The upf1-5 (SALK_112922) and upf3-1 (SALK_025175) mutants used in this study were previously described (Hori and Watanabe, 2005; Arciga-Reyes et al., 2006). We generated upf1-5 upf3-1 double mutants by crossing upf1-5 and upf3-1 mutants and confirmed their homozygosity by PCR-based genotyping using AtUPF1-1F and AtUPF1-2R primers for upf1-5 and AtUPF3-4F and AtUPF3-6R primers for upf3-1 (Hori and Watanabe, 2005; Riehs-Kearnan et al., 2012; Supplementary Table 4). For RNA-seq and qPCR analyses, 8-day-old wild-type (WT) Columbia-0 (Col-0), upf1-5, upf3-1, and upf1-5 upf3-1 double mutants grown at 23°C under standard LD conditions (16:8 h light: dark) were used. Total leaf numbers and days to flowering were used for the measurement of flowering time. Total leaf numbers were counted when the primary inflorescences reached about 5 cm. Flowering time data are presented as a box plot (Spitzer et al., 2014). In our box plots, the center lines show the medians and plus signs (+) show the mean value; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range (IQR) from the 25th and 75th percentiles, and outliers that exceeded the 1.5X IQR are represented by ovals. The number of plants measured is shown above each genotype in the box plot.

The upf1-5 and upf3-1 single mutants and upf1-5 upf3-1 double mutants were vernalized by incubating their imbibed seeds at 4°C for 4 weeks, prior to sowing. For a non-vernalized control, the upf mutants and WT Col-0 seeds were imbibed at 4°C for 4 days. The vernalization response was recorded as differences in flowering time between vernalized and non-vernalized mutants. The flowering time difference in response to vernalization was statistically assessed using Student's t-test.

Plants were grown for 8 days on MS plates at 23°C under standard LD conditions (16:8 h light: dark). About 50–100 seedlings were harvested at Zeitgeber Time 16 (ZT16) and pooled for RNA extraction, which was done using the Plant RNA Purification Reagent (Invitrogen). RNA sequencing was performed with a single biological replication for each sample. Because we did a single RNA-seq experiment, we verified the results via qPCR by using the same growth conditions. For RNA sequencing, library preparation was performed with an Illumina TruSeq Stranded Total RNA Sample Prep kit (Illumina), according to the manufacturer's protocols. Briefly, after removal of rRNA from the total RNA (~700 ng) using the rRNA removal kit, the RNA was cleaned using RNA purification beads. After the rRNA was removed, the remaining RNA was fragmented using fragment mix (EPH) at 94°C for 6 min. The fragmented RNA was primed with random hexamers and transcribed to first-strand cDNA using reverse transcriptase and random primers at 25°C for 10 min, 42°C for 15 min, and then 70°C for 15 min. Then a replacement strand was synthesized by incorporating dUTP in place of dTTP to generate double-stranded cDNA using polymerase at 16°C for 1 h. After cleanup of cDNA using sample purification beads, a single “A” nucleotide was added to the 3′ ends of the blunt fragments using A-tailing mix reagent by incubating at 37°C for 30 min and then at 70°C for 5 min. Indexing adapters were ligated to the ends of the DNA fragments using ligation mix two reagent at 30°C for 10 min. After washing with sample purification beads twice, PCR was performed to enrich DNA fragments having adapter molecules on both ends. Thermocycler conditions were as follows: 95°C for 3 min, 8 cycles of: 98°C for 20 s, 60°C for 15 s, and 72°C for 30 min, with a final extension at 72°C for 5 min. Finally, quality and band size of libraries were assessed using Agilent 2100 Bioanalyzer (Agilent). Libraries were quantified by qPCR using CFX96 Real Time System (Bio-Rad). After normalization, sequencing of the prepared library was conducted as paired-end reads on an Illumina HiSeq2000 sequencer. The resulting raw RNA-seq data are available at NCBI Gene Expression Omnibus (GEO) database under the accession number GSE87851.

The raw sequence reads were processed for removal of adapter sequences, followed by qualitative analysis of raw reads using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). The high-quality reads were then mapped against the Arabidopsis thaliana reference genome TAIR10 (downloaded from http://arabidopsis.org) using TopHat 2.0.6 (https://ccb.jhu.edu/software/tophat/index.shtml) and Bowtie 2 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) using default parameters.

The resulting mapped reads were then assembled and normalized transcript abundances were determined using Cufflinks version 2.2.1 (Trapnell et al., 2012; http://cole-trapnell-lab.github.io/cufflinks) with default settings. Cuffdiff version 2.2.1 was used for the identification of DEGs between WT plants and upf mutants with default parameters, including a false discovery rate of 5% (Trapnell et al., 2012). The output of Cuffdiff was further analyzed using R package “CummeRbund” (Goff et al., 2012). Genes with fold-change (FC) values of two or more and FPKM (fragments per kilobase of transcript per million mapped reads) values of one or more were considered as differentially expressed genes.

To get insight into the flowering time changes seen in upf mutants, we studied expression of 306 flowering time regulator genes acquired from the Flowering Interactive Database (FLOR-ID; Bouche et al., 2016). For the analysis of genes that regulate FLC, we collected a total of 58 genes that potentially regulate FLC expression (Xu et al., 2013; Bouche et al., 2016; Supplementary Table 2) and analyzed their expression patterns. For representation of differential gene expression in upf mutants, three heatmaps were generated: one for overall DEGs found in upf mutants, a second for expression of all 306 flowering time regulators, and a third map for 58 genes that were identified as FLC regulators.

Quantitative real-time PCR (qPCR) was employed to validate the transcriptome data of the floral integrators and their regulators. Total RNA was extracted from 8-day-old Arabidopsis seedlings sampled at ZT16, using Plant RNA purification reagent (Invitrogen). The extracted RNA (~2 μg) was reverse transcribed into cDNA using the Transcriptor First Strand cDNA Synthesis kit (Roche). Expression analysis was performed using SYBR Green I Master mix (Roche) in a LightCycler 480 (Roche). The data were normalized against two stable reference genes, PP2AA3 (AT1G13320) and a SAND family gene (AT2G28390) (Hong et al., 2010). All the qPCR data are presented as the mean of two biological replicates with three technical replicates each and the error bars indicate the standard deviation. Statistical significance of differences of gene expression levels between the samples was assessed by using Student's t-test. P < 0.05 were considered as significant. Information on the primers that were used for qPCR in this study is shown in Supplementary Table 4.

For detection of potential NMD targets, we used the CummeRbund package to individually analyze each flowering time gene to identify genes that show accumulation of aberrant transcripts in NMD-deficient mutants. The shortlisted potential target transcripts were then detected using RT-PCR using PP2AA3 as an internal control (Hong et al., 2010). Sequences of normal and aberrant transcripts from RNA-seq data were extracted and analyzed for the presence of PTCs between the normal and aberrant transcripts. Putative cDNA sequences of both normal and aberrant transcripts were translated in silico to determine the effect of the PTCs on the translated protein.

To investigate the effect of the loss of NMD function on flowering time, we measured flowering time of upf1-5 and upf3-1 single mutants and the upf1-5 upf3-1 double mutants at 23°C under standard LD conditions. Both upf1-5 and upf3-1 single mutants flowered later than wild-type plants (Figures 1A,B), such that upf1-5 and upf3-1 mutants flowered with 22 ± 1.6 and 23 ± 2.3 leaves, respectively, consistent with a previous study that showed upf1-5 and upf3-1 mutants flowered late under both LD and SD conditions and had narrow, jagged rosette leaves (Arciga-Reyes et al., 2006). A previous study also reported that upf3-1 mutants showed a more severe phenotype and grew more slowly than upf1-5 mutants (Shi et al., 2012). We also generated upf1-5 upf3-1 double mutants and measured their flowering time to investigate whether UPF1 and UPF3 act redundantly in flowering time. The upf1-5 upf3-1 double mutants exhibited a severe bushy phenotype with no well-defined leaves in the vegetative phase (Figure 1C). Because multiple inflorescences emerged almost simultaneously from the axillary meristems of upf1-5 upf3-1 double mutants after the transition to the reproductive phase (Figure 1C), we were not able to count the number of leaves of the primary inflorescence of upf1-5 upf3-1 mutants; instead, we measured days to flowering. By this measure, the upf1-5 and upf3-1 mutants flowered later than wild-type plants under LD conditions. The upf1-5 and upf3-1 mutants flowered 33.5 ± 3.8 and 35.25 ± 3.6 days after germination (Figure 1D), respectively, whereas wild-type plants flowered 23.4 ± 1.7 days after germination. The upf1-5 upf3-1 double mutants flowered 40.6 ± 2.4 days after germination, indicating that the double mutation had an additive effect in delaying flowering time. This flowering time analysis indicated that the upf1-5 and upf3-1 mutants are late flowering and their combined mutations have an additive effect on flowering.

To gain insights on the transcript levels of flowering time genes in upf mutants, we performed RNA sequencing (RNA-seq) using upf1-5, upf3-1, and upf1-5 upf3-1 mutants with wild-type plants as a control (see Supplementary Table 1 for expression levels of all genes). Heat map analysis was performed to identify differentially expressed genes between upf mutants and wild-type plants (Figure 2A). Based on their expression patterns, we divided the DEGs into five major groups (Figure 2A). Group I included DEGs whose expression levels increased in both upf1-5 and upf3-1 mutants, but further increased in upf1-5 upf3-1 mutants. Group II included DEGs whose expression levels increased in upf3-1 mutants and upf1-5 upf3-1 mutants. Group III included DEGs whose expression levels increased only in upf1-5 upf3-1 mutants. Group IV included DEGs whose expression levels decreased in upf3-1 mutants and further decreased in upf1-5 upf3-1 mutants. The genes whose expression level changes were difficult to classify were placed in group V (Figure 2A). This heatmap analysis suggested that the expression patterns of DEGs in upf3-1 mutants were more similar to upf1-5 upf3-1 mutants than upf1-5 mutants. Furthermore, scatter plot distribution of gene expression also revealed the similarities in DEGs between upf3-1 single and upf1-5 upf3-1 double mutants (Supplementary Figure 1).

We then selected individual DEGs after applying specific filters (genes with at least 2-fold difference in expression levels and FPKM value of 1 or more) from our RNA-seq data. The results showed that 2,254 genes were differentially expressed in upf1-5 mutants (1,707 were up-regulated and 547 were down-regulated), 4,404 genes were differentially expressed in upf3-1 mutants (2,947 were up-regulated and 1,457 were down-regulated; Figures 2B,C). In upf1-5 upf3-1 mutants, 4,541 genes were differentially expressed (3,103 up-regulated and 1,438 down-regulated). A total of 964 genes were commonly up-regulated among upf single mutants and upf1-5 upf3-1 double mutants. upf1-5 and upf1-5 upf3-1 double mutants shared 1,270 up-regulated genes, whereas upf3-1 and upf1-5 upf3-1 mutants shared 2,049 up-regulated genes (Figure 2B). In contrast, the upf1-5 and upf1-5 upf3-1 double mutants shared 212 down-regulated genes, whereas upf3-1 and upf1-5 upf3-1 mutants shared 658 down-regulated genes (Figure 2C). These indicated that in terms of number of DEGs, the upf3-1 mutants were closer to the upf1-5 upf3-1 mutants, compared with the upf1-5 mutants. Considering that the upf1-5 upf3-1 mutants showed an additive delayed flowering phenotype, these overall similarities in the number of DEGs between upf3-1 mutants and upf1-5 upf3-1 double mutants suggested that UPF3 exerted a stronger effect than UPF1 in the regulation of UPF-mediated flowering time.

To examine the biological functions of the DEGs, we performed Gene Ontology (GO) analysis using the ClueGO Cytoscape plug-in (Bindea et al., 2009). The “GO term fusion” option of ClueGO was used to reduce redundancy in GO terms. As expected, a variety of biological functions were affected by the NMD deficiency in upf mutants (Supplementary Figure 2). Most of the DEGs showed high enrichment for cellular metabolism of different biomolecules, metabolic process, gene expression regulation, transportation, protein localization, and response to different stimuli (Supplementary Figure 2), consistent with a previous finding (Rayson et al., 2012).

To identify potential candidate genes responsible for the delayed flowering of upf mutants, we looked for DEGs related to the regulation of flowering time. Among such genes (Supplementary Figure 3), we found that the mRNA levels of FLOWERING LOCUS C (FLC), a floral repressor that is important for the vernalization response (Michaels and Amasino, 1999), increased in upf1-5, upf3-1, and upf1-5 upf3-1 mutants (Figure 3A, Supplementary Table 2). Our RNA-seq data showed that FLC mRNA levels increased by 2.92-fold in upf1-5 mutants and by 2.45-fold in upf3-1 mutants, whereas FLC mRNA levels increased by more than 5-fold in upf1-5 upf3-1 double mutants, compared to wild-type plants, indicating that UPF1 and UPF3 have an additive effect in regulating FLC expression (Figure 3A, Supplementary Table 2).

As our global scale RNA-seq analysis revealed the elevated levels of FLC in the upf mutants that we tested (Figure 3A, Supplementary Table 2), we performed qPCR analyses to confirm the finding. FLC mRNA expression levels increased in upf1-5 and upf3-1 mutants (5.8- and 4-fold, respectively; Figure 3B). This suggested a role of FLC in delayed flowering of NMD-deficient mutants. Consistent with the further delayed flowering seen in upf1-5 upf3-1 mutants (Figure 1), FLC mRNA levels further increased in upf1-5 upf3-1 mutants (8.1-fold; Figure 3B).

Consistent with the increased levels of FLC, we also found that the mRNA levels of FLOWERING LOCUS T (FT; Kardailsky et al., 1999) and SOC1 (Lee et al., 2000), which were reported as major targets of FLC (Helliwell et al., 2006), decreased (Supplementary Table 2). Our RNA-seq analysis showed down-regulation of FT in upf1-5 and upf3-1 mutants (1.1 and 2.63-fold, respectively). In the upf1-5 upf3-1 double mutants, the mRNA levels of FT decreased by 5.1-fold, suggesting that the upf1 and upf3 mutations showed an additive effect (Figure 3C, Supplementary Table 2). Similar to FT, the mRNA levels of SOC1, another important floral integrator that is repressed directly by FLC (Hepworth et al., 2002), were reduced in upf1-5 and upf3-1 single mutants (1.7- and 4.5-fold, respectively), whereas the upf1-5 upf3-1 double mutants showed a 6.3-fold decrease (Figure 3E, Supplementary Table 2). Our qPCR results were consistent with our RNA-seq data that showed decreased levels of FT (Figure 3D) and SOC1 mRNAs (Figure 3F) in upf mutants. We also analyzed SHORT VEGETATIVE PHASE (SVP) expression levels, as FLC interacts with SVP to synergistically repress flowering (Li et al., 2008). However, we did not find apparent changes in SVP mRNA levels in upf mutants (Supplementary Figure 4). These results showed that FLC mRNA levels increased, whereas FT and SOC1 mRNA levels decreased in upf mutants, raising the possibility that FLC up-regulation represses FT and SOC1 expression in upf mutants, which eventually causes late flowering in upf mutants.

After confirming the elevated levels of FLC in upf mutants, we next investigated whether FLC mediates the late-flowering phenotype of upf mutants. For this purpose, we analyzed whether vernalization treatment accelerated flowering of upf mutants, as FLC-mediated repression of flowering can be recovered by prolonged cold treatment (Michaels and Amasino, 1999; Sheldon et al., 1999). We transferred upf mutant seeds to 4°C for 4 weeks for vernalization treatment. Flowering time measurement indicated that vernalization accelerated the flowering of upf1-5 and upf3-1 mutants. Vernalized upf1-5 mutants flowered with 15.0 ± 1.5 leaves, whereas non-vernalized upf1-5 mutants flowered with 22.7 ± 2.2 leaves (Figures 4A,C). Similarly, vernalized upf3-1 mutants flowered with 16.3 ± 1.0 leaves, whereas non-vernalized upf3-1 mutants flowered with 23.3 ± 2.1 leaves (Figures 4B,C). The non-vernalized Col-0 plants flowered with 14.3 ± 1.3 leaves and the vernalized Col-0 plants flowered with 12.4 ± 1.2 leaves (Figure 4C). These observations confirmed that wild-type Col-0 plants, which contain a non-functional FRI allele (Lee and Amasino, 1995), showed weak responsiveness to vernalization treatment. As leaf numbers could not be counted for upf1-5 upf3-1 mutants, we also measured days to flowering of vernalized upf1-5 upf3-1 mutants under LD conditions to determine the vernalized response of upf1-5 upf3-1 mutants. The vernalized wild-type plants flowered 22.4 ± 2.3 days after germination, compared to the non-vernalized wild-type plants, which flowered 23.7 ± 1.6 days after germination. Vernalized upf1-5 mutants flowered 24.7 ± 3.5 days after germination, whereas non-vernalized upf1-5 mutants flowered 33.4 ± 3.8 days after germination (Figure 4E). Similarly, vernalized upf3-1 mutants flowered 28.1 ± 3.4 days after germination, whereas non-vernalized upf3-1 mutants flowered 35.6 ± 3.6 days after germination. In addition, vernalized upf1-5 upf3-1 mutants flowered 33 ± 4.3 days after germination, whereas non-vernalized upf1-5 upf3-1 mutants flowered 40.7 ± 2.4 days after germination (Figures 4D,E). These analyses indicated that vernalized upf1-5, upf3-1, and upf1-5 upf3-1 mutants flowered ~9, 7.5, and 7.7 days earlier than non-vernalized mutants, respectively. These observations suggested that FLC repression by vernalization likely accelerated flowering in upf mutants, suggesting a key role of FLC in the late flowering of upf mutants.

Although it seemed that the late-flowering phenotype of upf mutants was largely mediated by FLC, we next looked for an aberrant FLC transcript that contains a PTC. In addition to AT5G10140.1, the predominant transcript that produces functional FLC protein, we found four alternatively spliced isoforms of FLC in our RNA-seq (Supplementary Figure 5A). Among them, AT5G10140.2, AT5G10140.3, and AT5G10140.4 were already reported. We found an unreported alternatively spliced isoform and named it AT5G10140.5. AT5G10140.5 had a similar structure to the fourth splice variant of FLC (i.e., AT5G10140.4), except for its untranslated regions (UTRs), as the novel AT5G10140.5 transcript had longer 3′- and 5′-UTRs. We then measured the transcript levels of each alternatively spliced isoform. The abundance of these four isoforms of FLC was not apparently altered in upf mutants (Supplementary Figure 5B). Moreover, we could not find a PTC from any of the alternative spliced isoforms. These results suggested that FLC is not a direct target of NMD in the regulation of flowering time in upf mutants.

As FLC may not be the direct target of UPF-mediated regulation of flowering time, we analyzed expression of genes that act upstream of FLC. For this purpose, we measured the transcript levels of 58 genes that regulate flowering time (Bouche et al., 2016) from our RNA-seq data (Figure 5A), to understand the positive regulation of FLC in the upf mutants. Detailed expression data for of all FLC regulators can be found in Supplementary Table 3. Among 58 genes that potentially regulate FLC expression, we found a few genes whose expression patterns in the mutants were similar to that of FLC (Figures 5B–I). They were: (1) MADS AFFECTING FLOWERING 3 (MAF3), a homolog of FLC that inhibits flowering by up-regulating FLC expression and repressing FT and SOC1 expression (Gu et al., 2013; Suter et al., 2014); (2) TBP-ASSOCIATED FACTOR 14 (TAF14), which is also known as HOMOLOG OF YEAST YAF9 B (YAF9B) and promotes FLC expression through chromatin modification together with AtYAF9A (Zacharaki et al., 2012; Bieluszewski et al., 2015); and (3) NF-YA2 and NF-YA5, which encode factors that may have dual functions (as a floral promoter and a floral repressor) in the regulation of flowering time (Xu et al., 2013; Hou et al., 2014).

RNA-seq analysis showed that mRNA levels of MAF3 were high in upf1-5 (2.9-fold that of wild type) and upf3-1 (2.8-fold) mutants. The mRNA level of MAF3 increased by 3.3-fold in upf1-5 upf3-1 double mutants (Figure 5B, Supplementary Table 3). Similarly, qPCR analysis showed increased transcript levels of MAF3 in upf1-5 (3.3-fold) and upf3-1 mutants (~4-fold; Figure 5C). The upf1-5 upf3-1 double mutants showed a 10.2-fold increase, suggesting that both upf mutations had an additive effect on MAF3 expression (Figure 5C). The mRNA levels of TAF14 also increased in upf1-5 (3.2-fold) and upf3-1 (2.0-fold) mutants, as well as in upf1-5 upf3-1 (2.1-fold) double mutants, (Figure 5D, Supplementary Table 3). The increased mRNA levels of TAF14 revealed by qPCR (Figure 5E) were consistent with RNA-seq data. The RNA-seq data showed that the mRNA level of NF-YA2 increased by 1.5- and 2.1-fold in upf1-5 and upf3-1 mutants, respectively, and was 3.3-fold higher in upf1-5 upf3-1 double mutants (Figure 5F, Supplementary Table 3). The qPCR results agreed with the RNA-seq data of elevated NF-YA2 mRNA levels (Figure 5G). The upf1-5 and upf3-1 mutants showed high levels of NF-YA5 mRNA (3.5- and 5.6-fold increase, respectively), and upf1-5 upf3-1 mutants showed an 8.7-fold increase in NF-YA5 mRNA (Figure 5H, Supplementary Table 3), suggesting that upf1 and upf3 mutations contribute additively to NF-YA5 mRNA levels in upf1-5 upf3-1 double mutants. The qPCR analysis of NF-YA2 also showed a similar expression pattern (Figure 5I).

We also analyzed the expression of genes that regulate flowering time in an FLC-independent manner to identify possible regulators of flowering time in NMD-deficient mutants. From our RNA-seq data, we found two floral repressors that showed consistent expression changes in upf mutants: B-BOX DOMAIN PROTEIN 19 (BBX19), which encodes a floral repressor that interacts with CO, thereby preventing CO from inducing FT expression (Wang et al., 2014), and ATC, which encodes a floral repressor that interacts with FD to affect the expression levels of the floral meristem identity gene AP1 (Mimida et al., 2001; Huang et al., 2012; Figures 5J,L). qPCR analysis revealed that BBX19 expression was significantly up-regulated with a fold change of ~2 in upf1-5 and upf3-1 mutants and ~3-fold in upf1-5 upf3-1 double mutants (Figure 5K). The transcript levels of ATC were similar between wild-type plants and upf1-5 mutants (Figure 5M); however, the ATC expression levels were elevated over 2-fold in upf3-1 and upf1-5 upf3-1 double mutants (Figure 5M). Thus, the elevated levels of BBX19 and ATC in upf mutants suggested an FLC-independent contribution to the delayed flowering of upf mutants.

To identify the direct targets of NMD in the regulation of flowering time, we analyzed the isoforms of all flowering time regulators. Among these genes, we found that MAF3 and BBX19 produced aberrantly spliced transcripts (Figures 6A,B). As these transcripts were not listed in TAIR, we then named them AT5G65060.3 (from MAF3) and AT4G38960.4 (from BBX19). The AT5G65060.3 transcript differed from AT5G65060.1, the predominant MAF3 transcript, by the presence of longer 3′- and 5′-UTRs and the absence of 38 nucleotides within the 4th exon. The 38-nucleotide deletion caused a predicted out-of-frame splicing event, which created a PTC in the same exon, leading to formation of a truncated MAF3 protein (Figure 6A). The full-length alignment of the transcripts and the proteins can be found in Supplementary Figures 6, 7. The AT4G38960.4 transcript generated from BBX19 had a long 5′-UTR, skipped the first two BBX19 exons, and retained part of the third intron, like the AT4G38960.2 transcript, a splice variant of BBX19. Importantly, seven nucleotides were absent in the 6th exon of the AT4G38960.4 transcript, compared to AT4G38960.3, the normal functional transcript of BBX19. The 7-nucleotide deletion likely created a PTC in the 5th exon to generate truncated BBX19 proteins (Figure 6B, Supplementary Figure 9). Considering that the AT5G65060.3 and AT4G38960.4 transcripts had a PTC and long UTRs (Figures 6A,B, Supplementary Figures 6, 8), a characteristic of aberrant transcripts that are destined for NMD under normal circumstances, these AT5G65060.3 and AT4G38960.4 transcripts could be potential targets of UPF-mediated regulation of flowering time.

RNA-seq analysis showed that FPKM levels of AT5G65060.3 and AT4G38960.4 transcripts increased in upf mutants. The FPKM levels of the AT5G65060.3 transcript increased by 20.3, 23.9, and 25.9-fold in upf1-5, upf3-1, and upf1-5 upf3-1 mutants, respectively (Figure 6C). The FPKM levels of the AT4G38950.4 transcript increased by 6.5, 2.6, and 5.8-fold in upf1-5, upf3-1, and upf1-5 upf3-1 mutants, respectively (Figure 6D). We then carried out conventional RT-PCR to determine whether we could detect the aberrant transcripts identified in our RNA-seq analysis. We successfully detected amplicons using primers to specifically amplify the aberrant AT5G65060.3 and AT4G38960.4 transcripts (Figure 6E). Furthermore, the levels of AT5G65060.3 and AT4G38960.4 apparently increased in upf mutants, compared to wild-type plants, suggesting that AT5G65060.3 and AT4G38960.4 transcripts were likely substrates of NMD. These analyses suggested that MAF3 and BBX19 are potential targets of NMD in the regulation of flowering time in upf mutants.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.