RING-Type E3 Ubiqitin Ligase Barley Genes (HvYrg1–2) Control Characteristics of Both Vegetative Organs and Seeds as Yield Components

Previously, studies on RING-type E3 ubiquitin ligases in cereals were preferentially focused on GW2 genes primarily controlling seed parameters in rice and wheat. Here we report cloning two HvYrg genes from barley that share significant homology with rice GW2 gene. In antisense genotypes efficiency of gene silencing varied between genes and transgenic lines: ASHvYrg1: 30–50% and ASHvYrg2: 20–27%. Reduced activity of both genes altered shoot system with increasing number of side shoots. Changes in leaf width, weight, or plant weight and height reached significant levels in some transgenic lines. Lowering expression of the two barley HvYrg genes caused opposite responses in spike development. Plants with ASHvYrg1 gene construct showed earlier heading time and prolonged grain-filling period, while plants from ASHvYrg2 genotype flowered in delay. Digital imaging of root development revealed that down-regulation of HvYrg1 gene variant stimulated root growth, while ASHvYrg2 plants developed reduced root system. Comparison of seed parameters indicated an increase in thousand grain weight accompanied with longer and wider seed morphology. In summary we conclude that in contrast to inhibition of GW2 genes in rice and wheat plants, down-regulation of the barely HvYrg genes caused substantial changes in vegetative organs in addition to alteration of seed parameters.


Introduction
Insuring the yield potential and stability of small-grain cereals, such as wheat (Triticum species), rice (Oryza sativa L.), and barley (Hordeum vulgare L.) is a priority for global food security. Demand for cereals is expected to rise in coming decades because of increase in population and income, furthermore the continuous reduction of arable land worldwide. Currently, barley is the fourth most important cereal in the world, and barley grains are mainly used for animal feed and in the production of alcoholic reduced expression of HvYrg1 and HvYrg2 genes were used for quantification of phenotypic parameters of green organs, roots, and mature seeds. Despite a variable degree of gene silencing, the presented data support the conclusion that RING-type E3 ubiquitin ligases encoded by HvYrg1 and HvYrg2 are involved in regulation of both vegetative and generative organ size and developmental processes as timing heading or seed maturation in barley.

Two Homologs of HvYrg Genes Encoding RING-Type Protein as E3 Ubiquitin Ligases in Barley
Early discovery of the GW2 gene encoding a RING-type protein as E3 ubiquitin ligase regulating seed size parameters in rice [3] has initiated wide research activities on different plant species including the major cereal crops [4,5]. The phylogenetic relations-based on their amino acid sequences-between the orthologs among the grass species are presented in Figure 1. The present work extends functional characterization of two orthologs of rice GW2 gene from barley representing different genes with 62.1% identity at nucleotide level. They were annotated under gene name or accession numbers as HvYrg1 (EU333863) and AK250398 as HvYrg2 in the DDBJ/EMBL/GenBank Data Libraries. Amino acid sequence identity between the two RING protein encoded by the HvYrg1 and HvYrg2 genes was only 47.7% that can reflect functional divergence (see Supplementary Figure S1). Comparison of the two RING-finger domains revealed differences between these proteins at the conserved histidine (His) residue. The HvYrg2 protein lacks this residue that is required for metal ligands (ML) to chelate two zinc atoms and define a cross-brace secondary structure for binding to the ubiquitin-conjugated E2 enzyme [4]. This structural characteristic was also recognized in two Arabidopsis RING-type E3 proteins (RING-C2 and Ring-v) as published by [15]. According to Song et al. because of the lack of a histidine residue these structures cannot be categorized as a RING domain [3].

Production of T 2 Generation Transgenic Barley Plants with Silenced Expression of the HvYrg Gene Variants
Antisense transgene-mediated gene silencing systems are widely used in plant research as powerful reverse genetic methods for studying gene function [16]. In the present studies, we generated HvYrg1 and HvYrg2 transgenic barley lines constitutively expressing DNA fragments from Yrg1 and Yrg2 genes in antisense direction. The T 2 progenies of the hygromycin resistant T 1 plants were genotyped by PCR analyses. As shown by Table 1, transgenic barley plants from four ASHvYrg1 and two ASHvYrg2 lines were included into detailed characterization. Reduction in expression levels was higher (30-50%) in leaves of ASHvYrg1 plants (except: ASHvYrg1/1: 20% reduction), than in leaves of ASHvYrg2 plants (20-27%) as shown by Table 1. Similar range of silencing efficiencies was reported for the wheat TaGW2 gene [8].

Altered Vegetative Organ Parameters in Transgenic Barley Plants with Silenced HvYrg Genes
Since antisense expression of the barley HvYrg gene fragments was based on the highly active constitutive ubiquitin promoter [17], we were able to analyze possible alterations in vegetative growth and organ sizes in addition to seed parameters. Here we report results of phenotyping from two different cultivation methodologies. In the first case the barley plants were grown in a traditional greenhouse, while the second experiment was carried out by using a semi-automated phenotyping platform for monitoring root traits [18,19]. As shown by Table 1, significant increase in flag leaf weight could be recorded in antisense transgenic plants from ASHvYrg1/1, ASHvYrg1/2, and ASHvYrg2/1 lines. Silencing both HvYrg variants altered the shoot architecture. Golden Promise (GP) plants served as references produced preferentially primary shoots, while in the transgenic plants number of side shoots was higher than of main shoots (Table 1). These phenotypic responses may reflect changes in hormonal status as results of lowered activity of HvYrg genes. RING-type E3 ligases are important regulators of ABA and ethylene signaling [6]. In grasses, the axillary meristems are capable of giving rise to side tillers and the increased branching in transgenic barley plants may originate from a lowered dormancy effect of ABA. Several barley mutants (granum-a (gra-a), grassy tillers (grassy), intermedium-c (int-c), many noded dwarf1 (mnd1), and many noded dwarf6 (mnd6) exhibit enhanced tillering [20]. In the present case, we can see an increased number of side shoots in transgenic barley plants with reduced expression of both HvYrg gene variants. In attempts to interpret global consequences of silencing the HvYrg RING-type E3 ligase genes, we have to consider relationships between seed and tillering parameters that is proposed by studies on several barley mutants [21]. At the end of growing period, silencing the variant HvYrg1 gene caused 14-16% increase in biomass expressed as stem dry weight (Table 1). These alterations in plants from ASHvYrg1/1, and ASHvYrg1/4 lines are in accordance with the higher tiller number and increased plant heights. In contrast to these observations, transgenic rice plants with antisense of GW2 gene did not show alteration in plant height and flag leaf width [3]. Another type zinc finger gene (NbZFP1) encoding C3HC4-type RING finger proteins from Nicotiana benthamiana was silenced by VIGS technique without phenotypic changes [22].

Down-Regulation of Barley HvYrg Genes Modulates Growth and Developmental Characteristics of Transgenic Plants
Quantitative parameters presented in the Table 1, highlight essential changes in plant architecture as consequences of reduction in RING-type E3 ubiquitin ligase function. Figure 2 shows characteristic phenotypes of Golden Promise and antisense-transgenic plants at vegetative growth phase. Changes in shoot size and number can be recognized on the presented examples. Table 1. Reduced expression of HvYrg gene variants in leaves of antisense transgenic barley plants and parameters of vegetative traits. The mean ± SD were calculated from the data of 9-20 plants per genotype. Based on Student's T-test, statistically significant events compared with Golden Promise (GP) mean value are indicated below with mean labels as p ≤ 0.001 ***, p ≤ 0.01 ** and p ≤ 0.05 *.

Genotype
Golden  At vegetative growing phase, we could recognize increase in leaf width only in ASHvYrg2 transgenic plants (Table 2). In contrast, ASHvYrg1 plants developed leaves with similar wideness or narrower as the wild GP plants. These silencing effects can discriminate between the two HvYrg gene variants. In later developmental phases, we monitored the induction of flowering and inflorescence development. Using reference Golden Promise plants, the down-regulation of the two barley HvYrg gene variants resulted in opposite responses. Spikes of plants from ASHvYrg1 transformants emerged from the boot earlier within 5-7 days. In contrast, the ASHvYrg2 plants showed delayed transition to flowering ( Figure 3). Boden et al. [23] reported that the early flowering and vegetative growth phenotypes of the barley elf3 mutant can be related to gibberellin (GA) biosynthesis. The SCF type E3 ligases have been linked to GA pathway [6]. In rice, mutation in the Heading date Associated Factor 1 (HAF1) gene can cause a later flowering date. This factor was identified C3HC4 RING domain-containing E3 ubiquitin ligase [24]. Alterations in activities of HvYrg1 gene variant resulted not only in earlier heading dates, but also affected grain-filling duration. As shown by the Figure 4. despite of the earlier spike development, the grain-filling period was prolonged in ASHvYrg 1/2 and ASHvYrg 1/4 plants. In cereals, the duration of grain filling can influnce grain weight [25]. Since considerable reduction in HvYrg1 gene expression was recorded in plants from these genotypes (Table 1), essential changes in the hormonal status of spike could be expected as consequence of the lowered RING-type E3 ubiquitin ligase function [6]. Spike development is an important factor in the yield production of barley that is under the control of hormonal crosstalk [26].

Down-Regulation of the HvYrg Gene Variants Can Differentially Modify Development of Root System Monitored in a Semi-Automated Phenotyping Platform
As an outcome of intensified plant phenotyping research, several alternative methods exist for the non-destructive imaging of root systems grown in either soil-free medium or rhizotrons filled with soil [27]. In the present study barley plants were grown in plexiglass columns that allowed to photograph the root system from four different side positions and from the bottom. The root-related white pixels were identified by subtracting the black soil background from the images. This methodology provided root density information about the development of roots during the growing period. As shown by Figure 5A, growth of root systems became intensified after eighth week of growing period with different rates between various genotypes according to side images. Down-regulation of ASHvYrg2/1 gene variant stimulated root biomass in comparison to GP or ASHvYrg1/2 plants. The later ones developed significantly reduced root system. Developmental differences can be seen in Figure 6 that also presents delay in growth of root system of ASHvYrg1/2 plants. Root parameters were also quantified by images taken from the bottom side. Transgenic plants accumulated less root biomass at the bottom of plexiglass columns than the Golden Promise plants ( Figure 5B). were identified by subtracting the black soil background from the images. This methodology provided root density information about the development of roots during the growing period. As shown by Figure 5A, growth of root systems became intensified after eighth week of growing period with different rates between various genotypes according to side images. Down-regulation of ASHvYrg2/1 gene variant stimulated root biomass in comparison to GP or ASHvYrg1/2 plants. The later ones developed significantly reduced root system. Developmental differences can be seen in Figure 6 that also presents delay in growth of root system of ASHvYrg1/2 plants. Root parameters were also quantified by images taken from the bottom side. Transgenic plants accumulated less root biomass at the bottom of plexiglass columns than the Golden Promise plants ( Figure 5B).   As reviewed by [5] the RING-type E3 ligases play a key role in the control of different root systems. In rice plants overexpressing mutant EL5 proteins that are impaired in E3 activity exhibited rootless phenotype accompanied by cell death in root primordia [28]. The presented data provide additional insight into functional differentiation between the two barley HvYrg gene variants.

Silencing the HvYrg Barley Genes Can Alter Seed Parameters in Antisense Transgenic Plants
Transgenic barley plants carrying antisense constructs for HvYrg1 gene variant were also characterized for seed production-related traits (Table 3). We selected this gene variant since in the T 2 generation we could identify PCR+ and PCR− segregants. Data presented in the Table 3, shows comparison of these genotype categories. Considering the significant alterations in number of main and side shoots (Table 1) spike parameters were separately recorded in case of these different shoots. As number of kernels per spike is considered, silencing HvYrg1 resulted in variable, transgenic line-dependent changes. The PCR+ plants from HvYrg1/ 1 line produced significantly more seeds per spike from the main branches than the PCR− variants. Similar trend could not be seen in the main branches of plants from HvYrg1/4 genotype. Reduction of kernel number/spike was characteristic for PCR+ plants from this genotype. Opposite, no significant changes could be seen in spikes from side shoots Comparison of this trait from PCR+ and PCR− plants of HvYrg1/2 showed no difference between PCR+ and PCR− genotypes. Based on SD values, seed number/spike values exhibited considerable variation between plants from the same genotype. Reference data are available from studies on rice and wheat transformants with silenced GW2 gene [3,8]. In wheat no significant differences in grain number per spike, while in rice GW2 loss-of-function reduced the grain number per spike. Comparison of kernel weight/spike between PCR+ and PCR− plants presented increase in transgenic plants of the genotypes (ASHvYrg1/2 and ASHvYrg1/4) with silenced HvYrg1 gene. This alteration reached significant level only in the case of ASHvYrg1/1 plants (Table 3). This trend cannot be seen on the spikes from the side tillers. Table 3. Comparison of grain yield-related traits from transgenic (PCR+) and non-transgenic (PCR−) segregants in T2 generation of antisense transformation of HvYrg1 gene of barley. The mean ± SD were calculated from the data of 14 (PCR+) and 5 (PCR−) plants per genotype. Based on Student's T-test, statistically significant events compared with Golden Promise mean value are indicated below with mean labels as p ≤ 0.001 ***, p ≤ 0.01 ** and p ≤ 0.05 *.

Genotypes
Origin Of all the factors that influence yield, kernel weight measured as thousand kernel weight (TKW) is in the center of barley improvement. In the present study, we have recorded significant positive changes in this trait in main spikes of PCR+ plants from ASHvYrg1/4 genotype compared to the PCR− segregants. Interestingly, side heads of ASHvYrg1/2 and ASHvYrg1/4 of barley produced larger seeds. The TKW values from side shoots were higher even in comparison to the Golden Promise seeds. Since in these transgenic plants the silencing effects were significant (see Table 1) we may postulate the negative regulatory role of HvYrg1 gene in control of seed development. Increased TKW parameters were described for antisense GW2 transformants of rice [3]. This modification was accompanied with reduction in grain number per main panicle. Similar trend can be seen in spikes of ASHvYrg1/4 of barley plants.
As an additional grain parameter, the weights of single grains were also characterized. Because of the significant variation in this quantitative trait, we present distribution curves in addition to the average values. Figure 7A clearly shows larger kernels as shown by the TKW values for the transgenic plants carrying silenced variants of both HvYrg genes in T 2 generation in comparison to GP kernels. The average TKW values from main and side spikes was 29.69 g for GP, 35.53 g for ASHvYrg1/1 and 34.94 g for ASHvYrg2/2 plants. Detection of PCR+ and PCR− segregants in the same plant population allowed direct comparison of phenotypic differences caused by the reduction of gene expression. Figure 7B shows a shift of the distribution curve towards larger kernels for N • 8 plants with silenced HvYrg1 gene expression in comparison to the PCR− N • 3 plants.
Plants 2020, 9, 0 11 of 15 Of all the factors that influence yield, kernel weight measured as thousand kernel weight (TKW) is in the center of barley improvement. In the present study, we have recorded significant positive changes in this trait in main spikes of PCR+ plants from ASHvYrg1/4 genotype compared to the PCR− segregants. Interestingly, side heads of ASHvYrg1/2 and ASHvYrg1/4 of barley produced larger seeds. The TKW values from side shoots were higher even in comparison to the Golden Promise seeds. Since in these transgenic plants the silencing effects were significant (see Table 1) we may postulate the negative regulatory role of HvYrg1 gene in control of seed development. Increased TKW parameters were described for antisense GW2 transformants of rice [3]. This modification was accompanied with reduction in grain number per main panicle. Similar trend can be seen in spikes of ASHvYrg1/4 of barley plants.
As an additional grain parameter, the weights of single grains were also characterized. Because of the significant variation in this quantitative trait, we present distribution curves in addition to the average values. Figure 7A clearly shows larger kernels as shown by the TKW values for the transgenic plants carrying silenced variants of both HvYrg genes in T 2 generation in comparison to GP kernels. The average TKW values from main and side spikes was 29.69 g for GP, 35.53 g for ASHvYrg1/1 and 34.94 g for ASHvYrg2/2 plants. Detection of PCR+ and PCR− segregants in the same plant population allowed direct comparison of phenotypic differences caused by the reduction of gene expression. Figure 7B shows a shift of the distribution curve towards larger kernels for N • 8 plants with silenced HvYrg1 gene expression in comparison to the PCR− N • 3 plants. Since increase in kernel weight of NIL(GW2) in rice plants was primarily due to increased grain width, followed by grain thickness and length [3], we also quantified the seed size parameters in barley AS genotypes. As described in the Material and Methods grain length and width parameters were quantified by the Seed Size Analysis Program v0.95. Pixels for the color of seed surface were used for prediction of seed size. As shown by Table 4 grains from ASHvYrg1/3, ASHvYrg1/4, and ASHvYrg2/1 lines were significantly longer than the Golden Promise kernels. We also measured the length of individual grains (Supplementary Figure S2). Frequency of longer kernels was increased in the case of ASHvYrg1/4 plants.
In agreement with ASGW2 transgenic rice seeds [3] the wideness of kernels from three barley transformed lines (ASHvYrg1/3; ASHvYrg1/4, and ASHvYrg2/1) was also increased according to pixel based calculation (Table 4). Plants from the ASHvYrg1/2 genotypes showed significant reduction in this trait.
According to data of single kernels produced by plants from ASHvYrg1/1 and ASHvYrg2/2 lines developed wider kernels than of GP plants (Supplementary Figure S3). Comparison of kernels from Since increase in kernel weight of NIL(GW2) in rice plants was primarily due to increased grain width, followed by grain thickness and length [3], we also quantified the seed size parameters in barley AS genotypes. As described in the Material and Methods grain length and width parameters were quantified by the Seed Size Analysis Program v0.95. Pixels for the color of seed surface were used for prediction of seed size. As shown by Table 4 grains from ASHvYrg1/3, ASHvYrg1/4, and ASHvYrg2/1 lines were significantly longer than the Golden Promise kernels. We also measured the length of individual grains (Supplementary Figure S2). Frequency of longer kernels was increased in the case of ASHvYrg1/4 plants.
In agreement with ASGW2 transgenic rice seeds [3] the wideness of kernels from three barley transformed lines (ASHvYrg1/3; ASHvYrg1/4, and ASHvYrg2/1) was also increased according to pixel based calculation (Table 4). Plants from the ASHvYrg1/2 genotypes showed significant reduction in this trait.
According to data of single kernels produced by plants from ASHvYrg1/1 and ASHvYrg2/2 lines developed wider kernels than of GP plants (Supplementary Figure S3). Comparison of kernels from single PCR+ and PCR− plants showed similar alteration in this trait ( Figure 7B). In interpreting these alterations, we can rely on studies on rice ASGW2 transgenics [3]. The outer parenchyma cell layer contained substantially more cells, while endosperm cells showed increased size without changes in the cell number. The presented ASGW2 transgenic plants can serve as an experimental material for future studies on involvement of RING-type E3 ligases in cell division control in plants. Despite of variable efficiency in silencing of HvYrg gene variants, the present study focuses the attention on a central role of RING-type E3 ligase pathway in controlling cereal plant architecture, growth and development. The presented alterations detected in phenotypic traits of antisense barley plants are similar to those described in studies on seed characteristics of rice and wheat plants with additional insight to response of vegetative organ including shoot and root system. Role of HvYrg gene variants in regulation of heading time and grain filling period is supported by this analysis. The present use of constitutive promoter allowed to gain basic functional information, but it can be limiting factor by generating multiple alterations in different traits in a breeding project. Therefore, based on the presented results, it is advisable to induce gene specific mutations by genome editing tools for improvement of agronomic traits.

Construction of Antisense HvYrg RING-Type E3 Expression Vectors and Barley Transformation
Genomic DNA samples were isolated with a CTAB-based extraction method according to [29] from "Golden Promise" (GP) barley variety. BLAST analysis was carried out to identify the barley orthologues of GW2 RING-type E3 genes (HvYrg1 and HvYrg2) in the barley genome sequence data base [30]. The presented primer sequences amplified 512 bps of HvYrg1 (Yrg1_Forward 5 -GGGAGCTTTATGCCTTTTGAGCAACC-3 , Reverse5 -GTGTGCGTTCTACCATGAGCTTCTGC-3 ) and 539 bps of HvYrg2 (Yrg2_Forward 5 -ATAGGTGCCGTGCCACCAACAC-3 , Reverse 5 -TACCGCCAAGCTAACGCTGGAG-3 ) fragments. In an additional PCR cycle, these fragments were extended by SpeI and BamHI restriction sites. The fragments were digested with the appropriate enzymes (Thermo Scientific, Waltham, USA) and were cloned at BamHI and SpeI sites in the first cloning site of pUbi-AB intermediate vector [31]. This expression cassette was transferred into the p6d35s binary plant expression vector [32] with double enhanced CaMV35S promoter using SfiI (New England Biolabs, Hitchin, Hertfordshire, UK) digestion and ligation [31]. Plasmid DNA was purified with GenElute™ HP Plasmid Miniprep Kit (Sigma, St. Louis, MO, USA) and the nucleotide sequences of the GW2 RING-type E3 gene constructions were determined by the ABI 3100 Genetic Analyzer from Applied Biosystem (Foster City, CA, USA).
For the stable transformation of immature barley embryos (GP) we used the LBA4404 Agrobacterium tumefaciens strain by following protocol from [33]. Transgenic barley plants were tested for the integration of GW2 Ring-type E3 gene fragments and hygromycin resistance genes (Forward 5 -CCTGAACTCACCGCGAC -3 , Reverse 5 -GCTCATCGAGAGCCTGC -3 ). We were able to establish four independent transgenic lines producing T2 generation plants with the ASHvYrg1 construct and two lines with the ASHvYrg2 constructs.

Growth Conditions for Barley Plants in Greenhouse or Semi-Automatic Phenotyping Platform
Regenerated hygromycin resistant plantlets were cultured in vitro and later transferred into soil in greenhouse. The T 1 seeds were collected and sown into plastic pots (diameter 16 cm) containing a mixture of soil and sand (2:1, v/v) under a cycle of 12 h illumination (250 mmol m_2 s_1)/12 h dark. The T 2 plants from the analyzed transgenic lines were tested for the presence of HPT selective marker gene and the HvYrg Ring-type E3 gene fragments in antisense orientation.
At the end of vegetative growing period, we measured parameters as plant height, dry biomass of whole plants, and of three flag leaves. After harvesting, we determined seed number and weight per spike, furthermore thousand kernel weight (TKW). For comparative quantification of individual seeds (seed size, length, and width) we used pixel-based imaging. From each genotype, 80-80 seeds were measured individually with Ohaus Model EP114C analytical balance by 0.1 mg accuracy. For characterization of root growth of barley plant from GP and ASHvYrg1 and ASHvYrg2 lines we used a semiautomatic phenotyping platform described previously [18,19,34]. The T 2 seeds were sown into radio-tagged plexiglass columns with a mixture of 80% Florimo peat soil and 20% sandy soil. Five plexiglass columns surrounded with polyvinyl chloride tubing were placed on a metal rack. Three racks were used for each genotype with random arrangement. In the case of roots, the plexiglass columns were photographed from four different side positions and from the bottom. The root-related white pixels were identified by subtracting the black soil background from the images. Pixel numbers were converted to millimeters using 95-mm diameter pots captured in the images. To characterize the root area appearing at the surface of the chamber, the metric values of the area of the four side view projections (90 • rotation) are summarized and the metric value of the area of the bottom view.

Quantitative Real-Time PCR (qRT-PCR) for Measurement of Expression of HvYrg1 and HvYrg2 Gene Variants in Transgenic Barley Plants
The total RNA samples were isolated from young leaves of two individuals of GP and T 2 generation barley plants according to the AGPC (acid guanidinium thiocyanate-phenol chloroform) method [35]. For the cDNA synthesis 1 µg of total RNA was used, following the First Strand cDNA Synthesis Kit manual (Fermentas Life Sciences, Vilnius, Lithuania). For the quantitative RT-PCR reactions we used the following primers: ASHvYrg1_Forward 5 -TGCAGCACATCCTATTCAGC-3 , Reverse 5 -GAATGGAAAGACCGCATGTT-3 , ASHvYrg2_Forward 5 -GTTTCCTCTTGTGCGTGACA-3 , Reverse 5 -TATCCCACCAGTCCCTATGC-3 .
These were designed by the Primer Express Software from Applied Biosystems (Foster City, USA). The qRT-PCR reactions were carried out in the ABI PRISM 7000 Sequence Detection System using the SYBR Green PCR Master Mix and the reaction conditions were the same as described by [36]. The 2 −∆∆C(T) method was used to analyze the real-time PCR data [37] and the expression of our examined genes was normalized to the reference gene (18S RNA, Forward 5 -GTGACGGGTGACGGAGAATT-3 , Reverse 5 -GACACTAATGCGCCCGGTAT-3 ).

Data Management
The statistical significance of the results was determined using the Microsoft Excel 2003 software (Microsoft Inc., Redmond, WA, USA) Student's T-test.