Introduction

Plants are sessile organisms that have developed a wide array of response pathways to meet the challenges imposed by the environment. Characterization of these response pathways holds a key to genetic resources that can be utilized in molecular breeding. Among all signalling pathways, the ubiquitin proteasome system (UPS) mediated degradation of selected substrates has emerged as a powerful regulatory mechanism for switching between growth and stress responses (Sadanandom et al. 2012). The UPS system is based on tagging substrate protein with a chain of ubiquitin proteins to act as a signal for 26S proteasome mediated degradation. For the attachment of ubiquitin, a cascade of three enzymatic reactions are required, by ubiquitin activating enzyme (E1), ubiquitin conjugating enzyme (E2), and ubiquitin ligase enzyme (E3). Genetic analyses in plants have proposed that the UPS plays a vital role in numerous of signalling pathways; including floral homeostasis (Ni et al. 2004), pathogen defence (Gou et al. 2017), stress responses (Tian et al. 2015; Adler et al. 2017), and phytohormone signalling (Gray et al. 1999; Tan et al. 2007). Apart from the hormone receptors, the majority of the UPS signalling pathway networks still remain to be elucidated.

The E3 ligases are the most abundant UPS components; there are more than a thousand E3 ligase genes in Arabidopsis genome (Stone et al. 2005; Jiménez-López et al. 2018). The E3 ligases found in plants belong to one of seven subtypes that can be divided in two: genes with either a Homology to E6-AP C-Terminus (HECT) or Really Interesting New Gene (RING)/U-box domain (Sadanandom et al. 2012; Jiménez-López et al. 2018). The RING-type E3 proteins are the most abundant among the single subunit E3 ligases and can be divided in seven RING finger domain classes distributed in 30 groups (Stone et al. 2005; Jiménez-López et al. 2018). However, only few E3 ligases have been functionally analysed and their substrates identified.

Spatio-temporal gene expression patterns are informative in assigning proteins to their signalling and regulatory pathways. Recently, we have curated the Arabidopsis RING-type ubiquitin E3 ligase protein sequence number to 509 and associated 60 out of them to floral organ or flowering time regulation based on their expression patterns in the Genevestigator database (Hruz et al. 2008; Pavicic et al. 2017). Through these associations, we identified four novel flowering time or flower development regulators. In the current study, we analysed gene expression patterns of a collection of RING-type ubiquitin E3 ligase genes. Several of the genes were responding to various perturbations and signalling pathways, such as abiotic and biotic stress factors and phytohormones. These results support the hypothesis that ubiquitin E3 ligases play multiple roles in and integrate different signalling pathways (Sadanandom et al. 2012; Stone 2014). To further investigate the role of these genes in hormone pathways, we analysed their expression patterns in abscisic acid (ABA) and gibberellic acid (GA) during germination. To confirm the functional importance of the hormonal responses at gene expression level, we developed a hormone response assay using (ABA) as an early regulator of germination and seedling emergence (Vishal and Kumar 2018).

We utilized the National Plant Phenotyping Infrastructure (NaPPI, https://www.helsinki.fi/en/infrastructures/national-plant-phenotyping) for developing an experimental design and workflow for image-based high throughput in vitro screen. Various platforms and sensors have been used previously for germination analysis, including RGB, VNIR and SWIR (Joosen et al. 2010; Whan et al. 2014; Dumont et al. 2015; Jahnke et al. 2016; Ugena et al. 2018). Here, we introduce a novel approach that utilizes a conveyor belt-based system to deliver thousands of seeds under an imaging station to be scored in one experimental round. Due to the small size of Arabidopsis seeds a chlorophyll fluorescence-based imaging system was used to detect the emerging cotyledons (principle growth stage 0.7 of seed germination according to Boyes et al. 2001) as indication of successful seed germination on ABA containing media. Through screening the selected mutant collection, we identified several RING-type ubiquitin E3 ligases mutant lines with altered ABA responses compared to wild type Col-0. One mutant showed resistant response to ABA, while other four were susceptible to ABA, suggesting that the E3 ligases can act as both positive and negative regulators of ABA responses during germination and seedling emergence. These observations support the emerging versatile roles for RING-type E3 ligases in regulation of ABA response components, further expanding the current signalling models of ABA responses (Yu et al. 2016). The developed system can be easily extended to screening responses to any hormones and natural or chemical compounds or treatments.

Materials and methods

Bioinformatic analysis

Genevestigator gene expression database software was used to associate RING-type ubiquitin E3 ligases with ABA signalling (https://genevestigator.com/gv/) by selecting a set of microarray experiments where Arabidopsis seedlings or organs were treated with ABA (AT-00110, AT-00218, AT-00231, AT-00241, AT-00420, AT-00433, AT-00468, AT-00539, AT-00541 and AT-00637). Expression data from these experiments were retrieved and sorted for 36 RING-type ubiquitin E3 ligases, and 15 of these with more than fourfold of change (LOG2 > 2) where selected for further analysis. For the fifteen ABA responding RING-type ubiquitin E3 ligase genes, gene expression data of a time course analysis of Arabidopsis seed treatment with gibberellic acid (GA), was extracted from GSE5701 dataset (Goda et al. 2008) in the Gene Expression Omnibus (GEO) database.

Plant material and hormone treatments

In total, 18 Arabidopsis Col-0 mutant lines representing the 15 ABA responding genes were obtained from the NASC stock centre. Genotyping of the mutant lines (Table 1) was performed as described in Pavicic et al. (2017). For germination assays, seed stocks were harvested from three individual plants per line grown together in a growth chamber set at 16 h light/8 h darkness, 22 °C, and 60% relative air humidity. These three seed stocks per line were considered as biological replicates. In vitro plants were grown on ½ MS (Duchefa Biochemie, The Netherlands, www.duchefa-biochemie.com), 500 mg/L MES buffer, 10 g/L sugar, 0.8 g/L agar, pH 5.7 adjusted with KOH. For the ABA screening assay development, media was supplemented with 0 nM, 250 nM, 500 nM or 1000 nM of ABA (Sigma-Aldrich, Germany, www.sigmaaldrich.com) in 10 cm petri dishes (Sarstedt, Germany, www.sarstedt.com). Fresh ABA stock was used for every experiment. 30 seeds per plate was used to ensure that the emerging cotyledons did not touch or overlap upon germination on the taken images. Seeds were stratified for three nights at + 4 °C in dark, and transferred to the growth chamber (PhytoScope, PSI, Czech Rep., www.psi.cz) to 16 h light/8 h darkness, even 22 °C, and 60% relative air humidity. The light condition was set and controlled at 130 µmol m−2 s−1 photon flux of photosynthetically active radiation (PAR), using a LI-190R Quantum Sensor coupled to LI-250A light meter (LI-COR, Germany, www.licor.com).

Table 1 List of genes and mutant lines analysed in this study
Full size table

For a progressive drought analysis seeds were sown in high density (~ 60 seeds per 8 cm × 8 cm pots) directly on soil with 50% 1–2 mm vermiculite (Vermipu Oy, Finland) and 50% peat (Kekkilä kylvöseos W HS R8017, Kekkilä Oy, Finland), and watered until saturation. The pots were placed on trays with plastic domes to keep high humidity and cold stratified at 4 °C in dark for three nights. After stratification, trays were transferred to growth chamber with 12 h light/12 h darkness (PhytoScope, PSI, Czech Rep., www.psi.cz). Light intensity, temperature and relative humidity were the same as stated above. Four days after stratification (DAS), tray domes were removed. Seven days after stratification, seedlings were transferred to individual pots watered until saturation and separated into two groups: well-watered and progressive drought. Well-watered group was set to keep watering to 55% water content based on pot weight, while progressive drought was left to dry until drought symptoms became clear. Imaging trays of 20 plants each (7 cm distance between plants) were lifted few centimetres from the shelves to ensure an even pot drying by allowing air circulation through tray bottom holes. Soil was covered with blue sand (Komodo CaCo Royal Blue Sand, www.Komodoproducts.com) to improve automated plant recognition. Quenching analysis of chlorophyll fluorescence and RGB images were taken to assess drought progression (Abid et al. 2017; Tschiersch et al. 2017). Imaging, image processing and data analysis of these plants were performed as described in Pavicic et al. (2017).

RT-qPCR gene expression

Two weeks old in vitro grown Col-0 wild type seedlings were collected and incubated in 0 µM or 100 µM ABA solution in ½ MS with gentle agitation for 0, 1.5, 3 and 6 h. Three biological replicates were prepared for each ABA concentration and time point. RNA was extracted using InviTrap® Spin Plant RNA Kit (STRATEC Molecular, Germany, www.invitek-molecular.com), cDNA was prepared with SuperScript® IV Reverse Transcriptase (Thermo Fisher Scientific, Inc. www.thermofisher.com), and qPCR was performed using Roche Lightcycler® 480 Instrument II with LightCycler® 480 SYBR Green I Master (Roche Diagnostics, Switzerland. www.roche.com). The fold up values were calculated using the 2 −ΔΔCT method according to Schmittgen and Livak (2008), using one of the biological replicates at 0 h as baseline. Reference genes in this study were two stable genes across many conditions according to Czechowski et al. (2005): TIP41 LIKE (AT4G34270) and AT4G33380. The primers for each gene and the thermal protocols are presented in Supplemental Table S1.

High throughput in vitro germination assays

The compact phenotyping facility NaPPI at the University of Helsinki Viikki campus greenhouses (https://www.helsinki.fi/en/infrastructures/national-plant-phenotyping/nappi-facilities) was used to design the high throughput in vitro germination assays. In one experimental round, 30 seeds were sown on a petri dish, one per treatment and biological replication. In the ABA germination assays, in total 5400 seeds representing the 18 mutant lines and Col-0 wild type, were sown and the plates were fully randomized on the 27 trays, each holding 9 dishes (Fig. 1a). The germinating seeds were imaged daily by an overhead pulse amplitude modulated (PAM) chlorophyll fluorescence camera for 7 days (FluorCam, PSI, Czech Rep., www.psi.cz, Fig. 1a).

Fig. 1
figure 1

Work flow of the high throughput in vitro chlorophyll fluorescence imaging system. a Day 1: surface sterilization of the seeds—plating on petri dishes with solidified MS or MS supplemented with ABA—stratification of seeds in cold room—in darkness for three nights. Days 4–11: Three petri dishes with 30 seeds per genotype distributed on the PlantScreen trays—daily imaging for chlorophyll fluorescence (ChlF). Days 12–15: collection of .fimg images—image processing for germinated seedling count—time course data analysis to determine % of germination per genotype. b Image processing steps in FIJI program, from left to right, first: fluorescence max (Fm) .fimg image of all germinated Arabidopsis seedlings on a petri dish, second: pixel thresholding based on maximum fluorescence values, third: binary image generated by assigning 0 value to background pixels and 255 to seedling pixels, fourth: automatic seedling count

Full size image

Image processing

Detection of germination was done based on chlorophyll fluorescence emitted from the emerging cotyledons from the germinating seed. The imaging capture was done using a Fv/Fm imaging protocol of a FluorCam system (PSI, Czech Rep., www.psi.cz) that generates parameter images of minimum fluorescence Fo and maximum fluorescence Fm (Murchie et al. 2013; Tschiersch et al. 2017). This technique is based on dark-adaptation of plants (15 min) to open the photosystem II (PSII) centres. Then, plants are irradiated with a soft measuring light pulse that is not able to trigger photosynthesis, but it is enough to generate a minimum of chlorophyll fluorescence. At this moment an image is taken and saved as F0 (minimal fluorescence). A saturating light pulse generates a maximum peak of chlorophyll fluorescence. Here, a second image is taken and stored as Fm (maximum fluorescence; Murchie et al. 2013; Tschiersch et al. 2017). Fm parameter is very effective to detect small cotyledons due to their chlorophyll content. The background fluorescence is almost non-existent and easily removed, making these images suitable for automated object detection and counting (Fig. 1b). To extract pixel information, we processed the Fm images with “Fiji is just ImageJ” (FIJI) software (Schindelin et al. 2012; Rueden et al. 2017). The images were imported to FIJI software using “import RAW” command and saved as.tiff images for post processing. The pixel cut off to select only the green pixelated seedlings from the Fm pictures were determined each time the camera settings were modified to avoid over or under representation. To discard any noise from the pictures, all pixels with values below 180 were filtered out by the function “threshold” (Fig. 1b). Binary images containing two values for each pixel, 0 for the background and 255 for the object-of-interest, were generated (Fig. 1b). On binary images, a rectangle was drawn around each plate in Fm images and seedlings were counted using FIJI function “particle analysis” (Fig. 1b). The process from thresholding to particle analysis was automated using FIJI macro record function and process folder template from FIJI script editor. Scripting the functions allowed processing of dozens of pictures in couple of minutes, task that would take hours or days when done manually. The output of the image analysis was a CSV file with a column for the image name and another column with the seedling count. All these processes were summarized in three scripts and they are available at GitHub repository (https://github.com/mipavici/PCTOC-Pavicic-2019).

The image name contained information about the experiment number, measuring round, experiment name, hormone concentration and biological replicate number. The image files were imported to R (https://www.r-project.org/), where the image name was separated in multiple columns for better access to the information and a new column was added giving position names to each measurement from A1 to C3. This spreadsheet was bound to another containing mutant lines names by the function “inner_join of dplyr” package using position and tray names as binding references. This process created the final data file that was used for further analyses. Once the process pipeline was assembled, it was possible to go from 500 raw images per experimental round to exploratory plots in few minutes.

Manual validation of the germination screen

To verify the reliability of the automated cotyledon emergence screening, a manual germination count was performed by sowing seeds on 24 petri dishes containing ½ MS media without ABA and 24 plates supplemented with 250 nM ABA. The cotyledon emergence was counted manually for 7 days. Parallel, Fm pictures of all the petri dishes were recorded daily. Fm images were used for automatic cotyledon emergence counting as it was described in the previous section. Altogether, germination of a total of 336 individual seeds were counted for manual and automated assays and the percentage of germination was calculated. To establish the relationship between manual and automated counting a plot of manual versus automated percentage of cotyledon emergence was generated and a linear regression model was fitted.

Statistical analysis of the data

Rstudio (Version 1.1.383) Software was used to process and statistically analyse the percentage of germination per treatments, plates and trays (https://www.r-project.org/). Daily differences of percentage of germination and maximum germination differences were tested using ANOVA test with aov function of R base package. The experiment was repeated two times and the repetition was included in the model as a block. Statistical differences were calculated using Dunnett post hoc test with the function ghlt from multcomp package (Hothorn et al. 2008). Probability values were corrected taking into consideration the number of comparisons.

Results

Data mining for ABA responsive RING-type ubiquitin E3 ligases

RING-type Ubiquitin E3 ligases are a large protein family with diverse additional domains determining the upstream regulation and downstream targets (Stone et al. 2005; Jiménez-López et al. 2018). Through the complex structural characteristics RING-type ubiquitin E3 ligases are involved in many developmental or signalling processes, including ABA (Sadanandom et al. 2012; Yu et al. 2016). Here, we searched for ABA signalling components by submitting 36 RING-type ubiquitin E3 ligase genes to ABA related microarrays in Genevestigator (Hruz et al. 2008). As a control we included ABI3 INTERACTING PROTEIN 2 (AIP2), a RING-type ubiquitin E3 ligase gene negatively regulating ABA signalling during germination (Zhang et al. 2005). Of 36 RINGs, 15 showed responses to ABA treatments (Fig. 2a). ABA responses were highly variable depending on the gene, plant organ or duration of the treatment. Some were highly suppressed or selectively responsive such as ARABIDOPSIS TOXICOS IN LEVADURAS 8 (ATL8), XB3 ORTHOLOG 4 IN ARABIDOPSIS (XBAT34) and ARABIDOPSIS TOXICOS IN LEVADURAS 42 (ATL42), while ABA INSENSITIVE RING PROTEIN 2 (AIRP2) was up-regulated in most treatments. Most of these RING-type ubiquitin E3 ligase genes showed up-regulation after 3-h ABA treatment in seedlings, adult plants or guard cells (Fig. 2a). The gene expression data for the same 15 genes was analysed for GA responses during seed germination but no differential expression patterns compared to Col-0 were observed (Fig. S3).

Fig. 2
figure 2

Bioinformatic analysis of RING-type ubiquitin E3 ligase gene expression in ABA treatments. a Heat map of gene expression patterns of 15 RING-type ubiquitin E3 ligase genes in ABA treatments in a set of microarray experiments from Genevestigator database. Expression data with more than fourfold of change (LOG2 > 2) where retrieved and sorted. Yellow and blue colours representing up- and down-regulation of fold-change between 2 and − 2, respectively. b Relative expression of 9 ubiquitin E3 ligase genes in 2-week old in vitro Col-0 seedlings treated with 0 and 100 µM ABA for 0, 1.5, 3 or 6 h The fold-up was calculated to 0 h (n = 3, error bars = SE)

Full size image

Validation of RING-type ubiquitin E3 ligases by RT-qPCR

To validate the expression data extracted from Genevestigator, 9 of the 15 RING-type ubiquitin E3 ligase genes were selected and analysed by RT-qPCR for their expression profiles (Fig. 2b). As seen in the Genevestigator data, the gene expression of AT2G77770, ATL42, AIP2, AIRP2, RING-H2 FINGER A1B (RHA1B), WAV3 HOMOLOG 1 (WAVH1) and RING AND DOMAIN OF UNKNOWN FUNCTION 1117 1 (RDUF1) was up-regulated, while ARABIDOPSIS TOXICOS EN LEVADURAS 31 (ATL31) and XBAT34 were down regulated by ABA (Fig. 2a, b). AIRP2, RDUF and WAVH1 showed an early ABA response, being upregulated at 1.5 h of treatment. In turn, AT1G77770, ATL42, AIP2, WAWH1 and RHA1B expression peaked moderately at 3 h of ABA treatment. Conversely, XBAT34 was downregulated after 3 h of ABA treatment. Interestingly, 0 µM ABA treatment induced a strong downregulation of AIRP2. The highest differential expression in this analysis was observed in AIRP2, WAVH1 and XBAT34. These results confirmed these RING-type ubiquitin E3 ligase genes as ABA responsive genes.

Progressive drought responses of aip2-1 and wavh1-3

To test if ABA responses were time point specific, aip2-1, as ABA responding control with no drought response and wavh1-3 with similar, but stronger ABA response gene expression pattern to aip2-1 (Fig. 2b) were subjected to progressive drought analysis. We measured the projected rosette area (PRA), maximum quantum yield of photosystem II (QY max), effective quantum yield of photosystem II (ϕPSII), non-photochemical quenching (NPQ), ratio of fluorescence decay (Rfd) and rosette temperature (Supplemental Fig. S2). No differences were observed between mutant lines and Col-0, indicating that the role of these two RING-type ubiquitin E3 ligases is linked to germination and not drought tolerance.

High throughput platform for cotyledon emergency using chlorophyll fluorescence imaging

To confirm the observed ABA responses at gene expression level, we analysed germination of the 18 T-DNA insertion lines corresponding to 15 RING-type ubiquitin E3 ligase genes at different concentrations of ABA. In this assay development, we used cotyledon emergence as a proxy for seed germination. For the first assay we tested four ABA concentrations: 0, 250, 500 and 1000 nM. The highest ABA concentration inhibited germination, while in 500 nM ABA concentration only few seedlings germinated after day 6 (data not shown). Therefore, 500 and 1000 nM ABA were omitted from the analysis and later assays. Thirty seeds per seedstock were sown in round plates in a 5 times 6 array and randomized on trays with nine plates (Fig. 1a, left and middle panels).

To automate the detection of germination, we used Fv/Fm chlorophyll fluorescence imaging. Using this technique, the plates were imaged daily for 7 days after stratification. To automate the data extraction from Fm images, all post-processing steps were registered using FIJI macro recorder and then applied to a folder containing all images generated during the experiment. The first step in the macro was to import raw Fm fimg images generated by PSI FluorCam system and convert them into TIFF format. Afterwards, pixels belonging only to seedlings were selected and background pixels were filtered out based on their Fm values (Fig. 1b). After pixel thresholding, the images were converted into binary images and a selection rectangle was drawn around each plate and FIJI plugin Analyse Particles was run. This plugin recognizes and counts the pixel areas of seedlings within the drawn rectangle. The counts were saved in comma separated value (CSV) files with the image name that served as an identification code for posterior linking of the seedlings with their genotype and measuring time. In our assays, this process lasted about 15 s, allowing rapid and efficient analysis of images.

Manual validation of the automated system

To evaluate the performance of the automated system, the percentage of germination was measured automatically as described above, and manually using a stereoscopic microscope. In total, 24 plates 0 nM and 24 plates with 250 nM ABA were analysed for 7 days, giving 336 germination measurements. In the manual estimation, seeds with cotyledons completely separated from the seed coat were considered as germinated (Boyes et al. 2001). In the automated percentage of germination measurement was estimated using the same parameters described above (seedlings with pixel value in the Fm images above 180). In addition, we tested the effect of different pixel thresholding values on the automated image processing. The pixel threshold of 110 generated binary images which were biased towards manual counting, and a threshold of 250 biased the analysis towards automatic counting (data not shown). Thus, a pixel threshold of 180 generated the most similar automated percentage of germination to manual estimation (Fig. 3b). A linear model was fitted to a plot of manual counting versus automatic counting showing R2 value of 0.9643 and the slope different than zero with high significance (p value < 0.0001). This analysis indicates that the automated percentage of germination measurement was highly representative of the manual estimation (Fig. 3c).

Fig. 3
figure 3

Percentage of germination on different ABA concentrations. a Cumulative percentage of germination for six of the 18 knockout lines in media supplemented with 0 or 250 µM ABA. Results represent the mean of six individual measurements of two experimental repetition. Faded point in the background represent individual measurements with a soft jittering to prevent point overlapping. b Maximum percentage of germination in medium supplemented with 250 µM ABA. c Correlation of automated and manual percentage of germination estimation. Asterisks show mutant lines (a and b) and days b where differences statistically significant (p-value < 0.05). Differences were calculated using ANOVA followed by Dunnett’s post hoc test, adjusting p-values to the number of comparisons. Error bars: standard error of the mean

Full size image

Hormone germination assays of RING E3 ligases

ABA treatment delayed germination, and therefore cotyledon emergence, in a concentration dependent manner (Fig. 3a). At day 1 after stratification no seedlings were detected in any of the treatments. On 0 nM ABA plates the first seedlings started to show green cotyledons after 2 days in all lines (Fig. 3a). For Col-0, applying 250 nM of ABA delayed the first cotyledon emergence to day 3. ABA concentration of 500 nM was enough to suppress the cotyledon emergence in all lines, only few seedlings emerged after 6 days post-stratification. Taken together, the best separation between Col-0 and mutant lines was observed on 250 nM ABA. Of the 18 lines, five lines showed sensitive (aip2-1, at5g41400-1, atl8-1, rha1b-1 and wavh1-3) and one line showed insensitive (atl42-2) responses to ABA on germination rate in comparison to Col-0 (Fig. 3a, Supplemental Fig. S1 and Supplemental Table S2). Lines wavh1-3 and atl42-2 showed strong differences on 250 nM ABA, having, respectively, less or more germinated seedlings than Col-0 (Fig. 3a).

For each line the maximum percentage of germination on day 7 was calculated and mutant lines aip2-1, at5g41400-1, atl8-1, rha1b-1 and wavh1-3 showed below 25% of maximum percentage of germination (p-value < 0.05, Fig. 3b and Supplemental Table S3).

Despite the strong resistance to ABA treatment of atl42-2 mutant on days 4–6, it was not different than Col-0 for maximum percentage of germination at day 7 (Fig. 3b and Supplemental Table S3). However, the trend shows still a higher average of percentage of germination for atl42-2.

Discussion

Gene expression and germination screens revealed early ABA response RING-type ubiquitin E3 ligases

Plants such as Arabidopsis have evolved hundreds of RING-type ubiquitin E3 ligases to regulate their development, growth and stress response (Stone et al. 2005; Sadanandom et al. 2012; Pavicic et al. 2017; Jiménez-López et al. 2018). In our study, ABA was selected as an early developmental regulator to screen a mutant collection to identify possible new ABA response RING-type ubiquitin E3 ligases. Gene expression patterns of the selection of 36 RING-type ubiquitin E3 ligases were screened on ABA treatment in the Genevestigator transcriptome database. Altogether 15 out of the 36 genes were up or down regulated by the variety of ABA treatments (Fig. 2a). Among the 15 ABA response candidates, several known ABA response ubiquitin E3 ligases could be identified, such as AIP2, AIRP2, PROTEIN WITH THE RING DOMAIN AND TMEMB_185A 1 (PPRT1) and RDUF1 (Zhang et al. 2005; Cho et al. 2011; Kim et al. 2012; Pei et al. 2019).

Our working hypothesis was that mutants of ubiquitin E3 ligases, which were up-regulated upon ABA treatment, would be more resistant to ABA than Col-0. This implies that the ubiquitin E3 ligase substrate would be a repressor of the given ABA response pathway (Fig. 4). If the substrate would be an activator, then the respective ubiquitin E3 ligase mutant would show ABA hypersensitivity. To test the working hypothesis the ABA response RING-type ubiquitin E3 ligases were subjected to germination screen on ABA concentration gradient. As a control line aip2-1 behaved susceptible to ABA treatment in accordance to Zhang et al. (2005), validating the set up (Fig. 3a, b). Five ABA response RING-type ubiquitin E3 ligases were identified as responding differentially compared to Col-0 wild type: at5g41400-1, atl42-2, atl8-1, rha1b-1 and wavh1-3 (Fig. 3a, b). With 250 nM ABA atl42-2 line was able to germinate at 60% rate, while Col-0 was below 50% (Fig. 3a). In contrast, the other four mutant lines showed lower germination levels than Col-0 (Fig. 3). These results suggest that our screen was able to capture both types of responses proposed in our working hypothesis. In the RT-qPCR analysis AIP2, ATL42, RHA1B and WAVH1 showed early ABA responses, being upregulated after 1.5 and peaking at 3 h after treatment. ATL42 belongs to the RING H2-type ARABIDOPSIS TOXICOS EN LEVADURAS (ATL) family and it has not been characterized. However, eFP Browser data suggests that it is a root specific gene induced by heat and cold (Schmid et al. 2005; Kilian et al. 2007; Klepikova et al. 2016). A mutant line of another member of the ATL family, ATL43, showed resistant phenotype to ABA treatments in germination (Serrano et al. 2006). ATL42 and ATL43 share high sequence similarity and same germination phenotype in ABA, suggesting that they could share similar functions (data not shown). Double knockout lines atl42/atl43 would help to elucidate their relationship during seed germination. ATL8, was described previously by Serrano et al. (2006), no homozygous mutant could be identified, and it was suggested that ATL8 could be an essential gene for Arabidopsis development. In this study, homozygous atl8-1 mutant was isolated and RT-qPCR results suggest that it is a full knockout mutant, with ABA susceptible phenotype (Fig. 3a). RHA1B phenotype has not been characterized previously and expression analysis shows that it is highly expressed in dry seeds (Schmid et al. 2005; Klepikova et al. 2016). Double mutants of RING-H2 FINGER A2A (RHA2A) and RHA2B, two closely related genes to RHA1B, are resistant to ABA treatments in cotyledon greening phenotype (Li et al. 2011). The mutant rha1b-1 showed susceptibility to ABA treatment suggesting that it could play opposite role to RHA2A and RHA2B (Fig. 3a, b). The gene AT5G41400 is a RING H2-type gene with a very similar amino acidic sequence to RHA1B (data not shown). Expression data shows that AT5G41400 is induced by salt, drought and heat, suggesting that it could be a stress related gene (Kilian et al. 2007). WAVH1 gene also contains a RING H2-type domain and has been associated with the very first leaf primordia in young seedlings, but its association with germination has not been stablished previously (Sakai et al. 2012). The expression patterns of these genes are variable across Arabidopsis development. However, as all these genes contain a RING H2-type domain, our results suggest that this type of proteins could be hormone and stress related.

Fig. 4
figure 4

Hypotheses for ubiquitin E3 ligase function in ABA response pathways. a In the absence of ABA, the ABA signalling pathway is not activated in Col-0 due to active repression. b In the presence of ABA, the ABA response repressors that are ubiquitin E3 ligase substrates are ubiquitinated and targeted for degradation, resulting in release of ABA response. c Ubiquitin E3 ligases can act on ABA response pathway participating in degradation of activators or enhancers of ABA responses. d In a situation where the E3 ligase substrate is ABA response activator, the ABA supply will inactivate ubiquitin E3 ligase and prevent degradation of the activator, which leads to hypersensitivity to ABA

Full size image

UPS involvement in ABA signalling

Germinating at proper time and in optimal conditions is critical for offspring survival and preventing pre-harvest sprouting. Defects in primary dormancy will result in pre-harvest sprouting which will lead to the loss of grain weight and quality of kernels in cereals such as wheat, rice and maize (Fang and Chu 2008). ABA has pleiotropic effects on plant development and growth, such as inhibition of seed germination and seedling growth. Additionally, ABA is a stress related hormone, and the ABA signalling pathways are overlapping with drought, cold and immune response (Fujita et al. 2006; Cao et al. 2011), all of which are enormous challenge to crops growing in the field (Gray et al. 1999; Pauwels et al. 2010). There is accumulating evidence that the UPS signalling and individual RING-type ubiquitin E3 ligases, such as AIP2, KEEP ON GOING (KEG), AIRP2, RHA2A and RHA2B, are involved in ABA signalling (Zhang et al. 2005; Stone et al. 2006; Lyzenga and Stone 2011; Li et al. 2011; Cho et al. 2011; Yu et al. 2016). From these known ABA related ubiquitin E3 ligases we included AIP2 in the analysis as control of ABA response. On the germination screen atl42-2 mutant seeds were resistant to ABA reaching about 60% germination, while at5g41400-1, atl8-1, rha1b-1 and wavh1-3 were susceptible with germination below 25%, in ABA treatment. Therefore, we speculate that in Col-0 plant, ATL42 would target repressors, and while the other four RING H2-type proteins could target activators of ABA response for proteosomal degradation. Thus, the mechanism of action of ABA regulation in Col-0 would be that without supply of ABA there is no ABA response ubiquitin E3 ligase induction. Therefore, substrates (here, ABA response repressors) remain active and repress ABA responses (Fig. 4a). In the other hand, activators of ABA pathway would be constantly degraded to prevent strong responses of this hormone (Fig. 4c). In the presence of ABA, ATL42 will target repressor substrates for degradation, leading to normal ABA response (Fig. 4b). In this same scenario, ABA pathway activators would be induced together with their respective E3 ubiquitin E3 ligases (such as WAVH1), allowing the right amount of activators to accumulate, and thus lead to a controlled ABA response (Fig. 4d).

The ATL42 knock out mutant where is unable to target repressor substrates for degradation, and therefore they are stabilized, and ABA responses are not activated despite ABA supply. In ubiquitin E3 ligase mutants like wavh1-3, which targets ABA response activators, the opposite outcome is expected. Because the ubiquitin E3 ligase is not induced upon ABA supply, ABA response activators accumulate and remain active to promote strong ABA response. Therefore, such mutant is hypersensitive to ABA like shown for aip2-1 and wavh1-2. In the ABA germination screen, aip2-1 and wavh1-2 were ABA hypersensitive. Indeed, a known AIP2 target protein ABI3 promotes ABA signalling and represses germination of seeds (Zhang et al. 2005). Substrates for WAVH1 E3 ligase are not known. We tested the mutant wavh1-3 and aip2-1 for progressive drought assays with not significant differences from the control plants. These results suggest that for those two genes their ABA responses are specific to germination.

Automated HTP germination assay to facilitate gene function discoveries

Seed quality is an important agronomic trait that ensures uniform germination rate and seedling vigour (Rajjou et al. 2012). Recently, several approaches have been developed to automate or facilitate the seed quality assessment or analysis of the germination process (Dumont et al. 2015; Jahnke et al. 2016; Ligterink and Hilhorst 2017; Zhang et al. 2018). In these applications, facilities with high precision imaging tools provide means for analysis of seed size, volume, colour and internal quality. Eventually automation can be utilized sorting the individual seeds based on the assigned quality control (Jahnke et al. 2016). Automation can also be extended to delivery of the imaging targets to camera (this study), or camera to the imaging targets. In combination with different levels of automation high throughput plant phenotyping facilities would allow assessing the germination processes in time course fashion. Here, to allow fast screening of the tens of mutant lines for germination on several concentrations of ABA, a high throughput imaging-based assay was established. To this end, Arabidopsis seeds germinating on the petri dishes containing different concentrations of ABA hormone were imaged by chlorophyll fluorescence sensor. The cotyledon emergence was used as an indicator of germination to allow detecting the tiny Arabidopsis seedlings by the chlorophyll fluorescence signal (Fig. 1). For larger seeds, such as wheat and legumes, RGB imaging and colour-based detection can be utilized instead of chlorophyll fluorescence (Zhang et al. 2018). The image processing workflow allowed detecting the very first fluorescence signals that the emerging cotyledons emitted, and these signals were used for scoring the germination rate of the mutant lines. Automation of this process allowed screening of hundreds of pictures in few minutes. This data extraction together with R script pipeline allowed to go from raw images to exploratory plots in less than 30 min. Thus, establishing high throughput germination screens in the automated imaging-based phenotyping facilities will facilitate analysis of seed quality traits such as germination rate, dormancy and seed longevity but also bioassays with hormones and compound libraries.

UPS mediated protein degradation has been found to participate in numerous of signalling pathways in plants, including phytohormone mediated plant growth and development (Gray et al. 1999; Pauwels et al. 2010). The automated high throughput germination screen allowed us to discover two novel ubiquitin E3 ligases associated with ABA signalling pathways. The two novel ABA signalling RING-type ubiquitin E3 ligases identified here are currently being functionally characterized to push forward current understanding of ubiquitin E3 ligases as a molecular switch from growth and development to stress responses. Future work will aim at identification of their molecular networks, including the substrates.