the S1096-phosphorylated form of UPF1, and remobilizes phospho-UPF1-bound NMD target transcripts into P bodies for degradation (Fukuhara et al., 2005). Other models suggest that SMG7-mediated decay occurs in the cytoplasm, whereas P bodies are only storage compartments for translationally repressed mRNAs (Stalder and Muhlemann, 2009). The SMG5, SMG6 and SMG7 proteins also control the dephosphorylation and recycling of UPF1 (Ohnishi et al., 2003). Human proline-rich nuclear receptor coregulatory protein 2 (PNRC2) can also bind phosphorylated UPF1 and trigger the decapping-mediated decay of the phospho-UPF1-bound transcripts (Cho et al., 2009).

Little is known about plant NMD. It was shown that long 3′–UTRs, 3′–UTR-located introns and upstream open reading frames (uORFs) can trigger NMD in plants (Kertesz et al., 2006; Kerenyi et al., 2008; Nyiko et al., 2009; Benkovics et al., 2011). Plant orthologs of the UPF1, UPF2, UPF3 and SMG7 NMD factors were experimentally identified, whereas orthologs of SMG5, SMG6 and SMG1 have not been found in Arabidopsis (Hori and Watanabe, 2005; Yoine et al., 2006; Wu et al., 2007; Kerenyi et al., 2008). Indirect evidence suggests that the early steps of NMD are conserved between plants and other eukaryotes (Kerenyi et al., 2008).

The late phase of plant NMD has not been systematically studied yet. Previously, we showed that in a tethering assay used to model the late steps of NMD, UPF1 and SMG7 trigger target transcript degradation (Kertesz et al., 2006; Benkovics et al., 2011). Thus we postulated that UPF1 and SMG7 play critical roles in the late phase of plant NMD.

As a first step towards understanding the late phase of plant NMD, we have functionally mapped UPF1 and SMG7. UPF1 and SMG7 mutants were studied in two different assays: (i) in a depletion-complementation test that measures the overall NMD competency of a given construct; and (ii) in a tethering assay that indicates the role of the tested protein in the late phase of NMD. Our data suggest that in plants NMD targeted transcripts are predominantly degraded by an SMG7-mediated, XRN4-independent pathway.


The N- and C–terminal regions of UPF1 play redundant but essential roles in NMD

UPF1 consists of a highly conserved cysteine–histidine-rich (CH) domain and ATPase/helicase domain, and less conserved N- and C–terminal S/TQ-rich regions. The function of different UPF1 regions was studied in mammals. The CH domain binds UPF2 and eRF3, whereas the exact role of the helicase domain in NMD is not known (Ivanov et al., 2008). The ATPase activity of UPF1 is required to disassemble the NMD complex target transcript ribonucleoproteins (Franks et al., 2010). The phosphorylated N- and C–terminal regions of UPF1 recruit SMG6 and SMG5–SMG7, respectively (Okada-Katsuhata et al., 2012). Both terminal regions are essential for mammalian NMD.

As Arabidopsis UPF1 contains several S/TQ motifs in its C–terminal region, and SMG7 harbors a 14-3-3-like phosphoserine-binding domain (Figures S1 and S2), we previously hypothesized that in plants, as in mammals, SMG7 binds C–terminally phosphorylated UPF1 and connects NMD complex formation with target transcript degradation (Kerenyi et al., 2008). In this case, the C–terminal region of UPF1 should be essential for NMD. To test this, we studied the NMD competency of a truncated Arabidopsis UPF1 construct lacking the whole C–terminal region, including all S/TQ motifs (U1ΔC, note that it is a HA-tagged construct) in a previously described virus-induced gene silencing (VIGS) agroinfiltration-based transient gene depletion–complementation assay (for details of the assay, see Figure S3; Benkovics et al., 2011). Briefly, VIGS was used to generate UPF1-silenced Nicotiana benthamiana plants (VIGS:U1), and then the U1ΔC construct was co-agroinfiltrated with a GFP-based NMD-sensitive reporter construct (Gc–I; Figure 1a) into the UPF1-silenced leaf to test whether the U1ΔC protein could complement the UPF1 deficiency of the silenced leaf. The other patches of the same UPF1-silenced leaf were infiltrated with controls. As negative controls, Gc–I was infiltrated alone (–) or was co-infiltrated with UPF2 (Figure 1b, left upper panel). As a positive control, Gc–I was co-infiltrated with hemagglutinin (HA)-tagged UPF1. (Note that the P14 RNA silencing suppressor, which does not inhibit either NMD or VIGS, but suppresses the agroinfiltration-triggered RNA silencing, is always co-infiltrated with each sample in NMD competency and tethering assays. Co-infiltrated P14 also serves as an internal control for RNA gel blot assays. P14 is shown in figures, but will not be mentioned in the main text. For further details about the role of P14 in these assays see Figure S3.)

Figure 1.

Functional mapping of UPF1. (a) Schematic, non-proportional representation of the constructs used. (b–f) Nonsense-mediated mRNA decay (NMD) competency tests of truncated UPF1 constructs. To functionally map UPF1, Gc–I NMD reporter plus P14 silencing suppressor were co-infiltrated into UPF1-silenced leaves (VIGS:U1), with UPF1 deletion constructs lacking C-, N-, N- and C–terminal, or N–terminal and CH regions (U1ΔC, λN–U1ΔN, λN–U1ΔNΔC, λN-ΔCH). As controls, Gc–I+P14 leaves were infiltrated alone (–), or Gc–I+P14 leaves were co-infiltrated with λN or UPF2 negative controls, or with differentially tagged UPF1-positive control constructs (UPF1 and λN–U1). P14 was co-infiltrated to suppress RNA silencing and served as an infiltration control for RNA gel-blot assays. (b) Bright fluorescence shows that NMD is inactive, whereas weak fluorescence indicates that co-infiltration of the UPF1 mutant complemented the NMD deficiency of the UPF1-silenced leaf. (c–f) RNA gel blots were hybridized with GFP and P14 probes. To quantify RNA samples, at each lane the signal of the Gc–I band was normalized to the corresponding P14 signal. Mean values were calculated from three independent samples (± SD, n = 3) and compared. The mean Gc–I/P14 ratio of (–) samples was taken as 1. Photos and RNA samples were taken at 3 dpi (days post inoculation) (g, h) UPF1 N- and C–terminal regions are phosphorylated. UPF1 (λN–U1) and different UPF1 deletion mutants (U1ΔC, λN–U1ΔN and λN–U1ΔNΔC) were immunoprecipitated with HA antibody from protein extracts derived from agroinfiltrated leaves. (g) From each elute 1- and 3–μl volumes were separated on SDS-PAGE, and the gels were stained with control or phospho-specific stains. (h) From a different infiltration, U1ΔC and λN–U1ΔNΔC immunoprecipitation (IP) samples were treated with alkaline phosphatase (AP, +) or with buffer (–), separated on gel, and phosphor- or control-stained. Note that λN–U1 and λN–U1ΔN could not be tested in phosphatase assays, as these samples were degraded in AP buffer.

As NMD does not function in UPF1-silenced leaves, agroinfiltration of the NMD-sensitive Gc–I reporter leads to strong expression. Indeed, the green fluorescence was strong and the Gc–I transcript level was high in Gc–I infiltrated (–) and in Gc–I+UPF2 co-infiltrated patches (Figure 1b, left upper panel, and Figure 1c, lanes 1 and 2). These data show that UPF1 silencing was efficient in the test leaf, and that Gc–I expression is not decreased aspecifically by the co-infiltration of UPF2. Previously we showed that UPF1 co-infiltration complements the UPF1 deficiency of the silenced leaf, and restores NMD (Kerenyi et al., 2008). In line with these results, co-infiltration of UPF1 resulted in weak green fluorescence and low Gc–I mRNA levels, indicating that the restored NMD efficiently degraded the Gc–I trancripts (Figure 1b left upper panel, and Figure 1c, compare lanes 1 and 3). Surprisingly, U1ΔC co-infiltration reduced Gc–I mRNA expression with a comparable efficiency to the UPF1 construct (Figure 1b, left upper panel, and Figure 1c, compare lanes 3 and 4). These results indicate that U1ΔC fully complemented the NMD deficiency of the UPF1-silenced leaf. Thus in plants, unlike in mammals, the C–terminal S/TQ-rich region of UPF1 is not essential for NMD.

In plants, S/TQ motifs are also present in the N-terminal region of UPF1 (Figure S1). Therefore, we hypothesized that SMG7 binds the N–terminal S/TQ-rich region of UPF1, and that the N–terminal region is essential for NMD. However, the N–terminal deletion UPF1 construct (λN–U1ΔN) complemented the NMD deficiency of UPF1-silenced leaves with comparable efficiency to the similarly tagged UPF1 (λN–U1) (Figure 1b, right upper panel, and Figure 1d). Thus, we concluded that the N–terminal region of UPF1 is not required for NMD. (Note that in this assay and in several further experiments λN- and HA-tagged UPF1 was used. The λN tag does not play a role in complementation, but is required for tethering assays; Benkovics et al., 2011; also see below.)

Our results suggested that either both the N- and C–terminal regions of UPF1 are dispensable for NMD, or their function is redundant. To differentiate between these two possibilities, NMD competency of a UPF1 construct lacking both terminal regions (λN–U1ΔNΔC) was tested. In spite of the comparable protein expression levels (Figure S4), the λN–U1ΔNΔC construct complemented UPF1 deficiency much less efficiently than the λN–U1ΔN control construct (Figure 1b, left bottom panel, and Figure 1e). Taken together, our data indicated that the N- and C–terminal S/TQ-rich regions of UPF1 play a redundant but essential role in NMD.

Phosphorylation of UPF1

Phosphorylation of UPF1 is a prerequisite of the late steps of mammalian NMD (Okada-Katsuhata et al., 2012). As the plant UPF1 contains S/TQ motifs both in its N- and C–terminal regions (Figure S1), and these regions play a redundant role in NMD, we hypothesized that both regions are phosphorylated. To test this, constructs expressing full-length UPF1 (λN–U1) or expressing N-, C-, or N- and C–terminally truncated proteins (λN–U1ΔN, U1ΔC and λN–U1ΔNΔC, respectively) were agroinfiltrated. The expressed proteins were immunoprecipitated from the leaf extract by HA antibodies (each construct was HA-tagged), separated on SDS-PAGE and stained with ProQ-Diamond phosphoprotein gel-stain or PAGE-Blue non-specific control stain. Interestingly, the full-length λN–U1 as well as the λN–U1ΔN and U1ΔC truncated proteins were heavily phosphorylated relative to the λN–U1ΔNΔC sample (Figure 1g). These results support the assumption that both the N- and the C–terminal regions of UPF1 are phosphorylated.

To confirm that the phospho-staining was specific, the U1ΔC and λN–U1ΔNΔC samples were treated with phosphatase or as a control with phosphatase buffer (the λN–U1 and the λN–U1ΔN constructs could not be further analyzed, as these samples were degraded in the buffer). The phosphatase treatment significantly reduced the strong phospho-staining of the U1ΔC, but did not alter the weak phospho-staining of the λN–U1ΔNΔC sample (Figures 1h and S5). These data suggest that λN–U1ΔNΔC shows only non-specific background phospho-staining, whereas the strong phospho-staining of U1ΔC is specific. These results confirm that the N–terminal region of plant UPF1 is phosphorylated.

Taken together, our data strongly indicate that the N–terminal UPF1 region is phosphorylated, and suggest that the C–terminal region is also phosphorylated.

The N- or C–terminal regions of UPF1 are required for target-level reduction in tethering assays

Next we wanted to test whether the redundant N- and C–terminal UPF1 regions are involved in the early or in the late steps of NMD. To test this, the activity of the λN–U1ΔNΔC construct was studied in a tethering assay (for a detailed description, see Figure S6), in which the NMD factor studied is artificially bound to the 3′–UTR of the target mRNA. The tethering assay bypasses the early steps of NMD and models the late phase of NMD. If the tethering of an NMD factor leads to target transcript decay, it is interpreted as an indication that the given protein plays a role in the late steps of NMD.

The tethering assay harnesses the specific and strong binding of the λN peptide to boxB RNA sequences. A tethering target construct was created by inserting five boxB sequences into the 3′–UTR of a GFP reporter (G–3′bB, Figure 2a), and then it was infiltrated alone (–) or was co-infiltrated with the λN–U1 positive control or with the λN–U1ΔNΔC test construct. Confirming our previous result that the tethering of UPF1 triggers rapid target mRNA degradation (Kertesz et al., 2006), we found that G–3′bB transcript levels were strongly reduced in λN–U1 co-infiltrated patches (Figure 2b, right upper panel, and Figure 2c, compare lanes 1 and 2). Importantly, co-infiltration of λN–U1ΔNΔC did not reduce the expression of G–3′bB (Figure 2b, right upper panel, and Figure 1c, compare lanes 2 and 3). By contrast, tethering of either the N–terminal (λN–U1ΔN) or the C–terminal deletion construct (λN–U1ΔC) induced target transcript reduction almost as efficiently as the λN–U1 control (Figure 2b, left upper panel, and Figure S7). These data indicate that the presence of either the N- or C–terminal region of UPF1 is essential to induce degradation of the tethering target transcript, and suggest that these regions play a redundant but essential role in the late steps of NMD.

Figure 2.

Tethering assays of UPF1 mutants. (a) Schematic, non-proportional representation of the constructs used (see also Figure 1). (b–e) Wild-type leaves were infiltrated with G–3′bB tethering target construct+P14 (–), or G–3′bB+P14 leaves were co-infiltrated with λN-negative or λN-tagged UPF1-positive control (λN–U1) or with λN-tagged test constructs (λN–U1ΔNΔC, λN–U1ΔNP4, λN–U1ΔNP9, λN–U1ΔNP13, λN–U1DN and λN–U1ΔCH). Weak fluorescence and low G–3′bB mRNA level suggests that the co-infiltrated construct induced degradation of the G–3′bB tethering target transcripts. To quantify target transcript levels, G–3′bB/P14 signal ratios were calculated from three samples. The G–3′bB/P14 ratio of the λN (c, d) or U1 (e) sample was taken as 1.

To further map which segment of the long C–terminal region is involved in NMD, C–terminal deletion constructs were generated from the λN–U1ΔN construct, and then these mutants were tested in tethering and NMD competency assays. The constructs contained 13 (λN–U1ΔNP13), nine (λN–U1ΔNP9), four (λN–U1ΔNP4) and zero (λN–U1ΔNΔC) C–terminal (and zero N–terminal) S/TQ motifs, respectively (Figure 2). Interestingly, tethering target levels correlated with the length of the C–terminal region. Whereas tethering of the λN–U1ΔNΔC construct did not reduce the expression of the target transcripts, tethering of the λN–U1ΔNP4, λN–U1ΔNP9 and λN–U1ΔNP13 constructs showed gradually increasing tethering target degradation activity (Figure 2b, right upper panel, and Figure 1c). Extension of the UPF1 C–terminal region gradually restored NMD competency as well (Figure S8). Whereas the λN–U1ΔNΔC construct failed to restore the NMD activity of UPF1-silenced leaves, the λN–U1ΔNP4 construct partially complemented NMD deficiency. The λN–U1ΔNP9 and λN–U1ΔNP13 constructs complemented the NMD deficiency of the UPF1-silenced leaves almost as effectively as the λN–U1ΔN positive control construct. The findings that the studied C–terminal segments show additive effects in both assays suggest that these segments play similar roles in NMD. We speculate that the longer UPF1 C–terminal region leads to more extensive phosphorylation, and in turn, to more efficient NMD target degradation.

The CH and helicase domains of UPF1 are required for the early steps of NMD

Next, we studied the role of the CH domain by expressing a truncated UPF1 lacking the N–terminal region and the CH domain (λN–U1ΔCH). In an NMD competency test, λN–U1ΔCH could only partially complement UPF1-silenced leaves (Figure 1b, right bottom panel, and Figure 1f). However, in the tethering assay it decreased G–3′bB expression with comparable efficiency to λN–U1 (Figure 2b, left bottom panel, and Figure 2d). These results suggest that the CH domain is involved only in the early phase of NMD.

The helicase domain of plant UPF1 is essential for NMD. If a conserved arginine residue, which is essential for helicase activity of human UPF1, is changed to cysteine (R844C) in the plant UPF1 (U1DN), it has a dominant-negative effect on NMD (Kertesz et al., 2006). To test whether helicase activity is essential for early and/or late steps of NMD, a λN-tagged version of U1DN (λN–U1DN) was co-infiltrated with the G–3′bB tethering reporter construct. Surprisingly, λN–U1DN reduced the G–3′bB mRNA level with comparable efficiency to the wild-type λN–UPF1 (Figure 2b, right bottom panel, and Figure 1e). This result suggests that the helicase domain of UPF1, like the CH domain, is required only for the early phase of NMD.

The N–terminal domain of SMG7 is required for the early steps of NMD, and the C–terminal domain is required for the late steps of NMD

The N–terminal region of the mammalian SMG7 comprises a 14–3–3-like domain and binds phospho-UPF1, whereas its C–terminal region is responsible for P–body localization and the induction of target mRNA decay (Fukuhara et al., 2005). A 14–3–3-like domain was also identified in the N–terminal region of plant SMG7; however, the C–terminal region is divergent between plants and mammals (Figures S2 and S9; Benkovics et al., 2011; Riehs et al., 2008). To functionally map the plant SMG7 protein, SMG7 mutants were tested in complementation and tethering assays.

As the N- and C–terminal deletion mutants of SMG7 (S7–C and S7–N, respectively) failed to complement the NMD deficiency of SMG7-silenced plants (Figure 3a, the left panel of 3b and e), we suggest that both domains are essential for the NMD function of SMG7.

Figure 3.

Functional mapping of SMG7. (a) Schematic, non-proportional representation of the SMG7 constructs. (b, e and f) NMD competency tests of SMG7 mutants. SMG7-silenced leaves (VIGS:S7) were infiltrated with Gc–I NMD reporter plus P14 (–), or Gc–I+P14 constructs were co-infiltrated with full-length SMG7 (S7), N- or C–terminal deletion mutants (S7–C and S7–N) or with an SMG7 point mutant (S7mut). (c, d, g and h) Tethering assays of SMG7 constructs. Leaves were co-infiltrated with G–3′bB (3′ tethering target construct) plus P14, and with different SMG7 mutants. As negative controls, λN and the λN-free versions of the test constructs were used. RNA levels were quantified as described in Figures 1 and 2.

Next we wanted to study the role of the putative phosphoserine-binding, 14–3–3-like N–terminal domain of SMG7 in plant NMD. We created a point mutant (S7mut) by changing two conserved amino acids (K77E and R185E, see Figures S2 and S9) that participate in phosphoserine-binding in 14–3–3 proteins (Fukuhara et al., 2005), and then S7mut was studied in an NMD competency assay. We found that S7mut failed to complement the NMD deficiency of SMG7-silenced plants (Figure 3b right panel and f), indicating that these amino acids are essential for the NMD function of SMG7. As the amino acids corresponding to K77 and R185 play a key role in the phosphoserine binding of 14–3–3-like proteins, we hypothesize that the N–terminal region of SMG7 has phosphoserine binding activity that is essential for plant NMD.

Next, we compared the target transcript degradation activity of λN-tagged SMG7 constructs in tethering assays. The tethering of SMG7, S7mut and the C–terminal region of SMG7 (S7-λN, S7mut-λN and S7-C-λN, respectively) resulted in strongly reduced G–3′bB mRNA levels, whereas the tethering of the N–terminal domain (S7-N-λN) did not induce tethering target reduction (Figure 3c,d,g and h). These data (and our previous results) show that the C–terminal region is responsible for target decay, inducing the activity of the tethered SMG7, and suggest that the C–terminal region of SMG7 plays a role in the target degradation step of plant NMD (Benkovics et al., 2011). Presumably, SMG7 binding to the NMD complex is a late irreversible step of plant NMD, as its C–terminal region can trigger the rapid decay of the NMD complex-bound transcript.

UPF1 and UPF2 are not required for SMG7 tethering-induced mRNA reduction

If SMG7 binding is an irreversible step of plant NMD, SMG7 tethering should induce mRNA decay, even in the absence of UPF factors. Indeed, tethering of SMG7 strongly decreased the G–3′bB transcript level in UPF1- and UPF2-silenced test leaves, similarly to Phytoene desaturase (PDS)-silenced control leaves (Figures 4a,c and S10). These data suggest that the UPF factors are not required for SMG7 tethering-induced transcript decay. (The efficiency of VIGS is never 100%, thus we cannot formally exclude the alternative explanation that UPF1 and UPF2 are required for SMG7 tethering-induced target transcript degradation, and the remaining UPF1 and UPF2 proteins can fulfill this function in the silenced leaves. However, as NMD was impaired in both UPF1- and UPF2-silenced leaves, we think that this is an unlikely scenario. For further discussion see Figure S11.)

Figure 4.

UPF1 and SMG7 tethering assays in different silenced plants.(a and c) SMG7 tethering in PDS-, UPF1- and UPF2-silenced leaves (VIGS:P, U1 and U2). Silenced leaves were co-infiltrated with G–3′bB+P14 and with λN or SMG7 (S7) negative control constructs, or with a λN-tagged SMG7 (S7-λN) test construct. (b and d) UPF1-tethering in UPF2- and SMG7-silenced leaves (VIGS:U2 and S7). (e) UPF1 and SMG7 tethering in XRN4-silenced leaves (VIGS:X4). (f) The 3′ cleavage product of NMD target transcript is not detected in XRN4-silenced leaves. PDS- or UPF1-silenced control and XRN4-silenced test leaves (VIGS:P, U1 and X4) were co-infiltrated with 5′U2–G NMD reporter construct and P14.

In summary, the results of SMG7 functional studies support the model that the N–terminal domain of SMG7 is required for the early steps of NMD, whereas the C–terminal domain is involved in the late steps of NMD by inducing target transcript decay.

UPF2 and SMG7 are not essential for UPF1 tethering-induced mRNA reduction

Tethering of either UPF1 or SMG7 to the 3′–UTR of a tethering target transcript leads to target transcript degradation. As UPF1 is not required for SMG7 tethering-induced target decay, we postulated that UPF1 acted in tethering assays by recruiting SMG7, and that the decay of target mRNA was directly induced by SMG7. This model predicts that UPF1 tethering should not trigger target decay in SMG7-silenced leaves, but should function in UPF2-silenced plants. However, UPF1 tethering led to strongly reduced target levels in both SMG7- and UPF2-silenced leaves (Figures 4b,d and S11), suggesting that UPF1 tethering-induced transcript decay is independent of both UPF2 and SMG7.

XRN4 is not essential for plant NMD

Tethering of either UPF1 or SMG7 to a target mRNA initiates its degradation (Kertesz et al., 2006; Benkovics et al., 2011). However, it is not known if UPF1 and SMG7 activate the same degradation pathway or recruit different decay systems. In mammals, SMG7 tethering induces DCP- and XRN1-dependent target degradation (Unterholzner and Izaurralde, 2004). To test whether XRN4, the plant homolog of XRN1 (Kastenmayer and Green, 2000), is required for SMG7- and/or UPF1-induced decay, tethering assays were carried out in XRN4-silenced leaves. Importantly, in XRN4-silenced leaves, SMG7 tethering induced strong target transcript reduction, whereas the tethering of UPF1 did not alter the expression of the target mRNA (Figure 4e, compare lanes 3 and 6). These data suggest that UPF1 tethering triggers an XRN4-dependent degradation pathway, whereas SMG7 tethering activates an XRN4-independent degradation pathway.

To study whether the XRN4-dependent decay pathway is required for the degradation of NMD target transcripts, two different NMD reporter constructs (5′U2–G and PHAm) were infiltrated into PDS-silenced negative and UPF1-silenced positive control leaves, and into XRN4-silenced test leaves (different biological assays showed that XRN4 silencing was effective, see Figure S12 for details). 5′U2–G contains an NMD-inducing uORF in the 5′–UTR, whereas PHAm contains an NMD-inducing PTC (Figures 4f and S12c,d; Kertesz et al., 2006; Nyiko et al., 2009). As expected, the 5′U2–G and the PHAm NMD target transcripts accumulated to low levels in the PDS-silenced negative control, and to high levels in the UPF1-silenced positive control leaves. Of relevance, both NMD reporters accumulated to low levels in XRN4-silenced leaves (Figures 4f and S12c,d). These results indicate that NMD is functional in XRN4-silenced leaves, and suggest that these NMD reporter mRNAs are degraded independently of XRN4. Thus, we propose that XRN4 is not essential for plant NMD.

Previously we showed that SMG7 is a direct target of plant NMD (Kerenyi et al., 2008). If XRN4 is required for the degradation of endogenous NMD targets, SMG7 transcripts should be overexpressed in XRN4-deficient plants. However, SMG7 transcripts did not overaccumulate in an XRN4-deficient Arabidopsis line, whereas SMG7 mRNAs were upregulated in upf31 and upf51 NMD mutants (Figure S13). These data also suggest that XRN4 is not essential for plant NMD.

In animals, NMD target degradation starts with SMG6-mediated transcript cleavage close to the PTC (Huntzinger et al., 2008). The 3′ cleavage fragment can be detected in XRN1-deficient cells. It is not known how plant NMD target degradation is initiated. We have postulated that if degradation of PTC containing transcripts is initiated in plants with an endonucleolytic cleavage close to the PTC, the 3′ cleavage fragment should be detectable in XRN4-silenced leaves, because no other 5′ → 3′ cytoplasmic exonuclease was found in plants. However, we could not detect the 3′ cleavage product of either PHAm or 5′U2–G transcripts in XRN4-silenced leaves (Figures 4f and S12c,d). These results suggest that plant NMD is not initiated with an endonucleolytic cleavage close to the PTC.

SMG7 relocalizes UPF1 into P bodies

In mammals, overexpressed SMG7 relocalizes phospho-UPF1 (and the bound transcripts) into P bodies, thereby connecting the NMD complex formation and target degradation steps of NMD (Unterholzner and Izaurralde, 2004). To test whether SMG7 might play a similar role in plants, we studied the localization of SMG7 and the effect of SMG7 co-expression on UPF1 localization.

Nicotiana benthamiana leaves were infiltrated with SMG7-YFP or were co-infiltrated with SMG7-YFP and the DCP1-CPF P–body marker (Xu et al., 2006). We found that SMG7 co-localized with DCP1 (Figure 5d), suggesting that plant SMG7 accumulates in P bodies (and in the nucleus). It is unlikely that the DCP1-CFP construct modified the localization of SMG7, as SMG7 localization is similar when it is infiltrated alone (Figure 5b). Moreover, SMG7-YFP complemented SMG7-silenced leaves, showing that the fusion construct is biologically active (Figure S14).

Figure 5.

SMG7 and UPF1 localization studies.(a–c) Localization of DCP1, SMG7 and UPF1. (d1–4) DCP1 and SMG7 are co-localized. (e1–4) UPF1 is a cytoplasmic protein that partially co-localizes with the DCP1 P–body marker. (f1–4) SMG7 relocalizes UPF1 into P bodies. (g1–4) UPF1 and SMG7 interact in cytoplasmic foci. The merged images show the YFP signal in green, the CFP signal in red and the overlay of the two signals in yellow. Scale bar: 20 μm. Agroinfiltrated Nicotiana benthamiana leaves were analyzed at 3 dpi.

Next, we tested the effect of SMG7 co-expression on UPF1 localization. A YFP-UPF1 fusion construct was infiltrated alone (Figure 5c), or was co-infiltrated with DCP1-CFP (Figure 5e) or with DCP1-CFP plus a non-fluorescent SMG7 (SMG7-FLAG; Figure 5f). As reported, UPF1 was predominantly cytoplasmic, but also accumulated in discrete cytoplasmic foci (Figure 5c), presumably corresponding to P bodies (Kim et al., 2009). Indeed, we found that YFP-UPF1 shows partial co-localization with DCP1-CFP (Figure 5e). Importantly, SMG7-FLAG co-expression dramatically altered UPF1 localization, UPF1-YFP accumulated almost exclusively in P bodies (and occasionally in the nucleus) when it was co-expressed with DCP1-CFP+SMG7-FLAG (Figure 5f). These data suggest that overexpressed plant SMG7 relocalizes UPF1 into P–bodies.

Next, we wanted to test whether the plant SMG7 can interact with UPF1. We failed to co-immunoprecipitate SMG7 with UPF1; hence, we performed bimolecular fluorescence complementation studies to test whether UPF1 and SMG7 accumulate in the same complexes (Walter et al., 2004). Leaves were co-infiltrated with an SMG7 construct fused to the C–terminal domain of YFP (SMG7-YFP–C) and a UPF1 construct fused to the N–terminal domain of YFP (UPF1-YFP–N) (Figure 5g). As controls, SMG7-YFP–C and YFP–N or UPF1-YFP–N and YFP–C were co-infiltrated, respectively. As expected, no YFP expression was observed in the negative control leaves. In contrast, YFP expression could be detected in cytoplasmic foci – probably in P bodies – in SMG7-YFP-C+UPF1-YFP-N co-infiltrated leaves (Figure 5g). These observations indicate that in plants SMG7 and UPF1 are present in the same complex.


To understand the late steps of plant NMD, UPF1 and SMG7 NMD factors were functionally analyzed by using NMD competency tests, tethering assays and co-localization studies. Our data support a model that NMD target plant mRNAs are degraded by an SMG7-induced, XRN4-independent exonucleolytic pathway.

UPF1 and transcript degradation

UPF1 plays a key role in the early as well as in the late steps of NMD. Functional mapping studies suggest that the CH- and helicase-domains act during the early phase of NMD, whereas the terminal UPF1 regions are involved in the late steps of NMD.

The results showing that the highly conserved CH and helicase domains of UPF1 are essential for plant NMD, but are dispensable for UPF1 tethering-induced target decay (Figures 1 and 2), indicate that these domains play a role only in the early steps of NMD. It is possible that in plants, like in animals, the CH domain is required for UPF2 and/or eRF3 binding, whereas the helicase domain might remodel the NMD complex–target transcript RNP complex (Ivanov et al., 2008; Shigeoka et al., 2012).

The terminal regions of UPF1 play a critical but redundant role in the late steps of plant NMD. This conclusion is supported by the findings that: (i) constructs with either the N- or C-terminal region were fully active in the NMD competency test, whereas the UPF1 mutant lacking both terminal segments failed to complement UPF1-silenced leaves; (ii) the presence of either the N- or the C–terminal region was required (and sufficient) to trigger target transcript reduction in the tethering assay.

In mammals, the SMG1-mediated phosphorylation of both terminal UPF1 regions at specific sites (N–terminal T28 and C–terminal S1096) is essential to trigger NMD target decay (Okada-Katsuhata et al., 2012). Our data suggest that the terminal segments of plant UPF1 are also phosphorylated. As either the N- or the C–terminal region is required for UPF1 tethering-induced target reduction, we propose that either of the phosphorylated UPF1 terminal segments can trigger plant NMD target degradation.

We found that shortening the C–terminal region (in the absence of the redundant N–terminal segment) led to a gradual loss of NMD competency and tethering target degradation activity. If in plants, as in mammals, phosphorylation of a specific C–terminal S/TQ site was required for NMD, the difference between the deletion constructs should be pronounced: the construct containing the critical phosphorylation site would function, whereas the construct lacking that site would fail to complement. Thus, we hypothesize that the C–terminal region of plant UPF1 is phosphorylated at multiple, functionally redundant sites.

The role of SMG7

The NMD competency assays have shown that both the N- and C–terminal regions of SMG7 are required for NMD, and that the putative phosphoserine-binding amino acids of the 14–3–3-like N–terminal domain are also essential for plant NMD. Moreover, tethering assays revealed that only the C–terminal part of SMG7 is involved in tethering target degradation. We speculate that the 14–3–3-like domain of plant SMG7 binds the phosphorylated N- and C–terminal UPF1 regions, and then the C–terminal domain of SMG7 triggers degradation of the phosho-UPF1-bound transcript.

Our co-localization studies revealed that plant SMG7 is a P–body component that can relocalize UPF1 into P bodies (Figure 5). We propose that binding of plant SMG7 to the phospho-UPF1 component of the NMD complex facilitates the relocalization of the whole ribonucleoprotein complex (mRNP), including the PTC-containing transcript into the P body, where mRNAs are degraded. The result that SMG7 overexpression led to almost complete relocalization of UPF1 into P bodies apparently conflicts with our model, suggesting that SMG7 specifically binds phosho-UPF1. It is possible that in plants, unlike in mammals, UPF1 is predominantly phosphorylated, thus SMG7 overexpression could lead to the efficient relocalization of UPF1 into P bodies. Alternatively, plant SMG7 binds phospho-UPF1 with high affinity, but it can also bind non-phosphorylated UPF1 with low affinity. Indeed, in mammals it was shown that SMG6 and SMG5–SMG7 proteins bind phospho-UPF1 with strong preference, but these proteins also bind non-phosphorylated UPF1 with low affinity (Okada-Katsuhata et al., 2012).

Plant NMD target transcripts might be degraded by an SMG7-induced XRN–4-independent exonucleolytic pathway

XRN1 plays an important role in the elimination of NMD target yeast and mammalian transcripts. Full-length NMD target transcripts accumulate to high levels in XRN4-deleted yeast cells, whereas mammalian XRN1 eliminates the 3′ products of SMG6-mediated NMD target cleavage (Eberle et al., 2009; Hu et al., 2010).

In plants, XRN4 plays a less important role in the degradation of NMD target transcripts. Our findings that NMD reporter transcripts accumulate to wild-type levels in XRN4-silenced leaves, and that SMG7 endogenous NMD target transcripts accumulate to wild-type levels in xrn4 mutant plants (Figures 4f and S13) strongly suggest that XRN4 is not essential for plant NMD.

Our data indicating that the 3′ cleavage product of NMD target transcripts was not detected in XRN4-silenced leaves (Figures 4f and S12) suggest that degradation of plant NMD target mRNAs is not initiated with an endonucleolytic cleavage. Consistently, neither 3′ cleavage products nor 3′ degradation intermediaries of the endogenous NMD target SMG7 transcript were identified in high-throughput tilling and degradome arrays of xrn4 null mutants (Rymarquis et al., 2011). Moreover, no plant ortholog of the mammalian NMD endonuclease SMG6 was identified (Kerenyi et al., 2008). Although we could not exclude alternative explanations, the most straightforward interpretation of our results is that plant NMD targets are degraded by an XRN4-independent exonucleolytic pathway. As SMG7 triggers XRN4-independent target decay, whereas UPF1 triggers XRN4-dependent target decay, and because in our assays XRN4 was dispensable for NMD target degradation, we hypothesize that PTC-containing plant transcripts are degraded predominantly by an SMG7-induced, XRN4-independent exonucleolytic pathway.

The role of the UPF1-XRN4 pathway

Tethering of UPF1 triggers SMG7-independent, XRN4-dependent target degradation, whereas SMG7 tethering triggers UPF1- and XRN4-independent target decay. These results suggest that alternative NMD target degradation pathways function in plants. Alternative NMD degradation pathways also exist in yeast (in the 5′ → 3′ major and 3′ → 5′ minor exonucleolytic pathways) as well as in mammals (SMG6, SMG5–SMG7 and PNRC2-mediated pathways; Cao and Parker, 2003; Nicholson and Muhlemann, 2010). It is likely that in all eukaryotes multiple NMD degradation pathways evolved to avoid the trapping of NMD components and ribosomes on PTC-containing transcript when the main NMD degradation pathway operates with low efficacy.

XRN4 is not essential for plant NMD; therefore, we postulate that the UPF1-XRN4 pathway could act only as a minor NMD degradation pathway. It might operate as an SMG7-independent temporary backup mechanism to rescue stalled ribosomes and the components of the NMD complex if SMG7 is not available. Alternatively, the UPF1-XRN4 pathway degrades a subset of NMD targets or acts only under specific physiological conditions.

The UPF1-XRN4 decay pathway might also be used by mRNA binding proteins, similarly to the mammalian Staufen protein. Staufen destabilizes its target transcript by recruiting UPF1, which in turn activates the phospho-UPF1-PNRC2-decapping pathway to degrade Staufen-bound mRNAs (Cho et al., 2012).

Model of the late phase of plant NMD

Based on our results we propose a speculative model for the late steps of plant NMD (Figure 6). We suggest that plant UPF1 is present in the functional NMD complex as N- and/or C–terminally phosphorylated protein. SMG7 binds phospho-UPF1 and induces P–body relocalization of the mRNP. NMD target transcripts might be degraded in the P body by XRN4-independent exonucleases. In the absence of SMG7, phospho-UPF1 might activate the UPF1-XRN4 pathway to degrade the NMD complex-bound transcript.

Figure 6.

Model of late steps of plant nonsense-mediated mRNA decay (NMD). For details, see the main text.

Experimental procedures

Plasmid constructs

For agroinfiltrations, genes were cloned into Bin61S binary vector or into the derivatives of Bin61S. P14, GFP, PHAm, Gc–I, G–3′bB, UPF1, U1DN, TRV–P, TRV-P–U1, TRV-P–U2, TRV-P–S7, 5′U2–G, S7, S7–N, S7–C, S7-λN, S7-N-λN and S7-C-λN clones have been described previously (Kertesz et al., 2006; Kerenyi et al., 2008; Nyiko et al., 2009; Benkovics et al., 2011). Other constructs are described in detail in Appendix S1.

Agroinfiltration assays

Agroinfiltration and GFP detection were described by Kertesz et al. (2006). Wild-type or silenced N. benthamiana leaves were infiltrated with a mixture of cultures (OD600 of each culture was 0.4, or in the case of P14, 0.2). GFP fluorescence was detected by using a handheld long-wave ultraviolet lamp (UVP,

VIGS agroinfiltration experiment

To trigger VIGS, approximately 21–day-old N. benthamiana plants were co-infiltrated with a mixture of three Agrobacterium cultures. One expressed P14, the second expressed TRV RNA1 and the third expressed TRV RNA2 containing segments from N. benthamiana PDS or PDS+UPF1, PDS+UPF2, PDS+SMG7 or PDS+XRN4 (TRV–P, TRV-P–U1, TRV-P–U2, TRV-P–S7 and TRV-P–X4). When the upper leaves started to bleach (indicating that PDS silencing was efficient), leaves under the bleaching leaves were agroinfiltrated with a mixture of cultures (Kerenyi et al., 2008). To clone PDS-XRN4, a approximately 600–nt-long N. benthamiana XRN4 sequence was RT-PCR amplified (with primers XRN4VIGSF and XRN4VIGSR), and cloned into TRV-PDS vector.

RNA gel blot analysis

RNA methods and quantifications were as described by Kertesz et al. (2006). PCR fragments labeled by the random priming method were used for northern analyses. Phosphorimage measurements were used to quantify mRNA expression. RT-PCR was carried out with Qiagen OneStep RT-PCR Kit (Qiagen,

Western blot, immunoprecipitation and phosphatase assay

For immunoprecipitations (IPs), anti-HA affinity matrix (Roche, was used as described by Kerenyi et al. (2008). Rabbit polyclonal P14 antibody and monoclonal antibodies (anti-HA peroxidase, Roche; anti-Flag M2-peroxidase, Sigma-Aldrich, were used for chemiluminescence protein detections, according to the manufacturer's instructions (ECL, Amersham Biosciences, now GE Healthcare, For alkaline phosphatase (AP) assay, IP pellets were incubated with AP (FastAP, Fermentas, for 30 min on 37°C.

Microscopic analysis

The localization of fusion proteins was studied at 3 dpi (days post inoculation) with an Olympus Fluoview FV1000 confocal laser scanning microscope (Olympus, Digital images and artworks were processed with ImageJ and Olympus Fluoviewer FV1000 software.


We are grateful to C. Lacomme (University of Edinburgh), D. Baulcombe (University of Cambridge) and S.P. Dinesh-Kumar (Yale University) for TRV vectors, to F. Kerenyi and L. Szabadkai for sharing their unpublished results, and to Aman Y. Husbands for language corrections. This research was supported by grants from the OTKA (K60102 and C77086) and ICGEB (CRP/HUN09-01). T. Nyikó and A. Hangyáné Benkovics were graduate students of ELTE ‘Classical and Molecular Genetics’ and of Corvinus University of Budapest ‘Viticulture’ PhD programs, respectively. Z. Merai was supported by the EMBO short term and the Marie-Curie (PIEF-GA-2009-253075) fellowship program. The authors declare no conflict of interest.