Introduction

Plants have a complex set of signaling pathways that enables them to interpret and respond to their surroundings. One mechanism by which plants alter their growth and development in response to environmental cues is through changes in gene transcription. An important group of transcriptional regulators involved in this process are members of the AT-HOOK MOTIF NUCLEAR LOCALIZED (AHL) family [1,2,3]. In Arabidopsis thaliana (Arabidopsis), AHL29/ SUPPRESSOR OF PHYTOCHROME B4-#3 (SOB3) has an important function regulating growth and development in response to light, affecting processes such as hypocotyl growth, flowering, and senescence [1, 4,5,6,7,8,9,10]. SOB3 was identified as a repressor of hypocotyl growth in an activation-tagging screen for suppressors of the long-hypocotyl phenotype exhibited by the phytochrome B-4 (phyB-4) photoreceptor mutant grown in white light [1, 11,12,13]. Further investigation revealed that SOB3 regulates hypocotyl growth in a light-dependent manner, affecting growth specifically in dim light [1]. Therefore, SOB3-Dominant (SOB3-D) seedlings, which are derived from the aforementioned activation-tagging experiment, have enhanced expression of SOB3 and exhibit short hypocotyls compared to wild-type Col-0 plants [1, 5]. In addition to SOB3, other AHLs act in a semi-redundant fashion with this family member to regulate hypocotyl growth in light [6].

The AHL gene family is characterized by two conserved elements important for its function: a PLANT AND PROKARYOTIC CONSERVED/DOMAIN OF UNKNOWN FUCTION #296 (PPC/DUF296), and one or two AT-hook motif(s) [6, 14, 15]. SOB3 interacts with itself, other AHLs, and non-AHL transcription factors via the PPC domain [6]. On the other hand, the AT-hook motif(s) bind DNA sequences rich in adenine and thymine [4, 6, 16,17,18]. Notably, a single missense mutation within the central AT-hook domain of SOB3 in the SOB3-D background, known as the sob3-6 allele, produces extremely tall seedlings [1]. This phenotype is attributed to a dominant-negative effect of a non-functional SOB3-6 protein being overexpressed, i.e. the SOB3-6 protein can engage in protein–protein interaction as it normally would, but cannot bind to DNA [6]. This leads to the formation of unproductive complexes that fail to inhibit seedling growth. From this research, it became clear that the AT-hook and PPC domain are both important for the function of SOB3 and other AHLs.

AHLs regulate growth in both constant light and long days (LD; 16 h light/8 h dark) at least partially by repressing the auxin pathway. In seedlings grown in constant light, SOB3 binds directly to the promoters of the auxin biosynthesis gene YUCCA8 (YUC8) and auxin signaling genes belong to the SMALL AUXIN UP RNA 19 (SAUR19) subfamily and represses their transcription [5]. However, overexpression of SAUR19 only partly recovers the short-hypocotyl phenotype of SOB3-D, suggesting that the effect of AHLs on growth is not fully explained by its effect on the auxin pathway. In a different study using seedlings grown in LD, Lee and Seo found that AHLs repress YUC9 transcription by promoting deposition of the H2A.Z histone variant on the promoter of this gene [8]. This likely represses auxin biosynthesis and consequently, hypocotyl growth in older seedlings. Additionally, results from another study using LD-grown plants suggest that AHLs inhibit petiole growth largely by acting on the auxin pathway [18]. In this study, it was found that AHLs limits petiole elongation by antagonizing the growth-promoting PHYTOCHROME-INTERACTING FACTORs (PIFs). Specifically, AHLs repress PIF-mediated activation of genes involved in hormone-mediated growth, such as ACS8, which is involved in ethylene biosynthesis, NPY1, which is involved in auxin transport, YUC8, and SAUR24, thus accounting for their negative effect on petiole growth.

The majority of the research on AHLs has been performed in continuous white light or LD. Not only is constant light uncharacteristic of a natural growth environment, but there are likely different mechanisms that regulate growth in short days (SD; 8 h light/16 h dark), compared to continuous light or LD [18]. Additionally, it is unknown if AHLs have different effects on gene expression throughout the day. Therefore, in order to investigate these questions, we examined how AHLs regulate hypocotyl elongation in SD, a condition in which little is known about how this family of transcription factors affects growth.

Results

AHLs repress hypocotyl growth in SD

We investigated how AHLs regulate hypocotyl elongation in short days (SD) (8 h light/16 h dark), a condition in which little is known about how this family of transcription factors affect growth. Phenotypic analysis of mutants for SOB3 revealed that SOB3-D exhibits reduced hypocotyl growth in SD, while sob3-6 seedlings are extremely tall (Fig. 1). These phenotypes indicate that SOB3 and/or other AHLs repress hypocotyl growth in SD, similar to what has been reported for seedlings grown in constant light or long days [1, 4,5,6, 8, 18].

Fig. 1
figure 1

AHLs repress hypocotyl growth in SD. Hypocotyl measurements of sob3-6, WT Col-0, and SOB3-D five-day-old seedlings grown in SD. Error bars represent standard error of the mean. In a Welch’s t test (unpaired two-tailed t test with unequal variance) compared with the wild type: P ≤ 0.0001 = ****

Full size image

AHLs regulate the expression of many genes in a time point-specific manner in SD

Next, in order to gain insight into the molecular mechanisms by which AHLs affect hypocotyl growth in SD, we identified nuclear-encoded genes that are differentially expressed between SOB3-D and sob3-6. Specifically, we performed RNA-seq using seedlings harvested at three different time points: at ZT4 (four hours after dawn) and ZT9 (one hour after dusk) in five-day-old seedlings, and at dawn in six-day-old seedlings (ZT24) (Fig. 2A; Table S1). The three time points of ZT4, ZT9, and ZT24 were specifically chosen based on what is known about the regulation of hypocotyl growth as well as PIF activity in SD. In SD, hypocotyl growth mainly occurs at the end of the night, when high expression of PIF4 and PIF5 coincides with minimal post-translational repression of the PIFs by phytochromes and other inhibitory factors [19,20,21]. Therefore, we chose to test the ZT24 time point to investigate if AHLs might be preventing excessive growth at this time point by binding to and regulating the expression of PIFs and their target genes. On the other hand, we chose the other two time points because they represent times during the day (ZT4) or evening (ZT9) when PIF activity and growth are minimal. We reasoned that AHLs might be contributing to repression of growth at one or both of these time points, possibly by inhibiting PIF transcription and/or the expression of PIF-activated genes.

Fig. 2
figure 2

AHLs repress the expression of growth-promoting genes while also activating growth repressors. RNA-seq analysis of SOB3-D and sob3-6 seedlings in SD. A SD photoperiod diagram depicting the time points used to harvest seedings for RNA-seq. The diagram starts at the beginning of day five and ends at the beginning of day six. B, C AHL-repressed (B) or -induced (C) nuclear-encoded genes identified from the RNA-seq data based on having higher or lower expression in sob3-6 than SOB3-D, respectively. In (B), AHL-repressed PIFs, PIF-target genes, and BR-associated genes are named. In (C), AHL-induced growth repressors are named

Full size image

EdgeR analysis of our RNA-seq data for SOB3-D and sob3-6 revealed 2,276 differentially expressed genes at ZT4, 2,560 at ZT9, and 1,480 at ZT24 (Fig. 2B, C), indicating that SOB3 and/or other AHLs might have a greater impact on gene expression during the middle of the daytime and early during the evening than at dawn. Of these differentially expressed genes, substantially more, 2,835, were found to be repressed by AHLs (Fig. 2B) with only 1,654 induced by AHLs (Fig. 2C). This suggests that AHLs might act more often as transcriptional repressors than transcriptional activators in SD. Furthermore, the majority of genes are mis-regulated exclusively at only one time point. Among the AHL-repressed genes, 651 are uniquely repressed at ZT4, 798 at ZT9, and 435 at ZT24, which represents 23%, 28%, and 15% of all repressed genes, respectively (Fig. 2B). Consistent with the overall smaller number of AHL-induced compared to AHL-repressed genes, there are fewer uniquely induced genes at all time points: 452 at ZT4, 567 at ZT9, and 202 at ZT24 (Fig. 2C). The uniquely induced genes at ZT4 and ZT9 make up 28% and 26% of all induced genes respectively, whereas ZT24 represents only 13%. Interestingly, there are relatively few genes mis-regulated at all three time points, particularly for the AHL-induced genes. Among induced genes, only 4% are mis-regulated at all three time points, whereas 13% are mis-regulated at all three time points for AHL-repressed genes (Fig. 2B, C). Collectively, these data suggest that the expression of most AHL-regulated genes is only affected by these transcription factors at certain times during the day.

AHLs regulate the expression of PIFs and their target genes to control growth of the hypocotyl in SD

At least one way by which AHLs inhibit growth is by reducing the expression of PIF-activated genes [18]. Therefore, we investigated how the expression of several PIF target genes are affected in our RNA-seq data, focusing on PIF-activated genes highlighted in a recent review [21]. Indeed, our transcriptomic data indicate that PIF target genes are repressed by AHLs throughout the day in SD-grown seedlings (Fig. 2B). Some PIF-activated genes were identified as AHL-repressed at all three time points, such as YUC8, IAA19, and two SAURs (Fig. 2B), which are all growth promoting genes associated with the auxin pathway that have been also previously identified as genes directly repressed by the AHLs [5, 18]. Notably, we found that in addition to IAA19, the similarly functioning and PIF-target gene IAA29 [22,23,24] is also repressed by the AHLs at all three time points (Fig. 2B). The PIF-activated genes PIL1 and PRE1 [25, 26] are also found among the AHL-repressed genes at all three time points (Fig. 2B). Notably, PRE1 promotes hypocotyl growth, likely at least in part by enhancing PIF protein activity [26,27,28,29,30,31]. Other PIF targets were only identified as AHL-repressed at one or two time points. For example, the growth-promoting gene CDF5 [32] is found among the AHL-repressed genes at ZT9 and ZT24, but not at ZT4 (Fig. 2B). Additionally, some genes that are differentially regulated between SOB3-D and sob3-6 at all three time points exhibit different fold changes in expression at the different time points, such as SAUR22 (greatest fold change at ZT4), SAUR19 (greatest fold change at ZT9), and IAA19 (greatest fold change at ZT24), (Table S2). Thus, AHLs appear to have quantitatively different effects on the expression of some PIF-regulated genes throughout the day.

In addition to finding that many PIF targets are repressed by AHLs, our data also indicate that AHLs repress the expression of PIFs themselves in SD. Specifically, PIF4 is among the AHL-repressed genes at ZT4 and ZT9, while PIF5, 7, and 8 are repressed exclusively at ZT9 (Fig. 2B). These data suggest that AHLs regulate the expression of PIFs during the daytime and early in the evening, but not late in the night. Remarkably, this is in contrast to the effect of AHLs on the expression of PIF target genes described in the previous paragraph, which appears instead to persist throughout the entire day. Therefore, AHLs likely inhibit hypocotyl growth in SD at least in part by reducing the expression of PIF-activated genes, which is likely partially a downstream consequence of AHLs repressing expression of the PIFs themselves.

AHLs repress brassinosteroid biosynthesis genes and activate the expression of growth repressors to regulate hypocotyl growth in SD

Our RNA-seq data also indicate that AHLs likely repress brassinosteroid (BR) biosynthesis downstream of the PIFs as well as BR signaling. CPD and BR6OX2, both PIF4-activated genes that encode enzymes involved in brassinosteroid biosynthesis [33, 34], are found among the ZT4 AHL-repressed genes, while BR6OX2 also falls into this group at ZT24 (Fig. 2B). We further found that like CPD, BES1, which encodes a key transcription factor involved in BR-mediated hypocotyl growth [35, 36], is specifically repressed by AHLs at ZT4 (Fig. 2B). Taken together, these data suggest that in SD, the repressive effect of AHLs on hypocotyl growth is likely caused in part by them inhibiting the BR pathway.

Besides repressing the expression of genes that promote hypocotyl growth in SD, our RNA-seq data also indicate that AHLs activate the expression of growth repressors. Specifically, eight growth repressors highlighted in the Favero et al. review [21] are among the AHL-induced genes we identified from our RNA-seq experiment (Fig. 2C). Of these eight genes, six were identified as AHL-induced only at a single time point. Four growth repressors, including HY5 and BBX22, are AHL-induced specifically at ZT4. Notably, HY5 and BBX22 proteins are protected from COP1-mediated degradation only in the presence of light [37, 38], thus AHLs appear to activate HY5 and BBX22 specifically when their encoded proteins are able to contribute to hypocotyl growth repression, i.e. during the daytime [39,40,41]. BBX22’s close homolog, BBX21, on the other hand, is one of two genes we found to be AHL-induced only at ZT24; however, BBX21 is unlikely to contribute to growth repression at the end of the night, since it, like BBX22, is largely degraded in a COP1-dependent manner in darkness [41, 42]. In addition to these six growth repressors, two were identified as AHL-induced at two different time points: ELF4 at both ZT4 and ZT9 and IAA17 at both ZT4 and ZT24. Collectively, our RNA-seq data suggest that AHLs inhibit hypocotyl growth in SD by not only repressing the expression of PIF-activated, growth-promoting genes, but also by increasing the expression of growth repressors, particularly during the daytime.

Changes in SOB3 binding do not explain the differential regulation of genes at ZT4, ZT9, and ZT24 in SD

Since our RNA-seq data show that AHLs regulate somewhat different sets of genes at different times during the day, we next investigated if SOB3 binding to DNA changes throughout the day in SD by performing ChIP-seq at the same three time points used for RNA-seq (Fig. 2A; Table S3). Genes bound by SOB3 in at least two of the three replicates for a given time point were considered SOB3-bound genes for subsequent analyses (Fig. 3, yellow outline), which resulted in 14,718 genes at ZT4, 14,226 genes at ZT9, and 14,778 genes at ZT24. When these genes were compared to each other, we found that most genes are bound by SOB3 at all three time points (Fig. 4). Interestingly, there are a smaller number of genes that are only bound at one or two time points (Fig. 4). These observations suggest that SOB3 is mostly binding to similar genes at these time three points.

Fig. 3
figure 3

SOB3 binds a similar number of genes at ZT4, ZT9, and ZT24. Overlap between genes found to be bound by SOB3 in different replicates of the ChIP-seq experiment at A) ZT4, B) ZT9, or C) ZT24. ChIP 1 is the first replicate, ChIP 2 is the second replicate, and ChIP 3 is the third replicate. Yellow outlines indicate genes that were used for subsequent analysis, i.e. those bound in at least two replicates within a time point

Full size image
Fig. 4
figure 4

Most SOB3-bound genes are bound at ZT4, ZT9, and ZT24. Overlap between genes bound by SOB3 in at least two ChIP-seq replicates at each time point

Full size image

Next, we looked at the distribution of SOB3 binding sites in order to evaluate how SOB3 is functioning in the genome. We observed that the majority of binding sites are in close proximity to a transcription start site (promoter-TSS) (Fig. 5). Additionally, SOB3 has few binding events within genes bodies, with 5% of binding events occurring in exons and 3% occurring in introns. Favero et al. similarly observed that SOB3 mainly binds genomic regions located outside of gene bodies, with a high proportion of binding events occurring in promoter-TSS regions [18]. Interestingly, each time point has the same percentage of binding events in each of the five categories. The high proportion of binding events observed in promoter-TSS regions at all three time points lends further support to previous findings which indicate that SOB3 functions as a transcription factor [1, 4, 6, 14, 16, 17, 43,44,45,46,47,48,49,50,51], and further suggests that this role of SOB3 remains consistent throughout the day. In addition, when the binding patterns from the three time points are compared, we observe nearly identical binding of SOB3 at all three time points (Fig. 6), suggesting that this transcription factor largely binds to similar loci throughout the day. Therefore, it is unlikely that a change in binding to target genes explains the different effects of AHLs on gene expression at ZT4, ZT9 and ZT24.

Fig. 5
figure 5

SOB3 binds the same types of genomic regions at different time points. SOB3 binding pattern distribution determined from the ChIP-seq data based on annotations obtained using HOMER software. Distribution of SOB3 binding at A ZT4, B ZT9, and C ZT24. Promoter-TSS is defined as -1,000 bp to + 100 bp in relation to a transcription start site. TTS is defined as -100 bp to + 1000 bp in relation to a transcription termination site

Full size image
Fig. 6
figure 6

SOB3 exhibits similar binding to target genes at ZT4, ZT9, and ZT24. Distribution of SOB3 binding, as determined from the ChIP-seq data, to genes identified as SOB3-bound at one, two, or three time points

Full size image

Motifs associated with TCPs and PIFs are highly enriched within the SOB3 ChIP-seq data

We then used MEME-ChIP to conduct motif enrichment analysis of the ChIP-seq data [52]. This analysis showed the GGHCCA motif, which resembles sequences bound by TCPs [53,54,55,56], as the most enriched motif in all replicates for all time points (Fig. 7). This motif is also much more highly enriched compared to all other motifs in every replicate and at every time point, which suggests an intimate relationship between AHLs and the TCPs in SD. The GGHCCA motif was also seen as the most enriched motif in a recent paper by Favero et al. that examined the effects of SOB3 in long days on petiole growth [18]. As noted in their paper, this motif is likely the most enriched in this data set because TCPs and AHLs have been shown to physically interact [6, 53, 54, 57]. As such, this motif is likely bound by TCPs that are in a SOB3 protein-DNA complex. Our MEME-ChIP analysis also revealed enrichment of other types of motifs at all three time points (Fig. 7). One such motif is CACRYG, which is similar to sequences bound by PIFs [25, 26, 54, 56, 58,59,60]. This is particularly interesting given our findings from the RNA-seq data that the expression of several PIF-activated genes are repressed by the AHLs. Therefore, it is likely that AHLs bind directly to the promoters of PIF targets and regulate their expression in SD, which also appears to be happening in the context of petiole growth [18]. Various AT-rich motifs are also highly ranked at all three time points. Some of these, including AWATAAWA, AWAATAW, TATWTWW, ATAWWATA, and TATTWTW, exclusively contain adenine and thymine residues and are similar to motifs that have been demonstrated to be bound by other AHLs, including AHL12, AHL20, and AHL25, in vitro [4, 6, 14, 16, 17, 51, 54, 56]. These motifs may be bound by SOB3 itself and/or other AHL family members found in complex with SOB3. Notably, the motifs ARAGAVA or ARASAVA are also present in all replicates from all time points and could be bound by AHLs since they are also AT-rich. However, this possibility seems unlikely, since from a structural perspective, four consecutive A/T base pairs appear to be needed to accommodate binding by an AT-hook [61].

Fig. 7
figure 7

Motifs associated with TCP-binding, PIF4-binding, and AHL-binding are all enriched in regions bound by SOB3. Top 4 enriched motifs from the ChIP-seq data at each time point: A ZT4, B ZT9, and C ZT24. Enriched motifs were identified in the 300 bp regions surrounding SOB3 peaks using CentriMo and DREME software in MEME-Suite.

Full size image

GO term enrichment shows AHLs directly induce genes associated with similar processes at ZT4 and ZT9, but not at ZT24

Next, to identify genes that are directly regulated by AHLs, we compared our RNA-seq and ChIP-seq results. We identified 1,042 genes as directly repressed by AHLs at ZT4, 1,183 at ZT9, and 771 at ZT24. Interestingly, we found fewer genes were directly induced by AHLs with 595 at ZT4, 604 at ZT9, and 301 at ZT24 (Fig. 8). Thus, it appears that AHLs are likely acting as both repressors and activators of transcription in SD. However, it also appears that AHLs have a bias towards acting as repressors, which is in agreement with what we had inferred earlier based on the RNA-seq data alone. Notably, this bias of AHLs towards acting as transcriptional repressors seems to be particularly high at ZT24. Comparing SOB3-bound AHL-regulated genes at each time point, we see that 64% of them are repressed at ZT4 and 66% at ZT9, while 72% of them are repressed at ZT24 (Fig. 8).

Fig. 8
figure 8

SOB3-bound genes are more frequently repressed than activated by AHLs in SD. Overlap between AHL-repressed or AHL-induced genes identified from the RNA-seq data and SOB3-bound genes identified from ChIP-seq at each of the three time points tested: A ZT4, B ZT9, and C ZT24

Full size image

To further investigate differences in the genes directly regulated by AHLs based on time point, we compared GO term enrichment among the different time points for the genes we identified as direct targets of AHLs. The top enriched GO terms among genes that are directly repressed by AHLs largely include those involved in responses to the external environment and by extension, the internal responses that would follow (Fig. 9). Terms such as “response to endogenous stimulus”, “response to organic substance”, “response to stimulus”, and “response to hormone stimulus” are among the top repressed GO terms for all time points. However, there are some notable differences when ZT24 is compared to ZT4 and ZT9. For example, “catalytic activity” appears as the second highest GO term among directly repressed genes at ZT24, but is much lower for ZT4 and is not present within the top 25 for ZT9. There are also other interesting time point-specific findings within the GO terms of repressed genes, such as “response to auxin stimulus,” which is highly enriched at ZT4 and ZT24, but not at ZT9. Additionally, “response to brassinosteroid stimulus”, “protein phosphorylated amino acid binding”, and “phosphoprotein binding is highly enriched at ZT4. Contrastingly, “light-harvesting complex” and “chlorophyll binding” are both highly enriched specifically at ZT9. The top GO terms among genes directly induced by AHLs for both ZT4 and ZT9 time points are “chloroplast part”, “plastid” or “plastid part”, and “chloroplast”, but the top GO terms for the ZT24 time point are entirely different and include, “transcription factor activity”, “transcription regulator activity”, and “response to hormone stimulus”. Moreover, the top functions unrelated to chloroplast function that are associated with directly induced genes at ZT24 are mostly functions that are also highly enriched among directly repressed targets at this time point. Taking these data together, it seems that there are somewhat specific processes being regulated by the AHLs at the different time points, but that ZT4 and ZT9 share more similarities when compared with ZT24.

Fig. 9
figure 9

AHLs directly regulate the expression of somewhat different types of genes at ZT4, ZT9, and ZT24. Top 25 most enriched GO terms among genes identified as SOB3-bound and AHL-repressed (A), or SOB3-bound and AHL-induced (B) at ZT4, ZT9 or ZT24. The relative P value for each GO term is indicated by color, with darker shades representing smaller P values

Full size image

Changes in binding of SOB3 to target genes do not explain differential regulation of genes by AHLs at ZT4, ZT9, and ZT24

Finally, we returned to the question of whether differences in SOB3 binding at ZT4, ZT9, and ZT24 might explain the time point-specific effects of the AHLs on gene expression. First, we examined if and at which time points the AHL-regulated genes identified from our RNA-seq data and specifically highlighted in Figs. 2B and 2C are bound by SOB3. We found that nearly all of these genes are bound by SOB3, based on our ChIP-seq data, at all three time points (Table S4). Importantly, genes such as CPD, BES1, HY5, and BBX22 are bound by SOB3 at ZT4, ZT9, and ZT24, despite being subject to AHL regulation at only one of the three time points (Fig S1). These data suggest that time point-specific regulation of target genes by AHLs is not achieved through differences in transcription factor binding. To further test this hypothesis, we examined the binding distribution of SOB3 specifically to genes identified as AHL-repressed (Fig. 10) or AHL-induced (Fig. 11) at only one or two time points. Using this approach, we again observed similar binding profiles for SOB3 at all three time points. Lastly, we examined if changes in the level of SOB3 enrichment on target genes could explain the time point-specific effects of AHLs on gene expression (Figs S2 and S3). However, we did not observe changes in SOB3 enrichment between the three time points that were likely to explain the time point-specific effects of the AHLs on the expression of target genes. This further indicates that time of day-specific effects of the AHLs on gene expression are not caused by differences in binding of these transcription factors to their target genes.

Fig. 10
figure 10

SOB3 exhibits similar binding to genes identified as directly repressed by AHLs at ZT4, ZT9, and ZT24. Distribution of SOB3 binding, as determined from the ChIP-seq data, to genes identified as both SOB3-bound and AHL-repressed at only one or two time points

Full size image
Fig. 11
figure 11

SOB3 exhibits similar binding to genes identified as directly induced by AHLs at ZT4, ZT9, and ZT24. Distribution of SOB3 binding, as determined from the ChIP-seq data, to genes identified as both SOB3-bound and AHL-induced at only one or two time points

Full size image

Discussion

AHLs regulate growth as well as the expression of specific genes throughout SD

SOB3 and other AHLs have been shown to be involved in light-mediated hypocotyl development in continuous white light and LD [1, 4,5,6, 8, 10]. Firstly, we showed here that SOB3 is also able to affect growth of the hypocotyl in SD (Fig. 1), as has been reported previously for AHL22 [4]. Next, we conducted RNA-seq at three different time points (ZT4, ZT9, and ZT24) to identify mis-regulated genes that may be involved in growth and development of the hypocotyl (Fig. 2). By testing three different time points throughout SD, we were hoping to uncover mis-regulation of genes that may be overlooked by only testing one time point. In fact, our data suggest that AHLs have a greater impact on gene expression during the middle of the daytime and early during the evening than at the end of the night (Fig. 2 B, C), although it is possible that there are simply fewer genes expressed in seedlings at ZT24 than at ZT4 and ZT9. Additionally, we found that the majority of genes are mis-regulated at exclusively one time point. These data suggest that AHL-regulated genes are largely affected at specific times during the day.

AHLs regulate PIFs and PIF-target genes, including the newly identified PRE1, PIL1, HFR1, CDF5, and XTR7, in SD

In order to understand more about the specific effects SOB3 and other AHLs may be having on growth, we focused on PIFs and PIF-target genes. We found growth-promoting genes known to be activated by PIFs and involved in auxin biosynthesis (YUC8, [62]) or signaling (SAURs 19/22, [25, 26, 63, 64], and IAAs 19/29, [22,23,24]) are repressed by AHLs at all time points (Fig. 2B). On the other hand, the PIFs themselves (4, 5, 7, and 8), and BES1, which is involved in BR signaling [35] are only repressed at one or two time points (Fig. 2B). Notably, BES1 and PIF4 are known to function together in a complex that activates the expression of growth-promoting genes [26, 34], and our data suggest that AHLs reduce activity of this complex particularly at ZT4 by repressing the expression of both BES1 and PIF4 at this time point (Fig. 2B). Additionally, we identified the PIF-regulated BR biosynthesis genes CPD and BR6OX2 [33, 34] as directly repressed at ZT4 (Fig. 2B, Table S4), suggesting that AHLs likely further reduce BES1-PIF4 complex activity at this time point by reducing BR biosynthesis, consequently decreasing the quantity of active BES1.

Besides CPD and BR6OX2, we have also identified five PIF target genes that are repressed and have not been previously described as repressed by AHLs: PRE1, PIL1, HFR1, CDF5, and XTR7 (Fig. 2B). PRE1 and PIL1 are repressed at all time points. PRE1 promotes PIF- and BR-mediated regulation of cell elongation [26,27,28,29,30,31]. On the other hand, PIL1, recently renamed PIF2 [65], is a PIF that functions in an atypical manner. PIL1 represses activation of gene expression, and consequently growth, by other PIF family members within a negative feedback loop [66,67,68]. In contrast to PRE1 and PIL1, HFR1 is only repressed at ZT9 (Fig. 2B). HFR1 also represses PIF-mediated activation of gene expression and growth as part of a negative feedback loop [69,70,71,72]. CDF5 and XTR7 are both repressed by SOB3 at ZT9 and ZT24, but not at ZT4 (Fig. 2B). CDF5 promotes the expression of some PIF-activated genes, including YUC8 and IAA19, and consistently also plays a positive role in hypocotyl growth [32]. XTR7 (also known as XTH15) is a cell wall-modifying enzyme that has been shown to positively regulate petiole elongation [73]. Given the many similarities between the molecular mechanisms that regulate hypocotyl and petiole growth [21], it is likely that XTR7 also promotes hypocotyl elongation; thus, the effect of the AHLs on hypocotyl growth in SD may be caused in part by their repressive effect on XTR7.

Additionally, microtubules have been shown to be involved in elongation of the hypocotyl in Arabidopsis [74, 75]. In fact, several genes encoding microtubule associated proteins (MAPs) were found to be repressed by SOB3 in our RNA-seq data. For example, MAP65-6, whose protein localizes to the mitochondria and binds microtubules [76] was repressed at ZT4 and ZT9 and bound by SOB3 at all time points. A microtubule stabilizing gene, WDL4 [77, 78], was repressed at ZT9 and bound by SOB3 at all time points. It fact, it was recently discovered that WDL4 acts to modulate auxin maxima by regulating PIN trafficking during apical hook formation [79]. Therefore, our data show that SOB3 could partially control hypocotyl elongation in SD by repressing MAPs.

Our RNA-seq data also indicate that AHLs activate the expression of growth repressors in SD. We specifically identified growth repressors discussed in Favero et al. 21, such as HY5, BBX21-22, ELF4, and IAA17, as AHL-induced (Fig. 2C). However, it is interesting to note that while Favero et al. [18] observed higher expression of BIN2 in SOB3-D compared to sob3-6 using 14-day-old plants grown in LD and harvested at ZT4, we did not find BIN2 to be differentially regulated between the two mutants at ZT4, ZT9, or ZT24. Therefore, while AHLs might inhibit growth by promoting BIN2-mediated phosphorylation and subsequently degradation of PIF3 and PIF4 [80, 81] by enhancing the expression of BIN2 in juvenile rosettes [18], AHLs likely do not inhibit growth through this pathway in seedlings, at least those growing in SD. Given that we and the Favero et al. [18] study identified RGA as up-regulated in SOB3-D compared to sob3-6 at ZT4 (Fig. 2C), AHLs might repress hypocotyl growth in SD partly by promoting DELLA-mediated PIF degradation and inhibition of PIF binding to target genes [82,83,84] specifically at this time point.

An important question is if PIF target genes are actually regulated directly by AHLs, as is suggested by both our ChIP-seq data and that from the Favero et al. [18] study. It is conceivable that the down-regulation of PIF targets that is observed in SOB3-D compared to sob3-6 might simply be a downstream consequence of less PIF activity in SOB3-D, given that we found PIF4, 5, 7, and 8 are all AHL-repressed (Fig. 2B) and AHLs might promote DELLA-mediated degradation of PIFs, as described above. However, while these mechanisms might reduce PIF activity and consequently the activation of PIF target genes at ZT4 and ZT9, they do not appear to explain the negative effect of AHLs on the expression of these genes at ZT24. Another way the AHLs could theoretically affect PIF activity at ZT24 is by increasing phyB transcription, thereby enhancing phyB-mediated degradation of these transcription factors during the night, since phyB only slowly reverts back to its inactive Pr form after dusk [85,86,87,88,89,90,91,92]. However, the only photoreceptor found to be differentially regulated in our RNA-seq data was UVR8. Therefore, it seems likely that at least at ZT24 in SD, AHLs are directly repressing the expression of PIF target genes.

SOB3: A transcription factor and a potential chromatin remodeling protein

The SOB3 binding patterns seen in our data could indicate that SOB3 is acting as a chromatin remodeling protein. The AT-hook motif was first described in the high mobility group (HMG) proteins [93,94,95], which are a family of non-histone chromosomal proteins found in a wide variety of eukaryotic organisms [96, 97]. HMGA proteins, similarly to SOB3, contain an AT-hook motif through which they bind AT-rich DNA [95, 98, 99]. Furthermore, HMGA proteins are also involved in the organization of chromatin [97] and the formation of the enhancesome [100, 101]. HMGA proteins compete with histone H1 for binding to linker DNA, inducing a loosening of chromatin structure [95, 102,103,104]. Additionally, HMGA proteins have been shown to bind to the H2A/H2B/H3/H4 nucleosome [105], indicating a role of these proteins in eviction and/or mobilization of core histones during transcriptional regulation [95]. HMGA proteins may also be involved in DNA looping and chromatin rearrangements that bring enhancesomes/promoters together so that transcription can begin [95, 100, 101]. Our SOB3 ChIP-seq binding data could indicate that SOB3 is acting similarly to HMGA proteins, where its role would be to passively regulate transcription by modifying chromatin. It is further possible that SOB3 can act as either a transcription factor or a chromatin remodeling protein depending on the specific environment in which binding occurs.

It is possible that SOB3 itself is not a chromatin remodeler, but influences chromatin remodeling through other mechanisms, such as interacting with members of chromatin remodeling complexes. These complexes are evolutionarily conserved and contain multiple subunits. Chromatin remodeling complexes can regulate chromatin structure by changing the composition of nucleosomes [106, 107]. There are four major known subfamilies of chromatin remodeling ATPases: INO80/SWR1, CHD, ISWI, and SWI/SNF [108,109,110,111,112,113,114]. SWR1 exchanges histone H2A for H2A.Z. This is particularly of interest, as warm temperature promotes eviction of H2A.Z from chromatin, which enhances the activation of PIF-targeted genes such as YUC8 [18, 62, 115,116,117,118,119]. Additionally, there is some evidence that SOB3 represses transcription of YUC9 by recruiting by increasing H2A.Z deposition on its promoter via recruitment of the SWR1 complex [8]. Interestingly, actin-related protein 9 (ARP9), which has been shown to be involved in SWI/SNF chromatin remodeling [120, 121], was found to be induced at all time points in our RNA-seq data. Additional experiments are needed to determine if SOB3 affects chromatin remodeling by interacting with chromatin remodeling complexes or by influencing the expression of chromatin remodeling proteins.

Roles of AHLs in plant development: agricultural implications

Previous research has indicated that overexpression of SOB3 and AHL20 in Arabidopsis and Camelina delays flowering [9, 10]. Further research on newly discovered SOB3-regulated genes, such as CDF5, may elucidate additional mechanisms by which SOB3 is regulating flowering time. For example, loss-of-function cdf5-1 mutants have been shown to have earlier flowering when compared to WT plants [122]. The ability to change or regulate flowering time within crop species can have many valuable advantages, such as allowing for different or increased planting seasons. Further experiments evaluating the role of CDF5 in AHL-regulated flowering time may elucidate additional mechanisms for manipulating this advantageous agronomic trait.

Conclusions

In summary, our data revealed that at three different time points in SD-grown seedlings, genes involved in growth of the hypocotyl are repressed, while growth repressors are induced by AHLs (Fig. 2). Additionally, we found that although somewhat different sets of genes appear to be regulated by AHLs at different time points, the binding patterns of SOB3 largely remain the same (Figs. 4, 5, 6, 10, and 11). It is important to note that we cannot rule out the possibility that some of the differences in gene expression observed between SOB3-D and sob3-6 in our RNA-seq experiment are caused indirectly by the phenotypic difference in hypocotyl length between the two genotypes, and thus are not directly regulated by AHLs at the time points used in this experiment. However, we suspect that at least some of the time point-specific effects of the AHLs on gene expression identified in this experiment are attributable to the composition of AHL-containing protein complexes differing at ZT4, ZT9, and ZT24. The activities of AHL complexes are perhaps modified through interactions with different non-AHL transcription factors that are important for regulating gene expression and ultimately hypocotyl growth at specific time points in SD. Alternatively, changes in the specific AHL family members present in AHL complexes at the different time points might explain the different effects of these transcription factors on gene expression at the three time points. It is also possible that AHLs regulates gene expression in SD by acting as a chromatin remodeling proteins, since the AT-hook motif is also found in the chromatin remodeling HMGA proteins. Alternatively, rather than AHLs acting as chromatin remodeler themselves in SD, they might engage members of chromatin remodeling complexes to regulate gene expression. Future studies should focus on testing these hypotheses, in order to clarify the precise molecular mechanisms that govern regulation of gene expression and hypocotyl growth by SOB3 and other AHLs.

Methods

The aim of this study was to undercover unique molecular events that promote growth at different time points during SD conditions in Arabidopsis. More precisely, we aimed to identify specific genes that are bound by SOB3 and are mis-regulated between sob3 mutants that may be involved in growth and development of the hypocotyl in Arabidopsis. The overall design contains two main experiments: ChIP-seq and RNA-seq, which are described in detail below.

Plant materials and growth conditions

The Arabidopsis (Arabidopsis thaliana) lines used in the study are all in the Columbia (Col-0) background and have been described previously: SOB3-D and sob3-6 [1]; ProSOB3::SOB3-GFP sob3-4 [5]. Seeds had been surface-sterilized and were grown on full-strength MS medium containing 0.6% Gelzan CM (Sigma-Aldrich) and 1% sucrose (w/v). Seeds were stratified in the dark at 4 °C for 3 days. Plants were grown in an MLR-351 plant growth chamber. The chamber was kept at a temperature of 22 °C and a short-day photoperiod (8 h light/16 h dark) for five-six days. 3LS was used as the light setting during the daytime, producing white fluorescent light with a fluence rate of approximately 35–55 μmol/m2/sec. For the RNA-seq and ChIP-seq experiments, seedlings were harvested at one of three time points, 4 h after dawn (ZT4), 1 h after lights off (ZT9), or at dawn the next day (ZT24), and immediately frozen in liquid nitrogen.

Arabidopsis Locus Identifiers:

  • PIF4: AT2G43010

  • CPD: AT5G05690

  • BES1: AT1G19350

  • SAUR24: AT5G18080

  • BR6OX2: AT3G30180

  • PRE1: AT5G39860

  • SAUR19: AT5G18010

  • SAUR22: AT5G18050

  • YUC8: AT4G28720

  • PIL1: AT2G46970

  • IAA19: AT3G15540

  • IAA29: AT4G32280

  • HRF1: AT1G02340

  • PIF5: AT3G59060

  • PIF7: AT5G61270

  • PIF8: AT4G00050

  • CDF5: AT1G69570

  • XTR7: AT4G14130

  • ELF4: AT2G40080

  • HY5: AT5G11260

  • BBX22: AT1G78600

  • RGA: AT2G01570

  • PRR9: AT2G46790

  • IAA17: AT1G04250

  • ELF3: AT2G25930

  • BBX21: AT1G75540

Hypocotyl measurements

Seedlings used for quantification of the sob3 mutant phenotypes were grown in the same conditions as described above. WT seedings are of Arabidopsis thaliana (L.) Col-0 ecotype. Five-day-old seedings were transferred to transparencies, which were digitized with a flatbed scanner. The transparencies included a ruler for measuring a 1 mm length to set the parameters for measurements in ImageJ (NIH). hypocotyls were measured from the top of hypocotyl to the beginning of the roots using the segmented line tool. The measurements were transferred to an Excel spreadsheet for analysis. A two-tailed Welch's t-test was used to check for significant differences in hypocotyl length between the mutants and the wild type.

RNA sequencing (RNA-seq) and data analysis

20–35 whole seedlings were harvested in triplicate at the indicated time points, with plants for different biological replicates grown on different plates. Samples for different genotypes within the same biological replicate set were harvested from the same plate(s). This experimental design was used to minimize the impact of plate-to-plate variation on gene expression. Total RNA was extracted using the RNeasy Plant Mini Kit (QIAGEN) according to manufacturer’s instructions, including on-column DNase digestion to eliminate genomic DNA. Isolated RNA was subjected to library preparation using the KAPA Stranded mRNA-Seq Kit (Kapa Biosystems) with NEBNext Multiplex Oligos for Illumina (New England Biolabs) used as adapters and Agencourt AMPure XP (Beckman Coulter) beads used instead of KAPA Pure Beads. Libraries were pooled and 84- or 86-bp, single-read sequences were obtained with an Illumina NextSeq500 sequencer. Raw data files (bcl format) were converted to fastq files by bcl2fastq (Illumina). Over 50% of reads were mapped to the Arabidopsis TAIR10 cDNA reference using Bowtie [123] with the following parameters: ‘–all –best –strata’. The total number of mapped reads per sample was 6–12 million. Differentially expressed transcripts between pairs of samples were identified using the edgeR package in R/Bioconductor [124] with an FDR cutoff of 0.05. AHL-regulated genes were defined as nuclear-encoded genes for which at least one transcript was differentially regulated between sob3-6 and SOB3-D. Venn diagrams were generated using an online tool available at http://bioinformatics.psb.ugent.be/webtools/Venn/.

Chromatin Immunoprecipitation sequencing (ChIP-seq)

The ChIP procedure included three biological replicates for each time point. Chromatin immunoprecipitation for ChIP-seq was performed essentially as described previously [18]. 1.5 g of ProSOB3::SOB3-GFP sob3-4 [5] whole seedlings were harvested at ZT4 and immediately frozen using liquid nitrogen. Samples were ground to a fine powder using an MB1200 Multi-beads Shocker (Yasui Kikai) and cross-linking performed in 1% formaldehyde for 10 min, after which the nuclear fraction was isolated. Chromatin suspended in SDS lysis buffer (50 mm Tris–HCl [pH 7.8], 1% SDS, 10 mM EDTA) was diluted approximately five-fold with ChIP dilution buffer (50 mm Tris–HCl [pH 7.8], 0.167 M NaCl, 1.1% Triton X-100, 0.11% sodium deoxycholate, 1 Roche cOmplete EDTA-free tablet/50 mL solution) and sheared at 5 °C for 15 min using a Covaris S2 Focused-ultrasonicator with milliTUBE 1 mL AFA Fiber tubes (Covaris) and the following settings: duty cycle 5%, intensity 4, and cycles per burst 200. This produced an average fragment size of approximately 100–300 bp. Sonicated chromatin was immunoprecipitated using an anti-GFP antibody (ab290, Abcam) and Dynabeads Protein A (Thermo Fisher Scientific) at 4C. Following incubation with antibody, samples were washed once with low salt RIPA buffer (50mMTris-HCl [pH 7.8], 150mMNaCl, 1mMEDTA, 0.1% SDS, 1% Triton X-100, 0.1% sodium deoxycholate, 1 Roche cOmplete Ultra tablet/50 mL solution), twice with high salt RIPA buffer (50mMTris-HCl [pH 7.8], 500mMNaCl, 1mMEDTA, 0.1% SDS, 1% Triton X-100, 0.1% sodium deoxycholate, 1 Roche cOmplete Ultra tablet/50 mL solution), once with LNDET (10 mM Tris–HCl [pH 7.8], 250 mM LiCl, 1% IGEPAL CA-630, 1% sodium deoxycholate, 1 mM EDTA) and once with TE buffer. Chromatin was eluted from Dynabeads and cross-linking reversed at 65 °C in a solution containing 10 mM Tris–HCl [pH 7.8], 300 mM NaCl, 5 mM EDTA and 0.5% SDS. Phenol–chloroform extraction followed by ethanol precipitation was used to isolate DNA. Sequencing libraries were prepared essentially as described in Rymen et al. (2019). Isolated DNA was quantitated using the Qubit dsDNA HS Assay Kit with the Qubit 2.0 Fluorometer (Thermo Fisher Scientific), and 3.65 ng of ChIPed or input (for control libraries) DNA was subjected to library preparation. Libraries were prepared using the KAPA Hyper Prep Kit for Illumina (KAPA Biosystems) and Illumina compatible adaptors (New England Biolabs). Libraries were pooled and 86-bp, single-read sequences were obtained with an Illumina NextSeq500 sequencer.

Chromatin Immunoprecipitation sequencing (ChIP-seq) data analysis

Raw data files (bcl format) were converted to fastq files by bcl2fastq (Illumina). Over 70% of reads were uniquely mapped to the Arabidopsis TAIR10 reference using Bowtie [123] with the setting ‘‘-m 1.’’ The total number of uniquely mapped reads per sample was 11–26 million. Peaks were called by comparing ChIP samples with the input using the ‘‘callpeak’’ command in MACS2 [125]. Fold-enrichment bdg format peak files were generated by using the treatment pileup and control lambda output files generated from ‘‘callpeak’’ as inputs for the MACS2 ‘‘bdgcmp’’ command with the setting ‘‘-m FE.’’ These fold-enrichment bdg files were then converted to bigWig files, which were in turn used to generate the figures showing SOB3 binding distribution or relative binding on target genes via the computeMatrix and plotHeatmap or plotProfile commands, respectively, in deepTools [126]. Peaks were visualized using the Integrative Genomics Viewer (IGV) version 2.3.88 [127, 128]. Motif analysis was performed using MEME-ChIP [129] with Arabidopsis PBM motifs selected as known motifs [56]. 300 bp nucleotide sequences centered at the SOB3-GFP ChIP-seq peak summits were used as primary input sequences for MEME-ChIP analysis. Peaks were annotated using HOMER [130]. Venn diagrams were generated using an online tool available at http://bioinformatics.psb.ugent.be/webtools/Venn/.

Quantification and statistical analysis

Where indicated, a two-tailed Welch’s t test was performed where P ≤ 0.05 = *; P ≤ 0.005 = **; P ≤ 0.0005 = ***; and P ≤ 0.00005 = ****

Gene ontology enrichment analysis was performed via BINGO [131]. GO term results were viewed and heatmaps were created using MeV 4.8.1 [132].