Starch is a vital carbohydrate reserve in plants and serves as a key component of the food chain for animals. It is synthesised in plastids, including chloroplasts in leaves and amyloplasts in tubers and grain. Starch consists of two α-glucans: amylose and amylopectin. The amylose mainly presents as a linear glucan structure comprised of α-1,4 glycosidic bond with a limited α-1,6-linked branching (less than 1%). This glucan is synthesised by the concerted action of ADP-glucose pyrophosphorylase (AGPase) and granule-bound starch synthase I (GBSSI). In contrast, amylopectin is much more complex since it harbours highly branched structures generated by the action of AGPase, soluble starch synthases (SSs), starch branching enzymes (SBEs), and starch debranching enzymes (DBEs) (Ball and Morell, 2003; Tappiban et al., 2019). In addition, the starch degrading process in chloroplasts also relies on several enzymes, including β-amylases (BAMs), disproportionation enzyme 1 (D-enzyme1 or DPE1) and debranching enzymes (DBEs), which generate residual sugars, glucose and maltose, which are exported to the cytosol to be further metabolised (Zeeman et al., 2010).

Debranching enzymes (DBEs), members of glycoside hydrolase family 13 (GH13), are found in plants and microorganisms. They could be categorised into two major types: isoamylase (EC 3.2.1.68, ISA) and pullulanase (PUL, limit dextrinase, R-enzyme, EC 3.2.1.41) according to their peptide sequences and substrate specificities. Although they both can act on α-1,6 glucosidic linkages (Panpetch et al., 2018b), only PUL shows high specific activity against pullulan, α-1,6-linked maltotriose units, and liberates maltotriose final product. DBEs are required for both starch synthesis and degradation in plants. Generally, plants contain 3 isoforms of ISA (ISA1, ISA2, and ISA3) but only 1 pullulanase (Zeeman et al., 2010). The ISA isoforms 1 and 2 (ISA1 and ISA2) exist as either heterooligo- or homooligo-meric structures, depending on the plant species (Hussain et al., 2003; Panpetch et al., 2018b). They have been proposed to play important roles mainly in starch synthesis by the removal of redundant branches of immature amylopectin. This produces an appropriate structure that is competent to form highly structured granular starch, based on “the glucan trimming model” (Myers et al., 2000; Nakamura, 2002). Loss of DBEs in starch biosynthesis results in the accumulation of phytoglycogen (Delatte et al., 2005), while in starch degradation, a starch-excess or sex phenotype arises (Caspar et al., 1991). Further, PUL together with ISA3 participates in the debranching of so-called β-limit dextrin, starch residues arising from the action of BAM (Zeeman et al., 2010). Deficiency of these debranching enzymes promotes the accumulation of branch-maltooliogosaccharides and sex phenotypes, resulting in a slow growth rate (Delatte et al., 2006). To date, plant starch metabolism has been mainly evaluated in model plants such as Arabidopsis thaliana; additional information on starch metabolism in other plants, however, is also required.

Currently, only plant PUL from barley (Hordeum vulgare) (HvLD), a monocotyledon, has been biochemically and structurally characterised (Møller et al., 2015). The HvLD was classified into GH13 subfamily 13. In addition to the 3 major domains, including N-, C-, and catalytic domains, HvLD structure also haboured an additional carbohydrate-binding module 48 (CBM48) located between N-terminal domain and catalytic domain (Stam et al., 2006; Machovič and Janeček, 2008; Vester-Christensen et al., 2010b; Drula et al., 2022). Here is less data for PUL, especially from dicotyledonous plants. Only partially purified PUL from potato (Solarium tuberosum L.) (StPUL) (Ishizaki et al., 1983) and purified PUL from spinach (Spinacia oleracea L.) (SoPUL) (Renz et al., 1998) were biochemically characterised to date. Moreover, the PUL from cassava, a tropical dicotyledonous plant, has not been studied before (Tappiban et al., 2019).

Cassava (Manihot esculenta Crantz), one of the most important economic crops in Asia, Africa and South America, in particular, possesses a very high starch content in its tubers (Tappiban et al., 2019). Although cassava starch has been widely used for many years, its starch-modifying enzymes have been little studied. Some of these enzymes have been cloned, expressed, and characterised (Tappiban et al., 2019), such as DPE1 (Tantanarat et al., 2014), ISA1/ISA2 (Panpetch et al., 2018b), and ISA3 (Panpetch et al., 2018a). Notably, most studies of plant pullulanases have been conducted in vivo (Dinges et al., 2003; Fujita et al., 2009; Li et al., 2019), which is complicated by the concerted action of several other enzymes and, as a result, specific action information about individual enzymes has rarely been investigated. This study aimed to biochemically characterise pullulanase from Manihot esculenta Crantz (MePUL), with a view to determining its biological role in starch metabolism which remained unclear.

The protein sequence of rMePUL was identical to the cassava pullulanase in the Phytozome database [Manes.10G051700.1.p]. Comparison of the MePUL to those of other plants and bacteria showed some sequecne similarity - around 37 – 74%, including to Hordeum vulgare pullulanase (HvLD; GenBank AAD04189.1, 65%), Oryza sativa pullulanase (OsPUL; GenBank BAA09167.1, 60%), Solanum tuberosum pullulanase (StPUL; GenBank XP_006361707.1, 71%), Arabidopsis thaliana pullulanase (AtPUL; GenBank NP_196056.2, 72%) Zea mays pullanase (ZmPUL; GenBank NP_001104920.1, 74%), and Spinacia oleracea pullulanase (SoPUL; GenBank CAA58803.1, 71%), and Klebsiella pneumoniae pullulanase, a bacterial pullulanase (KpPUL: PDB 2FGZ, 37%). The sequence alignments are presented in Figures S2, S7 . However, MePUL showed very low similarity compared with other types of Manihot esculenta debranching enzymes (MeISA1: GenBank AUZ20772.1, 23%; MeISA2: GenBank AUZ20773.1, 24%; and MeISA3: GenBank AUZ20774.1, 29%) shown in Figure S8 . Although they are members of the same Glycoside hydrolase (GH) 13 family, ISAs are subcategorised into subfamily 11. The MePUL belongs to GH 13 subfamily 13 (pullulanase subfamily) based on sequence similarity (Machovič and Janeček, 2008). The MePUL also comprised 4 domains of N-terminal domain, CBM48, catalytic domain, and C-terminal domain similar to that of HvPUL as shown in Figure S2 (Vester-Christensen et al., 2010b).

Fortunately, although the MePUL comprised a total of 14 cysteine residues in the mature protein sequence, it was successfully expressed in soluble form in E. coli SoluBL (DE3) transformant. Noticeably, protein MW obtained from gel filtration chromatography under non-denaturing condition was ~ 2-fold higher than the MW shown by SDS-PAGE analysis ( Figure 1 ), which implied that MePUL could be naturally present as a homodimer. Unlike other debranching enzymes from cassava, the MeISA1 and MeISA2 essentially exist as catalytically active hetermeric enzymes with various molar ratios (Panpetch et al., 2018b), while MeISA3 is a monomeric enzyme (Panpetch et al., 2018a). Additionally, SoPUL was reported as a monomeric protein (Renz et al., 1998) and partially purified StPUL from potato tuber present as a dimer analysed under nondenaturing condition however 3 protein bands were shown on SDS-PAGE gel (Ishizaki et al., 1983).

An in vitro biochemical assay of rMePUL disclosed that rMePUL has pH optimum in the slightly acidic pH range and is a metal ion-dependent enzyme ( Figure 2 ). Surprisingly, the optimum temperature of rMePUL was at 50°C ( Figure 2A ), while for rMeISA3 it was at only at 37°C (Panpetch et al., 2018a) even though they both originate from the same organism. rMePUL exhibited the highest activity in acetate buffer pH at 6.0 ( Figure 2B ) like rMeISA3. The activity dropped by up to 50% at pH 5.0 and 7.0, with a sharp drop of activity below 5 or above 7, similar to the reported optimum pH of isoamylases from potato and maize (Ishizaki et al., 1983; Doehlert and Knutson, 1991; Rahman et al., 1998). Nevertheless, both optimal pH and temperature between MePUL and StPUL were similarly (Ishizaki et al., 1983). Moreover, only Ca²⁺ ion could promote the activity of rMePUL ( Figure 2C ) corresponded to the presence of embedding of Ca²⁺ in HvPUL crystal structure (Vester-Christensen et al., 2010b), while the opposite was true for StPUL where Ca²⁺ ion lowered the enzyme activity by around 40%. However, enzyme activity was completely inhibited by Cu²⁺ and Zn²⁺ ( Figure 2C ), correlating with previous work of Iwaki and Fuwa (1981) (Iwaki and Fuwa, 1981), which reported that 1 mM Cu²⁺, Zn²⁺, Ag²⁺, or Cd²⁺ inhibited rice endosperm debranching enzyme. Moreover, we also found that rMePUL activity could be enhanced after treating with EDTA. Possibly, there was contamination of some metal ion inhibitors in gene expression and protein purification processes

To evaluate the substrate specificity of rMePUL, the enzyme was incubated with various substrates ( Figure 2 ). rMePUL exhibited the highest activity toward pullulan ( Figure 2A ). This confirmed that this debranching enzyme was pullulanase-type, while isoamylase-type debranching enzyme clearly showed no activity on pullulan at all (Panpetch et al., 2018b; Panpetch et al., 2018a). Notably, rMePUL was unable to cleave substrates harbouring highly branched structures, especially amylopectin, cassava starch, and glycogen. However, compared to no activity on cassava starch, rMePUL could digest potato starch weakly, with ~20% of its activity on pullulan. This possibly resulted from less complexity of the intrinsic structure of potato starch than that of cassava. Hence, in vitro results could support the trimming model hypothesis of starch granule synthesis, proposing that PUL could partially substitute for ISA1 activity (Dinges et al., 2003; Wattebled et al., 2005; Fujita et al., 2009). Importantly, rMePUL also preferentially digested β-limit dextrin at a high level as shown in Figure 2D . This might be because its structure is rather similar to that of α-limit dextrin or pullulan-like polymers found in plants, which are both produced by amylase action in starch degradation. Moreover, the results from protein localisation indicated that MePUL was specifically localised to the plastid organelles ( Figure 8 ) such as chloroplast and amyloplast, where starch degradation occurs in plant cells (Zeeman et al., 2010). Taken together, these data along with some previous reports on target localisation to plastid organelles support the notion that MePUL principally participates in starch degradation processes (Zeeman et al., 2010). Additionally, unlike other PULs from cereals, such as rice (Yamasaki et al., 2008) and barley (Mcdougall et al., 2004), transglycosylation activity could not be observed in rMePUL.

Enzyme kinetic analysis with rMePUL acting on pullulan and β-limit dextrin was fitted with Hill plots, indicating the cooperative manner of the enzyme. Conversely, it was still unclear why addition of some maltooligosaccharides such as maltose and maltotriose could obviously change the kinetic pattern of rMePUL (from Hill to Michaleis-Menten). Notably, maltotriose, the main product of pullulan hydrolysis, was a potential positive effector of the rMePUL action, since it could clearly promote binding affinity (k or K_m ) of the enzyme to pullulan substrate. However, maltose and maltotriose could also increase k_cat around 1.2 - 1.7 folds for β-limit dextrin hydrolysis ( Figure 3B , Supplementary Figure S3 , and Table 1 ). It might be that these maltooligosaccharides produced during starch degradation (Zeeman et al., 2010) act as allosteric modulators to some enzymes participating in the starch degradation process.

Although MePUL exhibited very low sequence similarity to KpPUL (Mikami et al., 2006) ( Supplementary Figure S7 ), their product patterns of on pullulan degradation were rather similar, mainly reflecting the release of maltotriose and maltotriose-containing oligosaccharides ( Figure 4B ). However, pullulan does not entirely consist of maltotriose building blocks (Catley and Whelan, 1971; Carolan et al., 1983); it also contains maltooligosaccharides with different DPs, such as maltose (DP2) and maltotetraose (DP4), which could also be detected in HPAEC analysis from rMePUL and KpPUL reactions. rMePUL specifically hydrolysed α-1,6 linkage between maltoriose units but no hydrolytic activity on linear α-1,4-linked oligosaccharides was observed ( Figure 4A ), strongly indicating that MePUL is a pullulanase type I (Møller et al., 2016). Cyclic oligosaccharides such as β-cyclodextrin (β-CD) and α-1,6 branched β-cyclodextrins (β-CD-G1 and β-CD-G2) were not cleaved by rMePUL. Furthermore, the pullulan cleavage was almost completely inhibited by β-CD as judged by DNS assay. This corresponded to the competitive inhibition of debranching activities of KpPUL and HvLD by CDs, where the cyclic structure of cyclic oligosaccharides can productively bind at subsites +1, +2, and 0′, as shown in superimposition structures of MePUL and HvLD (Vester-Christensen et al., 2010a; Møller et al., 2015) ( Supplementary Figure S9 ). Although β-CD-G1 and β-CD-G2 contained an α-1,6 branch, the glucosyl and maltosyl branches were not appropriately positioned at catalytic cleavage site ( Supplementary Figure S9 ), corresponding to complex structure of HvLD with 6-S-(α-D-maltosyl)-6-deoxy-6-thiocyclomaltoheptaose (Møller et al., 2015). Therefore, β-CD-G1 and β-CD-G2 could not be hydrolysed by rMePUL. Interestingly, rMePUL could not digest the α-1,6 linkage of glucosyl-maltotriose (G1G3), panose or isopanose. The experimental data demonstrated that the active site of rMePUL required longer substrates in order to make productive contacts with the longer binding site. This information acorrelated with computational analysis, which showed that the MePUL-branched maltohexaose complex exhibited high binding energy at +2, +1, -1, -2, and -3 sites ( Figure 6C ). The computational and bioinformatic data revealed that the residues W365, Y367, N368, H414, C449, V450, L484, W522, F524, D551, R554, F563, H651, D652, N653, D708, Y710, and K737 were potentially essential for substrate binding based on the results of the MM/GBSA method and conserved sequence analysis. Most of these residues, which bound with ligands ( Figure 6B ), were highly conserved at the same positions in the 3D-structures of MePUL, HvLD, (Møller et al., 2015), and KpPUL (Mikami et al., 2006) even though they did not show high similarity of overall protein sequences. Notably, two out of those residues, F563 and D708 located near the branch point of branched maltohexaose, showed 80% lower score of the cutoff conservation. Perhaps these variable residues might play a role in imposing the substrate specificity. Proposed oligosaccharide binding subsites of MePUL are shown in Figure 6D .

To analyse the relationship between MePUL and MeISA3 debranching activities in starch degradation, mixed-enzyme reaction was performed on β-limit dextrin substrate ( Figure 7 ). The results suggested that rMePUL preferentially cleave longer chains in broad ranges of maltooligosaccharides and meanwhile rMeISA3 liberated only short-chain maltooligosaccharides ( Figure 7 and Figure S4 ). This suggested that MePUL might play a major role in starch debranching of cassava starch catabolism. Interestingly, when rMeISA3 and rMePUL were present together in the same reaction, the debranching activity of short-chain oligosaccharides (DP 2 - 4) was apparently increased ( Figure 7 and Figure S6 ). This result implied that MeISA3 and MePUL worked synergistically, possibly by providing favourable substrates for each other by removal of some unfavourable branched points due to their different substrate preferences. Importantly, considering the prospect of an heterocomplex between rMeISA3 and rMePUL, no multimeric rMeISA3/rMePUL complex was observed by gel filtration chromatography ( Supplementary Figure S10 ). This demonstrated that they should work separately for their debranching activities.

Some previous studies in vivo revealed that deficiency of PUL in rice (Fujita et al., 2009) and maize (Wong et al., 2003) slightly effected on an accumulation of water-soluble glucan. Although MeISA3 and MePUL seemed to exhibit overlapping activities in debranching processes, cassava starch catabolism may not be efficient when only either PUL or ISA3 are present, which contrasts with some previous reports from other plant species (Wong et al., 2003; Delatte et al., 2006; Fujita et al., 2009). Importantly, both PUL and ISA3 have been conserved in most plants. Previous in vivo studies showed the relationship between these two enzymes, but they did not clearly unravel the reason for the conservation of both enzymes. Nevertheless, our results show that both PUL and ISA3 are necessary for efficient starch degradation in vitro: their synergistic action may account for why these two types of debranching enzymes are typically present in plant cells.

Pullulanase from Manihot esculenta Crantz ‘KU50’ was successfully cloned, expressed, and biochemically characterised. This debranching enzyme was clearly categorised as pullulanase type I due to its preference for cleaving α-1,6 linkages of pullulan and β-limit dextrin. The debranching action of PUL is conserved in most plants, both mono- and di-cotyledons, and is located in plastids. The present work demonstrates the synergistic action between MePUL and MeISA3, which is important for efficient starch degradation in cassava and which may have implications for biotechnology innovation in relation to harnessing starch for food and materials applications.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

KW, TC, and PP: conceptualization. KW, TC, MK, and PP: methodology. KW, TC, and PP: formal analysis. KW, TC, MK, RF, and PP: investigation. TC and PP: resources. TC and PP: data curation. KW, TC, and MK: writing - original draft. KW, TC, RF, and PP: writing - review & editing. KW, TC, MK, and PP: visualization. KW and PP: supervision. PP: funding acquisition. PP: Project Administration. All authors contributed to the article and approved the submitted version.