Comparative Analysis of Circadian Transcriptomes Reveals Circadian Characteristics between Arabidopsis and Soybean

Abstract

The circadian clock, an endogenous timing system, exists in nearly all organisms on Earth. The plant circadian clock has been found to be intricately linked with various essential biological activities. Extensive studies of the plant circadian clock have yielded valuable applications. However, the distinctions of circadian clocks in two important plant species, Arabidopsis thaliana and Glycine max (soybean), remain largely unexplored. This study endeavors to address this gap by conducting a comprehensive comparison of the circadian transcriptome profiles of Arabidopsis and soybean to uncover their distinct circadian characteristics. Utilizing non-linear regression fitting (COS) integrated with weights, we identified circadian rhythmic genes within both organisms. Through an in-depth exploration of circadian parameters, we unveiled notable differences between Arabidopsis and soybean. Furthermore, our analysis of core circadian clock genes shed light on the distinctions in central oscillators between these two species. Additionally, we observed that the homologous genes of Arabidopsis circadian clock genes in soybean exert a significant influence on the regulation of flowering and maturity of soybean. This phenomenon appears to stem from shifts in circadian parameters within soybean genes. These findings highlight contrasting biological activities under circadian regulation in Arabidopsis and soybean. This study not only underscores the distinctive attributes of these species, but also offers valuable insights for further scrutiny into the soybean circadian clock and its potential applications.

1. Introduction

Residing on Earth, a multitude of organisms have developed a circadian clock system to synchronize their biological activities with the environment for their growth and well-being. These organisms span prokaryotes like cyanobacteria and eukaryotes such as yeast, plants, and animals. The circadian clock within them is pivotal for adaptation to adverse conditions. Particularly in plants, the circadian clock enables them to anticipate environmental shifts like temperature fluctuations and pathogen invasions [1,2,3]. By integrating external signals with internal cues, the plant circadian clock orchestrates diverse biological events, ensuring their occurrence at the optimal times throughout the day [4,5,6]. Given that the circadian clock governs the timing of biological activities throughout a plant’s life, understanding of its function has proven instrumental in enhancing plant vitality and prosperity.

At its core, the plant circadian system comprises central oscillators, input pathways, and output pathways [7,8]. Research on the circadian clock of the model plant Arabidopsis has predominantly focused on the central circadian oscillator and its interplay with input and output pathways [9]. Currently, over twenty clock or clock-associated components have been identified in Arabidopsis [7]. Within this circadian system, the central oscillator encompasses a series of interlocked transcription–translation feedback loops, while the input and output pathways facilitate the transmission of signals between external and internal environments to regulate various processes. Distinct clock proteins become active at specific times of the day, mutually influencing the expression of other clock genes [4,10]. The core interconnected feedback loop of the Arabidopsis circadian clock involves CCA1/LHY and TOC1 [11,12,13,14,15]. CCA1 and LHY peak at dawn, while TOC1’s expression peaks around dusk. CCA1/LHY and TOC1 mutually suppress each other’s expression [16,17]. Following a reduction in CCA1/LHY protein levels post-dawn, the repression on TOC1 subsides, resulting in increased TOC1 protein levels. With elevated TOC1 protein levels, CCA1 and LHY expression are further restrained. The inverse process unfolds during the night. Some clock proteins can also repress their own expression when overexpressed. Furthermore, the clock system incorporates greater complexity, featuring multiple loops and numerous other proteins such as PRR5, ELF3, RVE1 etc. [18,19,20].

The Arabidopsis circadian clock has been extensively investigated. Likewise, research on the circadian clock of various crops has been expanding [21,22]. For instance, it has been reported that OsCCA1 regulates ABA signaling pathways to enhance rice’s abiotic tolerance [23]. TaELF3 can impact the heading dates of wheat through its effects on photoperiodic responses in circadian oscillators [24]. In maize, the evening complex of the circadian clock promotes flowering and adaptation to temperate regions [25]. Furthermore, studies have indicated tight correlations between circadian rhythms and abiotic stress in soybeans [26]. Despite these scattered documentations of the function of the circadian clock system in different crops, systematic comparative studies of circadian rhythms between Arabidopsis and crops remain limited. Given our scarce understanding of crop circadian clocks, such comparisons may produce valuable insights into the circadian clock’s role within crops.

The analysis of expressed genes through transcriptome profiling has proven integral in plant circadian clock research [27,28,29]. Specifically, time-course circadian transcriptome profiling offers a comprehensive approach for identifying a wide array of circadian rhythmic genes [26,30,31]. In this study, we undertook a comparative analysis of circadian transcriptome profiles in Arabidopsis and soybean to unveil their unique circadian attributes. Through the comparison of rhythmic genes and their corresponding circadian parameters, we discovered the distinctive characteristics of Arabidopsis and soybean. Additionally, an examination of core circadian clock genes identified inherent differences in phase, period, and amplitude between the two organisms. Finally, we unveiled that translation activities are more likely under circadian regulation in Arabidopsis, while photosynthesis activities are more prone to circadian control in soybean. These findings provide insights into the soybean circadian clock and establish a foundation for future engineering of the soybean circadian clock to improve the yield of soybean.

2. Results

2.1. Comparison of Circadian Rhythmic Genes Unveils Distinctive Circadian Parameter Characteristics between Arabidopsis and Soybean

We retrieved Arabidopsis and soybean circadian time-course transcriptome profiles from two independent studies which share consistent sampling schemes (Figure 1) [26]. In both studies, plants were initially cultivated in a light–dark cycle (LD) for a duration of 9 to 11 days. Subsequently, the plants were transferred to a continuous light condition (LL). Under the LL condition, timepoints were denoted using zeitgeber time (ZT), starting from ZT0. One day after transferring plants into the LL condition, samples of Arabidopsis and soybean were collected over the course of two consecutive days at 4 h intervals. These samples were then subjected to RNA-seq analysis, and the resulting raw sequencing reads were processed to generate the raw count matrix of gene expression data.

Following the pre-processing of the expression data, a total of 21,013 genes for Arabidopsis and 33,543 genes for soybean were retained and considered expressed genes. The good reproducibility of samples from the same timepoint is evident in the sample correlation heatmap (Figure S1). Notably, samples collected at the 24 h interval exhibit a higher correlation compared to those collected at the 12 h interval (Figure S1). Leveraging non-linear regression fitting (COS) integrated with weights analysis, oscillatory parameters were estimated for all expressed genes. This comprehensive dataset of expressed genes and their corresponding oscillatory parameters served as the foundation for subsequent analyses.

Prior to delving into the classification of circadian rhythmic genes, a thorough examination of the estimates of circadian oscillatory parameters was performed. Firstly, we defined circadian oscillation correlation as an index used to assess rhythm robustness and detect rhythmic genes. Circadian oscillation correlation refers to the correlation coefficient of observed data and predicted data from COS fitting. Higher circadian oscillation correlation indicates better oscillation of genes. The distribution of circadian oscillation correlations among the expressed genes demonstrates an apparent disparity between Arabidopsis and soybean (Figure 2a). Notably, for circadian oscillation correlation exceeding 0.8, Arabidopsis exhibits a higher density compared to soybean. Conversely, the absolute number of rhythmic genes in soybean consistently surpasses that of Arabidopsis across a range of correlation cutoffs (Figure 2b). This outcome is anticipated given the substantially larger number of soybean genes. However, when the proportion of rhythmic genes relative to the total expressed genes was considered, no statistically significant difference was observed between the two organisms (Figure 2c). This suggests that during the recent conversion from the tetraploid to the diploid genome [32], circadian rhythmic genes were not significantly retained or depleted in the soybean genome.

Previous studies have applied a circadian oscillation correlation of no less than 0.7 as the threshold in their research, and this has proven to be a robust cutoff [33,34]. Therefore, we adopted a circadian oscillation correlation of no less than 0.7 as a threshold to detect circadian rhythmic genes. This yielded a percentage of 44.98% and 42.51% of circadian rhythmic genes among the expressed genes of Arabidopsis and soybean, respectively. The yielded number of rhythmic genes are enough for down-stream analysis.

A comprehensive comparison of the circadian rhythmic genes of Arabidopsis and soybean was executed through an in-depth analysis of their oscillatory parameters. Phase24 is a period-corrected estimation of the peak expression time of circadian rhythmic genes [26]. The distribution of the phase24 of circadian rhythmic genes in Arabidopsis manifests a major peak at 10.97 h, with two minor peaks at 1.57 h and 22.60 h. However, the peak at 11.62 h is dramatically reduced, while the peak at 22.3 h becomes more dominant in soybean circadian rhythmic genes (Figure 3a). Consistently, a Kolmogorov–Smirnov test suggests statistically significant different distribution patterns of the phase24 of Arabidopsis and soybean circadian rhythmic genes (p < 0.001). The period distribution of soybean rhythmic genes is also different from that of Arabidopsis. Soybean rhythmic genes have significantly longer periods than those of Arabidopsis genes (p < 0.001, Mann–Whitney test, Figure 3b). Additionally, soybean rhythmic genes exhibit a higher amplitude than those of Arabidopsis (p < 0.001, Mann–Whitney test, Figure 3c). Notably, the average expression level of soybean rhythmic genes in is lower than that of Arabidopsis genes (p < 0.001, Mann–Whitney test, Figure 3d). These statistical analyses unequivocally indicate significant distinctions across all four key oscillation parameters. Importantly, these conclusions remain the same when alternative circadian oscillation correlation cutoffs, such as ≥0.5/0.6/0.8/0.9, are adopted (Figures S2–S5). Collectively, these findings reveal that the Arabidopsis and soybean rhythmic genes display drastically different oscillatory characteristics.

2.2. Comparative Analysis of Homologous Circadian Clock Genes Implies Intricate Circadian Regulatory Mechanisms Distinguishing Soybean from Arabidopsis

To delve deeper into the variations in circadian regulation, we conducted an in-depth comparative analysis of the core clock genes in Arabidopsis and their corresponding homologous genes in soybean. The details of Arabidopsis gene names and IDs, alongside their counterparts in soybean, are outlined in Table S1. Previous studies have indicated that Arabidopsis CCA1 and LHY constitute partially redundant genes essential for circadian rhythm maintenance [14,35]. We identified their homologous genes in soybean. By comparing their oscillation parameters, we observed that no GmCCA1/LHY genes exhibit significantly changed periods compared to CCA1 (Figure 4b). However, all GmCCA1/LHY genes, except GmCCA1/LHY_5, demonstrate lower amplitudes than Arabidopsis CCA1 (Figure 4c). The oscillations of CCA1, LHY, and all GmCCA1/LHY genes are well pronounced (Figure 4a and Figure S6a), with GmCCA1/LHY_4 displaying an apparent phase shift (Figure 4a and Figure S6a). These findings underscore diverse changes across various GmCCA1/LHY genes when compared with CCA1.

Given that CCA1 and LHY are fundamental constituents of the morning loop within the central oscillators of Arabidopsis, while TOC1 serves as the core component of the evening loop, we extended our comparison to TOC1 and its homologous genes in soybean. Notably, there is no discernible difference in period between TOC1 and its soybean-homologous genes (Figure 4e). While the amplitudes of GmTOC1_1 and GmTOC1_2 closely mirror that of TOC1, GmTOC1_3 displays a lower amplitude, whereas GmTOC1_4 exhibits a higher amplitude (Figure 4f). Generally, all GmTOC1 genes exhibit better oscillations than TOC1, with a negligible disparity in phase (Figure S6b).

Next, we directed our attention to the key clock genes within the Arabidopsis central oscillators (Figure 5). Among the analyzed soybean homologous genes corresponding to the 15 Arabidopsis clock genes, only GmPRR3, GmZTL and GmELF3 exhibited a noteworthy alteration in period (Figure S7). Meanwhile, among the soybean homologous genes linked to 13 of the Arabidopsis clock genes, there was a notable finding: for each of these genes, at least one ortholog exhibited significant alterations in amplitude (Figure S8). With the exception of GmPRR9, the remaining soybean homologous genes demonstrated a significant phase shift (Figure S9). These shifts in the circadian expression patterns of circadian clock genes imply intricate and distinct regulatory mechanisms governing the central oscillators between Arabidopsis and soybean.

2.3. Circadian Parameter Alterations of Genes Provide Insights into the Circadian Control of Flowering and Maturity in Soybean

Numerous genetic loci, including those designated as the E series and J (long juvenility gene), have been cloned and determined to exert regulatory control over the flowering and maturation processes in soybean [36]. Remarkably, among these genes, E2, E3, E4, and J have been recognized as homologous counterparts of Arabidopsis clock genes [37,38,39,40].

Locus J is the ortholog of Arabidopsis EARLY FLOWERING 3 (ELF3), and J can repress the transcription of E1, which is a legume-specific flowering repressor [41], so as to promote flowering under short days [40]. In this study, we found that the phase of J (GmELF3_1, GLYMA_04G050200) shifts from 15.87 (phase of Arabidopsis ELF3) to 14.11, and J (GmELF3_1) shows better oscillation than ELF3 (Figure 6a and Figure S9n). Moreover, the period of J shows no difference from that of ELF3 (Figure S7n). However, the amplitude of J (GmELF3_1) is significantly higher than that of ELF3 (Figure S8n). The four orthologs of ELF3 show diverse circadian expression patterns (Figure 5n), implying potentially different functions.

Locus E2 is the ortholog of Arabidopsis GIGANTEA (GI), and E2 can repress flowering in soybean [38]. In this study, we found that the phase of E2 (GmGI_1, GLYMA_10G221500) significantly shifts from 9.02 (phase of Arabidopsis GI) to 7.74, and E2 shows better robustness of oscillation than GI (Figure 6b and Figure S9o). However, the period of E2 is no significantly different from GI (Figure S7o). Additionally, the amplitude of E2 (GmGI_1) is no different from GI (Figure S8o).

Loci E3 and E4 are orthologs of Arabidopsis phytochrome A (PHYA) [37]. E3 and E4 could respond to different degrees of light under LDs and repress soybean flowering [42]. Details regarding the corresponding gene name and gene ID of PHYA and its orthologs in soybean are listed in Table S1 (under “PHYA” tab). E3 (GmPHYA_3, GLYMA_19G224200) and E4 (GmPHYA_1, GLYMA_20G090000) present obvious phase shifts compared with PHYA (Figure 6c). An interesting phenomenon is that E4 (GmPHYA_1) oscillates better than PHYA, while E3 (GmPHYA_3) oscillates much worse than PHYA. Besides, E4 (GmPHYA_1) shows no difference in period compared to PHYA and has higher amplitude than PHYA (Figure S10). However, E3 (GmPHYA_3) shows a shorter period and lower amplitude than PHYA (Figure S10).

These findings showcase a multitude of divergent traits between Arabidopsis clock genes and their corresponding orthologs implicated in the regulation of flowering and maturity within soybean. These results imply that the crucial roles played by circadian clock genes in governing the intricate processes of flowering and maturation may be the result of alterations of key circadian parameters.

2.4. Circadian Control of Physiological Activities in Arabidopsis and Soybean

While the abovementioned findings have allowed us to compare the attributes of central oscillators in Arabidopsis and soybean, the associated physiological activities regulated by these circadian clock genes remain to be investigated. To maximize the information derived from COS analysis, we conducted a gene set enrichment analysis (GSEA) utilizing the circadian oscillation correlation from both Arabidopsis and soybean. The enriched gene ontology (GO) terms were ranked by normalized enrichment score (NES). A higher NES implies a greater concentration of genes associated with the GO term at the top of the gene list, indicating a higher circadian oscillation correlation and better rhythmicity. We observed that both organisms feature enriched GO terms associated with rhythmic process-and photosynthesis-related functions (Figure 7). However, when it comes to molecular function (MF) and cellular component (CC) terms, Arabidopsis rhythmic genes are also enriched with many ribosomal-associated terms, while soybean rhythmic genes are still primarily enriched with GO terms associated with the photosystem (Figure 7). The expression profiles of rhythmic genes associated with the enriched GO term “cytosolic large ribosomal subunit” in Arabidopsis are presented in Figure S11, while those associated with the enriched GO term “photosystem” in soybean are profiled in Figure S12. This disparity suggests that rhythmic genes in Arabidopsis are more linked with translation processes in addition to photosynthesis, whereas their counterparts in soybean appear to correlate predominantly with photosynthesis.

2.5. Differential Circadian Regulation of Physiological Activities in Arabidopsis and Soybean

To enable a better dissection of the distinctive circadian regulations on physiological activities in Arabidopsis and soybean, we employed a tailored approach to filter GO terms. We compared the relationships among these GO terms and categorized overlapping terms into four groups based on their NES.

Initially, we employed FDR ≤ 0.05 as a threshold, yielding 581 enriched GO terms in Arabidopsis and 454 enriched GO terms in soybean. Of these, 194 terms were found to overlap between two organisms (Figure 8a). Notably, a significant positive correlation between the NESs derived from Arabidopsis and those from soybean is evident in these overlapping terms (p < 0.0001, F test, Figure 8b). This suggests that Arabidopsis and soybean share a consistent circadian regulation on the physiological activities associated with these terms.

To allow a more exhaustive survey of the distinctions of the circadian regulations between Arabidopsis and soybean, we reduced the threshold to p-value ≤ 0.05, since the multiple comparison corrections used to derive FDR are not strictly required due to the inherent algorithm difference between GSEA and a traditional GO enrichment analysis [43]. This resulted in 1377 GO terms for Arabidopsis and 1067 GO terms for soybean. Among these, 444 terms are shared by both organisms (Figure 8c). Intriguingly, eight GO terms were classified in the upper-left region of the NES diagram (Figure 8d), indicating positive Arabidopsis NESs but negative soybean NESs. These eight terms are predominantly associated with translation, as outlined in more details in

Table 1. 8 GO terms from 444 overlapped terms between Arabidopsis thaliana and Glycine max.

			Arabidopsis thaliana			Glycine max
Type	GO ID	Description	NES 1	p Value	Rank 2	NES	p Value	Rank
BP	GO:0042274	ribosomal small subunit biogenesis	2.7329	0.0001	5671	−1.5852	0.0405	20,587
CC	GO:0005852	eukaryotic translation initiation factor 3 complex	2.3767	0.0001	5596	−1.6797	0.0131	15,219
BP	GO:0055075	potassium ion homeostasis	1.7333	0.0126	9684	−1.6109	0.0496	8780
CC	GO:0000502	proteasome complex	1.5905	0.0146	10,956	−2.2152	0.0120	16,030
CC	GO:0005839	proteasome core complex	1.6584	0.0166	11,279	−2.2368	0.0013	16,579
CC	GO:1905369	endopeptidase complex	1.5202	0.0229	10,956	−2.4388	0.0169	16,030
BP	GO:0010499	proteasomal ubiquitin-independent protein catabolic process	1.5683	0.0364	11,279	−2.0977	0.0033	16,579
CC	GO:0070993	translation preinitiation complex	1.4947	0.0488	5486	−1.7524	0.0045	15,219

¹, NES, normalized enrichment score. ², Rank, order of the gene which corresponds to the enrichment score in the descending list.

. This indicates that translation-related processes are enriched with circadian rhythmic genes in Arabidopsis but depleted in soybean, suggesting differential involvement of circadian regulations on translation in these two organisms.

To further reveal the distinctions in the circadian regulation of physiological activities between these Arabidopsis and soybean, we carried out an enrichment map analysis targeting the BP terms of the 387 GO terms uniquely enriched in Arabidopsis, as well as the BP terms of the 260 GO terms specifically enriched in soybean (GO terms from Figure 8a). The enrichment map pertaining to Arabidopsis unveiled an apparent concentration of pathways related to translation, response to stimuli and nucleobase biosynthetic processes (Figure 8e, and the whole map elaborated in Figure S13). On the other hand, the enrichment map of soybean identifies enrichment of activities associated with the regulation of nuclear division and cysteine metabolic processes (Figure 8f, with the whole map information in Figure S14).

Finally, we looked into all the 6462 GO terms from Arabidopsis and 5135 GO terms from soybean without any specific cutoffs to allow the most relaxed and inclusive comparison. Among these, 4603 terms are shared by Arabidopsis and soybean (Figure S15a). These terms can be separated into four regions in the NES diagram (Figure S15b).

Within each of these regions, we systematically sorted the terms based on the ascending order of p values. Subsequently, we extracted the top five terms from the upper-left region indicating positive Arabidopsis NESs but negative soybean NESs (Figure 9a–e). Impressively, the predominant theme of these five terms is associated with ribosome and translation processes. This compelling observation suggests a heightened likelihood of translation activities being regulated by the circadian clock in Arabidopsis, whereas Glycine max might deviate from circadian regulation in this context. In parallel, we identified the top five terms from the lower-right region signifying positive soybean NESs but negative Arabidopsis NESs (Figure 9f–j). Interestingly, these five terms are notably related to phosphogluconate dehydrogenase (decarboxylating) activity and related functions. This pattern implies that specific enzyme activities in soybean are more prone to circadian regulation, potentially differing from the norm in Arabidopsis. Further specifics about these ten terms are outlined in

Table 2. Comparison of 10 GO terms from GSEA results between Arabidopsis thaliana and Glycine max.

	Arabidopsis thaliana				Glycine max
Term	NES 1	p Value	FDR 2	Rank 3	NES	p Value	FDR	Rank
ribonucleoprotein complex subunit organization	1.965	0.000	0.003	5744	−1.933	0.333	0.674	20,531
ribonucleoprotein complex assembly	1.995	0.000	0.003	5744	−1.943	0.333	0.674	20,531
ribosomal small subunit biogenesis	2.745	0.000	0.003	5671	−1.633	0.037	0.199	20,587
cytoplasmic translation	2.741	0.000	0.003	5696	−1.311	0.185	0.494	18,680
maturation of SSU-rRNA from tricistronic rRNA transcript	2.491	0.000	0.003	7533	−1.331	0.083	0.323	20,587
cytidine to uridine editing	−0.743	0.842	0.944	17,268	2.032	0.001	0.022	5953
phosphogluconate dehydrogenase (decarboxylating) activity	−0.882	0.598	0.819	11,493	1.780	0.008	0.075	5914
amyloplast	−0.939	0.520	0.769	10,958	1.752	0.013	0.104	11,820
carbon-oxygen lyase activity, acting on phosphates	−0.732	0.847	0.948	15,240	1.721	0.015	0.112	8267
protein kinase CK2 complex	−1.005	0.422	0.703	13,202	1.684	0.019	0.131	4358

¹, NES, normalized enrichment score. ², FDR, false discovery rate. ³, Rank, order of the gene which corresponds to the enrichment score in the descending list.

.

3. Discussion

The circadian clock is a fundamental timekeeping mechanism that allows organisms to synchronize their biological activities with the external environment, ensuring optimal growth, development, and adaptation. Arabidopsis is an important model organism extensively studied to dissect the intricate molecular underpinnings of circadian rhythms. Glycine max, also known as soybean, is a major crop utilized in many aspects of human life. This study employed a comprehensive comparative analysis of circadian time-course transcriptome profiles to uncover the circadian characteristics that differentiate these two species, shedding light on the investigations and applications of the intricate regulatory mechanisms in soybean’s circadian clock.

Firstly, with a circadian oscillation correlation no less than 0.7 as a threshold, we obtained a percentage of 44.98% and 42.51% of circadian rhythmic genes among the expressed genes of Arabidopsis and soybean, respectively. The percentage in Arabidopsis is a little higher than that in soybean. However, both percentages are in common regions, as reported in previous study which observed that about 5.2% to 55.9% of genes show significantly rhythmic expression across species in Archaeplastida [44]. Besides, some studies have reported that about 50% of genes are rhythmic in mammalian animals [45,46]. Therefore, the percentage of rhythmic genes in Arabidopsis and soybean are somehow consistent with that in broad species.

Then, with the analysis of circadian rhythmic genes, there were significant variations in the expression patterns, phase24, period, and amplitude of these genes. Soybean homologous genes of Arabidopsis rhythmic genes were different in their expression patterns and generated distinct rhythmicity. For example, Arabidopsis PHYA (circadian oscillation correlation = 0.97) and soybean GmPHYA_1 (circadian oscillation correlation = 0.96) present good rhythmicity, while soybean GmPHYA_4 (circadian oscillation correlation = 0.59) is not rhythmic. Besides, our analysis also showed the differentiation of homologous clock genes in phase24, period, and amplitude. These differences underscore the intricate nature of the circadian regulatory mechanisms within each species. More comprehensive details of the differentiation phenomenon remain a concern for future investigation.

Further analysis of the circadian clock genes provided deeper insights into the circadian control of flowering and maturity. Homologous genes like GmGI and GmELF3 of Arabidopsis clock genes in soybean were previously found to play pivotal roles in the regulation of these critical developmental processes [38,40]. The circadian parameter alterations of these homologous genes in soybean emphasize the importance of circadian genes in the regulatory pathways across different species. In addition, previous studies have reported that the period and phase shifts of clock genes were relevant to the domestication of cultivated tomato [47]. It cannot be denied that other homologous genes of Arabidopsis clock genes in soybean may also play roles in a wide range of applications.

Delving into the enrichment map analysis, this study uncovered the specific biological processes subjected to circadian regulation in each organism. Interestingly, while some processes showed overlapping patterns of regulation, such as rhythmic processes and photosynthesis-related terms, others exhibited distinct patterns. The variations in circadian regulation were evident in processes associated with translation, response to stimuli, and nucleobase biosynthetic processes in Arabidopsis, while in soybean, activities tied to regulation of nuclear division and cysteine metabolic processes stood out. These differential regulatory patterns likely reflect the main roles of rhythmic genes in each species.

A recent study identified starch metabolism as a clock-controlled pathway in hexaploid bread wheat, and provided important targets for future wheat breeding [31]. In this study, one of the remarkable findings was the divergent behavior of translation activities between Arabidopsis and soybean. While translation activities in Arabidopsis were more likely to be regulated by the circadian clock, circadian rhythmic genes were actually depleted in translation-related pathways in soybean. Another finding was that some activities, especially enzyme-associated activities such as phosphogluconate dehydrogenase (decarboxylating) activity, are more likely under circadian regulation in soybean. By separating common GSEA GO terms between Arabidopsis and soybean into four groups (Figure 8a–d and Figure S14), we identified terms showing opposite NES. Larger positive NES means the genes behind the term tend to concentrate in the top region of the whole gene list and tend to be more rhythmic. On the one hand, the results confirmed that translation activities in Arabidopsis were more likely to be regulated by the circadian clock (Table 1 and Table 2). On the other hand, the results indicated some activities, especially enzyme-associated activities, are more prone to circadian regulation in soybean (Table 2). This observation underscores the specific adaptability of circadian regulation in soybean and provides more directions for research about soybean circadian rhythms.

In conclusion, this comprehensive comparative analysis of circadian time-course transcriptome profiles provides a multifaceted understanding of the circadian characteristics that distinguish soybean from Arabidopsis. The findings extend insights into rhythmic genes and highlight the variations of biological processes between Arabidopsis and soybean, implying the complexity between circadian rhythms and biological activities. As our understanding of circadian regulation deepens, insights from this study may pave the way for targeted interventions in crop development, enhancing agricultural productivity and sustainability in a changing world.

4. Materials and Methods

4.1. Plant Materials and Growth Conditions

The Arabidopsis plants used in the study were wild-type, and of the Columbia (Col-0) ecotype. Col-0 seeds were sown on MS medium plates and cold-treated (4 °C, darkness) for 3 days. The MS plates were transferred to 22 °C under LD conditions (12 h light/12 h dark cycles; 50 μmol⋅m⁻²⋅s⁻¹ of white light) for 11 days, and then placed under LL conditions (24 h light; 50 μmol⋅m⁻²⋅s⁻¹ of white light).

The soybean seedlings used in the study were wild-type, and of the soybean cultivar Williams 82. Williams 82 seedlings were grown in soil under LD conditions (16 h light/8 h dark, 100 μmol⋅m⁻²⋅s⁻¹, 28 °C, 50% relative humidity) for 9 days. On the tenth day, the light was switched to LL (24 h light; 100 μmol⋅m⁻²⋅s⁻¹).

After a one-day transition under LL conditions, samples were harvested at 4 h intervals for 2 continuous days starting at ZT24. Samples of a total of 12 timepoints were collected. For Arabidopsis, each timepoint contains 2 biological replicates. For soybean, each timepoint contains 3 biological replicates. Samples were subjected to RNA extraction and RNA-seq. More details of Arabidopsis [30] and soybean [26] can be inferred from corresponding studies.

4.2. Sequencing Reads Acquisition, Read Alignment, and mRNA Quantification

Raw RNA-seq reads generated from Arabidopsis and soybean in this study can be retrieved from EMBL-EBI’s ArrayExpress (accession no. E-MTAB-7933) [30] and National Center for Biotechnology Information’s Gene Expression Omnibus (accession no. GSE94228) [26], respectively.

First, raw sequencing reads were analyzed using fastp [48] to remove low-quality reads and trim adapter sequences. Then, the qualified reads were aligned to the reference genome by TopHat2 (v2.1.1) [49] and assembled by StringTie (v2.1.5) [50] to obtain the quantified raw count matrix of gene expression.

4.3. Pre-Processing of Expression Data

The R packages limma [51] and edgeR [52] were used to process the raw count matrix of gene expression. First, function filterByExpr was selected to retain expressed genes which have sufficient large counts for downstream analysis. Then, scaling factors were calculated to normalize the raw library size for each sample. At the same time, the correlation matrix was calculated to check the reproducibility of samples. Lastly, the function voomWithQualityWeights was used to calculate weights for each sample to reduce the effect of variable samples on downstream statistical analysis. At the same time, normalized expression levels of genes were calculated as log₂(CPM), derived from the voomWithQualityWeights function.

4.4. Weighted Estimation of Rhythmic Parameters of Genes

After pre-processing of the expression data, non-linear regression fitting (COS) integrated with weights was applied to each gene. The following formula was used in fitting.

$$\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{l}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{l}=\mathrm{A}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{u}\mathrm{d}\mathrm{e}\times \mathrm{c}\mathrm{o}\mathrm{s}(\frac{2\mathsf{\pi }}{\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}}\times \mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}-\frac{2\mathsf{\pi }}{24}\times \mathrm{P}\mathrm{h}\mathrm{a}\mathrm{s}\mathrm{e}24)+\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{t}$$

(1)

Each gene has 12 timepoints. For each gene, the rhythmic parameters were estimated using the fitting formula with sampling time as the independent variable and the corresponding sample’s expression level as the dependent variable. The fitting was carried out in the following way:

(1)

Remove the slope trends by fitting a linear regression model to the gene expression data and keep residuals.

(2)

Perform Fast Fourier Transformation (FFT) of the time series data and keep main signals.

(3)

Estimate the initial period using FFT transformed data.

(4)

Perform non-linear regression fitting. The fitting was applied with weights obtained in the pre-processing. Amplitude was constrained to be non-negative. Period was constrained to be greater than 12 h but less than 36 h. Phase24 and Constant have no constraints. Phase was normalized as phase24, which is more than 0 but less than 24. Constant indicates the average expression level of the gene.

(5)

Predict the best-fit data and calculate circadian oscillation correlation between observed data and predicted data.

(6)

Genes without convergent fits were considered arrhythmic. Genes with the resulting best-fit Amplitude, Period, Phase24, Constant and their standard error and degree of freedom were used for downstream statistical analysis.

4.5. Analysis of Homologous Genes

Homologs of Arabidopsis circadian clock genes in soybean were identified according to the method described previously [26]. Time-course expression levels of the Arabidopsis circadian clock gene and its homologous genes in soybean were profiled together using normalized data from the pre-processing. The rhythmic parameters of homologous genes in soybean were statistically tested with those of Arabidopsis circadian clock genes.

4.6. Phase24 Plots

Phase24 plots were shown as radial plot indicating differences in phase24 and oscillation robustness between Arabidopsis genes and their homologous genes in soybean. Phase24 indicates the gene’s phase normalized into a period of 24 h, and is plotted as the angular coordinate. Robustness is indicated by −log₁₀(p) with a larger −log₁₀(p) representing better oscillation. Lines with arrows indicate the phase24 shift and robustness change from the Arabidopsis genes to their homologous genes in soybean.

4.7. Gene Set Enrichment Analysis

Gene set enrichment analysis (GSEA) is a method to determine whether a group of specific genes tends to occur toward the top or bottom of the ranked whole gene list. In this study, GSEAs were performed using the R package clusterProfiler [53]. For Arabidopsis GSEA, the database org.At.tair.db was used. For soybean GSEA, database AH85411 from package AnnotationHub was used. The gene list was generated by sorting genes with their circadian oscillation correlation in descending order for Arabidopsis and soybean, respectively. Larger circadian oscillation correlation indicates better oscillation. The maximum size of gene set for analysis was set as 800. The minimum size of gene set for analysis was set as 3. Enriched GO terms including BP, MF, and CC were analyzed using the function gseGO in clusterProfiler. The network showing relationships and significance of BP terms was generated by the emapplot function in the package enrichplot [54]. The GSEA enrichment plots were generated using the gseaplot2 function in the package enrichplot.

4.8. Statistical Analysis

The statistical methods and details are indicated in the methods or figure legends. Statistical tests and analysis were performed using RStudio software version 4.0.4.