Abstract
Background Limitations in molecular oxygen reduce ATP production and dramatically curtail energy demanding processes in eukaryotes. Little is known of the influence of hypoxia on nuclear regulatory mechanisms and their integration with mRNA accumulation and translation. Here we apply multomic technologies to evaluate epigenetic to translational regulation in response to hypoxic stress in seedlings of Arabidopsis. We focus on hypoxia-responsive (HRG) and RIBOSOMAL PROTEIN (RP) genes that encode actively and poorly translated mRNAs under hypoxia, respectively.
Results Evaluation of hypoxia-induced dynamics in eight chromatin readouts including chromatin accessibility, histone modifications, RNA polymerase II (RNAPII) activity plus three RNA populations (nuclear, polyadenylated, and ribosome-associated) identified distinct patterns of nuclear regulation. HRGs coordinately increased promoter accessibility, Histone 2A.Z eviction, Histone 3-lysine 9 acetylation, and RNAPII engagement. Many HRG promoters were bound by HYPOXIA-RESPONSIVE ETHYLENE RESPONSIVE FACTOR (ERF) 2 (HRE2) or had cis-elements targeted by related ERFs. Hypoxia sustained RP transcription but the transcripts were largely retained in the nucleus. We discovered heat and oxidative stress genes with pronounced hypoxia-induced RNAPII engagement accompanied by elevated nuclear and ribosome-associated but not polyadenylated transcripts. These heat stress genes had cis-elements recognized by HEAT SHOCK FACTOR transcriptional activators, 5' biased histone marks, less H2A.Z eviction and Histone 3-lysine 4 trimethylation than genes bound by HRE2 and coordinately upregulated from transcription through translation. Genes of the circadian cycle, photosynthesis, and development also displayed notable nuclear regulation.
Conclusions Hypoxia triggers dominant patterns in nuclear regulatory control that differentiate cohorts of genes associated with stress responses and growth.
Background
The regulated expression of protein coding genes in eukaryotes involves processes within the nucleus including the remodeling of chromatin, recruitment of RNA polymerase II, co-transcriptional processing and the export of mature mRNA to the cytoplasm, after which it may be translated, sequestered or degraded. These processes are modulated in response to conditions that necessitate alterations in metabolism and management of energy reserves to maintain cell viability including oxygen deficiency. In Arabidopsis thaliana, cellular hypoxia promotes activation of anaerobic metabolism to sustain cell viability. This involves rapid activation in transcription followed by selective translation of a subset of cellular mRNAs, while others are transiently sequestered and stabilized whereas others are degraded [1–4]. The mRNAs that are well translated under hypoxic stress are not characterized by a sequence element or feature [1, 3, 4] but are likely to be transcriptionally upregulated by the stress.
Transcriptional activation in response to hypoxia involves the evolutionarily conserved group VII ethylene response factor (ERFVII) transcription factors (TFs) that are required for regulation of metabolism and survival of hypoxia [5–13]. The five members of this family are stabilized under hypoxia, due to attenuation of their oxygen-stimulated targeted proteolysis via the N-end rule pathway [7, 8, 12, 14]. The constitutively synthesized ERFVIIs RELATED TO APETALA 2.2, 2.3, and 2.12 (RAP2.2, RAP2.3, RAP2.12) transactivate hypoxia-responsive gene (HRG) promoters in protoplasts [13, 15, 16] through a conserved cis-acting hypoxia-responsive promoter element (HRPE) identifiable in ~50% of the 49 HRGs expressed across cell types [7, 15, 17]. The two additional ERFVIIs, HYPOXIA-RESPONSIVE ERF 1/2 (HRE1/2) only weakly transactivate via the HRPE in protoplasts [15, 18], suggesting they interact with a distinct cis-element or may otherwise influence gene transcription. Both HRE1/2 are upregulated by hypoxia along with ALCOHOL DEHYDROGENASE 1 (ADH1) and PLANT CYSTEINE OXIDASE 1/2 (PCO1/2) that are required for anaerobic metabolism and negatively regulate the anaerobic response by catalyzing the oxygen-stimulated turnover of ERFVIIs, respectively [14, 16, 19, 20].
Transcriptional activation requires access of TFs to specific cis-regulatory elements that are typically located near the 5’ transcription start site (TSS) of genes and facilitate assembly of an RNAPII initiation complex. The distribution of nucleosomes across a genome can be assessed by DNAse I hypersensitivity site mapping [21, 22] or the Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) [21]. ATAC-seq leverages a Tn5 transposon that simultaneously inserts adaptor sequences and cleaves DNA in nucleosome-depleted chromatin regions [23]. The depletion of nucleosomes relative to the 5’ TSS of each gene is controlled by a cadre of ATP dependent chromatin remodelers and can be influenced by interactions with certain TFs. The spatial distribution of nucleosomes with specific histones in their octamer core is routinely determined by chromatin immunopurification and sequencing (ChIP-seq). TF binding and RNAPII activity are influenced by and influence histone modifications and variants [24, 25]. Tri-methylated Histone 3-lysine 27 (H3K27me3) and H3K4me3 are prevalent in the gene body of lowly and actively transcribed genes, respectively [25, 26]. Acetylation of Histone H3 lysine residues (H3K9ac, H3K14ac) that reduce interactions between DNA and histones are associated with ongoing transcription [25]; these marks are also correlated with the active transcription of genes during environmental stress in plants [27–29].
Constitutive high levels of the Histone 2A variant H2A.Z are associated with stress-activated transcription and its presence at the first nucleosome within the gene body may reduce the energy required for commencement of transcriptional elongation on genes activated by elevated temperatures [30–34]. To date, there has been no genome-scale evaluation of coordinated dynamics in chromatin accessibility, histone alterations, transcription, and translation in a plant or other eukaryotes.
As transcription commences, the phosphorylation of specific residues within the heptad repeats of the carboxyl terminal domain (CTD) of RNAPII orchestrate interactions with factors that facilitate transcription-coupled histone modifications as well as co-transcriptional 5’ capping, splicing, and polyadenylation of the nascent transcript [35, 36]. RNAPII CTD phosphorylation at Serine 2 (Ser2P) demarks active elongation [37]. In animals, pausing of RNAPII downstream of the TSS is common among genes activated by heat stress [38]. Global nuclear run-on sequencing (GRO-seq) on Arabidopsis seedlings under optimal growth conditions found limited evidence of promoter-proximal pausing of RNAPII [39] but the mapping of the 5’ end of nascent transcripts by Native Elongating Transcript sequencing (NET-seq) [40] indicated that elongation can be rate-limiting shortly after initiation. Both studies found transcription is rate-limited as RNAPII pauses just beyond the site of cleavage and polyadenylation. Nuclear splicing, polyadenylation and export to the cytoplasm is regulated during development and in response to environmental cues, and hypoxia promotes alternative splicing [2, 3, 17, 41], alternative polyadenylation [42], and nuclear retention of transcripts [43] in plants. We hypothesized that the strong hypoxia-induced upregulation of HRGs and their prioritization for translation may be associated with specific features of their chromatin or transcriptional activation.
Here, we monitored dynamics in histones, chromatin accessibility, HRE2 binding and RNAPII-Ser2P distribution along gene bodies in response to hypoxic stress in seedlings. These nuclear readouts were compared with dynamics in nuclear RNA (nRNA), polyadenylated RNA (polyA RNA) and ribosome-associated RNA. An integrative analysis exposed distinct patterns of chromatin signatures, cis-element enrichment, and temporal regulation of nRNA accompanied by modulation of polyA or translated transcripts. This multiomic analysis provides a rich resource for evaluation of gene activity from epigenetic state and RNAPII activity through translation in a model organism.
Results
A multiscale dataset for analysis of dynamics from chromatin to translation
To better understand dynamics in gene regulatory variation following hypoxic stress, we developed chromatin and RNA-based genome-scale datasets in triplicate for seedlings treated with non-lethal hypoxic stress (2 h, 2HS; 9 h, 9HS) or normoxia (2NS, 9NS) consistent with prior high-throughput analysis of cell-type specific and translational regulation [2, 3, 17] (Fig. 1a; Figure S1a). The position and abundance of H3K9ac, H3K14ac, H3K4me3, H3K27me3, and H2A.Z was surveyed by ChIP-seq. The capture of nuclei by Isolation of Nuclei Tagged in Specific Cell Types (INTACT) was coupled with mapping of chromatin accessible regions by ATAC-seq [44, 45] and ChIP-seq to monitor binding of the TF HRE2 and elongating RNAPII-Ser2P. Three transcript populations were compared including nRNA obtained by INTACT (rRNA subtracted nascent transcriptome [46]), polyA RNA obtained by oligo dT affinity purification (transcriptome) and polyA RNA associated with ribosomes obtained by Translating Ribosome Affinity Purification (TRAP RNA; translatome) [3, 47]. The sequence reads mapped across annotated gene transcripts for each library type and treatment was reproducible but distinguishable by t-Distributed Stochastic Neighbor Embedding analysis (Fig. 1b) and the distribution of read abundance (Fig. 1c; Figure S1b-c). Because the ATAC-seq reads mapped primarily to non-transcribed regions (Fig 1g) these data were assayed separately. These assays of gene activity were used collectively to explore the integration of nuclear and cytoplasmic gene regulatory processes following hypoxic stress.
Stress- and growth-associated genes contrast in chromatin accessibility, histone, modifications and RNA modulation under hypoxia
For the nucleosome level evaluation, H3K9ac, H3K14ac, H3K4me3, H2A.Z, and H3K27me3 were plotted along each annotated protein-coding gene and global averages were compared for each condition (Fig. 1d; Figure S2a). H3 modifications associated with transcription (H3K9ac, H3K14ac, H3K4me3) and stress-activated transcription (H2A.Z) were enriched just 3’ of TSS and tapered off at the primary site of polyadenylation (TES), whereas H3K27me3 was distributed more evenly across gene bodies. Hypoxia had little effect on these modifications at the global level, with the exception of a slight elevation of H3K14ac and reduction of H3K4me3 near the TSS. Meaningful alterations were apparent when histone abundance and distribution was surveyed for two gene cohorts with opposing transcript accumulation and translation under hypoxia, the induced and translated HRGs (n=49) and stable but poorly translated cytosolic RPs (n=246). HRGs significantly increased in H3K9ac and decreased in H2A.Z association at 2HS, with limited change in these marks on RPs (Fig. 1e, f; Figure S3a-c). We extended the stress to 9 h (9HS) to evaluate H3K9ac and the more slowly changing H3K14ac associated with active transcription. At the global scale, hypoxia increased H3K9ac on 863 and 3,646 genes and H3K14ac on 2 and 1,216 genes and after brief and prolonged hypoxia, respectively (Figure S1b; Table S2). Notably, H3K14ac but not H3K9ac levels progressively increased on RPs (Fig. 1f; Figure S1c; Figure S3a, c).
Next, we evaluated dynamics in chromatin accessibility in genic regions by use of ATAC-seq. Accessibility in 5’ promoter regions is associated with depletion of nucleosomes and binding of TFs and other transcriptional machinery [45, 48, 49], whereas accessibility in 3’ flanking regions is associated with formation of chromatin loops and transcriptional termination [50, 51]. Our ATAC-seq reads mapped almost exclusively within 1000 bp 5’ of the TSS and ~300 bp 3’ of the major TES of genes in all four conditions (Fig. 2c). At 2HS the summative ATAC-seq peak height was higher than at 2NS, with the opposite trend observed for 9NS and 9HS (Fig. 2a). With respect to the opposingly regulated HRGs and RPs, the average ATAC signal within promoters rose by 1.8- and 1.4-fold by 2HS, respectively (Fig. 2b). Chromatin remodeling by the stress was evident from the expansion into more 5’ and 3’ flanking regions of the HRGs at both 2HS and 9HS. We also performed peak-calling to recognize Tn5 hypersensitive insertion sites (THSs) in order to quantify localized dynamics in chromatin accessibility, identifying both constitutively present (8,072) and stress-specific THSs (25,795) (Fig. 2c-f; Table S1a).
To complete our integrative survey of gene activity, we compared the distribution of the log2 fold change (FC) (2HS/2NS) values of the chromatin, RNAPII and RNA readouts genome-wide and for the HRGs and RPs (Fig. 3a; Table S2a). We found that stress-induced RNAPII-Ser2P engagement was positively correlated with increased H3K9ac and negatively correlated with H2A.Z eviction as demonstrated by the HRGs (global R = −0.27, HRGs R = −0.79) (global R = 0.43; HRGs R = 0.17; Fig. 3 b, c; Figure S4). RPs displayed minor changes in RNAPII-Ser2P association at 2HS, yet H3K14ac and nRNA abundance was elevated at the majority of these genes by 9HS (Fig. 3a; Figure S3c; Table S2a). These results demonstrate patterns of stress-induced epigenetic regulation distinct to stress-induced and growth-associated genes.
Regulation at nuclear and cytoplasmic scales can be concordant or discordant
To explore possible coordination of transcriptional and posttranscriptional gene regulation in response to hypoxia we performed a systematic analyses of the Ser2P, nRNA, polyA and TRAP RNA datasets. First, we surveyed whether significantly up- and downregulated genes (DRGs) in one readout were similarly regulated in other readouts. Second, we performed gene clustering to group genes that were co-regulated based on one or more readout.
A four-way comparison of DRGs identified nRNA with the greatest (1,722) and polyA RNA the fewest (602) up-DRGs (Fig. 4a). The vast majority (92%) of the polyA up-DRGs were upregulated in at least one other readout, whereas only 213 (13%) were significantly elevated in all four readouts. Notably, the coordinately upregulated genes included 38 of the 49 HRGs. A parallel analysis found 72% of the polyA down-DRGs were significantly reduced in at least one other dataset, with nRNA again displaying the greatest number of DRGs (down-DRGs: Ser2P [206], nRNA [2,608], polyA [291], TRAP [578]) (Fig. 4b). The lack of consensus in nRNA and polyA transcript dynamics suggests these RNA populations of are not equivalent. Indeed, a pairwise comparison of nRNA and polyA RNA abundance in the 2NS and 2HS samples identified transcripts enriched in the nRNA (nucleus-localized transcripts) and polyA RNA fractions (Figure S5a, b; Table S3).
Patterns of differential gene regulation were resolved further by identifying genes that were co-regulated in response to 2HS. Log2 fold change (FC) values for all DRGs identified in at least one of the four datasets were sorted into 16 groups by partition around medoids (PAM) clustering and evaluated by Gene Ontology (GO) term enrichment (Fig 4c; Table S4a). This identified four clusters of coordinately upregulated (clusters 1-4) and one cluster of coordinately downregulated (cluster 16) genes. Other clusters differed in dynamics in one or more of the readouts. These analyses confirm that hypoxia impacts gene regulatory processes in both the nucleus and cytoplasm. This finding motivated us to consider whether changes in histones were characteristic of the patterns observed in RNAPII-Ser2P and RNA accumulation.
Patterns of histone alteration are associated with concordant and discordant regulation
To explore associations between regulatory variation at the chromatin, RNAPII and RNA scales, the average signal value for each data type of each cluster was plotted along genic regions (Fig. 4d; Figure S6). To appreciate the dynamic regulation of these genes over time, the same clusters were plotted using the 9NS and 9HS chromatin and RNA data (Fig. S7). We found the coordinate (concordant) upregulation of cluster 1-3 gene transcripts at 2HS was accompanied by enhanced chromatin accessibility, eviction of H2A.Z, as well as elevation of H3K9ac and RNAPII-Ser2P across the gene body. The coordinate rise in TRAP RNA level indicated these transcripts were translated in proportion to their increased abundance, as determined previously for the HRGs by high-resolution ribosome footprinting analyses [3]. Nearly all HRGs were present in the first two clusters, which were enriched for the the GO categories decreased oxygen level (cluster 1, p-adj. <6.73e−22) and general stress (cluster 2, <1.06e−21) (Table S4a). Genome browser views of two representative HRGs, ADH1 and PCO2, illustrate the coordinate upregulation across the assays of gene activity (Fig. 5a, b).
Cluster 4-8 genes were not coordinately upregulated, indicating they are targets of post-transcriptional regulation. The rise in cluster 4 nRNA coincided with elevation of H3K9ac and RNAPII-Ser2P that is indicative of increased transcription, yet this was not accompanied by a similar rise in polyA or TRAP RNA (Fig. 4c; Figure S6). This discordant pattern could reflect nuclear retention or cytoplasmic destabilization of these transcripts. Clusters 6 and 7 genes were elevated primarily in nRNA at 2HS, a pattern that could reflect nuclear retention of previously synthesized transcripts in response to the stress. Yet another pattern was observed for Cluster 8, which was enriched for genes encoding RPs (<8.06e−16). As noted for the RP cohort (Fig 3a, b; Figure S1b), these genes had limited change in H3K9ac but a rise in H3K14ac and nRNA at 2HS and 9HS (Figures S6 and S7), as illustrated by RPL37B (Fig. 3a; Fig. 5d; Figure S1b).
One of the most intriguing clusters was 9, which was notably distinct for the hypoxia-induced engagement of RNAPII-Ser2P without a similar rise in nRNA or polyA RNA. By comparison to cluster 1, cluster 9 had lower and more 5’ skewed H2A.Z at 2NS that was only slightly reduced at 2HS (Fig 4d; Figure S9a). In addition, H3K4me3 association was significantly lower for cluster 9 than cluster 1 at 2HS (Figure S9c, d). The status of cluster 9 gene regulation changed between 2HS and 9HS, at which time their promoter region accessibility was elevated, along with H3K9ac, H3K14ac, nRNA and polyA RNA (Figure S7). Both cluster 2 and 9 genes were enriched for heat (cluster 2, <7.74e−14; cluster 9, <6.20e−13) and oxidative stress (cluster 2, <1.13e−12; cluster 9, <8.23e−11). The cluster 2 gene HEAT SHOCK PROTEIN 70-4 (HSP70-4), as an example, displayed a 3.4-fold increase in RNAPII-Ser2P across the genic region with a very limited increase in polyA RNA at 2HS but a pronounced increase by 9HS (Fig. 5c). Notably, TRAP mRNA levels were higher than polyA RNA for HSP70-4 at 2HS, as observed for ADH1 and PCO2.
Highly downregulated genes have distinct stimulus-regulated histone modifications
Survey of the down-DRG genes (clusters 10-16) also showed stimulus-driven dynamics of chromatin, RNAPII and RNAs (Fig. 4a; Figure S6). These clusters displayed two dominant patterns. Cluster 10-14 were distinguished by significant reductions in nRNA at 2HS to 9HS that were accompanied by progressive increases in H3K14ac but limited changes in H2A.Z and H3K9ac. By 9HS, levels of H3K14ac had clearly increased in all but cluster 13, particularly near the TSS (Figure S6; Figure S7). Cluster 15 and 16 genes were strongly and coordinately reduced in RNAPII-Ser2P engagement and all three RNA readouts at 2HS (Fig. 4a; Figure S6). This was accompanied by slight elevation of H2A.Z incorporation and a strong decline in H3K9ac, all antithetical to the changes observed for cluster 1-3 (Figure S9a, c). These genes also had limited change in chromatin accessibility. Of the down-DRGs, only clusters 15 and 16 were strongly enriched for GO categories, including root development (cluster 15, <3.7e−8) and regulation of transcription and RNA biosynthesis (cluster 16, <7.78e−7). The reduced transcription and translation of these genes would contribute to energy conservation during hypoxia, as reported for the RPs [2]. Despite their opposing regulation, we noted that the most highly up-(cluster 1-2) and down-DRGs (cluster 15-16) were characterized by H3 modifications (H3K4me3, H3K9ac, H3K14ac) broadly distributed along the gene body, as opposed a more 5’ biased distribution of H3 and H2A.Z in the other clusters. This indicates that H2A.Z distributed across a gene body limits chromatin accessibility and engagement of RNAPII under non-stress conditions.
H3K14 acetylation accompanied by limited H3K9 acetylation is characteristic of synthesized but nuclear-retained transcripts
To assess the impact of prolonged sub-lethal hypoxic stress we monitored the progressive change in H3K9ac, H3K14ac, nRNA, and polyA RNA at 2HS and 9HS by evaluation of log2 FC distribution again using multiple methods to visualize these genome-scale data (Fig. 6a, b; Figure S8; Table S2a). H3K9ac and H3K14ac levels were progressively and distinctly altered for the HRG and RP cohorts (Fig. 6a). We observed a coordinated rise in H3K9ac, nRNA and polyA RNA at 2HS and 9HS for clusters 1b-2b, identifying over 360 genes (38 HRGs) that were progressively upregulated. By contrast, we resolved patterns of discordant up- and downregulation between clusters 4b-11b, confirming that regulation of these changed as hypoxia was prolonged. Many genes with elevated nRNA at 9HS were enriched for H3K14ac near their TSS (clusters 2b-8b) (Figure S8), as exemplified by the RPs (Fig. 1f). Of the 228 cytosolic RPs monitored, 149 significantly increased in nRNA abundance at 9HS. Their progressive increase in nRNA demonstrates that genes scored as unchanged or downregulated by polyA transcriptomics may be actively transcribed during the stress but retained in the nucleus as near-full length (non-polyadenylated) nascent or mature transcripts. Other genes were more up-(cluster 6b) or downregulated after 9HS (clusters 8b-11b). Pronounced downregulation at 9HS included the GO categories photosynthesis (cluster 8b, <5.8e−7 and <1.5e−6), root development (cluster 10b, <1.7e−6) and transcription (cluster 11b, <7.6e−6) (Table S4b). A key finding was that regulation of transcript abundance may be distinct within the nucleus and the cytoplasm, where the vast majority of polyA RNA is located and undergoes translation, sequestration, or degradation.
Notably, genes associated with phasing of the circadian clock were perturbed by prolonged hypoxic stress. This included maintenance of mRNAs with peak abundance towards the end of the light cycle (cluster 6b; evening genes, circadian rhythm <7.9e−10) and reduced upregulation of morning gene transcripts. For example, TIMING OF CAB EXPRESSION (TOC1) and other evening gene polyA RNAs remained elevated after 9HS (Fig. 6c, d). By contrast, transcripts of the key morning expressed clock regulators CIRCADIAN CLOCK ASSOCIATED 1 (CCA1) (cluster 16b) and LATE ELONGATED HYPOCOTYL (LHY) (cluster 15b) as well as the midday PSEUDO RESPONSE REGULATOR 7/9 (PRR7/9) were dampened in nRNA and/or polyA RNA at 9HS. Thus, hypoxia imposed in at the end of the day extremely delays or arrests the phasing of the circadian cycle. Low energy stress is known to extend the circadian cycle via SnRK1-mediated enhancement of PPR7 transcriptional activation just prior to dawn [52]. This mechanism, however, does not seem to be relevant when hypoxia is initiated at the end of the light cycle and continued overnight, as PPR7 nRNA and polyA RNA was significantly lower at 9HS relative to 9NS. This difference between the effect of hypoxia and conditions that invoke an energy stress could be our imposition of hypoxia at the end of the light cycle or inclusion of sucrose in the medium.
Many coordinately hypoxia-upregulated genes are targets of ERFVII regulation
We hypothesized that the 215 coordinately up-DRGs might be transcriptionally activated by the low-oxygen stabilized ERFVIIs. We therefore searched for the presence of the HRPE in genes of each cluster resolved in the 2HS analysis (Fig. 4c) and found it was significantly enriched in cluster 1 promoters (Fig. 7a; Table S5), including the genes with the strongest and most coordinate upregulation. The constitutively expressed ERFVIIs (RAP2.2, RAP2.3 and RAP2.12) activate transcription via the HRPE in transfected protoplasts assays and based on targeted ChIP-qPCR [15, 16], whereas the hypoxia-induced ERFs HRE1 and HRE2 do not. Of these, HRE2 is encoded by an HRG and is stabilized in seedlings after 2HS [7]. To gain more insight into the role of HRE2, ChIP-seq was performed on 2HS seedlings overexpressing a stabilized version of this ERFVII. Over 75% of the reads mapped within 1 kb upstream of TSSs, with clear enrichment of binding within 500 bp of the TSS (Figure S10; Table S2a). Binding of HRE2 to promoter regions was significant for clusters 1-3 (Fig. 7b). An unsupervised search for motifs enriched in HRE2 peak regions discovered an overrepresented multimeric 5’-GCC-3’ element (p-value <1e−351) (Fig. 7c; Table S5). This motif did not correspond to the HRPE, rather it resembled the GCC-rich EBP box bound by ERFs including ERFVII RAPs. Indeed, RAP2.12 bound a nearly identical motif based on the in vitro DNA affinity purification and sequencing (DAP-seq) assay [53] and stabilized RAP2.3 associated in vivo with promoter regions with this element based on ChIP-qPCR [54] and bound in vitro to DNA oligos with this sequence in electrophoretic mobility shift assays [55]. Although the HRE2 motif was enriched in cluster 1, it was more prevalent than the HRPE in other clusters (Fig. 7c), leading us to consider HRE2 binding to a promoter and HRPE presence over the HRE2 motif presence as relevant to hypoxia. The binding of HRE2 to the promoter was observed for 60 of the 49 coordinately upregulated genes, of which 23 contained an HRPE (Fig. 7e; Table S2a). Overall, 50% of the 215 coordinately up-DRGs were bound by HRE2 in their promoter region or are likely RAP-ERFVII targets based on the presence of an HRPE in their promoters (Fig. 7e). These results support the conclusion that many of the coordinately up-DRGs are transcriptionally activated by one or more ERFVII.
Hypoxic-stress progressively activates genes in heat and oxidative stress networks
The regulatory variation of oxidative stress and heat response genes displayed at 2HS in clusters 2-3 and 9 was of interest as genes in these categories are upregulated by rapid onset anoxia or prolonged hypoxia [2, 3, 17, 41, 56]. Both the onset of hypoxia and reoxygenation trigger a reactive oxygen species (ROS) burst that contributes to signaling [57]. The upregulation of RESPIRATORY BURST OXIDASE (RBOH) genes including RBOHD contributes to survival of hypoxia, submergence and post-submergence recovery [58–61]. The relationship between hypoxia and heat stress is not clearly understood but a brief heat shock protects Arabidopsis seedlings from anoxia via cooperative action of the HEAT SHOCK FACTOR (HSF) transcriptional activators HSFA1A/B and HSFA2 [56, 62]. Elevation of RNAPII-Ser2P without a commensurate increase in polyA RNA was characteristic of HSPs, RBOHs and transcriptional co/activators associated with heat and oxidative stress (i.e., MULTIPROTEIN BRIDGING FACTOR 1C (AtMBP1C), DEHYDRATION-RESPONSIVE 2A (DREB2A), ZINC FINGER PROTEIN 12 (ZAT12)).
Because strong and coordinate upregulation of genes in response to hypoxia was associated with activation by ERF-VIIs, we considered that the more gradual activation of cluster 9 genes may be mediated by members of HSF transcriptional activators. This led to the discovery that clusters 2 and 3 included several HSFs that were elevated in TRAP RNA at 2HS (HSFA2, HSFA4A, and HSFA7A) and HSF cis-elements (HSEs) were significantly enriched in promoters of cluster 2, 3 and 9 genes (Fig. 7d). The pronounced increase in RNAPII-Ser2P engagement on these genes at 2HS corresponded with a rise in nRNA but limited elevation of polyA RNA at either 2HS and 9HS (Figure S11). These genes were distinguishable from the HRE2-bound up-DRGs by a (1) 5’ bias in nucleosomes across gene bodies, (2) more 5’ localized H2A.Z that was less dramatically evicted by 2HS, (3) reduced H3K4me3, (4) increased H3K14ac that preceded the rise in H3K9ac, (5) and an increase in nRNA not necessarily mirrored in polyA RNA. Collectively, these results resolve patterns of nuclear gene regulatory control that distinguish stress-activated genes regulated by different transcriptional activators.
Discussion
Gene regulation in response to hypoxic stress involves regulation of chromatin, transcription, and post-transcriptional processes in the nucleus and cytoplasm
This study assembled a dataset of genome-wide measurements of chromatin state and intermediary steps of gene expression in seedlings exposed to sub-lethal hypoxic stress and identified time-dependent chromatin and transcript dynamics, the latter not captured by standard transcriptomics. We found that hypoxic stress modulates the (1) position and degree of open chromatin near the TSS, (2) trimethylation and acetylation of H3 lysines, (3) eviction of H2A.Z, (4) engagement of RNAP II-Ser2P, and (5) the abundance of nuclear, polyadenylated and ribosome-associated gene transcripts. Here, we demonstrate that steady-state nuclear and polyA transcriptomes are distinct, as shown by a similar comparison for rice [46]. Our findings of translational regulation are also consistent previous reports [1, 63, 64]. The use of RNAPII-Ser2P association as a proxy for transcriptional elongation suggests that global levels of transcription were not dramatically dampened by 2HS (Fig. 1c), although there was considerable regulation of individual genes (Fig 1c; Fig. 4c). Remarkably, we discovered that less than 10% of all DRGs were coordinately up- or downregulated at 2HS from transcription (RNAPII-Ser2P) through translation. Four patterns of discontinuous gene regulation were apparent: (1) partial upregulation, characterized by high RNAPII-Ser2P engagement and delayed elevation of polyA RNA, as observed for many heat and oxidative stress genes, (2) compartmentalized downregulation, characterized by maintained RNAPII-Ser2P engagement and increased nRNA abundance with limited change in polyA RNA, as observed for RPs; and (3) enhanced or reduced translational status, measured by comparison of TRAP RNA and polyA RNA abundance as described previously [2, 3, 17]. We also identified variations in nuclear regulation (chromatin accessibility, histone modification, RNAPII engagement and nRNA abundance) that were a talisman of changes in polyA RNA abundance and translation.
Coordinate upregulation of hypoxia-responsive genes is characterized by nucleosome dynamics associated with histone modification, histone variant eviction, and increased chromatin accessibility
Hypoxia affected the position, composition, and modifications of specific histones of nucleosomes. The genes that were coordinately upregulated at the level of RNAPII-Ser2P engagement through ribosome association included the HRGs critical to low-oxygen stress survival [7, 13, 15–17, 65]. The strong and coordinate upregulation of HRGs was characterized by pronounced eviction of H2A.Z and enrichment of H3K9ac across the gene body as well as elevation of chromatin accessibility RNAP-Ser2P engagement, nRNA, polyA and ribosome-associated mRNAs (Fig. 8). The co-transcriptional histone acetyltransferase (HAT)-catalyzed placement of a positive charge on the N-terminus of H3K9 is considered a reliable signature of active transcription [27] and is common for stress-activated genes [28, 29, 66].
In animals, the H3K9ac modification is promoted by the presence of H3K4me3 and stimulates the recruitment of the super elongation complex for release of poised RNAPII complexes [38]. Although H3K4me3 generally rose on HRGs in response to 2HS, it was not as strongly correlated with RNAPII-Ser2P engagement as H3K9ac (Fig 3b; Figure S4d), raising the possibility that the co-upregulated genes may be activated by releasing polymerase pausing. This could relate to the stress-activated eviction of H2A.Z that is broadly distributed across these gene bodies at 2NS. The presence of H2A.Z in nucleosomes of gene bodies is associated with limited transcription [30, 31] and its eviction is characteristic of temperature-responsive genes [30–34].
The mechanisms of H2A.Z deposition, eviction, and the impact on transcriptional regulation are complex and not fully understood. This process requires ACTIN-RELATED PROTEIN 6 (ARP6), a component of the SWR1 remodeling complex complex, and other factors in Arabidopsis [26, 67]. Yet disruption of H2A.Z deposition in an arp6 mutant has limited effect on regulation of HRGs under aerated growth conditions [34]. Other chromatin remodeling proteins or specific TFs could be important. In response to small increases in temperature; displacement of H2A.Z was facilitated by HSF1A in the activation of the heat response network [32]. H2A.Z eviction from HRGs could involve ERFVIIs. Indeed, RAP2.2 reportedly interacts with the SWI-SNF remodeler BRAHMA leading to the hypothesis that it that may contribute to H2A.Z eviction or chromatin remodeling associated with increased chromatin accessibility [68]. Of the >200 coordinately up-DRGs, 69 had one or more HRPE within 1 kb of their TSS and many bound by HRE2 at 2HS (Fig. 7e; Table S2). RAP2.12 is stabilized and nuclear localized as external oxygen levels decline below 10 kPa [69]. The stabilization of RAP2.12 and other ERFVIIs may coordinate RNAPII initiation and facilitate H2A.Z eviction to promote productive transcription that is coupled with H3K9 acetylation. If ERFVIIs interact with the ATP-dependent chromatin remodeling machinery, they could also contribute to the stabilization or expansion of regions of accessible chromatin to facilitate high levels of transcriptional activation, resulting in transcripts that are efficiently processed, exported and translated under hypoxia. This continuum of effective upregulation from transcription to translation is reminiscent of the production of translationally competent mRNAs during nutrient starvation in yeast that is associated with specific transcription factors [70].
Genes associated with heat and oxidative stress are progressively activated by hypoxia
The stress distinctly activated genes associated with heat and oxidative stress. The progressive upregulation of genes in cluster 9 (Fig. 4c; Figure S7) became more evident when their activity was compared at 2HS and 9HS with two chromatin and two RNA readouts (Fig. 7g). A significant proportion of these genes had HSEs within their promoters (Fig. 7d). Their pattern of regulatory variation included a 5’ bias in histone marks and moderate to very high RNAPII-Ser2P engagement at 2HS accompanied by elevated nRNA and TRAP RNA, but little to no fulllength polyA RNA (Fig. 8). The increase in H3K9ac, H3K14ac and nRNA confirms these are progressively upregulated by hypoxia, whereas the pronounced 3’ bias in polyA RNA suggests susceptibility to 5’ to 3’ degradation, as evident for HSP70-4 (Fig. 5) and other members of cluster 9 (Figure S11).
Previously, heat stress-responsive genes were recognized as highly induced in Arabidopsis seedlings directly transferred to anoxia [56, 62]. Remarkably, a brief pre-treatment with heat stress increased the resilience to anoxia of wild-type seedlings but not loss-of-function hsfa2 or hsf1a hsf1b mutant seedlings [71]. This is enigmatic because high temperatures do not typically precede flooding in natural environments, but the availability of HSPs could reduce cellular damage under extreme stress conditions such as anoxia. Indeed, HSPs were plentiful in cluster 9. The limited synthesis of ATP-dependent chaperones during early hypoxia by nuclear sequestration/physical compartmentalization of their transcripts could minimize demands on limited energy reserves. However, as the stress is prolonged or upon reoxygenation, the synthesis of chaperones and proteins that provide protection from ROS could aid survival.
The nucleosome dynamics of cluster 2 and 9 genes included similar modifications with notable exceptions (Fig. 4c, d; Figures S7 and S11). Genes of both clusters underwent a loss of H2A.Z near the TSS of the gene body in conjunction with increased H3K9ac, but cluster 9 showed a loss of H3K4me3 proximal to the TSS that was not evident for cluster 2 or other genes with strong coordinate upregulation. The decrease in H3K4me3 may be an indication of non-productive RNAPII-Ser2P engagement, as this modification is associated with RNAPII elongation [72, 73]. We found a significant enrichment of HSEs in the promoters of both rapidly and progressively upregulated genes (Fig. 7d). The heat-mediated upregulation of HSFA2 (cluster 2) but not HSFB1 (cluster 9) is dependent on recruitment of ANTI-SILENCING FUNCTION 1 that was associated with RNAPII engagement [74]. We predict that recruitment of specific TFs and chromatin modifying enzymes may influence histone or RNAPII CTD modifications that have ramifications on the rate of RNAPII elongation, termination, or post-transcriptional processing, thereby tuning the timing and production of cohorts of stress induced transcripts. The progressive upregulation of heat and oxidative stress networks could be mediated by the upregulation of TF genes encoding HSFs (i.e., HSFB2B) and ZINC FINGER PROTEIN 10 (ZAT10), and others during the early hours of the stress. Our datasets provide an opportunity for further meta-analyses of heat and oxidative stress gene regulatory networks.
Transcription of genes associated with major cellular processes during hypoxia
Intriguing patterns of transcriptional and RNA regulation of genes associated with cellular housekeeping became evident from our multi-tier evaluation (Fig. 8). Our prior comparisons of the transcriptome and translatome determined that hypoxic stress limits the investment of energy into the production of abundant cellular machinery such as ribosomes. mRNAs encoding RPs and some other abundant proteins are maintained but poorly translated under hypoxic stress [1–4]. This stress-limited translation of RPs is associated with their sequestration in UBP1C-associated macromolecular complexes [4]. Upon reoxygenation, RP mRNAs became rapidly associated with polysomes [2–4]. Although nuclear run-on transcription assays demonstrated that transcription of housekeeping genes continues during hypoxia in roots of maize seedlings [75], it was not known if RPs or other UBP1C-sequestered mRNAs continue to be synthesized during hypoxia in Arabidopsis seedlings. Here, the profiling of RNAP-Ser2P and nRNA confirmed that genes encoding RPs, metabolic enzymes and some photosynthetic apparatus are synthesized and retained as partially or completely processed transcripts in the nucleus during hypoxia (Fig. 6, e.g., cluster 8b), regulatory variation characterized by progressive increase in H3K14ac but not H3K9ac. The cause of the limited increase in polyA RNA is unknown but could be due to limited cleavage and polyadenylation. The uncoupling of the H3K9ac and H3K14ac marks associated with active transcription but limited polyA RNA accumulation was reported in pluripotent embryonic stem cells of mice on a cohort of promoters deemed inactive but primed for future activation [76]. We propose that many housekeeping genes continue to be transcribed and accumulate in the nucleus during the stress. This compartmentalized downregulation may poise cells for the observed rapid recovery of the transcriptome and translatome upon reoxygenation [2, 4].
Our genome-scale analysis support evidence of nuclear retention of transcripts that was visualized in response to hypoxic stress in Arabidopsis and lupine [43]. Moreover, mRNAs associated with the cell cycle are enriched in the nucleus under control conditions in Arabidopsis [77]. Our comparison of the nRNA and polyA RNA transcriptomes (Figure S5) validated the bias in nuclear enrichment of transcripts associated with the cell cycle (Figure S5), firmly indicating nuclear export is a point of regulation under control and stress conditions. During hypoxia, enhanced nuclear compartmentalization of the abundant transcripts associated with ribosome biogenesis, primary metabolism and photosynthesis would contribute to energy conservation by limiting translation, similar to the cytoplasmic sequestration associated with UBP1C. Nuclear retention and cytoplasmic sequestration could provide two reservoirs of transcripts that can be deployed upon reaeration, enabling rapid restoration of cellular processes. The RNA helicase eIF4AIII [78] as well as the RNA binding proteins UBP1A/B/C and the HRG-encoded CML38 form macromolecular ribonucleoprotein complexes that are primarily cytoplasmic but also present within the nucleus in response to hypoxia [4, 79]. These complexes could orchestrate the nuclear co- and post-transcriptional processes during the stress. Hypoxia promotes alternative splicing including intron retention [3] and alternative polyadenylation site selection [42] and some of these transcripts are found in association with ribosomes. Nonetheless, there is no clear evidence of a general disruption of splicing or polyadenylation by hypoxia [3, 42, 78]. Further experimentation is required to understand the mechanisms responsible for the distinctive elevation in H3K14ac and nuclear-retained transcripts observed for many genes under hypoxic stress.
An ERFVII hierarchy drives transcriptional activation in response to hypoxia
This study included the first genome-wide analysis of the cis-element targets of an ERFVII. Enigmatically, ChIP-seq of the hypoxia-induced ERFVII HRE2 led to the identity of a multimeric GCC cis-element that did not correspond to the HRPE found in many HRGs and sufficient for transactivation by RAP ERFVIIs in protoplasts [15]. The HRE2 binding motif strongly resembled the double GCC motif of the EBP box first defined for an ERF of Nicotiana tabacum [80].
Strikingly, RAP2.3 was shown to bind the ACID INSENSITIVE 5 (ABI5) promoter in the region of its two EBP boxes and transactivation of the ABI5 promoter was demonstrated in protoplasts by all three constitutively expressed ERFVIIs (RAP2.2/2.3/2.12) in an EBP box-dependent manner [54]. It remains to be determined if HRE2 binds directly to the HRPE in planta, as neither HRE1 or HRE2 interacted with a 33 bp region containing the HRPE in a Yeast-1 hybrid assay, and both provided 40-fold lower transactivation via the HRPE in protoplasts than the RAP-ERFs [15]. This supports the conclusion that these these two subgroups of ERFVIIs are not genetically redundant. Yet the predicted HRPE contains 5’-GCC-3’ followed by a lower frequency 5’-GCC-3’ and perhaps more significantly, the 33 bp sequence used to define the HRPE contained a tandem 5’-GCC-3’. In addition we found that promoter regions with high scores for the HRPE and HRE2 motif coincided in 40 of the 213 coordinately upregulated genes, of which 13 were bound at 2HS by HRE2. Finally, the region of the Arabidopsis ADH1 promoter that is necessary and sufficient for hypoxic upregulation contains both of these motifs, as discussed by [15]. Our data do not conflict with the model that the low-oxygen-mediated stabilization of the constitutively-expressed RAP-type ERFVIIs is responsible for rapid transactivation of HRGs that include HRE2 [81, 82], which is synthesized and stabilized under hypoxia [7]. It seems likely that HRE2 production and binding to promoters may bolster upregulation of HRGs, as an hre1 hre2 double mutant fails to maintain the upregulation of these genes after 4 h of anoxia [83]. Still, further analyses are required to more fully comprehend the independent or overlapping roles of the low abundant HRE2 and RAP-type ERFVIIs in the temporal activation of transcription in response to hypoxia.
In mammalian systems, the response to hypoxia involves regulation of histone modifications and transcriptional activity [84]. The similarity to plants is remarkably analogous. Hypoxia-responsive genes are activated by the two-subunit Hypoxia Inducible Factor (HIF) transcription factor complex, that includes the oxygen-destabilized HIF1A subunit [85]. Similar to the three RAP-ERFVIIs, HIF1A is a constitutively synthesized protein that is unstable when oxygen is replete. It is the oxygen-dependent activity of a group of prolylhydroxlases that govern the hydroxylation of specific residues of HIF1A that triggers its oxygen-dependent ubiquitylation and protoeasome-mediated turnover. In animals, the chromatin landscape is additionally regulated through chromatin accessibility [86] and the post-translational modification of histones [87] in an oxygen-sensitive manner. It has been shown that HIF signaling itself directly mediates histone modification of target genes through the interaction with chromatin modifying enzymes [84]. Additionally, several histone modifying enzymes are directly targeted by HIF transcriptional activation, and HIF1A expression itself is regulated by histone acetylation [88], confirming the integration of epigenetic regulation. The analogy of low-oxygen stabilized transcriptional regulators and their coordination with chromatin modifications hints of convergent mechanisms of hypoxia signaling in plants and animals. Indeed, this may extend in a concerted manner beyond transcriptional regulation as cytoplasmic sequestration and selective translation of mRNAs is also observed in plants and metazoa [2, 3, 89, 90]. It remains to be explored if hypoxia influences the synthesis and accumulation of nascent transcripts associated with growth in animals, as demonstrated in this multiomic analysis of Arabidopsis.
Conclusions
This deep investigation into gene regulatory processes in hypoxic stress revealed distinct regulation within the nucleus that fine tune cellular and physiological acclimation to transient deficiencies in oxygen availability. We determined that rapid regulation of the epigenome in response to short-term hypoxic stress continues as the stress becomes more severe.
Coordinate regulation from chromatin accessibility, to RNAPII engagement, polyadenylated mRNA accumulation, and translation is observed for over 200 hypoxia-upregulated genes, while more discontinuous upregulation of nascent transcript production, export to the cytoplasm, and recruitment to ribosomes is observed for other gene cohorts. These findings uncover dynamic patterns of nuclear gene activity control that operate in response to an environmental challenge.
Methods Genetic Material
The following genotypes were used for genome wide experiments: Col-0, Histone modification ChIP-seq, RNAP II ChIP-seq; pHTA11:HTA11-3xFLAG [32], H2A.Z ChIP-seq; pUBQ10:NTF/pACT2:BirA [21], ATAC-seq, nRNA-seq, polyA mRNA-seq; p35S:HF-RPL18 [47], TRAP-seq, mRNA-seq; and HRE ChIP-seq p35S:C2A-HRE2-HA [7].
Growth and Treatment Conditions
Arabidopsis seeds were surface sterilized by incubation in 70% EtOH for 5 min, followed by incubation in 20% (v/v) bleach, 0.01% (v/v) TWEEN-20, followed by three washes in ddH2O for 5 min, in triplicate. Sterilized seeds were placed onto 1x MS media (1.0x Murishige Skoog (MS) salts, 0.4% (w/v) Phytagel (Sigma-Aldrich), and 1% (w/v) Suc, pH 5.7) in 9 cm2 Petri dishes and stratified by incubation at 4°C for 3 d in complete darkness. Following stratification, plates were placed vertically into a growth chamber (Percival) with 16 h light / 8 h dark cycle at ~120 μmol photons•s−1•m-2, at 23 °C for 7 d.
Seedling Treatments
For hypoxic stress, seedlings were removed from the growth chamber at Zeitgeber time (ZT) 16 and were subjected to hypoxic stress by bubbling argon gas into sealed chambers in complete darkness for 2 or 9 h at 24 °C. The chamber set-up was as described by [2]. Oxygen partial pressure in the chamber was measured with the NeoFox Sport O2 sensor and probe (Ocean Optics). Hypoxia ([O2] < 2%) was attained within 15 min of initiation of the stress. Control samples were placed in an identical chamber that was open to ambient air under the same light and temperature conditions. Post hypoxia aeration was achieved by removing plates from the chamber and placing them into identical chambers in ambient air. Tissue was rapidly harvested into liquid N2 and stored at −80°C.
Chromatin Immunopurification (ChIP)
ChIP was performed according to [44] with minor modifications. Seedlings were grown on Petri dishes and treated as described above. For the HRE2 ChIP experiments, seedlings were pre-treated by flooding with 10 mL of 100 μM Calpain inhibitor IV (American Peptide) and 1% (v/v) DMSO for 2 h prior to the hypoxia treatment. For all genotypes, following hypoxia treatment, seedlings were immediately cross-linked in 1% (v/v) formaldehyde nuclei purification buffer (NPB: 20 mM MOPS, pH 7.0, 40 mM NaCl, 90 mM KCl, 2 mM EDTA, and 0.5 mM EGTA) in a vacuum chamber under house vacuum for 10 min. The reaction was quenched by addition of 5M glycine to reach a concentration of 125 mM followed by vacuum infiltration for 5 min. This was followed by three washes in 30 mL ddH2O. Seedlings were blotted dry and then pulverized under liquid N2. To isolate nuclei, 0.5 g tissue was thawed to 4°C in 10 mL NPB that additionally contained 0.5 mM spermidine, 0.2 mM spermine, and 1 X Plant Protease Inhibitor Cocktail (Sigma-Aldrich P9599). Nuclei were pelleted by centrifugation in 4°C at 1200g for 10 min. The nuclei were washed in 10 mL NPBt (NPB, 0.1% Triton X-100) and pelleted by centrifugation in 4°C at 1200g for 10 min in triplicate. Following the final NPBt wash, the supernatant was aspirated and the nuclei pellet was resuspended in 120 µL NPB, and lysed with the addition of 120 µL 2X nuclei lysis buffer (NLB: 100 mM Tris, pH 8.0, 20 mM EDTA, 2% [w/v] SDS, and 2X Plant Protease Inhibitor Cocktail) by vortexing for 2 min at 24 °C. The chromatin was sheared into 200 to 600 bp fragments by sonication (Diagenode, Denville, NJ) with 33 cycles of 30 s ON and 30 s OFF at 4°C. The sample was cleared by centrifugation at 16,000g at 4°C for 2 min and the supernatant was diluted ten-fold with dilution buffer (DB: 16.7 mM Tris, pH8.0, 167 mM NaCl, 1.1% (v/v) Triton X-100, 1.2 mM EDTA). The entire chromatin fraction was precleared by incubation with uncoupled Protein G Dynabeads (ThermoFisher) for 30 min followed by collection of the supernatant. Three hundred microliters of ththe supernatant was remove input chromatin (1 mL for HRE2) was incubated with 3 µL of anti-H3K4me3 (ab8580, Abcam), anti-H3K27me3 (07-449, EMD Millipore), anti-H3K9ac (ab4441, Abcam), anti-H3K14ac (ab52946, Abcam), anti-p-CTD RNA Polymerase II (ab5095, Abcam), anti-FLAG (F1804, Sigma), or anti-HA (h3663, Sigma) for 4 h (overnight for HRE2) while rocking at 4°C. Protein G Dynabeads (30 µL) were washed in DB, added to the chromatin fraction and allowed to incubate for 2 h while rocking at 4°C. Beads were magnetically captured and washed sequentially in 1 mL for 5 min with four buffers: low NaCl2 wash buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 0.1% [w/v] SDS, 1% [v/v] Triton X-100, 2mM EDTA), high NaCl2 wash buffer (20 mM Tris, pH 8.0, 500 mM NaCl2, 0.1% [w/v] SDS, 1% [v/v] Triton X-100, 2mM EDTA), LiCl wash buffer (10 mM Tris, pH 8.0, 250 mM LiCl, 1% [w/v] sodium deoxycholate, 1% [v/v] Nonidet P-40, 1 mM EDTA), and standard TE buffer (10 mM Tris, pH 8.0, 1 mM EDTA). The beads were then washed twice with 10 mM Tris, pH 8.0 and resuspended in 25 µL of tagmentation reaction mix (10 mM Tris, pH 8.0, 5 mM MgCl2, 10% [w/v] dimethylformamide) containing 1 µL of Tagment DNA Enzyme (Illumina) and incubated at 37°C for 1 minute. Beads were washed twice with low NaCl2 wash buffer and then once in standard TE buffer. The chromatin was eluted from the beads by heating for 15 min at 65°C in elution buffer (EB: 100 mM NaHCO3 and 1% [w/v] SDS) and reverse cross-linked by the addition of 20 µL 5 M NaCl2 with heating at 65°C overnight. Following reverse cross-linking, Proteinase K (0.8 units; New England Biolabs) was added and the sample was incubated at 55°C for 15 min. The final tagged ChIP-DNA sample was purified using Qiagen MinElute columns according to the manufacturer’s instructions and eluted with 14 µL of EB.
ChIP-seq Library Preparation
Library preparation for short-read sequencing (ChIP-seq) with tagmentation was performed according to [92], with minor modifications. Final library enrichment was performed in a 50 µL reaction containing 12 µL ChIP DNA, 0.75 µM primers, and 25 µL 2X NEBNext PCR Master Mix. To determine the appropriate amplification cycle number, a qPCR reaction was performed on 1 µL of tagmented ChIP DNA in a 10 µL reaction volume containing 0.15 µM primers, 1X SybrGreen (ThermoFisher), and 5 µL 2X NEBNext PCR Master Mix (New England Biolabs) with the following program: 72°C for 5 min, 98°C for 30 s, 24 cycles of 98°C for 10 s, 63°C for 30 s, 72°C for 30 s, and a final elongation at 72°C for 1 min. Libraries were amplified for n cycles, where n is equal to the rounded up Cq value determined in the qPCR reaction. Amplified libraries were purified and size selected using SPRI AMPure XP beads (Beckman). AMPure XP beads were added at a 0.7:1.0 bead to sample ratio, and the remaining DNA was recovered by the addition of AMPure XP beads at a 2.0:1.0 bead to sample ratio and eluted in 13 µL of EB. Quantification of the final libraries was performed with Quant-iT PicoGreen (ThermoFisher), and library fragment distribution was evaluated by use of the the Agilent 2100 Bioanalyzer using the high sensitivity DNA chip. Final libraries were multiplexed to >5 nM final concentration and sequenced on the HiSeq 3000/4000 at the UC Davis DNA Technologies Core.
Isolation of Nuclei Tagged in Specific Cell Types (INTACT)
INTACT was performed according to [44] with modifications. Frozen pulverized tissue (0.5 g) of pUBQ:NTF/pACT2:BirA seedlings was thawed in 10 mL of cold NPB buffer and filtered through 30 µm filters (Sysmex). Nuclei were pelleted by centrifugation at 1200g for 7 min at 4°C, and the nuclear pellet was resuspended and washed twice in NPB containing 0.1% (v/v) Tween-20 (NPBt), and the washed pellet was resuspended in 1 mL of NPB. M280 Streptavidin Dynabeads (10 µL; Invitrogen) washed with 1 mL NPB and resuspended in the original volume were added to the nuclei and the sample was allowed to rock for 30 min at 4°C. The bead-nuclei mixture was diluted to 14 mL with NPBt and incubated with rocking for 30 s at 4°C. The beads were magnetically collected, the supernatant removed, the beads washed twice with NPBt, and resuspended in 1 mL of NPBt. A 25 µL fraction was removed to quantify nuclei following addition of 1 µL of 1.0 µg/µL Propidium Iodide following incubation on ice for 5 min before visualization and quantification with a C-Chip hemocytometer (Incyto) using a fluorescence microscope.
Assay for Transposase Accessible Chromatin (ATAC)
Fifty-thousand INTACT-purified nuclei from root tissue were magnetically captured. Any remaining NPB was removed and the nuclei were resuspended in 50 µL of transposition mix (1X TD buffer, 2.5 µL TDE1 transposase) and were incubated for 30 min at 37°C with occasional mixing. The transposed DNA was purified with Qiagen MinElute PCR purification columns, and the purified DNA was eluted in 11 µL of EB. For library construction, a reaction was prepared to amplify the sample in a two step process. The reaction was assembled containing transposed DNA (10 µL), 5 µM Ad1.1 and an indexing primer [23], 1X NEBNext High Fidelity PCR mix and cycled in the following program: 72°C 5 min; 98°C 30 s; 3 cycles of 98°C 10 s, 63°C 30s, 72°C 1 min; 4°C and the samples were placed on ice. Then a qPCR was performed using an aliquot of the amplified library with the following program: 98°C for 30 s; 20 cycles of 98°C for 10 s, 63°C for 30 s, 72°C for 1 min. The original PCR reaction was resumed for n additional cycles, where n is the cycle at which the qPCR reaction was at 1/3 of its maximum fluorescence intensity.
Amplified DNA was then purified with Qiagen MinElute PCR purification columns and eluted in 20 µL of EB. Amplified libraries were purified and size selected for fragments between 200 to 600 bp using SPRI AMPure XP beads (Beckman). AMPure XP beads were added at a 0.55:1.0 bead to sample ratio, and the remaining DNA was recovered by the addition of AMPure XP beads at a 1.55:1.0 bead to sample ratio and eluted in 13 µL of EB. Library concentration was determined using the NEBNext (New England Biolabs) library quantification kit. Final libraries were multiplexed to >5 nM final concentration and sequenced on the HiSeq 3000/4000 at the UC Davis DNA Technologies Core.
INTACT followed by Nuclear RNA Extraction
Fifty-thousand INTACT-purified nuclei were collected magnetically and the supernatant was removed, and resuspended in 20 µL of NPB. Nuclear RNA was extracted and purified using the Qiagen RNeasy Micro kit, and eluted in 20 µL of H2O. The RNA was then DNaseI digested with with 1 µL (2U/µL) of Turbo DNAseI (Ambion) and incubated for 30 min at 37°C. DNaseI was inactivated by adding EDTA to 15 mM, and heated at 75°C for 10 min, centrifuged at 2000g for 5 min and transferred to a new tube. The RNA was then purified using AMPure XP beads, and eluted in 15 µL H2O. Ribosomal RNA was depleted from the sample using the plant Ribo-Zero rRNA removal kit (Illumina) according to the manufacturer’s instructions.
Total RNA Isolation and PolyA RNA Affinity Purification
Total RNA was extracted from 50 mg frozen pulverized tissue by addition of 800 µL Trizol (Life Sciences) and incubation for 5 min at room temperature. Chloroform (200 µL) was added and the sample briefly vortexed and incubated at room temperature for 3 min. Following incubation, the samples were centrifuged at 12,000g for 15 min at 4°C, and the clear phase was transferred to a new tube. RNA was precipitated by addition of 500 µL isopropanol, vortexing briefly, and incubation at room temperature for 10 min, and pelleted by centrifugation at 12,000g for 10 min at 4°C. Purified RNA was resuspended in 50 µl EB and DNAseI digested with 2.5 µL of TURBO DNaseI (Ambion) for 30 min at 37°C. DNaseI was inactivated by addition of EDTA to 15 mM and heat treatment at 75°C for 10 min. The RNA was pelleted by centrifugation at 2000g for 5 min and transferred to a new tube. The RNA was then purified using AMPure XP beads, and eluted in 50 µL H2O. For polyA RNA selection, the total RNA sample was heated for 2 min at 65oC and placed on ice, after which 50 µl of Dynabeads Oligo (dT)25 pre-washed with wash buffer (WB: 10 mM Tris, pH 7.5, 150 mM LiCl, 1 mM EDTA) were added and the sample incubated at room temperature for 15 min with agitation. The beads were magnetically separated and washed twice with WB. For RNA elution, 50 µl EB was added and the sample heated for 2 min at 80°C and the eluted RNA was transferred to a new tube. The polyA RNA selection was repeated a second time and eluted in a volume of 16 µl.
Translating Ribosome Affinity Purification (TRAP)-seq
TRAP of mRNA-ribosome complexes was performed following the procedure of [17]. Briefly, pulverized tissue (1 mL) from the 35S:HF-RPL18 genotype was added to 5 mL of polysome extraction buffer (PEB: 200 mM Tris, pH 9.0, 200 mM KCl, 25 mM EGTA, 35 mM MgCl2, 1% PTE, 1 mM DTT, 1 mM PMSF, 100 µg/mL cycloheximide, 50 µg/mL chloramphenicol) containing 1% detergent mix (20% (w/v) polyoxyethylene(23)lauryl ether, 20% (v/v) Triton X- 100, 20% (v/v) Octylphenyl-polyethylene glycol, 20% (v/v) Polyoxyethylene sorbitan monolaurate 20) and homogenized with a glass homogenizer on ice The homogenized mixture was allowed to stand for 10 min on ice and then centrifuged at 16,000g at 4°C for 15 min.
Following centrifugation, the supernatant was transferred to a new tube and filtered through one layer of Miracloth (Millipore) to produce the clarified extract.
Protein G Dynabeads (50uL; ThermoFisher) were prewashed twice with 1.5 mL of wash buffer (WB-T: 200 mM Tris, pH 9.0, 200 mM KCl, 25 mM EGTA, 35 mM MgCl2, 5 mM PMSF, 50 µg/mL cycloheximide, 50 µg/mL chloramphenicol) by suspension and magnetic collection, with final re-suspension in the original volume. These were added to the clarified tissue extract and incubated at 4°C for 2 h with gentle rocking. The beads were collected magnetically, the supernatant was removed, and the beads were gently resuspended in 6 mL WB-T for 5 min at 4°C with rocking. This wash step was repeated two additional times, after which the beads were resuspended in 1 mL WB-T and transferred to a new tube and the supernatant was removed.
RNA was purified from the sample and the reserved clarified extract by addition of 105 µl of LBB (LBB: 100 mM Tris, pH 8.0, 500 mM LiCl, 10 mM EDTA, 1% (w/v) SDS, 5 mM DTT, 1.5% (v/v) Antifoam A, 5 µl/mL 2-mercaptoethanol) followed by vortexing for 5 min. Samples were incubated at room temperature for 10 min, centrifuged at 13,000g for 10 min and transferred to a new tube. Poly(A)+ RNA selection was performed by addition of 1 µl of 12.5 µM biotin-20nt-dT oligos (Integrated DNA Technologies) to 200 µl of the TRAP or RNA lysate sample, followed by incubation at room temperature for 10 min. In a separate tube 20 µl magnetic streptavidin beads (New England Biolabs) were washed with 200 µl LBB. The lysate was added to the washed beads and incubated at room temperature for 10 min with gentle agitation. The beads were magnetically separated and washed sequentially with wash buffer A (WBA:10 mM Tris, pH 8.0, 150 mM LiCl, 1 mM EDTA, 0.1% (w/v) SDS), wash buffer B (WBB:10 mM Tris, pH 8.0, 150 mM LiCl, 1 mM EDTA), and low salt buffer (LSB: 20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA). Following washes, the pellet was resuspended in 16 µl 10 mM Tris, pH 8.0 containing 1 mM 2-Mercaptoethanol and heated at 80°C for 2 min. The beads were magnetically collected, the supernatant was transferred to a new tube, and the poly(A)+ RNA selection process was repeated, and the samples combined in a new tube before storage at − 80°C.
RNA-seq Library Preparation
RNA-seq library preparation for nRNA, polyA, and TRAP RNA was performed as described in Kumar et al., 2012 [93]. In brief, RNA was fragmented and primed for first strand synthesis with random hexamers (Invitrogen). First strand synthesis was performed, followed by second strand synthesis, end repair, and A-tailing. Universal sequencing adapters were ligated to cDNA molecules. An initial enrichment and adapter extension PCR was first performed on 0.25x volume of the ligated products for each sample to determine amplification cycle number to prevent overamplification of sequencing libraries. A final enrichment and adapter extension PCR was performed with 0.25x starting volume for the specified cycle number (13-15 cycles) for each sample. Following enrichment, purification, size selection of libraries, and sequencing was performed as described for ChIP-seq.
Bioinformatic Analyses
For all high throughput outputs, short reads were trimmed using FASTX-toolkit to remove barcodes and filter short and low quality reads prior to alignment to the TAIR10 genome with Bowtie2/Tophat2. Read quality reports were generated using fastqc. For all datasets, counting of aligned reads was performed on entire transcripts [94] using the latest Araport 11 annotation (201606) and reads per kilobase of transcript per million mapped reads (RPKM) values were calculated. Differentially expressed genes were determined by edgeR, |FC| > 1, FDR < 0.01.
Data visualization and file generation
Bigwig files for Integrated Genome Viewer (IGV) visualization were generated from aligned bam files and normalized by RPKM values using the deepTools command bamCoverage with the -normalizeUsingRPKM specification. Within IGV, all chromatin based outputs were normalized to the same scale and all RNA outputs were normalized to a separate scale.
Peak calling
For ChIP-seq and ATAC-seq datasets, peak calling was performed using independent replicates in combination as input for HOMER using the findPeaks command [95]. For downstream analyses, peaks identified from each time point comparison were combined and counting was performed on individual replicates. Peaks were annotated using the HOMER command annotatePeaks.pl. Differential regulation of peaks was performed using edgeR.
Motif discovery
HRE2 peaks identified by HOMER command findPeaks were used as an input for the HOMER command findMotifsGenome.pl. The matrix for the top HRE2 motif (p: 1e-69) was used as input the MEME command ceqlogo to generate the motif logo.
DAP-seq motif analysis
Sequences of promoter regions spanning 1 kb upstream to 500 bp downstream of the TSS were extracted for genes within each cluster. The promoter sequences were compared against each TF, per TF family, using motif matrices identified by [53]. The number of significant motifs identified within the promoters of each cluster were quantitated and normalized to the number of genes within each cluster.
t-distributed stochastic neighbor embedding (t-SNE)
t-SNE analysis was performed according to [96]. Briefly, a principal component analysis was performed using RPKM values for each replicated dataset and timepoint. t-SNE was then performed on PCs 1-10 and the results were plotted using ggplot2.
Wilcoxon signed rank sum test
Wilcoxon signed rank sum test was performed by comparing the RPKM values of one genomic output for each gene within a cluster to the genes of another cluster. Each cluster was compared individually to all other clusters to determine significance, p < 0.05.
Fisher’s exact test
Fisher’s exact test was performed by comparing the number of motifs / peaks associated with genes within a cluster to the total number of motifs / peaks identified for all DRGs. Significance
Bioinformatic Tools
Analyses were performed with Bioconductor R packages particularly the Next Generation Sequencing analysis software of systemPipeR [97]. Programs used within Bioconductor included: BiocParallel [98], BatchJobs [99], GenomicFeatures [94], GenomicRanges [94], edgeR [100], gplots [101], ggplot2 [102], RColorBrewer [103], Dplyr [104], biomaRt [105], ChIPseeker [106], rtracklayer [107], Rtsne [108], Pheatmap [109], e1071 [110], cluster [111], ShortRead [112], and Rsamtools [113]. The Publically available UNIX/Python/Perl packages used included Tophat [114], fastx_toolkit [115], Fastqc [116], MEME [117], HOMER [95], deepTools [118], circos [119], Bedtools [120], samtools [121].
Declarations
Ethics approval and consent to participate
Not applicable
Consent for publication
Not applicable
Availability of data and material
The datasets generated and/or analysed during the current study are available in the NCBI GEO repository
Competing interests
The authors declare that they have no competing interests.
Funding
This work was supported by U.S. National Science Foundation (Grant Nos. IOS−1121626, IOS-1238243, MCB-1716913) and the U.S. Department of Agriculture, National Institute of Food and Agriculture - Agriculture and Food Research Initiative (Grant No. 2011−04015) to J.B.-S..
Authors' contributions
T.L. and J.B.-S. designed research; T.L. performed research; T.L. and J.B.-S. analyzed data; and T.L. and J.B.-S. wrote the paper.
Authors' information (optional)
The current location of T.L. is: Plant Biology Laboratory, The Salk Institute for Biological Studies, La Jolla, CA 92037, USA
Supplemental Tables
Table S1: Tabulation of ATAC and HRE2-ChIP peaks on chromatin
Table S2: Survey of histones, RNAPII, HRE2 and three RNA sub-populations in response to two durations of hypoxic stress
Table S3: nRNA Enrichment and Gene Ontology Analysis
Table S4: Gene Ontology Analysis for two durations of hypoxic stress in Table S2 Table S5: List of Arabidopsis Blacklisted Chromatin (ABC)
Supplemental Note
Arabidopsis blacklisted chromatin
The variety of genome-wide outputs in this study recording dynamics at the level of chromatin and RNA has permitted the identification of Tn5 bias/genomic regions with high nonspecific background and/or high signal/read counts in Arabidopsis irrespective of experiment, similar to regions identified by the ENCODE project for other organisms [91]. The identification and filtering of genomic blacklisted regions is especially important for calculation of genome wide Pearson correlations and the generation of signal tracks used for accurate browser views. The Arabidopsis Blacklisted Chromatin (ABC) regions (Table S5) can be used to remove these defined regions that create statistical artefacts of the nuclear genome prior to performing analyses of chromatin-based data.
Acknowledgements
We thank Mauricio Reynoso, Roger Deal, Marko Bajic, Seung-Cho Lee, Maureen Hummel, Lauren Dedow, Sonja Winte, Jérémie Bazin, Thomas Girke, and Thomas Eulgem for helpful discussions. This work was supported by U.S. National Science Foundation (Grant Nos. IOS−1121626, IOS-1238243, MCB-1716913) and the U.S. Department of Agriculture, National Institute of Food and Agriculture - Agriculture and Food Research Initiative (Grant No. 2011−04015) to J.B.-S..
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.
- 7.↵
- 8.↵
- 9.
- 10.
- 11.
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.
- 60.
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.↵
- 107.↵
- 108.↵
- 109.↵
- 110.↵
- 111.↵
- 112.↵
- 113.↵
- 114.↵
- 115.↵
- 116.↵
- 117.↵
- 118.↵
- 119.↵
- 120.↵
- 121.↵
- 121.