Introduction

Oxygen (O2) is a central molecule of eukaryotic metabolism, produced in plants, algae and bacteria through photosynthesis, and required for aerobic respiration and a multitude of biochemical reactions1,2. Acute hypoxia (reduced environmental O2 availability) is the central problem of flooding and waterlogging, agriculturally important abiotic stresses that are increasing due to climate change3. Hypoxia is also an important component of normal plant development (chronic hypoxia for example, in apical meristems) or induced by pathogens1,2. A mechanism for oxygen-sensing in plants was discovered more than 10 years ago4,5, and is analogous to that of the animal Hypoxia Inducible Factor (HIF) system, including proteasomal destruction of transcription factors following covalent attachment of oxygen via dioxygenase enzymes, though the mechanisms are not related2. Work in mammalian systems initially showed that the Arg/N-degron pathway acts as a sensor for both oxygen and nitric oxide (NO)6,7 Plant oxygen-sensing was shown to require the O2 (and NO)-sensitive destruction of substrate proteins that initiate with amino-terminal Cysteine (Nt-Cys), including the group VII ETHYLENE RESPONSE FACTOR (ERFVII) transcription factors, VERNALIZATION (VRN)28 and LITTLE ZIPPER (ZPR)29, through the PLANT CYSTEINE OXIDASE (PCO) branch of the PROTEOLYSIS (PRT)6 N-degron pathway, hereafter the PCO N-degron pathway (Fig. 1a). Current understanding of the mechanism is that Nt-Cys is revealed in these Met1-Cys2 initiating substrates after cleavage of the methionine by MetAP activity. Following oxidation of Nt-Cys by PCO enzymes, oxidised Cys is arginylated by ARGINYL TRANSFERASEs (ATEs), allowing recognition by the E3 ligase PRT6, ubiquitylation and degradation via the Ubiquitin Proteasome System (UPS). Reduced oxygen availability inhibits PCOs10 resulting in accumulation of substrates. Methionine excision of substrates has not been shown in plants, though is predicted from the substrate specificity of highly conserved MetAP activity11. Biochemical analysis in vitro using recombinant enzymes purified from Escherichia coli showed that PCOs add molecular O2 to Nt-Cys to give Nt-Cys(SO2-) (sulfinate), that can be arginylated by ATEs12. PCOs were originally identified in plants13, and more recently an N-terminal cysteine oxidase equivalent, ADO, was also identified in animals14. Although there is an absolute requirement for NO for substrate degradation in vivo through this pathway6,15, the site of action within the pathway remains unknown16. Previously it was shown that the stability of ERFVIIs can be increased genetically by removing components of the PCO N-degron pathway including PCOs13, ATE5 and PRT64,5. Stability is also increased through this pathway in the absence of endogenous NO17 or by perturbing mitochondrial function18. In addition, environmental reduction of either oxygen or NO, increased ethylene, low temperature, heavy metal stress and herbicides also enhance ERFVII stability8,15,19,20,21. ERFVII degradation is also controlled by other E3 ligases, in PCO N-degron pathway independent, mechanisms22,23.

Fig. 1: ERFVIIs as determinants of plant hypoxia tolerance.
figure 1

a Schematic of the currently accepted PCO N-degron pathway. The changing N-terminus of ERFVIIs is shown in relation to enzyme activities. MetAP, METHIONINE AMINOPEPTIDASE; PCO, PLANT CYSTEINE OXIDASE; ATE, ARGINYL TRANSFERASE; PRT6, PROTEOLYSIS6. The conversion of amino-terminal Cys to Cys-sulfinate (SO2-) is shown. Residues indicated by three letter code. b Root tip survival of wild-type (Col-0) and mutant seedlings subjected to 4 h hypoxia and 3 days recovery. MA-ERFVIIs indicates individual promERFVII:C2A-ERFVII transgene in Col-0 background. Data are presented as mean ± SD (n = 9) significant differences denoted with letters for one-way ANOVA (p < 0.05). c Germination under different levels of ambient oxygen for wild-type (Col-0) and mutant seeds. Data are presented as mean ± SD (n = 3). d Images of four-week-old wild-type (Col-0) or erfVII plants grown in sandy-clay loam soil in control conditions or upon waterlogging treatment for 7 days. Scale bar is 1 cm. e Weights of rosettes and roots of four-week-old wild-type (Col-0) and erfVII mutant with or without waterlogging treatment of the duration of 7 days. Data are presented as mean ± SD (n = 8), t-test, ***p-value < 0.0002 and ****p-value < 0.0001, ‘ns’ not significant.

Both O2 and NO-sensing through the PCO N-degron pathway are important for ERFVII (and other substrate) regulated tolerance to multiple abiotic stresses. Sensing of low oxygen enhances plant tolerance of submergence and waterlogging, and is an important component of plant agricultural and natural environmental adaptation to acute hypoxia24,25,26,27,28 as well as chronic hypoxia9,29,30,31,32,33. Sensing of NO through ERFVIIs contributes to tolerance to salinity, drought and heat17 and has also been shown to pre-acclimate plants to low oxygen stress19. Oxygen-sensing through ERFVIIs has a role in environmental adaptation. The rice ERFVII SUBMERGENCE (SUB)1A, a key locus for improving rice submergence tolerance, was shown not to be a substrate of the PCO N-degron pathway, despite being a Met1-Cys2 initiating protein because the carboxyl terminal structure prevents its degradation4,34. Recently it was shown that adaptation to geographical aridity is associated with differences in alleles of two cis-elements upstream of the ERFVII RELATED TO AP (RAP)2.12 specifically within the species Arabidopsis thaliana27, and within angiosperms alteration of the sensitivity of oxygen-sensing is associated with adaptation to altitude26.

Importantly, although many components of the plant oxygen-sensing mechanism have been defined, several key aspects remain unresolved, un-investigated or inferred from previous work in mammalian systems. For example, it has not been shown that substrates are degraded through the proteasome via ubiquitylation, or the relative activity of the PCO N-degron pathway in relation to substrate degradation in response to hypoxia. Although in vitro analysis using recombinant PCOs and ERFVII oligopeptides showed oxidation of amino-terminal Cys to Cys-sulfinate (SO2-)12, in vivo data from substrates purified from mouse L cells showed that the amino-terminal Cys of mammalian Regulator of G protein Signalling (RGS)4 was Arg-Cys(SO3-) (sulfonate), not sulfinate35. More recently work in mammalian systems has suggested that Cys(SO3-) may also be a product of oxidative stress, resulting in mammalian substrate degradation through p62-mediated autophagy36. Finally, although ERFVIIs have been shown to activate hypoxia-related gene expression, their combined effect on the regulation of genome expression is not known. Here we experimentally address important gaps in understanding of the known plant oxygen-sensing system, defining the role of ERFVIIs in regulating gene expression though the PCO N-degron pathway, their influence in root responses to acute hypoxia, their modulation through environmental and genetic factors, and the influence of oxygen on the chemical nature of the amino-terminus.

Results

ERFVIIs non-autonomously enhance tolerance to acute hypoxia

To analyse the influence of ERFVIIs on physiological responses to different forms of acute hypoxia we used a genetic approach removing the activity of all five A. thaliana ERFVIIs in the rap2.12 rap2.2 rap2.3 hypoxia responsive erf (hre)1 hre2 pentuple mutant (hereafter abbreviated to erfVII) and prt6 erfVII28. Comparison of prt6 erfVII with the single E3 ligase mutant prt6 allows an understanding of the effect of constitutive stabilisation of ERFVII proteins (in prt6) against a background without ERFVII activity (prt6 erfVII). Analysis of tolerance of seedling root tips to severe rapidly evoked hypoxia (<0.5% O2) showed that prt6 greatly enhanced survival, as was previously shown19, but removal of all ERFVII activity in the prt6 background (prt6 erfVII) strongly reduced tolerance (Fig. 1b, Supplementary Fig. 1a). In comparison, erfVII root tips showed a similar tolerance to wild-type (accession Col-0) seedlings but tolerance was reduced compared to prt6 erfVII. To determine the individual contribution of each ERFVII we analysed root tip survival of hypoxia in the prt6 genetic background in which only one of the five ERFVIIs was functional (for example, prt6 RAP2.12 rap2.2 rap2.3 hre1 hre2). This analysis showed that no single ERFVII was capable of enhancing survival to the level of prt6 (where all ERFVIIs are stable) (Fig. 1b). Conversely, each individual constitutively stabilised ERFVII (through conversion of Cys2 to Ala [C2A] derived from a promoterERFVII:MA-ERFVII transgene28) in a background of WT ERFVII activity could significantly increase root tip hypoxia survival compared to Col-0 untransformed control (Fig. 1b). C2A-ERFVIIs could not enhance hypoxia-related gene expression (with the exception of SUS4 for MA-RAP2.12 (Supplementary Fig. 1c). Both RAP2.12 and RAP2.2 when individually stabilised by the prt6 background did enhance expression of the three core hypoxia transcripts analysed, and RAP2.3 one transcript (Supplementary Fig. 1c). Hypoxia negatively affects seed germination and a role for ERFVIIs through oxygen-sensing was previously proposed4. We analysed the role of ERFVIIs in regulating seed germination under hypoxia (Fig. 1c, Supplementary Fig. 1b), that showed a key role specifically for RAP2.2 in enhancing germination potential at low oxygen. The influence of ERFVIIs on aerial and root biomass response to prolonged (7 days) waterlogging in natural soil was also investigated. Both dry and fresh weights of rosettes and roots of wild-type plants were reduced in response to treatment, however reduction of biomass was far greater in the erfVII mutant (Fig. 1d, e). To understand the effect of waterlogging on root development and 3D-architecture in situ in the soil we used non-invasive X-ray computed micro-tomography (µCT) imaging of wild-type and erfVII roots of 4-week-old plants grown in well-watered (control) or waterlogged mixed sandy-clay loam soil for 7 days (Fig. 2a, b, Supplementary Fig. 2, Supplementary Movies 1 and 2). µCT imaging of control well-watered soil showed no appreciable difference in root development, whereas in waterlogged soil root total volume, maximum depth and surface area (referred as region of interest, ROI) were significantly reduced both in wild-type and erfVII plants, with a more drastic reduction observed in the erfVII mutant. In particular, total root volume and surface area in waterlogged conditions decreased by 99 and 92% reduction respectively in erfVII compared to 29 and 26% reduction in Col-0.

Fig. 2: ERFVII influence on root structure in soil.
figure 2

a Representative 3D rendered X-ray computed micro-tomography front view images of four-week-old wild-type (Col-0) and erfVII mutant roots grown in sandy-clay loam soil under control conditions or waterlogging treatment conditions for 7 days. Scale bar is 15 mm. b Root total volume (mm3), root region of interest (ROI) surface area (mm2) and maximum root depth calculated from X-ray computed micro-tomography images of wild type (Col-0) and erfVII mutant roots with or without waterlogging treatment, as described in (a). Data are presented as mean ± SD (n = 4). For µCT imaging, four replicates were imaged, data were analysed by two-way ANOVA and different letters indicate significant differences (p-value < 0.05). Scale bar is 15 mm.

Defining the influence of ERFVIIs on the seedling transcriptome

To understand the influence of ERFVIIs on global transcript abundance we compared transcriptomes using RNAseq data derived from 7-day-old seedlings grown in continuous light of Col-0 (wild-type), with prt6 and prt6 erfVII (Supplementary Data 13). Principal Component Analysis (PCA) showed that in comparison to prt6, removal of ERFVII genetic activity in prt6 erfVII strongly influenced gene expression in the prt6 background, suggesting that ERFVIIs promote a large effect on the transcriptome of prt6-1 (Supplementary Fig. 3a). In prt6-1 (where all substrates of the PRT6 N-degron pathway are stabilised) only ~2.5% of the genome was differentially regulated compared to Col-0, as opposed to 25% compared to prt6 erfVII, and 28% comparing Col-0 to prt6 erfVII (Supplementary Data 13). Interestingly comparison of prt6 vs prt6 erfVII with Col-0 vs prt6 erfVII showed a very high level of similarity of differentially (both up and down)-regulated genes (Fig. 3a) suggesting that ERFVIIs are at least partially active in Col-0 under normoxia. Analysis of Gene Ontologies demonstrated that the major functional groups associated with ERFVII genetic activity in the comparison of prt6 to prt6 erfVII were related to biological processes of response to hypoxia (Fig. 3b, Supplementary Data 4, 5). Analysis of genes previously shown through meta-analysis of available transcriptome datasets to be differentially regulated in response to hypoxia37 showed that almost half were shared with transcripts differentially regulated in prt6 compared to prt6 erfVII (Fig. 3c). The majority of a set of 49 core hypoxia-responsive transcripts, previously shown to be upregulated in diverse cell populations of A. thaliana seedlings38 were also more highly expressed in prt6 than prt6 erfVII, more highly in prt6 than Col-0 and in hypoxia compared to normoxia15 (Fig. 3d, Supplementary Data 6). Out of 49 core hypoxia genes 30 were differentially regulated in prt6 compared to prt6 erfVII and also shown previously through genome wide chromatin immunoprecipitation (ChIP) to be bound by the ERFVII HRE2 under hypoxic stress (Supplementary Data 6)39. These included PCO1 and PCO2, core components of the oxygen sensing pathway (Fig. 1a). Moreover, many genes differentially expressed between prt6 and prt6 erfVII contain copies of the Hypoxia Responsive Promoter Element (HRPE40) in the −1000bp promoter region or genomic locus region (Supplementary Data 710). Genes upregulated in prt6 that contain the HRPE and were shown to bind HRE2 in vivo showed stronger over-representation of GO terms associated with “hypoxia” than those upregulated in prt6 erfVII (Supplementary Fig. 3B, Supplementary Data 11).

Fig. 3: Influence of ERFVIIs on the seedling transcriptome.
figure 3

a Proportional Venn diagram representation of gene sets for Col-0 (wild-type) vs. prt6 erfVII and prt6 vs. prt6 erfVII. In each case numbers of total differentially regulated genes (up or down) are shown. b Top 10 most significant GO terms associated with transcripts differentially regulated in prt6 vs. prt6 erfVII, Fisher’s one-tailed test (p < 0.05). c Proportional Venn diagram representation showing overlapping gene sets between genes differentially expressed between prt6 vs prt6 erfVII and hypoxia vs normoxia37. Total number of differentially regulated genes are shown next to comparisons, in set identities within the intersecting set. d Differential expression of a set of core 49 hypoxia upregulated genes38 in transcriptome comparisons, sorted by fold difference in prt6 vs prt6 erfVII.

Genetic and environmental determinants controlling ERFVII stability

Although several genetic and physiological components have been shown to increase stability of ERFVII proteins no direct comparison of the relative influence of these components has been carried out. Here we analysed how ERFVII protein steady-state level is influenced by genetic and environmental factors, using as an exemplar RAP2.3. Using a 35S:RAP2.33xHA transgene construct15, immunolocalization showed that RAP2.33xHA was undetectable in control 4-day-old seedling roots, but was strongly stabilised and localised in the nucleus after 3 h treatment with the proteasome inhibitor bortezomib (Fig. 4a). Using a ubiquitin fusion technique (UFT41,42) construct previously designed for analysis of protein stability in animal systems43, we developed a plant transformation vector that provides expression of a single protein cleaved by constitutive deubiquitinase activity to reveal C-terminal Cys2-RAP2.33xHA (Supplementary Fig. 4a). This construct has the advantage of monitoring RAP2.3 stability without the requirement for MetAP activity to remove the initiating Met1. In wild-type Col-0 accession seedlings grown under constant light RAP2.33xHA abundance was very low, but was significantly enhanced in the presence of the proteasome inhibitor bortezomib, where clear addition of high molecular weight polyubiquitin moieties was visible (Fig. 4b, Supplementary Fig. 4b). In comparison in the prt6-1 background RAP2.33xHA was also stabilised, but not to the same degree as in the presence of bortezomib. Addition of bortezomib increased stability in prt6-1 and also resulted in a polyubiquitin ladder (Fig. 4b).

Fig. 4: Modulation of RAP2.3 stability by genetic and environmental factors.
figure 4

a In situ immunolocalization of RAP2.33xHA in control or bortezomib (50 μM) treated roots of 4-day-old seedlings, scale bars 20 μm. b Western blot analysis of RAP2.33xHA abundance in control or bortezomib treated WT or prt6-1 mutant seedlings. c Time-course Western blot analysis of RAP2.33xHA abundance in WT seedlings in response to submergence or hypoxia treatment. d Time-course of expression of ADH1 in WT seedlings in response to submergence or hypoxia treatment. Data are presented as mean ± SD (n = 3). f Western blot analysis of RAP2.33xHA abundance in WT seedlings in response to constant contact or transfer to media with no, low or high supplemented Fe2+. e Western blot analysis of RAP2.33xHA abundance in WT seedlings following transfer from hypoxia (60 min) to normoxia. g Western blot analysis showing relative abundance of RAP2.33xHA in response to bortezomib (BZ), cPTIO (NO scavenger), hypoxia or submergence. h Western blot analysis showing relative abundance of RAP2.33xHA in low altitude accession Col-0 in comparison to high altitude Sha at either 21% or 15% ambient oxygen. BZ treatment for light-grown seedlings. Ponceau or CBB staining of Western blots is shown.

The influence of two forms of reduced oxygen availability on RAP2.33xHA stability were assayed, either by submergence in the dark or dark incubation in a chamber with reduced oxygen (hypoxia). Whereas the former results in entrapped intracellular ethylene the latter does not. RAP2.33xHA protein was rapidly stabilised by both treatments in comparison to control seedlings left in the dark, for submergence that occurred within 10 minutes (Fig.4c, Supplementary Fig. 4c). ERFVII-regulated ADH1 (Supplementary Data 3) was rapidly induced under both conditions after the stabilisation of RAP2.33xHA (Fig. 4d), as were ERFVII-regulated PCO1 and SUS4 (Supplementary Fig. 4c). To determine the influence of MetAP activity (Fig. 1a) we analysed protein accumulation in submergence for the 35S:RAP2.33xHA transgenic line used for immunolocalization (Fig. 4a). Stabilisation of RAP2.33xHA in this line showed very similar kinetics to RAP2.33xHA from the UFT transgene, indicating that MetAP activity does not influence accumulation kinetics (Supplementary Fig. 4e). Following return to normoxia, RAP2.33xHA abundance reduced to very low levels within 30 to 60 min (Fig. 4e). Previously it was shown that PCO enzymes require Fe2+ to oxidise amino-terminal Cys10 and zinc deficiency was shown to enhance ERFVII stability21. Constant growth of seedlings or transfer to media +Fe or -Fe showed that under Fe2+ starvation conditions RAP2.33xHA was stabilised (Fig. 4f). In comparison to bortezomib, RAP2.33xHA abundance in response to physiological treatments that reduce oxygen or NO was lower, indicating either incomplete inactivation of the PCO N-degron pathway or alterative mechanisms for degrading ERFVIIs (Fig. 4g). Abundance was higher in the presence of the NO scavenger cPTIO than under reduced oxygen conditions.

Finally, we analysed the influence of genetic adaptation to altitude in the regulation of ERFVII stability in relation to atmospheric oxygen. We analysed two accessions of A. thaliana originally collected at 100 m above sea level (Col-0 pO2 21 kPa) or 3400 m (Sha pO2 13.9 kPa). In etiolated seedlings at pO2 21 kPa RAP2.33xHA was more abundant in Col-0 than Sha. However, under lower ambient oxygen (15%, equivalent to 14.7 kPa) RAP2.33xHA was as abundant in high altitude accession Sha as in low altitude Col-0 (Fig. 4h).

Characterisation of ERFVII amino terminal modifications

We investigated the chemical changes that occur in vivo at the amino-terminus of ERFVIIs as a result of the action of the PCO N-degron pathway. To do this we employed a wheat germ extract commonly used for in vitro coupled transcription/translation. Unlike the rabbit reticulocyte lysate (RRL), this system was previously shown not to contain an active UPS44. Therefore, whereas Cys2 proteins are synthesised and degraded in the RRL through the oxygen-dependent action of the Arg/N-degron pathway6,7,45, this does not occur in the wheat germ extract. We used this plant-derived extract to synthesise ERFVIIs and study chemical changes at the amino terminus. Following coupled transcription/translation in wheat-germ lysate and anti-HA-based pull-down of RAP2.33xHA and RAP2.123xHA bands were visible on Coomassie-stained SDS-PAGE gels at the expected sizes of the two proteins (Supplementary Fig. 5a). Excision, destaining and trypsin digestion of these bands, followed by LC-MS/MS analysis and database searching, identified them unambiguously as originating from RAP2.3 and RAP2.12 12 based on the presence of 9 matching peptides from each protein (Supplementary Fig. 5b, Supplementary Data 12). Examination of the MS and MS/MS data revealed the same extensive alterations at the amino termini of both proteins, including methionine excision, and arginylation. In addition, a Cys-sulfonate (SO3-) modification only at position 2 was also observed. High quality MS/MS spectra containing a clear series of b and y fragment ions were recorded for the two tryptic peptides to confirm this observation (Fig. 5a, b). Interestingly, no signal for Nt-Arg-Cys-sulfinate (SO2-) was detected. As a negative control for artefactual Cys oxidation, RAP2.3HIS was synthesised in E. coli, which does not possess the enzymatic machinery for Nt-Cys oxidative processing. As expected, no oxidation or subsequent arginylation was detected, and the unmodified Nt-tryptic peptide was found using the LC-MS/MS methodology with the Met1 removed (from the action of endogenous E. coli MetAP activity; Supplementary Fig. 5c).

Fig. 5: Analysis of the chemical structure at the amino-termini of RAP2.3 and RAP2.12 in a plant extract.
figure 5

MS/MS spectra of the Nt tryptic peptides of (a) RAP2.3 ([M + 2H]2 + = m/z 856.4) and (b) RAP2.12 ([M + 2H]2 + = m/z 823.9) following expression of RAP2.33xHA and RAP2.123xHA proteins, respectively, in a wheat germ extract showing the presence of Cys-sulfonic acid (C3ox) and arginylation at the amino-terminus (R-) in each case.

Previously it was shown that the two oxygens added to amino-terminal Cys by PCOs originate from molecular oxygen not water12. To establish whether the third oxygen seen in the Cys-sulfonic acid residue of RAP2.3 originated from water or molecular oxygen, RAP2.33xHA was synthesised using the wheat germ lysate in 40% (v/v) H218O. If the additional oxygen originated from water, then a pair of Nt-peptide isotopologues bearing 16O:18O in the ratio 60:40 would be expected. LC-MS analysis of the tryptic digest of RAP2.33xHA synthesised under these conditions showed that 18O from H218O was incorporated into the RAP2.3 peptide (Supplementary Fig. 5d). MS/MS performed on the [M + 2H]2+ precursor ions at m/z 856.5/857.5 showed that 18O was actually incorporated into the carboxylate group of the Asp-9 residue and that no 18O was incorporated into Cys2 (Supplementary Fig. 5e). We infer from this result that the third oxygen atom seen in the Cys sulfonate residues of both RAP2.3 and RAP2.12 is derived from molecular oxygen.

Discussion

The ERFVII transcription factors are key components of the plant oxygen sensing response, regulating the expression of genes associated with tolerance to low oxygen1,2. In order to understand the influence on the genome of the entire ERFVII cohort we developed the pentuple erfVII mutant that removes all five ERFVII genetic functions. Combination with the prt6 E3 ligase mutation (prt6 erfVII) allows an analysis of the specific role of ERFVIIs in a genetic background where all other PRT6 substrates are stable (including ZPR2, VRN28,9 and also other as yet unknown substrates16). These two genetic resources, in combination with lower order permutations for ERFVII mutations allowed an analysis of their role(s) in regulating response to low oxygen at increasing physiological levels of complexity. Enhanced root tip tolerance to acute hypoxia of the prt6 mutant was shown to require ERFVII function, as tolerance was strongly reduced in prt6 erfVII, similarly previously described for lower order combinations of erfVII mutations19. Interestingly removal of ERFVII function in the prt6 background does not reduce tolerance to the same level as erfVII, suggesting that other PRT6 substrates contribute to this response. In a prt6 background with all other ERFVIIs genetically removed each individual WT ERFVII was unable to enhance hypoxia tolerance above the level shown by prt6 erfVII. However, individually stabilised ERFVII proteins (through transgenes containing C2A) in a genetic background with WT endogenous ERFVIIs were able to strongly enhance root tip survival, all to a similar level. This indicates that this sub-family functions in a non-autonomous fashion, each ERFVII requiring the activity of others to enhance root tip hypoxia survival. Conversely hypoxia gene expression in normoxia was more greatly affected by individual stabilised ERFVIIs in the prt6 background than by individual C2A-ERFVIIs in a wild-type background, that may indicate that perception of hypoxia itself is also important for ERFVII action in enhancing tolerance. Although RAP2.12 has been shown to be the predominant ERFVII regulating response to submergence in A. thaliana accessions adapted to humid environments (including Col-0)27, germination tolerance to low oxygen was shown to be dependent on active RAP2.2. This suggests some functional diversification of ERFVIIs in the hierarchy of regulation of low oxygen responses at different stages of plant development, where different responses to hypoxia may be required.

The most important interface between plants and water availability are roots, that also face the challenge of growth under the soil in potentially hypoxic conditions when water levels rise due to flooding46,47. The role of oxygen-sensing in regulating lateral root formation was previously investigated in plants grown on agar medium, where it was shown that ERFVIIs repress auxin signalling thereby inhibiting lateral root production30. Bending of primary roots in response to external hypoxia was also reported to be dependent on ERFVIIs and alteration of auxin signalling, again in plants grown on agar medium48. However, no analysis has previously been undertaken to understand how ERFVIIs influence root 3D development in natural soils. To understand the role of ERFVIIs in roots growing in soil we used non-invasive 3D X-ray µCT imaging. This allowed an analysis of the tridimensional root architecture under non-stressed and waterlogged conditions, highlighting the strong reduction of root development (more than 90%) that occurs in natural soil in the absence of ERFVIIs (Fig. 2a, b, Supplementary Fig. 2, Supplementary Movie 1, 2). Analysis of aerial rosette growth showed a very strong response to root waterlogging, including reduced growth and increased anthocyanin synthesis, once again with a more severe effect in the erfVII mutant compared to wild-type. A slightly reduced growth, although not significant, was also observed in erfVII roots and rosettes under non-stressed conditions, suggesting a role of ERFVIIs in plant growth and development, worthy of further investigation. This observation is in agreement with the recent finding that lack of functional ADH1 or PDC1 leads to a growth penalty under normal (aerobic) growth conditions49. These results highlight the importance of ERFVIIs for root function in the soil and suggests that evolution of ERFVIIs was an important determinant in the ability of roots to survive in hypoxic soil environments.

Having shown the strong effect of ERFVIIs on physiological responses to low oxygen, we analysed total ERFVII-regulated gene expression by comparing prt6 with prt6 erfVII. Analysis of differentially expressed genes highlighted the importance of ERFVIIs for regulating the expression (both positively and negatively) of transcripts involved in response to low oxygen (Fig. 3, S3 and Supplementary Data 4, 5). ERFVIIs were shown to regulate the expression of around 25% of the genome in seedlings. In addition, ten times the number of genes were differentially expressed in comparisons including prt6 erfVII, compared to Col-0 vs. prt6, indicating the profound influence of stabilised ERFVIIs on gene expression. Interestingly the transcriptomic state of Col-0 is much more similar to prt6 than prt6 erfVII, a feature highlighted by analysis that showed a strong overlap of similarly differentially expressed genes (Fig. 3a). This suggests that in Col-0 seedlings grown in continuous light ERFVIIs are at least partially active (and therefore stable), as was recently observed for dark grown etiolated seedlings26 and suggested by the repression of oxygen-sensing by mitochondrial retrograde signalling during normal development18. Previous work has highlighted the importance of environmental stress in early growth stages of plant development50,51, suggesting that some level of stabilisation of ERFVIIs during seedling establishment may have adaptive advantage. In line with this suggestion, it has also been shown that ERFVIIs are more active at early growth stages compared to later development in A. thaliana29,52. Approximately 50% of hypoxia-associated transcripts were differentially expressed in prt6 compared to prt6 erfVII, suggesting that ERFVIIs may not control the complete response to low oxygen, and that other mechanisms contribute to the hypoxia-related transcriptome53,54,55,56. The HRPE, previously identified in the LBD41 promoter40, and shown to be a key cis-element associated with ERFVII function39,40 was present in the promoter or gene regions of many genes whose transcripts are differentially regulated by ERFVIIs, suggesting direct regulation of these genes. In addition, there was a strong over-representation of hypoxia-related transcripts in the overlapping set representing genes differentially regulated by ERFVIIs that contain the HRPE and that also were shown to bind HRE2 in genome-wide ChIP analysis39. In combination with previous datasets analysing the role of individual or combinations of ERFVIIs19,40,57, these data highlight the overall importance of ERFVIIs in regulating the expression of the genome in response to low oxygen, as a response to inhibition of the PCO N-degron pathway.

Direct visualisation of ERFVII abundance through immunolocalization and Western blot analysis allowed an investigation of the relative influences of environmental and genetic factors. Stabilisation of RAP2.33xHA in the presence of bortezomib and addition of polyubiquitin moieties demonstrate that degradation of ERFVIIs requires the proteasome (Fig. 4a, b, Supplementary Fig. 4). Genetic removal of PRT6 activity also resulted in stabilisation, though not as strongly as with bortezomib, and addition of bortezomib to prt6 seedlings revealed high molecular weight products, suggesting the activity of other E3 ligases in regulating ERFVII stability. In addition to PRT6, two other A. thaliana proteins carry a UBR-like motif (required for Arg N-degron substrate recognition), BIG and the product of AT4G2386016 and other E3 ligases unrelated to the N-degron pathway also regulate ERFVII stability22. Not unexpectedly, because PCO activity requires Fe2+, growth on low iron medium resulted in increased RAP2.33xHA abundance, suggesting that ERFVII function might be an important component of the plant low iron response as shown for the response to zinc excess21. Both hypoxia treatment and submergence resulted in rapid stabilisation of RAP2.33xHA, for submergence within 10 minutes, preceding accumulation of RNA for ERFVII-regulated transcripts. Such a rapid onset of stabilisation indicates that under dark submergence intracellular oxygen is used up very rapidly resulting in strong inhibition of PCO activity. A model for the function of RAP2.12 has been proposed whereby under normoxia a population of molecules is tethered to the plasma membrane and released under hypoxia to migrate to the nucleus and activate gene expression5. Whilst not precluding this possibility, our results indicate that a simpler explanation only requires that ERFVIIs are stable after translation following very rapid intracellular oxygen depletion and subsequently translocate to the nucleus and activate transcription (Fig. 6). RAP2.12 has been studied as the most influential ERFVII in A. thaliana, a recent study showed that this is because a subset of accessions of this species, including Col-0, are adapted to humid environments through specific cis-element changes that enhances constitutive expression of RAP2.1227. This indicates that in other species RAP2.12 orthologues may not necessarily take a predominant role. We also observed that none of the five ERFVIIs could autonomously enhance root hypoxia tolerance, that required a combination of ERFVII action (Fig. 1b).

Fig. 6: Model for plant oxygen sensing through the PCO N-degron pathway based on data presented in this work.
figure 6

ERFVII transcription is shown at the top of the model. Subsequent enzymatic steps on the protein are shown, either oxygen-independent (MetAP) or oxygen dependent. Activation of nuclear gene expression by stabilised ERFVIIs in the absence of oxygen, through the Hypoxia Responsive Promoter Element (HRPE) is indicated. “?” indicates an unknown component required for the addition of a third oxygen to form Cys sulfonate.

Previous work showed that recombinant PCOs can oxidise amino-terminal Cys to Cys-sulfinate (SO2)- that can act as a substrate for ATE10,12, but when purified from mouse L cells the amino-terminus of Cys2 protein RGS4 was shown to be in the form Arg-Cys(SO3)35, that was also shown to be the case in response to oxidative treatments of different cell types36. Using a wheat-germ extract we observed the formation of Cys(SO3)- and arginylation of both RAP2.3 and RAP2.12. Analysis in the wheat germ extract system precludes oxidation through ROS or other chemical oxidising agents as synthesis is carried out in the presence of DTT. PCOs are dioxygenases and thus catalyse the incorporation of two oxygen atoms from molecular oxygen into the amino-terminal Cys residue. Although it is known that this degree of oxidation is sufficient for recognition by ATE1, and hence arginylation of amino-terminal Cys proteins12, in our analyses using wheat germ extract we were only able to detect arginylated peptides possessing the sulfonate (SO3) level of oxidation. This raises the question of the origin and significance of the third oxygen atom in the sulfur acid. As the third oxygen was derived from molecular oxygen, we suggest that it is most likely introduced by the action of a yet uncharacterised enzyme. This monooxygenase/oxidase presumably acts after PCO oxidation, but whether it precedes arginylation or not remains an open question given that ATE1 is able to act on proteins bearing either amino-terminal Cys sulfinic or sulfonic acids12. As the pathway is highly conserved in eukaryotes2 (with ADO acting as the oxygen sensing enzyme in mammals14) it is possible that such an enzyme also exists in non-plant systems.

Here we investigated the influence of ERFVIIs on outputs of the PCO N-degron oxygen-sensing pathway. Our observations provide a holistic understanding of the action of these transcription factors in influencing genome expression through PRT6 gated stability, and on root physiological response to low oxygen at increasing environmental complexity and on how their activity is modulated. In particular, we provide evidence of the importance of ERFVIIs in maintaining root growth in hypoxic soil conditions associated with waterlogging. We defined exact chemical changes at the amino-terminus of ERFVIIs suggesting components that are required for oxygen-sensing through this pathway in both plant and animal systems. These data indicate avenues for future research to completely define oxygen-sensing through the PCO N-degron pathway and biotechnological approaches to enhance tolerance of crops to flooding and waterlogging.

Methods

Plant genetic resources

Several A. thaliana genetic resources used in this investigation were reported previously including prt6-1 and ate1 ate258, erfVII and prt6 erfVII28. In all cases wild-type refers to the Col-0 accession of A. thaliana. Plants containing various combinations of ERFVIIs with T-DNA insertions were obtained in the process of generating the prt6 erfVII sextuple mutant and contain the same alleles as this mutant. Accession Shakdara (Sha; NASC ID: N929, originally collected at 3400 metres above sea level at Pamiro-Alay: Tajikistan) containing pUFT:RAP2.33HA was obtained by direct transformation of the accession59.

Analysis of plant growth and development

Plants were grown and seeds collected as described60. Analysis of root tip survival following hypoxia treatment (<0.5% ambient oxygen) and waterlogging were carried out as previous described8,19. Trypan blue staining was performed on roots to investigate cell death; seedlings were incubated with 0.004% Trypan blue solution for 10 mins then decolorized for 10 minutes in distilled water. Images were taken using the ZEISS Axio Observer microscope. The root tip map was imaged using a 20× objective under the brightfield function and fluorescence of trypan blue visualized by excitation of the samples with a 633 nm laser as described61. For soil-based root growth assay, wild-type and erfVII mutant seeds were germinated directly on soil, after three days of cold-treatment at 4 °C in the dark. Plants were grown for three weeks, followed by one week of waterlogging treatment, in controlled environmental conditions with photoperiod of 12:12 h light:dark. Hypoxia and submergence treatments for Western blot analysis of RAP2.33HA abundance were carried out as previously described8. Seed germination under hypoxia was performed as previously described4.

Iron treatment

Iron treatment was done as described previously62,63 using the 35S:RAP2.33xHA transgene. +Fe and −Fe plates were made with macronutrients and micronutrients at 2 mM Ca(NO3)2, 0.75 mM K2SO4, 0.65 mM MgSO4, 0.1 mM KH2PO4, 10 µM H3BO3, 0.1 µM MnSO4, 0.05 µM CuSO4, 0.05 µM ZnSO4, 0.005 µM (NH3)6Mo7O24, 1 mM MES, 0.6% agar, adjusted to pH 6.0 and supplemented with 50 µM Fe(III)-EDTA for +Fe plates and ferrozine, an iron chelator [3-(2-pyridyl)−5,6-diphenyl-1,2,4-triazine sulfonate for low Fe (150 µM) and −Fe (300 µM) media.

Soil material and X-ray µCT imaging

Sandy-clay loam soil was collected from the University of Nottingham farm at Bunny, Nottinghamshire, UK (52.52°N, 1.07°W), air dried and then passed through a 2 mm sieve to homogenise, as previously described64. Polyvinylchloride (PVC) columns (8 cm length × 3 cm diameter) were filled with the soil to a bulk density of ~1.1–1.2 g cm−1. Four-week-old roots from control and waterlogged plants were imaged using a Phoenix vǀtomeǀx M 240 high resolution X-ray µCT system (Waygate Technologies (a Baker Hughes business), Wunstorf, Germany) at the Hounsfield Facility, University of Nottingham, UK. The scanning parameters were optimized to allow balance between a large field of view and a high resolution. Prior to µCT scanning the soil columns were placed on a layer of drying paper for 2 days, to reduce soil moisture and thus enhancing contrast between root and the soil matrix. Each sample was then imaged using a voltage and current of 112 kV and 168 µA, respectively, at a voxel size resolution of 23 µm. The scanner stage rotated through 360 degrees at a rotation step increment of 0.125 degrees over a period of 48 minutes, acquiring a total of 2880 projection images using a fast (no image averaging) multiscan of three sections with an exposure time of 333 ms. Each scan was reconstructed using DatosRec software (Waygate Technologies (a Baker Hughes business), Wunstorf, Germany). Radiographs were visually assessed for sample movement before being reconstructed in 32-bit depth volumes with a beam hardening correction of 8, no filtration was applied to the image. Reconstructed sections were processed in VGStudioMAX (version 2.2.0; Volume Graphics GmbH, Heidelberg, Germany) with manual alignment of the 3 sections to combine to a single 3D volume.

Image analysis

Merged volumes were processed in VGStudioMAX (version 2.2.0; Volume Graphics GmbH, Heidelberg, Germany) to identify plant root material. This was done via a 3D region growing algorithm to extract root architecture, as previously described in ref. 65. Once the segmented ROI was created, VGStudioMax was used to assess the rooting characteristics between treatments. The morphological properties measured were root volume (mm3), root surface area (mm2) and maximum in-situ rooting depth (cm).

Affinity purification of RAP2.33XHA

For affinity purification of RAP2.33XHA, seven-day old seedlings grown in continuous light were treated for 2 h with 50 μM bortezomib (dissolved in DMSO). Total protein was extracted by grinding 0.5 g tissue in liquid nitrogen with 700 μl of IP buffer (50 mM Tris-Cl pH 8.0, 150 mM NaCl, 0.05% IGEPAL, 50 μM BZ, 100 mM PMSF). Extract was centrifuged at 13,000 rpm for 30 mins and supernatant (600 μl + 200 μl TBST) was incubated with 25 μl of Pierce™ Anti-HA Magnetic Beads (Thermo Fisher, UK) for 30 mins at room temperature. After incubation flow through was collected and beads were washed three times for 2 mins each with 300 μl of PBST (Tris Buffered Saline Pack (25 mM Tris, 0.15 M NaCl; pH 7.2, Thermo Fisher, UK; 0.1% (v/v) Tween-20). Finally, beads were washed with 300 μl molecular grade water. Elution of the beads was achieved by adding 100 μl of SDS-PAGE Lane Marker Non-Reducing Sample Buffer (2X), (0.3 M Tris-HCl, pH 6.8, 5% (v/v) SDS) (Thermo Fisher, UK) to the tube containing beads, followed by incubation at 95 °C for 10 min. Two μl of β-mercaptoethanol was added before loading onto an SDS-PAGE gel. Western blot analysis of RAP2.33XHA abundance were carried out as previously described8. Anti-ubiquitin (Agrisera AS08 307 1:4000 dilution) and anti-HA antibodies (Sigma, H3663-200UL; 1:10,000 dilution) were used for Western immunodetection.

Seedling transcriptome analyses

For RNA-seq analysis, seedlings of Col-0, prt6-1 and prt6 erfVII were grown under continuous light on 1/2 strength MS agar. Four biological replicates of 7-day-old seedlings of each genotype were harvested and frozen in liquid nitrogen. RNA was extracted using RNeasy plant mini kit Qiagen as per manufacturer’s instructions. RNA was sent to Deep-Seq Next Generation Sequencing Facility (University of Nottingham) for further analysis. RNA concentrations were measured using the Qubit Fluorometer and the Qubit RNA BR Assay Kit (ThermoFisher Scientific; Q10211) and RNA integrity was assessed using the Agilent 4200 TapeStation and the Agilent RNA Screentape (Agilent; 5067-5576). Poly(A) selection was performed on 1 μg of total RNA, using the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB; E7490L). Indexed sequencing libraries were then prepared using the NEBNext Ultra Directional RNA Library Preparation Kit for Illumina (NEB; E7420) and NEBNext Multiplex Oligos for Illumina, Index Primer Sets 2, 3 and 4 (NEBNext; E7500S, E7710S and E7730S). PolyA selection, cDNA synthesis and library generation were carried out using the Biomek 4000 Automated Laboratory Workstation (Beckman Coulter). Libraries were quantified using the Qubit Fluorometer and the Qubit dsDNA HS Kit (ThermoFisher Scientific; Q32854). Library fragment-length distributions were analysed using the Agilent TapeStation 4200 and the Agilent High Sensitivity D1000 ScreenTape Assay (Agilent; 5067-5584 and 5067-5585). Libraries were pooled in equimolar amounts and final library quantification performed using the KAPA Library Quantification Kit for Illumina (Roche; KK4824). The library pool was sequenced on the Illumina NextSeq500 using two NextSeq500 High Output 150 cycle kits (Illumina; 20024907), to generate around 30 million pairs of 75-bp paired-end reads per sample.

Low quality reads and adaptor sequences were removed from raw reads using default parameters of skewer v0.2.2 software66. Filtered reads were mapped onto the reference genome (TAIR10) using the default parameters of Hisat2 v2.1.0 software67. Only uniquely and correctly mapped reads were kept in bam files for the further analysis. Read counts for each gene are calculated using featureCounts v1.6.0. PCA was used to identify that the biological replicates are close enough to each other. DESeq2 v1.24.0 was used to detect the differentially expressed genes for each comparison. and we used p-value < 0.5 as the DESeq thresholds. Gene ontology analysis of the differentially expressed genes was done using GOSeq v1.4.0 with default settings. Quantitative rtPCR was carried out as previously described18 using a selection of known hypoxia up-regulated genes68. A list of oligonucleotide primers used for qRT-PCR is given in Supplementary Data 13. Gene Ontology (GO) representation analyses were carried out using the G:Profiler web resource (https://biit.cs.ut.ee/gprofiler/gost). Search for hypoxia-related cis-elements was carried out using the HRPE (GCCVCYGGTTTY) sequence69 at MEME Suite 5.4.1 FIMO (https://meme-suite.org/meme/tools/fimo).

Western blot and immunolocalization analyses of ERFVII stability

Western blots were carried out as previously described26. Primary antibodies used were anti-HA (Sigma, H3663-200UL; 1:1000 dilution), anti-FLAG (Sigma, F1804-200UG, 1:2000 dilution) and secondary antibody Goat anti-Mouse IgG1, HRP from Thermo Fisher Scientific, PA1 74421, (1:10000 dilution). Full scan uncropped Western blots are provided (Source Data 1). Immunolocalization of RAP2.3 was carried as described70. Four-day-old seedlings containing the 35S:RAP2.33XHA transgene were treated for 3 h with 50 μM bortezomib or DMSO (control). Anti-HA primary antibody (Roche, Lewes, UK) and Alexa-Fluor-488-coupled anti rat secondary antibody (1:200 dilution) (Molecular Probes, Carlsbad, CA) were used for detection. Seedlings were counter stained using Propidium Iodide and visualised using confocal microscopy.

Generation of ubiquitin fusion technique construct

We developed a Ubiquitin Fusion Technique construct based on the plasmid pKP496 protein coding region FLAGDHFRHA-UBIQUITIN-FLAG43. We replaced epitope tags to produce FLAGDHFR-UBIQUITIN-3xHA (pCD1). As described in43 this provides a SacII restriction site downstream of the UBIQUITIN moiety that allows cloning in frame of test proteins with defined amino-terminal residues. In this case we cloned RAP2.3 starting at Cys2, so that upon constitutive cleavage in vivo by deubiquitinating enzymes Cys2-RAP3.33xHA is released from the pre-protein. This construct was subcloned into plant transformation gateway vector pH7wg271 providing ectopic constitutive expression from the 35S CaMV promoter to generate the construct pUFT:RAP2.33HA. Plants were transformed as previously described59.

Affinity purification of in vitro produced ERFVII

Protein coupled in vitro transcription/translation was carried out using a TnT® SP6 High-Yield Wheat Germ Protein Expression System (HYWGE) (Promega, USA) according to the manufacturer’s recommendations. Reaction containing template pTNT-RAP2.33xHA /RAP2.123xHA HA-tagged constructs4 contained 30 μl HYWGE with 20 μl of 300 ng/μl construct purified by QIAprep Spin Miniprep Kit (QIAgen, Crawley, UK)), reaction mixtures were incubated for 2 h at 25 C. Three reactions of TnT® SP6 High-Yield Wheat Germ Protein Expression System (HYWGE) (Promega, USA) were added separately to 50 μl of Pierce™ Anti-HA Magnetic Beads (Thermo Fisher, UK), and each topped up with 700 μl of freshly filtered (0.2 μm pore size) BupH Tris Buffered Saline Pack (25 mM Tris, 0.15 M NaCl; pH 7.2, Thermo Fisher, UK)). Samples were placed on an orbital rotor at 90 rpm. After 30 min samples were placed on a magnetic stand, the supernatant was removed, and beads were washed 3 times with 700 μl BupH Tris buffer (pH 7.2) with 0.1% Tween 20 (Fisher Bioreagents, UK). Each wash was for 2 min and was followed by magnetic precipitation of the beads. Beads were washed with Milli-Q water. Elution of the beads was achieved by adding 30 μl of SDS-PAGE Lane Marker Non-Reducing Sample Buffer (2X), (0.3 M Tris-HCl, pH 6.8, 5% SDS) (Thermo Fisher, UK) to the tube containing beads from the first reaction, followed by incubation at 96 °C for 10 min. Following precipitation, the supernatant was removed and transferred to the tube containing beads from the second reaction and the heating and precipitation steps repeated. Finally, the supernatant was transferred to the third tube, and after heating and precipitation, the supernatant was once again recovered and 1 μl of β-mercaptoethanol was added before loading onto an SDS-PAGE gel.

Production of RAP2.33xHA in H2 18O-containing medium

An aliquot of pTNT:RAP2.33xHA plasmid (20 μl) was vacuum centrifuged until dry and resuspended in H218O (20 μl). This was then added to 30 μl of HYWGE to give a 40% H218O/ H216O solution (v/v). The cell free expression system was incubated and HA-tagged protein product purified as described above.

SDS-PAGE and gel band excision

Samples from bead elution were loaded onto a 1.5 mm wide 12% gel SDS-PAGE gel. Gels were run for 20 min at 80 V until proteins passed from the stacking gel to the resolving gel, and then for 100–120 min at 120 V (for separation in the resolving gel). Staining was carried out in ~15 mL of Coomassie InstantBlue™ Protein Stain (Exepedeon, UK) for 1 h on an orbital shaker with rotation at 80 rpm. Gels were washed with Milli-Q water (×3) and left overnight for de-staining in Milli-Q water on an orbital shaker at 80 rpm. The band corresponding to purified protein of interest was excised with a sterile scalpel and cut into small pieces ready for in gel protein digestion.

In gel protein digestion

Excised SDS-PAGE band slices were destained in 50 μl 50% acetonitrile (MeCN) (Fisher Scientific, Loughborough, UK) and 50% 100 mM ammonium bicarbonate (AmBic) (Sigma Aldrich). After 10 min incubation gel pieces were dehydrated by the addition of 450 μl of 100% MeCN with vortexing for 3 min (dehydration step). The supernatant was discarded, and gel pieces were treated with 50 μl of 10 mM dithiothreitol (DTT) (Sigma Aldrich) dissolved in 100 mM AmBic, and incubated at 55 °C for 30 min (DTT treatment). The dehydration step was repeated using 100% MeCN as above and gel slices were incubated with 50 μl of 10 mM iodoacetamide (Sigma Aldrich) in 100 mM AmBic for 30 min in the dark in order to block free Cys groups with a carbamidomethyl group. MeCN dehydration was repeated and additional DTT treatment was performed in order to consume excess IAA. After the final dehydration step gel slices were swollen with 50 μl 100 mM AmBic solution containing trypsin (0.4 ng μl−1) and incubated at 37 °C overnight. After digestion samples were vortexed, centrifuged and supernatant was moved to a new 0.5 mL Eppendorf tube. Finally, samples were dried using a Savant DNA120 OP SpeedVac Concentrator (ThermoElectron, UK). Before LC-MS analysis, dried peptide samples were reconstituted in 15 μl of 70:30 H2O:MeCN containing 0.1% formic acid and centrifuged (5000 RCF for 2 min) prior to transfer of the supernatant into tapered plastic LC vials (ThermoFisher, UK).

LC-MS and LC-MS/MS analysis of protein digests

Digests were analysed on a Dionex U3000 nanoLC coupled to a ThermoFisher LTQ FT Ultra mass spectrometer (a hybrid linear ion trap-Fourier transform ion cyclotron resonance (FTICR) instrument equipped with a 7 Tesla magnet). Samples were loaded onto a C18 Pepmap300 cartridge (10 mm, 300 Å, 5 μm particle size) and separated using a Thermo Scientific C18 Pepmap300 column (150 mm × 75 μm, 300 Å, 5 μm particle size) attached to a 30 μm Picotip emitter (New Objective) in the nanoESI MS source. The mobile phases A and B consisted of 95:5 water:MeCN (v/v) and 5:95 water:MeCN (v/v), respectively, and both contained 0.1% formic acid. Samples (3 µL) were injected in load-trapping mode. Peptides were eluted using a 30 min linear gradient of mobile phase B from 0 to 55% followed by 5 min at 90% B and 15 min column re-equilibration. The LTQ FT Ultra mass spectrometer was equipped with a standard ThermoFisher nanoESI source through which a 1.8 kV voltage was applied to the Picotip emitter. The inlet capillary of the mass spectrometer was held at 275 °C with a tube lens value of 145 V.

For the identification of peptides/proteins the mass spectrometer was operated in data dependent acquisition (DDA) mode on the five most intense ions per each survey scan. Ions were isolated in the linear ion trap within a window of m/z 8 and subjected to collision induced dissociation. Charge-state rejection of +1 ions was employed. The data were submitted to Search GUI database searches and results visualised in Peptide Shaker72,73. Parameters within Search GUI were as follows: MS Convert was used to generate.mgf data files using a 0.5 minimum peak spacing setting; X! Tandem was used for database searching with the following settings: Enzyme = Trypsin, Missed Cleavages = 2, Fixed Modifications = None, Variable Modifications = Carbamidomethylation of C, Oxidation of M, Precursor Tolerance = 0.5 Da, Product Tolerance = 0.5 Da, Fragment Ion Type = b/y, Charge State Range = +1 to +5. A FASTA decoy database file containing 250 sequences (forward and reverse) of common contaminant proteins together with the sequences of RAP2.3 and RAP2.12 was used for searches. For detailed examination of N-terminal modification of RAP2.3 and RAP2.12, data were analysed manually using Xcalibur software (ThermoFischer). Predicted peptides of interest masses and fragmented ion masses were calculated with:

http://db.systemsbiology.net:8080/proteomicsToolkit/FragIonServlet.html

Statistical analyses

No statistical methods were used in predetermining sample size and experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. For experimental analyses at least three independent replicates with different original biological material are reported for each experiment and each experiment (excluding immunolocalizations) was repeated at least twice. All graphs were produced using Graphpad software (Version 9), which was also used to calculate standard deviation, t-test and one-way ANOVAs.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.