Viral Perturbation of Alternative Splicing of a Host Transcript Benefits Infection

Sugarcane mosaic virus infection alters alternative splicing patterns of maize phytoene synthase1 transcripts to promote virus infection.

Abstract

Pathogens disturb alternative splicing patterns of infected eukaryotic hosts. However, in plants it is unknown if this is incidental to infection or represents a pathogen-induced remodeling of host gene expression needed to support infection. Here, we compared changes in transcription and protein accumulation with changes in transcript splicing patterns in maize (Zea mays) infected with the globally important pathogen sugarcane mosaic virus (SCMV). Our results suggested that changes in alternative splicing play a major role in determining virus-induced proteomic changes. Focusing on maize phytoene synthase1 (ZmPSY1), which encodes the key regulatory enzyme in carotenoid biosynthesis, we found that although SCMV infection decreases total ZmPSY1 transcript accumulation, the proportion of splice variant T001 increases by later infection stages so that ZmPSY1 protein levels are maintained. We determined that ZmPSY1 has two leaf-specific transcripts, T001 and T003, distinguished by differences between the respective 3′-untranslated regions (UTRs). The shorter 3′-UTR of T001 makes it the more efficient mRNA. Nonsense ZmPSY1 mutants or virus-induced silencing of ZmPSY1 expression suppressed SCMV accumulation, attenuated symptoms, and decreased chloroplast damage. Thus, ZmPSY1 acts as a proviral host factor that is required for virus accumulation and pathogenesis. Taken together, our findings reveal that SCMV infection-modulated alternative splicing ensures that ZmPSY1 synthesis is sustained during infection, which supports efficient virus infection.

Successful viral infection of a plant depends on complex molecular interactions between host and pathogen. Viruses have small genomes with limited protein-coding capacities. To compensate for this, viruses modulate plant gene expression and co-opt host factors to support their replication, cell-to-cell movement, and systemic movement (Hull, 2014; Wang, 2015). Modulation of host gene expression can lead to development of disease symptoms in plants (Mandadi and Scholthof, 2013; Hull, 2014; Wang, 2015). An important form of post-transcriptional gene regulation in eukaryotes is RNA splicing (Syed et al., 2012; Reddy et al., 2013), but this has been relatively understudied in the context of host–virus interactions (Boudreault et al., 2019).

Alternative splicing of primary host transcripts allows generation of multiple transcripts from a single eukaryotic gene, which enhances transcript and protein diversity (Marquez et al., 2015; Laloum et al., 2018). The transcripts of 33% to 61% of multiexonic plant genes undergo alternative splicing, often in response to biotic and abiotic stresses (Marquez et al., 2012; Mandadi and Scholthof, 2015; Jiang et al., 2017; Laloum et al., 2018). In the case of the dominant N resistance gene, alternative splicing is important in generating distinct N protein variants needed for tobacco (Nicotiana tabacum) to resist tobacco mosaic virus (Dinesh-Kumar and Baker, 2000). Additionally, mis-splicing of primary transcripts for eukaryotic translation initiation factor 4E in Brassica rapa contributes to recessive resistance to turnip mosaic virus (TuMV; Nellist et al., 2014). In contrast, however, the role of alternative splicing of host transcripts, especially when variation in splicing is triggered by a pathogen in a susceptible host, is less clear. But changes in alternative splicing of host transcripts have been seen in plants infected by bacteria, fungi, oomycetes, viruses, and viroids, suggesting that modification of splicing activity represents an important battleground in the arms race between pathogens and their hosts (Dinesh-Kumar and Baker, 2000; Nellist et al., 2014; Mandadi and Scholthof, 2015; Liu et al., 2016; Huang et al., 2017; Jiang et al., 2018). Determining whether the observed changes improve adaptation to stress or promote viral spread/infection remains an intriguing topic (Mach, 2015). So far, however, the importance of alternatively spliced host transcripts in determining the outcomes of virus–host interactions has not been clear.

Many studies have dissected host responses to virus infection at transcriptional and translational levels (for review, see Zanardo et al., 2019; Alexander and Cilia, 2016). However, few comparative studies have closely examined differences between the presymptomatic and steady-symptom stages of systemic infection. In this context we define a steady-symptom stage as the point at which systemic symptoms have become clearly visible. This dearth of studies is surprising, because plant organs at different developmental stages undergo obvious dynamic changes in gene expression, especially for plants under stress (Chen et al., 2017a; Hickman et al., 2017; Yi et al., 2019; You et al., 2019). Thus, it is reasonable to assume that infection with a virus will modulate plant gene expression differentially. For example, at the presymptomatic stage, plants may attempt to combat viral replication and multiplication, whereas host responses at the steady-symptom stage could be more general; for example, plants showing mosaic symptoms might exhibit downregulation of chloroplast-related genes/proteins (Zhao et al., 2016). Hence, dissecting the molecular network of pathogen–host interplay throughout the sequence of systemic infection will enhance our basic knowledge of this overall process and aid the development of new strategies to block and/or control pathogen infections.

Recently, we established a maize (Zea mays)–sugarcane mosaic virus (SCMV) experimental pathosystem to identify and mechanistically characterize plant–virus interactions at the molecular level (Chen et al., 2017a, 2017b; Yuan et al., 2019). Maize is one of the world’s most important cereal crops. SCMV is a maize pathogen of global importance that causes severe dwarf mosaic disease, and in China this disease can cause losses of up to 50% in yield (Chiu, 1988; Jiang and Zhou, 2002; Fan et al., 2003; Zhu et al., 2014). When SCMV occurs in mixed infections with maize chlorotic mottle virus, this causes maize lethal necrosis disease, which results in 100% losses (Redinbaugh and Stewart, 2018). SCMV is a monocot-infecting Potyvirus that causes mosaic symptoms, usually apparent as yellow streaks, on maize leaves. The replication site of SCMV is unknown, but because it is a positive-sense RNA virus it is likely to recruit host membranes to form replication organelles (Nagy, 2020). Because other potyviruses, including TuMV, maize dwarf mosaic virus, tobacco etch virus, and potato virus Y, are reported to recruit chloroplast outer membranes for replication sites, we suspect that SCMV also does this (Mayhew and Ford, 1974; Gadh and Hari, 1986; Gunasinghe and Berger, 1991; Wei et al., 2010).

In this study, by integrative analysis of RNA-sequencing (RNA-seq) and iTRAQ-based proteomics datasets, we found that SCMV infection extensively modifies maize gene expression from the presymptomatic stage. Subsequently, we observed clear post-transcriptional modulation of RNA metabolism after SCMV infection. Focusing on a SCMV infection-altered spliced gene required for carotenoid metabolism, maize phytoene synthase1 (ZmPSY1), we discovered that SCMV infection alters the accumulation dynamics of two leaf-specific ZmPSY1 splice variants (T001 and T003), which prevents ZmPSY1 protein levels from decreasing. Furthermore, we determined that ZmPSY1 is a proviral host factor required for maximal SCMV infection and for symptom induction. Taken together, we propose that pathogen-stimulated differential alternative splicing is not simply an incidental effect of infection, but instead provides a mechanism by which a pathogen maintains efficient expression of an essential host factor.

RESULTS

SCMV Infection Causes Dynamic Changes in the Transcriptome and Proteome of Maize Seedlings

To understand the dynamic modulation of maize gene expression by SCMV, we first characterized the progress of SCMV systemic infection in maize seedlings. We mechanically inoculated 8-d-old B73 seedlings with the SCMV-Beijing isolate (Fan et al., 2003) and followed systemic symptom development in the noninoculated leaves. No symptoms were observed until 5-d post inoculation (dpi) when the majority of SCMV-inoculated seedlings (86 of 95) started to show mosaic symptoms on the first systemically infected leaves. By 9 dpi, all of the first and the second systemically infected leaves exhibited mosaic symptoms (Fig. 1A). Compared with the equivalent leaves from mock-inoculated plants, chlorophyll and carotenoid contents significantly decreased in leaves of SCMV-infected plants by 7 dpi and these decreases continued at least until 9 dpi (Fig. 1B). SCMV RNA was detectable at 3 dpi, and by 9 dpi, coat protein (CP) was also detectable (Fig. 1C). Based on these differences in the dynamics of symptom induction, viral RNA, and CP accumulation, we defined, for the purposes of this study, 3 dpi as the pre-symptomatic stage of SCMV systemic infection and 9 dpi as the steady-symptom stage of systemic infection.

Figure 1.

Development of systemic symptoms in maize seedlings infected with SCMV. A, Infection progress and mosaic/streaking symptoms on the first and second systemically infected leaves of SCMV-inoculated maize seedlings and equivalent leaves of mock-inoculated plants (Mock) at 3, 5, 7, and 9 dpi. The white dashed boxes indicated regions showing mosaic/streaking symptoms as they expanded between 5 and 9 dpi. Scale bars = 0.5 cm. B, Levels of photosynthetic pigments in the first systemically infected leaves over the course of infection. Data were presented as means ± se (n = 3 to 5). Asterisks indicate the significant differences between mock-inoculated and SCMV-infected plants as evaluated by two-tailed Student’s t tests. C, Accumulation of SCMV RNA and CP levels in the first systemically infected leaves of mock-inoculated and SCMV-infected maize seedlings at 3 and 9 dpi. RT-qPCR data (top) are presented as means ± se (n = 4) and relative to ZmUbi (ubiquitin). Asterisks indicate significant differences between 3 and 9 dpi in SCMV-infected plants determined by the two-tailed Student’s t test. Immunoblots (bottom) show SCMV CP (arrowhead) accumulation at 3 and 9 dpi. Actin was used as a loading control.

To identify transcriptomic and proteomic changes that occurred by the presymptomatic (3 dpi) and steady-symptom (9 dpi) stages, we performed RNA-seq and iTRAQ-based proteomic analyses of B73 seedlings responding to SCMV infection (Fig. 2A; Supplemental Figs. S1 and S2). We adopted the generalized linear model approach (see “Materials and Methods” for details; Hickman et al., 2017) to identify differences in gene expression caused by SCMV infection. SCMV infection led to the differential accumulation of ∼88% differentially expressed genes (DEGs) starting from 3 dpi (DEGs 1+3/total DEGs = 8,893/10,134; Fig. 2B; Supplemental Table S1). Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of the DEGs indicated enrichment of pathways for protein processing in the endoplasmic reticulum (ER), for fatty acid biosynthesis, and for RNA splicing (Supplemental Figs. S3 and S4). Among the genes that were only differentially expressed at the steady-symptom stage (DEG 2 set; Supplemental Table S1), gene ontology (GO) revealed changes affecting proteins of the thylakoid and other plastid-related cellular components (Supplemental Fig. S3A).

Figure 2.

Workflow and global analysis of DEGs and DEPs in maize seedlings responding to SCMV infection at the presymptomatic (3 dpi) and steady-symptom (9 dpi) stages of systemic infection. A, Workflow for the integrated transcriptomic and proteomic analyses. Images of aboveground tissues from representative mock-inoculated or SCMV-infected maize seedlings at 3 and 9 dpi are shown. Maize aboveground tissues (n = 3) were harvested, pooled, and used to extract total RNA. RNA-seq allowed identification of DEGs and differentially spliced transcripts. The same aboveground samples were subjected to iTRAQ labeling (n = 2), followed by differential expression analysis. Changes in expression at the post-transcriptional level were assessed by cross-correlating transcriptomic and proteomic data with alternative splicing analysis, followed by biological function analysis of the ZmPSY1 transcripts that responded to SCMV infection. B, The Venn diagram indicates overlap between DEGs affected by only x₁ (item of Infection Status) itself and DEGs affected by x₁ × x₂ (interaction item of Infection Status and Infection Stage) represented in a green circle and a blue circle, respectively. Numbers and percentages of DEGs 1, 2, and 3 are shown. DEG 1 did not change with x₁ × x₂ and contained two types shown in pink-colored schematic line graphs below. These colored schematic line graphs (“fictional models”) were used as a classification method. The y-axes of the line graphs represent gene expression level (TPMs). Asterisks indicate significant differences between mock-inoculated and SCMV-infected samples at 3 or 9 dpi. The DEG 1 was affected only by x₁ and contained 413 upregulated DEGs and 1,296 downregulated DEGs. The DEG 2 was affected only by x₁ × x₂, and contained two types at 9 dpi (93 upregulated and 1,148 downregulated DEGs). The DEG 3 was affected by both x₁ and x₁ × x₂, and contained two types (5,627 upregulated and 1,557 downregulated genes) at 3 dpi. The degree of difference for DEG 3 was variable at 9 dpi due to x₁ × x₂ effect. C, Venn diagrams showing the overlaps between the upregulated (left) and downregulated (right) DEPs at 3 and 9 dpi.

The iTRAQ-based proteomic analyses produced a total of 1,173,943 high-resolution tandem mass spectra (MS). From these spectra, 33,252 peptides were unique, and from these unique peptides, a total of 7,530 proteins were identified (Supplemental Table S2). Among the identified proteins, 6,161 were quantified with protein and peptide false discovery rate (FDR) < 0.01 (Supplemental Table S2). Using a quantification ratio > 1.2 or < 0.83 as the upregulated cutoff or the downregulated cutoff, respectively, we identified 75 (of which 32 were upregulated, 43 downregulated) and 301 (178 upregulated, 123 downregulated) differentially expressed proteins (DEPs) at 3 and 9 dpi, respectively (Fig. 2C; Supplemental Table S3). KEGG pathway enrichment analyses showed that the DEPs at 3 dpi were enriched for annotation related to the regulation of autophagy, protein processing in the ER, DNA replication, and other functions (Supplemental Fig. S5), while the DEPs at 9 dpi were enriched for proteins annotated as related to metabolic pathways, protein processing in the ER, spliceosomes, and biosynthesis of secondary metabolites (Supplemental Fig. S6).

In plant cells, correlations between mRNA transcript and protein levels are generally considered to be positive (Ponnala et al., 2014); however, it is now appreciated that mRNA–protein correlations are weak in organisms that are responding to biotic and abiotic stresses (Ponnala et al., 2014; Jiang et al., 2017; Sharma et al., 2017; Ye et al., 2017). To test this, we explored associations between changes in mRNA level and corresponding proteins during SCMV infection. Corresponding transcripts were detectable for 97% (5,983 of 6,161) of the quantified proteins. However, when we examined the putative mRNA transcripts for all DEPs, we found that only ∼57% (43 of 76) of the DEPs at 3 dpi and ∼62% (188 of 301) of the DEPs at 9 dpi showed the same trends in regulation between the mRNA and protein levels (Supplemental Table S4). In addition, ∼55% of non-DEPs were correlated with DEGs. These results indicated that SCMV-induced changes in protein expression do not correlate simply to changes in mRNA accumulation, which is suggestive of post-transcriptional regulation occurring.

SCMV Modulates Transcription and Splicing of Host Transcripts

Because changes in host mRNA splicing patterns have recently been reported in virus-infected plants (Mandadi and Scholthof, 2015; Martin et al., 2016; Ghorbani et al., 2019), we analyzed our data sets for the relative proportions and subtypes of alternative splicing events in mock-inoculated or SCMV-infected plants. We observed that, for the number of alternative splicing events, there were no significant changes at 3 dpi while an obvious decrease by 9 dpi occurred in SCMV-infected plants compared with that of mock-inoculated plants (Supplemental Fig. S7A). Using the generalized linear model as described above, we compared the differentially alternatively spliced genes (DASGs) with DEGs, which identified 5,026 transcripts originating from 467 genes as DASGs after SCMV infection. These DASGs were divided into three groups: DASG 1 (108 genes) and DASG 3 (186 genes) starting from the presymptomatic stage, while the genes in DASG 2 (173 genes) were only differentially spliced at the steady-symptom stage (Supplemental Fig. S7B; Supplemental Table S5). Approximately 64% of DASGs (DASGs 1+3/DASGs = 314/489) occurred in SCMV-infected plants at 3 dpi, suggesting that SCMV infection had altered alternative splicing patterns from the presymptomatic stage. Only 14% of DASG 1 (DASG 1 versus DEG 1), 2% of DASG 2 (DASG 2 versus DEG 2), and 58% of DASG 3 (DASG 3 versus DEG 3) were associated with changes both in total mRNA level and alternatively spliced transcript frequency. We noticed that, as previously reported (Mandadi and Scholthof, 2015; Chen et al., 2018), the accumulation of specific transcripts produced by alternative splicing patterns did not simply correlate with overall transcription of a gene. For instance, the overall expression of Zea mays phytoene synthase1 (ZmPSY1, Zm00001d036345) was downregulated by 9 dpi (Supplemental Fig. S8A; Supplemental Table S1), but the accumulation of one of the ZmPSY1 splice variants, transcript T001, was significantly upregulated at this time point (Fig. 3A, the counting bin E007 that was indicated by vertical dotted line). These results suggest that SCMV infection causes changes in both transcription and splicing that can result in either up- or downregulation of total transcript accumulation.

Figure 3.

SCMV infection modifies the accumulation dynamics of ZmPSY1 transcripts but maintains ZmPSY1 synthesis. A, Differential counting bin usages identified ZmPSY1 as a gene giving rise to differentially alternatively spliced transcripts. From this maize genome annotation (http://plants.ensembl.org/Zea_mays/Info/Index), ZmPSY1 has two annotated transcripts, T001 and T002, involving seven exons (gray rectangles) with the sixth having alternative boundaries. Counting bins (shaded boxes) were constructed as depicted, and the sixth exon of variable length was split into two bins (counting bin E006 and E007). The arrowheads indicate alternative splicing sites. The y axis represented normalized read counts from RNA-seq in corresponding counting bins. The counting bin E007 that exhibited significantly differential usage between SCMV-infected and mock-inoculated (Mock) is indicated by a magenta dashed vertical line. Data of E007 shown in the larger magenta dashed box were higher magnification of the boxed area with a magenta dashed vertical line. Data for four treatments, including 3-dpi mock, 3-dpi SCMV, 9-dpi mock, and 9-dpi SCMV, are shown as red, green, blue, and purple lines, respectively. Schematic organizations of alternatively spliced ZmPSY1 transcripts are shown with the coding sequences in red rectangles and the UTRs in black boxes. B, Schematic organization of the newly discovered alternatively spliced ZmPSY1 transcript, T003. Probe used in C is complementary to the common region to the ZmPSY1 transcripts. C, RNA gel blot analysis of ZmPSY1 transcripts responding to SCMV infection at 5 and 9 dpi. This experiment was repeated twice with similar results. D and E, Relative expression levels of ZmPSY1 total transcripts (D) and T003 (E) were separately determined using RT-qPCR of RNA in the first systemically infected leaf from SCMV-infected and the equivalent leaf from mock-inoculated maize seedlings at 3, 5, 7, 9, and 12 dpi. Data were presented as means ± se (n = 6). Asterisks indicate significant differences between values for mock-inoculated and SCMV-infected plants at the same time point as evaluated by two-tailed Student’s t tests. F and G, The protein levels of ZmPSY1 were analyzed by immunoblotting at 3, 5, 7, 9, and 12 dpi in SCMV-infected plants or equivalent samples from mock-inoculated plants. The arrowhead indicates the position of the ZmPSY1 band in (F). The relative amount of ZmPSY1 (G) was calculated using the software ImageJ, and the lanes of 3-dpi mock were set to 1.0. Actin was used as gel loading control. Data were presented as means ± se (n = 3). P values evaluated by two-tailed Student’s t test analysis indicated no significant differences between the relative ZmPSY1 levels for mock-inoculated and SCMV-infected plants within time points.

SCMV Infection Alters the Accumulation Dynamics of ZmPSY1 Transcripts but Maintains ZmPSY1 Protein Levels

Recent studies (Mandadi and Scholthof, 2015; Martin et al., 2016; Ghorbani et al., 2019) and our data shown above provide evidence for modulation of splicing of host transcripts during virus infection, but it is unknown whether or not these differentially spliced transcripts are beneficial for virus infection. Based on the following facts, we focused on the gene ZmPSY1. Firstly, PSY1 catalyzes the committed step of carotenoid biosynthesis, which is required for normal chloroplast activity and morphology (Havaux, 1998; Li et al., 2008b; Shumskaya et al., 2012; Liu et al., 2014). Secondly, SCMV infection of maize is characterized by development of typical mosaic symptoms on systemically infected leaves accompanied by significant decreases in chlorophyll and carotenoid content (Fig. 1, A and B). This suggests that SCMV, like many other viruses, disrupts chloroplast function (Zhao et al., 2016; Bhattacharyya and Chakraborty, 2018). Thirdly, our integrative omics and alternative splicing analyses revealed that ZmPSY1 showed differences in overall transcriptional and splicing patterns but not at the protein accumulation level after SCMV infection (Fig. 3A; Supplemental Fig. S8). However, there was no previously known role for ZmPSY1 in responses to virus infection or to any other biotic stresses.

To identify the alternatively spliced transcripts of ZmPSY1, we performed northern blotting, molecular cloning, and analyses of RNA-seq reads. Two splicing variants, T001 and T003, were identified from the primary transcript of ZmPSY1 in maize leaves (Fig. 3; Supplemental Fig. S9). The 5′-untranslated region (UTR) and coding regions of the mature T001 and T003 transcripts are identical, while T003 possesses a longer 3′-UTR (Fig. 3, A and B). Transcript T001 has been previously annotated in the current version of the maize genome (http://plants.ensembl.org/Zea_mays/Info/Index), while T003 was newly identified in this study (Supplemental Fig. S9). With a probe targeting the coding region of ZmPSY1, northern blotting confirmed T001 and T003 in maize leaves and showed that T001 was more abundant than T003 (Fig. 3, B and C; Supplemental Fig. S9B). We did not detect another transcript, T002, which was previously reported as the major ZmPSY1 transcript in maize embryos (Buckner et al., 1996), in leaves with primers F1219 and R2037 targeting the common sequences of 3′-UTRs in T002 and T003 (Supplemental Fig. S9, A and C). Furthermore, we found two alternative polyadenylation regions (1,634 to 1,736 and 1,985 to 2,121 nucleotides) for T001 and T003, respectively, using 3′ Rapid Amplification of cDNA Ends (RACE; Supplemental Fig. S10). The full length of T003 was cloned from maize leaves and deposited in the GenBank database (accession no. MN128624). Furthermore, we found that 35 and 90 clean reads completely cover the specific sequence of T003 3′-UTR in our RNA-seq data from mock-inoculated and SMCV-infected maize leaves, respectively (Supplemental Figs. S11 and S12). These results clearly demonstrated that T003 is steadily transcribed, and accumulates in maize leaves.

The above northern blotting (Fig. 3C) showed that both T001 and T003 levels were reduced in SCMV-systemically infected leaves compared with those in mock-inoculated plants at 5 and 9 dpi, and both T001 and T003 levels at 9 dpi were higher than that at 5 dpi. For more precise quantification of the accumulation levels of T001 and T003, we performed reverse transcription quantitative PCR (RT-qPCR) with primers targeting both T001 and T003 or specific to T003. As the T001 RNA sequence is completely contained within the T003 RNA, we could not directly quantify the T001 level but we could infer the T001 level in maize leaves by quantifying T003 and subtracting from the total ZmPSY1 transcript accumulation level, i.e. the difference between total ZmPSY1 and T003. Compared with mock-inoculated plants, overall ZmPSY1 transcript accumulation levels were significantly reduced to 77%, 33%, 59%, and 63% in SCMV-infected plants at 3, 5, 7, and 9 dpi, respectively (Fig. 3D). At 12 dpi, the overall ZmPSY1 transcript accumulation levels were significantly increased by ∼1.4-fold in SCMV-infected plants compared with mock-inoculated plants. For T003, the accumulation level was decreased (reduced to 68%, 69%, 67%, 55%, and 65%) over the time course with the lowest level seen at 9 dpi in SCMV-infected plants (Fig. 3E). Similar expression patterns of T003 were observed over the time course in SCMV-infected and mock-inoculated plants (Fig. 3E). However, the patterns of overall ZmPSY1 transcript accumulation levels were clearly different between SCMV-infected and mock-inoculated plants (Fig. 3D). Specifically, the overall ZmPSY1 transcript accumulation in mock-inoculated plants peaked at 5 dpi and gradually decreased between 5 and 12 dpi, presumably reflecting changes in ZmPSY1 gene expression during leaf development, while in SCMV-infected plants the overall levels of ZmPSY1 transcripts showed only slight changes by 3 to 12 dpi (Fig. 3D; Supplemental Fig. S13). Given that transcript T001 was more abundant than T003 in the total ZmPSY1 transcript population, these results show that SCMV infection alters splicing of the primary ZmPSY1 transcripts leading to preferential accumulation of T001 occurring by later infection stages.

We wondered how SCMV-induced differences in T001 and T003 accumulation would impact ZmPSY1 mRNA activity and thence accumulation of ZmPSY1 protein. Our iTRAQ-based proteomic data showed that ZmPSY1 protein accumulation was not significantly altered by SCMV infection at either 3 or 9 dpi (Supplemental Fig. S8B). To verify this, we generated a polyclonal antiserum against ZmPSY1 and performed immunoblotting. As expected, changes in ZmPSY1 protein accumulation over the time course in SCMV-infected plants were similar to that in mock-inoculated plants (Fig. 3F). Specifically, in mock-inoculated plants, the changes in ZmPSY1 protein accumulation were positively correlated with the expression patterns of ZmPSY1 transcripts over the time course (Fig. 3, D and G, blue lines). In SCMV-infected plants, ZmPSY1 protein accumulation is similar to the accumulation in mock-inoculated plants over the time course, although the total ZmPSY1 transcript accumulation decreased sharply at 5, 7, and 9 dpi compared with ZmPSY1 transcript in mock-inoculated plants (Fig. 3, D and G, red lines). Altogether, these results suggested that SCMV infection altered alternative splicing in a way that maintained ZmPSY1 protein accumulation (Fig. 3G).

Maize Mutants Carrying Nonsense Mutants of ZmPSY1 Are Less Susceptible to SCMV Infection

To determine the possible contribution of ZmPSY1 during SCMV infection, we identified two nonsense mutants of ZmPSY1, d13 and d14, from the Maize EMS-induced Mutant Database (Lu et al., 2018). PCR and sequencing confirmed replacements of C by T at nucleotide 3,144 (ZmPSY1^C3144T) in d13 and of G by A at nucleotide 3,179 (ZmPSY1^G3179A) in d14, respectively (Fig. 4A; Supplemental Fig. S14, C and D). The ZmPSY1^C3144T mutation in d13 results in premature termination of ZmPSY1 translation (Q280 mutated to a stop codon). The ZmPSY1^G3179A mutation in d14 causes a splice site alteration leading to retention of the fourth intron, which also causes premature termination of ZmPSY1 translation (Fig. 4A; Supplemental Fig. S14E).

Figure 4.

Mutation of the ZmPSY1 gene ameliorates SCMV-induced pathogenesis. A, Schematic representation of the ZmPSY1 transcript T001, annotated with the positions of mutations in the two psy1 mutant maize lines d13 and d14. Red rectangles and interconnecting black broken lines represent coding region and introns, respectively, and red outlined boxes indicate UTRs. B and C, the left shows mosaic/streaking symptoms in systemically infected leaves of SCMV-infected seedlings of the d13 and d14 mutant lines and those in wild-type (WT) seedlings at 5 dpi (B) and 9 dpi (C). Middle representation in (B) and (C), accumulation of viral RNA was measured in wild-type, d13, and d14 leaves by RT-qPCR using ZmUbi as the internals standard. Error bars represent means ± se (n = 3). Asterisks indicated significant difference as evaluated by two-tailed Student’s t test analysis. The right shows immunoblot analysis of accumulation levels of SCMV CP in the first systemically infected leaf of wild-type, d13, and d14 seedlings. Arrowheads indicate the position of SCMV CP. The relative intensity of each band detected with anti-CP was quantified and the level of CP in wild-type plants was taken to be 1.0. Values (means ± se) are graphically presented on a histogram (n = 3). The different letters above each bar in the right of B and C indicate statistically significant differences as determined by a one-way ANOVA followed by Tukey’s multiple test (P < 0.05). Scale bars = 0.5 cm. D, TEM of chloroplasts in the first systemically infected leaves of mock-inoculated (left) and SCMV-infected (center) wild-type, d13, and d14 seedlings at 9 dpi. Grana are indicated with red arrows and their thicknesses were quantified for chloroplasts in both mock-inoculated and SCMV-infected plants (histograms to the right of the electron micrographs). Formation of peripheral vesicles (red arrowheads) was only observed in chloroplasts of SCMV-infected wild-type plants but not in SCMV-infected d13 and d14. Data are presented as means ± se (n = 9 or 10 chloroplasts). Significant differences were identified by a one-way ANOVA followed by Tukey’s multiple test and are shown by the different letters above each bar (P < 0.05). Scale bars = 0.5 μm.

Plants of both mutants, d13 and d14, displayed a pale-green phenotype in their leaves (Supplemental Fig. S14, A and B), which is similar to a previously described ZmPSY1-frameshift mutant, y1-8549 (Li et al., 2008b). Chlorophyll and carotenoid levels in the first true leaves of d13 and d14 plants were significantly reduced compared to those in wild-type B73 plants. There were 57.4% and 56.1% decreases in chlorophyll a, 43.9% and 36.9% decreases in chlorophyll b, and 40.7% and 41.4% decreases in levels of carotenoids in d13 and d14, respectively (Supplemental Fig. S14, A and B, right). These data confirmed that no ZmPSY1 expression occurs in d13 and d14 mutant plants.

After SCMV infection, d13 and d14 seedlings showed much milder mosaic symptoms than wild-type seedlings at both 5 and 9 dpi (Fig. 4, B and C). RT-qPCR and immunoblotting showed that accumulation levels of both SCMV RNA and CP were significantly lower in d13 and d14 plants than in wild-type plants at both time points (Fig. 4, B and C). These results indicated that ZmPSY1 is required for supporting efficient SCMV multiplication and normal SCMV-induced pathology in maize.

Because PSY is required for the normal function and structure of chloroplasts (Havaux, 1998; Shumskaya et al., 2012; Liu et al., 2014), and in light of the observed differences in photosynthetic pigment content in the d13 and d14 mutant plants (Supplemental Fig. S14, A and B), we imaged the ultrastructure of chloroplasts by transmission electron microscopy (TEM). Grana were thinner in chloroplasts of both mutants than in wild-type chloroplasts (Fig. 4D). These changes are consistent with the paler color and decreased photosynthetic pigment content of these mutants. In SCMV-infected plants, at 9 dpi we found that chloroplasts in SCMV-infected wild-type seedlings had significantly thinner grana, and in SCMV-infected d13 and d14 mutant plants granal thickness was also decreased (Fig. 4D). Vesicles that appeared to be chloroplast-derived were observed around chloroplasts in SCMV-infected wild-type plants; however, vesicles were not seen around chloroplasts in cells of SCMV-infected d13 and d14 mutant plants (Fig. 4D). These results indicate that the mutant plants lacking functional ZmPSY1 have altered chloroplast physiology and reduced susceptibility to SCMV infection.

The 3′-UTRs of ZmPSY1 Transcripts Control mRNA Translation Activity

The results shown above indicate that SCMV infection modulated post-transcriptional regulation of ZmPSY1 to maintain its translational level. We wondered how SCMV-induced differences in T001 and T003 accumulation would impact ZmPSY1 mRNA activity. Because the only difference between ZmPSY1 transcript variants T001 and T003 is that T003 has a longer 3′-UTR (Fig. 3, A and B), we further dissected the structures and functions of the 3′-UTRs for the alternative splicing products, T001 and T003. A bioinformatic analysis of RNA folding of the 3′-UTRs of ZmPSY1 transcript variants showed that the T003-3′-UTR, but not the T001-3′-UTR, is predicted to fold into a highly branched structure (Fig. 5A). This suggests that the 3′-UTRs of T001 and T003 might condition the differences in mRNA translation efficiency.

Figure 5.

The ZmPSY1 T001 3′ UTR enhances translation efficiency more than that of T003. A, RNA structural analyses of the 3′UTRs of ZmPSY1 T001 (1,349 to 1,776 nucleotides) and T003 (1,349 to 2,121 nucleotides). The structures were colored by base-pairing probabilities (from 0 to 1). For unpaired regions, the colors denoted the probability of being unpaired. The 3′-UTRs are shown in blue boxes. B, Sensor constructs comprising the EGFP-coding sequence fused to 3′-UTRs. The EGFP coding sequence was fused with the native 3′-UTRs (construct T001-3′-UTR and T003-3′-UTR). 35S, CaMV 35S promoter; NOS, NOS terminator. C to G, EGFP constructs were transiently expressed in N. benthamiana leaves and examined via observation using a confocal microscope with the same laser parameters for EGFP fluorescence (C and D), by immunoblot analysis (E and F), and by RT-qPCR analysis (G) of leaves infiltrated with the sensor constructs. EGFP fluorescence was quantified in (D) using the LAS Application Suite X imaging system (Leica; n = 14 leaves from three biological replicates). Asterisks in (D) and (F) indicate significant differences evaluated by two-tailed Student’s t tests. Error bars are means ± se. Quantification of EGFP mRNA levels was performed with five independent experiments (n = 5); error bars are means ± se (G). The relative intensities of bands detected by immunoblotting were quantified using the software ImageJ, and the value for T001 in each biological repeat (n = 3) was arbitrarily set to 1.0. Gels were stained with Coomassie Blue R-250 as loading control; band of the large subunit of Rubisco is shown. Scale bar = 1.0 cm on the leaf image; scale bars = 250 μm in microscopic images. H and I, Reporter transcripts bearing ZmPSY1 T001-3′-UTR and T003-3′-UTR exert different translation efficiency in the wheat germ in vitro translation system. The EGFP (including stop codon) with variant 3′-UTRs of ZmPSY1 transcripts were separately transcribed in vitro. In vitro translation was performed in a wheat germ lysate system with equimolar amounts (0.01 nmol) of mRNA species at 25°C for 1 h. After boiling in SDS sample buffer, translation products were analyzed by immunoblotting. (−) Negative control, adding sterile water in wheat germ lysate system; (+) positive control at a 1:1,000 dilution, N. benthamiana leaves transiently over-expressing EGFP with no UTR construct. Red arrowhead indicates the bands of EGFP. Gels were stained with Coomassie blue R-250 as loading control. The relative intensity of each band detected with anti-EGFP was quantified and the level of the EGFP band for T001-3′-UTR taken to be 1.0. Values were presented as means ± se (n = 3). Significant differences were identified using two-tailed Student’s t tests.

A previous study reported that, in contrast to maize ZmPSY1, Arabidopsis (Arabidopsis thaliana) AtPSY splicing variants have differently sized 5′-UTR sequences that affect translation efficiency (Álvarez et al., 2016). Here, to see if the structural differences between the 3′-UTRs of the ZmPSY splice variants affect translation, we performed functional analyses of the T001 and T003 3′-UTRs. For this, we expressed the full-length 3′-UTRs of T001 and T003 with each fused to a sequence encoding the reporter enhanced-GFP (EGFP-T001-3′-UTR, and EGFP-T003-3′-UTR) separately by agroinfiltration in Nicotiana benthamiana, following previously reported methods (Fig. 5B; Wachter et al., 2007; Klein-Cosson et al., 2015). At 72-h post-infiltration, EGFP fluorescence, as well as EGFP protein accumulation, was monitored (Fig. 5, C to F). As expected, N. benthamiana leaves infiltrated with the EGFP-T001-3′-UTR construct consistently showed strong fluorescence signals and accumulation of EGFP protein, while the samples with the EGFP-T003-3′-UTR construct showed significantly weaker fluorescence signals and EGFP protein accumulation that was ∼33% lower than that of the EGFP-T001-3′-UTR construct (Fig. 5, C to F). RT-qPCR revealed no significant difference in EGFP transcript accumulation between tissues infiltrated with either the EGFP-T001-3′-UTR or EGFP-T003-3′-UTR constructs (Fig. 5G). In addition, the presence or absence of the tombusvirus RNA silencing suppressor protein P19 had no influence on these results (Supplemental Fig. S15). This showed that differences in the susceptibility to RNA silencing of the two 3′-UTR sequences and consequent effects on mRNA stability did not explain the effects on protein accumulation.

To further confirm the differences in translation efficiency conferred by the alternative 3′-UTRs, the effect of both 3′-UTRs on EGFP expression was examined using wheat (Triticum aestivum) germ in vitro translation assays with equimolar quantities of RNA molecules synthesized from in vitro transcription reactions. Translation products were analyzed by SDS-PAGE followed by immunoblotting. The transcript T001-3′-UTR gave rise to significantly greater quantities (∼1.7-fold) of translation product than T003-3′-UTR (Fig. 5, H and I), demonstrating that ZmPSY1 3′-UTRs control mRNA translation activity. These results confirm that the T001-3′-UTR sequence enhances translation more efficiently than the T003-3′-UTR sequence. Collectively, the high abundance and translation efficiency of T001 plays a major role in maintaining ZmPSY1 protein level.

The ZmPSY1 Splicing Variant T001 Plays a Critical Role in Virus Infection

To identify the functions of ZmPSY1 transcripts in supporting SCMV infection, we used a virus-induced gene silencing (VIGS) vector (cucumber mosaic virus [CMV]-based gene silencing vector pCMV-2b_N81) to decrease ZmPSY1 transcript accumulation (Wang et al., 2016). Because the longer T003 transcript sequence also comprises the complete sequence of T001, sequences common to both transcripts or specific to the T003-3′-UTR could be used, respectively, to silence all ZmPSY1 transcripts or to specifically silence transcript T003 (Fig. 6A). Therefore, for VIGS of all ZmPSY1 transcripts, a 200-bp fragment from the ZmPSY1 open reading frame (ORF) was inserted into the pCMV201-2b_N81 vector, resulting in construct CMV-PSY1 (Fig. 6A, in green). For specific silencing of T003, a 103-bp fragment specific to the T003 3′-UTR was inserted into the pCMV201-2b_N81 vector, resulting in construct CMV-T003 (Fig. 6A, in blue).

Figure 6.

VIGS-mediated knockdown of ZmPSY1 transcripts significantly decreases SCMV accumulation. A, Location of sequences subcloned into the CMV vector to stimulate VIGS of all ZmPSY1 transcripts (CMV-ZmPSY1) or only of the alternatively spliced transcript T003 (CMV-T003). Red rectangles and interconnecting black broken lines represent coding region and introns, respectively; black blank boxes show UTRs; green and blue boxes sequences are complementary to target sequences. Arrows marked QF4/QR4 and QF10/QR10 show the positions of RT-qPCR primers used in C and D. B to G, Silencing of ZmPSY1 and T003 decreased SCMV infection, viral RNA, and CP accumulation. RT-qPCR assays were used to quantify ZmPSY1 transcripts and SCMV RNA accumulation in the first systemically infected leaves of silenced seedlings and control seedlings infected with CMV-GFP (C–E). Data in C to E are presented as means ± se (n = 5). Significant differences were identified by one-way ANOVA followed by Tukey’s multiple test. The different letters above each bar indicate statistically significant differences (P < 0.05). Immunoblot analysis of CP (band indicated with a black arrowhead) accumulation in F and G. The relative intensity of each CP band was quantified. Error bars represent means ± se (n = 3). Statistical differences were determined by two-tailed Student’s t tests. Scale bar = 1.0 cm.

Maize plants preinfected with the VIGS or control vectors were inoculated with SCMV to determine how knockdown of ZmPSY1 transcript accumulation impacted SCMV infection and/or pathogenesis. SCMV produced typical mosaic symptoms on upper leaves of plants preinfected with the control CMV vector and produced a slightly milder mosaic symptom on plants preinfected with the CMV-T003-specific vector. Significantly, the mildest symptoms, consisting only of relatively few chlorotic spots occurring in limited areas of upper leaves, were observed upon silencing plants preinfected with CMV-PSY1 (Fig. 6B). In SCMV-infected plants, there was an ∼60% decrease in ZmPSY1 mRNA levels in plants preinfected with CMV-PSY1 (Fig. 6C, middle column). Plants preinfected with CMV-PSY1 supported significantly less accumulation of SCMV RNA (∼80% reduction, Fig. 6E) and CP (∼60% reduction, Fig. 6F). T003 levels were not significantly reduced in plants preinfected with CMV-PSY1 (Fig. 6D, middle column). Meanwhile, in plants preinfected with CMV-T003, there was an ∼40% decrease in ZmPSY1 total transcript accumulation and a slight decrease for T003 levels (Fig. 6, C and D, right columns), which resulted in an ∼40% decrease in SCMV RNA accumulation and a slight decrease in SCMV CP accumulation (Fig. 6, E and G). These results clearly demonstrated that ZmPSY1 T001 plays positive roles in supporting efficient SCMV multiplication and pathogenesis in maize.

DISCUSSION

SCMV Induces Extensive Changes in Host Gene Expression from the Presymptomatic Infection Stage

Pathogens have evolved diverse strategies to modulate host gene expression by effects on transcriptional and post-transcriptional processes, resulting in altered cellular metabolism to facilitate their own multiplication (Mandadi and Scholthof, 2013). In this study, we performed integrative analyses of maize transcriptomics data for presymptomatic and steady-symptom stages after SCMV infection, separately investigating the differences in gene expression caused by two independent variables (developmental and viral infection) as well as exploring their interactions. We adopted the approach of Hickman et al. (2017), who used a generalized linear model to identify jasmonate-induced changes in gene expression in Arabidopsis, in which both the time after treatment and the treatment itself were considered as factors. This provided detailed insights into the architecture and dynamics of this regulatory network (Hickman et al., 2017). In our analyses, we considered the effect of seedling development on gene expression using this linear model and found that ∼85% (19,210 to 22,564) of quantified genes were differentially expressed (Supplemental Table S6). Among these quantified genes ∼45% (10,134 of 22,564) were also DEGs implicated in SCMV infection (Supplemental Table S1). Subsequently, we focused on the DEGs induced by SCMV infection. We found that most (∼88%) DEGs occurred at the presymptomatic stage, which is consistent with the concept that early infection is an opportune window for pathogens to reprogram the function and/or accumulation of host transcription factors and splicing factors (Mandadi and Scholthof, 2015). Moreover, annotations relating to protein processing in the ER, and to fatty acid biosynthesis, were enriched among our SCMV-responsive DEGs and DEPs, implying that SCMV, similar to another potyvirus, TuMV, remodels host cellular membranes to promote successful infection (Cotton et al., 2009; Grangeon et al., 2012; Movahed et al., 2019).

Our linear-modeling–based correlation analyses of transcriptomic and proteomic data from the infection time course uncovered a complex reprogramming of RNA splicing activity occurring in maize after SCMV infection. It should be noted that this process in particular (and we suspect processes mediated by alternative splicing more generally) would be missed entirely or significantly underestimated if our analytical approach considered DEGs and/or DEPs only. Conceivably, starting from the presymptomatic stage, SCMV can cause profound effects on maize gene expression at both the transcriptional and post-transcriptional levels.

Alternative Splicing: A New Battleground in Plant–Pathogen Interactions

We identified 467 DASGs in maize seedlings responding to SCMV infection (Supplemental Table S5). Though extensive impacts on host alternative splicing have been reported for several cases of pathogen infection, little is known about the mechanism(s) by which plant pathogens trigger alterations in RNA splicing (Huang et al., 2017; Jiang et al., 2018). During premRNA splicing in the spliceosome, transcript variants occur due to differential usage of splice sites. Splice-site selection is determined not only by core spliceosome components but also, and to a large extent, by other RNA-binding proteins, predominantly Ser/Arg-rich proteins and heterogeneous nuclear ribonucleoproteins (Erkelenz et al., 2013; Laloum et al., 2018). There are demonstrated instances of pathogen-triggered changes in Ser/Arg-rich protein expression and of direct physical interaction between pathogen proteins and host Ser/Arg-rich proteins (Huang et al., 2017; Jiang et al., 2018). Probably the clearest example to date is the Phytophthora sojae effector protein PsAvr3c, which interacts with and recruits the soybean (Glycine max) Ser/Arg-rich GmSKRPs to manipulate host RNA splicing (Huang et al., 2017). Our observation that a putative premRNA splicing factor (DEAH5, Zm00001d031581) was among the differentially spliced transcript species suggested that SCMV might reprogram splicing in maize cells to enhance its multiplication (Supplemental Table S5). The DEGs and DEPs detected in our time-course experiments were enriched for annotations relating to the RNA-splicing pathway, and can thus be assumed to impinge on splicing homeostasis during maize–SCMV interactions (Supplemental Fig. S4).

It was notable that the majority of detected DASGs did not correlate with DEGs. This distinction supports previous observations and highlights that transcription-level expression changes are often, perhaps usually, inconsistent with changes to transcript splicing patterns (Mandadi and Scholthof, 2015; Chen et al., 2018). For instance, transcription of Zm00001d043911, a drought-induced factor with two annotated transcripts (T001 and T002), was significantly upregulated at 3 dpi. However, the expression of Zm00001d043911_T001 was significantly downregulated at 3 dpi (Supplemental Fig. S16). These discrepancies indicate that elaborate regulatory mechanisms balance changes in transcription and generation of splice variants as plants respond to pathogen infection. Moreover, our demonstration that SCMV infection can selectively favor accumulation of one alternative splicing variant (ZmPSY_T001) over another, suggests that such regulatory mechanisms may represent a previously unappreciated battleground for plant-pathogen interactions.

SCMV Infection Interferes with ZmPSY1 Alternative Splicing Patterns for Efficient Multiplication

Until now, the function of altered splicing transcripts by pathogen infection remains largely unknown. In this study, we focused on ZmPSY1 transcripts for functional research on SCMV infection. We found that the overall transcriptional level of ZmPSY1 decreased upon SCMV infection, but the accumulation of T001 relative to T003 was altered, with T001 becoming predominant by later stages of infection. Using VIGS to suppress accumulation of ZmPSY1 transcripts, but in particular by suppressing T001 transcript accumulation, decreased SCMV titers and alleviated the severity of mosaic symptoms. Furthermore, experiments with nonsense ZmPSY1 mutants also revealed decreased SCMV replication level and milder symptoms, confirming that ZmPSY1 is a proviral factor that facilitates viral multiplication and pathogenesis. Altogether, these results demonstrated that SCMV can modulate alternative splicing of ZmPSY1 primary transcripts to preferentially enhance accumulation of T001 to maintain efficient SCMV infection. Given that T001 is the major transcript of ZmPSY1 in maize leaves and possesses higher translational efficiency than T003, SCMV infection modulates the alternative splicing of ZmPSY1 to produce more T001—thus maintaining sufficient ZmPSY1 protein accumulation to support infection.

As the rate-limiting enzyme for carotenoid biosynthesis, PSY is highly regulated and its expression level often correlates with the quantities of carotenoids or their derivatives formed (Álvarez et al., 2016). Maize encodes three PSY isozymes (i.e. ZmPSY1, ZmPSY2, and ZmPSY3) with a shared protein identity of 64.7% (Gallagher et al., 2004; Li et al., 2008a, 2008b). ZmPSY2 and ZmPSY3, like most PSYs in other plants (e.g. Arabidopsis PSY), localize to plastoglobuli in plastids (van Wijk and Kessler, 2017). ZmPSY2 plays a key role for photomorphogenesis and is induced by light in photosynthetic tissue. ZmPSY3 is associated with root carotenogenesis responding to drought and salt stresses (Li et al., 2008b; Shumskaya et al., 2012). ZmPSY1, with dual localization to stroma and plastid membranes, not only is related to endosperm carotenogenesis, but also plays a key role in leaf carotenogenesis, especially in the dark and under heat-stress conditions (Li et al., 2008b). It is unclear how ZmPSY1 contributes to SCMV multiplication. Previous studies suggest that ZmPSY1 appears to be essential for heat-stress–induced biosynthesis of carotenoid antioxidants that protect the plastid membrane (Davison et al., 2002; Havaux et al., 2007; Li et al., 2008b). Because potyviruses, including TuMV, maize dwarf mosaic virus, tobacco etch virus, and potato virus Y, need to recruit chloroplast membranes for their replication (Mayhew and Ford, 1974; Gadh and Hari, 1986; Gunasinghe and Berger, 1991; Wei et al., 2010), maintaining chloroplast function and preventing the disruption or destruction of chloroplasts is critical for potyvirus replication. This is also supported by the poor replication of SCMV in d13 and d14 mutants, which express nonsense ZmPSY1 transcripts and are associated with decreased integrity of chloroplast envelopes and disruption of granal structure. In the course of SCMV infection, mosaic symptoms are associated with disrupted chloroplast structure and decreased accumulation of carotenoids and the photosynthetic pigments chlorophyll a and chlorophyll b. Based on the above evidence, we speculate that, to achieve efficient replication, SCMV modulates alternative splicing of ZmPSY1 to produce relatively more T001, maintaining ZmPSY1 synthesis and thereby preserving functional chloroplast structures.

Differences among Transcriptional, Post-Transcriptional, and Translational Regulation of PSY Gene Expression in Maize and Arabidopsis

In Arabidopsis, alternative splicing of PSY primary transcripts results in differences in 5′-UTR sequences that control PSY mRNA translation efficiency (Álvarez et al., 2016). We discovered that maize synthesizes ZmPSY1 transcripts with identical 5′-UTRs and ORFs, but with different 3′-UTRs, resulting from alternative splicing. In maize seedling leaves, we discovered the T001 and T003 variants, while no T002 was detected. These findings are consistent with two previous studies that reported the presence of the T002 variant in maize embryos, whereas two other putative ZmPSY1 variants, a 1.8-kb band that may be T001 and another 2.0-kb band (which might be T003) were detected by northern-blot hybridization of leaf RNA from 1-week-old seedlings (Buckner et al., 1990; 1996). Differences among transcripts T001, T002, and T003 only exist at their respective 3′-UTRs. Moreover, we found there are multiple polyadenylation sites for T001 and T003, similar to what is known for T002 in embryos (Buckner et al., 1996). Collectively, these findings from multiple studies reemphasize our earlier point that alternative splicing and accumulation of ZmPSY1 variants are differentially orchestrated by developmental as well as stress factors.

In this study, we found that the shorter T001 3′-UTR has higher mRNA activity than that of T003. The construct containing the 3′-UTR of T001 yielded more EGFP accumulation levels in vivo and in vitro than that of T003, independent of expression levels (Fig. 5). Several studies showed that alternative splicing in 5′-UTRs regulates translation efficiency of mRNA variants (Remy et al., 2014; Álvarez et al., 2016). In the case of a zinc-induced facilitator2 (ZIF2), intron retention (IR) in 5′-UTR is important in generating the variant ZIF2.2 needed for protecting Arabidopsis against excess zinc (Remy et al., 2014). When the 5′-UTR intron is retained, an imperfect stem loop is predicted to form—one that enhances ZIF2 translation (Remy et al., 2014). Interestingly, secondary structure prediction of the ZmPSY1 T003 indicated the formation of a highly branched structure in 3′-UTR, but this structure was lost in the T001 variant with its shorter 3′-UTR (Fig. 5A). Moreover, our bioinformatics analyses revealed that transcribed EGFP mRNAs bearing the T001-3′-UTR or T003-3′-UTR possess different RNA folding structures, particularly in their 3′-UTRs (Supplemental Fig. S17). Accordingly, we found these structural changes of the 3′-UTR affected EGFP translation efficiency (Fig. 5). Secondary structures within 3′-UTRs affect riboswitches, suggesting that this might regulate translation of mRNAs with different 3′-UTR structures and thus affect translation inhibition or translation permissiveness (Wachter et al., 2007; Manzourolajdad and Arnold, 2015). Thus, the presence of relatively more T001 would increase the translation of total ZmPSY1 transcripts, resulting in relatively higher ZmPSY1 protein accumulation levels under stress conditions. It might be interesting to investigate the role of T003. Though T003 does not seem to respond to SCMV infection it may function in other stress conditions as, intriguingly, it possesses a complicated 3′-UTR structure (Wachter et al., 2007). Taken together, these findings suggested that alternative splicing in the 3′-UTR can fine-tune gene expression levels to help plants responding to stresses.

In summary, our study provides a novel perspective for discussing how changes in splicing of UTRs promote synthesis of proviral factors. We demonstrated that the T001 and T003 transcripts possess identical 5′-UTRs and encode identical ZmPSY1 proteins—and differ only in the length and structure of their 3′-UTRs. The 3′-UTR sequence of T001 makes it the more efficient of the two ZmPSY1 mRNAs occurring in maize leaves. Due to the dynamic changes in the relative amounts of T001 and T003 over the course of infection, we speculate that although maize leaf cells may attempt to counteract SCMV infection by decreasing overall ZmPSY1 transcription early in infection, the virus infection causes changes in splicing that lead to preferential accumulation of the T001 variant. Because T001 is a more efficient mRNA that yields more ZmPSY1 protein, this outcome permits the SCMV to continue to manipulate chloroplast function and, by means yet to be determined, this promotes SCMV infection.

MATERIALS AND METHODS

Plant Growth Conditions and Virus Inoculations

Maize (Zea mays ‘B73’) and Nicotiana benthamiana plants were grown in controlled environment chambers set at 16-h light/24°C and 8-h dark/22°C. The SCMV-Beijing isolate (Fan et al., 2003) was propagated in maize plants, and infected leaf extracts were used to inoculate the first true leaf of 8-d-old maize seedlings, as described in Chen et al. (2017a). Maize seedlings of the same age inoculated with phosphate buffer were used as mock-inoculated control plants.

Photosynthetic Pigment Quantification

Pigments were extracted using 95% (v/v) ethanol from fresh leaf tissue (∼0.1 g) of maize seedlings. At least three separate samples from three plants were measured for each treatment. Levels of chlorophyll a, chlorophyll b, and carotenoids were determined using a UV/VIS Spectrophotometer (model no. 721N; INESA Instruments) by measuring absorbance at 665, 649, and 470 nm, respectively (Wang et al., 2016). Pigment contents were calculated as described in Lichtenthaler (1987).

RNA Extraction and RNA-Seq

Three independent biological replicate samples of aboveground shoots were separately collected from SCMV-infected maize plants at 3 and 9 dpi and from equivalent tissues of mock-inoculated plants, following the sampling method of Mandadi and Scholthof (2015). Total RNA was isolated by use of the TRIzol reagent (Invitrogen) and further purified with the RNase-Free DNase Set (Qiagen) and the RNeasy Micro Kit (Qiagen). RNA quality was checked using a model no. 2100 Bioanalyzer (Agilent) and NanoDrop ND-2000 (Thermo Fisher Scientific). A total of 3 μg of RNA per sample was used as input material for library construction. All procedures, including mRNA purification, complementary DNA (cDNA) preparation, ends repair of cDNA, adapter ligation, and cDNA amplification, were carried out for preparing RNA-Seq libraries as instructed by the manufacturer (Illumina). For each of the four treatments, three biological replicates were sequenced via the HiSeq2500 platform (Illumina), generating ∼30 million reads with 125-bp paired ends per sample.

Sequencing Data Processing and Transcriptomic Analysis

Sequence data were filtered with the tool Seqtk (https://github.com/lh3/seqtk) to remove adaptor sequences and low-quality sequences, and clean reads were mapped to the B73 reference genome (v.4) using the program STAR v.2.4.2a (Dobin et al., 2013) with an Ensembl gene annotation (AGPv4) as transcript index. Transcripts were assembled using the program Cufflinks (v.2.2.1; Trapnell et al., 2010). Gene expression values were normalized to transcripts per million reads (TPMs) using the software Salmon (Patro et al., 2017). In this experiment, the two independent variables (Infection Status and Infection Stage) were dummy variables including two levels: mock-inoculated and SCMV-infected (the presymptomatic [3 dpi] and steady-symptom [9 dpi] stages of systemic infection, respectively). Genes that were significantly differentially expressed were further identified using a generalized linear model (Gene Expression ∼ Infection Status + Infection Stage + Infection Status: Infection Stage) with the software Sleuth (Pimentel et al., 2017). In this model, the dependent variable y represents gene expression level (TPM). The independent variable “x₁” represents infection status including Mock and SCMV assigned 0 and 1, respectively. The independent variable “x₂” represents Infection Stage including 3 or 9 dpi assigned 0 and 1, respectively. The “x₁ × x₂” represents interaction item of Infection Status and Infection Stage. The values β₀, β₁, β₂, and β₃ are regression coefficients. To assess x₁ and x₂ effect on y for each gene, the full model was separately compared with three reduced models (in the software Sleuth [Hickman et al., 2017; Pimentel et al., 2017]). Any genes with a mean abundance of <0.5 TPM in each treatment were not included in DEG analysis (Brinton et al., 2018). Remaining genes were screened with adjusted P value ≤ 0.05 as DEGs. To assess the impact of SCMV infection on gene expression, DEGs were divided into the subsets DEG 1, 2, and 3. DEG 1 (1,709 genes) includes genes with matching expression trends at both 3 and 9 dpi; DEG 2 (1,241 genes) represents genes with expression altered exclusively at 9 dpi; and DEG 3 (7,184 genes) covers genes altered in expression at 3 dpi, regardless of their trends at 9 dpi (Fig. 2A; Supplemental Table S1). Additionally, GO term enrichment analyses were performed with the software BLAST2GO v.5.2 (Conesa and Götz, 2008). The significance of enrichment of DEGs or proteins in each GO term was determined at a threshold of adjusted P ≤ 0.05. The threshold of P ≤ 0.05 was applied to identify the KEGG pathway with significant enrichment (Kanehisa et al., 2012).

Protein Extraction and iTRAQ Labeling Proteomic Analysis

Two independent biological replicate samples were collected from infected maize plants at 3 and 9 dpi. These were the same plants used for RNA-seq. Proteins from SCMV-infected and mock-inoculated plants were isolated by precipitation in trichloroacetic acid/acetone (Sheoran et al., 2009). Further processing of proteins and peptides analyzed was performed as described in Chen et al. (2017a), unless otherwise stated. Briefly, all procedures including iTRAQ labeling, reverse-phase chromatography, and triple quadrupole time-of-flight tandem MS, were carried out.

Database Searching, Protein Quantification, and DEP Analysis

Tandem MS data were extracted by the software ProteoWizard (Chambers et al., 2012) without charge state deconvolution. All MS data were searched using the software Mascot (http://www.matrixscience.com/) in the Ensembl Z. mays database (http://ensembl.gramene.org/Zea_mays/Info/Index) assuming the digestion enzyme trypsin. Mascot was set up with following parameters: a fragment ion mass tolerance of 0.020 D and a parent ion tolerance of 10.0 ppm; carbamidomethyl of Cys, iTRAQ 4-plex of Lys, and the N terminus as fixed modifications; and oxidation of Met, acetylation of N-termini, and iTRAQ 4-plex of Tyr as variable modifications. Proteins with peptide and protein FDRs ≤ 0.01 were identified. These identified proteins achieved FDR ≤ 0.01, which were quantified by the Scaffold Q+ Local FDR algorithm (http://www.proteomesoftware.com/products/qplus/; Supplemental Table S2). Among the quantified proteins, DEPs were obtained with a quantification ratio > 1.2 as the upregulated threshold and a ratio < 0.83 as the downregulated threshold (Supplemental Table S3).

Alternative Splicing Event Characterization and Differential Exon Usage Identification

The various types of alternative splicing events, including alternative transcript start and termination, single IR, multi-IR, skipped exon, multiexon skipping, alternative exon usage, and other events, were characterized via the software ASprofle (Florea et al., 2013). To identify genes with differential exon usage, the program DEXSeq (Anders et al., 2012) was used to fit a generalized linear model and compare it with reduced models to assess DASGs due to x₁ and x₂. The DASGs were screened with adjusted P value ≤ 0.05. To consider the effects of virus infection on exon usage, the significant differential exon usages (DEUs) were divided into three types with biological explanation: (1) DASG 1—the significant DEU did not change with infection stage (x₂). (2) DASG 2—the significant DEU was only at 9 dpi with no significant DEU at 3 dpi. (3) DASG 3—the significant DEU was at 3 dpi and the degree of difference might or might not change at 9 dpi (Supplemental Table S5).

Northern Blotting

Different length transcripts of ZmPSY1 were determined by northern-blot analysis. Total RNA was extracted from leaf tissues and fractionated via electrophoresis using a 1.5% (w/v) agarose gel containing 7% (v/v) formaldehyde and then transferred to Hybond N+ nylon membranes (GE Healthcare). For analyses of different length transcripts of ZmPSY1, probes were prepared using a PCR DIG Probe Synthesis Kit (Roche) with specific primers located in a conserved region (Supplemental Table S7). Hybridization was detected by a DIG-High Prime DNA II Labeling and Detection Starter Kit (Roche) according to the manufacturer’s instructions.

Cloning of ZmPSY1 Transcripts and 3′ RACE

Near-full length ZmPSY1 transcripts were cloned using RT-PCR with pairs of specific primers that target 5′- and 3′-UTRs, respectively (Supplemental Table S7). Three independent, nearly full-length cDNA clones of ZmPSY1 transcript T003 were sequenced. RACE reactions were performed to amplify the 3′-UTRs (Buckner et al., 1996). The PCR amplification of the polydT-generated cDNA was performed with ZmPSY1 transcript-specific primers and adaptor primers. The purified PCR products were cloned into pMD19-T vector (TaKaRa).

RT-qPCR analysis

Synthesis of cDNA was performed with 2 μg of total RNA from maize leaf tissues in each 20-μL reaction, then primed with a polydT primer and catalyzed by M-MLV reverse transcriptase (Promega). RT-qPCR analysis was performed using 10-fold diluted cDNAs, gene-specific primers, and a FastSYBR mixture (CWBIO) on an ABI QuantStudio 6 Flex Real Time PCR system (Thermo Fisher Scientific). Transcript- and gene-specific primer sequences are listed in Supplemental Table S7. ZmUbi (ubiquitin) transcript accumulation was used as an internal control for assays using maize RNA. NbActin was selected as the internal control for experiments with N. benthamiana RNA. Relative expression levels of the target genes were calculated using the 2^−ΔΔCT method (Livak and Schmittgen, 2001). All experiments were repeated at least three times.

Immunoblot Analysis

Total leaf protein extraction and the immunoblot analysis were performed as described in Chen et al. (2017a). The polyclonal antibody for ZmPSY1 was generated in rabbits by immunization with a ZmPSY1 N-terminal sequence-specific peptide (PVLDARPQDMDMPR) at GenScript. ZmPSY1 expression was detected with the polyclonal antibody diluted at 1:5,000. Detection of SCMV CP was performed as described previously (Chen et al. 2017a). Anti-β actin mouse monoclonal antibody (TransGen Biotech) at a 1:5,000 dilution was used for detection of the beta actin internal control (Chen et al., 2017a). EGFP was detected with a monoclonal antibody (TransGen Biotech) at a 1:5,000 dilution. The signal was detected using an Azure c400 Imager (Azure Biosystems) and quantified using the software ImageJ (http://imagej.net/; Schindelin et al., 2015).

Agroinfiltration and GFP Imaging

Cells of Agrobacterium tumefaciens strain C58C1 carrying plasmids with 35S-driven expression cassettes were cultured in Luria–Bertani medium (containing 100 mg L⁻¹ kanamycin and 25 mg L⁻¹ rifampicin) for 16 h at 28°C by shaking at 180 rpm. Bacteria were collected by centrifugation at 4,000g for 10 min, followed by resuspension in infiltration buffer (10 mm of MgCl₂, 100 μm of acetosyringone, and 10 mm of 2-ethanesulfonic acid at pH 5.6), and standing at room temperature for 4 to 6 h before infiltration. To evaluate the functions of various 3′-UTR derivatives of ZmPSY1 transcripts on EGFP expression, different combinations of bacteria suspensions with a combined optical density (at 600 nm) of 0.5 were infiltrated into abaxial leaf surfaces of 4- to 5-week-old N. benthamiana plants with a needle-free syringe (Wachter et al., 2007; Zhang et al., 2017). At 72-h post infiltration, leaves were observed using a model no. SP8 binocular microscope (Leica; http://www.leica-microsystems.com/) to detect fluorescence. Images of GFP fluorescence driven by different EGFP-3′-UTR fusions were obtained using the same laser parameters to allow comparison (Wachter et al., 2007; Klein-Cosson et al., 2015).

In Vitro Transcription/Translation Assays

The EGFP (including stop codon) with variant 3′-UTRs of ZmPSY1 transcripts were separately cloned into the vector pSPT19 (Roche) under the control of the T7 promoter (Supplemental Table S8). In vitro transcription was performed with 1 μg of EcoR I-linearized plasmids using the T7-RiboMAX kit (Promega) in a reaction volume of 60 μL at 37°C for 3 h. The reaction products were purified using the RNase-Free DNase Set (Qiagen). In vitro translation was performed in a wheat germ lysate system (Promega) with equimolar amounts (0.01 nmol) of mRNA species at 25°C for 1 h. Translation products were boiled in SDS loading sample buffer (Laemmli, 1970) for analysis by SDS-PAGE and immunoblotting.

TEM

For TEM, the leaf blades of wild-type, d13, and d14 psy1 mutants were prepared as described in Liu et al. (2014). The samples were fixed in 2.5% glutaraldehyde in 0.1 m of sodium phosphate buffer at pH 7.2. After dehydrating in an acetone series and embedding in Spurr’s resin, the ultrathin sections of samples were obtained with an Ultracut E Ultramicrotome (Leica). A model no. CM 100 Transmission Electron Microscope (Philips) was used to observe chloroplast ultrastructure.

CMV-Induced Silencing of ZmPSY1 Transcripts

A 200-bp fragment (representing nucleotides 627 to 826 from the ATG start codon) amplified from the ZmPSY1 ORF and a 103-bp fragment specific to the 3′-UTR of T003 were separately inserted into pZMBJ-CMV201-2b_N81 (Wang et al., 2016) to generate the VIGS vectors CMV-PSY1 and CMV-T003 (Supplemental Table S8). A. tumefaciens strain C58C1 carrying pCMV101, pCMV301, and pCMV201:GFP; pCMV201:ZmPSY1; or pCMV201:T003 were prepared and infiltrated into N. benthamiana leaves. At 4 dpi, virus crude extract was isolated from the agroinfiltrated leaves and inoculated maize seeds by the vascular puncture inoculation method as described in Wang et al. (2016).

Accession Numbers

The names and accession numbers of genes can be found in Supplemental Table S9. RNA-seq data sets for the transcriptome analysis have been deposited under accession number CRA001815 and are available from the Genome Sequence Archive (http://bigd.big.ac.cn/gsa) of the Beijing Institute of Genomics Data Center (Wang et al., 2017). The raw iTRAQ-MS data sets are available in the ProteomeXchange database (http://www.proteomexchange.org/) under accession number PXD014670 (Deutsch et al., 2017).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Cluster analysis for quantified genes in maize seedlings under four treatments.

• Supplemental Figure S2. Correlation analysis for quantified proteins in maize seedlings under four treatments.

• Supplemental Figure S3. Functional enrichment analysis of DEGs in maize seedlings responding to SCMV infection.

• Supplemental Figure S4. DEGs enriched in the RNA-splicing pathway.

• Supplemental Figure S5. Functional enrichment analysis of DEPs in maize seedlings responding to SCMV infection at 3 dpi.

• Supplemental Figure S6. Functional enrichment analysis of DEPs in maize seedlings responding to SCMV infection at 9 dpi.

• Supplemental Figure S7. Global alternative splicing analysis in mock-inoculated and SCMV-infected maize seedlings.

• Supplemental Figure S8. Transcription and protein level dynamics of ZmPSY1 expression in response to SCMV infection.

• Supplemental Figure S9. Detection of alternatively spliced ZmPSY1 transcripts.

• Supplemental Figure S10. 3′ RACE analyses for T001 and T003 using 11-d-old maize seedling leaves.

• Supplemental Figure S11. Clean reads (35) from RNA-seq data of mock-inoculated maize leaves completely cover the specific sequence of the T003 3′-UTR.

• Supplemental Figure S12. Clean reads (90) from RNA-seq data of SMCV-infected maize leaves completely cover the specific sequence of the T003 3′-UTR.

• Supplemental Figure S13. Overall accumulation of ZmPSY1 transcripts in mock-inoculated (Mock) and SCMV-infected plants.

• Supplemental Figure S14. Identification of the psy1 mutations in maize lines d13 and d14.

• Supplemental Figure S15. Expression of EGFP-3′-UTR with or without coexpression of the silencing suppressor protein P19.

• Supplemental Figure S16. Transcription- and alternative splicing-level dynamics of Zm00001d043911 responding to SCMV infection.

• Supplemental Figure S17. RNA structural analyses of EGFP with the 3′-UTRs of ZmPSY1 T001 and T003.

• Supplemental Table S1. TPMs for all quantified genes and DEGs across the time series after SCMV/Mock treatment.

• Supplemental Table S2. Proteomics raw data at 3 and 9 dpi.

• Supplemental Table S3. DEPs analysis at 3 and 9 dpi.

• Supplemental Table S4. DEPs corresponding to mRNA expression level.

• Supplemental Table S5. DASGs across the time series after SCMV/Mock treatment.

• Supplemental Table S6. DEGs under the effect of seedling development.

• Supplemental Table S7. Primers used for PCR analyses.

• Supplemental Table S8. Primers used for plasmid construction.

• Supplemental Table S9. The names and accession numbers of genes.