Cassava shrunken-2 homolog MeAPL3 determines storage root starch and dry matter content and modulates storage root postharvest physiological deterioration

Key message Among the five cassava isoforms (MeAPL1–MeAPL5), MeAPL3 is responsible for determining storage root starch content. Degree of storage root postharvest physiological deterioration (PPD) is directly correlated with starch content. Abstract AGPase is heterotetramer composed of two small and two large subunits each coded by small gene families in higher plants. Studies in cassava (Manihot esculenta) identified and characterized five isoforms of Manihot esculenta ADP-glucose pyrophosphorylase large subunit (MeAPL1–MeAPL5) and employed virus induced gene silencing (VIGS) to show that MeAPL3 is the key isoform responsible for starch and dry matter accumulation in cassava storage roots. Silencing of MeAPL3 in cassava through stable transgenic lines resulted in plants displaying significant reduction in storage root starch and dry matter content (DMC) and induced a distinct phenotype associated with increased petiole/stem angle, resulting in a droopy leaf phenotype. Plants with reduced starch and DMC also displayed significantly reduced or no postharvest physiological deterioration (PPD) compared to controls and lines with high DMC and starch content. This provides strong evidence for direct relationships between starch/dry matter content and its role in PPD and canopy architecture traits in cassava. Electronic supplementary material The online version of this article (10.1007/s11103-020-00995-z) contains supplementary material, which is available to authorized users.


Introduction
Cassava (Manihot esculenta Crantz) is widely grown for its starchy storage roots across the tropics and subtropics, with an estimated annual production of over 292 million metric tonnes (MT) in 2017 (FAOSTAT, accessed 12/17/2019). Africa accounts for 61% of this production, with Nigeria being the world's largest producer at 59.4 million MT. Cassava is grown by 100 s millions of smallholder farmers as a subsistence crop. However, dependence on the crop in sub-Saharan Africa is threatened by biotic stresses, especially cassava mosaic disease (CMD), cassava brown streak disease (CBSD), cassava bacterial blight (CBB) and cassava green mites (CGM) (Bull et al. 2011;Bart and Taylor 2017). Cassava production and utilization is also limited by inherent susceptibility of storage roots to rapid postharvest deterioration after removal from the soil (Beeching et al. 1998). This commences 24-48 h after harvest and is known as postharvest physiological deterioration (PPD). Globally, postharvest losses of cassava storage roots are estimated at about 19% (Zainuddin et al. 2018), and in Africa may reach 29% (Djabou et al. 2017). PPD has been classified as primary and secondary in nature. The former is considered enzymatic, and happens early in the process, while the latter involves mainly microbial deterioration of affected storage root tissues (Booth 1976). Studies have shown association of PPD with increased respiration and water loss (Marriott et al. 1978), elevated activities of enzymes involved in wound response, production of phenols and polyphenols (Rickard 1985) and changes in sugar and starch content (Booth 1976). Oxidative burst, believed to be an early event that leads to PPD, has been causally linked to the process of cyanogenesis (Zidenga et al. 2012). Studies of transcripts (Reilly et al. 2007), the proteome (Vanderschuren et al. 2014) and metabolites (Uarrota and Maraschin 2015) have documented potential genes and pathways involved in the progression of PPD. However, details of its molecular basis and the underlying sequential biochemical steps remain elusive, making development of affordable and sustainable mitigation measures an ongoing challenge.
Cassava is mainly grown for its starch-rich storage roots, although young leaves are consumed as a nutritious vegetable in some countries (Howeler et al. 2013). Preceded only by cellulose, starch is the most abundant polymer on earth (Geigenberger 2011) and used as food, feed, biofuel and in many industrial applications (Smith 2008;Geigenberger 2011;Streb and Zeeman 2012). Starch is the predominant metabolite in cassava storage roots, accounting for 65-91% on dry weight (DW) basis (Sánchez et al. 2009). Starch is synthesized and stored in the chloroplasts of photosynthetic organs in a transitory manner, and in the longer term in amyloplasts of heterotrophic tissues such as cassava storage roots, potato tubers and cereal endosperm (Streb and Zeeman 2012). The sequential steps and genes involved in starch metabolism have been elucidated in plants (Geigenberger 2011;Streb and Zeeman 2012). Computational approaches based on known plant starch metabolism genes have been used to document cassava genes potentially involved in starch metabolism (Saithong et al. 2013;Tappiban et al. 2019). This knowledge has been exploited to tailor the type and quality of starch produced in cassava storage roots (Raemakers et al. 2005;Zeeman et al. 2010;Bull et al. 2018;Wang et al. 2018). The first committed step in starch biosynthesis is catalyzed ADP-glucose pyrophosphoryalase (AGPase). AGPase is a heterotetramer, composed of two large and two small subunits in higher plants (Copeland and Preiss 1981). These enzymes are coded by a small gene family of two to three genes for the small subunit, and three to five genes for the large subunit in most species studied to date (Ballicora et al. 1998(Ballicora et al. , 2005Crevillén et al. 2003). The small subunit plays a catalytic role, while the large subunit plays both catalytic and regulatory roles. The catalytic role of the large subunit was demonstrated in rice large subunit 2 (Tuncel et al. 2014), and in A. thaliana, APL1 and APL2. This catalytic function has been lost in APL3 and APL4 (Ventriglia et al. 2008). The different isoforms of APL genes differ in their kinetic and regulatory properties conferred to the hetetrotetrameric enzyme. It has also been hypothesized that different hetrotetramers are formed in various plant tissues depending on tissue specific expression patterns of the different isoforms, thereby controlling the rate of starch synthesis in these organs (Crevillén et al. 2003).
The importance of AGPase in starch biosynthesis has been described through characterization of mutants. For example, the maize brittle-2 (bt2) and shrunken-2 (sh2) are mutants for the small and large subunits, respectively (Bhave et al. 1990;Bae et al. 1990). Rice mutants osagps2-1 and osagpl2-1 lack isoforms of the small (S2b) and large (L2) subunits (Lee et al. 2007). Activated glucose (ADP-glucose) is the substrate used in starch biosynthesis, with synthesis of ADP-glucose from glucose-1-phosphate and ATP by AGPase considered to be the rate limiting step (Preiss 1984;Stark et al. 1992). The catalytic activities of AGPase are under allosteric regulation, whereby the net catalytic activity is increased by activator 3-phosphoglycerate (3-PGA) and inhibited by inorganic phosphate (Pi) (Ballicora et al. 2004). Other mechanisms including redox, phosphorylation and transcriptional regulation of the tetramer have also been described (Geigenberger 2011).
In cassava, homologs of bt2 and sh2 have been cloned and their expression pattern determined by northern blot analysis in different tissues (Munyikwa et al. 2001). Presence of multiple isoforms of APL genes in cassava is expected (Saithong et al. 2013;Dong et al. 2019;Tappiban et al. 2019), and would be consistent with monocot and dicot plant species (Ballicora et al. 2004). In the present study, we identified five isoforms of cassava APL genes (MeAPL1-MeAPL5) from the cassava genome and characterized their functional significance in starch biosynthesis in storage roots using virus induced gene silencing (VIGS). Data is presented to show that among the five cassava APL isoforms, MeAPL3 is critical for starch accumulation in cassava storage roots. Study of stable transgenic MeAPL3 co-suppression lines displaying varying levels of storage root starch/dry matter content provide evidence for a strong and direct relationship between starch, dry matter content and cassava PPD. Reduction in starch and dry matter content was also associated with changes in plant architecture, notably petiole leaf angle.
Virus induced gene silencing (VIGS) vectors were generated by modification of the DNA-A component of East African cassava mosaic virus (EACMV-K201) (Beyene et al. 2017b). VIGS constructs targeting the five MeAPL genes were generated using 451-452 bp products amplified from cDNA (see section 'RNA extraction, RT-PCR and Real-time quantitative PCR'). Forward and reverse primer pairs (Supplementary Table 1) used to amplify the respective MeAPL genes carried the Nhe I and Sbf I restriction sites. Amplified PCR products were cloned into pCR-Blunt II-TOPO vector (Invitrogen) and verified by sequencing. DNA fragments were digested with Nhe I and Sbf I from the TOPO cloning vector and sub-cloned into Nhe I and Sbf I digested EACMV-K201 based VIGS vector as described by Beyene et al. (2017b). Cloned VIGS vectors were named MeAPL1-VIGS, MeAPL2-VIGS, MeAPL3-VIGS, MeAPL4-VIGS and MeAPL5-VIGS, respectively, after the corresponding genes targeted. In addition, a VIGS-vector (Patatin-VIGS) targeting the 421 bp 3′-region of the patatin promoter driving expression of both crtB and DXS transgenes in pEC20 lines (Beyene et al. 2018) was amplified by introducing restriction sites Nhe I and Sbf I as described above, and sub-cloned into the infectious EACMV-K201 based VIGS vector.
A modified pCAMBIA2300 binary vector was used for production of stable transgenic cassava plants. Modifications included replacement of the nptII plant selectable marker gene with hygromycin phosphotransferase (hptII) driven by the Agrobacterium tumefaciens nopaline synthase promoter (NOS), with the resulting vector named p8384. The p8384 binary vector was used directly (empty-vector control) or after addition of an expression cassette of MeAPL3 under control of the Cauliflower mosaic virus promoter (CaMV 35S) (Kay et al. 1987). The MeAPL3 coding sequence was amplified from cDNA prepared from leaves of cassava variety TME 7S (Beyene et al. 2017b) using primers shown in Supplementary Table 1, with Eco RI and Bam HI sites introduced for subsequent sub-cloning. The amplified MeAPL3 coding sequence was cloned into pCR-Blunt II-TOPO vector (Invitrogen) and verified by sequencing. The expression cassette harboring CaMV35S::MeAPL3::35S-terminator was assembled in an intermediate vector and moved to p8384. The resulting binary vectors named p8388 (CaMV 35S::MeAPL3::35S-terminator/p8384) and the empty vector control p8384 (NOS::HPTII::NOS-terminator) were electroporated into A. tumefaciens strain LBA4404 and used to transform friable embryogenic callus of the CMDsusceptible cassava cultivar TME 7S. Stable transgenic cassava plants were generated following Chauhan et al. (2015). Recovered transgenic plants were characterized for presence of the transgene by PCR using primers described in Supplementary Table 1, and positive lines propagated as in vitro plantlets and established in the greenhouse (Taylor et al. 2012).

RNA extraction, RT-PCR and real-time quantitative PCR
Level of VIGS-mediated suppression of each targeted MeAPL gene and GFP control was quantified by RT-qPCR. Total RNA was extracted from leaves and storage root samples collected from plants after 16-17 weeks growth in the greenhouse. Fully expanded leaves were collected from the third through fourth positions below the shoot-tip and frozen immediately in liquid nitrogen. Storage root samples were obtained by cleaning harvested roots in running tap water, slicing transversely and peeling. Samples were placed in 50 mL Falcon tubes and immediately flash frozen in liquid nitrogen. Samples were then freeze dried and used for total RNA extraction (Beyene et al. 2018). Total RNA was extracted from ~ 50 mg of fresh frozen leaf or freezedried storage root powder using the cetyltrimethylammonium bromide (CTAB) method (Doyl and Doyle 1990). Genomic DNA was removed using the TURBO DNA-free Kit (Ambion) and synthesized cDNA used as template for RT-PCR and RT-qPCR. RT-qPCR analysis was performed following Ogwok et al. (2015). The cassava serine/threonine-protein phosphatase 2A (PP2A) housekeeping gene was used for normalization of expression values (Moreno et al. 2011). RT-qPCR primers used for quantification of targeted MeAPL genes are listed in Supplementary Table 1.

Phenotypic assessment of greenhouse-grown plants
Leaf angle was measured at the junction of petiole and stem using a protractor 8-10 weeks after transfer to the greenhouse. Leaf/petiole angle was assessed on three leaves at positions 8, 9 and 10 nodes counted downwards from the top-most leaf. Three to four plants per transgenic line were assesssed. At harvest, 16-17 weeks after planting, data on storage root number and storage root fresh weight was collected.

Dry matter, starch, sucrose and glucose determination
Storage root dry matter content was determined as described by Beyene et al. (2018). Free sugars (glucose, fructose, sucrose, lactose, maltose and raffinose) (Kakehi and Honda 1989) and starch (AACC 2000) determinations were performed at the Agricultural Experiment Station Chemical Laboratories (AESCL), University of Missouri-Columbia, USA, using approximately 200 mg of freeze-dried cassava storage root flour.

Iodine staining of storage root
Iodine staining of cassava storage roots was performed using Lugol's (iodine-potassium iodide, IKI) staining solution. Harvested storage roots were cut transversely into 1-2 cm thick sections. Sections were incubated in Lugol's solution in a Petri dish for one minute and then rinsed with distilled water. Sections were tapped dry with a paper towel. Images were captured using an iPhone6 Plus.

PPD determination
Greenhouse grown cassava plants were removed from pots 16 to 17 weeks after planting and storage roots were harvested by cutting with pruning shears at the stem/root junction. Storage roots were washed clean of soil debris under running tap water and blotted dry, making sure not to inflict wounds or bruises. Roots were placed in paper bags and stored at 25 °C day/night temperature and RH of 60% in the dark. Assessment for PPD was performed by slicing storage roots transversely at proximal, middle and distal ends to produce approximately 2 cm thick segments along the root length. PPD was assessed visually by assigning a score of 0-100% as described by Salcedo and Siritunga (2011).
To confirm that MeAPL5 is coded by a single loci, cDNA covering the coding sequence and corresponding genomic region was amplified and sequenced from cassava variety TME 7S. This sequence has been deposited under GenBank MN734216 for cDNA and MN734217 for genomic DNA to represent MeAPL5. Coding sequences of the five MeAPL genes share 62-89% identity at the nucleotide level and 58-90% deduced amino acid sequence identity (data not shown). All MeAPL genes identified have conserved domains shared by APL genes of other plant species, as determined by searching the conserved domain database (CDD) available at NCBI (https ://www.ncbi.nlm. nih.gov/Struc ture/cdd/wrpsb .cgi) (Marchler-Bauer et al. 2014) and reported by Dong et al. (2019). Alignment of the deduced amino acid sequences of the five cassava APL genes with APL from other plant species revealed conserved amino acid residues Arginine (R), Lysine (K) or Glutamine (Q) at position 102, and K or Threonine (T) at position 112 (numbering based on Arabidopsis APL1 amino acid) that are responsible for catalysis (Fig. 1a), and conserved amino acid K at position 271 ( Fig. 1b) within Glucose-1-phophate (Glc-1-P) binding sites (Ballicora et al. 2005;Ventriglia et al. 2008).
Preferential expression of the five MeAPL genes in fibrous roots, storage roots, stems and leaves was assessed using the cassava ATLAS (https ://shiny .danfo rthce nter.org/cassa va_atlas ) (Wilson et al. 2017). Expression of MeAPL1 was minimal (0.2-3.3 FPKM) in fibrous roots, storage roots, stem and leaf tissues, while MeAPL2, MeAPL4 and MeAPL5 are preferentially expressed in leaves, with minimal or no expression in storage roots (Fig. 1c). MeAPL3 was clearly the most expressed isoform in cassava storage roots, fibrous roots and stems, and found abundantly expressed at 10-times higher levels (612 FPKM) in storage root tissues (Fig. 1c).

EACMV-VIGS is efficacious for suppressing gene expression in cassava storage roots
We previously showed functionality of VIGS in cassava based on use of a virulent infectious clone of the East African cassava mosaic virus (EACMV-K201) (Beyene et al. 2017b). VIGS-mediated gene silencing in shoot tissues was achieved for cassava phytoene synthase (MePSY) (Beyene et al. 2017b) and cassava SPINDLY (MeSPY), generating a visible phenotype on leaves and shoot-tips respectively (Beyene et al. 2017b). In order to demonstrate the ability of EACMV-based VIGS to silence genes in cassava storage roots, we utilized cassava plants accumulating provitamin A resulting from co-expression of DXS and crtB transgenes (Beyene et al. 2018). In these plants expression of both transgenes is driven by the predominantly storage organ-specific patatin promoter, resulting in orange coloration of storage root parenchyma, elevated ß-carotene and concomitant reduction in dry matter content (Beyene et al. 2018). Silencing DXS and crtB by targeting the patatin promoter using Patatin-VIGS in transgenic plant lines EC20-08 and EC20-11 abolished accumulation of carotenoids in storage roots (Fig. 2). Storage roots of the two transgenic lines inoculated with the control vector, GFP-VIGS, which has no target in the cassava genome, remained orange colored, while storage roots of the two transgenic lines challenged with Patatin-VIGS were white in color, in a manner similar to the non-transgenic wild-type control (Fig. 2c). Patain-VIGS-mediated suppression of transgene expression was further supported by measurement of total carotenoids and Fig. 2 Functionality of East African cassava mosaic virus (EACMV-K201) based virus induced gene silencing (VIGS) in cassava storage roots. a Dry matter, b total carotenoid content and c transverse slices of transgenic storage roots harvested from EC20 plants and wild-type TME 7S after challenge with GFP-VIGS and Patatin-VIGS construct. EC20-08 and EC20-11 transgenic events co-express crtB and DXS transgenes, each driven by the patatin promoter (Beyene et al. 2018). Silencing of crtB and DXS expression resulting from targeting the patatine promoter with Patatin-VIGS abolishes carotenoid accumulation in storage roots (b, c) and restores DMC to wild-type levels (a), GFP-VIGS has no target in the cassava genome and does not affect carotenoid accumulation or DMC. Bars show SD (n = 4-6) dry matter content (DMC) (Fig. 2a, b). Silencing of crtB and DXS by targeting the patatin promoter reduced total carotenoids from 73 and 66 ppm to 2.0 and 2.4 ppm, respectively, in plants of EC20-08 and EC20-11, a level comparable to that seen in non-modified wild-type TME 7S (Fig. 2b). As expected, VIGS-mediated silencing of crtB and DXS also resulted in elevation of storage root DMC to wild-type levels, increasing from 23.0% to 40.7% in plants of EC20-08, and from 28.6 to 39.0% in EC20-11. The DMC of wild-type TME 7S remained unchanged at 40% in plants challenged with GFP-VIGS and Patatin-VIGS (Fig. 2a). With effectiveness of the EACMV-K201 based VIGS shown for functional studies of gene expression in cassava storage roots, this technique was employed to target and downregulate MeAPL genes in cassava plants.

MeAPL3 is critical for starch and dry matter accumulation in cassava storage roots
In order to determine the relative importance of each MeAPL gene in starch biosynthesis, the five MeAPL genes (MeAPL1-MeAPL5) were individually targeted for silencing using VIGS. VIGS constructs were generated to target each of the five family members. Four-six-week-old wild-type plants of the CMD susceptible cultivar TME 7S were inoculated with the respective VIGS vectors and grown in the greenhouse for an additional 12 weeks, after which the storage roots were harvested and assessed for DMC, starch and soluble sugar concentrations. CMD symptoms became apparent on leaves 7-15 days after inoculation. These remained mild to moderate (score 1.5-3.0 on a scale of 0-5) on all plants and did not compromise plant growth or development (Beyene et al. 2017b). To confirm that MeAPL genes were downregulated by VIGS, RT-qPCR analysis was performed for each of the MeAPL target genes and compared with the GFP-VIGS control in storage roots and leaf tissues 12 weeks after inoculation. Data presented shows a 40-78% reduction of expression in leaves for all five targeted MeAPL genes in inoculated plants compared to inoculated GFP-VIGS controls. In storage roots expression of MeAPL1, MeAPL3 and MeAPL5 was reduced by 66%, 80% and 81%, respectively ( Fig. 3a-e). Expression of MeAPL2 and MeAPL4 was not detected in cassava storage roots by RT-qPCR assays, and was therefore consistent with data extracted from the cassava atlas (Wilson et al. 2017) (Fig. 1c).
Storage roots harvested from VIGS inoculated plants were analyzed for DMC, starch and free sugars. A significant (p < 0.001) reduction from 41.6% to 30% (28% reduction) in DMC was observed in storage roots of MeAPL3-VIGS plants compared with the GFP-VIGS control (Fig. 4a). Likewise, starch content in storage roots infected with MeAPL3-VIGS was significantly (p < 0.01) reduced by 33% (Fig. 4b), and soluble sugars, glucose and sucrose ( Fig. 4c, d) significantly (p < 0.01) elevated by 200-300% compared to GFP-VIGS controls. No significant differences were observed for DMC, starch or soluble sugars between the GFP-VIGS control and plants infected with the other three MeAPL-VIGS clones (Fig. 4a-d). Similarly, no significant differences were observed for DMC and starch content in plants inoculated with MeAPL5-VIGS and GFP-VIGS control, although soluble sugars were slightly elevated in MeAPL5-VIGS infected plants (Supplementary Fig. 1). This data indicates that MeAPL3 is the predominant MeAPL gene involved in starch biosynthesis in cassava storage roots.

Transgenic co-suppression lines of MeAPL3 have increased storage root number and fresh weight
Transgenic plants were generated in which MeAPL3 was co-suppressed by constitutively overexpressing the MeAPL3 coding sequence driven by the duplicated CaMV 35S promoter. A total of 26 independent transgenic lines were recovered and confirmed to be PCR positive for presence of the MeAPL3 transgene (data not shown). Additionally, five transgenic lines were recovered using the empty vector construct p8384. Nine MeAPL3 transgenic lines, plus empty-vector and non-transgenic wild-type TME 7S control plants were established in the greenhouse. RT-qPCR analysis indicated that most p8388 lines selected for characterization in greenhouse showed suppression of MeAPL3 transgene and endogenous MeAPL3 mRNA expression, indicating characteristics of co-suppression, also known as posttranscriptional gene silencing or RNA interference (RNAi) (Supplementary Fig. 2). Six p8388 lines showed significant reduction in endogenous MeAPL3 mRNA expression, while two lines expressed endogenous MeAPL3. Three transgenic lines 8388-01, 8388-2 and 8388-4 showed expression of transgene-derived MeAPL3, and accumulation of MeAPL3 derived siRNA ( Supplementary Fig. 2C). Transgenic lines 8388-06 and 8388-09 did not show expression of the MeAPL3 transgene, endogenous MeAPL3 or siRNA expression derived from these sequences. Expression of endogenous MeAPL3 was not affected in plants transgenic for the empty vector control and was comparable to that in wild-type TME 7 ( Supplementary Fig. 2).

Transgenic co-suppression lines of MeAPL3 have reduced starch, dry matter content and increased soluble sugars and modified leaf angle
At harvest 17 weeks after planting, storage roots from seven of the nine co-suppression lines showed a significant reduction in DMC, reaching only 14-27%, (a 32-65% reduction) compared to wild-type and empty vector controls at 40% DMC (Fig. 6a). A distinct phenotype was observed associated with petiole angle in MeAPL3 co-suppressed transgenic plants. Petioles of control plants were held close to the horizontal. In contrast, in MeAPL3 suppressed plants the petioles of mid-and lower-canopy leaves were held downwards. Measurement of petiole angles on four low DMC lines and  (Fig. 6b, c). In all cases the petiole/lamina junction adapted to cause the leaf surface to be held horizontally (Fig. 6c). Measurement of starch and soluble sugars on MeAPL3 co-suppression lines 8388-02, 8388-04 and 8388-10 showed highly significant reduction in starch content of 23%, 26% and 28%, respectively, as compared to starch content of the wild-type and the empty vector controls each at 72% (Fig. 7a). This was consistent with reduction in DMC observed in these lines (Fig. 6a). Storage root glucose content was elevated by 200%, sucrose by 400-500% and raffinose by 65-100% in these lines compared to controls (Fig. 7c, d). Staining of storage root sections with Lugol's IKI solution confirmed reductions in starch content, with controls staining noticeably darker than MeAPL3 co-suppressed lines (Fig. 7e). CMD susceptible plants of cassava variety TME 7S were challenged with VIGS vectors and plants harvested 10 weeks later. Bars are SD of five biological replicates per line; ** and *** stand for significant difference, respectively, at p ≤ 0.01 and p ≤ 0.001. Student's t-test compared to the GFP-VIGS control

MeAPL3 co-suppression lines have reduced PPD
Storage root PPD was assessed in MeAPL3-co-supression and control lines after three days of storage at 25 °C and 60% RH. Visual scoring for PPD was carried out as described by Salcedo and Siritunga (2011) on transverse slices obtained from proximal, middle and distal portions of the storage root. Data indicate that lines with reduced starch and DMC had no, or very low, incidence of PPD, while transgenic lines with high starch and DMC had very high PPD incidence (Fig. 8). PPD incidence was significantly (p < 0.01) higher at 87-100% in high DMC/high starch lines including wild-type TME 7S and transgenic lines 8384 (empty vector control), 8388-05 and 8388-11. In contrast, PPD incidence was lower at 25-33% in lines 8388-06, 8388-10 and 8388-12 and very low at 0-3% in the remaining three transgenic Fig. 5 Effect of silencing of cassava MeAPL3 on storage root production. a Cassava storage root number, b storage root yield, c images of storage roots from different transgenic lines, empty-vector and wild-type TME 7S controls. Transgenic lines were generated using p8384 an empty vector control (p8384) and p8388 that expresses MeAPL3 (sense). Plant were harvested 17 weeks after planting in the greenhouse, assessed for storage root number and yield. Bars are SD of five biological replicates per line; * and ** stand for significant difference, respectively, at p ≤ 0.05 and p ≤ 0.01. Student's t-test compared to the wild-type TME 7S control MeAPL3 co-suppression lines at the end of the three day assessment period (Fig. 8).

Discussion
In this study, functional characterization of five MeAPL genes that code for the large subunit of ADP-glucose pyrophosphorylase is presented, showing their preferential expression in cassava leaves and storage roots. Silencing each of these genes individually using VIGS showed that starch synthesis and accumulation in cassava storage roots is facilitated very largely by MeAPL3. This isoform is by far the most abundantly expressed isoform in cassava storage roots, fibrous roots and stems compared to the other MeAPL isoforms identified in this study. By generating stable transgenic co-suppression lines of MeAPL3, we showed that DMC and starch were significantly reduced (by up to 65%), thereby substantiating data generated by VIGS. We also demonstrated that transgenic cassava lines with reduced starch and DMC had low incidence of PPD, showing a strong, direct relationship between starch, DMC and incidence of PPD. We also further showed that transgenic MeAPL3 co-suppression lines with reduced starch and DMC presented a droopy leaf phenotype characterized by increased leaf angle. Fig. 6 Effect of silencing MeAPL3 on DMC and plant phenotype. a DMC b leaf angle in selected transgenic lines and wild-type control c plant shows partial petiole drooping of lower leaves in transgenic line 8388-01 and severe drooping in 8388-04 compared to petiole angles of wild-type TME 7S. Transgenic lines were generated using an empty vector control (p8384) and p8388 that express MeAPL3 (sense). Plants were harvested from the greenhouse and assessed for DMC 17 weeks after planting. Average petiole angle of the middle canopy per plant was determined 10 weeks after planting by measuring leaf angle from position 8th, 9th and 10th counted downwards from the top-most leaf. Four plants per line was evaluated. Bars in (a) show SD of five biological replicates and Bars in (b) show SD of four biological replicates per lines. *, ** and *** stand for significant difference, respectively, at p ≤ 0.05, p ≤ 0.01 and p ≤ 0.001. Student's t-test compared to the wild-type TME 7S control Dong et al. (2019) recently described gene structures of cassava AGPase genes that included four large and three small subunits and their conserved domains. The MeAPL3 identified and characterized in the present study is most likely the same as cassava shrunken-2 (sh2) (Munyikwa et al. 2001), the only differences being (1) the deduced amino acid sequence described here is 527 aa residues compared to 531 by Munyikwa et al. (2001) or 528 aa by Dong et al. (2019), and (2) abundance of expression of this gene is higher in storage roots than in leaves, as shown in Fig. 3c and expression data obtained from cassava Atlas (Fig. 1c) (Wilson et al. 2017). We identified at least five loci that code for MeAPL genes in the v6.1 cassava genome. Four of these genes had full gene structure and ORF and are annotated as such (Dong et al. 2019). The MeAPL5 (GenBank MN734216) identified here is represented by two partial, but overlapping sequences, one on chromosome 18 with the remaining partial sequence on an un-anchored scaffold (Manes.S107700). This could be due to assembly related errors in the cassava reference genome.
Multiple APL genes have been reported from different plant species, four in rice (Lee et al. 2007), four in corn (Huang et al. 2014), three in potato (Van Harsselaar et al. 2017) and four in Arabidopsis (Crevillén et al. 2003(Crevillén et al. , 2005. Here, we show presence of not four as reported by Dong et al. (2019), but five isoforms of APL genes in cassava  (8384) and wild-type TME 7S. Plants were harvested from the greenhouse for starch determination 17 weeks after planting as described in Fig. 5. Bars show SD of three biological replicates (Fig. 1). Different isoforms of APL genes are known to be expressed preferentially in different tissues and organs. For example, in rice APL2 and APS2b are endosperm specific, while APS2a and APL3 are expressed preferentially in leaves (Hirose et al. 2006;Lee et al. 2007). Similarly, in A. thaliana APL1 and APS1 are preferentially expressed in photosynthetic (source) tissues while APL3 and APL4 are expressed preferentially in roots and sugar-inducible in leaves (Crevillén et al. 2005;Streb and Zeeman 2012). In heterotrophic tissue of potato tubers APL3 (StAPL1 in this study, Fig. 1a and b) is preferentially expressed, while APL1 (StAPL3 in this study, Fig. 1a and b) is expressed mainly in leaves (Van Harsselaar et al. 2017). Expression assays in leaves and storage roots of VIGS-challenged plants (Fig. 3) coupled with data from the cassava transcriptome Atlas (Wilson et al. 2017) (Fig. 1c) clearly show preferential expression of the different MeAPL isoforms in different organs. Specifically, isoform MeAPL3 is abundantly expressed in storage roots, fibrous roots and stems (Figs. 1c and 3c). In leaves, the closely related MeAPL2 and MeAPL4 (89.85% identity at amino acid level) are expressed at higher levels than MeAPL3. Tissue specific expression of different AGPase genes is suggested to be due to functional specialization in source and sink tissue, and as such, might have different sensitivity to allosteric regulations (Ballicora et al. 2005;Nakata and Okita 1995). The MeAPL3 characterized in this study codes for conserved amino acid Q and T at position 102 and 112, respectively of the catalytic region (Fig. 1a) in a manner consistent with group III large subunit from different plant species that preferentially express in the tissues of sink organs such as tubers, stem and roots (Ballicora et al. 2005). In contrast, MeAPL1, MeAPL2, MeAPL4 and MeAPL5 have conserved residues amino acids R and K at position 102 and 112, respectively within the catalytic region common to group I and group II types based on phylogenetic classification by Ballicora et al. (2005). These genes tend to express in source tissues.
By presenting evidence from transient studies using VIGS and multiple stable co-suppression plants lines we showed that MeAPL3 is the major cassava APL isoform involved in starch accumulation in storage roots (Fig. 4, Supplementary  Fig. 1, Figs. 6 and 7). This is consistent with other studies of shrunken-2 in maize (Bhave et al. 1990), rice APL2 (Lee et al. 2007), and RNAi suppression lines (Xu et al. 2019). An interesting associated observation was an increase in average storage root number per plant in lines with reduced starch and DMC. Transgenic potato expressing antisense cassava (Munyikwa et al. 2001) and potato AGPase small subunits (Müller-Röber et al. 1992) also displayed increased tuber number, tuber fresh weight and reduction in tuber DMC.
In the current study, we demonstrated that silencing MeAPL3 using VIGS resulted in approximately 30% reduction in storage root DMC and starch content (Fig. 4).
Transgenic MeAPL3 co-suppression lines showing reduced expression of MeAPL3 ( Supplementary Fig. 2) also had significantly reduced starch and DMC (Figs. 6 and 7). Lines with reduced DMC and starch content displayed increases in soluble sugars (Figs. 4 and 7), a phenotype that we previously reported in transgenic cassava lines accumulating carotenoids (Beyene et al. 2018). These characteristics have also been reported in transgenic RNAi pea lines (Weigelt et al. 2009), and in transgenic potato expressing an antisense of potato small subunit of AGPase that resulted in increased soluble sugars and reduction in tuber DMC (Müller-Röber et al. 1992).
Leaf angle is an important plant trait that determines light interception and productivity (Drewry et al. 2014;Truong et al. 2015). Transgenic cassava lines with significantly reduced DMC and starch content displayed a unique droopy leaf phenotype on the middle and lower canopies. On the same plants newly formed, fully expanded leaves occupying the upper quarter of the canopy grew as wildtype (Fig. 6b, c). This change in petiole/stem angle has not been reported in cassava, but correlates with studies describing droopy leaf phenotype associated with mutation of droppy leaf (drl) genes in Zea mays (Strable et al. 2017). The mechanism of how starch content relates to leaf angle is not clear, but it is possible that starch at the stem/ petiole junction might play a role as shown previously in which amyloplast displacement (starch content) affected gravitropism (Weise et al. 2000;Morita and Tasaka 2004). Further investigations are needed to determine the molecular basis and physiological mechanism of this phenotype.
A predominant problem in production and utilization of cassava is the very short shelf-life of harvested storage roots, with 20-30% of yields estimated to be lost due to PPD every year (Djabou et al. 2017;Zainuddin et al. 2018). This limits the market potential and value of the crop across all its growing regions. Previous studies have implicated, but not proven, a positive correlation between high DMC and severity of PPD in cassava storage roots (Sánchez et al. 2006). We previously reported that transgenic cassava lines accumulating carotenoids have significantly reduced PPD (Beyene et al. 2018). In addition to reduced DMC, these plants were characterized by elevated levels of soluble sugars, triacylglycerols and total fatty acids, and reduced starch content (Beyene et al. 2018), making it problematic to determine what might be influencing reduced severity of PPD. Conventionally bred orange-fleshed cassava varieties also have low or no PPD incidences (Sánchez et al. 2006;Morante et al. 2010). Previous reports were inconclusive as to whether the reduction of PPD incidence was due to altered metabolite changes including reduced DMC and starch, or the direct effect of carotenoid accumulation on PPD.
Harvested cassava roots accumulate elevated levels of reactive oxygen species (ROS) (Reilly et al. 2003;Iyer et al. 2010;Zidenga et al. 2012). Reduced PPD in high betacarotene accumulating cassava storage roots has been suggested to be due to the anti-oxidant properties of carotenoids (Morante et al. 2010;Zidenga et al. 2012). In line with this hypothesis, transgenic elevation of APX (Vanderschuren et al. 2014) and alternative oxidase AOX (Zidenga et al. 2012) and co-expression of copper/zinc superoxide dismutase (Xu et al. 2013) reduced PPD due to the antioxidant properties of these enzymes. In the present study, a strong relationship is shown between incidence of PPD and dry matter/starch content, where lines with high dry matter and starch content showed 70-100% PPD after three days of storage, while lines with reduced DMC and starch content had reduced or no incidence of PPD. This clearly demonstrates PPD incidence and severity to be tightly related with storage root dry matter/starch content. Although previous studies have implicated the existence of a relationship between DMC and PPD, these relied on obsevations across different genotypes with different genetic backgrounds (Van Oirschot et al. 2000;Sánchez et al. 2006;Morante et al. 2010) or treatments like pruning (Van Oirschot et al. 2000). Sugars are known for their antioxidant properties, either by serving as signaling molecules that trigger gene expression involved in ROS scavenging (Couée et al. 2006) or directly where high concentration of sugars can function as ROS scavengers (Van Den Ende and Valluru 2009;Bolouri-Moghaddam et al. 2010). Our study shows that silencing of the MeAPL3 increases the pool of soluble sugars (both reducing and non-reducing). It remains unknown whether these sugars are critical in perturbing the incidence of PPD observed in Fig. 8 Effect of silencing MeAPL3 on cassava postharvest physiological deterioration (PPD). a transverse slices of storage roots of transgenic lines showing low or no PPD (left panel) and transgenic and wild-type with high PPD (right panel) b PPD score of storage roots. Transgenic lines were generated using an empty vector control (p8384) and p8388 that expresses MeAPL3 (sense). Storage roots were harvested from the greenhouse 17 weeks after planting, washed and dried with paper towels, and stored at 25 o C in paper bags for three days. Roots cut transversely at proximal, middle and distal ends for PPD evaluation. PPD at day 0 is not shown as no PPD was detected at that time. Bars are SD of three biological replicates this study or in transgenic and conventionally bred orangefleshed cassava.
In conclusion, MeAPL3 has been identified as the main isoform of the large subunit genes reponsible for starch biosynthesis in cassava storage roots. We also show a direct relationship between storage root DMC/starch content and cassava PPD. Open questions remain, concerning the molecular and biochemical bases linking high starch to enhanced PPD and how low starch leads to droopy phenotype in leaves.