A seed resource for screening functionally redundant genes and isolation of new mutants impaired in CO2 and ABA responses

Over 14000 amiRNA-expressing Arabidopsis plants were generated, to be made publicly available, enabling screening of redundant genes. Known and new genes were identified.


Introduction
The presence of large gene families in plants, including Arabidopsis (Arabidopsis Genome Initiative, 2000), leads to functional genetic redundancies or partial functional overlap among closely related genes. Functional overlap and partial or complete redundancy between different family members has been proposed to provide a buffer for loss or gain of function mutation events and mechanistic robustness of cellular networks (Wagner, 2005). This is considered to be a main reason for the lack of observable phenotypes in single-gene deletion mutants and increasing severity of phenotypes in higher order mutants of homologous genes (Ma et al., 2009;Park et al., 2009). Identification and characterization of functionally overlapping genes in genetic screens is limited, as is evident from the relatively low number (591 of all Arabidopsis genes) of genes not associated with a single mutant phenotype (Lloyd and Meinke, 2012). Analysis of genome-wide gene family definitions showed that the Arabidopsis genome includes over 22 000 genes belonging to gene families (Hauser et al., 2013). Strategies and tools have been developed to enable screens of the functionally redundant gene space. Recently, an artificial microRNA (amiRNA)-based computational design approach was introduced (Hauser et al., 2013). AmiRNAs designed to specifically target diverse combinations of gene family members or combinations of subfamily members enable the screening of partial overlapping homologous gene functions at a genome-wide scale. The presented platform also provides an approach for the capture of homologous gene silencing phenotypes, for which higher order loss of function mutants would lead to lethality, as illustrated by a mutant identified here.
Here, we report the generation of over 11 000 T2 amiRNA lines and 3000 additional amiRNA lines by transformation of Arabidopsis Col-0 with a previously published amiRNA library (Hauser et al., 2013) and screening of T2 amiRNA lines for abscisic acid (ABA)-insensitive seed germination phenotypes or plants with low-CO 2 -insensitive high-leaf-temperature phenotypes. Methods are described to identify robust amiRNA mutants and how to avoid pitfalls of this approach. The screen rapidly identified two amiRNAs that target three PYR/RCAR ABA receptor- (Ma et al., 2009;Park et al., 2009) or six SNF1-related kinase-(SnRK2; Mustilli et al., 2002;Yoshida et al., 2002;Fujii and Zhu, 2009) encoding genes known to be involved in ABA-mediated control of seed germination. One candidate line that shows an ABA-insensitive seed germination phenotype contains an amiRNA that targets three genes of unknown function, which are annotated as Avirulenceinduced gene 2 (AIG2A; AT3G28930), Avirulence-induced gene 2-like protein A (AIG2LA; AT5G39720), and Avirulence-induced gene 2-like protein B (AIG2LB; AT5G39730). One amiRNA that causes a low-CO 2 -insensitive high-leaf-temperature phenotype targets two genes encoding proteasomal α2-subunits, annotated as PAB1 (AT1G16470) and PAB2 (AT1G79210), for which double mutation causes lethality. New amiRNA lines that target the gene for proteasomal α7-subunit, annotated as PAG1 (AT2G27027), were constructed resulting in a similar stomatal phenotype. Together these observations indicate a rate-limiting role of the intact proteasome for stomatal opening responses.
Large scale transformation with the amiRNA library pools (Hauser et al., 2013) was performed as described elsewhere (Cutler et al., 2000) with the following modifications. One microgram of DNA of each amiRNA sublibrary (Hauser et al., 2013) was electroporated into a total of 500 µl electrocompetent A. tumefaciens cells. The 20 bp and 21 bp amiRNA sublibrary variants for each pool were individually electroporated and combined at this stage. After 2 h at 30 °C in non-selective Luria-Bertani-Miller medium (LB, Teknova), the cells were spread on 20 LB plates (1.5% agar; 150 mm diameter) containing all the appropriate antibiotics (rifampicin, gentamycin, tetracycline, spectinomycin) and grown for 3 d at 30 °C. The bacteria were scraped from the plates, resuspended in 5 ml LB and concentrated by centrifugation for 20 min at 5855 g. Plants were transformed by spraying the flowers with this suspension of the bacteria in infiltration medium (adjusted to an optical density at 600 nm of 0.5) twice with 1 week between the treatments. T1 plants were selected on plates supplemented with 75 µM phosphinotricin or directly on soil by spraying diluted herbicide (1000× dilution, Finale ® ; Bayer, Research Triangle Park, NC, USA) four times with 2-7 d between the treatments. Herbicide-resistant plants were transferred to soil and grown to full maturity and T2 seeds collected from individual plants. When appropriate, media for growth of bacteria or plant selection contained the following concentrations of antibiotics (mg ml −1 ): carbenicillin 100, gentamycin 25, kanamycin 30, rifampicin 50, spectinomycin 100, tetracycline 10, and phosphinotricin 15.
Screen for abscisic acid-insensitive cotyledon emergence phenotype T2 plants were screened individually for insensitivity of seed germination to ABA in 96-well plates (100 µl 0.1% agarose supplemented with 2 µM (±)-ABA, Sigma-Aldrich). Approximately 10-20 seeds were used from each T2 plant. For the pooled screening, approximately 10-50 seeds of 90 individual T2 plants were mixed, surface sterilized and sprinkled onto agar plates (3 μM (±)-ABA; Sigma-Aldrich). As control for ABA insensitivity, abi4-1 (ABRC, CS8104) or abi5-1 (ABRC, CS8105) was used and Col-0 was used as a wildtype control. A putative ABA-insensitive phenotype was scored in a binary manner for similarity to the abi4-1 or abi5-1 phenotype and difference from wild-type after 5-8 d using green cotyledons as indicator (Kuhn et al., 2006). For lines that showed a putative ABA insensitivity, the seed germination assay was repeated by propagating individual T2 seedlings to the next generation (T3) and using seeds of the T3 generation for ABA sensitivity assays. This time, seeds were placed on plates with and without ABA (2 µM (±)-ABA; Sigma-Aldrich) and images were taken daily for 7 d and emergence of radicles and cotyledons was counted manually using Fiji (Schindelin et al., 2012). For candidates of the individual screen the T2 seeds were used for the repetition of the germination assay.
For candidates of the pooled screen ABA-insensitive seedlings were transferred to plates containing 75 µM phosphinotricin, and after 7-10 d resistant seedlings were transferred to soil, grown up to full maturity, and the T3 seeds used for the validation of the ABA-insensitive germination phenotype.

Screen for CO 2 -insensitive leaf temperature phenotype
Seeds of T2 plants were germinated in 96-pot-flats (254 mm×508 mm; East Jordan Plastics, East Jordan, MI, USA) on soil with each pot containing seeds from one plant. After 7 d, seedlings were sprayed with a 1000× dilution of Finale ® (Bayer), and 2-3 d later pale seedlings were removed and only one healthy dark green seedling was left per pot. After 19 d under standard growth conditions, the plants were exposed to 150 ppm CO 2 for 2 h in a Percival growth chamber. A first set of thermal images was taken with a FLIR A320 thermal imaging camera (FLIR, Wilsonville, OR, USA). Subsequently the plants were exposed to ≥ 800 ppm CO 2 and after 2 h a second set of thermal images was taken. Control plants included in the experiments were ht1-2 (Hashimoto et al., 2006), ost1-3 and wild-type Col-0. Thermal images were converted into Flexible Image Transport System format (fits) using the ExaminIR software (FLIR). For the screen using the 96-pot-flat format, the temperature of plant leaves and the surrounding soil were measured using Fiji (Schindelin et al., 2012). The soil temperature served as a location-specific reference to compensate for temperature variation depending on the position in the 96-pot flat. Either the temperature difference between plant leaves and surrounding soil or the average temperature of plant leaves was used as a quantitative measure. Plants with more than 1 °C difference from soil were considered as primary candidates and subject to further testing. The high-temperature leaf phenotype of ht1-2 was used as a reference for CO 2 insensitivity. To test the reproducibility of the CO 2dependent leaf temperature phenotype of putative candidates, T2 plants were grown in triplicate and assayed again alongside with ht1-2 and wildtype control plants.

Identification of amiRNA sequences and testing reproducibility
Genomic DNA from candidates with a robust and reproducible phenotype was isolated as described elsewhere (Edwards et al., 1991) and the sequence of the amiRNA present was determined by sequencing of the PCR product (primers pha2804f and pha3479r; see Supplementary  Table S1 at JXB online). Using the Target Search function available on the WMD3 website (Ossowski et al., 2008), putative amiRNA target genes were predicted. For independent confirmation of the phenotype, independent lines were generated by cloning the identified amiRNA into pFH0032 (Supplementary Table S2; (Hauser et al., 2013) and transforming it into Arabidopsis Col-0. Confirmed phenotypes were further analysed by using single knock-out mutants, higher order mutants generated by crossing and/or generating amiRNAs that target subsets of the initial target genes (see Supplementary Table S3).

Gas exchange analyses
Stomatal conductance of H 2 O (g s ) was measured in leaves of 5-to 6-week-old plants using portable gas exchange systems (LI-6400 and LI-6800, LI-COR, Lincoln, NE, USA), starting 2 h after growth chamber light onset. For intact single leaf ABA treatments, a LED light source set at 150 μmol m −2 s −1 (10% blue) and a chamber temperature of 21 °C was used. Leaves were equilibrated for 1 h at a relative humidity of 70-72%, airflow of 200 rpm and CO 2 concentration of 400 ppm. After 1 h, steadystate stomatal conductance was recorded 10 min before the addition of ABA to the petioles submerged in water at the indicated concentration. For light-response measurements, plants were kept in the dark for 18 h prior to experiments. Stomatal conductance of a single intact leaf in the dark was recorded for 10 min, followed by red light treatment of 600 μmol m −2 s −1 . After 20 min of red light treatment, additional blue light was applied at 10 μmol m −2 s −1 . The incoming air humidity was kept at 62-65% and air flow at 200 rpm. For stomatal conductance measurements of single intact leaf CO 2 responses, incoming relative air humidity was kept at 62-65% and the imposed changes in CO 2 concentration were applied as indicated. Leaves were attached to intact plants and were equilibrated for 1 h before the measurements. The data presented represent n≥3 leaves with each leaf from independent plants per genotype per treatment.

qRT-PCR analysis
Total RNA (500 ng) was reverse transcribed using the first-strand cDNA synthesis kit (GE Healthcare). qRT-PCR analyses were performed using 3-fold-diluted cDNA (Maxima SYBR Green Rox/qPCR Master Mix; Thermo Fisher Scientific) on a CFX Connect PCR cycler (BioRad). The housekeeping PDF2 gene was used as an internal control (Czechowski et al 2005). The threshold cycle (C T ) was determined by the instrument (CFX Manager Software, BioRad), and the ΔΔC T method was used to calculate the fold change in each gene (Livak and Schmittgen, 2001). For RAB18 gene expression measurements, total RNA was extracted from 2-week-old seedlings that were treated with ABA for 9 h and final concentration of 20 µM.

Generation of amiRNA library plants
We have previously described the generation of an amiRNA library consisting of 10 sublibraries that represent 22 000 individual amiRNA designs (Hauser et al., 2013). Deep sequencing of these 10 sublibraries showed that ≥95% of the designed amiRNAs were present in these sublibraries (Hauser et al., 2013). The amiRNA library was transformed first into Agrobacterium tumefaciens and then into Arabidopsis Col-0. Over a period of over 4 years, the amiRNA library consisting of 10 sublibraries was transformed and T1 seeds harvested. Using plate-or soil-based selection methods, herbicide-resistant T1 plants were grown and T2 seeds from over 14 000 individual plants were harvested ( Table 1). The transformation rate varied over a range from 0.08% to 0.76% with an average of 0.25%. During the course of this research, approximately 3000 additional T2 lines (Zhang et al., 2018) were generated expressing amiRNAs that target homologous transporter-encoding gene family members. These 3000 lines will also be made available to the ABRC, such that in the end over 14 000 total T2 lines will be submitted for use by the community.

Screen for ABA-insensitive seed germination phenotype
In total over 2500 T2 amiRNA lines were screened individually and over 5000 T2 amiRNA lines were screened in pools for ABA-insensitive germination phenotypes ( Fig. 1). In the primary screen using individual plants in a 96-well plate format, 59 putative candidates were identified. In the primary screen using pools of 90 plants with 25-80 seeds per line, approximately 340 putative candidates representing an unknown number of lines were identified (Fig. 1).
These candidates were subjected to further analysis in a secondary screen (Fig. 2). The cotyledon emergence phenotype of 24 T3 seedlings from a total of 76 retested plants showed a more reduced ABA sensitivity that was clearly different from wildtype and less severe than the abi4-1 and abi5-1 controls ( Fig. 2A). From the 59 putative candidates identified using the individual screening approach, the amiRNA line p8l1257 showed a reproducible partial insensitivity to ABA in the T3 generation (Fig. 3). Only the amiRNA in candidates with the most robust phenotypes was determined by sequencing. Two of the amiRNAtargeted gene sets identified in 24 seedlings with reproducible phenotypes are known core components of the ABA signal transduction cascade ( Fig. 2; Table 2). These include amiRNA lines that target the three ABA receptors PYR1 (RCAR11), PYL4 (RCAR10), and PYL6 (RCAR9) (Fig. 2B, C; Table 2). Furthermore, amiRNA-expressing plants that target six members of the SnRK2 protein kinase family Yoshida et al., 2002;Fujii and Zhu, 2009) were isolated in this screen, including the three SnRK2 protein kinases, SnRK2.2, SnRK2.3, and SnRK2.6 (OST1), that are known to be required for abscisic acid signaling (Fig. 2B, C; Table 2; Mustilli et al., 2002;Yoshida et al., 2002;Fujii and Zhu, 2009).
Notably, Fig. 2D shows a strong variation in the cotyledon emergence phenotype among plants expressing the same amiRNA that targets six SnRK2 kinase transcripts. This variation might be responsible for the high number of variable candidates that did not show a robust phenotype following the primary screen. Additional amiRNA lines were isolated as putative mutants and the amiRNA sequence was determined (Supplementary Table S4). Although some of the predicted targets might be expected to affect abscisic acid responses, rescreening of these putative mutants did not show consistently robust reproducible phenotypes. Thus, amiRNAs appear to produce phenotypes that may be variable even within the same line. These findings   Hauser et al. (2013), the number of amiRNAs designed for each library, and the number of individual T2 amiRNA transformants that were generated here led us to develop a protocol in which: (i) only putative mutants that showed a consistent phenotype when screening seeds from the next generation of plants were selected, and (ii) Only lines that showed similar phenotypes upon re-transformation with new amiRNAs that are predicted to target the same genes were selected. Furthermore, based on the variation observed here in the secondary screen it is advisable to investigate over 10 independent transformed lines (Schwab et al., 2006;Hauser et al., 2013) in the future to determine which amiR-NAs produce phenotypes that can be carried forward. The isolation of amiRNA lines targeting functionally overlapping PYR/RCAR ABA receptor and SnRK2 protein kinase genes, which could not be isolated in traditional forward genetic mutant screens, provides a proof of principle that functionally redundant genes can be isolated in forward genetic screening using this new amiRNA resource. The inclusion of control lines and the validation steps described above should enable screening for diverse phenotypes using the lines generated here that are being provided to ABRC.
AmiRNA lines targeting three avirulence-induced genes show partial insensitivity to ABA inhibition of seed germination but not to ABA-induced stomatal closure The amiRNA in line p8l1257 isolated in the present screen targets a new set of three genes (Fig. 3A, 3B). Previous research  Table 2). Wild-type control (Col-0) and abi4-1 and abi5-1 as reference for insensitivity are shown in the other two rows. (C) Time course of cotyledon emergence in the presence of ABA for wild-type control (wt), abi4-1 as a reference for ABA insensitivity and two representative amiRNA lines that target a set of six SnRK2 protein kinase genes (amiRNA-SnRK2) or a set of three PYR/RCAR ABA receptor genes (amiRNA-PYR/ RCAR). Per genotype 74 ± 46 seeds were phenotyped. (D) Bar plot of variation of cotyledon emergence phenotype (day 6; 2 µM ABA) in the T3 generation of plants isolated in the primary screen that were selected as candidates based on their seed germination phenotype in the T2 generation. Sequencing of the amiRNA confirmed that all 18 of these amiRNA-expressing plants contain an amiRNA that targets a set of six SnRK2 kinase genes (amiRNA-SnRK2; see Table 2). Wild-type control (wt, Col-0) and abi4-1 as reference for ABA insensitivity are shown.
annotated these genes based on their mRNA upregulation in a transcriptomic study after infection with Pseudomonas syringae pv maculicola carrying avrRpt2 (avrRpt2-induced gene, AIG2) (Reuber and Ausubel, 1996). However, these genes have not been previously described to be involved in ABA-mediated control of seed germination or other phenotypic responses in plants.
The line p8l1257 was named amiRNA-AIG here and was further tested by analysing seed germination with additional T2 generation seeds from the original p8l1257 stock (Fig. 3). Germination properties were compared with a control amiRNA line targeting the human myosin 2 (amiRNA-HsMYO), which has no targeted genes in Arabidopsis (Hauser et al., 2013). After 12 d on plates containing 2 µM ABA, the  amiRNA-AIG line showed cotyledon greening in contrast to the control amiRNA-HsMYO line (Fig. 3A). The effect of the amiRNA-AIG on the expression of a known ABA-induced gene, RAB18, was analysed by qRT-PCR (see Supplementary  Fig. S1). The ABA-mediated induction of RAB18 expression was substantially reduced in the amiRNA-AIG line indicating a role of the targeted avirulence-induced genes (AIG2s) in ABA signal transduction.
Since two out of the three genes are tandemly repeated, generation of double mutants using T-DNA insertion knockouts would be limited. Therefore, five new amiRNA lines were generated that target subsets of genes targeted by the original amiRNA-AIG to verify the relevance of the predicted AIG2 target genes. AmiRNA1, 2 and 3 targeted each a single AIG2 ( Fig. 3B; Supplementary Table S3). AmiRNA4 targeted two tandem-repeat AIG2 genes and amiRNA5 targeted all three AIG genes targeted in the original amiRNA-AIG line, but with a different amiRNA sequence ( Fig. 3B; Supplementary  Table S3 for amiRNA sequences). When the T2 seeds expressing these five new amiRNAs were tested in a seed germination assay with 0.5 µM ABA, only the amiRNA4 and amiRNA5expressing lines showed less sensitivity to ABA compared with the control amiRNA-HsMYO line in cotyledon greening (Fig. 3C). The expression of all three putative target genes (AT5G39720, AT5G39730, AT3G28930) was analysed using qRT-PCR in the originally isolated amiRNA-AIG line and in all the amiRNA lines 1-5 (see Supplementary Fig. S2). The amiRNA efficiency of transcriptional inhibition varies between the lines, target transcript(s) and amiRNA sequence. Note that microRNA silencing in plants occurs via two mechanisms, (i) the degradation of transcripts and (ii) inhibition of translation (Brodersen et al., 2008). Thus, quantification of targeted transcripts may not fully show the degree of silencing of target genes. Combined, these data provide evidence that the original amiRNA-AIG phenotype is attributable to silencing of more than one AIG2 gene, suggesting overlapping homologous gene functions.
The original amiRNA-AIG line was also investigated to determine if it affects ABA-induced stomatal closure using an intact leaf gas exchange analysis approach. When ABA was applied to the transpiration stream of intact leaves at a final concentration of 2 µM, both the control amiRNA-HsMYO line and the amiRNA-AIG line showed an ABA-induced decrease in stomatal conductance to H 2 O (g s , Fig. 4A). Normalization of the stomatal conductance data showed no dramatic difference in ABA-induced stomatal closure between amiRNA-HsMYO and amiRNA-AIG (Fig. 4B). Together, the present data show that the isolated amiRNA-AIG line is less sensitive to ABA inhibition of seed germination.
The AIG2 genes are functionally annotated as putative γ-glutamyl cyclotransferases (GGCTs, EC:4.3.2.9) based on their similarity to the human orthologue (HsGGCT; O75223). AIG2LA and AIG2LB share only 16% and 17% identity, respectively, to the human orthologue. GGCTs have been described to cleave γ-glutamyl-amino acid dipeptides to release the free amino acid and 5-oxoproline (Oakley et al., 2008). Further research will be required to determine the mechanism by which AIG2s affect ABA inhibition of seed germination.

Screen for CO 2 -insensitive leaf temperature phenotype
In total, over 2500 T2 amiRNA lines were screened individually for an altered leaf temperature response to a low CO 2 concentration (150 ppm) by infra-red thermal imaging (Fig. 5). Leaf temperature depends on various parameters including radiation absorption, air temperature, and humidity . Low ambient CO 2 concentration leads to stomatal opening in Arabidopsis, causing an increased transpiration rate and thus a decrease in leaf temperature compared with the surrounding air. Mutants impaired in CO 2 -induced stomatal opening appear warmer compared with wild-type plants. In the screen, we used the soil temperature as reference to compensate for the local temperature differences due to various factors including humidity of the soil. Wild-type plants and plants of the HIGH LEAF TEMPERATURE1-2 (ht1-2) mutant (Hashimoto et al., 2006) were included in all trays as controls. Based on visual inspection of the thermal images, plants with relatively higher leaf temperature under low [CO 2 ] compared with the other plants in the same image were selected and the difference between the average leaf temperature and the surrounding soil was determined. The difference between leaf temperature and soil temperature was determined as reference for overall temperature and to compensate for local temperature differences. A set of 106 plants with more than one degree difference between the leaf temperature and the surrounding soil was defined as initial putative candidates for further testing (see Methods for details). For the rescreening of putative mutants, we set a high threshold for temperature differences in the selection of mutants compared with the wild-type strain of 1 °C. The constitutive CO 2 response mutant ht1-2, when exposed to low [CO 2 ], shows a delta temperature above 1 °C between leaf and soil. Rescreening of these candidates in the T2 generation revealed an amiRNA line (p9l22) with a robust and reproducible impaired response to low CO 2 (Fig. 5B).
After exposure to low [CO 2 ], the leaf temperature of the p9l22 line was compared with wild-type (Col-0) and with the constitutive high-CO 2 -response mutant ht1-2 ( Fig. 6A; Hashimoto et al., 2006). The leaves of the p9l22 line had a higher temperature than wild-type leaves and a similar temperature to ht1-2 leaves (Fig. 6A). Stomatal index (SI) and density (SD) were calculated for wild-type, the control amiRNA-HsMYO, and p9l22 lines. No noteworthy differences were found between the genotypes (Supplementary Fig. S3; amiRNA-HsMYO versus p9l22 line, one-way ANOVA, P>0.05 for SI and SD).
To measure [CO 2 ] responses in a time-resolved fashion, we measured stomatal conductance (g s ) using a gas exchange analyser. In the amiRNA-HsMYO control line, the shift from ambient (400 ppm) to low (150 ppm) [CO 2 ] led to a rapid increase in stomatal conductance (Fig. 6B). AmiRNA line p9l22 responded to the same treatment with a lower magnitude of stomatal opening (Fig. 6B). Both lines showed stomatal closure in response to high (800 ppm) [CO 2 ] exposure at similar rates (Fig. 6B). To test whether line p9l22 is defective in response to other stimuli that cause stomatal opening, light-induced g s responses were investigated (Fig. 6C). The control amiRNA-HsMYO and p9l22 lines were kept in the dark for 18 h prior to the experiments and steady-state g s was measured. When red light (at 600 µmol m −2 s −1 ) was applied, the p9l22 line showed a reduced rate of g s increase when compared with the control line. The same was observed when blue light (at 10 µmol m −2 s −1 ) was superimposed on the red light background (Fig. 6C). Thus, the amiRNA line p9l22 causes reduced responses to low CO 2 concentration, red light, and blue light.
The amiRNA in the p9l22 line was sequenced and is predicted to target two homologous proteasomal subunit genes (PAB1, AT1G16470; and PAB2, AT1G79210). PAB1 and PAB2 are the sole genes that encode the 20S proteasome α2 subunit (Baumeister et al., 1998). First, we attempted to isolate a double mutant (pab1 pab2) using T-DNA insertion lines (SALK_099950 and SALK_144987; Alonso et al., 2003). After genotyping over 100 plants in the F2 generation, no homozygous double mutant was recovered. We concluded that the double mutant is very likely lethal.
Alternatively, a new amiRNA sequence targeting solely the PAB1 gene was cloned and transformed into the pab2-1 single mutant (SALK_144987). This new amiRNA line, pab2-1mut pab1amiRNA, was investigated in stomatal conductance analyses of [CO 2 ] responses (Fig. 7). Leaves were first exposed to high (900 ppm) [CO 2 ] for 1 h and steady-state g s was recorded. Shifts to low (150 ppm) [CO 2 ] led to an increase in g s in both the pab2-1mut pab1amiRNA line and the control amiRNA-HsMYO line (Fig. 7A). The normalized stomatal conductance data show that the pab2-1 amiRNA-PAB1 line responds to low [CO 2 ] with a reduced magnitude compared with the control line (Fig. 7B).
Initial experiments were pursued to determine if modifications in the α-ring of the 20S proteasome might be linked to the above phenotypes, or whether this mutation is specific to only α2 subunit mutations of the proteasome. The α-ring of the 20S proteasome is composed of seven α-subunits (Kurepa and Smalle, 2008). The p9l22 amiRNA targets the only two genes that encode the α2 subunit of the proteasome (Fig. 7, inset highlighted in red). To determine whether other α-subunits also affect the response to low [CO 2 ], a second amiRNA line was generated, which targets the PAG1 gene (α7 subunit, inset in Fig. 7 highlighted in green), Fig. 5. Thermal imaging screen for mutants with impaired response to low CO 2 identified the amiRNA line p9l22. (A) Thermal images of amiRNA lines in the primary screen after exposure to low CO 2 (150 ppm). Plants were grown in 96-pot flats under ambient CO 2 and after 26 d exposed to low (150 ppm) CO 2 for 2 h and then thermal images of the entire flat were taken in eight separate images per flat. An image is shown from the primary screen in which the plant p9l22 (white box) was flagged as a candidate with a putative altered response to CO 2 based on the leaf temperature. The average leaf temperature was computationally calculated across all rosette leaves for flagging putative mutants. (B) Differences of leaf temperature (T plant ) and surrounding soil (T soil ) for putative mutants exposed to low [CO 2 ] (150 ppm) for 2 h. Average leaf temperature was computed by image analysis of the most fully expanded rosette leaves. Bars show average ±SD (n=3-5 leaves). WT (Col-0) (orange) and ht1-2 (green) were used as control.
named amiRNA-PAG1. The α7 subunit is encoded by a single gene in Arabidopsis (Kurepa and Smalle, 2008). When a amiRNA-PAG1 line was tested in g s responses to [CO 2 ] shifts, it showed a lower rate of stomatal opening when compared with the control amiRNA-HsMYO line (Fig. 7C, D; one-way ANOVA P>0.05). The expression levels of PAB1, PAB2, and PAG1 were analysed in the p9l22 amiRNA line and also in pab2-1mut pab1amiRNA and amiRNA-PAG1 lines using qRT-PCR (see Supplementary Fig. S4). With the exception of the severely reduced PAB2 expression in the pab2-1mut pab1amiRNA line when compared with control lines, no clear evidence for knock down at the transcriptional level could be detected in the amiRNA lines, which may point to amiRNAmediated inhibition of translation (Brodersen et al., 2008).
The present findings show that the p9l22 amiRNA line is also partially impaired in red light-induced stomatal opening. Red light mediates stomatal opening in part via activation of photosynthesis and the resulting drop in internal concentration of CO 2 (C i ) (Roelfsema et al., 2002;Matrosova et al., 2015).  Inserts show a representation of the 26S proteasome, with the α2 subunits highlighted in red and α7 subunits highlighted in green. The initial slope of g s response for amiRNA-HsMYO and amiRNA-PAG1 was calculated and one-way ANOVA was used to compare the values (P>0.05).
In addition, the p9l22 line is also partially impaired in blue light-induced stomatal opening. This suggests that a general regulator of stomatal opening is impaired in this line. As the proteasome mediates the degradation of proteins and reduced functions of α-ring subunits are predicted to increase protein levels, it is tempting to speculate that the phenotype observed might be correlated with an increased abundance of a negative regulator of stomatal opening. Further research will be required to test this or other hypotheses. In other studies, the 26S α2 subunit, when overexpressed, enhanced thermotolerance and adaptation in rice and Arabidopsis, suggesting that proteasomal subunits can have rate-limiting roles in regulating plant physiological responses (Li et al., 2015).

Summary and future use
A library of over 11 000 plus 3000 additional T2 generation amiRNA lines was created as a new resource to screen the redundant gene space in Arabidopsis. These amiRNAexpressing lines are being provided as individual lines to the Arabidopsis Biological Resource Center (ABRC). Given the observations and findings in the present study, lines will be available for high-throughput screening in pools of 90 lines per pool with approximately 25-50 seeds per individual amiRNA line in each pool. In each pool, the pooled seeds for screening will originate from one of the 10 sublibraries that target gene family members with defined functional annotations (Table 1; Hauser et al., 2013). This approach will increase the probability of identifying interesting putative mutants in future screens despite the biological variability in amiRNA silencing lines found here (Fig. 2D).
The screen for ABA-insensitive seed germination phenotypes identified two amiRNAs targeting PYR/RCAR ABA receptor genes and SnRK2 genes, which are both known groups of redundant key genes and proteins required for ABA signal transduction (Ma et al., 2009;Park et al., 2009). Isolating amiRNA lines in these known components serves as proof of principle for our approach. Moreover, screening this amiRNA population enables the identification of mutants that require co-silencing of homologous gene family members, which are less likely to be found in forward genetic screens of ethyl methanesulfonate or T-DNA mutagenized seed populations.
Overall the presented amiRNA screen shows that amiRNA lines are prone to showing a high rate of variable candidates with weak or non-robust phenotypes. Nevertheless, as shown here new mutants can be isolated. Furthermore, during the course of this research, this amiRNA library resource has been used to isolate long-sought functionally redundant auxin transporter genes (e.g. ABCB6, ABCB20; Zhang et al., 2018). Approaches to circumvent the inherent limitations of forward genetic screening with amiRNAs were developed in the present study. As a first step, it is recommended to rescreen the next generation to identify robust and reproducible phenotypes in individually isolated putative mutant lines. As a second step, the amiRNA sequence of confirmed mutant lines needs to be determined (see Methods). AmiRNA sequences linked to the observed phenotypes are retransformed, and testing over 10 independent lines for the phenotype is recommended. Alternatively, amiRNA on one line without break that target a subset of the initially predicted targets can be used to narrow down the causative genes (e.g. Fig. 3). For cases where only two to three genes are targeted, T-DNA lines or CRISPR/Cas9 mutants can be used to narrow down the genes relevant for the phenotype.
Over 95% of the amiRNAs in this library were designed to target only two to five genes (Hauser et al., 2013), meaning that identification of causal genes is facilitated. Using the above approach, we report on two newly identified mutants: (i) amiRNA lines targeting three genes encoding avirulence induced gene (2-like) proteins show an ABA-insensitive seed germination phenotype; and (ii) amiRNA lines targeting two proteasomal subunits show insensitivity to low-CO 2 -induced stomatal opening. Further analyses of the two targeted genes in this amiRNA line suggest that stronger T-DNA alleles result in lethality. This indicates the usefulness of the generated amiRNA lines for forward genetic isolation of higher order mutants that would be lethal upon knock-out. Our data suggest that the wild-type expression levels of two α-subunits of the 20S proteasome, α2 and α7, are required for fully functional stomatal opening mediated by physiological stimulation. This indicates that the proteasomal subunits are likely controlling an unknown general negative regulator of stomatal opening. The amiRNA seed resource generated here provides a new and potent tool to identify redundant genes and also lethality causing higher order mutants in many biological processes in Arabidopsis.
In conclusion, the amiRNA library resource is well suited for screening of phenotypes that can be easily verified in subsequent generations. This population may be best suited for screens that permit high throughput or medium throughput screening for phenotypes with a large dynamic range. Many such powerful screens have been performed in classical Arabidopsis mutant screens that were previously not designed to identify functionally redundant genes.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. The induction of RAB18 gene expression by ABA is lower in the amiRNA-AIG lines. Fig. S2. The expression of AIG2 genes in amiRNA-AIGs lines. Fig. S3. The p9l22 amiRNA line has normal stomatal indices and density. Fig. S4. The expression of PAB1, PAB2, and PAG1 genes in amiRNA lines. Table 1. List of relevant primers used in this study. Table 2. List of relevant plasmids used in this study. Table 3. List of new amiRNAs designed and cloned in this study. Table 4. AmiRNA sequences and predicted target genes found in candidate plants.