Developmental dynamics of stem starch accumulation in Sorghum bicolor

Abstract Sweet sorghums were identified that accumulate up to ~9% of their total stem dry weight as starch. Starch accumulated preferentially in stem pith parenchyma in close proximity to vascular bundles. Stem starch accumulated slowly between floral initiation and anthesis and more rapidly between anthesis and 43 days post‐anthesis before declining in parallel with tiller outgrowth. Genes involved in stem starch metabolism were identified through phylogenetic approaches and RNA‐seq analysis of Della stem gene expression during the starch accumulation phase of development. Genes differentially expressed in stems were identified that are involved in starch biosynthesis (i.e., AGPase SS/LS, starch synthases, starch‐branching enzymes), degradation (i.e., glucan‐water dikinase, β‐amylase, disproportionating enzyme, alpha‐glucan phosphorylase) and amyloplast sugar transport (glucose‐6‐P translocator). Transcripts encoding AGPase SS and LS subunits with plastid localization were differentially induced during stem starch accumulation indicating that ADP‐glucose for starch biosynthesis is primarily generated in stem plastids. Cytosolic heteroglucan metabolism may play a role in stem sucrose/starch accumulation because genes encoding cytosolic forms of the disproportionating enzyme and alpha‐glucan phosphorylase were induced in parallel with stem sucrose/starch accumulation. Information on the stem starch pathway obtained in this study will be useful for engineering sorghum stems with elevated starch thereby improving forage quality and the efficiency of biomass conversion to biofuels and bio‐products.

requirements for metabolism, growth, and/or seed filling are satisfied.
This allows continued photosynthesis with minimal feedback inhibition reducing the cost/damage associated with interception of light by leaves in the absence of a sink for energy utilization. The nonstructural carbohydrates that accumulate in stems provide a reservoir of sugars that enhance plant resilience and enable grain filling under adverse environmental conditions that cause inhibition of photosynthesis (Czedik-Eysenberg et al., 2016). In sorghum, stem sucrose and starch levels decline in parallel with onset of tillering post-grain maturity indicating that stem reserves also contribute to traits such as ratooning that are often associated with perennial sorghum (McKinley, Rooney, Wilkerson, & Mullet, 2016a).
Sugarcane breeders have focused on increasing the purity, concentration, and yield of sucrose in sugarcane stems (Waclawovsky, Sato, Lembke, Moore, & Souza, 2010). As a consequence of strong selection, the sucrose concentration in stems of elite sugarcane has reached an apparent plateau (~0.7 M) and recent gains in yield are correlated with increases in plant size and biomass (Waclawovsky et al., 2010;Wu & Birch, 2007). The stem sucrose concentration plateau has been attributed to constraints caused by elevated turgor associated with high sucrose concentrations in pith parenchyma cells, sucrose-mediated feedback inhibition of photosynthesis, and carbohydrate sensing and signaling mechanisms (Wu & Birch, 2007).
Sucrose concentrations in sweet sorghum are similar to sugarcane; however, as sorghum is not a large commercial source of sucrose, selection for sucrose purity has been less intense which may explain why significant levels of stem starch are observed in sweet sorghum genotypes (McKinley et al., 2016b).
Our group has been developing and designing high-biomass sorghum hybrids that can be used for forage and production of biofuels and bio-products (Mullet et al., 2014;Rooney, Blumenthal, Bean, & Mullet, 2007). Stems account for approximately 80% of this crop's harvested biomass; therefore, stem composition is an important attribute of biomass value and utilization. High stem sucrose yield is a useful trait for biofuels production because conversion of sucrose/ sugars to biofuels and bio-products is low cost and highly efficient (Gnansounou & Dauriat, 2010;Wu et al., 2010). However, stem sugars are unstable following plant harvest due to metabolic and microbial degradation, therefore, commercial production of sucrose from sugarcane stems requires efficient harvesting, transport, and processing (Wu et al., 2010). The instability of stem sugars prevents longterm storage of harvested biomass restricting mill operation to times of the year when these crops can be harvested. These logistical obstacles could be reduced if sugars that are transported to stems were used to synthesize more stable polysaccharides such as starch or mixed linkage glucans. Higher stem starch could also be useful in forage crops to increase biomass digestibility. Therefore, one motivation for this study was to characterize the dynamics of stem starch accumulation during sorghum development and to increase knowledge of this pathway to enable selection and engineering.
During periods of carbohydrate surplus, leaves shunt hexoses into the synthesis of starch and starch is often the primary storage carbohydrate in tubers and grain. Starch metabolism has been extensively studied in leaves, potato tubers, rice and maize endosperms, as well as other systems (Sonnewald & Kossmann, 2013;Stitt & Zeeman, 2012). In many dicot species, starch synthesis in heterotrophic tissues involves the import G-6-P into plastids via the glucose-6-phosphate/phosphate translocator (GPT) followed by isomerization to glucose-1-phosphate by phosphoglucomutase (PGM).
The heterotetrameric enzyme ADP-glucose pyrophosphorylase (AGPase) catalyzes the conversion of glucose-1-phosphate to ADPglucose, the committed step in starch biosynthesis (Hädrich et al., 2011). In leaves, this reaction is rendered irreversible by the action of plastidial alkaline pyrophosphatase which hydrolyzes inorganic pyrophosphate into orthophosphate (Stitt & Zeeman, 2012). In grasses, a cytosolic AGPase has been identified that is responsible for the synthesis of 95% of the ADP-glucose in the grain of maize (Denyer, Dunlap, Thorbjornsen, Keeling, & Smith, 1996). The cytosolic ADP-glucose is then imported into the plastid by the action of Brittle-1, an adenylate translocator (Shannon, Pien, Cao, & Liu, 1998). Plastid localized ADP-glucose is then utilized by starch synthases (SS) to generate starch polymers with α-1,4-glycosidic linkages (Stitt & Zeeman, 2012). Following linear polysaccharide synthesis, starch-branching enzymes (BEs) hydrolyze an oligosaccharide from the terminal end of the starch polymer just below the granule surface and attaches it deeper into the growing granule, catalyzing the formation of an α-1,6-linked branch point (Tetlow & Emes, 2014).
This action leads to the formation of semi-crystalline amylopectin and increases granule density, crystallinity, and stability, and the number of nonreducing ends available for chain elongation by starch synthase (Tetlow & Emes, 2014). The action of starch synthase in combination with glucan trimming by debranching enzymes (DBE)/ isoamylases ISA1 and ISA2 fine tunes the structure and crystallinity of the granule and its physical properties which are compatible with species specific enzymes involved in starch degradation (Geigenberger & Fernie, 2014;Lorberth, Ritte, Willmitzer, & Kossmann, 1998;Tetlow & Emes, 2014).
The synthesis of starch creates a molecule that is highly stable (Hejazi et al., 2008). Because of this stability, industrial processing and degradation of starch require high heat to gelatinize starch. The addition of heat adds thermal energy to the polysaccharide chains of the molecule increasing the physical space between them allowing for solvation followed by hydrolysis by starch hydrolytic enzymes (Biliaderis, Maurice, & Vose, 1980). As plants cannot utilize heat to disrupt granule crystallinity, destabilization in vivo is thought to be achieved by increasing the negative charge density of the granule surface through phosphorylation thereby disrupting crystallinity and separating the polymer chains which increases enzyme accessibility (Hejazi et al., 2008). Phosphorylation of the granule surface occurs initially by glucan-water dikinase (GWD) which phosphorylates the C-6 position of terminal glucose monomers, followed by phosphorylation by phosphoglucan-water dikinase (PWD) which phosphorylates the C-3 positions of different glucosyl residues (Ritte et al., 2006). Phosphorylation by PWD is strictly dependent on prior phosphorylation by GWD (Ritte et al., 2006). After the localized swelling introduced by GWD and PWD, a phosphoglucan phosphatase encoded by SEX4 subsequently removes the phosphate groups as these groups impede the accessibility of starch degradation enzymes to the granule surface (Oa et al., 2009). Hydrolysis begins once the reducing ends of the amylopectin chains are accessible to β-amylases (exo-amylases that liberate maltose). One member of the debranching enzyme family, ISA3, seems to be involved in hydrolysis of branch points facilitating degradation (Delatte et al., 2006). Additional degradation of oligosaccharides liberated from the granule is performed by the α-amylases (endo-amylases) and disproportionating enzymes (DPE1 and DPE2) (Critchley, Zeeman, Takaha, Smith, & Smith, 2001). Transport of glucose and maltose released by starch degradation from the amyloplast can occur through glucose transporters (pGlcT) and maltose exporters (MEX1) (Stitt & Zeeman, 2012).
The results of this study showed that sweet sorghum genotypes can accumulate up to 9% of their stem biomass as starch and that starch accumulates primarily in stem pith parenchyma. Stem starch accumulated slowly between floral initiation and anthesis then more The genotypes in this study were at variable stages of post-grain maturity development. Three plants per plot were harvested, avoiding the end of the plots to avoid edge effects, and these three plants were used in subsequent analyses. After harvesting, leaves and leaf sheaths were removed from the stems. A stem section spanning three internodes of which the middle internode was located at 50% of the total stem length was removed by excision at the nodes.
These internode sections were cut into smaller sections and dried in a forced air oven at 60°C for 3 days; the tissue was coarsely ground in a Wiley Mill (Thomas Scientific, Inc.) and subsequently finely ground in a Cyclone Sample Mill (Udy Corporation, Fort Collins, Colorado, USA) until the tissue exhibited the consistency of a powder.
Then, this powered tissue was used for starch quantification.

| Harvest of the sweet sorghum Della during development for NSC quantification
Della (Reg. no. CV-130, PI1566819), a sweet sorghum developed from a cross of Dale and ATx622 (Harrison & Miller, 1993), was planted at the Texas A&M field station near College Station, Texas in 2012 and 2013. Before planting, ammonium nitrate (32-0-0) was applied at a rate of 140 kg N/ha. In both years, the experimental design consisted of two ranges of 10 plots each of Della planted in adjacent regions of the field. Additionally, two boarder rows were planted on either side of the experiment to prevent edge effects.
The spacing between rows was 76 cm, and plants were thinned to 10-cm spacing at the five-leaf stage. Basal tillers were removed until anthesis when basal tillering ceased. Tiller removal was performed to maintain a planting density of one main culms per 10 cm. During the experiment, plants were irrigated as needed. Five plants were harvested from each experimental replication for a total of 10 plants harvested per time point for a total of 10 plants per sample. Plants were harvested a minimum of two feet from the end of the plot to avoid edge effects. After harvesting, leaves and leaf sheaths were removed from the stems. A three internode section, of which the middle internode was located at 50% of the total stem length, was removed by excision at the node. These internode sections were cut into smaller sections and dried in a forced air oven at 60°C for 3 days. Next, the internode tissue was coarsely ground in a Wiley Mill (Thomas Scientific, Inc.) and subsequently finely ground in a Cyclone Sample Mill (Udy Corporation, Fort Collins, Colorado, USA) until the tissue exhibited the consistency of a powder. Then, this powered tissue was used for NSC quantification.

| Della stem tissue harvest for RNA-seq
To obtain tissue for RNA-seq, Della was grown in a greenhouse in College Station during the fall of 2011. The day length was maintained at 14 hr, and the mid-day PAR of the green house was 1,100-1,200 μE/m 2 /s. Plants were germinated in five gallon pots containing Metromix potting soil. Osmocote was added to the soil at the beginning of the experiment, and additional fertilization was supplemented with Peter's nutrient solution every 30 days. The plants were watered with the use of an automated watering system to eliminate the possibility of drought stress. Leaves were numbered as they appeared insuring that internode 10 was harvested at each of the eight sampling dates. Three biological replicates were harvested in the morning on each harvest date. To harvest internode samples, leaves and leaf sheaths were removed from the stem and Internode 10 was excised at the nodes to insure that a complete internode was harvested for analysis. The internode was quickly sectioned into smaller pieces with a razor for storage and ease of processing after freezing. The internode sections were placed into a 50-ml conical tube, frozen in liquid nitrogen, and stored at −80°C until processing.
To prepare tissue for RNA extraction, the frozen tissue of each internode was ground to a fine powder using a mortar and pestle and stored in 50-ml conical tube until extraction.

| Soluble carbohydrate quantitation
To quantify the abundance of stem nonstructural carbohydrates, previously, oven dried samples from field experiments were redried MCKINLEY ET AL. | 3 overnight using a forced air oven at 60°C to insure accuracy of dry weight measurements. Next, 200 mg of finely ground biomass was weighed (±0.5 mg) using an analytical balance and transferred to a 15-ml conical tube. Alternatively, to quantify nonstructural carbohydrates in tissue that gave rise to the RNA-seq data, approximately 0.5 g of frozen, pulverized tissue was placed into a previously weighed 15-ml conical tube and lyophilized until desiccated. The tube and tissue were reweighed, and tube weight subtracted to obtain the tissue weight. Both tissue types were subjected to the following extraction protocol. Water-soluble NSCs were extracted in 10 ml of water/Na azide (200 mg/L) solution at 50°C for 48 hr with agitation. This length of incubation was experimentally determined to be optimal for this experimental set-up. This extraction time

| Starch quantification
The starch in stem biomass was quantified using the NREL Laboratory Analytical Procedure for extracting starch from solid biomass (Sluiter & Sluiter, 2005). After the soluble nonstructural carbohydrate extraction, samples were washed four times in methanol-chloroformwater extraction solvent at room temperature for 15 minutes then centrifuged at 1,640 g for 15 min to sediment structural biomass and starch granules. After the MCW washes, the samples were oven dried at 65°C for 48 hr. Next, DMSO was added to the dry samples and incubated overnight. Overnight imbibition insures that the DMSO, the gelatinization solvent, has fully penetrated the biomass and is capable of fully gelatinizing the starch which is essential to fully digest all the starch. After imbibition, the samples were incubated at 100°C for 10 min with agitation every two minutes to gelatinize the starch. Then, MOPS buffer and thermostable α-amylase were added to digest the starch to maltose at 95°C for 1 hr with agitation to insure full digestion. Subsequently, sodium acetate buffer and amyloglucosidase were added to hydrolyze the maltose to glucose at 55°C for 24 hr. To insure that starch was fully digested, digestion completion samples were included in the assay and sampled after digestion with amylase and amyloglucosidase and stained with Lugol's iodine solution. Glucose was quantified with high performance anion exchange-pulsed amperometric detection (HPAE-PAD) chromatography with the use of a Carbopac PA1 analytical column (Dionex, Sunnyvale, CA). Curve validation samples were randomly interspersed among the analytical samples during chromatography to validate the glucose, fructose, and sucrose standard curves. The percentage of starch in the dry biomass was calculated from the glucose released during digestion.

| Stereomicroscopy of internode cross sections
To visualize the location of starch in the stem by stereomicroscopy, field grown plants were harvested at anthesis and post-grain maturity. A 1-mm slice of stem tissue was excised with a razor blade from the middle of internode 11. These 1-mm internode sections were infused with Lugol's solution in a vacuum. After infusion, the internode sections were stored in this solution overnight. Excess staining solution was removed by incubating the tissue slice in water for 24 hr, changing the solvent every 6 hr. The sections were imaged under white light using a Zeiss M 2 Bio Fluorescence Combination Zoom Stereo/Compound Microscope coupled with a Zeiss AxioCam digital camera (Kramer Scientific). Photographs were captured using the Zeiss AxioVision 3.0.6 software.

| Template preparation and sequencing
RNA extraction was performed using the Trizol extraction method (https://www.mrcgene.com/rna-isolation/tri-reagent). The high salt precipitation step was included because of the potential high concentration of soluble carbohydrates in the sample. Total RNA was further purified using the RNeasy kit (Qiagen, Venlo, the Netherlands) which isolates RNAs >200 nucleotides in length. An on column DNase1 digestion was performed using the RNase-free DNase set (Qiagen, Venlo, Netherlands) to remove any contaminating DNA.
The quality of the RNA libraries was assessed using an Agilent Bioanalyzer. The mean RNA integrity number (RIN) from this quality control step was 8.9, and the minimum RIN was 8.3 indicating that the quality of the RNA libraries was high. RNA libraries were

| Transcriptome analysis
The 70 bp reads were aligned to the Sorghum bicolor V3.1 genome using the HISAT2 aligner (Daehwan, Langmead, & Salzberg, 2015;McCormick et al., 2018). Expression was quantified using the String-Tie version 1.3 software (Pertea et al., 2015). Analysis of gene expression was performed on TPM normalized data. The prepDE.py script was used to convert nucleotide coverage data from StringTie into reads that could be used by differential expression statistical packages that use conventional raw reads. Differential expression and the FDR-adjusted p-values were calculated using the edgeR package. The fold-change displayed in the tables is the absolute value of the maximum fold-change value between any two time points. The FDR-adjusted p-value represents the p-value associated with the maximum fold-change values shown in the gene tables. In some cases, zero expression at one time point, when being compared to another time point, led to a very high differential expres-

| Phylogenetic analysis of gene families to identify orthologs
Phylogenetic analysis was used increase the accuracy of gene family annotation. This approach was based on the identification of genes in other species with known previously validated function. The primary protein sequences were obtained Phytozome or NCBI. The protein sequences of these validated genes and the sorghum candidate homologs were then clustered using the CLUSTAL Omega algorithm using default settings.

| Identification of chloroplast targeting peptide sequences
Peptide sequences were obtained from Biomart, hosted by the Phytozome database. The primary transcript was used for the analysis.
The ChloroP algorithm (http://www.cbs.dtu.dk/services/ChloroP/) was used to predict the presence of chloroplast targeting sequences in peptides sequences.

| Variation in sorghum stem starch accumulation in diverse sorghum accessions
The level of starch present in stems of field grown grain and sweet sorghum genotypes was analyzed after grain maturity. A stem segment spanning three internodes located approximately

| Distribution of starch in stems
The amount of starch and soluble carbohydrates in each of the elongated internodes of Della stems was assayed at grain maturity (Figure 3). Most of the stem internodes had similar nonstructural carbohydrate concentration profiles. However, sucrose concentrations were lower (and monosaccharides somewhat higher) in internodes located near the base of the plant (internode 8), internodes just below the peduncle (internodes 16, 17), and the peduncle. Excluding the peduncle, starch accounted for >5% of the internode dry biomass of every internode at grain maturity and internodes 9-11 has the highest starch levels (~12.5% of dry weight) (Figure 3).

| Expression of genes involved in stem starch metabolism during plant development
One overall goal of this study was to identify and characterize the

| Expression of genes encoding AGPases
Two sorghum genes encoding AGPase small subunits (SSs) and four genes encoding AGPase large subunits (LSs) were identified in the sorghum genome (Supporting information Figure S1; Supporting information Table S1). Several different transcripts were produced from these genes. One gene encoding AGPase SS (Sobic.002G160400, APS1) and  Table S1). Sobic.009G245000 (APL2) was differentially expressed 47fold higher in stems post-anthesis and Sobic.002G160400 (APS1) eightfold higher (Figure 6a; Supporting information Table S1). Both of these genes were expressed at~five-tenfold lower levels in leaves compared to stems; however, expression in panicles was similar to stems (Supporting information Figure S1).
Many grasses synthesize some portion of the ADP-glucose used for starch synthesis in grain in the cytosol (Beckles, Smith, & Ap Rees, 2001). As the stem post-anthesis is a heterotrophic carbohydrate storage organ similar to grain (Emanuelsson, Nielsen, & von Heijne, 1999), an analysis was conducted to characterize the expression of genes/ transcripts involved cytosolic ADP-glucose synthesis during development (Supporting information Figures S1 and S2). Alignment and phylogenetic analysis of sorghum and maize AGPase gene homologs that are localized in the cytoplasm or plastid (Huang, Hennen-Bierwagen, & Myers, 2014) and ChoroP prediction of plastid transit sequences (Xuan et al., 2013) were used to identify sorghum transcripts that encode AGPase subunits that would localize in the cytoplasm. This analysis F I G U R E 1 A survey of the percent composition of starch in the stem of diverse sorghum genotypes grown in the field in College Station, TX and assayed after grain maturity. Genotypes represented by green bars are sweet sorghums, orange bars represent grain sorghums, and the blue bar represents a forage sorghum. Error bars represent SEM identified two transcripts (Sobic.007G101500.2, Sobic.007G101500. 3) that encode AGPase small subunits and three transcripts (Sobic. 003G230500.1/.2/.3) that encode AGPase large subunits predicted to produce proteins that localize in the cytoplasm (Supporting information Figure S1). These transcripts are expressed at lower levels in stems compared to other genes/transcripts that encode AGPase subunits with predicted plastid localization. Taken together, this indicates that a plastid localized AGPase is the primary source of ADP-glucose for starch synthesis in stem amyloplasts.

| Genes encoding stem starch synthases and starch-branching/debranching enzymes
ADP-glucose is the substrate of starch synthases, the enzyme family responsible for adding glucose monomers to the reducing ends of growing amylose and amylopectin chains (Hirose & Terao, 2004).
The starch synthase gene family consists of 10 genes in rice and maize (Supporting information Table S1). This family can be further divided into the soluble and granule bound starch synthases, with each of these groups containing unique subtypes (Supporting information Figure S3; Supporting information Table S1) (Hirose & Terao, 2004). Due to the functional complexity of the family, a phylogenetic analysis was conducted to align sorghum gene annotations with gene family members identified in rice and maize (Supporting information Figure S3) (Hirose & Terao, 2004). This analysis identified ten genes in the sorghum, the same number of genes in the rice starch synthase gene family (Supporting information Figure S3) (Hirose & Terao, 2004). Eight of the 10 genes in the family were expressed in sorghum stems. Most of these family members showed peak expression in stems post-anthesis (Supporting information Table S1)  The starch-branching enzymes and debranching enzymes tune starch granule structure to achieve specific physicochemical properties of the granule (Smith et al., 1997). Four genes encoding starch-branching enzymes were expressed in the sorghum stem with peak expression after anthesis (Supporting information Table S1). The most highly expressed gene/transcript was Sobic.006G066800.2 (max TPM = 865, 43 days after anthesis) a gene that was also differentially expressed during plant development (DE = 25-fold) (Figure 6a). Isoamylase 1-3 are starch-debranching enzymes. Isoamylases 1 and 2 are involved in starch biosynthesis (Supporting information Table S1), tuning the structure of amylopectin to facilitate crystallization (Hussain et al., 2003).
Three transcripts of ISA1 and ISA2 were expressed in stems with peak expression occurring 25 days after anthesis (Figure 6a; Table S1). The gene encoding isoamylase 3 (Sobic.002G233600) produced three transcripts that varied in abundance in stems with peak expression occurring post-anthesis (Supporting information Table S2).

| Genes involved in stem starch degradation
The expression of several genes that encode enzymes involved in starch turnover increased between floral initiation and grain maturity (Supporting information Table S2). To enhance starch degradation, the crystallinity of the starch granule is reduced through the action of GWD ( Figure 5). The expression of GWD (Sobic.010G143500.1) was induced~96-fold between floral initiation and grain maturity in parallel with the accumulation of starch in the sorghum stem (Figure 6b; Supporting information Table S2).
There were three splice variants derived from a gene expressed in the stem that encodes PWD (Supporting information Table S2). All the PWD splice variants were expressed at~10-fold lower levels compared to GWD, and only one transcript was differentially expressed (~fivefold) with peak expression post-anthesis (Table S2) Table S2).
Two genes encoding DPE1 and DPE2 were significantly upregulated in stems between floral initiation and grain maturity (Figure 6c; Supporting information Table S2). DPE2, which encodes an enzyme with predicted cytosolic localization, was more highly expressed and upregulated~10-fold between floral initiation and 43 days postanthesis. DPE1, which encodes an enzyme with predicted plastid localization, was expressed at lower levels, but also induced during this phase of plant development. Two genes encoding heteroglucan phosphorylase (PHS1, PHS2) were expressed at similar levels, and gene expression was induced~21-fold and~37-fold, respectively, with peak expression 43 days post-anthesis (Figure 6b and c; Supporting information Table S2). SbPHS1 has a predicted localization in plastids whereas SbPHS2 has a predicted localization in the cytosol (Figure 6b and c; Supporting information

| Differential expression of starch pathway genes in stems and leaves
The expression of genes involved in starch metabolism in leaves and stems was compared to see if members of gene families were regulated in an organ-specific manner. The expression of pairs of genes from the same gene family was compared using RNA isolated from leaves and stems at the same time in the morning during four stages of plant development (Figure 7 and Supporting information Figure S4). The analysis showed that expression of Sobic.  Table S3). This gene was highly expressed (max TPM = 2,689) and showed strong differential expression during development (FDR-adjusted p-value = 0.00021) (Figure 6d; Supporting information Table S3). Induction of this highly expressed gene suggests that glucose-6-phosphate is a predominant metabolite transported to and from stem amyloplasts. A gene PLASTIDIAL GLU-COSE TRANSPORTER (pGlcT) encodes a protein that transports glucose released from starch degradation out of the plastid. This gene showed increased expression in stems post-anthesis; however, the relative expression of pGlcT (max TPM = 150) was much lower than GPT2 (TPM = 2,689) (Figure 6c, Supporting information Table S3).
The plastid ATP/ADP-transporter translocates adenylates across plastid membranes thus maintaining plastidial ATP supply (Figure 6c; Supporting information Table S3) (Tjaden, Mohlmann, Kampfenkel, Henrichs, & Neuhaus, 1998). One member of this gene family was expressed in stems (max TPM = 120) although expression did not increase in parallel with starch accumulation during development (Supporting information Table S3). The gene MALTOSE EXCESS 1 (MEX1) encodes a protein that transports maltose out of plastids.
This gene was expressed at low levels in stems (max TPM = 8) and was not differentially expressed between floral initiation and 43 days post-anthesis (Supporting information Table S3).

| DISCUSSION
The overall goal of this study was to learn more about the regulation of starch accumulation in sorghum stems because increasing stem starch levels could improve biomass yield and the composition of high-biomass sorghum that is used for forage and production of biofuels and bio-products. While prior studies have noted the presence of starch in sorghum stems (Gutjahr et al., 2013;McBee, Waskom, & Creelman, 1983;McKinley et al., 2016b) (Gutjahr et al., 2013;Paul & Foyer, 2001) and improve the efficiency and lower the cost of biomass conversion to biofuels and bio-products.
Starch accumulation was highest in stem pith parenchyma cells adjacent to vascular bundles in the subepidermal region of the stem where vascular bundle density was highest. The observation that starch accumulates in this region of the stem suggests that sucrose/ sugar availability is high near vascular bundles. Starch also accumulated in pith parenchyma cells immediately adjacent to vascular bundles located in the central region of the stem. Interestingly, in the center of the stem where vascular bundle density is lower, pith parenchyma cells located furthest from vascular bundles accumulated lower levels of starch suggesting that sucrose/sugars derived from the phloem are used preferentially by pith parenchyma cells closest to the phloem and that starch accumulation in stem pith cells is responsive to sucrose availability.
Leaves are the primary source of sucrose that is used for synthesis of starch in stem pith parenchyma cells. Prior studies have examined the transport of sucrose from sorghum leaf mesophyll cells to stems (Sowder, Tarpley, Vietor, & Miller, 1997) and identified transporters that are involved in this pathway (Bihmidine, Baker, Hoffner, & Braun, 2015;Bihmidine, Julius, Dweikat, & Braun, 2016;Milne et al., 2017;Mizuno, Kasuga, & Kawahigashi, 2016). studies have shown that the cytosolic forms of these enzymes are involved in the synthesis/turnover of cytosolic heteroglucans (Fettke et al., 2005(Fettke et al., , 2009. The function of the cytosolic heteroglucans has been investigated using knock-outs/RNAi constructs targeting PHS2 and DPE2 (Chia et al., 2004;Duwenig, Steup, Willmitzer, & Kossmann, 1997;Schopper et al., 2015). Mutants that lack DPE2 are compromised in starch turnover in darkness and show a starch excess phenotype (Chia et al., 2004). In potato, reduction of PHS2 expression results in early flowering and increased sprouting, possibly indicating modification of sugar signaling (Schopper et al., 2015).
In sorghum stems, genes encoding DPE2 and PHS2 show 37-fold and 10-fold increases in expression in parallel with stem sucrose and starch accumulation, DPE and PHS may be involved in synthesis and remobilization of sucrose from the stem, where they catalyze intermediate steps between maltose metabolism and sucrose synthesis (Lu & Sharkey, 2004).

| Regulation of stem starch gene expression and accumulation
The sucrose transported to stems is utilized for many purposes that depend on growth and development. During rapid stem elongation, sucrose is used principally for cell wall biosynthesis and only low levels of sucrose and starch accumulate in elongating internodes under good growing conditions. Floral initiation reprograms the shoot apical meristem from production of nascent internodes to production of the peduncle and panicle. Nascent internodes produced prior to floral transition grow out between floral transition and anthesis. Stem growth and secondary cell wall formation continues until~7 days before anthesis maintaining a sink for sucrose transported to stems. However, during the transition from rapid stem growth/secondary cell wall formation and rapid grain filling, there is a reduction in these sink activities that is correlated with the accumulation of stem sucrose and starch. An early step in the transition of stems to becoming heterotrophic storage depots involves the downregulation of vacuolar invertase gene expression starting at floral initiation so that sucrose can accumulate in vacuoles (McKinley et al., 2016a). Stem sucrose sometimes begins to accumulate prior to anthesis but usually reaches maximal levels between anthesis and grain maturity. Some internodes accumulate as a percentage of their biomass,~10%-12% starch, and~30% sucrose. Preferential accumulation of sucrose is logical because sequestration and remobilization of sucrose are energetically less costly compared to starch. However, as starch is sequestered as a large polymer with low osmotic activity, accumulation of starch can occur in pith cells that already have high levels of vacuolar sucrose providing an additional, longer term store of sugar for grain filling or tillering (ratooning) post-anthesis.
As the induction of the starch biosynthesis pathway in stems is correlated with the accumulation of sucrose, it is likely that the expression of genes involved in stem starch metabolism is regulated by stem sucrose levels directly or indirectly. Follow on studies will be needed to determine whether sucrose/glucose-sensing pathways (i.e., T6P, SnRK1, TOR, hexokinase) regulate the expression of genes in the stem starch pathway. Additional levels of regulation that occur in chloroplasts and amyloplasts (i.e., metabolic feedback, redox) (Cho et al., 1999;Mikkelsen et al., 2005;Valerio et al., 2011) may also be important modulators of starch accumulation and turnover in stems.

| Summary
In this study the temporal dynamics and spatial distribution of starch accumulation in stems of the sweet sorghum Della was characterized. Transcriptome and phylogenetic analyses helped identify differentially expressed genes involved in stem starch metabolism and sugar transport through the outer membrane of amyloplasts. Specific gene family members involved in stem starch metabolism were differentially expressed in stems and leaves.

ACKNOWLEDG MENTS
We would like to thank Vlad Panin for use of his HPLC for carbohydrate quantification. We would also like to thank Keerti Rathore for the use his microscope. wrote the manuscript.