Discrete and cell‐specific hypoxic responses in Arabidopsis roots resolved by single‐nuclei transcriptomics

Summary

With the intensification of unpredictable flooding events because of global warming, there is a need to understand how root cells perceive and respond to oxygen deprivation. The use of high‐throughput single‐nuclei RNA‐sequencing (snRNA‐Seq) allows the examination of gene expression profiles in discrete cell types. Root tip segments of Arabidopsis seedlings subjected to normoxic (21% oxygen) or hypoxic (4% oxygen) treatments for 4 h were analyzed by snRNA‐Seq. Hypoxia induces a major transcriptome rewiring, most prominent in the meristematic root cells, with the exclusion of the quiescent center (QC), which is vulnerable to low oxygen. Changes in carbon and nitrogen metabolism by low oxygen were centered around increasing demand for carbohydrate to drive glycolytic fermentation, reduction of nitrate to sustain energetic processes, and the bypass of the TCA cycle via aspartate aminotransferase upregulation. The reduction of nitrate and nitrite in hypoxic cells contributes to amino acid metabolism and the utilization of NO in the phytoglobin–nitric oxide cycle to decrease the redox level and sustain energy production. In conclusion, the use of snRNA‐Seq provides a high‐resolution atlas of gene expression events defining root cell‐specific responses to low oxygen.

	Content
	Summary	2652
I.	Introduction	2652
II.	Materials and Methods	2653
III.	Results	2654
IV.	Discussion	2660
V.	Conclusions	2664
	Acknowledgements	2664
	References	2664

I. Introduction

Oxygen deprivation through flooding, metabolic status, or structurally based restricted oxygen availability impairs plant growth that can eventually lead to plant death. The diminished oxygen supply during root flooding results in changes in energy‐generating metabolic pathways, transcriptional alterations and modifications in protein turnover (reviewed in Bailey‐Serres & Voesenek, ²⁰⁰⁸; van Dongen & Licausi, ²⁰¹⁵). Changes in carbon and nitrogen metabolism are prominent in the acclimation process. Increases in pyruvate decarboxylase (PDC) and alcohol dehydrogenase (ADH) to produce ethanol occur early in the acclimation, resulting from the need to shift from oxidative phosphorylation to substrate phosphorylation. To provide the carbon supply required for this less efficient route for ATP production, there are increases in sucrose synthase activity and accompanying changes in carbon and nitrogen metabolism (reviewed in Cho et al., ²⁰²¹). This undoubtedly results from the upregulation of a core of anaerobic response genes that include several glycolytic and nitrogen metabolism genes (Lee et al., ²⁰¹¹). In maize, however, there is not a simultaneous induction of anaerobic genes associated with glycolysis (Hake et al., ¹⁹⁸⁵). Nitrate reduction during hypoxia leads to the replacement of oxygen by nitrite as the terminal electron acceptor, forming NO that is further scavenged by class 1 phytoglobin (Pgb1) (Igamberdiev & Hill, 2004). This sequence of reactions, defined as the phytoglobin–nitric acid (Pgb–NO) cycle, prevents overreduction of the hypoxic cell and yields limited amounts of ATP. On the other hand, nitrite reduction to ammonia also remains active under hypoxia, leading to the activation of amino acid metabolism (Gupta et al., ²⁰¹²). PGB1 expression is tissue‐specific, which affects cell differentiation and stem cell functionality (Mira et al., ^{2020, 2023}). It is also linked to the changes in the hormonal status of root meristematic cells, including the effects on auxin, jasmonic acid (Rathnayaka Pathiranage et al., ²⁰²³), and especially ethylene (Mira et al., ^{2024, 2025}), which is important for determining the fate of cells under hypoxic conditions. The extensive rewiring of carbon and nitrogen metabolism for keeping efficient energy production is mediated through the coordinated action of Pgb1, NO, and phytohormones.

The hormonal and signal transduction events that trigger the hypoxic response have been studied at several levels. Group VII (ERFVII) ethylene‐responsive factors coordinate the hypoxic response (Xu et al., ²⁰⁰⁶; Papdi et al., ²⁰⁰⁸; Hinz et al., ²⁰¹⁰; Licausi et al., ²⁰¹⁰). Under normoxic conditions, their levels are controlled by protein degradation via the N‐degron pathway that utilizes cysteine oxidases to modify the protein for processing by ubiquitin E3 ligases (Gibbs et al., ²⁰¹¹). Cysteine oxidases require oxygen (Weits et al., ²⁰¹⁴), and the degradation of the group VII ethylene transcription factors via the N‐degron pathway is dependent on NO (Gibbs et al., ²⁰¹⁴). Under hypoxic conditions, limited oxygen and the consumption of NO by oxyphytoglobin1 limit substrate, inhibiting protein degradation. The presence of Pgb1, with its strong oxygen‐binding activity (Trevaskis et al., ¹⁹⁹⁷), would also contribute to further reducing available oxygen.

Root hypoxia has generally been studied by examining specific sections of the organ to account for the root being at different stages of development during stress exposure. Limits on root oxygen diffusion can result in a hypoxic core, even under normal growth conditions (Berry & Norris, 1949; Thomson & Greenway, ¹⁹⁹¹; Mira et al., ²⁰²³; Iida et al., ²⁰²⁵) that can further complicate analyses. Since a section of a root consists of several cell types, each can experience different oxygen tensions. Moreover, different cell types can invoke different transcriptional programs in response to hypoxia (Mira et al., ²⁰²⁴), though demonstrating these changes on a global scale has been hindered by the need for high‐resolution RNA‐sequencing methods. Hypoxia‐induced changes in the ribosome‐bound fraction of mRNA in different cell types in response to low oxygen revealed the occurrence of global and cell‐type specific stress responses (Mustroph et al., ²⁰⁰⁹). The authors were able to isolate specific root cell types or, in some cases, clusters of diverse cell types to capture changes in transcription and translation associated with low oxygen stress. To expand that study and provide a better spatial resolution and genome coverage here we define transcriptome changes in single nuclei isolated from root tips of Arabidopsis seedlings exposed to normoxia or hypoxia for 4 h. We examine global transcriptomic variations and, more specifically, highlight changes in carbon and nitrogen metabolism genes, as well as hormone and signal transduction.

Our results revealed that hypoxia evokes major cell‐specific transcriptional changes in most cell types, and that this is most evident within the meristematic regions of the root. Our results also show that the transcriptomes of the Quiescent Center (QC) cells were unresponsive to the stress, though the QC itself is prone to degradation under hypoxia. Metabolically, the induction of key genes suggests coordination of carbon and nitrogen metabolism, reflecting the increased demand for carbohydrates to drive glycolytic fermentation. Activation of nitrogen metabolism processes might be part of the early hypoxic response defined by the generation of nitrite to drive developmental and/or energetic processes, as well as the bypass of the TCA cycle via aminotransferase upregulation. Finally, we provide evidence that the hypoxic response mediated by ethylene and regulated by ERFVII is well advanced after 4 h of hypoxic stress.

II. Materials and Methods

1. Plant material and growth conditions

Arabidopsis (Arabidopsis thaliana) Col‐0 seeds were sterilized in 70% ethanol containing 0.5% Triton X‐100 for 15 min followed by 95% ethanol for 15 min (Mira et al., ²⁰¹⁶). Seeds were plated in squared petri dishes containing agar germination medium (½‐strength Murashige & Skoog salt with vitamins, 2.5% sucrose (w/v) and 0.8% agar (w/v), pH 5.7), incubated for 24 h at 4°C in the dark, and then transferred to a growth cabinet (22°C, 16 h light : 8 h dark). Oxygen treatments were carried out on uniform 5‐d‐old seedlings as reported in Mira et al. (2024), with two replicates each. Briefly, plates were placed in acrylic desiccators flushed with either 4% oxygen (hypoxia) or 21% oxygen (normoxia) under dark conditions for 4 h. Each biological replicate consisted of at least 50 plants across five plates.

2. Analyses of reporter lines and microscopy

Reporter lines were those utilized in Mira et al. (2016). Seedlings were stained for 20 min with 10 μg ml⁻¹ propidium iodide, rinsed three times in water, and visualized by confocal microscopy. For the starch assay, roots of treated plants were fixed in 4% paraformaldehyde, dehydrated in ethanol, and embedded in historesin exactly as reported by Yeung (1999). Thin sections (3 mm) were stained with periodic acid‐Schiff reaction for visualization of carbohydrates (Yeung, 1999).

3. Nuclei isolation and library preparation

Nuclei were isolated from fresh root tissue (1 cm root tips) using a modified protocol (Guillotin et al., ²⁰²³). Roots were collected from five different plates and immediately transferred into a prechilled Petri dish containing 300 μl of lysis buffer (0.3 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 15 mM Tris–HCl, 20 mM MES, 10 mM MgCl₂, 60 mM KCl, 15 mM NaCl, 0.5 mM spermine and spermidine, 0.05% Triton X‐100, 5 mM DTT, 1 mM PMSF, 1% Protease Inhibitor Cocktail (Sigma P9599), 0.4 u μl⁻¹ Protector RNase Inhibitor (Sigma), 0.4% BSA, 1% DEPC) and chopped on ice with two razor blades for 2–3 min. The homogenate was transferred to a chilled microcentrifuge tube, gently ground with a plastic micro pestle, and incubated on ice for 2–3 min. The suspension was filtered through a 20 μm strainer into a chilled 1.5 ml Eppendorf tube, rinsed with 200 μl lysis buffer, and centrifuged at 1000  g for 10 min. The supernatant was gently removed, and the pellet resuspended in 300 μl of wash buffer (0.3 M sucrose, 15 mM Tris–HCl, 15 mM MES, 60 mM KCl, 15 mM NaCl, 2 mM EDTA, 0.5 mM spermine and spermidine, 0.05% Triton X‐100, 5 mM DTT, 1 mM PMSF, 1% Protease Inhibitor Cocktail, 0.2 u μl⁻¹ Protector RNase Inhibitor, 0.4% BSA) and centrifuged at 1000 g for 5 min. The pelleted nuclei were resuspended in 50–100 μl of final buffer (0.3 M sucrose, 15 mM Tris–HCl, 15 mM MES, 60 mM KCl, 15 mM NaCl, 0.5 mM spermine and spermidine, 5 mM DTT, 1% Protease Inhibitor Cocktail, 0.2 u μl⁻¹ Protector RNase Inhibitor, 1% BSA). The suspension was filtered through a 10 μm strainer into a fresh chilled tube. To count and assess the quality, the nuclei were stained with 1 μg ml⁻¹ DAPI for 5 min, loaded onto a hemocytometer, and viewed under an LSM 700 confocal microscope (Zeiss, Oberkochen, Germany).

Single‐nuclei RNA‐sequencing libraries were prepared using the 10× Genomics (Pleasanton, CA, USA) Chromium Next GEM Single Cell 3′ Reagents Kits v.3.1 (Dual Index) protocol without modification on the 10× Genomics Chromium X instrument (Pleasanton, CA, USA). A total of 902 million 100 bp paired‐end reads were sequenced on a NovaSeq 6000 (Illumina, San Diego, CA, USA) at Génome Québec (Montréal, QC, Canada). Raw data are available on NCBI's Gene Expression Omnibus under accession GSE308672.

4. Transcriptome analyses

Data processing was performed on HPC clusters supported by the Prairie DRI group and the Digital Research Alliance of Canada (www.alliancecan.ca). Reads were pseudoaligned to the TAIR 10 Arabidopsis thaliana reference transcriptome (Lamesch et al., ²⁰¹²) with scKB (Hsu et al., ²⁰²²), which uses kallisto (Bray et al., ²⁰¹⁶) and bustools (Melsted et al., ²⁰²¹). All downstream analyses were performed in R (R Core Team, ²⁰²³) with Seurat v.5.1.0 (Hao et al., ²⁰²⁴). High‐quality cells were filtered with COPILOT (Hsu et al., ²⁰²²), and doublets were removed with DoubletFinder (McGinnis et al., ²⁰¹⁹). Following a published analysis pipeline (Nolan et al., ²⁰²³), cells were normalized with SCTransform v.0.4.1 (Hafemeister & Satija, 2019), integrated with the Shahan et al. (2022) Arabidopsis root atlas, and cell annotations were transferred with Seurat's label transfer system (FindTransferAnchors, TransferData, AddMetaData). Annotated samples were then integrated into one Seurat object (SelectIntegrationFeatures, PrepSCTIntegration, FindIntegrationAnchors, IntegrateData).

Visualizations were generated with scCustomize (Samuel Marsh et al., ²⁰²¹) and ggplot2 (Wickham, 2011). Root micrograph heatmaps were generated with ggPlantMap (Jo & Kajala, 2024). All differentially expressed genes (DEGs) were calculated in Seuart (FindMarkers) on pseudobulked clusters with the DESeq2 test (Love et al., ²⁰¹⁴). Pseudobulk expression values were calculated by aggregating normalized counts for each cell type with Seurat (AggregateExpression). Enriched Gene Ontology (GO) terms were calculated with Fisher's exact test with topGO (Alexa & Rahnenfuhrer, 2010). Enriched transcription factor binding motifs were identified with SEA (Bailey & Grant, 2021) via Meme‐Suite (Bailey et al., ²⁰¹⁵) using motifs from CIS_BP 2.00 (Weirauch et al., ²⁰¹⁴); all detected genes were used as background control sequences. Cell perturbation scores were calculated with Augur (Skinnider et al., ²⁰²¹) using default settings. Metabolic pathways and the genes corresponding to each enzymatic reaction were retrieved from the KEGG database (Kanehisa et al., ²⁰²⁵), and enrichment scores were calculated with clusterProfiler (Yu et al., ²⁰¹²).

III. Results

1. Global transcriptomic changes evoked by hypoxia

To build a cell‐scale transcriptome atlas of Arabidopsis root responses to hypoxic conditions, we used the 10× Genomics Chromium X platform to sequence the transcriptomes of over 50 000 nuclei (Supporting Information Table S1). We harvested root tips (1 cm) from 5‐d‐old Arabidopsis seedlings exposed to normoxic (21% oxygen) or hypoxic (4% oxygen) treatments. Using two biological replicates per treatment, nuclei were isolated from root segments for analyses. To capture early hypoxic changes, we imposed 4 h of treatments, corresponding to the initial increased expression of the hypoxic marker genes PYRUVATE DECARBOXYLASE 1 (PDC1‐AT4G33070), ALCOHOL DEHYDROGENASE 1 (ADH1‐AT1G77120), and PHYTOGLOBIN 1 (Pgb1‐AT2G16060) (Fig. S1). After filtering out doublets and low‐quality nuclei that included > 5% mitochondrial reads or low UMI (unique molecular identifier) counts using a distribution‐based approach (Hsu et al., ²⁰²²), the final dataset was comprised of 45 978 nuclei. A median of 1403 genes was detected per nucleus; across the population of nuclei, 20 004 genes were detected in at least 100 cells, representing 73.0% of genes in the Arabidopsis genome (Table S1).

To annotate the nuclei with respect to cell type and developmental state, we leveraged published single‐cell and bulk Arabidopsis transcriptome datasets for which experimentally validated cell type annotations had been made (Shahan et al., ²⁰²²). By integrating the snRNA‐seq data from the present experiment with those from the published data, we were able to transfer the annotations based on transcriptome similarity. Using this approach, the sequenced nuclei were mapped to a broad range of developmental stages, including cells in meristematic, elongation, and mature cell states (Fig. 1a) and cell types (Fig. 1b). In downstream analyses, cells within the elongation and maturation zones were combined, as the transcriptomes of cells in these growth zones were highly correlated (Fig. S2). We ultimately classified the nuclei into eight meristematic and eight mature cell types based on the expression of established marker genes (Fig. 1c). These cell types represent 10 of the 14 cell types identified by the root atlas (Shahan et al., ²⁰²²) (Fig. S3). Notably, the root atlas had further divided the phloem, xylem, and pericycle into different developmental groups, while we instead categorized the epidermis, cortex, endodermis, and stele into meristematic and mature groups. We captured a similar number of nuclei for each annotated cell type in normoxic and hypoxic conditions and between biological replicates (Figs 1d, S4). This was expected as the treatment lasted for only 4 h, so a limited number of cell division events would have occurred. To confirm the annotations, we examined the expression of four genes: WOX5 (AT3G11260), SCR (AT3G54220), WER (AT5G14750), respective markers of the quiescent center (QC), endodermis, and epidermis, and the hypoxia‐responsive HRE2 (AT2G47520), preferentially expressed in the stele (Fig. S5). The hypoxia treatment induced HRE2, did not affect the expression of SCR, and only marginally (but not statistically) reduced the expression of WOX5 and WER; a trend concordant with the cell type assignments we had made. Based on these results, we are confident that the data are sufficient to distinguish between cell types, developmental states, and oxygen treatments.

Fig. 1.

snRNA‐seq dataset captures representative cell states across cell types and developmental zones of the Arabidopsis thaliana root tip. UMAP of snRNA‐seq transcriptomes colored by (a) developmental zone or (b) cell type. Root maps indicate where each annotated cell type is within the root tip. Transcriptomes from both normoxic and hypoxic conditions are included. (c) Expression of select marker genes within each cell type; the size indicates the percentage of cells expressing each gene, and the color indicates expression level. (d) Proportion of cells within each cell type and treatment. Two replicates are shown for each treatment. QC, quiescent center.

Next, we explored hypoxia responses within the transcriptomes of each of the annotated cell types. By performing pairwise contrasts between the normoxic and hypoxic cells for each cell type, we identified 586 genes that were differentially expressed in at least one cell type (FDR < 0.05; |LFC| > 0.5; Table S2). Half of the hypoxia‐responsive genes were differentially expressed in more than one cell type: 43 were differentially expressed in 10 or more cell types, 279 were differentially expressed in 2–9 cell types, and 264 were expressed in a single cell type (Fig. 2a). The epidermis, cortex initials, endodermis initials, and stele had the largest number of differentially expressed genes (DEGs), while the QC had the lowest number. The low number of DEGs in the QC was not due to a lower number of detected genes. This number (3289), while lower than the number of detected genes in most mature cell types, was comparable to the other cell types of the root cap: lateral root cap – 3446; columella – 3083 (Fig. S6A). Rather, the QC had fewer DEGs because of higher basal expression and smaller changes in expression of hypoxia‐induced genes (Fig. S6B). Most DEGs were upregulated by hypoxia, with 67.6% (396) of the DEGs being more highly expressed in hypoxic conditions than in normoxic conditions; the remaining 32.4% (190) DEGs were downregulated under hypoxia (Fig. 2b). No DEGs were divergently regulated across cell types (i.e. upregulated in one cell type and downregulated in another). We clustered DEGs that were differentially expressed in more than one cell type into three distinct patterns: core DEGs that were upregulated in at least 10 cell types (43 DEGs), genes that were upregulated in 2–9 cell types (208 DEGs), and genes that were downregulated in 2–9 cell types (71 DEGs). The core DEGs were enriched for gene ontology (GO) terms associated with hypoxia, anaerobic respiration, nitrite assimilation, and glycolytic processes (Fig. 2c; Table S3). Of these, 23 were identified as core hypoxia response genes (Mustroph et al., ²⁰⁰⁹). DEGs that were upregulated in few cell types (primarily the epidermis, cortex initials, endodermis initials, and stele) were enriched for GO terms associated with hypoxia, pentose‐phosphate shunt, gluconeogenesis, and glycolytic processes. Downregulated DEGs were enriched for GO terms associated with protein folding and degradation, regulation of fatty acid biosynthesis, and cell division.

Fig. 2.

Hypoxia induces a larger transcriptional response in the meristematic tissues of Arabidopsis thaliana roots. (a) Number of hypoxia‐induced DEGs in each cell type (FDR < 0.05; |LFC| > 0.5). Color indicates the number of cell types in which genes are differentially expressed. (b) Log₂ fold‐change (LFC) of DEGs in each cell type. Only DEGs that are differentially expressed in two or more cell types are shown. Color bars show grouping based on number of shared cell types, with the striped pattern indicating downregulated DEGs. (c) Top enriched Gene Ontology (GO) terms of each cluster in B (P < 0.05). (d) Root maps of exemplar DEGs, highlighting different expression patterns. Cell color represents the log₂ fold‐change between normoxia and hypoxia. (e) Cell perturbation quantified with Augur. A higher AUC represents greater transcriptome perturbation in hypoxia. (f) Top enriched GO terms of cell type specific upregulated DEGs in the epidermis, stele, and cortex initials (I) (P < 0.05). (g) Enriched transcription factor binding motifs in the promoter regions (−500: +100 bp) of DEGs upregulated in the epidermis, stele, and cortex initials (FDR < 0.05). TFs associated with the motifs are listed on the right. DEG, differentially expressed gene; AUC, area under the curve; FDR, false discovery rate; QC, quiescent center.

We identified exemplar DEGs that responded to hypoxia in almost every cell type (Fig. 2d). PLANT CYSTEINE OXIDASE 1 (PCO1) and 2 (PCO2) were upregulated in a variety of immature and mature cell types. We also identify DEGs localized to specific areas of the root, such as the developing tissues (BIP1), vasculature (HRM1, FSK2), mature cell types (CYP707A3), or individual cell types (AHG1). The heterogeneity of hypoxia responses amongst cell types revealed by this analysis emphasizes the value of cell type transcriptome analyses of stress responses.

In addition to the ubiquitously differentially expressed genes, we identified cell‐type‐specific DEGs. Hypoxia‐responsive DEGs unique to the epidermis (n = 51) were enriched for GO terms associated with transmembrane transport of nitrate and calcium ions and for transport between the cytoplasm and organellar compartments (Fig. 2f; Table S4). Hypoxia‐responsive DEGs unique to the stele (n = 41) were enriched for GO terms associated with oxidative stress and lipid metabolism. Hypoxia‐responsive DEGs unique to the cortex initials (n = 55) were enriched for GO terms associated with isoleucine, valine, and ketone biosynthesis. To determine if there are common regulators of the hypoxia responses in these cells, we identified enriched transcription factor binding motifs in the promoter regions (−500 to +100 bps) of DEGs upregulated in the epidermis, stele, and cortex initials (n = 351 genes; FDR < 0.05). This highlighted three significant motifs and their associated TFs (Fig. 2g; Table S5). The motif ‘AAATATCT’ was associated with six MYB TFs (LHY, AT2G28920, RVE4, RVE7L, RVE6, RVE1); these TFs are involved in interactions between the circadian clock, auxin signaling, and abiotic stress responses (Rawat et al., ²⁰⁰⁹; Kidokoro et al., ²⁰²¹). A ‘GA’‐repeat motif was associated with BPC6, a root tip‐specific developmental regulator that has been shown to be perturbed in ethylene responses (Monfared et al., ²⁰¹¹). Finally, the motif ‘TTTGTCTTT’ was associated with SGR5, a regulator of gravitropism that is involved with starch accumulation (Tanimoto et al., ²⁰⁰⁸). These results show that in addition to the widespread hypoxia responses that are critical for all cells responding to the oxygen deficit, different cell types within the root also have distinctive responses and regulatory programs that relate to specific processes that are perturbed in hypoxia.

One of the major limitations of identifying DEGs for different cell populations is that the size of a cell population strongly influences the likelihood of identifying DEGs. Larger cell populations have a greater statistical power for detecting DEGs; as such, responses in larger cell populations tend to be overrepresented in composite DEG lists. For this reason, we took a second approach to identify the cell types that were most perturbed by hypoxia exposure, using the tool Augur (Skinnider et al., ²⁰²¹). Briefly, this method uses a machine‐learning approach to identify cell types whose transcriptomes are most perturbed by a treatment. Through this analysis, we determined that the cortex initials and epidermis were the most perturbed by hypoxia exposure, followed by the elongating trichoblasts, endodermis initials, and stele (Fig. 2e). The transcriptomes of mature tissues, including the vasculature, lateral root cap, and columella, were least perturbed by hypoxia. Together, these analyses confirmed the strong effect of hypoxia on the meristematic tissues. Based on this, we focused downstream analyses on DEGs detected in epidermis, immature cortex, immature endodermis, and stele transcriptomes.

2. Carbon metabolism

To identify changes in metabolism and cellular functions induced by hypoxia, we next ran a gene set enrichment analysis (GSEA) on pathways in the KEGG database (Fig. 3a; Table S6) (Kanehisa, 2000; Kanehisa et al., ²⁰²⁵). Here, we focused on the KEGG carbon metabolism pathways as they were the most enriched. Among the enriched KEGG carbon metabolism pathways, ‘Glycolysis/Gluconeogenesis’ was significantly upregulated in most of the annotated cell types (FDR < 0.05), while ‘Carbon fixation by Calvin cycle’ and ‘Carbon metabolism’ were upregulated only in dividing and meristematic cell types with the exception of the QC (Fig. 3a). Oxygen‐consuming pathways including ‘Oxidative phosphorylation’ were downregulated in mature cell types and the lateral root cap. Accumulation of starch in treated root cells was also analyzed at a histochemical level. The presence of granules of starch, mainly noted in the root stele and cortex, decreased after 4 h of hypoxic treatment (Fig. 3b).

Fig. 3.

Hypoxia alters carbon metabolism in Arabidopsis thaliana roots. (a) Normalized enrichment scores (NES) of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways related to carbon metabolism. Significant pathways within each cell type are marked with an asterisk (FDR < 0.05). (b) Starch granules (arrows) in the cortex (C) and stele (S) of normoxic or hypoxic roots. Bars, 25 μm (left panels), 150 μm (right panels). n = 20. (c, d) Module score of genes within specified KEGG pathways: glycolysis (118 genes); tricarboxylic acid (TCA) cycle (64 genes). (e) Hypoxia‐responsive differentially expressed genes (DEGs) in the glycolysis pathway. Root maps are only shown for genes with at least one DEG. FDR, false discovery rate; QC, quiescent center.

We also mapped the expression of enzyme‐encoding genes in the glycolysis or TCA cycle KEGG pathways onto the UMAP (Fig. 3c,d). A single expression value was calculated for each cell, called a module score that shows the level of dysregulation of the gene set for each oxygen condition. The glycolysis module expression score incorporated data for 118 genes and the TCA module score incorporated data for 63 genes. Genes in the glycolysis pathway were most highly induced in cortex and endodermis initials in response to hypoxia; no such response was detected for the TCA cycle pathway. We next explored the response of each gene in the glycolysis pathway to hypoxia to determine in which cells each step was responding (Fig. S7). Most steps in the pathway have at least one enzymatic reaction that is upregulated in hypoxia, but the hypoxia‐induced responses are not consistent across all cell types in the developing root. For example, while two sucrose synthases (SUS1 and SUS4) are upregulated in most of the queried cell types, the enzymes that convert fructose 6‐P to fructose 1,6‐bisphosphate (PFK3 and PFK6) are predominantly upregulated in the mature pericycle and endodermis. Strikingly, the only glycolysis enzyme that was significantly induced by hypoxia in the QC is GAPC‐2 (Fig. S7).

We also mapped key enzymes of carbon metabolism (Fig. 3e). Hypoxia evoked a ubiquitous and strong induction of several genes in both meristematic and mature root tissue. These included the sucrose‐degrading enzymes SUCROSE SYNTHASE1 and 4 (SUS1 and 4), GLYCERALDEHYDE 3‐P DEHYDROGENASE 2 (GAPC‐2), converting glyceraldehyde 3‐P (GA 3‐P) to glycerate 1,3‐PP, and the two fermentation enzymes PYRUVATE DECARBOXYLASE 1 (PDC1) and ALCOHOL DEHYDROGENASE 1 (ADH1). With the exception of GAPC‐2, the genes were not induced in the xylem and QC. Low oxygen also moderately upregulated FRUCTOSE 1,6‐PP ALDOLASE 3 and 6 (FBA3 and 6) and PYRUVAYE KINASE 9 (PK9). The induction of PK9 by hypoxia mainly occurred in all meristematic cell types, with the exclusion of the columella cells and stem cell niche, as well as trichoblasts, endodermis, pericycle, and procambium of the mature zone (Fig. 3e).

The variation in the differential expression of several of the glycolytic genes amongst cell types is puzzling. The cortex and epidermis are among the most active root cells and may be more responsive to stress. Increased carbon supply (SUS genes) appears to be a common feature in all cell types, yet PFK is predominantly upregulated in only the mature pericycle and endodermis. A paper which primarily deals with the mitochondrial proteome in Arabidopsis cell cultures provides supplemental data on changes in the cells of the root proteome under hypoxia (Heazlewood et al., ²⁰⁰⁴; Table S2). Their results show similar cell specificity in the increased protein levels under hypoxia amongst specific cell types.

3. Nitrogen metabolism

The effects of hypoxia on genes in the KEGG nitrogen metabolism pathways were examined (Fig. 4a; Table S6). The ‘biosynthesis of amino acids’ pathway was upregulated solely in several meristematic cell types, including epidermis, cortex, and endodermis initials. Low oxygen also induced nitrogen metabolic processes in the atrichoblasts and glutathione metabolic events in the same cell types, as well as in the xylem. With the exclusion of the ‘ribosome’ pathway, induced by hypoxia exclusively in the epidermis, cortex, endodermis, and stele of the meristematic region, all other protein‐processing events were suppressed by the stress in the majority of the cell types within the mature and meristematic domains.

Fig. 4.

Hypoxia alters nitrogen metabolism in Arabidopsis thaliana roots. (a) Normalized enrichment scores (NES) of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in each cell type, grouped by processes related to nitrogen metabolism or protein processing. Significant pathways within each cell type are marked with an asterisk (FDR < 0.05). (b) Hypoxia‐responsive differentially expressed genes (DEGs) in the nitrogen metabolic pathway. Root maps are only shown for genes with at least one DEG. The arrow shows nitrite conversion to NO. (c) Scheme of the aminotransferase bypass of the tricarboxylic acid (TCA) cycle. It includes the amination reaction of glutamate dehydrogenase (GDH) and aspartate aminotransferase (ASPAT). Abbreviations of the TCA cycle enzymes: 2‐OGDH, 2‐oxoglutarate dehydrogenase complex; ACO, aconitase; CS, citrate synthase; FUM, fumarase; ICDH, isocitrate dehydrogenases (NAD‐ and NADP‐dependent); MDH, NAD‐malate dehydrogenase; SCS, succinyl‐CoA synthetase; SDH, succinate dehydrogenase. FDR, false discovery rate; QC, quiescent center.

Next, we mapped key genes of nitrogen metabolism (Fig. 4b). A strong induction by low oxygen occurred for PHYTOGLOBIN 1 (Pgb1) in all cell types except xylem and QC. A similar pattern was also observed for ASPARTATE AMINOTRANSFERASE 2 (ASP2), although the induction also included xylem cells. Low oxygen also upregulated NITRATE REDUCTASE 1 and 2 (NIA 1 and 2). Within the meristematic region, NIA1 was mainly induced in the lateral root cap, with no upregulation occurring in the QC and columella. The induction of NIA2 was more pronounced and extended in all meristematic cells, particularly the root cap, with the exclusion of the QC. Both genes were preferentially upregulated by low oxygen in the atrichoblasts of mature roots. The expression of NITRITE REDUCTASE 1 (NIR1) was moderately increased by the stress, especially in the meristematic cortical cells. Unlike many nitrogen‐related genes, low oxygen reduced the expression of GLUTAMATE DEHYDROGENASE 2 (GDH2) within the epidermal and cortical cells of the meristem. The function of some of the nitrogen metabolism genes, including ASP2, in a scheme bypassing the TCA cycle to generate aspartate, which will be discussed further, is shown in Fig. 4(c).

4. Hormones and signaling

Hormonal responses and cell signaling are influenced by hypoxia. The ethylene biosynthetic genes, S‐ADENOSYLMETHIONINE SYNTHASE 1 and 2 (SAM1 and 2), as well as AMINOCYCLOPROPANECARBOXYLATE OXIDASE 1 (ACO1), were strongly upregulated by oxygen deprivation (Fig. 5a). While the induction of SAMs was limited to the stele, cortex, and epidermis of the meristem, that of ACO1 encompassed all cell types with the exclusion of xylem and QC. Among the ethylene and low‐oxygen signaling genes, ETR2 was only mildly induced by hypoxia in the procambial cells, HRE1 was upregulated in the endodermis and pericycle of the mature root and meristematic cortex, whereas HRE2 was induced especially in the stele of the meristem and pericycle (Fig. 5b).

Fig. 5.

Hormone biosynthesis and signaling is altered by hypoxia in Arabidopsis thaliana roots. Root maps show the log₂ fold‐change of gene expression in hypoxia for genes related to: (a) ethylene biosynthesis, (b) ethylene and low‐oxygen signaling, (c) auxin biosynthesis, (d) auxin binding and transport, and (e) calcium signaling.

Low oxygen also increased the expression of several regulators of auxin biosynthesis (Fig. 5c). YADOKARI‐1 (YDK1), a repressor of auxin biosynthesis, was upregulated in atrichoblast cells, whereas INDOLE‐3‐ACETALDEHYDE OXIDASE 1 (AAO1), which catalyses the conversion of indole‐3 acetaldehyde to indole‐3‐acetic acid, was induced only mildly in the epidermis. The strongest induction by hypoxic stress occurred for two cytochrome P450 genes, TRYPTOPHAN N‐MONOOXYGENASE CYP79B2 and CYP79B3, converting tryptophan to indole‐3 acetaldoxime. The former exhibited an induction in the lateral root cap, whereas the latter in the pericycle. Among genes involved in auxin binding and transport, the auxin receptor TRANSPORT INHIBITOR RESPONSE 1 (TIR1) was upregulated in hypoxic epidermis, cortex, and stele, while the auxin transporter LAX2 was induced in the stele (Fig. 5d). A different trend occurred for the auxin efflux transporter AGRAVITROPIC 1 (AGR1), exhibiting a suppression in the stele.

We also examined the expression patterns of components of calcium signaling. CBL‐INTERACTING PROTEIN KINASE 7 and 15 (CIPK7 and 15), CALMODULIN 38 (CML38), as well as CYCLIC NUCLEOTIDE GATED CHANNELS 5 and 12 (CNGC5 and 12) exhibited an increased expression due to hypoxia (Fig. 5e). Except for CNGC12, which was upregulated in the inner cell layers of the mature roots, all genes were induced in one or more meristematic cell types. A similar induction pattern encompassing all meristematic cells except the QC and root cap, albeit of diverse degree, was noted for CIPKs, CML38, and CNGC12. This contrasted with the upregulation of CNGC5, which was limited to the meristematic endodermal cells. Low oxygen decreased the expression of CALMODULIN‐LIKE 8 (CAM8), especially in the trichoblasts and atrichoblasts, as well as that of CALNEXIN 1 (CNX1) and CALRETICULIN 1 (CTR1), which were suppressed in many meristematic cells with the exception of QC, columella and endodermis (CNX1) and QC and root cap (CRT1). A mild suppression also occurred for CNX1 in the pericycle of the mature root. (Fig. 5e).

IV. Discussion

Hypoxia triggers extensive transcriptome remodeling in many root cell types with an overall transcriptional induction of genes participating in diverse cellular and metabolic processes (Fig. 2a,b). While the consistent upregulation of the 49‐core hypoxia responsive genes in many cell types (within the meristematic and mature regions) confirms a ‘unified’ and ‘ubiquitous’ response of root cells to oxygen deprivation (Mustroph et al., ²⁰⁰⁹), greater responses occur in the meristematic root domain (Fig. 2a). This is in line with the important role attributed to the survival of the meristematic cells during hypoxia (Mira et al., ^{2016, 2020}). However, within the meristematic cells, the QC shows the least transcriptional changes to low oxygen, confirming the unique behavior of this domain acting as the organizing center of the root stem cell niche coordinating root growth (Dubrovsky & Ivanov, 2021). The low number of hypoxia‐induced DEGs in the QC is not due to a lower number of detected genes, but rather to smaller changes in expression (Fig. S6). The unresponsive state of the hypoxic QC is likely due to the fact that QC cells normally develop under low oxygen tension (Shukla et al., ²⁰¹⁹; Weits et al., ²⁰¹⁹; Mira et al., ²⁰²³) and therefore might already be adapted to hypoxia. Retention of quiescence in the QC has been identified as an adaptive survival strategy (Mira et al., ²⁰²⁰). Low oxygen compromises QC identity and in extreme cases activates and consumes the QC leading to meristem abortion (Mira et al., ²⁰²⁰). Furthermore, conditions favoring QC quiescence enhance hypoxic tolerance while those inducing precocious proliferation are conducive to susceptibility (Mira et al., ²⁰²⁰). An important observation emerging from all these studies is that the QC is extremely vulnerable to hypoxia, and retention of a metabolic quiescence might be essential for its functionality, safeguarding the integrity of the roots. However, the conditions making the QC vulnerable to hypoxia are unknown. The few DEG observed in the hypoxic QC suggest that its vulnerability is not the result of inherent transcriptome changes, but perhaps variations occurring in adjacent cells which might limit essential factors sustaining the QC. Mira et al. (2023) suggested that one of these limitations could be the reduction of nutrients, that is carbon and nitrogen, supplied by the apical, mature, hypoxic root cells to the QC. The highest responses to low oxygen observed in the four cell types apically located to the QC: the epidermis, cortex and endodermis initials, and stele, exhibiting induction of genes linked to several carbon and nitrogen metabolic processes (Fig. 2a–c), could be viewed as an attempt to sustain the provision of nutrients for the hypoxic meristem. Besides the activation of specific events related to glycolysis/gluconeogenesis and carbon fixation and metabolism (Fig. 3a), there is also a defined transcriptional induction related to ribosomal processes and amino acid synthesis (Fig. 4a) specific to these cell types. This transcriptome rewiring is concomitant to variations in regulatory processes including nitrate and calcium transport in epidermal cells, sodium homeostasis and oxidative stress responses in the stele, and amino acid synthesis and metabolism in the cortex initials (Fig. 2e,f). Thus, the activation of diverse processes, particularly related to carbon and nitrogen metabolism, within the meristematic cells subtended by the QC could be a critical acclimation strategy to hypoxia which we suggest influences QC identity, meristem functionality, and ultimately root growth.

The existence of a coordinated and interconnected role of carbon and nitrogen metabolism within the hypoxic root cells is undoubtedly evidenced by the almost identical and ubiquitous induction patterns of PHYTOGLOBIN 1 (PGB1), NITRATE REDUCTASE (NIA1 and NIA2), ASPARTATE AMINOTRANSFERASE (ASP2) (Fig. 4b), SUCROSE SYNTHASE (SUS1 and 4), PYRUVATE DECARBOXYLASE (PDC1), and ALCOHOL DEHYDROGENASE (ADH1) (Fig. 3e). These genes are highly induced by low oxygen in most cell types, with some exceptions. Cell‐specific changes in carbon and nitrogen metabolism will be discussed separately.

1. Carbon metabolism

The induction of SUS (1 and 4), PDC1, and ADH1 (Fig. 3e) is likely a consequence of incapacitated oxidative phosphorylation in the hypoxic cells and the need to increase starch breakdown (Fig. 3b) to sustain sugar supply and translocation (Baud et al., ²⁰⁰⁴; Yao et al., ²⁰²⁰) and drive substrate phosphorylation. Sucrose synthases are considered to be adenylate‐conserving mechanisms of respiration (Koch, 2004). Double sus1/sus4 mutants grow normally under aerobic conditions but show marked growth retardation under hypoxic conditions (Bieniawska et al., ²⁰⁰⁷). The strong induction of PDC1 and ADH1 after 4 h of hypoxic treatment (Fig. 3e), consistent with the initiation of fermentation to maintain substrate phosphorylation, may be a rapid and transient response. The transcript levels of these two genes in maize roots were, in fact, unchanged after 12 h of oxygen deprivation, despite an increase in ADH activity (Mira et al., ²⁰²³).

The glycolytic pathway of hypoxic roots was defined by the induction of several genes, especially GAPC‐2 and PK9 (Fig. 3e). The former was preferentially induced in the meristematic cells, with only a moderate upregulation in mature root domains. Interestingly, low oxygen did not induce GAPC‐2 in the phloem cells, likely enriched in photosynthates. PK9 was also induced in all cell types, with the exclusion of xylem, phloem, cortex and atrichoblasts in mature tissue, and columella and stem cell niche in the meristematic zone. The ubiquitous upregulation of GAPC2 by low oxygen raises questions as to whether this relates to glycolytic pathway needs, turnover of the enzyme, or its function in regulatory processes. H₂O₂ promotes the interaction of GAPC with PLDϒ, a membrane‐associated phospholipase D, activating it and inducing signaling responses that regulate NADPH oxidase activity (Zhang et al., ²⁰⁰⁹). Both GAPC‐1 and GAPC‐2 alter root development and affect sugar and amino acid balance (Munoz‐Bertomeu et al., ²⁰⁰⁹). GAPC‐1 is thought to be fundamental in roots and its activity may be required in root meristems and the root cap for normal primary root growth (Anoman et al., ²⁰¹⁵). Furthermore, both GAPCs affect abscisic acid signal transduction (Munoz‐Bertomeu et al., ²⁰¹¹).

Unlike the induction observed for several glycolytic and fermentation enzymes, hypoxia did not cause major transcriptional changes of enzymes involved in the TCA cycle. Suppression of the TCA cycle was in fact limited to atrichoblasts and procambium, while a depression in oxidative phosphorylation occurred in more cell types: trichoblasts, atrichoblasts, cortex, epidermis, and xylem of the mature tissue and lateral root cap within the meristem (Fig. 3a). While several pathways of pyruvate conversion are likely to occur during hypoxia, the conversion of pyruvate via the TCA cycle can take place when the reactions that are inhibited by the elevated redox level, for example the oxidation of 2‐oxoglutarate and other substrates (Antonio et al., ²⁰¹⁶), are bypassed by aminotransferase reactions and when nitrite is served as an alternative electron acceptor, substituting oxygen (Igamberdiev & Hill, 2004).

2. Nitrogen metabolism

Hypoxia induced nitrogen metabolism (Fig. 4b). Downregulation of GLUTAMATE DEHYDROGENASE 2 (GDH2) in the epidermal and cortical cells is a notable exception. In the highly reduced atmosphere of the hypoxic cell, it would be substrate‐limited and would be generating substrate to a hypoxia‐limited TCA cycle. Its operation in the reverse direction requires high millimolar concentrations of NH₄ ⁺ (Grzechowiak et al., ²⁰²³). Together with Pgb1, metabolizing NO via the Pgb–NO cycle (Igamberdiev et al., ²⁰⁰⁵), the strong upregulation of NIA1 and NIA2 suggests that nitrogen metabolism, and in particular reduction of nitrate to nitrite, plays an important role during hypoxia. The two NIA genes are mainly upregulated in the outer root tissues: atrichoblasts and lateral root cap (NIA1) and atrichoblasts, root cap and epidermis (NIA2), denoting increased provision (and reduction) of nitrate from the soil because of low oxygen.

The plausible accumulation of nitrite in hypoxic root tissues, because of the strong induction of NIA1 and NIA2, might also be facilitated by the more moderate induction of NIR1 in the stele, epidermis and cortex. Nitrite is likely to accumulate preferentially in the hypoxic root caps (as estimated by the pronounced induction of NIAs and the unchanged levels of NIR1), where it can become the source of NO under high redox levels typical of low oxygen environments (Igamberdiev & Hill, 2004). The slight enhancement of the expression of nitrite reductase NIR1 (AT2G15620) keeps the rate of conversion of nitrite to NH₄ ⁺ at the same or a bit higher level compared to normoxia. The excess of formed nitrite due to a higher expression/activity of nitrate reductases (NIA1/2) is readily converted in the Pgb–NO cycle, with the Km of 175 μM (Gupta et al., ²⁰⁰⁵). This avoids the accumulation of toxic millimolar concentrations of nitrite (Oke, 1966), while its lower levels are beneficial for plants under hypoxia, in particular, for the preservation of the mitochondrial structure and function (Gupta et al., ^{2017, 2020}). The excess of nitrite can be converted to NO in the side reaction of nitrate reductase (Rockel et al., ²⁰⁰²).

The root cap, where hypoxia induces NIA1 and NIA2, is the primary tissue of gravitropism perception and NO mediates gravitropism and root growth re‐directioning (Vazquez et al., ²⁰¹⁹; Mira et al., ²⁰²⁴), which is an important acclimation to hypoxia (Eysholdt‐Derzso & Sauter, 2017). Of note, NIA1 and NIA2 differentially contribute to the production of NO in Arabidopsis root cap during gravitropism responses, with nia1/nia2 mutants exhibiting gravitropism deviations (Vazquez et al., ²⁰¹⁹). In these responses, NO cooperatively works with auxin, which might also preferentially accumulate in the lateral root cap cells, as estimated by the induction of CYP79B2 (Fig. 5c).

The moderate induction of NIR1, NIA1, and NIA2 takes place in the meristematic cortical cells, where GDH2 is suppressed (Fig. 4b). In maize, the root cortex is a major site of amino acid synthesis (O'Neill & Lee, ²⁰²⁰). Besides a developmental function, nitrite can also be used as an alternative electron acceptor to oxygen under hypoxia in the Pgb–NO cycle, where the cytosolic NADH/NADPH is oxidized and ATP is generated at a limited rate under anaerobic conditions (Stoimenova et al., ²⁰⁰⁷). Thus, the ubiquitous and rapid upregulation of PGB1 observed here, which is consistent with immediate (30 min) induction of the barley PGB1 during hypoxia, might be required to convert NO back to nitrate via oxyphytoglobin (Igamberdiev & Hill, 2004). The Pgb–NO cycle (Hill et al., ²⁰²³) detoxifies NO and efficiently regulates redox state and energy balance in plants under the conditions of oxygen deficiency (Stoimenova et al., ²⁰⁰⁷).

Hypoxia also strongly upregulates ASPARTATE AMINOTRANSFERASE 2 (ASP2) in all tissues with the exception of the stem cell niche, indicating that oxaloacetate is actively directed to the formation of aspartate, with the bypass of several TCA cycle reactions. For the operation of this pathway, NH₄ ⁺ formed by nitrite reductase should be efficiently incorporated in metabolism. Under hypoxic conditions, the GS/GOGAT system, which requires ATP, is suppressed; however, an intensive nitrogen metabolism under hypoxia can generate high concentrations of ammonia and stimulate amino acid synthesis (Gupta et al., ²⁰¹²). A potential pathway involving components of the TCA cycle is shown in Fig. 4c. Such a pathway would be dependent on available nitrate, which is a factor in hypoxic stress survival (Arnon, 1937). The TCA cycle is bypassed by the reactions of aspartate aminotransferase and glutamate dehydrogenase (in the amination direction). For plant roots, Garcia‐Novo & Crawford (1973) proposed that under decreased soil aeration, this pathway may efficiently utilize NH₄ ⁺ formed during nitrate/nitrite reduction. Previously, it was shown that the isoform of glutamate dehydrogenase GDH1 can utilize ammonium upon its accumulation in plants (Melo‐Oliveira et al., ¹⁹⁹⁶).

Thus, under hypoxic conditions, the NIA‐mediated formation of nitrite might be required for developmental acclimation, that is, root bending via production of NO, as well as for metabolic adaptations by acting as an alternative electron acceptor to sustain energy production. NO, in turn, initiates the operation of the Pgb1–NO cycle that decreases redox level and results in limited ATP formation. The enzymatic and metabolic studies on the possible involvement of NADPH‐dependent amination of 2‐oxoglutarate under hypoxia can provide new evidence of the TCA cycle aminotransferase bypass.

3. Hormones and signaling

Events associated with the perception of hypoxia involve the production of ethylene and the prevention of protein turnover of ERFVII transcription factors by regulation of the N‐degron pathway through the activity of plant cysteine oxidases (PCO) (Hartman et al., ²⁰¹⁹). This pathway is fine‐tuned by mechanisms related to sugar availability (Kunkowska et al., ²⁰²³). Despite the induction of both PCO1 and 2 in many cell types (Fig. 2d), the potential consumption of both [O₂] and [NO] because of high Pgb1 levels might limit PCO activity, maintaining ERFVII levels and triggering hypoxic responses.

Ethylene biosynthetic genes are induced by hypoxia, with SAM1 and SAM2 moderately upregulated, especially in meristematic epidermis and cortex, and ACO1 more strongly upregulated in all cell types with the exclusion of xylem and stem cell niche (Fig. 5a). The upregulation of the oxygen‐requiring ACO1 in an oxygen‐deficient environment and under conditions of high and ubiquitous expression of Pgb1, which binds avidly to oxygen, raises the question of the function of ACO1 induction. The Km(O₂) of ACO1 is 67 μM (Mirica & Klinman, 2008) while the cortex [O₂] drops below 0.1 kPa (1.4 μM) within 2 h (Weits et al., ²⁰¹⁹). Pgb1 is 50% O₂‐saturated at 1.6 nM (Trevaskis et al., ¹⁹⁹⁷). From an oxygen perspective, it would appear that ACC oxidase could be severely substrate‐limited to produce ethylene. The rates of ethylene production in Arabidopsis plants are of the order of 0.2 nmol g⁻¹ FW h⁻¹ and production increases c. 20‐fold in a line where PGB1 has been silenced (Hebelstrup et al., ²⁰¹²). In Rumex species, ethylene levels reached saturation levels of 1 μl l⁻¹ within 1 h of submergence, but accumulation of ACC, with little conversion to ethylene, was observed after 2 d of submergence (Banga et al., ¹⁹⁹⁶). Based on the above, we suggest that the early induction of ACO1 in hypoxic roots might be a ‘growth resumption strategy’ required during re‐oxygenation. Re‐oxygenation after periods of hypoxia coincides with a rise in ethylene production necessary to mitigate the negative effects of oxidative stress (Hartman et al., ²⁰²⁰). Ethylene pretreatment enhances oxidative stress survival during reoxygenation (Liu et al., ²⁰²²).

HRE1 and HRE2 are not ethylene‐responsive (van Dongen et al., ²⁰⁰⁹) but do respond to low oxygen stress and are thought to be redundant in low oxygen signaling (Licausi et al., ²⁰¹⁰). The former is mainly induced by hypoxia in the cortex of the meristematic zone, as well as pericycle and endodermis of the mature root, while the latter is more markedly induced in the stele of the meristematic region (Fig. 5b), confirming previous studies (Eysholdt‐Derzso & Sauter, 2017). Hypoxia also induces HRE2 in more mature domains: trichoblasts, procambium and pericycle where it can contribute to adventitious rooting organogenesis (Eysholdt‐Derzso & Sauter, 2019). Interestingly, the hypoxic upregulation of HRE2 in the meristematic stele coincides with the upregulation of the auxin influx carrier LIKE AUXIN 1–2 LAX2 and the downregulation of the auxin efflux carrier AGRAVITROPIC 1 (AGR1) in the same cells (Fig. 5d). These changes might be related to the role of HRE2 in influencing bending of oxygen‐deficient roots through alterations in auxin flow (Eysholdt‐Derzso & Sauter, 2017). AGR1 mutation interferes with gravitropism by altering auxin flow, increasing auxin sensitivity, and decreasing sensitivity to ethylene (Chen et al., ¹⁹⁹⁸). Its regulation is mediated by kinase/phosphatase activity (Shin et al., ²⁰⁰⁵).

The different induction pattern of SAM1 and SAM2, restricted to hypoxic meristematic regions with the exclusion of root cap and stem cell niche, relative to ACO1 suggests that these genes might also fulfill functions not necessarily related to ethylene synthesis. SAM is also a key enzyme in the formation of polyamines that are important in abiotic stress tolerance (reviewed in Shabala et al., ²⁰¹⁶). ROS and polyamines integrate into signalling networks that are sensed by ion transport machinery. ROS‐triggered H₂O₂ activates Ca²⁺‐permeable hyperpolarization‐activated cation channels in a specific domain of plant roots. Changes in expression of Ca²⁺‐related genes occur primarily in hypoxic cells apically located to the root cap. These changes include the upregulation of CIPK15 and 7, downstream effectors of calcineurin B‐like proteins and Ca⁺² sensors (Ho et al., ²⁰¹⁷), the calcium channels CNGC5 and 12 (Tan & Zwiazek, 2019; Zhang et al., ²⁰²¹) and CML38, a regulator of gene silencing calmodulin‐like proteins (Lokdarshi et al., ²⁰¹⁶). The latter gene is highly induced by low oxygen and increases sensitivity to stress when mutated. Hypoxia also downregulated both the calnexin CNX1, a Ca²⁺−binding ER chaperone (Liu et al., ²⁰¹⁷), and CRT1, a calreticulin that restores ER Ca²⁺ levels (Probsting et al., ²⁰²⁰). Loss of function of CRT1 results in activation of the ethylene signaling pathway (Probsting et al., ²⁰²⁰), suggesting a possible interplay between Ca²⁺ networks and ethylene signaling.

V. Conclusions

The early hypoxic responses in Arabidopsis roots involve an extensive transcriptome rewiring defined by the consistent upregulation of the core hypoxia‐responsive genes. Excluding the QC showing early signs of degradation, these changes are pronounced in the meristematic cells. The induction of diverse processes, some of which are linked to carbon and nitrogen metabolism, within the meristematic cells might influence the viability of the QC and stem cell niche through the supply of carbon to fuel less efficient energy production mechanisms.

Metabolically, the ubiquitous and strong upregulation of PGB1, SUS, PDC1, and ADH1 in hypoxic roots suggests coordination of nitrogen and carbon metabolism. The minimal changes in transcription of glycolytic and TCA cycle genes and the universal increases in SUS in all root cells, but not the QC, relate to increased demand for carbohydrate to drive glycolytic fermentation. The significant and substantial increases in gene expression related to nitrogen metabolism indicate that it plays an equal role to glycolytic metabolism in early acclimation to hypoxic stress. Nitrite, produced via NIA activity, might be required not only for developmental acclimations but also as an alternative electron acceptor to sustain energy production in an energy‐limited environment. Availability of nitrate in hypoxic cells can contribute to the bypass of the TCA cycle with NO utilized in the Pgb1‐nitric oxide cycle to decrease the redox level and sustain energy production. Finally, the restricted upregulation of ERFVII to a few cell types of mature and meristematic regions and the strong upregulation of the final step of ethylene biosynthesis suggest that the response to the hypoxic stress is well advanced by 4 h. This agrees with the function of ERFVII transcription factor stability, through PCO activity, to modulate acclimations processes to low oxygen.

Competing interests

None declared.

Author contributions

CS, RDH, and OW planned and designed experiments. MM and SMR performed the experiments. SMR and OW analyzed the data. RDH, SMR, AUI, and CS wrote the manuscript with input from all authors. All the authors have read and approved the final version of this manuscript. RDH and SMR contributed equally to this work.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Supporting information

Fig. S1 Relative expression of selected hypoxia responsive genes.

Fig. S2 Correlations of transcriptomes in each growth zone.

Fig. S3 UMAP of all nuclear transcriptomes with labels from root atlas.

Fig. S4 UMAP annotations separated by conditions and replicates.

Fig. S5 Cell‐type expression of selected marker genes.

Fig. S6 Number of genes expressed and core hypoxia DEG expression of cell types.

Fig. S7 Dotplots showing cell‐type expression of genes in the glycolytic pathway.

Table S1 Sequencing metrics and cell statistics.

Table S2 Genes differentially expressed in each cell type under hypoxic conditions.

Table S3 Enriched Gene Ontology terms in each cluster of Fig. 2(b).

Table S4 Enriched Gene Ontology terms of DEGs specific to the Epidermis, Stele, or Cortex initials (I).

Table S5 Enriched transcription factor binding motifs in the promoters of DEGs from the Epidermis, Stele, or Cortex initials (I).

Table S6 Enriched KEGG pathways within cell types.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.