The spliceosome assembly factor GEMIN2 attenuates the effects of temperature on alternative splicing and circadian rhythms

Significance

RNA processing, an important step in the regulation of gene expression, is mediated by proteins and RNA molecules that are highly sensitive to variations in temperature conditions. Most organisms do not control their own body temperature. Therefore, molecular mechanisms must have evolved that ensure that biological processes are robust to temperature changes. Here we identify a protein that buffers the effect of temperature on biological timing by enhancing the assembly of the spliceosome, a large ribonucleoprotein complex involved in RNA processing in organisms ranging from yeast to humans, and thereby controlling the alternative splicing of clock genes.

Abstract

The mechanisms by which poikilothermic organisms ensure that biological processes are robust to temperature changes are largely unknown. Temperature compensation, the ability of circadian rhythms to maintain a relatively constant period over the broad range of temperatures resulting from seasonal fluctuations in environmental conditions, is a defining property of circadian networks. Temperature affects the alternative splicing (AS) of several clock genes in fungi, plants, and flies, but the splicing factors that modulate these effects to ensure clock accuracy throughout the year remain to be identified. Here we show that GEMIN2, a spliceosomal small nuclear ribonucleoprotein assembly factor conserved from yeast to humans, modulates low temperature effects on a large subset of pre-mRNA splicing events. In particular, GEMIN2 controls the AS of several clock genes and attenuates the effects of temperature on the circadian period in Arabidopsis thaliana. We conclude that GEMIN2 is a key component of a posttranscriptional regulatory mechanism that ensures the appropriate acclimation of plants to daily and seasonal changes in temperature conditions.

Circadian clocks allow organisms to coordinate physiological processes with periodic environmental changes. The core of all circadian systems, in organisms ranging from cyanobacteria to humans, is a network of multiple interlocked feedback loops that operate at the transcriptional, translational, and posttranslational levels to sustain oscillations with a period of ∼24 h, even in the absence of environmental cues. An increasing body of evidence links alternative splicing (AS) with the regulation of circadian networks across eukaryotic organisms (1–3). The core clock genes period in Drosophila, frequency in Neurospora, and BMAL2 in humans undergo AS to give rise to different mRNA isoforms (1, 2, 4). In Arabidopsis, several core clock genes, including TIMING OF CAB EXPRESSION 1 (TOC1) and CIRCADIAN CLOCK ASSOCIATED 1 (CCA1), also undergo extensive AS (5–7).

Interestingly, many of the alternative mRNA isoforms associated with the Arabidopsis core clock genes are abundant or alter their abundance upon changes in environmental conditions, suggesting that they have important physiological roles (5–7). For example, there is strong evidence that temperature regulation of CCA1 AS is critical for the proper functioning of circadian rhythms under cold conditions (8). Temperature also regulates the AS of frequency in Neurospora and period in Drosophila (1, 2), thereby promoting the proper functioning of circadian networks under the wide range of temperatures occurring throughout the seasons. Although our knowledge of the transcription factors that regulate clock function in different organisms has increased drastically over the last two decades, the splicing factors that modulate the AS patterns of core clock genes are only starting to be characterized (1). Splicing factors that mediate the effects of temperature on the AS of core clock genes are unknown.

Pre-mRNA splicing is catalyzed by the spliceosome, a large and dynamic molecular complex composed of five different small nuclear ribonucleoprotein (snRNP) particles (U1, U2, U4, U5, and U6 snRNPs) and over 150 additional proteins (9). Each snRNP consists of a specific small nuclear RNA and a number of core spliceosomal proteins. The regulation of AS has traditionally been associated with auxiliary splicing factors such as arginine–serine-rich (RS) proteins and heterogeneous nuclear ribonucleoproteins (hnRNPs), which repress or enhance the recruitment of snRNP particles to specific splice sites. More recently, interactions between the transcriptional machinery, chromatin structure, and core spliceosomal factors have also been shown to regulate AS (10). Furthermore, a systems-based analysis of the network of proteins that interact to regulate AS in mammalian cells suggested that the efficiency and/or kinetics of spliceosome assembly play a key role in the regulation of AS (11).

To investigate if modulation of spliceosome assembly links the regulation of AS to the control of circadian networks in plants, we characterized mutants with defects in genes encoding the main components of the survival motor neuron (SMN) complex, which controls the spliceosomal snRNP core assembly in eukaryotes (12–14). We found that GEMIN2, the only component of the SMN complex that is conserved from yeast to humans, controls the pace of the circadian clock under standard growth conditions in Arabidopsis by controlling the AS of TOC1 and other core clock genes. Furthermore, our results suggest that GEMIN2 attenuates the effects of temperature on the circadian period by modulating AS events associated with several core clock genes, most likely altering the overall balance required for proper temperature compensation of the clock.

Results

Arabidopsis GEMIN2 Is Required for Proper Biological Timing.

An evolutionary analysis of the SMN complex suggested that plants possess orthologs of both SMN and GEMIN2 (12–14). By conducting a more detailed phylogenetic analysis, we found that the Arabidopsis gene AT1G54380 is indeed an ortholog of mammalian GEMIN2 (Fig. S1A), but that the Arabidopsis SMN-like gene, AT2G02570, is more closely related to mammalian SPF30 than to SMN (Fig. S1B). Therefore, similar to Saccharomyces cerevisiae (budding yeast) (12), Arabidopsis lacks a true SMN ortholog and GEMIN2 is the only component of the mammalian SMN complex that is conserved from yeast to humans (9). GEMIN2 is essential for viability in all multicellular organisms characterized to date (15, 16), except for Arabidopsis (Fig. 1). Two different mutant alleles of GEMIN2 showed similarly mild growth and developmental alterations in Arabidopsis plants grown at 22 °C under long-day conditions (16 h light:8 h darkness), such as shorter petioles and smaller leaves, and these phenotypes disappeared when the mutant was complemented with a functional GEMIN2 gene (Fig. 1 A and B). We then characterized clock-dependent phenotypes in these plants. Flowering time is regulated by the circadian clock in Arabidopsis, and we found that both mutant alleles displayed an early flowering phenotype (Fig. 1C). Consistent with a role for GEMIN2 in the regulation of the circadian clock, both mutant alleles displayed a short-period phenotype for the circadian rhythms of leaf movement and gene expression (Fig. 1 D–F and Fig. S2 A and B). GEMIN2 expression cycled in wild-type plants under light/dark cycles, but circadian oscillations in GEMIN2 mRNA levels were not robust in plants transferred to constant light and temperature conditions, indicating that GEMIN2 is not a core component of the transcriptional feedback loops controlling clock function in Arabidopsis (Fig. S2 C and D). The temporal pattern of expression of all of the clock genes analyzed was consistent with the short-period phenotype of the mutant (i.e., the time of peak expression progressively advanced under constant light and temperature conditions), but overall mRNA levels of these clock genes were not altered in gemin2 mutants compared with wild-type plants, suggesting that the defects leading to period alterations most likely resulted from alterations at the posttranscriptional level (Fig. S2 E–H).

Fig. S1.

GEMIN2 is the main component of the SMN complex in Arabidopsis thaliana. (A) Multiple sequence alignment of GEMIN2 orthologs from diverse organisms. The sequence alignment was performed using the Praline program (www.ibi.vu.nl/programs/pralinewww/). Conservation of sequence is represented based on the BLOSUM score matrix. Partially conserved residues (conservation score 4–7) are highlighted as light gray; highly conserved residues (conservation score 8–10) are highlighted as dark gray. (B) Cladogram displaying Smn and Spf30 orthologs from diverse species, and homologs present in plants. Percentage bootstrap values are presented for each node.

Fig. 1.

A role for GEMIN2 in growth, development, and the circadian system of Arabidopsis. (A) Structure of GEMIN2. Introns are represented by lines and exons by boxes (white boxes indicate coding regions and the gray box represents the UTR). T-DNA insertions in gemin2-1 and gemin2-2 mutants are indicated. (B) Rosette phenotypes of wild-type (WT), gemin2-1 and gemin2-2 plants, and of the gemin2-1 mutant complemented with GEMIN2. (C) Flowering time measured as the number of rosette leaves at bolting in constant light (LL), long day (LD), and short day (SD) conditions (ANOVA and Tukey’s multiple comparison test, ***P < 0.001, n = 40–45). (D) Circadian rhythm of leaf movement (n = 8). (E and F) Bioluminescence analysis of CCR2::LUC (E) or TOC1::LUC (F) expression (n = 12). In D–F, plants were entrained under LD cycles and then transferred to constant light and temperature conditions (22 °C). Data represent average + SEM. Open and hatched boxes indicate subjective day and night, respectively.

Fig. S2.

GEMIN2 mutation shortens the period of multiple circadian rhythms. (A) Free-running period of leaf movement under LL, n = 8. (B) Free-running period of TOC1 and CCR2 expression under LL, n = 12. Data represent average + SEM. Period estimates were calculated with Brass 3.0 software (Biological Rhythms Analysis Software System); available from (millar.bio.ed.ac.uk/) and analyzed with the FFT-NLLS suite of programs, as described previously (see Materials and Methods). ANOVA test followed by Tukey's multiple comparison test were applied for comparisons (***P < 0.001). (C) GEMIN2 expression levels determined by RNA-seq analysis in plants grown under long days (LD; 16 h light:8 h darkness). Data were obtained from (39). (D) GEMIN2 expression measured by qPCR in plants grown under LL after entrainment under 12 h light:12 h darkness. Values (n = 3) are relative to PP2A and represent average + SEM. Open and closed boxes indicate light and dark periods, respectively. Lined boxes indicate subjective night. CCA1 (E), LHY (F), PRR9 (G) and PRR7 (H) expression measured by qPCR in seedlings kept under LL after entrainment under 12 h light:12 h darkness, n = 3. Values are expressed relative to PP2A and normalized to the maximum value of each gene. Data represent average + SEM.

Mutations in GEMIN2 Affect snRNP Levels and a Specific Subset of Splicing Events.

GEMIN2 is known to modulate the assembly of U1–U5 snRNPs (9, 13, 14, 17), each of which is composed of a common heptameric ring of seven Sm proteins, a specific snRNA, and several specific accessory factors. Defects in snRNP assembly destabilize uridine-rich snRNAs and, consistent with a role for Arabidopsis GEMIN2 in snRNP assembly, gemin2 mutants had severely reduced levels of U1 snRNA. We also observed slightly increased levels of U2, U4, and U5 snRNAs, as well as reduced levels of U6 snRNA (Fig. 2A). Similar alterations in snRNA stoichiometry have been reported in SMN-deficient mammalian cell lines, which correlate with tissue-specific defects in pre-mRNA splicing of a subset of splicing events in SMN-deficient mice (18).

Fig. 2.

GEMIN2 affects a specific subset of alternative splicing events, including retention of intron 4 of the clock gene TOC1. (A) Levels of snRNAs measured by qPCR (ANOVA and Tukey’s multiple comparison test, ***P < 0.001, **P < 0.01, *P < 0.05, n = 3). (B) Percentage of introns showing increased retention in gemin2-1 mutants relative to wild-type plants among alternatively or “constitutively” spliced introns. (C) Relative frequency of each type of annotated AS event among those present in the RNA-seq dataset of wild-type plants (total) or among those AS events altered in gemin2-1. The representation factor (RF) is the frequency of the event type among those altered in gemin2-1 divided by its frequency among all of the events evaluated in wild-type plants. (D) Read density map of TOC1 visualized with the Integrated Genome Browser. Intron 4 is highlighted with a box. (E and F) Circadian rhythm of leaf movement in constant light (LL) at 22 °C (E) and estimated circadian periods (ANOVA and Tukey’s multiple comparison test, ***P < 0.001, n = 8) (F). Data represent average + SEM.

We then conducted a genome-wide characterization of GEMIN2 effects on alternative and constitutive splicing, combining different experimental approaches such as tiling arrays, high-resolution RT-PCR panels, and RNA sequencing (RNA-seq) (Datasets S1–S3). We observed that GEMIN2 had a greater effect on alternative than on constitutive splicing, a phenomenon that has already been reported for several core snRNP components, such as SmB (19), LSm4 (20), or U1C (21) (Fig. 2B). Defects in AS were enriched for intron retention events, the most common AS event in Arabidopsis, but also included changes in exon skipping, as well as alternative donor and acceptor splice sites (Fig. 2C, Fig. S3, and Datasets S1–S3).

Fig. S3.

Genome-wide effects of GEMIN2 mutation on alternative splicing. (A) Alternative splicing events altered in gemin2-1 mutants, visualized with the Integrated Genome Browser (IGB). Green boxes indicate the event of interest. (B) Validation of affected alternative splicing events by conventional RT-PCR. PIR, percent intron retention, defined in (40); PSI, percent spliced in, defined in (41).

GEMIN2 Defects Alter the AS of Core Clock Genes.

Among the AS events detected with all three techniques, we found that gemin2 mutants displayed increased retention of the alternatively spliced intron 4 of the core clock gene TOC1 (Fig. 2D and Fig. S4A). Plants with mutations in the TOC1 gene have a short period phenotype (22), so we tested whether the gemin2 phenotype could be partially explained by altered levels of TOC1 functional mRNA by analyzing the effect of GEMIN2 on clock function in a null toc1 mutant background. Indeed, we found that, whereas gemin2 mutants showed a short-period phenotype when the mutation was present in an otherwise wild-type background, no additional effect on circadian rhythms was observed when gemin2 was in a toc1 mutant background (Fig. 2 E and F and Fig. S4 B and C). Thus, the gemin2 circadian phenotype requires functional TOC1 mRNA, suggesting that the aberrant phenotype could be due, at least in part, to increased retention of intron 4 of TOC1.

Fig. S4.

gemin2-1 mutants show alterations in alternative splicing of several clock genes. (A) The alteration in TOC1 alternative splicing confirmed by three independent techniques: tiling arrays, high resolution RT-PCR (HR RT-PCR), and RNA-seq. Data are normalized to the value determined for wild type. (B and H) Circadian rhythm of CCR2::LUC expression in constant light (LL) at 22 °C. (C and I) Estimated periods (ANOVA test followed by Tukey's multiple comparison test, *P < 0.05, ***P < 0.001, n = 6–11). Data represent average + SEM. Open and lined boxes indicate subjective day and night, respectively. (D–G) Read density map of TOC1, CCA1, ELF3, and PRR9 visualized with the IGB. Green boxes indicate the event of interest.

In addition to alterations in the AS of TOC1, we also found alterations in the AS of CCA1 and other core clock genes in gemin2, some of which might also contribute to the circadian phenotype of this mutant (Fig. S4 E–G). Indeed, we found that gemin2 enhances the retention of intron 6 of pseudo-response regulator 9 (PRR9), an alteration that should cause period lengthening. Consistent with this possibility, we found that the period-shortening effect of the gemin2 mutation was enhanced in the prr9 prr7 mutant background (Fig. S4 H and I), suggesting that the gemin2 circadian phenotype is most likely the result of integrating alterations in the splicing of several core clock genes.

Similar AS Patterns Are Observed in gemin2 Mutants Grown at 22 °C and in Wild-Type Plants Exposed to Cold Conditions.

Strikingly, it was previously reported that both retention of TOC1 intron 4 and exclusion of CCA1 intron 4 are AS patterns promoted by low temperature conditions in wild-type plants (5, 6) (Fig. S5). We therefore examined in more detail the overlap between GEMIN2 and the effects of low temperature on AS. For this examination, we first visually compared the effects of gemin2 on AS with those found in a publicly available RNA-seq dataset of plants exposed for 4 d to low temperature conditions (23). We found several AS events, including intron retention (IR), exon skipping (ES) and alternative 5′ splice sites (A5′SS), that were similarly affected by gemin2 at 22 °C and by low temperature conditions in wild-type plants (Fig. 3 A and B). Lack of replicates of this particular low temperature transcriptome prompted us to obtain an independent dataset with three biological replicates. For this purpose, we compared the transcriptome of plants grown for 9 d at 22 °C with that of plants grown at 22 °C for the 9 d and exposed at the end of this period for 1 or 24 h to 10 °C. Approximately 80% of a set of genes previously reported to be either induced or repressed in response to cold conditions in Arabidopsis also showed large and significant changes in expression in our experiment, validating our cold RNA-seq dataset (Fig. S6). We found a strong enrichment in genes associated with ribosome biogenesis, protein translation, and RNA processing and splicing among the genes up-regulated in response to cold conditions in wild-type plants (Datasets S4 and S5), supporting the notion that transcriptional regulation of the RNA processing machinery is an important component of the cold acclimation mechanism in plants. Interestingly, we also observed a strong overlap between the effects of cold conditions on pre-mRNA splicing in wild-type plants and the effects of GEMIN2 on pre-mRNA splicing at 22 °C, an overlap that was much larger than that expected to occur simply by chance (Fig. 3C). In particular, this extensive overlap was more significant for pre-mRNA splicing events affected in similar directions (i.e., increased inclusion or exclusion) by the mutation in GEMIN2 and exposure to cold stress (Fig. 3C).

Fig. S5.

Altered alternative splicing of core clock genes in gemin2-1 mutant resembles the effect of cold in wild-type plants. Read density map of TOC1 and CCA1 visualized with the IGB. Control and cold correspond to public RNA-seq datasets described and available at bioviz.org/quickload/A_thaliana_Jun_2009/cold_stress/. Green boxes indicate the event of interest.

Fig. 3.

Alternative splicing patterns of gemin2 mutants at 22 °C mimic those of wild-type plants exposed to low temperature conditions. (A) AS events similarly affected by a mutation in GEMIN2 at 22 °C and by exposure to 4 °C for 48 h in wild-type plants, visualized with the Integrated Genome Browser. Control and cold correspond to public RNA-seq datasets described in Gulledge et al. (23). Black boxes highlight the events of interest. (B) Affected events confirmed by RT-PCR. Cold-treated plants were subjected to 10 °C for 12 h. Space between lanes denotes that data were collected from noncontiguous lanes of the same gel. (C) Venn diagram showing the overlap between pre-mRNA splicing events affected in gemin2-1 mutants at 22 °C and those altered in wild-type plants exposed to 10 °C for 1 or 24 h.

Fig. S6.

Genes responsive to cold determined by RNA-seq overlap significantly with genes previously reported to be modulated by cold. Venn diagram showing genes induced or repressed by cold in wild-type plants as measured by RNA-seq and those shown to be modulated by cold in array expermients in (42).

GEMIN2 Maintains U1 snRNP Homeostasis and Buffers Low Temperature Effects on AS.

Based on the extensive overlap between pre-mRNA splicing events affected by low temperature conditions in wild-type plants and those affected in gemin2 at 22 °C, we hypothesized that cold exposure would inhibit U1 snRNP assembly. However, we found that U1 snRNA levels increased in wild-type plants exposed for 24 h to low temperatures (Fig. 4A). This finding is opposite to that expected if U1 snRNP assembly were compromised, because free U1 snRNA is unstable. Alternatively, cold conditions may reduce the functionality, rather than the levels, of U1 snRNP, and the increment observed in U1 snRNA levels may be required for a compensatory enhancement of U1 snRNP assembly. Interestingly, we found that a 24-h exposure to low temperatures increased GEMIN2 expression by almost 50%, suggesting that this effect might be associated with the cold-mediated enhancement of U1 snRNA levels (Fig. 4B). In agreement with this possibility, U1 snRNA levels did not accumulate to higher levels in response to low temperatures in gemin2 mutants.

Fig. 4.

GEMIN2 modulates low temperature effects on U1 snRNA levels, alternative splicing, circadian rhythm, and survival in Arabidopsis. (A) Levels of U1 snRNA measured by qPCR in wild-type and gemin2 mutant plants exposed to cold conditions for 1 or 24 h. Statistically significant differences are indicated by different lowercase letters (ANOVA and Tukey’s multiple comparison test, P < 0.001). (B) GEMIN2 expression determined by RNA-seq in wild-type plants. In A and B, data represent average + SEM (n = 3) and were normalized to the value present in wild-type plants at 0 h. (C) Read density (RD) map of U1-70K visualized with the Integrated Genome Browser. The alternatively spliced intron is highlighted with a box. (D) Relative expression (RD intron/RD gene) of the alternatively spliced intron from U1-70K. Data were normalized to the value observed in wild-type plants at the beginning of the experiment. (E) Response patterns of the two largest clusters of pre-mRNA splicing events whose relative abundance (RD alternatively spliced region/RD gene) was altered in response to cold exposures in gemin2-1 plants. Thin lines represent normalized values for individual alternatively spliced regions. Thick lines represent the average. The numbers appearing next to the cluster number represent the total number of splicing events present in that particular cluster. (F) Free-running periods of CCR2::LUC expression under constant light at different temperatures. Data represent the average of three independent experiments + SEM. (G) Five-day-old seedlings transferred to 10 °C for 4 wk.

Evidence suggesting reduced U1 snRNP functionality under cold conditions in wild-type plants was provided by the response to low temperatures of U1-70K, a key component of the U1 snRNP. Similar to recent findings in mammals, U1-70K mRNA is alternatively spliced in Arabidopsis, with one isoform encoding a functional U1-70K protein and another encoding a truncated protein (24). In mammals, compromising U1 snRNP function by interfering with another component of the U1 snRNP complex results in a compensatory increment in the functional U1-70K mRNA isoform (25). A strikingly similar situation is observed in gemin2 mutant plants defective in U1 snRNP assembly, which showed a dramatic shift in AS, leading to the complete depletion of the nonfunctional mRNA and to full accumulation of the functional U1-70K mRNA isoform (Fig. 4C). A 24-h exposure to low temperature conditions reduced the levels of the nonfunctional U1-70K mRNA isoform in wild-type plants (Fig. 4D). Together, these results support the idea that U1 snRNP functionality is negatively affected by low temperature conditions in wild-type plants, and that GEMIN2 compensates for this defect by enhancing U1 snRNP assembly.

To further characterize the connection between GEMIN2 and AS responses to low temperatures in Arabidopsis, we conducted a comparative genome-wide analysis of the effect of cold on the transcriptome of wild-type and gemin2 mutant plants. In agreement with the hypothesis that GEMIN2 attenuates the effects of low temperatures on pre-mRNA splicing, we found that most differences in splicing between gemin2 and wild-type plants were observed in plants exposed to cold conditions for 24 h (Datasets S6 and S7). Interestingly, most of the pre-mRNA splicing events that were strongly responsive to cold exposure in gemin2 were almost nonresponsive to low temperature in wild-type plants, indicating that GEMIN2 indeed mostly acts to buffer the effects of cold conditions on pre-mRNA splicing (Fig. 4E and Fig. S7). In addition, we found that most of the splicing events affected by cold exposure in wild-type plants showed a significantly altered response to low temperatures in gemin2 mutants, indicating that GEMIN2 is a strong modulator of the effects of cold conditions for a large subset of pre-mRNA splicing events (Fig. S8).

Fig. S7.

Clusters constructed from cold-sensitive splicing events in gemin2-1 mutants. Thin lines represent normalized values for individual splicing events. Thick lines represent the average. 1, 2, and 3 in ordinate indicate 0, 1, and 24 h at 10 °C for wild type; 4, 5, and 6 indicate 0, 1, and 24 h at 10 °C for gemin2-1 mutant. The numbers appearing next to the cluster number represent the total number of splicing events present in that particular cluster.

Fig. S8.

Clusters constructed from cold-sensitive splicing events in wild type. Thin lines represent normalized values for individual splicing events. Thick lines represent the average. 1, 2, and 3 in ordinate indicate 0, 1, and 24 h at 10 °C for wild type; 4, 5, and 6 indicate 0, 1, and 24 h at 10 °C for gemin2-1 mutant. The numbers appearing next to the cluster number represent the total number of splicing events present in that particular cluster.

GEMIN2 Is Required for the Acclimation of Physiological Processes to Low Temperatures.

Finally, we tested whether the GEMIN2 and temperature-dependent effects on pre-mRNA splicing were associated with alterations in the temperature-mediated regulation of different physiological processes. Consistent with this idea, we found that the effects of temperature on circadian period were strongly enhanced in gemin2 mutants, resulting in a disruption of temperature compensation (Fig. 4F). This defect could be due to temperature-dependent alterations in the overall balance of core clock components such as PRR9, PRR7, and CCA1, some of which undergo temperature-dependent changes in AS (5) and have been implicated in the temperature compensation of the Arabidopsis clock (26) (Fig. S9A). In addition, we observed that the survival rate of gemin2 mutants, but not of wild-type plants, was reduced after keeping the plants at 10 °C for 4 wk (Fig. 4G). This reduction in survival was most likely due to severe alterations in the AS of a large subset of events, which resulted in almost full retention of several introns (Fig. S10).

Fig. S9.

Circadian period of gemin2-1 mutants is determined by temperature-dependent and -independent alterations in alternative splicing of core clock genes. (A) Read density map of TOC1, CCA1, PRR7, PRR9, ELF3, and RVE8 visualized with IGB. Green boxes indicate the event of interest. Relative expression (RD intron/RD gene) of the alternatively spliced intron is shown. Data represent average + SEM. 22 °C, control plants always kept at 22 °C; 10 °C, plants grown at 22 °C and exposed to 10 °C for 24 h. (B) Model depicting the result of additive temperature compensation (TC) defects and temperature-independent period defects present in gemin2-1 mutants. Disruption of the temperature compensation property of the circadian system leads to a slower running clock at low temperatures in the mutant compared with the wild-type (Left). Additionally, a period defect fastens the pace of the clock in a temperature-independent manner (Middle). As a result of both defects, the difference in circadian period between gemin2 mutant and wild-type plants is larger at higher temperatures, decreases with temperature, and is expected to disappear at the lowest temperature conditions.

Fig. S10.

A set of alternative splicing events is completely impaired in gemin2-1 mutants exposed to cold. Alternative splicing events are visualized with IGB. Green boxes indicate the event of interest.

Discussion

Posttranscriptional mechanisms, including AS, control circadian rhythms across eukaryotic organisms, and the factors linking these regulatory networks are starting to be identified. We previously showed that PRMT5, an arginine methyltransferase that methylates Sm and LSm spliceosomal proteins, is important for the proper regulation of circadian rhythms in Arabidopsis and flies (27). Other splicing factors known to control clock function in plants are LSm4 and LSm5 (core components of the U6 snRNP) (20) and STIPL (28), an Arabidopsis protein with homology to a splicing factor involved in spliceosome disassembly in humans and yeast. In addition, the Arabidopsis homolog of the mammalian SKI interacting protein (SKIP), a splicing factor present in the spliceosomal NineTeen complex, also regulates the circadian period length in plants (29). For all these genes encoding splicing factors or regulators, the circadian phenotypes exhibited by the corresponding mutants are associated with alterations in specific subsets of AS events involving a few core clock genes, rather than with global defects in pre-mRNA splicing. Interestingly, most of these AS events are affected by temperature changes, and some are important for keeping the clock running at a proper pace at different temperatures. However, the splicing factors or regulators that modulate the effects of temperature on the AS of core clock genes have not been identified in any eukaryotic organism. Here we showed that GEMIN2, a spliceosomal snRNP assembly factor that is conserved from yeast to humans, controls several AS events in Arabidopsis plants grown at warm temperatures and modulates the effects of cold temperatures on a large subset of AS events, including many associated with core clock genes. These molecular events are tightly linked to the physiological role we identified for GEMIN2 in regulating the circadian period at warm temperatures, as well as in buffering the effects of temperature on the pace of the circadian clock in Arabidopsis. GEMIN2 is also essential for acclimation and survival under extended exposure to mild cold conditions.

Recent work suggests that changes in the level and/or activity of previously considered core spliceosomal proteins may influence AS by altering the kinetics and/or order of recruitment of core spliceosomal proteins during spliceosome assembly (11). In agreement with the idea that spliceosome assembly plays a critical role in regulating AS, depletion or inactivation of SMN in mammals results predominantly in defects in AS rather than in constitutive pre-mRNA splicing (18). Whereas initial studies on the regulation of snRNP biogenesis focused mainly on the role of SMN, recent crystallographic studies suggest that GEMIN2 has a more prominent role in controlling this process (9). Interestingly, we found that Arabidopsis lacks a true ortholog of SMN and, therefore, GEMIN2 is the only member of the SMN complex that is present in all organisms from yeast to humans. We found that gemin2 Arabidopsis mutants are viable and fertile, strongly suggesting that neither SMN nor GEMIN2 are essential under normal growth conditions. We did observe, however, that gemin2 has reduced levels of specific snRNAs, in particular U1, which strongly suggests that GEMIN2 is important for the assembly of normal levels of U1 snRNP, the most abundant of the snRNP particles. This defect in snRNP assembly correlates with significant defects in the pre-mRNA splicing of a number of genes in plants grown at 22 °C. Splicing defects are more frequent among AS events than among constitutively spliced introns, most of which are fully spliced, similar to what has been reported for SMN in mammals (18). Some of the AS events affected in gemin2 mutants are associated with core clock genes such as TOC1 and PRR9 and, indeed, the circadian period of gemin2 mutant results from the overall balance of its effects on several clock genes. Thus, GEMIN2 controls the assembly of some snRNP particles under normal growth conditions and thereby modulates several AS events and physiological processes.

Interestingly, the AS patterns of the TOC1 and CCA1 core clock genes in gemin2 mutants resemble those observed for these genes in wild-type plants under low temperature conditions. Furthermore, global RNA-seq analysis revealed a large overlap between the AS defects present in gemin2 plants grown at 22 °C and the AS changes induced by cold temperatures in wild-type plants. This overlap was not caused by reduced assembly of U1 snRNP under cold conditions in wild-type plants, because we observed increased rather than reduced U1 snRNA levels after 24 h at 10 °C. These observations suggest that U1 snRNP functionality might be limiting at low temperatures and plants may compensate for this by enhancing U1 snRNA synthesis and snRNP assembly.

Two different observations support the idea that spliceosomal function may be impaired at low temperatures in wild-type plants. First, we found that the changes in the AS of U1-70K in wild-type plants exposed to cold conditions resembled those found in mammalian cells with reduced functional U1 snRNP levels (25), as well as in gemin2 mutant plants, which are impaired in U1 snRNP assembly. Indeed, in both mammals and Arabidopsis plants, the AS of U1-70K mRNA, a core component of the U1 snRNP, results in two different mRNAs, one encoding a fully functional protein and the other introducing a premature termination codon that targets the mRNA to NMD pathway. Depleting the levels of another U1 snRNP component in mammalian cells, U1C, triggers compensatory increases in the levels of the functional U1-70K mRNA isoform that result in increased U1-70K protein levels (25). Given the conservation of this homeostatic mechanism from plants to mammals, we expect that the increased levels of the functional U1-70K mRNA isoform observed in wild type plant at low temperatures will be associated with higher levels of U1-70K protein. Taken together, these observations suggest that wild-type plants exposed to cold conditions have limiting levels of functional U1 snRNPs and react to this deficit with changes in AS, leading to higher levels of the functional U1-70K isoform, which, together with the increased levels of U1 snRNA, result in higher levels of U1 snRNP particles. Second, the expression of several RNA processing factors, including some involved in pre-mRNA splicing, was strongly enhanced in plants exposed to cold conditions for 24 h. Similar observations have been made in zebrafish, where exposure to cold affects the expression and splicing of many genes, including core clock genes, and is associated with an enrichment in genes encoding spliceosomal proteins (30). Therefore, several parallel mechanisms appear to operate simultaneously to enhance spliceosomal activity under cold conditions and we propose that GEMIN2 is a component of one of these compensatory mechanisms. Indeed, in the absence of functional GEMIN2, cold conditions affect the AS pattern of many genes that do not normally show changes in AS patterns in response to cold conditions in wild-type plants, and gemin2 mutants perish after being exposed to cold conditions for several weeks, whereas wild-type plants do not. Consistent with the idea that a key role of GEMIN2 is to enhance U1 snRNP assembly under cold conditions in eukaryotic organisms, the GEMIN2 ortholog in yeast, known as Bad Refrigeration Response 1 (BRR1), was originally identified in a genetic screen for cold sensitive pre-mRNA splicing mutants (31), and later shown to encode a key regulator of U1 snRNP assembly (32).

The period of circadian rhythms is temperature compensated, i.e., shows only modest changes over a broad range of temperatures, allowing circadian clocks to function as time measuring devices throughout the year (33). By contrast, temperature compensation is disrupted in the gemin2 mutant, resulting in significant period lengthening in response to cold conditions. Strikingly, GEMIN2 has similar effects on the AS of the clock gene TOC1 at both 22 °C and 10 °C (Fig. S9A). If defective AS of TOC1 were the main mechanism through which GEMIN2 affected the clock, the short period phenotype of gemin2 at 22 °C would also be observed at 12 °C. The finding that the circadian period is similar in gemin2 and wild-type plants at 12 °C strongly suggests that, at low temperatures, gemin2 affects the AS of other clock genes, which result in period-lengthening effects that balance the period-shortening effect associated with TOC1 intron 4 retention (Fig. 4F and Fig. S9B). Indeed, increased intron retention in PRR9 contributes to attenuate the period-shortening effect of gemin2 at 22 °C (Fig. S4). However, PRR9 and PRR7 are unlikely to be the targets of GEMIN2 that lengthen circadian period at low temperatures because their effects are larger at intermediate and high temperatures, but disappear at low temperatures (26). Thus, the clock genes that lengthen the period of gemin2 at low temperatures remain to be identified. Finally, we do not think that GEMIN2 is part of a specific molecular mechanism that evolved to modulate the effects of temperature on the circadian clock. Recent evidence indicates that temperature compensation in plants depends on the overall balance of temperature-dependent period-lengthening and -shortening effects (34). Here we propose that GEMIN2 acts as a global modulator of AS, particularly under cold conditions, and temperature compensation depends on GEMIN2 function because, in its absence, low temperatures drastically alter the AS of several clock genes, disrupting the proper balance of period-shortening and -lengthening effects.

Most organisms living on this planet are poikilothermic, i.e., they do not control their own body temperature, and buffering biological processes from daily and seasonal fluctuations in ambient temperature is essential for their survival. Splicing involves extensive remodeling of the interactions between spliceosomal snRNAs and pre-mRNAs (9). Low temperatures are expected to enhance these interactions, reducing the speed of the required rearrangements, and poikilothermic organisms must have mechanisms to compensate for the detrimental consequences of these effects. Our results indicate that GEMIN2 is a key component of one such mechanism that modulates the effects of low temperature on pre-mRNA splicing, helping plants to keep time accurately and survive the cold weather conditions they may face at different times of the day and year. These results are also consistent with the hypothesis that spliceosomal interactions with pre-mRNAs and the capacity to generate transcript variants may act as a thermometer that allows plants to adjust to changes in ambient temperature (35).

Materials and Methods

Plant Material.

All of the Arabidopsis lines used in this work were of the Columbia (Col-0) accession. The gemin2-1 (SALK_142993) and gemin2-2 (SAIL_567_D05) mutants were obtained from the Arabidopsis Biological Resource Center. Genotypes were confirmed by PCR using oligonucleotides listed in Dataset S8.

Growth Conditions.

Seeds were stratified for 4 d in the dark at 4 °C and then sown onto either soil or solid Murashige and Skoog medium containing 1% agarose. Seedlings were grown under different temperature and light regimes depending on the experiment.

Physiological Measurements.

Detailed information is in SI Materials and Methods.

qRT–PCR and RNA-Seq Analysis.

Detailed information is in SI Materials and Methods.

Full methods and any associated references are available in the SI Materials and Methods.

SI Materials and Methods

Plant Materials and Growth Conditions.

All of the Arabidopsis lines used in this work were of the Columbia (Col-0) accession. The gemin2-1 (SALK_142993), gemin2-2 (SAIL_567_D05), prr7 (prr7-3, Salk 03430), and prr9 (prr9-1, Salk 007551) mutants were obtained from the Arabidopsis Biological Resource Center (36). toc1-101 mutant was kindly provided by Peter Quail, University of California, Berkeley, Albany, CA. Seeds carrying the bioluminescence reporters were provided by Stacey Harmer, University of California, Davis, CA. Genotypes were confirmed by PCR using oligonucleotides listed in Dataset S8. Seeds were stratified for 4 d in the dark at 4 °C and then sown onto either soil or solid Murashige and Skoog medium containing 1% agarose. Seedlings were grown under different temperature and light regimes depending on the experiment.

Phylogenetic Analysis.

Homologs of human GEMIN2 were identified using BLASTP (www.ncbi.nlm.nih.gov/ and phytozome.jgi.doe.gov/pz/portal.html). Protein sequences were aligned using Praline (www.ibi.vu.nl/programs/pralinewww/) or Clustal Omega (www.ebi.ac.uk/Tools/msa/clustalo/). Maximum likelihood phylogenetic trees were built using SeaView Version 4 (available from doua.prabi.fr/software/seaview), with 1,000 bootstrap replicates.

Complementation of the gemin2-1 Mutant.

GEMIN2 (At1g54380) coding sequence, obtained from the Arabidopsis Information Resource (TAIR10), was synthesized de novo by GeneScript Corporation and then cloned into pDONR/Zeo (Invitrogen) through Gateway technology (Invitrogen). A binary vector containing this fragment was generated by LR recombination with destination vector pB7WGY2 (37). The constructs was transferred into Agrobacterium tumefaciens (strain GV3101), and gemin2-1 plants were transformed by the floral dip method.

Flowering Time Analysis.

Flowering time was estimated by counting the number of rosette leaves at the time of bolting. The experiments were performed in continuous light, long days (16 h light/8 h dark) and short days (8 h light/16 h dark) at a constant temperature of 22 °C. An ANOVA followed by Tukey’s multiple comparison test were used for comparisons among genotypes.

Circadian Rhythm Analysis.

For leaf movement analysis, plants were entrained in a 12:12 h light–dark cycle and then transferred to continuous white fluorescent light. The position of the first pair of leaves was recorded every 2 h for 6–7 d. For bioluminescence assays, wild-type plants and gemin2-1 mutants carrying the CCR2::LUC or TOC1::LUC reporter were grown in a 12:12 h light–dark cycle and transferred to continuous white fluorescent light for 7–9 d. Bioluminescence rhythms were detected with a microplate luminometer LB-960 (Berthold Technologies). Circadian period estimates were calculated with Brass 3.0 software (Biological Rhythms Analysis Software System, available from millar.bio.ed.ac.uk/) and analyzed with the FFT-NLLS suite of programs (22). An ANOVA followed by Tukey’s Multiple Comparison Test was used for comparisons among genotypes.

qRT-PCR Expression Analysis.

Total RNA was obtained from samples using TRIzol reagent (Invitrogen). One microgram of RNA was subjected to a DNase treatment with RQ1 RNase-Free DNase (Promega). cDNA derived from this RNA was synthesized using M-MLV (Invitrogen) and oligo-dT primer. To quantify snRNAs, cDNA synthesis was performed using random hexamers. The synthesized cDNAs were amplified with FastStart Universal SYBR Green Master (Roche) using the Mx3000P Real Time PCR System (Agilent Technologies) cycler. PP2A (AT1G13320) transcript was used as a normalization control. One-way ANOVA followed by Tukey’s multiple comparison test was used for comparisons of snRNA levels among genotypes. Primers used to amplify each gene are listed in Dataset S8.

High-Resolution RT-PCR Panels.

Plants were grown under continuous light for 3 wk without prior light or temperature entrainment and RNA samples were obtained as described earlier. First-strand cDNA synthesis, PCR protocol, and splicing analysis were conducted as previously described (27).

Whole Genome Tiling Array Analysis.

Plants were grown under continuous light for 3 wk without prior light or temperature entrainment. GeneChip Arabidopsis Tiling 1.0R Array (Affymetrix) was processed according to standard Affymetrix protocols. Splicing analysis was performed as previously described (27). Log transformed intensity data from wild types and mutants were fitted with a linear model: intensity = genotype + error, where the genotype term contrasts wild types and mutants. False discovery rates (FDRs) were determined by 20 permutations.

cDNA Library Preparation and High-Throughput Sequencing.

Seeds were sown onto Murashige and Skoog (MS) medium containing 0.8% agarose, stratified for 4 d in the dark at 4 °C, and then grown at 22 °C in continuous light. Whole plants were harvested after 9 d and total RNA was extracted with RNeasy Plant Mini Kit (Qiagen) following the manufacturer’s protocols. To estimate the concentration and quality of samples, NanoDrop 2000c (Thermo Scientific) and gel electrophoresis were used, respectively. Libraries were prepared following the TruSeq RNA Sample Preparation Guide (Illumina). Briefly, 3 μg of total RNA was polyA purified and fragmented, and first-strand cDNA was synthesized by reverse transcriptase (SuperScript II; Invitrogen) and random hexamers. This step was followed by RNA degradation and second-strand cDNA synthesis. End repair process and addition of a single A nucleotide to the 3′ ends allowed ligation of multiple indexing adapters. Then, an enrichment step of 12 cycles of PCR was performed. Library validation included size and purity assessment with the Agilent 2100 Bioanalyzer and the Agilent DNA1000 kit (Agilent Technologies). Samples were pooled to create multiplexed DNA libraries. For the initial wild type and gemin2-1 mutant comparison, samples were single-end sequenced on the Illumina GAIIx platform at Max Planck Institute for Development Biology, Tübingen, Germany. For the cold treatment experiment, samples were paired-end sequenced on the Illumina HiSEq 1500 at INDEAR, Argentina. On average, 16 million 100-nt single-end and 16 million 100-nt paired-end reads for each sample were obtained, respectively. Three replicates for each genotype and condition were sequenced.

Processing of RNA Sequencing Reads.

Sequence reads were mapped to Arabidopsis thaliana TAIR10 genome using TopHat v2.0.9 with default parameters, except for maximum intron length set at 5,000, as previously described (20). Count tables for different feature levels were obtained from bam files using custom R scripts and considering TAIR10 transcriptome.

Differential Gene Expression.

Before differential expression analysis, we decided to discard genes with less than 10 reads on average per condition. Differential gene expression was estimated using edgeR package version 3.4.2 and resulting P values were adjusted using FDR criterion, as previously described (20). Genes with FDR values lower than 0.05 and absolute log2FC greater than 0.585, were deemed as differentially expressed.

Differential Alternative Splicing.

For alternative splicing analysis, the transcriptome was partitioned into subgenic joint features called “bins” as proposed on DEXSeq, as previously described (20). The name of each bin indicates the relative position of the subgenic region analyzed in the forward strand of the corresponding locus. Due to our special interest in new intron retention events not only exons but also introns were considered in our analysis. The transcriptome was partitioned into 281,321 bins. A total of 152,631 corresponded exclusively to exonic regions, 128,700 to intronic regions, and 7,973 to DNA regions directly involved in alternatively spliced isoforms. We labeled these three kinds of bins as exon bin, intron bin or alternative-splicing (AS) bins, respectively. In addition AS bins were further classified as exon skipping (ES), 5′ or 3′ alternative (5′alt, 3′alt), intron retention (IR), or multiple (those including three or more different alternative splicing events in the same subgenic region) bins. For our analysis, we discarded bins from monoexonic genes and with mean count values lower than 5 reads per condition.

To provide a comprehensive summary of the calculated subgene features, separate tables were produced for introns and AS bins. We used edgeR exact test for the identification of differential use of bins corresponding to AS events or introns, and FDR corrected P values. We also computed read densities to have a relationship between the bin and its corresponding gene. A splicing index was calculated as bin read density/gene read density, and the splicing index ratio was calculated as splicing index in test condition/splicing index in control condition. Only genes with read densities greater than 0.05 in both conditions, and splicing indexes greater than 0.05 in at least one condition, were used for the analysis. AS events as well as all introns with an absolute Log2 splicing index ratio (splicing index in test condition/splicing index in control condition) greater than 0.58 and FDR values lower than 0.1, were deemed as differentially spliced.

Alternative Splicing Profile Clustering.

To analyze patterns of coordinated alternative splicing behavior, we performed a cluster analysis of exonic bins considering a correlation-based distance of bin read-density profiles (normalized over the corresponding gene read-density values). The cluster detection was performed in a two-stage procedure using an adaptive branch pruning of hierarchical clustering dendrograms, as implemented in the “dynamicTreeCut” R package (38). Preclusters were obtained in a first clustering round, aimed to find a rather low resolution partition by setting the deepSplit parameter to zero. Then, final clusters were obtained after merging similar preclusters. This last step involved a second clusterization performed with the deepSplit parameter set to an intermediate value (i.e., deepSplit = 2). We found that this two-stage clustering procedure resulted in a sensible partition of the original dataset amenable to further analysis and characterization.

GO Enrichment Analysis.

Functional categories statistically overrepresented in particular lists of genes compared with the entire genome were identified using the Singular Enrichment Analysis tool from AgriGO website (bioinfo.cau.edu.cn/agriGO/analysis.php). P values were calculated using the Fisher exact test (with FDR correction).

Chilling Tolerance Assay.

Seeds of relevant genotypes were germinated on MS agar medium and grown vertically for 5 d in continuous light at 22 °C. Then, the plates were transferred to 10 °C and grown vertically for 4 wk.

Statistical Analysis.

P values in Figs. 2 and 3 were calculated using the hypergeometric test.