Phosphoglucan-bound structure of starch phosphataseStarch Excess4 reveals the mechanism for C6 specificityDavid A. Meekinsa, Madushi Raththagalaa, Satrio Husodoa, Cory J. Whitea, Hou-Fu Guoa, Oliver Köttingb,Craig W. Vander Kooia,1, and Matthew S. Gentrya,1

aDepartment of Molecular and Cellular Biochemistry and Center for Structural Biology, University of Kentucky, Lexington, KY 40535-0509; and bInstitute ofAgricultural Sciences, Eidgenössische Technische Hochschule (ETH) Zürich, 8092 Zürich, Switzerland

Edited by Nicole Koropatkin, University of Michigan, Ann Arbor, MI, and accepted by the Editorial Board April 1, 2014 (received for review February 3, 2014)

Plants use the insoluble polyglucan starch as their primaryglucose storage molecule. Reversible phosphorylation, at theC6 and C3 positions of glucose moieties, is the only known naturalmodification of starch and is the key regulatory mechanismcontrolling its diurnal breakdown in plant leaves. The glucanphosphatase Starch Excess4 (SEX4) is a position-specific starchphosphatase that is essential for reversible starch phosphoryla-tion; its absence leads to a dramatic accumulation of starch inArabidopsis, but the basis for its function is unknown. Here wedescribe the crystal structure of SEX4 bound to maltoheptaose andphosphate to a resolution of 1.65 Å. SEX4 binds maltoheptaose viaa continuous binding pocket and active site that spans both thecarbohydrate-binding module (CBM) and the dual-specificity phos-phatase (DSP) domain. This extended interface is composed ofaromatic and hydrophilic residues that form a specific glucan-inter-acting platform. SEX4 contains a uniquely adapted DSP activesite that accommodates a glucan polymer and is responsible forpositioning maltoheptaose in a C6-specific orientation. We identi-fied two DSP domain residues that are responsible for SEX4 site-specific activity and, using these insights, we engineered a SEX4double mutant that completely reversed specificity from the C6 tothe C3 position. Our data demonstrate that the two domains actin consort, with the CBM primarily responsible for engaging glucanchains, whereas the DSP integrates them in the catalytic site for posi-tion-specific dephosphorylation. These data provide important insightsinto the structural basis of glucan phosphatase site-specific activityand open new avenues for their biotechnological utilization.

carbohydrate | Lafora disease | LSF2 | laforin

Starch is the primary carbohydrate storage molecule in plantsand is an essential constituent of human and animal diets.

Starch granules are composed of the glucose homopolymersamylose (10–25%) and amylopectin (75–90%) (1, 2). Amylose isa linear molecule formed from α-1,4-glycosidic–linked chains,whereas amylopectin is formed from α-1,4-glycosidic–linkedchains with α-1,6-glycosidic–linked branches (3, 4). Adjacentamylopectin chains interact to form double helices that causestarch granules to be water insoluble, which is an essential fea-ture for its function as a glucose storage molecule (1, 3, 5).However, the outer granular surface of transitory starch must besolubilized during nonphotosynthetic periods so that glycolyticenzymes can access and degrade starch glucans and meet themetabolic needs of the plant (6, 7). Plants regulate the solubilityof the starch granular surface via reversible starch phosphory-lation that results in a cyclic degradative process: phosphoryla-tion by dikinases, degradation by starch hydrolyzing amylases,and dephosphorylation by phosphatases (1, 8–11). Phosphory-lation of amylopectin chains causes helical unwinding and localsolubilization of the outer starch granule (12–14). The localsolubilization and helix unwinding permits degradation of sur-face, linear α-1,4 glucan chains by β-amylase, which sequentiallyremoves maltosyl units from the nonreducing end (1, 8, 15).Although glucan phosphorylation of the starch surface is nec-essary for degradation, the removal of these phosphate groups is

required because β-amylase is unable to degrade past the phos-phate (6, 15–17). Therefore, glucan phosphatases must releasephosphate from starch to reset the degradation cycle, allowingprocessive starch degradation (8, 16).Recent studies have established that plants use a two-enzyme

system for both starch phosphorylation and dephosphorylation.α-Glucan water dikinase phosphorylates the hydroxyl group ofstarch glucose at the C6 position. This event triggers phosphor-ylation of the hydroxyl group at the C3 position by phosphoglucanwater dikinase (18–21). Similarly, two glucan phosphatases releasephosphate from starch. Starch Excess4 (SEX4) preferentiallydephosphorylates the C6 position of starch glucose and Like SexFour2 (LSF2) exclusively dephosphorylates the C3 position (22–26). SEX4 activity is essential for starch catabolism and its mu-tation in Arabidopsis leads to an excess of leaf starch, a decreasein plant growth, and an accumulation of soluble phosphoglucansproduced by the activity of α-amylase 3 and the debranchingenzyme isoamylase 3 (16, 25, 27). Conversely, lsf2 mutant Arabi-dopsis plants display normal levels of leaf starch and plant growth,but the starch contains increased levels of C3-phosphate (22). Thedifference in plant vitality between sex4 and lsf2 mutants is likelydue to SEX4 possessing some compensatory C3-position phos-phatase activity (22). Cumulatively, the process of reversible phos-phorylation requires the concerted activity of dikinases and

Significance

Starch is the main carbohydrate storage molecule in plants andis ubiquitous in human life. Reversible starch phosphorylationis the key regulatory event in starch catabolism. Starch Excess4(SEX4) preferentially dephosphorylates the C6 position of starchglucose and its absence results in a dramatic accumulation of leafstarch. We present the structure of SEX4 bound to a phospho-glucan product, define its mechanism of specific activity, andreverse its specificity to the C3 position via mutagenesis. Theability to control starch phosphorylation has direct applicationsin agriculture and industrial uses of starch. These insights intoSEX4 structure and function provide a foundation to controlreversible phosphorylation and produce designer starches withtailored physiochemical properties and potentially widespreadimpacts.

Author contributions: D.A.M., C.W.V.K., and M.S.G. designed research; D.A.M., M.R., S.H.,C.J.W., H.-F.G., C.W.V.K., and M.S.G. performed research; D.A.M., O.K., C.W.V.K., and M.S.G.contributed new reagents/analytic tools; D.A.M., M.R., S.H., H.-F.G., O.K., C.W.V.K., andM.S.G. analyzed data; and D.A.M., C.W.V.K., and M.S.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. N.K. is a guest editor invited by the EditorialBoard.

Data deposition: The atomic coordinates and structure factors have been deposition inthe Protein Data Bank (PDB ID code 4PYH).1To whom correspondence may be addressed. E-mail: matthew.gentry@uky.edu or craig.vanderkooi@uky.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1400757111/-/DCSupplemental.

7272–7277 | PNAS | May 20, 2014 | vol. 111 | no. 20 www.pnas.org/cgi/doi/10.1073/pnas.1400757111

phosphatases with SEX4 activity being essential for normal patternsof starch metabolism and plant growth.Glucan phosphatases are members of the protein tyrosine

phosphatase (PTP) superfamily characterized by a conservedCx5R catalytic motif (24, 28, 29). The glucan phosphatases be-long to a heterogeneous subset of PTPs called dual-specificityphosphatases (DSPs), with some DSPs dephosphorylating p-Tyrand p-Ser/Thr residues of proteinaceous substrates and otherDSPs dephosphorylating lipids, nucleic acids, or glucans (30–32).In addition to SEX4 and LSF2, a glucan phosphatase calledlaforin has been identified that dephosphorylates glycogen andinfluences its solubility in vertebrates (24, 33, 34). Loss of laforinfunction in humans results in a fatal, neurodegenerative epilepsycalled Lafora disease (35–37). Due to their critical function incomplex carbohydrate metabolism, understanding the structuralbasis of glucan phosphatase activity is of particular interest.Toward this goal, we previously determined the ligand-freestructure of SEX4 and identified an extensive interdomaininterface between its DSP domain and carbohydrate-bindingmodule (CBM) that is maintained in part by a previously un-recognized C-terminal (CT) motif (38). However, the structuralmechanism for domain coupling, glucan interaction, and specificC6 dephosphorylation in SEX4 activity is unclear.Starch granule solubilization depends on phosphorylation of

starch glucose on the hydroxyl group at both the C6 and C3positions (6, 13). These phosphorylation events are critical fornormal transitory starch degradation, but also directly impact themelting temperature, viscosity, and hydration of starch in in-dustrial settings (39, 40). Developing a means to manipulatestarch phosphorylation patterns via enzymatic modification isrelevant to agricultural and industrial applications that use starchas a feedstock (9, 12, 14, 41). Therefore, understanding the basisfor the site specificity of glucan phosphatases is of particularinterest. We recently determined the structure of LSF2 witha glucan bound in a C3-specific orientation and identified uniquenoncatalytic surface-binding sites (SBSs) not found in otherglucan phosphatases (26). SEX4 lacks SBSs and preferentiallydephosphorylates the C6 position. The present study was designedto define the fundamental basis for SEX4 substrate binding andunderstand preferential C6-position specificity in SEX4.Herein, we elucidate the structural mechanism of SEX4-

specific activity by determining the structure of SEX4 bound to thephosphoglucan products maltoheptaose and phosphate. SEX4engages glucan chains via an extended interface of aromatic andhydrophilic residues that spans the CBM and DSP domains.Moreover, the SEX4 CBM is primarily responsible for glucanbinding whereas the SEX4 DSP active site is uniquely adapted toengage the phosphoglucan substrate, positioning it in a C6-specific

orientation. Structure-guided mutagenesis of DSP active-site res-idues resulted in a complete reversal from C6 to preferential C3dephosphorylation by SEX4. Cumulatively, this study establishesthe molecular basis for both SEX4 substrate engagement andSEX4 specificity and provides a method for engineering glucanphosphatase activity with modified site specificity.

ResultsStructure of SEX4 Bound to Maltoheptaose and Phosphate. The struc-ture of the glucan phosphatase Arabidopsis thaliana SEX4 [residues90–379, C198S (inactive mutant)], with maltoheptaose and phos-phate bound in the active site, was determined to a resolution of1.65 Å (Fig. 1A and Table 1). The crystallized SEX4 constructcontains the catalytic DSP domain, the CBM, and CT motif.Maltoheptaose is composed of seven glucose moieties with α-1,4-glycosidic linkages, and clear electron density allowed the modelingof six glucose units of the maltoheptaose chain (Fig. 1B). Themaltoheptaose chain is located within an extended pocket thatspans the DSP and CBM domains, with Glc1 located at the DSPand Glc6 located at the CBM (numbered from nonreducing toreducing end). In addition, a single phosphate molecule was foundwithin the catalytic site (PTP loop), directly below Glc2, at a dis-tance of 2.5 Å from the catalytic residue S(C)198. The DSP–CBMpocket is ∼9 Å deep and ∼33 Å long with a total contact area of610 Å2. Of this contact area, 40% of the interactions occur via theCBM domain and 60% via the DSP. Together, maltoheptaose andphosphate represent the product of dephosphorylation of the en-dogenous SEX4 phosphoglucan substrate, thus demonstrating thestructural basis for SEX4–phosphoglucan interactions.

Maltoheptaose at the SEX4 CBM. The SEX4 CBM interacts withthe maltoheptaose chain moieties Glc4–6 (Fig. 2A). The centralplatform for this interaction is a dual-tryptophan motif formedfrom W278 and W314, which combines with H330 to interactwith both faces of the glucan chain. In addition, N332 and K307 arepositioned behind maltoheptaose and form hydrogen-bondinginteractions with the O3 groups of Glc5 and Glc6, respec-tively. These five residues are highly conserved among SEX4orthologs (Fig. S1) and form a concerted glucan-interactinginterface.The SEX4 CBM belongs to the CBM48 family and the dual-

tryptophan platform (W278/W314) and K307 represent a con-served functional motif in CBM48 and the related CBM20 family(42, 43). The most similar CBM to that of SEX4 is found in theAMP-activated protein kinase β-subunit (AMPK-β) (ProteinData Bank ID code 1Z0N), which has a root-mean-square de-viation (rmsd) of 1.4 Å compared with the SEX4 CBM (44).AMPK-β contains all of the SEX4 CBM glucan-interacting res-idues except for a threonine where H330 is located in SEX4. A

Fig. 1. Crystal structure of At-SEX4 bound to aglucan chain and phosphate. (A) Surface/ribbon dia-gram of At-SEX4 [residues 90–379, C198S (catalyticallyinactive mutant)] bound to maltoheptaose (green)and phosphate (orange) determined to a resolution of1.65 Å. The SEX4 structure contains the DSP domain(blue) with the catalytic site (red), the CBM (pink), andthe CT motif (tan). The maltoheptaose chain (green) islocated in an extended pocket spanning the CBM andDSP domains, and a single phosphate molecule is lo-cated at the base of the catalytic site directly beneathGlc2. (B) Close-up view rotated 45° showing the 2Fo-Fcelectron density map (1.5 σ) of the maltoheptaosechain (green) and phosphate (orange) bound to SEX4.The density permittedmodeling of six glucosemoietiesin the maltoheptaose chain (numbered from non-reducing to the reducing end) and clear assignment ofglucan orientation. Two conformers were modeled forGlc1, differing in the orientation of the O6 group.

Meekins et al. PNAS | May 20, 2014 | vol. 111 | no. 20 | 7273

BIOCH

EMISTR

Y

comparison between the maltoheptaose-bound and ligand-freeSEX4 structures reveals that H330 undergoes a conformationalshift upon glucan binding, bringing H330 directly in line withGlc6 and W314 (Fig. S2A) (38). This glucan-bound configurationis ideally structured for positioning a linear glucan chain in theSEX4 CBM-binding interface.To determine the contribution of the CBM to overall SEX4

activity, we mutated the aforementioned CBM residues to alanineand tested the mutants’ ability to dephosphorylate native Arabi-dopsis starch, the endogenous substrate of SEX4 (Fig. 2B). Ala-nine point mutations of W278, K307, W314, H330, and N332resulted in a decrease of total phosphatase activity ranging from66% to 97%. The largest decrease resulted from the mutation ofW278, which is located closest to the interface between the CBMand DSP domains of SEX4. All CBM mutants had generic para-nitrophenyl phosphate (pNPP) dephosphorylation levels compa-rable to wild type, indicating that decreases in glucan phosphataseactivity were not due to aberrant folding (Fig. S3). Because CBMstypically engage substrates, these results suggest that a loss of starchphosphatase activity may indicate a decrease of starch binding.To test this, we incubated SEX4 proteins with amylopectin coupledto Con A Sepharose (GE Healthcare), washed the beads, andused Western analysis to determine if the protein was bound toamylopectin in the pellet or remained in the supernatant. Wefound that mutation of CBM residues resulted in a dramaticdecrease in amylopectin binding (Fig. 2C). Together, these datademonstrate that the SEX4 CBM is essential for glucan bindingand, consequentially, dephosphorylation of starch.

Maltoheptaose and Phosphate at the SEX4 DSP. The binding site atthe SEX4 CBM interface is continuous with a correspondinginterface in the DSP that guides Glc1–4 of the maltoheptaosechain directly over the catalytic site (Fig. 3A). The maltohep-taose chain has a curved configuration at the DSP active site.

The concave surface of the maltoheptaose chain interacts withF167, which was previously shown to be important for glucanphosphatase activity (38). The convex surface interacts withK237, F235, Y90, F140, and Y139. All of these residues are lo-cated in DSP subdomains whose variability among DSP familymembers corresponds with the specific substrate requirements ofeach particular phosphatase (45). Y90 is located in the Recog-nition Domain (residues 90–98), F140 and Y139 are in the var-iable (V) loop (131–157), F235 and K237 are located in the Rmotif (230–249), and F167 is located in the D loop (162–168)(Fig. S4). A comparison of the glucan-bound DSP with the non–glucan-bound SEX4 structure reveals that residues F167, F235,Y139, and K237 undergo a conformational shift upon glucanbinding to engage the glucan chain (Fig. S2B). Based on thesedata, we hypothesized that the DSP active site is also essentialfor dephosphorylation of starch glucan chains.To test this hypothesis, we generated alanine mutations of the

identified DSP residues and determined the mutants’ ability todephosphorylate Arabidopsis starch granules (Fig. 3B). Mutationof Y90, Y139, F140, F235, and K237 to alanine resulted in adecrease of total starch dephosphorylation ranging from 10% to80%. Each mutant had generic pNPP dephosphorylation levelscomparable to wild type, indicating that decreases were not dueto aberrant folding (Fig. S3). Interestingly, the average decreasein SEX4 starch dephosphorylation activity upon DSP mutation(38%) was lower than the average decrease upon CBM mutation(89%). Furthermore, we found that the DSP mutants maintainednear-wild type-binding to amylopectin (Fig. 3C). These datawere surprising, given that the DSP has a larger relative contactarea with the phosphoglucan substrate (60%) than the CBM(40%). However, the CBM is in contact with only the glucanwhereas the DSP is in contact with both phosphate and glucan.We hypothesized that the CBM functions in engaging glucanchains while the DSP integrates and positions a phosphoglucaninto the catalytic site, thereby dictating C6-specific activity.Indeed the maltoheptaose chain within the DSP domain of the

SEX4 structure is clearly positioned in a C6-specific orientation(Fig. 3D). The O6 group of Glc2 interacts with the phosphate inthe catalytic site at a distance of 2.6 Å, compared with 7.1 Å forthe O3 group. In addition, the glucose moieties upstream anddownstream of Glc2 are also oriented with the C6 position to-ward the catalytic site. Interestingly, the CBM mutants H330Aand K307A, which dramatically decreased overall glucan phos-phatase activity (Fig. 2B), still maintained a C6-position sitespecificity that is nearly identical to wild-type SEX4 (Fig. S5).

Table 1. Data collection and refinement statistics

Δ91 SEX4 (C193S); maltoheptaoseand phosphate

Data collectionBeamline APS 22-IDSpace group P212121Cell dimensions (a, b, c) 33.48, 77.89, 117.21Unique reflections 36,096Completeness, % 95.2 (91.0)Resolution, Å 20.0–1.65 (1.71–1.65)Rmerge % 8.0 (43.6)Redundancy 3.3 (2.8)I/σI 11.0 (2.3)

RefinementResolution limits, Å 19.43–1.65No. of reflections/no. to

compute Rfree

34,257/1,807

R (Rfree) 17.9 (21.8)No. of protein residues 293No. of solvent molecules 355No. of ligands 2B factorsProtein 17.3Ligand/ion 18.9 (maltoheptaose)

9.1 (phosphate)Water 27.5rmsdBond lengths, Å 0.009Bond angles, ° 1.52

Values in parentheses are for the highest resolution shell.

Fig. 2. Interaction of maltoheptaose with the SEX4 CBM. (A) Close-up ofGlc4–6 (green) bound to the SEX4 CBM (pink). CBM residues (yellow) interactwith Glc5 and Glc6 to form a binding interface. (B) Specific activity of CBMmutants against Arabidopsis starch. Error bars represent the ±SD of sixreplicates. Inactive mutant SEX4-C198S (SEX4 C/S) was used as a negativecontrol. Statistical comparison of wild-type and mutant activity demon-strates significant differences, P < 0.001, between all constructs. (C) Resultsof the cosedimentation assay of SEX4 and CBM mutants with amylopectin.Amylopectin-bound proteins are found in the pellet (P) and unbound pro-teins are found in the soluble (S) fraction.

7274 | www.pnas.org/cgi/doi/10.1073/pnas.1400757111 Meekins et al.

These data support the hypothesis that the DSP functions tocontrol phosphoglucan orientation, i.e., site specificity.

Structural Basis of SEX4 C6 Specificity. To determine the structuralbasis of the C6-specific orientation of maltoheptaose at the DSP,we examined the residues that form the surface of the SEX4active site (Fig. 4A) and compared them to LSF2, which is a C3-specific glucan phosphatase (Fig. 4B) (26). Although the overallactive sites of SEX4 and LSF2 are globally similar, we identifiedspecific conserved differences within the active sites. In SEX4,F235 in the R motif and F140 in the V loop that interact withGlc3 and Glc1 at a distance of ∼3.8 Å. In contrast, LSF2 containsG230 in its R motif and W136 in its V loop at the same positionsas SEX4 F235 and F140, respectively. In LSF2, G230 and W136form hydrogen bonds with the glucose moieties upstream anddownstream of the moiety at the catalytic site, whereas SEX4uses van der Waals interactions at these positions. Furthermore,analysis of the active site surface shows that SEX4 containsa distinct ridge at the R motif, whereas LSF2 contains a grooveat the same position (Fig. 4C). The ridge in SEX4 is created bythe β-carbon of F235, and the groove in LSF2 results from theabsence of a side chain at G230. We hypothesized that F235 andF140 in SEX4 combine to form a distinctive active site topologythat positions the phosphoglucan chain in a C6-specific orien-tation at the site of catalysis.To test this hypothesis, we mutated F235 and F140 in SEX4

to the corresponding residues found in LSF2. If these residuesinfluence substrate specificity in SEX4, then their mutationshould result in a shift from preferential C6-position dephos-phorylation to preferential C3-position dephosphorylation. Wegenerated single-point mutants (SEX4-F140W and SEX4-F235G)and a double mutant (SEX4-F140W/F235G) and determined therelative ability of each mutant to dephosphorylate the C6 andC3 position of Arabidopsis starch granules (Fig. 4 D and E).Wild-type SEX4 dephosphorylates the C3 position of starchglucose at a rate of 29% of total dephosphorylation. Strikingly,both the F140W and F235G mutations increased the rate of C3dephosphorylation to 51% of total dephosphorylation, effec-tively removing SEX4 C6 specificity. Even more remarkably, theF140W/F235G double mutant increased the ratio of C3 de-phosphorylation to 77%, fully reversing the substrate preferenceof SEX4 from the C6 to the C3 position. These results clearlysupport our hypothesis and indicate that F235 and F140 in theSEX4 DSP active site are responsible for preferential C6

dephosphorylation of starch granules. Moreover, the ability toreverse site-specific activity via DSP mutagenesis provides valu-able insights into the basis of substrate specificity for the glucanphosphatase family.

DiscussionReversible starch phosphorylation is the central regulatory eventgoverning the transition from starch synthesis to starch break-down in plant cells. SEX4 is critical to starch catabolism andthese data reveal the structural mechanism of its activity. Ourdata demonstrate the basis for the coordinated function of theSEX4 domains, with glucan engagement achieved via the CBMand phosphoglucan positioning and site specificity via the DSPdomain. We further demonstrate that two residues in the activesite control the specificity of SEX4. These results establish thestructural basis for the specific activity of glucan phosphatases.Within the SEX4 glucan-binding interface, the CBM interacts

with a nonphosphorylated glucan and the DSP integrates aphosphoglucan into the DSP active site. Although the DSPclearly positions the phosphoglucan, our starch-binding and de-phosphorylation data demonstrate that the DSP plays only aminor role in glucan binding and demonstrates that the CBM isrequired to engage the glucan. DSPs function via nucleophilicattack of the phosphorus atom by the catalytic cysteine followedby formation and then hydrolysis of a phosphoenzyme inter-mediate (28, 31, 32). The integration of the phosphoglucan in-to the catalytic site by the DSP, with minimal contribution tooverall glucan binding, is optimal for both formation of thespecific phosphoenzyme intermediate and disengagement of theproduct. Thus, the SEX4 mechanism is ideal for enhancingsubstrate/product turnover.Despite being composed of only α-linked glucose, starch is

a complex substrate containing two glucose polymers, α-1,6-branch points, variable chain lengths, and a helical secondarystructure. Thus, SEX4 must access phosphoglucans within thisheterogenous landscape. Although phosphate positioning withinglucan chains of amylopectin is currently unknown, our structuresuggests that multiple C6-phosphate modifications could be ac-commodated at Glc1, Glc2, Glc5, and Glc6. Additionally, thestructure indicates that an α-1,6-branch could be accommodatedat both glucan termini and at Glc5. Glucan chains within amy-lopectin also possess an inherent torsion as they wrap intodouble-helical structures (4, 6). It is of note that the boundglucan chain possesses both curvature and torsion, with the C6

Fig. 3. Interaction of maltoheptaose and phos-phate at the SEX4 DSP active site. (A) Close-up ofmaltoheptaose (green), phosphate (orange), andthe DSP (blue) active site (red). Side chains ofinteracting residues are colored yellow. (B) Specificactivity of DSP mutants against Arabidopsis starch.Error bars represent the ±SD of six replicates. Sta-tistical comparison of wild-type and mutant activitydemonstrates significant differences, P < 0.005,between all constructs. (C) Results of the cosedi-mentation assay of SEX4 and DSP mutants withamylopectin. Amylopectin-bound proteins are foundin the pellet (P) and unbound proteins are found inthe soluble (S) fraction. (D) Fo-Fc omit electron den-sity map (2.5 σ) of maltoheptaose and phosphateshows clear glucan orientation of the O6 grouppointed toward the catalytic site (red). O6 and O3groups of the Glc2 moiety are labeled. Catalytic tri-ad residues D166, R204, and S(C)198 are depictedin yellow.

Meekins et al. PNAS | May 20, 2014 | vol. 111 | no. 20 | 7275

BIOCH

EMISTR

Y

hydroxyls of Glc5–6 in the CBM pointed toward the solvent andaway from the protein, whereas they are pointed toward theprotein in Glc1–3 at the DSP (Fig. S6). Thus, SEX4 accom-modates a helical glucan and may influence the geometry ofthe chain during binding.The structure reveals that SEX4 accommodates six glucan

monomers within its extended active site, with Glc2 positionedover the active site. Because both ends of the bound glucanchain are directed toward the solvent, it is possible that SEX4can accommodate register shifts that would facilitate proc-essivity. Given the domain-specific contributions to glucanbinding and positioning, it is possible that SEX4 is able todiffuse along the glucan chain via the CBM and then integratea phosphoglucan into the catalytic site. In Arabidopsis leafstarch, the frequency of glucosyl phosphorylation is ∼1 in 2,000(18). Although phosphorylation is likely higher during starchdegradation, the challenges of locating a phosphorylated glu-cosyl residue may require this more coordinated, and possiblyprocessive, phosphoglucan engagement with the DSP–CBMoperating as an integrated unit.The DSP active site of SEX4 comprises DSP subdomains that

have been adapted to dephosphorylate starch via coordination ofaromatic and hydrophilic residues that form a glucan-bindingplatform. An aromatic platform was also found in LSF2, which is45% identical to the SEX4 DSP (26). Comparative analysis be-tween SEX4 and LSF2 reveals common features, as well asunique specificity-determining regions, in their active sites. Bothenzymes contain broad and shallow active sites that engage threeglucose moieties of an α-1,4-glycosidic–linked chain, but the activesite topology of each enzyme is finely tuned to direct glucan ori-entation. The LSF2 active site makes a substantial contribution toglucan binding, unlike the SEX4 active site (26). Thus, each glucanphosphatase contains common and specific elements that allow itto engage specific phosphoglucan substrates.The establishment of glucan-interacting motifs identified

in SEX4 and LSF2 provides a basis for comparison with thenoncatalytic DSP LSF1 and the human glycogen phosphataselaforin. In planta studies have demonstrated that LSF1 functionsin starch metabolism, although its precise role is unknown (46).

LSF1 is similar in domain architecture with SEX4, however thefailure of LSF1 to dephosphorylate glucan substrates may stemfrom the absence of glucan-interacting platforms found in LSF2and SEX4. Laforin dephosphorylates glycogen and possessesa CBM and DSP similar to SEX4, but the domains are in theopposite order (24, 33, 34). Mutations in the gene that encodeslaforin result in Lafora disease, in which cellular glycogen ishyperphosphorylated and forms amylopectin-like inclusions(47). Initial analyses suggest that some Lafora disease patientmutations are within putative glucan-interacting platformssimilar to those identified in SEX4. These aromatic- and hydro-philic-enriched, glucan-binding platforms represent a canonicaltheme in plant glucan phosphatases that may also be found inrelated enzymes.Reversible phosphorylation is critical for starch breakdown in

plants, and therefore represents a potential tool to modify thephysical properties of starch or increase its yield in industrial andagricultural settings (12, 41). The myriad food and nonfood starch-based industrial products require enzymatic and chemical modi-fication of starch to generate necessary starch-based feedstocks(48–50). Phosphorylation is the only known natural modificationof starch, and has direct influences on starch hydration, crystal-linity, freeze–thaw stability, viscosity, and transparency that arecentral to various commercial applications (39–41). Furthermore,it is clear that C6 and C3 phosphorylation of starch influence itsproperties differently, as studies show that C3 phosphate hasa more direct effect on starch granule solubilization (14). Thepresent study illustrates that the C6 and C3 specificity of SEX4 canbe altered via the engineering of a discrete number of DSP resi-dues and reversed site-specific activity in SEX4 would certainlyinfluence the pattern of starch metabolism in planta. EngineeredSEX4 could provide a means to generate designer starches withtailored patterns of phosphorylation and physical characteristicsuseful in industrial settings. Recently, GWD-silenced wheat plantswere shown to have a significant increase in starch yield (51), andsilencing or engineering SEX4 activity in crop plants may pro-duce similarly favorable results. Elucidation of the structuralbasis of SEX4 activity provides valuable insights necessary for itsapplication in biotechnology.

Fig. 4. Structural basis for C6 specificity in SEX4. (A) Activesite of SEX4 consisting of the R motif, PTP loop, and V loop.Glc2 of the maltoheptaose chain (green) is oriented withthe O6 position toward the phosphate (orange) in thecatalytic site. R-motif residues F235 and V-loop residueF140 interact with Glc3 and Glc1, respectively. A cross-sectionof the active site surface (red lines) is overlaid onto theequivalent positions in the model. (B) Equivalent active siteof glucan phosphatase LSF2 (40). Glc3 of the maltohexaosechain (green) is oriented with the O3 position toward thephosphate (orange) in the catalytic site. R-motif residueG230 and V-loop residue W136 interact with Glc2 and Glc4,respectively. A cross-section of the active site surface (purplelines) is overlaid onto the equivalent positions in the model.(C) Superimposed cross-sections of the active site surfacefrom SEX4 (red) and LSF2 (purple). Positions of F235 andG230 (F and G) and F140 and W136 (F and W) in SEX4 andLSF2, respectively, are denoted. (D) Relative specific activityof SEX4 active-site mutants at the C6 (blue) and C3 positions(yellow) of Arabidopsis starch granules represented as thepercentage of total position dephosphorylation per minuteper microgram of protein. Error bars represent the ±SD of sixreplicates. Statistical comparison of wild-type and mutant C6and C3 activity demonstrates significant differences, P <0.001, between all constructs. (E) Representation of datafrom D as a percentage of C3-dephosphorylation relative tototal dephosphorylation.

7276 | www.pnas.org/cgi/doi/10.1073/pnas.1400757111 Meekins et al.

Materials and MethodsCloning, expression and purification of A. thaliana Δ89-SEX4 wild type andmutants (38), as well as recombinant GWD and PWD (19, 21), were performedas previously described. A. thaliana Δ89-SEX4 C198S (inactive mutant) proteinused for crystallization was preincubated with 25 mM maltoheptaose (Sigma-Aldrich). The crystallization screen trials crystals were set up via hanging dropvapor diffusion using a Mosquito liquid-handling robot (TTPLabtech) using a200-nL drop with a 1:1 ratio of condition and 12 mg/mL protein. Ligand-boundcrystals were obtained in 0.2 MMgCl2 and 20% (wt/vol) PEG 3350. Crystals werebriefly soaked in mother liquor with 20% (vol/vol) glycerol and flash-frozen.A single crystal was used for data collection and structural determination. Datawere collected on the 22-ID beamline of the Southeast Regional CollaborativeAccess Team (Advanced Photon Source, Argonne National Laboratory,Argonne, IL) (Table 1) at 120 K at a wavelength of 1.00 Å. There was onemolecule in the asymmetric unit and the structure was determined using

molecular replacement with the B chain of the previously determined SEX4structure as the search model (38). Phosphatase assays using pNPP (24, 33, 35),phosphate-free Arabidopsis sex1-3 starch (16, 22, 26), and the glucan-bindingassay (52) were performed as previously described. Additional details regardingprocedures and data analysis are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Drs. Samuel Zeeman and Diana Santeliaand members of the M.S.G. and C.W.V.K. laboratories for fruitful discussions,as well as Drs. Carol Beach and Martin Chow for technical assistance. Thisstudy was supported in part by a National Science Foundation CAREER GrantMCB-1252345 (to M.S.G.), National Institutes of Health Grants R01NS070899(to M.S.G.) and P20GM103486 [to M.S.G. and C.W.V.K (core support)],Kentucky Science and Energy Foundation Grant KSE-2268-RDE-014 (toM.S.G.), a University of Kentucky College of Medicine startup fund (M.S.G.),Eidgenössische Technische Hochschule Zürich (O.K.), and Swiss South AfricaJoint Research Program Grant IZ LS X3122916 (to O.K.).

1. Streb S, Zeeman SC (2012) Starch metabolism in Arabidopsis.Arabidopsis Book 10:e0160.2. Tester RF, Karkalas J, Qi X (2004) Starch—composition, fine structure and architecture.

J Cereal Sci 39:151–165.3. Gallant DJ, Bouchet B, Baldwin PM (1997) Microscopy of starch: Evidence of a new

level of granule organization. Carbohydr Polym 32:177–191.4. Buléon A, Colonna P, Planchot V, Ball S (1998) Starch granules: Structure and bio-

synthesis. Int J Biol Macromol 23(2):85–112.5. Pohu A, Planchot V, Putaux JL, Colonna P, Buléon A (2004) Split crystallization during

debranching of maltodextrins at high concentration by isoamylase. Biomacromolecules5(5):1792–1798.

6. Blennow A, Engelsen SB (2010) Helix-breaking news: Fighting crystalline starch energydeposits in the cell. Trends Plant Sci 15(4):236–240.

7. Fettke J, et al. (2009) Eukaryotic starch degradation: Integration of plastidial andcytosolic pathways. J Exp Bot 60(10):2907–2922.

8. Silver DM, Kötting O, Moorhead GB (2014) Phosphoglucan phosphatase functionsheds light on starch degradation. Trends Plant Sci, 10.1016/j.tplants.2014.01.008.

9. Engelsen SB, et al. (2003) The phosphorylation site in double helical amylopectin asinvestigated by a combined approach using chemical synthesis, crystallography andmolecular modeling. FEBS Lett 541(1-3):137–144.

10. Kötting O, Kossmann J, Zeeman SC, Lloyd JR (2010) Regulation of starch metabolism:The age of enlightenment? Curr Opin Plant Biol 13(3):321–329.

11. Yu TS, et al. (2001) The Arabidopsis sex1 mutant is defective in the R1 protein, a generalregulator of starch degradation in plants, and not in the chloroplast hexose transporter.Plant Cell 13(8):1907–1918.

12. Blennow A, Nielsen TH, Baunsgaard L, Mikkelsen R, Engelsen SB (2002) Starch phos-phorylation: A new front line in starch research. Trends Plant Sci 7(10):445–450.

13. Blennow A, Bay-Smidt AM, Olsen CE, Møller BL (2000) The distribution of covalentlybound phosphate in the starch granule in relation to starch crystallinity. Int J BiolMacromol 27(3):211–218.

14. Hansen PI, et al. (2009) Starch phosphorylation—maltosidic restrains upon 3′- and 6′-phosphorylation investigated by chemical synthesis, molecular dynamics and NMRspectroscopy. Biopolymers 91(3):179–193.

15. Edner C, et al. (2007) Glucan, water dikinase activity stimulates breakdown of starchgranules by plastidial beta-amylases. Plant Physiol 145(1):17–28.

16. Kötting O, et al. (2009) STARCH-EXCESS4 is a laforin-like Phosphoglucan phosphataserequired for starch degradation in Arabidopsis thaliana. Plant Cell 21(1):334–346.

17. Takeda Y, Hizukuri S (1981) Studies on starch phosphate. Part 5. Reexamination of theaction of sweet-potato beta-amylase on phophorylated (1->4)-a-D-glucan. CarbohydrRes 39:151–165.

18. Ritte G, et al. (2006) Phosphorylation of C6- and C3-positions of glucosyl residues instarch is catalysed by distinct dikinases. FEBS Lett 580(20):4872–4876.

19. Kötting O, et al. (2005) Identification of a novel enzyme required for starch metabolismin Arabidopsis leaves. The phosphoglucan, water dikinase. Plant Physiol 137(1):242–252.

20. Baunsgaard L, et al. (2005) A novel isoform of glucan, water dikinase phosphorylatespre-phosphorylated alpha-glucans and is involved in starch degradation in Arabi-dopsis. Plant J 41(4):595–605.

21. Ritte G, et al. (2002) The starch-related R1 protein is an alpha -glucan, water dikinase.Proc Natl Acad Sci USA 99(10):7166–7171.

22. Santelia D, et al. (2011) The phosphoglucan phosphatase like sex Four2 dephos-phorylates starch at the C3-position in Arabidopsis. Plant Cell 23(11):4096–4111.

23. Hejazi M, Fettke J, Kötting O, Zeeman SC, Steup M (2010) The Laforin-like dual-specificity phosphatase SEX4 from Arabidopsis hydrolyzes both C6- and C3-phosphateesters introduced by starch-related dikinases and thereby affects phase transition ofalpha-glucans. Plant Physiol 152(2):711–722.

24. Gentry MS, et al. (2007) The phosphatase laforin crosses evolutionary boundaries andlinks carbohydrate metabolism to neuronal disease. J Cell Biol 178(3):477–488.

25. Niittylä T, et al. (2006) Similar protein phosphatases control starch metabolism inplants and glycogen metabolism in mammals. J Biol Chem 281(17):11815–11818.

26. Meekins DA, et al. (2013) Structure of the Arabidopsis glucan phosphatase like sexfour2 reveals a unique mechanism for starch dephosphorylation. Plant Cell 25(6):2302–2314.

27. Zeeman SC, Northrop F, Smith AM, Rees T (1998) A starch-accumulating mutant ofArabidopsis thaliana deficient in a chloroplastic starch-hydrolysing enzyme. Plant J15(3):357–365.

28. Tonks NK (2013) Protein tyrosine phosphatases—from housekeeping enzymes tomaster regulators of signal transduction. FEBS J 280(2):346–378.

29. Yuvaniyama J, Denu JM, Dixon JE, Saper MA (1996) Crystal structure of the dualspecificity protein phosphatase VHR. Science 272(5266):1328–1331.

30. Moorhead GB, De Wever V, Templeton G, Kerk D (2009) Evolution of protein phos-phatases in plants and animals. Biochem J 417(2):401–409.

31. Tonks NK (2006) Protein tyrosine phosphatases: From genes, to function, to disease.Nat Rev Mol Cell Biol 7(11):833–846.

32. Alonso A, Rojas A, Godzik A, Mustelin T (2003) The Dual-Specific Protein TyrosinePhosphatse Family (Springer, Berlin).

33. Worby CA, Gentry MS, Dixon JE (2006) Laforin, a dual specificity phosphatase thatdephosphorylates complex carbohydrates. J Biol Chem 281(41):30412–30418.

34. Tagliabracci VS, et al. (2007) Laforin is a glycogen phosphatase, deficiency of whichleads to elevated phosphorylation of glycogen in vivo. Proc Natl Acad Sci USA 104(49):19262–19266.

35. Gentry MS, Pace RM (2009) Conservation of the glucan phosphatase laforin is linkedto rates of molecular evolution and the glucan metabolism of the organism. BMC EvolBiol 9:138.

36. Serratosa JM, et al. (1999) A novel protein tyrosine phosphatase gene is mutatedin progressive myoclonus epilepsy of the Lafora type (EPM2). Hum Mol Genet 8(2):345–352.

37. Minassian BA, et al. (1998) Mutations in a gene encoding a novel protein tyrosinephosphatase cause progressive myoclonus epilepsy. Nat Genet 20(2):171–174.

38. Vander Kooi CW, et al. (2010) Structural basis for the glucan phosphatase activity ofStarch Excess4. Proc Natl Acad Sci USA 107(35):15379–15384.

39. Blennow A, Mette Bay-Smidt A, Bauer R (2001) Amylopectin aggregation as a func-tion of starch phosphate content studied by size exclusion chromatography and on-line refractive index and light scattering. Int J Biol Macromol 28(5):409–420.

40. Muhrbeck PEA (1991) Influence of the naturally-occuring phosphate esters on thecrystallinity of potato starch. J Sci Food Agric 55:13–18.

41. Santelia D, Zeeman SC (2011) Progress in Arabidopsis starch research and potentialbiotechnological applications. Curr Opin Biotechnol 22(2):271–280.

42. Janecek S, Svensson B, MacGregor EA (2011) Structural and evolutionary aspects oftwo families of non-catalytic domains present in starch and glycogen binding proteinsfrom microbes, plants and animals. Enzyme Microb Technol 49(5):429–440.

43. Cantarel BL, et al. (2009) The Carbohydrate-Active EnZymes database (CAZy):An expert resource for Glycogenomics. Nucleic Acids Res 37(Database issue):D233–D238.

44. Polekhina G, et al. (2005) Structural basis for glycogen recognition by AMP-activatedprotein kinase. Structure 13(10):1453–1462.

45. Andersen JN, et al. (2001) Structural and evolutionary relationships among proteintyrosine phosphatase domains. Mol Cell Biol 21(21):7117–7136.

46. Comparot-Moss S, et al. (2010) A putative phosphatase, LSF1, is required for normalstarch turnover in Arabidopsis leaves. Plant Physiol 152(2):685–697.

47. Gentry MS, Dixon JE, Worby CA (2009) Lafora disease: Insights into neurodegenerationfrom plant metabolism. Trends Biochem Sci 34(12):628–639.

48. Jobling S (2004) Improving starch for food and industrial applications. Curr Opin PlantBiol 7(2):210–218.

49. Tharanathan RN (2005) Starch—value addition by modification. Crit Rev Food Sci Nutr45(5):371–384.

50. Zeeman SC, Kossmann J, Smith AM (2010) Starch: Its metabolism, evolution, andbiotechnological modification in plants. Annu Rev Plant Biol 61:209–234.

51. Ral JP, et al. (2012) Down-regulation of Glucan, Water-Dikinase activity in wheatendosperm increases vegetative biomass and yield. Plant Biotechnol J 10(7):871–882.

52. McBride A, Ghilagaber S, Nikolaev A, Hardie DG (2009) The glycogen-binding domainon the AMPK beta subunit allows the kinase to act as a glycogen sensor. Cell Metab9(1):23–34.

Meekins et al. PNAS | May 20, 2014 | vol. 111 | no. 20 | 7277

BIOCH

EMISTR

Y