Genetic and environmental factors contribute to variation in cell wall composition in mature desi chickpea ( Cicer arietinum L.) cotyledons

Chickpea ( Cicer arietinum L.) is an important nutritionally rich legume crop that is con-sumed worldwide. Prior to cooking, desi chickpea seeds are most often dehulled and cleaved to release the split cotyledons, referred to as dhal. Compositional variation between desi genotypes has a significant impact on nutritional quality and downstream processing, and this has been investigated mainly in terms of starch and protein content. Studies in pulses such as bean and lupin have also implicated cell wall polysaccharides in cooking time variation, but the underlying relationship between desi chickpea cotyledon composition and cooking performance remains unclear. Here, we utilized a variety of chemical and immunohistological assays to examine details of polysaccharide composition, structure, abundance, and location within the desi chickpea cotyledon. Pectic polysaccharides were the most abundant cell wall components, and differences in monosaccharide and glycosidic linkage content suggest both environmental and genetic factors contribute to cotyledon composition. Genotype ‐ specific differences were identified in arabinan structure, pectin methylesterification, and calcium ‐ mediated pectin dimerization. These differences were replicated in distinct field sites and suggest a potentially important role for cell wall polysaccharides and their underlying regulatory machinery in the control of cooking time in chickpea.

Recently, Njoroge et al. (2014), Njoroge et al. (2015), and Njoroge et al. (2016) examined differences in pectic polysaccharides in relation to cooking time for common beans. Their results suggest that a hardto-cook variety generally had lower pectin solubility and more arabinans (suggesting higher amounts of branched pectin) with lower amounts of acetylation, but no significant difference in methylesterification compared with an easy-to-cook bean (Njoroge et al., 2014). In addition, Njoroge et al. (2016) concluded that the development of the hard-to-cook property was due to the release of Ca 2+ into the middle lamella where it cross-links low methoxyl pectin.

| Biological material and sample preparation
The 12 samples used in this study were selected from trials conducted by Pulse Breeding Australia (PBA) and Tamworth Agricultural Institute (New South Wales Department of Primary Industries) agronomy researchers in Northern NSW, Australia. They consisted of three environments (1997Spring Ridge, 1997Moree, and 2010 and seven genotypes including four cultivars (Amethyst, Norwin, Kyabra, and PBA HatTrick) and three breeding lines (Rounded isoline, Angular isoline, and a Cicer echinospermum derived line; Table S1). Two genotypes grown in both the Spring Ridge and Moree trials (Amethyst and Norwin) were selected based on differences in cooking times. The four cultivars and three breeding lines were included to examine a wider range of genetic diversity at the same site and year.
Seeds of each sample were repeatedly passed through the "pitter" component of an SK Engineering Mill (SK Engineering, India) to gently remove the seed coat and split the cotyledons, followed by aspiration, to produce dhal for investigation. Seeds and dhal were stored in sealed containers at 4°C prior to analysis.

| Cooking time determination
Cooking times of pulses are difficult to precisely quantify (Wood, 2016) but were estimated using two different methods. The first method was the tactile (finger and thumb) method APQ-102.1 (Burridge, Hensing, & Petterson, 2001;Williams, El-Haramein, Nakkoul, & Rihawi, 1988;Wood, 2016). Briefly, dhal (20 g) was placed in boiling water and a timer started. At regular time periods, dhal was withdrawn from the water and squashed between the finger and thumb. When the sample was close to being soft (i.e., cooked) 10 dhal were tested at a time, and the sample was deemed to be cooked when 90% of the dhal were soft to squash and showed no white core (Wood, 2016). If a dhal sample did not cook within 60 min, the test was stopped and the sample was labelled as "hard-to-cook." The second method was the Mattson Cooker method (Wang & Daun, 2005;Wood, 2016). This method is normally used for whole pulse seeds; however, on this occasion, the method was adapted to obtain cooking performance of dhal samples. Briefly, 25 individual dhal samples were placed in the apparatus saddles, centred with their convex side up. This orientation was preferred as the dhal was found to slip out from under the plunger more often when placed with their convex side down. A plunger (100 g weight; 2.0 mm rounded tip) was carefully positioned on the top centre of each individual dhal, before introducing the entire Mattson apparatus into a large vessel of boiling water. The time after immersion at which each plunger fell through the softening dhal was recorded as the cooking time for that individual dhal. The resulting 25 cooking time values for each dhal were then used to compare the cooking performance of each sample, such as the mean cooking time or the time taken to cook 80% or 90% of the dhal (Wood, 2016).
Both cooking methods were performed in duplicate, producing similar results. For comparative purposes, we arbitrarily classified the cooking times used in this work as slow (>40 min), medium (30-40 min), or fast (<30 min) cooking (see Table S1).

| Monosaccharide analysis by high performance liquid chromatography
To prepare the ground chickpea cotyledons for monosaccharide analysis, 2 × 20 mg aliquots of flour were weighed accurately into 2 ml tubes. One aliquot was used directly for monosaccharide analysis.
The remaining aliquot was washed with ethanol (500 μl, 70%) at 100°C for 15 min, followed by two further washes (100%) at room temperature. The ethanol washes were pooled and dried under vacuum.
Between two and four replicates of each sample were analysed.
Correlation coefficients and significant differences were calculated in R as described earlier (Wilkinson & Tucker, 2017), using one-way analysis of variance and the Tukey-Kramer test in Genstat or Student's t test in Microsoft Excel.

| Starch and cellulose analysis
Starch analysis was performed on 40 mg of flour using a scaled-down version of the Megazyme Total Starch assay (amyloglucosidase/αamylase method) for samples containing D-glucose (McCleary, Solah, & Gibson, 1994). The samples were initially washed with 80% ethanol at 85°C for 5 min, followed by a second wash with 80% ethanol at room temperature. Standards (Megazyme, 96% starch) were included with every batch and analysis was performed in duplicate. Cellulose was quantified on 75 mg of flour (duplicates) according to the Updegraff method (Updegraff, 1969).

| Glycosidic linkage analysis
To prepare alcohol insoluble residues (AIR) for glycosidic linkage analysis, 300 mg of flour was shaken three times in 10 ml of each of the following solvents, hexane and ethyl acetate (2 hr each); 80% ethanol for 8 hr; and acetone and methanol (20 min each). The samples were centrifuged at 3,000 g for 30 min after each wash, and the supernatant was discarded. The residues corresponding to the AIR preparations were vacuum dried before de-starching. Approximately 20 mg of each AIR sample was gently mixed with 0.5 ml of DMSO at 80°C for 1 hr. Another 0.5 ml of DMSO was added in each tube, and the AIR samples were successively placed at 100°C for 5 min and into a bath at 70°C. A solution of thermostable α-amylase (Megazyme, enzyme from Bacillus licheniformis, 100 U/ml in 100 mM sodium acetate buffer pH 5.0 containing 5 mM CaCl 2 ) was added (1.5 ml per tube), and the samples were gently stirred for 8 hr at 70°C. The destarched AIR residues were cooled to 50°C and 1 ml of 200 mM sodium acetate buffer pH 4.5 was added, followed by incubation with 1 ml of amyloglucosidase solution (Megazyme, 3 mg/ml in 200 mM sodium acetate buffer pH 4.5) at 50°C for 2 hr. The samples were dialysed (molecular weight cut-off 3,600 Da) against de-ionized water for 48 hr before freeze-drying. The cell wall material was subsequently precipitated in ethanol. Treatment with α-amylase revealed no significant difference in the yield of cell wall material isolated from the desi chickpea dhal samples. Uronic acids in the sample wall material (~2 mg) were converted to their 6,6-dideuterio neutral sugar counterparts using carbodi-imide activation at pH 4.75 followed by sodium borodeuteride (NaBD4) reduction at pH 7.0 (Kim & Carpita, 1992).
Glycosidic linkage analysis by methylation was performed as described in Xing et al. (2017) to produce permethylated alditol acetates. These derivatives were analysed using an Agilent 7890B/5977B GC-MS fitted with an Agilent J&W VF-23 ms GC (30 m × 0.25 mm, film thickness 0.25 μm) capillary column. Analysis was performed in duplicate.

| Immunolabelling and staining of cell walls
Immunohistochemical analysis of cell walls was performed according to Burton et al. (2010) with the exception of the immunodetection with the 2F4 antibody. Dilutions of the primary antibody (1:50) were applied to sections followed by a dilution (1:100) of the appropriate secondary antibody as listed in Table S2. For the 2F4 primary antibody, a similar method was followed using TcaS buffer (20 mM Tris-HCl, pH 8.2, 0.5 mM CaCl 2 , 150 mM NaCl) and skimmed milk for blocking, a 1:5 dilution of the primary antibody and 1:100 dilution of the secondary antibody (Guillemin et al., 2005). All images were captured using a Zeiss AxioImager M2 (Carl Zeiss, Oberkochen, Germany) equipped with an AxioCam MRm camera. All primary antibodies were from Plant Probes (Leeds, UK) and secondary antibodies from Invitrogen ™ (ThermoFisher Scientific, Australia). Appropriate negative controls were included to verify the absence of cross reactivity and eliminate false positives. Sections were also stained with the general stain Toluidine Blue (30 s with 0.01% [w/v] stain) and Pontamine Fast Scarlet (Sigma-Aldrich, Cat #: S479896; 20 min with 0.1% [w/v] stain) to detect cellulose.

| RESULTS
In this study, we aimed to identify differences in chickpea cotyledon composition, polysaccharide content, and structure that accompany differences in environment, genotype, and cooking time. Twelve desi chickpea samples were selected for analysis from different field trials (Table S1).

| Chickpea cotyledon flour is composed of diverse monosaccharides whose abundance vary in samples grown at different field sites
The sugar composition of chickpea cotyledons was analysed by acid hydrolysis and monosaccharide profiling as a first step to identify potential quantitative differences in polysaccharide content between the different samples. Four independent replicates were prepared for each of the 12 samples using total cotyledon flour (Table 1) which contributed approximately 43% (w/w) of the mass (Table 1 and Figure 1a). Much smaller amounts of arabinose, galactose, xylose, and galacturonic acid were also detected, contributing altogether~5% of the total mass (Table 1). In the soluble fraction, relatively low amounts of galactose and glucose were detected, contributing~7% of the total flour mass (Figure 1a).
Monosaccharide levels were compared to identify similar trends in abundance (Figure 1b). A significant positive correlation was identified between glucose and arabinose content, whereas galactose content appeared to vary independently of the other monosaccharides.
This may indicate that some changes in polysaccharide abundance are interrelated, even though the monosaccharides may not necessarily be derived from the same polymer.
The abundance of the three most prevalent monosaccharides was analysed to assess a putative association with cooking time variation Although details of all 12 samples are provided (Tables 1 and S3), for the remainder of this study, we focussed predominantly on   Table 2).
Detailed analysis of individual linkage types showed a large proportion of terminal arabinose (t-Ara) and 5-linked arabinose (5-Ara), consistent with the presence of branched arabinan polysaccharides (Tables 2 and 3). These may be present as branches on rhamnogalacturonan, which is part of the pectin fraction, although rhamnose (2-Rha and 2,4-Rha) was present only at low levels (Table 2). Low levels of 4-galacturonic acid (4-GalA;~1.2%) were assigned to rhamnogalacturonan based on the presence of 2,4-Rha branches, but the side chain composition is unclear, and it is also possible that the 4-linked galacturonosyl residues (4-GalA) arise from low levels of homogalacturonan (HG). Some substitution by galactan chains was also confirmed by the presence of terminal galactosyl residues (t-Gal), which may be derived from arabinogalactan proteins (AGPs), although only low levels of 3-Gal were present that are typical of arabinogalactans. The detection of 4-linked glucosyl residues (4-Glc) confirms the presence of linear 1,4-linked glucan that is likely to correspond to cellulose in these de-starched samples. This is further supported by the cellulose assays mentioned above. Some xyloglucan was also detected, as judged by the trace amounts of 4,6-glucosyl residues (4,6-Glc) and terminal xylosyl residues (t-Xyl; Tables 2 and 3). The linkage analysis is also consistent with a low abundance of heteromannan and heteroxylan in the samples (Table 3). In summary, the glycosidic linkage analysis indicates arabinan and cellulose are the main cell wall polysaccharides present within chickpea cotyledon flour.  Note. The average abundance of each linkage was determined across all Norwin and Amethyst samples and replicates. Polysaccharide composition was estimated following the protocol of Pettolino, Walsh, Fincher, and Bacic (2012). The asterisk indicates that the exact composition of several pectic polysaccharides is unclear due to a lack of data regarding the identity of side-chains on rhamnogalacturonman I.

| Genotypes showing differences in cooking time exhibit differences in the abundance of glycosidic linkages
The glycosidic linkage data were also used to identify relationships between polysaccharide compositional changes, field site, and genotypes showing differences in cooking time. First, differences between the samples were assessed in terms of site, irrespective of genotype.
Significant differences were identified in the abundance of t-Ara, 2,5-Ara, and 4-Gal (p < .05), indicating that environmental variation has an impact on chickpea polysaccharide structure. For these three linkages, levels were reduced in samples from Spring Ridge compared with Moree (Table 2). Second, differences between the samples were assessed in terms of genotype, irrespective of site. Significant hy differences were observed in the relative abundance of 2,3,5-Ara, t-Xyl, t-Gal, 4-GalA, and t-Glc residues (p < .05; Figure S1). The most abundant of these was 2,3,5-Ara, which varied from 4 to 9 Mol%, and was more abundant in the flour of cotyledons from fast-cooking Norwin compared with slow-cooking Amethyst, independent of field site (Table 2). By contrast, levels of t-Xyl were significantly higher in Amethyst compared with Norwin (similar to t-Gal, 4-GalA, and t-Glc), although overall abundance was low (varying from 1.2 to 2.8 Mol%).
These results suggest that differences in polysaccharide structure accompany differences in genotype in the two different environments tested. The 2F4 antibody, which detects dimeric association of pectic chains through calcium ions, labelled the cells in a similar pattern to LM19 ( Figure 2i). This is consistent with homogalacturonan molecules being linked together through calcium bridges, forming a robust matrix in the mature cotyledon cells. Finally, Pontamine Fast Scarlet staining was used to detect cellulose and showed an even distribution around the periphery of most cell types (Figure 2j; Figure S3). In summary, these data provide information regarding the location of polymer deposition and differences in labelling efficiency between different cell types.

| Immunolabelling confirms the presence of diverse polymers in chickpea cotyledon cell walls
These assays provide additional support for the chemical assays, suggesting the cell walls of chickpea cotyledons comprise a complex mixture of arabinan, AGPs, cellulose, and pectin that may influence the physicochemical properties of different cell types during growth and subsequent processing.

| Differences in cell wall labelling between genotypes and samples with different cooking times
The distinct patterns revealed by immunolabelling provided an opportunity to investigate specific differences in sample composition. to different antibody labelling efficiencies ( Figure S2). Samples were To consider the contribution of genotype to composition, we also investigated the Norwin and Amethyst genotypes, both of which were grown at two different field sites. As described above, these genotypes show distinct differences in cooking time, no consistent difference in monosaccharide composition, and small but significant differences in the abundance of several glycosidic linkage types, including 2,3,5-Ara (Figures 1 and S1 and Tables 2 and 3 Figure S3A-D′). In the majority of cases, labelling patterns appeared to be indistinguishable between samples. However, consistent differences were detected using 2F4 antibodies, whose labelling appeared to be more sporadic in fast-cooking Norwin compared with slow-cooking Amethyst (compare Figure 3i′,j′ to k′,l′). In addition, differences were observed with LM20 antibodies. In both Norwin samples, the use of LM20 antibodies revealed a punctate pattern in subepidermal cells but barely any labelling was detected in the epidermis (Figure 3q-r′). In contrast, the epidermis of both Amethyst samples was labelled with the LM20 antibody, but this was infrequent or lacking in the sub-epidermal cells (Figure 3s-t′). Despite no obvious difference in LM19 immunolabelling (HG; Figure 3m-p′), differences in LM20 signals may indicate changes in wall flexibility between the samples, whereas increased 2F4 labelling is potentially indicative of more calcium dimerization of non-methyl-esterified galacturonic acid blocks and potentially stiffer cell walls.  . Images are shown in the same orientation as Figure 2. Bar in g,i,k,m,o,f,h,j,l,n,p = 200 μm,in e',g',m',o',f',h',n',p',q,s,r,t = 120 μm,in i',k',j',l',q',s',r',t' = 80 μm. ep = cotyledon epidermis; MO = Moree; SR = Spring Ridge; S = slow-cooking; F = fast-cooking

| DISCUSSION
Chickpea is an important nutritionally rich legume crop that is consumed worldwide, particularly in the Indian subcontinent (FAO, 2017). Two distinct chickpea seed types are utilized for different purposes; kabuli chickpea are normally cooked and consumed whole, canned, or as hummus, whereas desi seeds are most often decorticated and cleaved to release the split cotyledons (referred to as dhal) prior to cooking (Wood & Grusak, 2007).
Here, we investigated compositional variation in desi chickpea genotypes that exhibit differences in cooking time. In particular, we focussed on the desi cultivars Norwin, Amethyst, and PBA HatTrick. The aim was to characterize the major cell wall-related polysaccharides in cotyledons, gain an understanding of their structure, identify their location within the cotyledon, and determine if any differences in deposition or abundance might associate with variation in cooking time.
The starch content of the desi chickpea dhal ranged from 44.3% to 51.5%, similar to previous studies (Wood et al., 2014a;Wood et al., 2014d;Wood & Grusak, 2007), whereas cellulose content ranged from 1.6% to 2.2%. The cellulose content of chickpea wholegrain flour has been reported to vary from 4% to 13% (Wood & Grusak, 2007), and the current results are consistent with the majority being derived from the seed coat. In addition to starch and cellulose, monosaccharide profiling indicated that other polysaccharides are present that contain arabinose, galactose, and xylose monomers. Small amounts of mannan are also likely to be present. This composition is similar to that of other legumes, including beans and lupins, where cotyledon primary cell walls contain a mixture of cellulose, arabinogalactan, arabinan, pectin, xyloglucan, and galactan (Shiga et al., 2009).
Although the cotyledon cell wall polymers make up only a small fraction (~5-6%) of the total mass compared with starch (~40-50%) and protein (~20-30%; Singh, 1985;Miao, Zhang, & Jiang, 2009;Wood et al., 2014a), studies in bean previously suggested a link between cell wall composition and differences in cooking time. For example, lower pectin solubility, higher arabinan content, and release of Ca 2+ into the middle lamella all correlate with the hard-to-cook defect (Njoroge et al., 2014). We considered several aspects of chickpea cell wall composition in terms of cooking time variation. Of these, monosaccharide abundance did not appear to be associated with differences in cooking time and tended to differ more between samples grown at different field sites. In terms of unfractionated cotyledon flour, variation in glucose content across field sites is consistent with previous studies showing the effect of environment on amylose content in pulses (Bhatty, 1988;Frimpong et al., 2009). The significant changes in arabinose content suggest a similar environmental effect on nonstarch polysaccharide composition. This is consistent with the effect of different environments on monosaccharide levels in Arabidopsis (Duruflé et al., 2017) and many other species.
Although variation in overall monosaccharide levels did not correlate with genotypic differences and cooking time, this does not exclude the possibility that cell wall polysaccharide structure might contribute to cooking-related properties. Indeed, specific glycosidic linkage types were identified that showed genotype-dependent variations in abundance. The fast-cooking cultivar Norwin showed higher levels of 2,3,5-Ara residues (likely derived from arabinan) and reduced levels of t-Xyl, t-Gal, t-Glc, and 4-GalA residues (possibly derived from mannan, arabinogalactan, rhamnogalacturonan, or homogalacturonan) relative to the slow-cooking cultivar Amethyst. Levels of these linkage types and the inferred polysaccharides were low relative to other cell wall components, but the results were consistent across multiple replicates and field sites. In the case of arabinan, increased branching may be present within the Norwin genotype. Arabinans are cell wall polysaccharides that show great structural diversity during development and between species, but in general contain a 1,5-arabinan main chain that is substituted at O-2 or O-3 by single arabinosyl residues or short side chains (Caffall & Mohnen, 2009). In bean, changes in arabinan branching are not associated with the development of the hard-to-cook property (Shiga et al., 2009). However, in some species such as apple, loss of branching in arabinans occurs in advance of the loss of firm texture (Peña & Carpita, 2004). Although increased branching appears to contrast the fast-cooking nature of Norwin compared with Amethyst, so little is known about the role of arabinans in modulating cell wall flexibility or stiffness (Verhertbruggen, Marcus, Chen, & Knox, 2013) that the effect of altered arabinan branching on chickpea cotyledon softening requires further investigation. It is interesting to note that arabinans were predominantly located to the desi cotyledon epidermis, as detected by LM6 antibody, which may suggest a cell type specific function in mechanical reinforcement.
In general, cell wall immunolabelling is a useful method to reveal the location of different cell wall polymers within complex tissues.
Although there are some caveats, particularly in regard to masking of polysaccharides by other wall polymers (Xue, Bosch, & Knox, 2013), different labelling efficiencies can also highlight important changes in polysaccharide distribution and abundance (Aditya et al., 2015;Chowdhury et al., 2014). Comparisons between the fast-cooking Norwin and slow-cooking Amethyst genotypes revealed no clear differences in the distribution of AGP, arabinan, cellulose, or HG epitopes in cotyledon sections. However, differences in LM20 (meHG) and 2F4 (Ca2 + HG) labelling were identified between Norwin and Amethyst samples, and these were conserved across different field sites. The LM20 antibody detected meHG epitopes in the epidermis of slowcooking Amethyst. This pattern was different in fast-cooking Norwin, where epitopes were sporadically detected in epidermal walls but more prevalent in sub-epidermal cotyledon cells. Different degrees of HG methylesterification impact the mechanical and physiological properties of pectin gels (Willats et al., 2001). In particular, stretches of un-methyl-esterified galacturonic acid residues may promote the formation of the so-called "egg-box" model structure through Ca 2+ cross-linking, which is assumed to induce gel formation and thus strengthen the wall (Liners, Letesson, Didembourg, & Van Cutsem, 1989). During cadmium stress in flax hypocotyls, an increase in blockwise de-esterified homogalacturonan and Ca 2+ cross-linking was detected by 2F4 immunolabelling and was proposed as a change that might oppose cell separation (Douchiche, Driouich, & Morvan, 2010). Consistent with this, the 2F4 antibody, which recognizes dimeric association of pectic chains through calcium ions, revealed a more intense and even distribution around sub-epidermal cells in slow-cooking Amethyst compared with fast-cooking Norwin. This is also consistent with studies in lentil that suggest the formation of hard-to-cook legume seeds may involve interactions among divalent cations, phytates, and pectic compounds (Galiotou-Panayotou, Kyriakidis, & Margaris, 2008). We propose a model whereby sub-epidermal cotyledon cell walls in Amethyst contain lower levels of HG methyl-esterification, thereby allowing more prevalent calcium-mediated associations between pectin molecules and formation of stronger cell walls. This may explain some differences between the slowcooking phenotype in Amethyst and fast-cooking phenotype in Norwin. In future studies, we plan to investigate these relationships in greater detail using a larger panel of cultivars showing differences in cooking time.
Taken together, the results from this study provide evidence of variation in arabinan structure, pectin methylesterification, and dimerization between fast-(Norwin) and slow-cooking (Amethyst) desi chickpea genotypes. The more prevalent pectin dimerization in Amethyst cell walls is likely to require a higher energy input (such as a longer cooking time) to weaken or break these intercellular bonds necessary for cotyledon softening. Superimposed over this, environmental factors influence multiple aspects of chickpea cotyledon polysaccharide composition, as indicated by monosaccharide abundance and immunolabelling efficiency. The specific effect of this variation on the properties of cotyledon cell walls and downstream food processing remains unclear at present, as does the genetic basis for variation in cooking time. However, these findings provide some support for theories that suggest despite being a minor component of the chickpea cotyledon, cell wall polysaccharides fulfil an important role in downstream processing-related applications.