Identification of genes involved in the formation of soluble dietary fiber in winter rye grain and their expression in cultivars with different viscosities of wholemeal water extract

Liudmil V. Kozlov, Alsu R. Nzipov, Oleg V. Gorshkov, Liliy F. Gilmullin, Olg V. Sutkin,Ntli V. Petrov, Oksn I. Trofimov, Sergey N. Ponomrev, Mir L. Ponomrev,Ttyn A. Gorshkov,*

a Kazan Institute of Biochemistry and Biophysics, Federal Research Center, Kazan Scientific Center of the Russian Academy of Sciences, 420111 Kazan, Russia

b Tatar Scientific Research Institute of Agriculture, Federal Research Center, Kazan Scientific Center of the Russian Academy of Sciences, 420059 Kazan, Russia

Keywords:Rye (Secale cereale)Kernel development Arabinoxylan Mixed-linkage glucan Viscosity

ABSTRACT The grain of rye(Secale cereale L.)used for baking contains a large amount of non-starch polysaccharides,making it an excellent component of functional foods.But rye grain intended for alcohol production and forage use should have a reduced content of these polysaccharides.A comprehensive parameter that can predict the best field of application for winter rye grain is the viscosity of its wholemeal water extract.However,our understanding of the genetic background underlying this key trait and associated features of rye grain is poor. By analyzing six Russian winter rye cultivars, we identified the most contrasting forms and characterized the peculiarities of their water-soluble carbohydrates capable of influencing the viscosity of water extracts. Then, using phylogenetic and transcriptomic analyses, we identified in the rye genome many genes encoding putative glycosyltransferases and glycosylhydrolases responsible for the synthesis and degradation of arabinoxylans, mixed-linkage glucans, cellulose, and some other polysaccharides.We determined the dynamics of mRNA abundance for these genes at three stages of kernel development. Comparisons of gene expression levels in two contrasting cultivars revealed specific members of multigene families that may serve as promising targets for manipulating non-starch polysaccharide content in rye grain.High-viscosity cultivars were characterized by up-regulation of many glycosyltransferases involved in the biosynthesis of arabinoxylans and other cell-wall polysaccharides,whereas low-viscosity cultivars showed up-regulation of several genes encoding polysaccharidedegrading enzymes.

1. Introduction

Rye (Secale cereale L.) is a mainly European cereal crop. It has superior winter hardiness and tolerance to drought and soil salinity and acidity in comparison with wheat plants.Rye grain is used in bread baking, animal feed, and alcohol and starch production[1,2]. In many aspects, rye is similar to wheat, but it has a greater content of dietary fiber(non-starch polysaccharides,including arabinoxylans[AXs],mixed-linkage glucans[MLGs],arabinogalactans,and cellulose,and also lignin)[3,4].Rye flour is composed of 56%–70%starch,8%–13%protein,2%–3%fats,and 15%–21%dietary fiber,only one fourth of which is water-soluble [5–7].

The AXs of cereals are often referred to as pentosans,because in the endosperm they are composed of five-carbon monosaccharides[8]. But it should be noted that(i) grain xylans, especially those of seed coats, do contain some hexoses, such as glucuronic acid and galactose, in their side chains [9]; and (ii) not all pentoses belong to the AXs,given that some arabinosyl and xylosyl residues belong to arabinogalactans, xyloglucans, and pectins [10].

A common point of view[11]considers AX the key regulator of water extract viscosity, meaning that its composition and content would determine the preferred sphere of grain use. Bread that is baked with rye flour having a higher AX content, and correspondingly higher extract viscosity, has a larger volume and gasretention ability, so that it hardens more slowly than does bread baked using rye with a lower AX content [4,12,13]. Accordingly,the breeding of rye cultivars designated for human food production is directed towards increasing the grain content of AXs (especially of their water-soluble fraction).

At the same time, however, a high AX content greatly restricts the use of rye in animal feed, especially for monogastric animals and poultry [14–17]. AXs slow down the bolus movement in the digestive tract,disturb the digestive process,reduce the absorption of nutrients,and consequently reduce gains in animal body weight.Too much AXs in rye grain is also unfavorable for alcohol and bioethanol production,given that AXs form a viscous sludge inside the distillation equipment that must be removed at extra cost,prolonging this technological process [18,19].

Thus, the challenges of breeding rye for baking, forage, and industry are not the same. Moreover, we still lack fundamental knowledge of the formation and function of non-starch polysaccharides of rye grain. Such knowledge could become the basis for further breeding and for making sound recommendations for the food industry or agriculture.

Recent progress made primarily in wheat, rice, and Arabidopsis revealed the main glycosyltransferases(GT)involved in the formation of AXs(Fig.1).The polysaccharide backbone is synthesized by an enzyme complex that includes three key proteins: IRX9 and IRX14 from the GT43 family [20–22] and IRX10 (XYS1) from GT47 [23,24]. Arabinose and xylose can undergo attachment to the xylan backbone, as side chains, by GT61 family members XAT, XAX, MUCI21, and XYXT [25–27], and likewise glucuronic acid by members of the GT8 family[28](Fig.1).Arabinose attached to the xylan backbone can be substituted with ferulic acids that form cross-links between xylan molecules [29]. The feruloylation of arabinoxylan engages BAHD acyltransferases, which harbor a PF02458 domain and cluster into a separate, so-called Mitchell’s clade [30] within a massive superfamily of coenzyme A-using transferases. The important properties of xylans, and their ability to interact with other polymers, are considerably modulated by acetyl groups attached to them by acetyltransferases from the TBL(TRICHOME-BIREFRINGENCE-LIKE)and RWA(REDUCED WALL O-ACETYLATION)families[31](Fig.1).Members of the GT2 protein family are involved in the formation of several cell wall polysaccharides including cellulose,MLGs,mannans,and xyloglucans[32].

The plant cell wall is a dynamic structure and myriad enzymes modify its existing polysaccharides in the course of various physiological processes, such as cell growth [33,34], seed development[35,36], and fruit ripening [37,38]. The action of such enzymes may influence the molecular mass of a polysaccharide and its degree of substitution and thereby affect its interaction with other cell wall components, resulting in solubility changes. In this way,such enzymes may contribute to both the yield and properties of rye grain’s water-extractable fraction [8].

A polysaccharide may be modified by exo-glycosylhydrolases that attack at nonreducing termini, to release monosaccharides;or by endo-glycosylhydrolases that act upon polysaccharides at any position except the termini,releasing oligosaccharide products[39] (Fig. 1). The backbone of AX can be degraded by β-xylanases and β-xylosidases, while α-arabinofuranosidases can remove its arabinose side chains. In planta, MLG can be modified by βglucanases and β-glucosidases (Fig. 1).

Fig. 1. Enzymes mediating the biosynthesis and degradation of main cell-wall polysaccharides of cereals. Blue arrows indicate the biosynthesis of a specific bond; pink arrows indicate its breakdown.Proteins are named according to the literature describing their activity(see the text for references).Protein family or clade within the family is given in brackets. GT, glycosyltransferases, GH, glycosylhydrolases. ?, plant GH9 enzymes have no evidence of in vivo action as cellulases. GH3 and GH51 are exoglycosylhydrolases, while GH9 and GH17 are endo-glycosylhydrolases.

None of the abovementioned proteins or corresponding genes has been characterized in rye. The rye genome of the inbred line‘Lo7’ was sequenced in 2017 [40], and the high-quality genome assembly for the same line was published in 2021 [41]. According to the latest model,57,222 gene models were predicted.However,genes whose expression determines the content of soluble dietary fiber have yet to be characterized. We were aimed to detect them and to check their expression in winter rye cultivars whose water extract viscosity varied substantially.We surmised that such genes would serve as suitable targets for further rye breeding.

2. Materials and methods

2.1. Plant material

Experiments were performed over four successive growing seasons(2016–2019)in the Laishev district,Tatarstan Republic,Russia(55.649°N,49.3083°E).Grain samples of winter rye were harvested from field trials at Tatar Research Institute of Agriculture, FRC Kazan Scientific Center of RAS, located in the forest–steppe area of the Volga region(Tatarstan Republic),carried out under uniform agronomic management(sowing date,rate,depth,soil preparation,application of fertilizer and herbicides) for winter rye production.The study was conducted on a gray forest soil with a humus content of 3.1%.

Six Russian commercial rye cultivars—Tatarskaya-1, Radon,Ogonek, Podarok, Marusenka, and Pamyati Kunakbaeva—were used.Plots(16 m2)of each cultivar were arranged in four replicates in the field trial in a randomized block design and were managed according to locally recommended cultural practices for winter rye production. Samples from each plot were collected and handled separately.

2.2. Viscosity of water extract, water-extractable, and total pentosan content

From each cultivar, 300 g of rye grain was milled into wholemeal flour (using an extraction rate of 96%–98%) using 0.8-mm sieve with a Perten Instruments Laboratory Mill 3100 (Perten Instruments, Stockholm, Sweden). The moisture content of flour samples was determined using an air–oven method (AACC 44–15.02).All samples of flour obtained from the four annual harvests had a dry matter content of about 88%, and were stored at room temperature until analysis.

The water extraction procedure was performed at 30°C following Boros et al. [42]. Water-extractable pentosan (WEP) content was measured by orcinol-chlorid method [43]as modified by Delcour et al. [44] for rye grain. The kinematic viscosity of water extract (VWE) was estimated at 30 °C using an extract volume of 2 mL per assay and a glass capillary viscometer,Labteh VPG 1(Labteh, Moscow, Russia); for more details, refer to Ponomareva et al.[45]. Total pentosan content (% of flour weight) was determined by near-infrared (NIR) spectroscopy with a Multi Purpose MPA II FT-NIR Analyzer (NIRS) (Bruker Optics, Billerica, MA, USA, ISO 9001)in the wholemeal flour samples from the four replicate plots,with two technical replications performed for each.

One-way analysis of variance (ANOVA) was performed for the kinematic viscosity of water extract, water-extractable and total pentosan content of wholemeal flour. All experiments were duplicated and the means of the six cultivars were compared using Duncan’s multiple range test (P ＜0.05). Pearson correlations between various wholemeal flour properties were calculated and considered significant at P ＜0.05.

2.3. Carbohydrate determination and chromatography

Water-extractable carbohydrates were obtained from the wholemeal flour from the six rye cultivars. Flour (200 mg) was incubated in 2 mL of MilliQ water for 1 h,at 40°C,with gentle agitation. Each extract was separated by centrifugation (10 min at 5000×g) and incubated for 10 min at 100 °C to inactivate endogenous enzymes. Ethanol precipitation at 4 °C overnight was performed to obtain the polysaccharides. The molecular-weight distribution of total and ethanol-precipitated fractions was characterized by size-exclusion chromatography (SEC), on an Agilent Infinity 1260 instrument (Agilent, Santa Clara, CA, USA) using a Shodex 806-M column (8 × 300 mm, Shodex, Tokyo, Japan) with a SB-G precolumn(6×50 mm,Shodex)and refractive index detection at 35 °C. UV absorption was recorded at respectively 280 and 320 nm for proteins and phenolic compounds using a WR 1260 diode array detector (Agilent). Elution was performed with MilliQ water at 50°C and a flow rate of 0.3 mL min-1.Pullulans(molecular weight range 0.342–800 kDa,Sigma-Aldrich,St.Louis,MO,USA)and D-glucose were used for column calibration. Agilent GPC/SEC software (version 1.2) (Agilent) come with the SEC system was used to characterize molecular weight distribution of analyzed samples.

Monosaccharide composition was characterized by highperformance anion-exchange chromatography with pulsed amperometric detection, on a Dionex DX-500 instrument (Dionex, Sunnyvale, CA, USA) equipped with a Carbo-Pac PA-1 (4 × 250 mm,Dionex) column. Eluents used were as follows: A, 200 mmol L-1NaOH; B, 100 mmol L-1NaOH in 1 mol L-1NaOAc; and C, MilliQ water.Elution followed Mikshina et al.[46]with one modification:in the first stage of elution the concentration of NaOH was 18 instead of 15 mmol L-1). Wholemeal flour water extract was diluted 100 times with MilliQ water and then injected into a column to estimate the content of free monosaccharides. Samples were degraded to monosaccharides after an aliquot was dried under airflow at 60 °C and subjected to hydrolysis for 1 h in 2 mol L-1trifluoroacetic acid(TFA) at 120 °C. To build the calibration, L-rhamnose, L-arabinose, D-galactose, D-glucose, D-xylose,D-mannose, D-galacturonic, and D-glucuronic acids (Merck, Darmstadt, Germany) subjected to a similar TFA treatment were used.PeakNet software (version 4.30) (Dionex) come with the HPAECPAD system was used to quantify the monosaccharide composition of analyzed samples. All experiments were performed using three independent replicates. ANOVA followed by Tukey test at α = 0.05 for mean separation was performed for the results of SEC analysis. T-test was used to compare means of Tatarskaya-1 and Marusenka for each monosaccharide in monosaccharide analysis.

2.4. Immunodot-binding assay

Water extraction and ethanol precipitation were performed as described above for the wholemeal flour samples of two rye cultivars: Marusenka and Tatarskaya-1. Their carbohydrates were applied to nitrocellulose membranes 0.2 μm(Cytiva,Marlborough,MA, USA) in gradually decreasing concentrations in 2 μL per cell.The membranes were allowed to air-dry for 30 min, after which they were washed for 2 min in PBST (phosphate-buffered saline(PBS) with 0.05% (v/v) Triton X-100), blocked for 45 min with PBS containing 1% (w/v) bovine serum albumin (Sigma-Aldrich),and then incubated for 40 min with a primary monoclonal antibody. AX1, LM11, LM27, LM28, LM21, LM12, LM5, and LM26 antibodies were applied in a 1:40 dilution in PBST and the BG1 antibody was applied in a 1:1000 dilution. After incubation with the primary antibody, the membranes were washed three times for 10 min with PBST and incubated for 40 min with secondary biotinylated antibody diluted to 1:10000 (Sigma-Aldrich) (antirat for LM11, LM27, LM28, LM21, LM12, LM5, and LM26, and anti-mouse for AX1 and BG1). Each membrane was washed again(three times in PBST for 10 min),incubated for 1 h in avidin conjugated with alkaline phosphatase (Rockland Immunochemicals,Pottstown, PA, USA) diluted to 1:15000 in 0.1 mol L-1Tris-HCl,0.15 mol L-1NaCl, pH 7.5 + 0.05% (v/v) Triton X-100, and developed with a nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate kit (Sileks, Moscow, Russia) in 0.1 mol L-1Tris,0.1 mol L-1NaCl, 0.05 mol L-1MgCl2, pH 9.5. Wheat flour arabinoxylan, barley flour mixed-linkage glucan (MLG), carob galactomannan, potato galactan (Megazyme, Bray, Ireland),beechwood xylan(Sigma-Aldrich),beetroot pectin(extracted with HCl from beetroot pulp [47] and the glucuronoarabinoxylanenriched fraction from maize mesocotyls (extracted with KOH[48]) served as controls (Table S1). Next, every membrane was imaged with a digital camera (Xiaomi Mi9, Xiaomi,Beijing, China)with 12 megapixel resolution. All these experiments were performed using three independent replicates. Obtained images were analyzed with ImageJ2 Fuji software (https://imagej.net/). Images were converted into black and white. Mean labeling intensity was normalized to spot area. Background intensity was measured in a clear area in proximity to each spot and subtracted from the normalized value. T-test was used to compare means of Tatarskaya-1 and Marusenka for each antibody.

2.5. RNA extraction and sequencing

For transcriptomic analysis, samples were taken from plants at three time points:Z 76,Z 86,and Z 90,following Zadoks et al.[49],which corresponded to milk, dough, and full kernel maturity stages.Four biological replicates used for analysis represented four different field plots for each cultivar. Rye kernels from cultivars Tatarskaya-1 and Marusenka from the middle part of the spikes(10–15 homogeneous spikes were used for each biological replicate) were collected into plastic tubes containing liquid nitrogen and stored at –80 °C. Sample collection of two cultivars was performed at the same dates, given the similar precocity of Tatarskaya-1 and Marusenka, avoiding weather biases. The collected samples were ground in a homogenizer (MM400, Retsch GmbH,Haan,Germany).Total RNA from kernels at the milk maturity stage was isolated using a CTAB extraction buffer (2% [w/w]acetyl trimethylammonium bromide, 1% [w/w] polyvinyl pyrrolidone, 0.1 mol L-1Tris-HCl, 1.4 mol L-1NaCl, 20 mmol L-1EDTA[pH 8.0]).After this extraction,the buffer solution was mixed with a chloroform:isoamyl alcohol(24:1[v/v])solution in 1:1(v/v)proportion. Following centrifugation, the upper phase was separated and mixed with both an RLT buffer (1/8 of the upper-phase volume)of the RNeasy Mini Kit(Qiagen,Hilden,Germany)and a Plant RNA Isolation Aid (Ambion, Vilnius, Lithuania) (1/20 of the upperphase volume). Further extraction followed the manufacturer’s instructions.

For kernel samples at the dough and full maturity stages, total RNA was isolated with ExtractRNA reagent(Evrogen,Moscow,Russia) combined with the RNeasy Mini Kit (Qiagen) and Plant RNA Isolation Aid (Ambion), using the same proportions as above and according to the manufacturer’s instructions. Total RNA isolated at all three stages of kernel development was treated with a TURBO DNA-free kit (Thermo Fisher Scientific, Vilnius, Lithuania), to remove any residual DNA, and all samples were then stored at–80 °C.

RNA integrity test (RIN ＞7), library construction, and sequencing were performed by the Novogene Corporation (Beijing, China;https://en.novogene.com/). The sequencing data for the samples were submitted to the NCBI database (Bioproject ID:PRJNA738161).Different protocols for cell lysis do not significantly change RNA-Seq results if samples were treated similarly during further steps of RNA purification and sequencing [50,51].

2.6. Processing and analysis of RNA-Seq data

RNA-Seq data processing and counting of the gene reads for each sample were performed with the BBDuk utility of BBTools v37.02 [52] (https://jgi.doe.gov/data-and-tools/bbtools/) and Kallisto [53] with specific parameters for processing the paired-end Illumina (San Diego, CA, USA) reads. The assembly of a reference rye genome and its annotations [41] were retrieved, from the e!DAL server[54](https://edal-pgp.ipk-gatersleben.de/).The Bioconductor package(tximport,[55])was used to import the count data produced by Kallisto into R [56] (https://www.r-project.org/) for normalization and further analysis of differential expression using DESeq2 [57].

The dataset consisting of 29,514 genes,with a mean normalized total gene read number(TGR)≥16 in at least one sample,was used for differential gene expression analysis. Three criteria were used to determine significant differential expression of genes between the two cultivars(i)TGR value ＞150 in at least one sample,(ii)fold change ＞1.5 in at least one comparison, and (iii) adjusted P value ＜0.01 for this comparison.Differential expression was determined for milk ripeness stage only.

2.7.Searching for genes encoding enzymes involved in the biosynthesis and modification of cell wall polysaccharides

Predicted full-length protein sequences of different protein families were recognized according to the presence of characteristic domains in cereal genomes.Protein sequences from the genome assembly of the winter rye inbred line Lo7 were used for this analysis[41].The protein sequences of Arabidopsis thaliana,rice(Oryza sativa L. japonica group), wheat (Triticum aestivum L.), sorghum(Sorghum bicolor (L.) Moench), maize (Zea mays L.), stiff brome(Brachypodium distachyon(L.)Beauv.),and barley(Hordeum vulgare L.) were retrieved from the Phytozome v12.1 database [58](https://phytozome.jgi.doe.gov/pz/portal.html), the Ensembl Plants database (release 45) [59] (https://plants.ensembl.org/index.html) and the Uniprot database (release 2020_04) [60](https://www.uniprot.org/) using gene name search. The HMMer v3.3 (http://hmmer.org/) program was used to identify various protein family members in the rye protein sequences, using Hidden Markov Model(HMM)domain profiles obtained from the Pfam v32.0 database [61] (http://pfam.xfam.org/). The protein families were distinguished by the presence of these domains: PF03552,PF00535, PF13632–GT2, PF01501–GT8, PF03360–GT43, PF03016–GT47, PF04577–GT61, PF00933, PF01915–GH3, PF00331–GH10,PF06964–GH51, PF02458–BAHD acyltransferases, PF07779–RWA acetyltransferases, PF14416, and PF13839–TBL acetyltransferases.The GT2 protein sequences are characterized by the presence of at least four transmembrane domains and two conserved motifs:DxD and QxxRW[62].Putative members of the rye GT2 family that lacked these features were discarded from further analysis. Transmembrane domain search was conducted on the TMHMM Server v2.0[63](http://www.cbs.dtu.dk/services/TMHMM/).Finaly,BioEdit software(version 7.0.5)[64]was used for local BLAST search of rye’s nearest homologs using the OsDARX1 and OsBS1 acetylesterases as queries.TGR values for all identified genes can be found in Table S2.

2.8. Phylogenetic analysis

The protein sequences obtained from the above plant species were subjected to multiple alignment with Clustal Omega [65](https://www.ebi.ac.uk/Tools/msa/clustalo/). Maximum-likelihood phylogenetic trees were constructed with IQ-TREE (version 1.6.9)[66]. The best-fitting model of amino acid sequence evolution was selected according to the Bayesian Information Criterion(BIC), as implemented in the ModelFinder method [67] of IQTREE.Ultrafast bootstrap branch support based on n=10,000 replicates[68]was used to build each dendrogram(clade support ≥95%were significant). These trees were then visualized with iTOL v6[69] (https://itol.embl.de/) and edited with Adobe Illustrator CC 2017(version v21.0.0)[70](Adobe,San Jose,CA,USA;https://adobe.com/products/illustrator). In some trees, the locus names are replaced by compact notation given by other authors;the full locus names for each phylogenetic tree are listed in Table S3.

3. Results

3.1. Viscosity and pentosan content

Kinematic viscosity of water extracts and water-extractable pentosan content in six winter rye cultivars were determined during four consecutive seasons (Table 1).The 2017 season was characterized by adverse weather conditions at the stage of kernelfilling.All the cultivars were characterized by lower than usual viscosity of water extracts;however,their overall trends were similar to those in other seasons. Two cultivars (Tatarskaya-1 and Radon)could be characterized as high-pentosan and as forming highly viscous water extracts (Table 1). Ogonek and Marusenka, in contrast,harbored low levels of water-extractable pentosans and their water extracts had the lowest viscosities among the studied cultivars. Podarok and Pamyati Kunakbaeva showed different properties among the seasons. Correlations between these two parameters (VWE and WEP) within one season ranged from high to very high, but the viscosity of water extract was much more variable than was the content of water-extractable pentosans as measured by the orcinol-chloride method. For example, the cultivar Marusenka showed a relatively stable level (1.22%) of waterextractable pentosans in its wholemeal flour in the 2016, 2017,and 2018 seasons while its water extract viscosity varied between 4.6 and 15.3 centistokes (cSt). Likewise, a greater than threefold difference in extract viscosity between the harvests of 2017 and 2018 was found for the Tatarskaya-1 cultivar,whereas its contents of water-extractable pentosans were similar across seasons.

Table 1 Kinematic viscosity of wholemeal flour water extract(VWE,cSt),water-extractable(WEP,%of flour weight),and total pentosan content(TP,%of grain weight)from six winter rye cultivars.

The content of total pentosans (TP) measured by NIR spectroscopy for the 2019 rye harvest did not correlate significantly with either water-soluble pentosan content or water extract viscosity (Table 1). This finding is exemplified by Tatarskaya-1, characterized by high WEP and VWE values but relatively low TP,while the opposite was found for Marusenka, the cultivar with low WEP and VWE and a relatively high TP. Water-extractable pentosans constituted only a portion of total pentosans,whose degree of solubility in water varied between the cultivars (of total pentosans,19.5%were water-soluble in Tatarskaya-1 but only 15.5%were soluble in Marusenka).

3.2. Water-extractable carbohydrates of winter rye flour

The total content of water-extractable carbohydrates was similar among the six studied cultivars and accounted for approximately 12% of their wholemeal flour weight (Table 2).

The SEC (size-exclusion chromatography) analysis showed that the carbohydrates of all cultivars were distributed over six peaks with apparent Mw(weight average molecular weight)values ranging from several hundred to several million Daltons(Fig.2A).More than half of the water-soluble carbohydrates detected were oligoand monosaccharides (peaks V and VI in Table 2 and Fig. 2A, B).Compared with the other four cultivars, Marusenka and Ogonek were characterized by a lower content of high-molecular-weight carbohydrates (peaks I and II) but a higher level of extracted oligo- and monosaccharides (peaks V and VI), and both cultivars showed the lowest values for the kinematic viscosity of wholemeal flour water extracts (Table 1). A high positive correlation was observed between the content of peak II (apparent Mw of 400 kDa)and water extract viscosity,but the latter showed a high negative correlation with the content of peak VI represented by mono- and disaccharides (Table 2). The highest UV absorption of Tatarskaya-1 and Marusenka water extracts at 280 and 320 nm was observed in a retention volume corresponding to peaks I, II and VI indicating that the major part of proteins and phenolic compounds coeluted with high-molecular-weight polymers and monosaccharides (Fig. S1). Tatarskaya-1 and Marusenka did not differ from each other either in total area of peaks recorded at both wavelenghths or in protein content,which constituted around 0.5%of their wholemeal flour weight (Fig. S1).

The yield of the ethanol-precipitable high-molecular-weight carbohydrates accounted for 2.7%–3.3% (w/w) of wholemeal flour from the six rye cultivars. The content of these water-extractable polysaccharides was correlated with water extract viscosity(Table 2). Ethanol-precipitated polymers showed significantly reduced UV absorption at both wavelengths in comparison with the total water soluble fraction (Fig. S1).

Because Tatarskaya-1 and Marusenka cultivars were the most contrasting with respect to their water-soluble pentosan content,water extract viscosity, and molecular weight distribution across seasons(Tables 1,2;[45]),they were chosen for further investigation. Besides the differing amounts of high-molecular weight carbohydrates (Table 2, peak I and ethanol-precipitated carbohydrates), a pronounced difference in Mw was also evidentbetween these two cultivars(Fig.2B).The apparent Mws of peaks II and V were lower for carbohydrates extracted from Marusenka flour. According to a monosaccharide analysis of total extracted carbohydrates(Fig.2C),of polysaccharides precipitated by ethanol(Fig. 2D), and free monosaccharides of water extracts (Fig. 2E),large differences were found in the levels of arabinose, glucose and xylose+mannose.Marusenka showed a higher content of glucose than Tatarskaya-1, whereas Tatarskaya-1 showed higher amounts of arabinose and xylose+mannose in total carbohydrates than Marusenka (Fig. 2C). Both arabinose and xylose + mannose were the major constituents of ethanol-precipitated polysaccharides (Fig. 2D). The differences between Tatarskaya-1 and Marusenka in their levels of glucose and xylose + mannose were significant. The free monosaccharide fraction of these cultivars’water-extractable carbohydrates contained mostly glucose and xylose + mannose (Fig. 2E).

Table 2 The yield and molecular-weight distribution of water-extractable carbohydrates for six winter rye cultivars (n = 3), from the harvest of 2019.

Fig.2. Molecular weight distribution and monosaccharide composition of water-extractable carbohydrates obtained from wholemeal flour of winter rye.(A)Size-exclusion chromatography profiles of water-extractable carbohydrates obtained from wholemeal flour of six winter rye cultivars.(B)Size-exclusion chromatography profiles of waterextractable carbohydrates obtained from wholemeal flour of Tatarskaya-1 and Marusenka cultivars. Roman numerals designate peaks; the mean apparent Mw (kDa) is indicated above each peak.Mean Mw values for each peak in(B)in red and blue fonts are those of Tatarskaya-1 and Marusenka,respectively.Gray triangles correspond to the elution times for pullulans (Mw: 805, 348, 200, 113, 48.8, 21.7, 10, 6.2, 1.32, 0.342 kDa) and glucose (Mw: 0.18 kDa) used for the column calibration. Monosaccharide composition in mol%of total carbohydrates(C),carbohydrates precipitated by ethanol(D),and free monosaccharides(E)in rye flour extracts.Mean values are presented,with SD given in (C–E); astrisks indicate significance of differences (t-test, n = 3; * P ≤0.05, ** P ≤0.01, *** P ≤0.001).

3.3. Immunoblot assay of water-extractable carbohydrates

The labeling pattern of water-extractable carbohydrates was similar for the two cultivars differing in water extract viscosity(Fig.3A). In both,heteroxylans (AX1,LM11,LM27,and LM28 antibodies)were recognized.However,labeling with antibodies specific for MLGs(BG1),heteromannans(LM21),galactans(LM5,LM26),and polymers containing ferulic acid (LM12) was also detected(Fig. 3A).

The ethanol-precipitable part of water-soluble carbohydrates was heavily labeled by AX1 antibody (Fig. 3B). Labeling intensity for both cultivars was four times that for the control polysaccharide (wheat arabinoxylan). Labeling of the ethanol-precipitable fraction by LM27 and LM28 antibodies (Fig. 3B) indicated that water-soluble high-molecular weight polysaccharides of rye flour contained glucuronoxylans and heteroxylans. Although the exact epitope for LM27 is not yet specified, these results suggested that water-soluble high-molecular weight xylans of wholemeal rye flour contain other substitutions than (feruloyl-) arabinose.

The iodine test was performed to check for the presence of starch in the extracted polysaccharides (Fig. 3C). Even high amounts of carbohydrates used for this analysis did not show marked blue staining, as would be expected if much starch from the rye flour were solubilized by water. Thus, the glucose in ethanol-precipitated carbohydrates (Fig. 2D) belonged mainly to MLG,whose presence was confirmed by labeling of this fraction by BG1 antibody (Fig. 3B).

Fig.3. Immunoblot assay(A,B)and iodine test(C)of water-soluble carbohydrates and their ethanol-precipitable fractions extracted from wholemeal flour of two winter rye cultivars.Antibody specificities:AX1,arabinoxylan,LM11,low-substituted xylan,LM27,grass heteroxylan,LM28,glucuronoxylan,LM12,feruloylated polysaccharides,BG1,mixed-linkage glucan,LM21,heteromannan,LM26,1,6 branched 1,4-galactan,LM5,galactan.The polysaccharides used as controls are specified on the right side of(A)and(B). AX, arabinoxylan, GAX, glucuronoarabinoxylan, MLG, mixed-linkage glucan. (A) Immunoblot assay of water-soluble carbohydrates. (B) Immunoblot assay of polysaccharides precipitated by ethanol from water-soluble carbohydrates.Values in(A)and(B)are means of labeling intensity relative to positive control,obtained in three biological replicates for 4 μg of applied carbohydrates. *, the only significant difference revealed between the two cultivars was LM28 labeling of the EtOH fraction (t-test,n = 3; P ≤0.05). (C) Iodine test.Total, water-soluble carbohydrates, EtOH, polysaccharides precipitated by ethanol from water-extractable carbohydrates. The +and – signs respectively denote the positive(amylopectin)and negative(birch wood xylan)controls for iodine test.Amount of applied carbohydrates in μg is specified above the panel in(C). T, Tatarskaya-1; M, Marusenka.

3.4. Genes encoding enzymes of AX biosynthesis

In rye genome, we identified ten genes whose predicted amino acid sequences included the PF03360 domain,the signature for the GT43 protein family encoding a putative participant in AX backbone synthesis (Figs. 1, 4A). Transcripts for seven of them were found in rye kernels (Fig. 4B). In the accompanying dendrogram construction, the SECCE2Rv1G0129990 and SECCE7Rv1G0509920 genes were positioned in the same clade with AtIRX14/AtIRX14L.AtIRX9 and AtIRX9L show respectively five and three putative homologs in the rye genome (Fig. 4A). All GT43 genes showed the highest level of transcripts at the milk ripeness stage with lower TGR values at further stages of kernel development in both cultivars. Among them, transcripts of SECCE7Rv1G0479050 (AtIRX9 homolog),SECCE7Rv1G0509920,and SECCE2Rv1G0129990(AtIRX14/AtIRX14L homologs) were the most abundant.

All five rye homologs of AtIRX10/AtIRX10L (Fig. 3A) revealed in the genome showed high expression in kernels of both cultivars especially at the milk stage of kernel development (Fig. 3B).IRX10(XYS1)(Fig.1)belongs to clade D(IRX10,FRA clade)of large GT47 family.A phylogenetic tree for clade D is presented in Fig.3A and a full tree of the GT47 family is presented in Fig. S2. The SECCE3Rv1G0209270 gene showed highest sequence similarity with AtIRX10/AtIRX10L (Fig. 3A) and highest transcript level among all rye members of this clade (Fig. 3B).

The GT61 family that includes proteins responsible for xylan backbone substitution (Fig. 1) was subdivided into three clades(Fig. 5A). Clade A included rye homologs for rice OsXAT2, OsXAT3,and wheat TaXAT1 (Fig. 5A); XATs are xylan α-1,3-arabinofurano syl-transferases[26].Clade A also included an OsXAX1 gene encoding an enzyme that adds β-1,2-xylose to the α-1,3-arabinose side chain of the xylan backbone [25], as well as an OsXYXT1 gene encoding a xylosyl-transferase capable of attaching β-1,2-xylosyl side chains to the xylan backbone [27] (Fig. 1).

Based on their expression pattern, genes of clade A could be divided into two groups:the first showed a high level of transcripts only at the milk stage of kernel development and the second group at the dough and full kernel maturity stages(Fig.5B).Maximal TGR values in both cultivars were detected for the rye homolog of the wheat TaXAT1 gene (SECCE6Rv1G0408940) at the milk stage,though this gene’s transcript abundance was diminished in later stages of kernel development. The first group also included rye homologs of rice OsXAT2 (SECCE1Rv1G0050610), OsXAT3 (SECCE1Rv1G0037300), and OsXAX1 (SECCE3Rv1G0147960, SECCE2Rv1G0125830) (Fig. 5B). The second group had generally lower TGR values and included at least one rye homolog of OsXAX1(SECCE4Rv1G0245490) (Fig. 5B).

Clade B included the AtMUCI21 gene (Fig. 5A), which is known to encode a β-1,2-xylosyl-transferase, a xylan branching enzyme of Arabidopsis seed mucilage[78](Fig.1).Two rye members of this clade(SECCE6Rv1G0433010 and SECCE6Rv1G0433020)were characterized by highest TGR values at the dough stage of kernel maturity(Fig. 5B). Other rye members of clade B showed the highest transcript abundance at the milk stage of ripeness (Fig. 5B).

We identified six rye homologs of AtGUXs from GT8 (Fig. S3).These genes encode xylan-glucuronosyltransferases [79] (Fig. 1).Only two of them, SECCE3Rv1G0204760 and SECCE1Rv1G003850,displayed a significant transcript abundance with a maximum at the milk ripeness stage in both cultivars (Fig. 5B).

We recognized 15 rye members of Mitchell’s clade (Fig. 1), the latter having 10 representatives in the rice genome (OsAT1–OsAT10)with just one in Arabidopsis(Fig.S4).Eleven rye members of Mitchell’s clade were expressed in developing kernels of both cultivars (Table S2). One of them (SECCE3Rv1G0161850) belongs to the same subclade on the phylogenetic tree as SvBAHD01 from Setaria viridis and BdBAHD01 from Brachypodium distachyon(Fig. S3). Silencing of both these genes reduced arabinoxylan feruloylation in model plants [80,81]. The expression of SECCE3Rv1G0161850 in rye kernels was the highest at the dough ripeness stage in both cultivars (Fig. 5B).

Fig.4. Phylogeny and expression of rye genes encoding proteins putatively involved in the synthesis of the xylan backbone.(A)Phylogenetic trees of GT43 genes and IRX10/FRA clade (clade D) of GT47 genes for rye (red text), Arabidopsis thaliana (black), rice (green), and wheat (orange). Names of rye genes displaying no expression in kernel samples are shown in gray text.The GT43 genes of A.thaliana are named following Wu et al.[20],while those of rice follow Chiniquy et al.[71]and Lee et al.[72].The wheat gene cluster as a triad from the A, B, and D subgenomes was named according to Pellny et al. [73] and Zeng et al. [74]. Two truncated genes of wheat (TaGT43_5B and TaGT43_10A in the IWGSC genome assembly,https://www.plants.ensembl.org)were not included in this phylogenetic analysis.The GT47 gene names are given according to Wu et al. [75] for A.thaliana, and according to Chen et al. [24]and Zhang et al. [76]for rice.The wheat gene cluster as a triad from the A, B, and D subgenomes was named according to Pellny et al.[77]and Zeng et al.[74].The full tree of GT47 members is shown in Fig.S2.Numbers indicate the ultrafast bootstrap support values for some of the nodes in each tree. (B) Expression of GT43 and GT47 genes in rye kernel samples, for which values are given in TGR (total gene reads) and presented in the red-white-blue heat map.Only genes with TGR values ＞150 in at least one analyzed kernel sample are shown.Note that TaGT43-4 gene was named TaGT43_1 and the TaGT47-13 gene was named TaGT47_1 in Pellny et al.[77].At,Arabidopsis thaliana;Os,Oryza sativa;Ta,Triticum aestivum.*,only expression of clade D members of GT47 protein family is shown.TGR values for all other members of GT47 can be found in Table S2.

In Arabidopsis, the TBL protein family (Fig. 1) has 46 members distributed over five subclades (Fig. S5) [82], and those of clade IV are known to perform xylan acetylation[83].Ten rye genes joined clade IV in phylogenetic tree (Fig. S5). Eight of them were found expressed in kernels of both cultivars (Fig. 5B). The other protein family,RWA,also contributes to the acetylation of cell wall polysaccharides (Fig. 1), and has four members in the Arabidopsis genome[31].Two RWA genes in the rye genome were recognized by the presence of the PF07779 domain,and both displayed the highest transcript abundance in the dough stage of kernel development(Fig.5B).

3.5. Genes encoding enzymes of the GT2 family that synthesize MLG,cellulose, and other polysaccharides

In rye, 51 members of GT2 were identified and subjected to phylogenetic analysis (Fig. 6A). At least one rye representative was found in each of the GT2 family subgroups,and the expression of 34 genes from the GT2 family was detected in kernels of both rye cultivars (Fig. 6B).

Fig. 5. Phylogeny and expression of rye genes encoding proteins putatively mediating the substitution of xylan backbone. (A) Phylogenetic tree of GT61 genes for rye (red text) Arabidopsis thaliana (black),rice (green), and wheat(orange; only two genes are shown). Names of rye genes displaying no expression in kernel samples are shown in gray text. Tree clades of GT61 were designated according to Anders et al. [26]. Some rye genes were divided in two or three parts because they encoded excessively long protein sequences containing several signature Pfam domains. Such parts are denoted by _a, _b, and _c suffixes. The protein sequences of these portions are listed in the Table S3. Numbers indicate ultrafast bootstrap support values for some of the nodes. (B) Gene expression levels, for which values are given in TGR (total gene reads) and shown in the red-white-blue heat map. Only genes with TGR values ＞150 in at least one analyzed kernel sample are shown. The putative function of these rye genes was assigned by their homology with enzymes with known functions using phylogenetic analysis.Dendrograms for GT8,Mitchel’s clade of BAHD,and TBL families can be found in Figs.S3,S4,S5 respectively.*,only the closest rye homolog of BdBAHD01 is shown among rye members of the BAHD protein family.TGR values for all other members of BAHD and GT8 families can be found in Table S2. At, Arabidopsis thaliana; AX, arabinoxylan; Bd, Brachypodium distachyon; Os, Oryza sativa; Ta, Triticum aestivum.

In the rye GT2 family, the CesA subgroup is the largest. It contains genes involved in cellulose synthesis. CesAs are further subdivided into those involved in the formation of the primary cell wall (homologous to AtCesA1, AtCesA3, and AtCesA6 [85]) or secondary cell wall (homologous to AtCesA4, AtCesA7, AtCesA8[86]).The AtCesA2,AtCesA5,and AtCesA9 are thought to be partly redundant with AtCesA6[85,87].Ten CesA genes were identified in the rye genome,of which seven were co-localized with the primary cell wall-related CesA genes of Arabidopsis in the phylogenetic tree,while the other three were closer to secondary cell wall-related Arabidopsis CesAs (Fig. 6A). The primary wall-associated CesAs had the highest levels of transcript at the milk ripeness stage in both cultivars. The expression levels of putative secondary cell wallassociated cellulose synthases in rye kernels did not exceed 150 TGR in any of the studied samples (Table S2).

CslA members are thought to mediate heteromannan synthesis,by adding β-1,4-linked Glc or Man residues to a polysaccharide backbone [88]. Seven genes from the CslA subgroup were expressed in rye kernel samples (Table S2), but only four of them exceeded the 150 TGR threshold(Fig.6B).The SECCE2Rv1G0132950 and SECCE6Rv1G0433320 genes were expressed almost exclusively at the milk ripeness stage(Fig.6B),while the remaining two members of the CslA subgroup showed similar expression levels at all stages of kernel development in both cultivars.

Members of the CslC subgroup of GT2 family possess xyloglucan backbone synthase activity[89].Three representatives of this subfamily(Fig.6A)were expressed at all tested stages of kernel development with the maximum in the milk stage (Fig. 6B).

The enzymes that synthesize MLG are encoded by genes belonging to the CslF, CslH, and CslJ subgroups of the GT2 family([90], Fig. 1). Of 10 rye genes grouped into the CslF clade, three were transcribed in the analyzed kernel samples(Table S2)including SECCE7Rv1G0490840 transcribed at the milk stage of kernel maturity with TGR values higher than 1000 in both cultivars(Fig. 6B). The CslH subgroup harbored two rye genes (Fig. 6A),one of which was actively expressed at the dough and full kernel maturity stages but not at the milk stage (Fig. 6B).

3.6. Genes encoding enzymes that modify cell wall polysaccharides

The large glycosylhydrolase 3 (GH3) family includes five plant proteins characterized by their exo-activities. Three of them are β-xylosidase (AtBXL4) [91] or bifunctional α-arabinofuranosida ses/β-xylosidases (AtBXL1 and AtBXL3) of Arabidopsis [92,93].Two others are β-1,3;1,4-glucosidases involved in the modification of MLG: EXG1 was shown to degrade MLG in maize seedling coleoptiles [94] and ExoI did so in germinating barley seeds[95,96]. These two groups form two rather distant clades in the phylogenetic tree of the GH3 protein family (Fig. S6). In rye, we detected 23 members of this GH3 protein family,of which 17 were expressed in the kernel samples: nine genes belonging to the BXL clade and eight genes to the ExoI clade (Fig. S6; Table S2). Expression levels of rye putative β-xylosidases/α-arabinofuranosidases(BXL clade) were highest at the milk stage of kernel development(Fig. 7).

The expression patterns of putative rye β-glucosidases (ExoI clade of GH3) were more variable than those shown by BXL clade members (Fig. 7). Two genes, SECCE5Rv1G0352930 and SECCE5Rv1G0353010, showed high expression at the milk stage but low levels at the dough and full maturity stages of kernel development. Two other ExoI clade members (SECCE5Rv1G0353020 and SECCE5Rv1G0353000) showed higher TGR values in the dough and full maturity stages than in the milk stage in both cultivars(Fig. 7).

The AX structure may also be influenced by αarabinofuranosidases, which belong to the GH51 family. We identified five genes in the rye genome encoding protein sequences with the characteristic alpha-L-AF_C C-terminal domain(PF06964) (Fig. S7). All of them were expressed in kernels of both cultivars(Table S2),but SECCE5Rv1G0318280 displayed the highest TGR values especially at dough and full maturity stages (Fig. 7).This gene is homologous with barley HvAXAH1(Fig.S7),which consumes wheat arabinoxylan and 1,5-α-L-arabinopentose at a high catalytic rate[97].The transcript level of SECCE4Rv1G0230850 gene was also high but only at the milk stage of kernel maturity(Fig.7).This gene is homologous with barley HvAXAH3, which is highly expressed in barley kernel during early stages of development(3–10 days after pollination [97]).

Plant β-xylanases belong to the GH10 family whose members harbor the glyco_hydro_10(PF00331)catalytic domain.We recognized 17 genes encoding protein sequences with glyco_hydro_10 domains in the rye genome (Fig. S8). Nine of these GH10 genes were expressed in kernels of both cultivars.Among them,the highest expression values were detected for the SECCE2Rv1G0075570 gene, positioned within the same clade as AtXYN1-3 (Fig. S8). This gene’s expression was almost the same at all stages of kernel development and in both cultivars (Fig. 7). A similar level of transcript abundance was found for SECCE7Rv1G0470630 at the milk ripeness stage of development, but only in cultivar Marusenka(Fig. 7).

Endolytic degradation of MLG can be mediated by cellulases(GH9) and 1,3;1,4-β-glucan endo-hydrolases (GH17) [90]. One of the characterized members of GH9 is KORRIGAN,a cellulase participating in cellulose synthesis [98]. The rye homolog of Arabidopsis KORRIGAN2 (SECCE2Rv1G0109590) exhibited higher TGR values at the milk stage of kernel development, as did many other cell wall-associated genes (Fig. 7). In contrast, the rye homolog of AtCEL1 (SECCE4Rv1G0249630) showed an opposite dynamic of transcript level, in that its lowest TGR value was detected at the milk stage, being highest instead at the full maturity stage of kernel development in both cultivars (Fig. 7).

GH17 is one of the most populated glycosylhydrolase families of cereals.Enzymes in this family may be involved in cleavage of various polysaccharides,such as callose and MLG[99].Transcriptomic data for all rye genes encoding GH17 enzymes are presented in Table S2. Fig. 7 shows TGR values only for SECCE1Rv1G0034990,which is homologous with Hordeum vulgare EI-EII encoding βglucanases that are characterized as β-1,3;1,4-glucanases[100,101].SECCE1Rv1G0034990 was highly expressed at the dough ripeness stage, but only in the low-viscosity cultivar Marusenka(Fig. 7).

Fig.6. Phylogeny and expression of rye genes encoding proteins of the GT2 family.(A)Phylogenetic tree of GT2 family for rye(red text),Arabidopsis thaliana(black)and rice(green).Names of rye genes displaying no expression in kernel samples are shown in gray text.The A.thaliana gene names are given according to Richmond and Somerville[32]and the rice genes according to Wang et al.[84].In the CesA clade,cellulose synthases putatively involved in secondary cell wall formation are outlined in blue;the other CesA genes are thought to be involved in primary cell wall cellulose synthesis. Numbers indicate values of the ultrafast bootstrap branch support. Branches with support values lower than 95 were deleted.(B)Expression of genes encoding proteins of the GT2 family in the kernels of two rye cultivars.Gene expression levels are shown in terms of TGR (total gene reads) and displayed in the red-white-blue heat map. Rye genes with TGR values higher than 150 in at least one analyzed kernel sample are shown. At,Arabidopsis thaliana; Os, Oryza sativa.

Fig.7. Expression of rye genes encoding proteins putatively involved in glucuronoarabinoxylan and mixed-linkage glucan degradation.The red-white-blue heat map shows the levels of gene expression for each family, in terms of TGR (total gene reads). Rye genes with TGR values ＞150 in at least one analyzed kernel sample are shown. The putative function of these rye genes was assigned by their homology with enzymes with known functions using phylogenetic analysis.The dendrograms of GH3,GH51,and GH10 families can be found in Figs. S6, S7, and S8 respectively. The close rye homologs of OsDARX1 and OsBS1 acetylesterases were revealed by local BLAST search implemented in BioEdit v7.0.5.*,only closest rye homologs for genes already characterized in other plant species are shown for GH9 family.**,only closest rye homolog for HvEI and HvEII gene is shown from GH17 family.TGR values for all other members of GH17 and GH9 families can be found in Table S2.At,Arabidopsis thaliana;Hv,Hordeum vulgare; Os, Oryza sativa; Zm, Zea mays.

BS1(brittle leaf sheath1)is an esterase that cleaves acetyl moieties from the xylan backbone [102], and DARX1 (DEACETYLASE ON ARABINOSYL SIDECHAIN OF XYLAN1)specifically targets acetyl groups located on arabinosyl residues [103]. Both belong to the GDSL esterase/lipase protein family (GELP). Rye homologs of OsDARX1 showed highest expression at the milk stage of kernel development in both cultivars. Rye homologs of OsBS1 were characterized by similar TGR values at all investigated stages of kernel maturity (Fig. 7).

3.7. Comparison of gene expression between high- and low-extract viscosity cultivars

The majority of genes in the studied families showed highest expression at the milk stage of kernel development. Fig. 8 shows the genes whose expression differed significantly between the two cultivars at this stage.

Six genes encoding proteins putatively involved in AX biosynthesis showed significantly higher expression in Tatarskaya-1 than in Marusenka(Fig.8).Among them were rye homologs of AtIRX14,two homologs of AtMUCI21,and two representatives of GT61 clade A (arabinosyl and xylosyl side chains of AX). Putative rye AXferuloyl transferase was also up-regulated in Tatarskaya-1 relative to Marusenka (Fig. 8).

Along with AX-related transferases,genes encoding several GTs responsible for the synthesis of other cell wall polysaccharides were up-regulated in Tatarskaya-1 relative to Marusenka (Fig. 8).All those genes represented the GT2 family and included four cellulose synthases (CesAs),one CslD (mannan backbone biosynthesis[104]), and one CslC (xyloglucan backbone biosynthesis [98])genes.

In contrast to genes encoding proteins related to polysaccharide biosynthesis, some GHs were up-regulated in low-viscosity cultivar Marusenka relative to Tatarskaya-1 (Fig. 8). Among these GHs were rye homologs of ZmXyl9,10, HvAXAH1, and HvExoI predicted to encode endo-xylanase (GH10), α-arabinofuranosidase (GH51),and β-glucosidase (GH3) respectively.

4. Discussion

4.1. Factors affecting the viscosity of grain flour water extracts

Fig.8. Differences in the expressions of cell wall-associated genes between high-and low-extract viscosity cultivars of winter rye. The red-white-blue heat map shows the levels of gene expression at milk ripeness stage in terms of TGR (total gene reads). Yellow background in fold change column indicates genes up-regulated in Tatarskaya-1,violet background indicates genes up-regulated in Marusenka. GAX, glucuronoarabinoxylan; MLG, mixed-linkage glucan; PCW, primary cell wall. Differentially expressed representatives of GH17 and GT8 families are not shown and can be found in Table S2.

The content of water-soluble cell wall polysaccharides of grains,and especially their AX levels, are considered the major determinants of the viscosity of wholemeal flour water extracts[42,105–107]. In accordance, rye cultivars whose flour produces highly viscous water solutions have a higher content of waterextractable pentosans compared with low-extract viscosity cultivars (Table 1). However, the total amount of water-soluble carbohydrates was the same among the six Russian cultivars studied(Table 2).

Besides the polymer concentration,the viscosity of the polymer solution is directly related to the polymer’s molecular weight [8].Strong positive relationships between water extract viscosity and the chain length of soluble AXs have been reported for wheat and rye [45,105,107]. We found that water-soluble highmolecular weight carbohydrates were more pronounced in the flour of high- than of low-extract viscosity cultivars (Fig. 2A).Moreover,we found a significant difference in the apparent molecular weight of at least two SEC peaks for the contrasting cultivars Tatarskaya-1 and Marusenka (Fig. 2B).

Other factors contributing to water extract viscosity could include the structure and distribution of the AX side chains.At least one report [105] for rye claims that a high viscosity of its grain’s water extract corresponds to a high amount of unsubstituted and a low amount of disubstituted xylose residues in the AX backbone,consistent with a lower degree of AX substitution. In waterextractable polymers,the ratio between Ara and Xyl,which is often used to characterize the degree of xylan backbone substitution,was close to 1(Fig.2D).This value is quite high,given that in cereals it is half that on average [8]. Interestingly, this ratio was similar between the low and high-extract viscosity cultivars. However,Ara/Xyl ratio cannot fully capture the peculiarities in AX structure,given that,in addition to the backbone,Xyl can be present in its side chains,while Ara may originate from arabinogalactan proteins and pectins.The water-soluble heteroxylans of rye flour are more complex than usually described in the literature. This conclusion follows from their labeling with LM27,LM28,and LM12(Fig.3A).

The degree and type of backbone substitution directly affects xylan solubility [9]. High-extract viscosity rye cultivars tend to have higher proportions of water-soluble AXs in their total AXs than do low-extract viscosity cultivars [105]. Similarly, in our study, a lower level of water-soluble pentosans was detected in Marusenka than in Tatarskaya-1, despite their total pentosan content showing the opposite pattern (Table 1).

One more factor that could contribute to water extract viscosity is the presence and amount of MLG.Treatment with lichenase and β-glucanase (MLG-degrading enzymes) reduced the water extract viscosity of wholemeal flour[105,108].It is important to note that the relative proportion of AXs and MLGs in rye grain may vary widely. For example, rye kernels are often reported to have half as much MLGs as water-soluble AXs[8,108],but in certain growing seasons and specific rye cultivars, the content of MLG may exceed that of water-soluble AXs [105,109]. Higher levels of glucose were revealed in total water-soluble carbohydrates and ethanolprecipitable polymers of Marusenka than of Tatarskaya-1(Fig.2C,D).However,immunoblot assay was not sensitive enough to detect the difference in labeling of total and ethanolprecipitated water-soluble carbohydrates by the anti-MLG antibody (Fig. 3).

A factor possibly influencing extract viscosity is the activity of polysaccharide hydrolases.Upon the milling and extraction of harvested grain,enzymes that were accumulated for defense or those needed for future starch mobilization encounter substrate polysaccharides in a water medium conducive for their reaction. If wheat water extracts were not inactivated by heat,their viscosity rapidly declined at room temperature,confirming the presence of endogenous enzymatic activities [8]. Endogenous xylanase activity has been shown to correlate negatively with water extract viscosity[109]. The amount of low-molecular weight derivatives present may be considered indirect evidence of higher endogenous enzymatic activity. However, it also should be noted that dry milling per se can lead to extensive polysaccharide fragmentation, as demonstrated for wheat and rye brans [110] and maize stems[111]. For this reason, we would expect only a portion of the observed low-molecular weight carbohydrates in wholemeal water extract to be produced by endogenous enzymes. Nevertheless, low-molecular weight carbohydrates were found present in higher amounts in the extracts of low viscosity cultivars (Fig. 2A;Table 2).

Taking these findings together,the Marusenka cultivar,in comparison with Tatarskaya-1, had a lower content of water-soluble pentosans, lower content of high-molecular weight waterextractable polysaccharides,and lower apparent molecular weight of some polysaccharides, and its pentosans showed lower solubility. Further, Marusenka was characterized by a higher proportion of glucose among its water-soluble carbohydrates and ethanol precipitatable polymers (Fig. 2C, D) and heavier labeling of the latter by the glucuronoxylan-recognizing antibody LM28 (Fig. 3B) than was Tatarskaya-1.So,these two cultivars chosen for transcriptomic analysis differed in the viscosity of their wholemeal flour water extracts and numerous features of their water-soluble polysaccharides that are rooted in the machinery of polysaccharide synthesis and degradation.

4.2. Gene complexes potentially involved in the formation of cell wall polysaccharides in rye kernel

Cell wall formation occurs mainly at the early stages of kernel development, during the processes of cellularization and endosperm cell differentiation. Kernel filling with reserve polymers,mainly starch, occurs later [112]. Accordingly, most of the genes involved in the biosynthesis of cell wall polysaccharides showed maximum expression at the milk stage of kernel development(Figs. 4–6). The major enzymes responsible for different types of linkages in xylans have been reliably identified in several plant organisms (reviewed by Smith et al. [113] and Amos and Mohnen [88]). In grasses, the corresponding gene sets are well described in wheat[77],rice[30,71],Brachypodium distachyon[114], and maize [115], but no such information was available for rye[116].The now revealed rye genes encoding GTs involved in AX biosynthesis make up the full set of known components of this enzymatic machinery(Figs.4,5).Genes for each necessary enzyme family were expressed in rye kernels.They included all three components known to be involved in xylan backbone formation:IRX10(XYS1)from GT47,together with IRX9 and IRX14 from GT43(Fig.4),numerous members of GT61 and GT8 that may participate in the addition of various side chains (Fig. 5), and specific members of BAHD, TBL and RWA families that are supposed to add non-glycan substituents (Figs. 1, 5).

In wheat kernels, the most abundant transcripts for genes involved in xylan backbone formation are TaGT47_2, TaGT43_1,and TaGT43_2, which are homologous with respectively IRX10,IRX14, and IRX9 of Arabidopsis [77]. The involvement of TaGT47_2 and TaGT43_2 in AX synthesis in wheat kernels has been demonstrated [21] by the implementation of RNA interference (RNAi)constructs under endosperm-specific promoter and analysis of transgenic lines. But synthase activity was confirmed in vitro only for IRX10; the other enzymes (IRX9 and IRX14) of the xylansynthesizing complex are thought to maintain the complex and participate in xylan synthesis, albeit in an unknown way [113].Recently,the involvement of TaIRX9 in controlling the xylan backbone’s degree of polymerization was shown [73].

In rye kernels we have found five homologs of AtIRX9/9L, two homologs for AtIRX14/14L, and five homologs of AtIRX10/10L expressed at significant levels (Fig. 4). One of the rye homologs of AtIRX9/9L showed a transcript abundance at the milk stage of kernel development at least tenfold higher than that of the other isoforms (Fig. 4). For rye homologs of both AtIRX10/10L and IRX14/14L, at least two isoforms with TGR values higher than 1000 were found. These genes could be targets for the breeding and development of cultivars distinguished by high or low xylan content.

In wheat, the suppression of GT43_2 (TaIRX9) and GT47_2(TaIRX10) isoforms led to a decrease in water extract viscosity[18]. This effect was explained by reduction of both the amount and the chain length of water-soluble AXs,with the former having a greater effect [18]. Arabidopsis irx9 and irx14 knockout mutants showed lower xylan content and chain length than wild-type plants, but these effects were partially restored by overexpression of rice IRX9/IRX9L and IRX14/IRX14L genes [71]. Two irx9+OsIRX9/9L lines showed xylosyltransferase activity exceeding that of the wild type[71].These findings suggest that manipulation of IRX9, 10, and 14 genes may well become a powerful tool for modifying AX content in cereal crops. Such advanced techniques as CRISPR/Cas9 knockout, artificial microRNA knockdown, and dominant suppression via overexpression of mutated isoforms were recently applied to Brachypodium distachyon [114]and Arabidopsis [117] to reduce their xylan content.

GT61 protein family members are thought to mediate the attachment of arabinosyl and xylosyl side chains to the xylan backbone. According to our phylogenetic and transcriptomic analyses,there are several close rye homologs of those already characterized for Arabidopsis, rice, and wheat GT61 genes, all of which were found actively transcribed in rye kernels. In wheat endosperm,the GT61 gene TaXAT1 is the most actively transcribed [73] and its corresponding protein is responsible for nearly all α-1,3-linked arabinosyl substitutions of xylan [26]. Suppression of TaXAT1 reduced wheat wholemeal water extract viscosity,although to a lower extent than did suppression of TaIRX9 and TaIRX10[18].Two rye homologs of TaXAT1 were transcribed in kernels of both rye cultivars, showing TGR values ＞3400 at the milk maturity stage, and three other isoforms of rye XATs were also characterized by large TGR values (Fig. 5).This finding may reflect the redundancy of AX-arabinosyltransferases present in the rye genome or their possible involvement in the production of different AX populations; for example, in different kernel tissues.

Other characterized members of clades A and B of the GT61 family add Xyl residues, as side chains, directly to the backbone(XYXT1, [27]; MUCI21, [78]), or onto the Ara of side-chains of AXs (XAX1, [25]). The transcript level of rye genes for putative xylosyltransferases from GT61 was considerably lower than that of putative arabinosyltransferases of the same family, at least at the milk stage (Fig. 5). The TGR values for some of these putative xylosyltransferases peaked at the dough and full ripeness stages(Fig.5).It is unclear whether this transcript abundance is reflected by the translation of active enzymes and whether these transferases exert any influence on the synthesis of water-soluble AXs of rye grain.Indeed,this is true for all of the genes.Yet,it is possible that the AXs in water extracts of rye wholemeal flour harbor xylosyl residues in their side chains. In studying the Arabidopsis muci21 knock-out mutant,Voiniciuc et al.[78]suggested that xylosyl branches of xylans participate in complex interactions between cellulose and pectins in seed mucilage.It is thus plausible that differential expression of rye MUCI21 and other xylosyl transferases adding xylose to side chains of AXs shift the ratio of watersoluble to water-insoluble AXs, making them more or less enmeshed in the cell walls. It is worth bearing in mind that GT61 families are quite large in cereal plants [90]. Because some separate clades of GT61 still have no characterized members (Fig. 5A),the activity of encoded enzymes is poorly predictable. Each gene of these proteins found highly transcribed in rye kernels can make its unique contribution to the structure of AXs or another polysaccharide modulating their properties and interaction abilities.

The critical structural feature of AX in grasses is the ferulic acid attached to the C5 hydroxyl of arabinosyl side chains.Both di-and triferulate bridges may crosslink AX molecules and influence their properties [118]. The gene family that mediates this ferulic acid attachment to AXs was proposed[30]following an analysis of ESTs overrepresented in cDNA libraries of grasses relative those of dicots.In this respect,identified rice genes,named OsAT1-10,were expressed at higher levels than in Arabidopsis and had relatively low similarity with Arabidopsis homolog [119]. Downregulation or overexpression of the OsAT1 homolog in B. distachyon (BdAT1)led to a corresponding decrease or increase in the ferulic acid content of cell wall polysaccharides[80].The RNA-interference silencing of homologs of rice OsAT9 SvBAHD01 and BdBAHD01 from S.viridis and B. distachyon led to a decrease of xylan feruloylation to a different extent [81]. The rye homolog (SECCE3Rv1G0161850) of these two genes was highly expressed at three analyzed stages of kernel maturity with TGR values exceeding 1000 (Fig. 5). Total and ethanol-precipitable polysaccharides of both cultivars were labeled with the feruloyl-recognizing antibody LM12 (Fig. 3).

MLG is the second basic polysaccharide constituent of primary cell walls in wheat, rye, oat, and barley endosperm [8]. In ethanol-precipitated polymers of the water-soluble fraction, glucose constituted ca.20%(Fig.2D),and virtually all of this belonged to MLG, given that starch was absent in this fraction and other glucose-containing polymers (e.g., heteromannan) were minor constituents (Fig. 3). Rye genes encoding MLG synthases and belonging to the CslF and CslH clades were expressed with different dynamics during kernel development(Fig.6B).This situation could give rise to two different populations of MLG in rye kernels that are likely necessary at different stages.

Although low amounts of cellulose (2%–4%) are consistently reported for cell walls of grain endosperm [8], a recent study[120] of cell walls from wheat endosperm revealed a substantial level of cellulose (up to 20%), by13C NMR, methylation analysis,and various types of microscopy. Cellulose is produced by members of the CesA clade in the GT2 family[32].By relying on homology with characterized Arabidopsis members of CesA clade, we divided rye CesAs into primary and secondary cell wall-associated isoforms (Fig. 6A). CesAs associated with primary cell wall cellulose synthesis were highly expressed in rye kernels, especially in the milk stage,when secondary cell wall-associated isoforms were transcribed at low levels and showed TGR values ＜150 in all samples (Table S2). High abundances of GT2 transcripts were also found in developing wheat endosperm [77]. Both transcript and protein abundance of primary cell wall-related CesAs of maize were comparable to or even higher than those obtained for ZmIRX9, 10, and 14 and ZmCslFs in maize endosperm, according to the Maize Genome Database [121] (https://www.maizegdb.org/). Active transcription and even translation of some genes do not necessarily imply the working product and realization of a specific enzymatic reaction in vivo. However, if the cellulose level in the endosperm is indeed underestimated, our understanding of cell wall architecture in grains should be reconsidered.

Using both phylogenetic and transcriptomic analyses, we identified in the rye genome many genes encoding putative GTs responsible for the synthesis of AXs, MLGs, and cellulose as well as some other polysaccharides(Figs.4–6).We anticipate that some of these genes will serve as promising targets for manipulating AX content and properties in rye grain.

4.3. Genes differentially expressed between high- and low-extract viscosity cultivars

High- and low-extract viscosity cultivars differed in expression of several genes encoding enzymes putatively involved in AX biosynthesis (Fig. 8). Transcription of rye homologs of IRX14 and MUCI21 as well as other AX-related transferases could be crucial for the formation of water-soluble AXs, which in turn determine the viscosity of grain water extracts. Arabidopsis irx14 and muci21 mutants showed lower levels of xylans in their stems and seed mucilage than did wild-type plants [71,78]. Together with AXrelated transferases, genes encoding several enzymes from GT2 family that may be responsible for the synthesis of cellulose,mannan, and xyloglucan were up-regulated in Tatarskaya-1 relative to Marusenka (Fig. 8). This finding may indicate an overall higher intensity of cell wall component biosynthesis in the highviscosity cultivar Tatarskaya-1 than in the low-viscosity Marusenka.

As distinct from the enzymes involved in polysaccharide biosynthesis, genes for various GHs including rye homologs of ZmXyl9,10 (endo-xylanase from GH10), HvAXAH1 (αarabinofuranosidase from GH51), and HvExoI (β-glucosidase from GH3) had higher expression levels in Marusenka as compared to Tatarskaya-1.HvAXAH1 is an α-arabinofuranosidase able to cleave arabinose residues from both AXs and arabinans [97]. HvAXAH1 was first purified from germinating barley extracts [122]. Both mRNA level and protein abundance of HvAXAHs were higher in the outer layers of barley kernels than in developing endosperm,indicating that these enzymes were preparing to facilitate cell wall mobilization at germination rather than to modify existing AXs during endosperm development [97]. Arabinofuranosidase action may reduce the degree of AX substitution, providing sites for further xylanase action.

Both imbibed seeds and seedlings of cereals were characterized by the presence of endoxylanase activity [123]. The endoxylanase ZmXyl is the most abundant protein on the surface of maize pollen[124]. It corresponds to ZmXyl7, according to nomenclature introduced by Hu et al.[125](Fig.S8)and represents the closest homolog of the rye xylanase differentially expressed between cultivars.All other plant xylanases characterized to date (Arabidopsis AtXYN-1; Populus PtxtXYN10A; maize WI5; rice OsXYN1) belong to another clade of the GH10 family(Fig.S8).For all of them,roles in secondary cell wall formation and vascular system differentiation have been shown [125–128]. ZmXyl effectively degrades oat,spelt,and birchwood xylans,albeit the latter at a much lower rate,and reduces the viscosity of xylan water solutions [124]. It seems plausible that a similar enzyme of rye expressed differently in contrasting cultivars (Fig. 8) should affect their wholemeal extract viscosity.

One more representative of genes differentially expressed between two cultivars is the rye homolog of HvExoI from the GH3 family (Fig. 8). HvExoI was also purified from germinating barley seeds and characterized as an exo-glucosidase having a preference for β-1,3 and β-1,4-linked terminal glucoside residues[129]but also able to cleave β-1,6 and β-1,2 linkages [95]. HvExoI was proposed [129] to participate in the hydrolysis of MLG fragments released by GH17 HvEI and HvEII enzymes during barley seed germination [130]. Interestingly, one of the rye homologs to both of these enzymes was also up-regulated in Marusenka relative to Tatarskaya-1, although at the dough stage of kernel development(Fig. 7; Table S2).

Glycosylhydrolases differentially expressed in Marusenka and Tatarskaya-1 may be accumulated in outer layers of kernels and be released upon seed germination, as has been shown for barley arabinofuranosidases [97] and xylanase [131]. Water extraction of wholemeal flour allows these proteins to encounter their target polysaccharides in aqueous medium.The higher proportion of lowmolecular weight carbohydrates in SEC profiles of Marusenka extracts than in those of Tatarskaya-1 provides indirect evidence for the higher activity of hydrolases (Table 2; Fig. 2B).

To summarize, a high-extract viscosity cultivar was characterized by an increased expression of genes involved in non-starch polysaccharide synthesis, whereas a low-extract viscosity cultivar displayed higher expression of several genes encoding glycosylhydrolases.

5. Conclusions

The presence of characteristic domains in predicted full-length protein sequences and phylogenetic analyses allowed the identification and characterization of genes encoding proteins responsible for biosynthesis and degradation of rye non-starch polysaccharides. Transcriptome analysis performed for the first time for rye kernels revealed specific isoforms involved in the formation of its cell wall polysaccharides.Immunodot analysis and RNA-Seq experiments suggested that the side chains of rye grain waterextractable AXs are more diverse than typically thought.The presence of xylose in AX side chains,both directly attached to the backbone and added to arabinose, can be inferred from the expression of genes whose homologs encode corresponding enzymes for xylosylation. High expression of GTs involved in the synthesis of MLG and cellulose suggests that their role in cell wall architecture of rye grain may be more pronounced than has been expected.Differential expression analysis of the revealed genes in high- and lowextract viscosity cultivars points to the balance between polysaccharide synthesis and degradation as the main factor that determines the viscosity of water extracts crucial for crop applications.

CRediT authorship contribution statement

Liudmila V. Kozlova:Conceptualization, Investigation, Formal analysis,Writing–original draft,Writing–review&editing,Visualization.Alsu R.Nazipova:Investigation,Formal analysis,Writing- original draft, Writing – review & editing, Visualization.Oleg V.Gorshkov:Investigation, Formal analysis.Liliya F. Gilmullina:Investigation, Formal analysis.Olga V. Sautkina:Investigation,Formal analysis.Natalia V.Petrova:Investigation,Formal analysis.Oksana I. Trofimova:Investigation, Formal analysis.Sergey N.Ponomarev:Investigation, Formal analysis.Mira L. Ponomareva:Conceptualization, Investigation, Formal analysis, Writing – original draft,Writing– review&editing.Tatyana A.Gorshkova:Conceptualization,Writing-original draft,Writing–review&editing,Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

Work was partially supported by Russian Foundation for Basic Research with project number of i_m_17-29-08023 (Liudmila V.Kozlova, Alsu R. Nazipova, Oleg V. Gorshkov, Liliya F. Gilmullina,Natalia V. Petrova, Sergey N. Ponomarev, Mira L. Ponomareva,Tatyana A. Gorshkova). Part of work (immunodot binding assay,Olga V. Sautkina; monosaccharide analysis, Oksana I. Trofimova;viscosity of water extract determination, Liliya F. Gilmullina) was performed with financial support from the government assignment for FRC Kazan Scientific Center of RAS. We thank Dr. Polina Mikshina (KIBB FRC Kazan Scientific Center of RAS) for help and advice. We thank Prof. James C. Nelson from Kansas State University for text editing.

Appendix A. Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2021.05.008.