A two-phase model for the non-processive biosynthesis of homogalacturonan polysaccharides by the GAUT1:GAUT7 complex

Homogalacturonan (HG) is a pectic glycan in the plant cell wall that contributes to plant growth and development and cell wall structure and function, and interacts with other glycans and proteoglycans in the wall. HG is synthesized by the galacturonosyltransferase (GAUT) gene family. Two members of this family, GAUT1 and GAUT7, form a heteromeric enzyme complex in Arabidopsis thaliana. Here, we established a heterologous GAUT expression system in HEK293 cells and show that co-expression of recombinant GAUT1 with GAUT7 results in the production of a soluble GAUT1:GAUT7 complex that catalyzes elongation of HG products in vitro. The reaction rates, progress curves, and product distributions exhibited major differences dependent upon small changes in the degree of polymerization (DP) of the oligosaccharide acceptor. GAUT1:GAUT7 displayed >45-fold increased catalytic efficiency with DP11 acceptors relative to DP7 acceptors. Although GAUT1:GAUT7 synthesized high-molecular-weight polymeric HG (>100 kDa) in a substrate concentration–dependent manner typical of distributive (nonprocessive) glycosyltransferases with DP11 acceptors, reactions primed with short-chain acceptors resulted in a bimodal product distribution of glycan products that has previously been reported as evidence for a processive model of GT elongation. As an alternative to the processive glycosyltransfer model, a two-phase distributive elongation model is proposed in which a slow phase, which includes the de novo initiation of HG and elongation of short-chain acceptors, is distinguished from a phase of rapid elongation of intermediate- and long-chain acceptors. Upon reaching a critical chain length of DP11, GAUT1:GAUT7 elongates HG to high-molecular-weight products.

Homogalacturonan (HG) 3 is a plant cell wall polysaccharide and glycan component of more complex polysaccharides and proteoglycans that contributes to the structure and mechanical strength of the wall and has roles in plant growth, development, morphology, and response to biotic and abiotic stress (1)(2)(3)(4). HG is a linear homopolymer of 1,4-linked ␣-D-galactopyranosyluronic acid (GalA) that may be partially methylesterified at O-6 and acetylated at O-2 and O-3. It is the simplest and most abundant glycan in the family of cell wall polysaccharides known as pectins, which include HG, rhamnogalacturonan I and rhamnogalacturonan II (1).
The pectic polysaccharides make contacts with each other (1), cellulose (5-7), hemicelluloses (1,5), and proteoglycans (2) within the wall through covalent and noncovalent interactions. The presence of pectin provides a barrier to the deconstruction of cellulosic biomass and partially obstructs degradative enzymes from accessing cellulose and hemicelluloses during the processing and saccharification of plant biomass (1, 8 -10). Reducing the content of pectin in woody and grass biofuel feedstocks has been shown to enhance biofuel production by increasing both biomass yield and the recovery of sugar (10 -12). Our current understanding of pectin synthesis and structure, however, is insufficient to explain its diverse functions in the plant or how reductions to pectin synthesis correlate with phenotypes such as increased plant growth and reduced recalcitrance to deconstruction.
The first gene identified encoding an HG biosynthetic enzyme was galacturonosyltransferase 1 (GAUT1), a CAZy (13) family 8 glycosyltransferase (GT). The in vitro synthesis of HG or homogalacturonan:galacturonosyltransferase (HG:GalAT) activity was mapped to GAUT1 following MS sequencing of Arabidopsis thaliana solubilized membrane proteins (1,14). Subsequent co-immunoprecipitation, MS sequencing, and bimolecular fluorescence complementation revealed that GAUT1 functions as a disulfide-linked heterocomplex with the homologous protein, GAUT7 (15). In vivo, GAUT1 is truncated at its N terminus by 167 residues and requires GAUT7 for localization in the Golgi (15). Based on these results, GAUT7 was proposed to be a noncatalytic membrane anchor for GAUT1 (14,15). A more detailed study of the function of the GAUT1:GAUT7 complex and the mechanism of HG synthesis required the establishment of a recombinant protein expression system that could co-express GAUT1 and GAUT7 and properly post-translationally fold and process the active GT complex.
Progress in studying plant cell wall biosynthesis has been hampered by difficulties associated with expressing and biochemically characterizing the relevant glycosyltransferases (16,17). Because plant cell wall biosynthetic GTs are typically large N-glycosylated proteins that contain hydrophobic transmembrane regions and may exist as multienzyme complexes, most successful purifications of such recombinant GTs have been achieved using eukaryotic expression systems, namely Nicotiana benthamiana (18 -21), Pichia pastoris (16,22,23), and HEK293 cells (14,24,25), with the latter being a particularly successful host for recombinant expression and enzymatic characterization of soluble forms of eukaryotic GTs (26 -32).
Here, we show that the HEK293 cell system can be used to express the heteromeric, disulfide-linked GAUT1: GAUT7 complex from A. thaliana in sufficient quantities to enable a comprehensive characterization of its enzymatic properties.
Heteromeric and homomeric GT complexes have been identified in glycan biosynthetic pathways shared by model eukaryotic organisms, including Homo sapiens, Saccharomyces cerevisiae, and A. thaliana (33). GT complexes have diverse biological functions and appear in N-and O-linked glycan, proteoglycan, and glycolipid synthesis pathways (33). In addition to the GAUT1:GAUT7 complex, at least six examples of proven and putative GT complexes involved in the synthesis of plant cell wall glycans are known (34 -36). Various hypotheses for the biological significance of GT complexes have been proposed, including enhancement of enzymatic activity, substrate channeling, and Golgi localization (33,34).
Here we demonstrate that the GAUT1:GAUT7 complex can synthesize high-molecular-weight (MW) HG polysaccharides in vitro, and we expand upon the original model of the GAUT1: GAUT7 complex (15). We propose a two-phase model of polysaccharide elongation in which short-chain acceptors are elongated slowly, followed by a rapid elongation phase once the glycans reach an intermediate, critical degree of polymerization. This model reconciles data showing that HG extracted from plant cell walls consists of long polymers containing Ͼ100 GalA units (37) with reports that in vitro HG elongation occurs through a nonprocessive mechanism (37)(38)(39)(40). The activity reported also provides a basis for the initiation of HG polysaccharides by the GAUT1:GAUT7 complex and suggests that GAUT7 has a previously unrecognized role in contributing to the synthesis of high-MW polysaccharides.

GAUT1 and GAUT7 form a complex when co-expressed in HEK293F cells
The co-immunoprecipitation of GAUT1 with GAUT7 from A. thaliana solubilized membranes revealed that GAUT1 and GAUT7 function as a GAUT1:GAUT7 heterocomplex covalently linked by disulfide bonds (15). In an effort to obtain sufficient amounts of purified, soluble GAUT1:GAUT7 complex, recombinant GAUT1 and GAUT7 constructs were heterologously co-expressed in HEK293F suspension culture cells, and the GAUT1:GAUT7 complex was characterized following purification.
Individual GAUT1 and GAUT7 constructs lacking their N-terminal transmembrane domains (GAUT1⌬167 and GAUT7⌬ 43) were generated as secreted fusion proteins harboring chimeric N-terminal fusion tags. GAUT1 purified from A. thaliana membranes is N-terminally truncated by 167 residues in vivo (15). The fusion tags were composed of a signal sequence, His 8 tag, AviTag, superfolder GFP, and tobacco etch virus (TEV) protease recognition site followed by the respective GAUT domains using a strategy previously employed for the expression of a large library of mammalian glycosylation enzyme expression constructs (32). The constructs were transfected into HEK293F cells, and the secreted fusion proteins were purified from the medium by Ni 2ϩ -Sepharose affinity chromatography.
Co-expression of GAUT1⌬167 and GAUT7⌬43 (hereafter referred to as GAUT1 and GAUT7) resulted in the production of secreted GAUT1:GAUT7 complex (Fig. 1A, lane 7), which required reducing conditions (ϩDTT) to separate GAUT1 and GAUT7 into monomers (Fig. 1A, lane 2). Treatment of the complex with peptide:N-glycosidase F (PNGase F) led to a reduction in the molecular weight of both GAUT1 and GAUT7, indicating that both proteins are N-glycosylated (Fig. 1A, lanes 3 and 5). From a 1-liter culture, a total of 47.5 mg of the enzyme complex was purified.
Expression of GAUT1 alone resulted in the secretion of a GAUT1 homocomplex that could be purified and identified by SDS-PAGE under nonreducing conditions (Fig. 1B, lane 5). However, heterologous expression of GAUT1 in the absence of GAUT7 also resulted in the formation of large amounts of truncation products and high-molecular weight aggregates (Fig. 1B,  lanes 3 and 5). Expression of GAUT7 alone led to no detectable secretion of the GAUT7 fusion protein into the culture medium. These results suggest that co-expression of GAUT1 and GAUT7 and the formation of a disulfide-bonded heterocomplex is required for proper folding and secretion of the active GAUT1:GAUT7 complex.

Recombinant GAUT1:GAUT7 catalyzes the transfer of GalA residues from UDP-GalA onto homogalacturonan acceptors
The GAUT1:GAUT7 complex isolated from A. thaliana membranes has previously been shown to transfer GalA from UDP-GalA onto HG acceptors (HG:GalAT activity) (15). Here, we employed similar HG:GalAT acceptor-dependent activity assays to examine HG extension by the recombinant GAUT1: GAUT7 complex produced in HEK293F cells. The co-ex-

Mechanism of homogalacturonan synthesis
pressed complex, containing intact N-terminal tags and N-glycosylation, was used in all assays.
The activity of the purified enzyme was determined by radioactive incorporation of 14 C-labeled GalA from UDP-[ 14 C]GalA onto the nonreducing ends of an HG acceptor mix enriched for HG oligosaccharides with a degree of polymerization (DP) of 7-23 (15,39,41). To assess the stability of the enzyme, GAUT1: GAUT7 activity was assayed immediately after Ni 2ϩ -Sepharose affinity purification and also after 3 days of storage at Ϫ80°C. GAUT1:GAUT7 retained 100% activity after storage at Ϫ80°C ( Fig. 2A). Single-thawed aliquots retained ϳ40% activity after storage for 15 months. The enzyme complex has a pH optimum of 7.2 ( Fig. 2B) and shows maximal activity in the presence of 0.1-1.0 mM MnCl 2 (Fig. 2C). A lower level of activity was obtained when GAUT1:GAUT7 was assayed with CoCl 2 (Fig. 2C). Manganese ion cofactors have been visualized in X-ray crystal structures of related GT8 enzymes, LgtC (42) and glycogenin (43), and make active-site contacts with UDP-sugar substrates. All of the other divalent and monovalent metal cations tested (NiSO 4 , FeCl 2 , CuSO 4 , CaCl 2 , ZnSO 4 , MgCl 2 , NaCl, and KCl) yielded less than 5% of the activity observed with MnCl 2 (Fig. S1).
Based on the assays outlined above, a standard reaction condition was defined: 5 min reactions containing 100 nM GAUT1: GAUT7, 1 mM UDP-GalA, 10 M HG acceptor, and 0.25 mM MnCl 2 in a pH 7.2 HEPES buffer containing 0.05% BSA. Using the standard conditions outlined, Michaelis-Menten kinetics were measured for the donor, UDP-GalA, and for the HG acceptor mixture with the nonvariable substrate held at saturating conditions. The GAUT1:GAUT7 complex shows a standard hyperbolic Michaelis-Menten curve for UDP-GalA with a K m of 151 M (Fig. 2D). Substrate inhibition was observed at HG acceptor concentrations Ͼ 5 M (Fig. 2E). Substrate inhibition at high concentrations of HG acceptor is expected if GAUT1:GAUT7 functions through an ordered kinetic mechanism in which the UDP-GalA donor binds to the active site prior to the HG acceptor (44,45). This ordered substrate binding scheme has also been shown in crystal structures and inhibition assays for homologous GT8-family enzymes (42,43,46,47). k cat values for the elongation of HG acceptors ranged from 0.92-1.99 s Ϫ1 . Results from six independent assays are summarized in Table S1.

GAUT1:GAUT7 synthesizes high molecular weight polysaccharides in vitro by elongation of medium-chain HG acceptors (DP > 11) using a distributive mechanism
The mechanism of HG backbone synthesis has not been clearly defined due to conflicting results from prior studies of this activity. HG:GalAT activity from detergent-solubilized microsomal membranes and from GAUT1:GAUT7 partially purified by immunoprecipitation has previously been shown to add between 1 to ϳ30 GalA residues onto DP13-15 acceptors (14,(37)(38)(39), without detection of high-MW polymeric HG. These results suggested that GAUT1:GAUT7 uses a distributive mechanism, in which single GalA units are added during each catalytic event, followed by the release of the HG acceptor, and also that GAUT1:GAUT7 does not synthesize high-MW HG polymers in vitro. In contrast, intact Nicotiana tabacum microsomal membranes produce an HG-containing product of ϳ105 kDa (39,48). However, in those experiments, it was not shown whether the high-MW product was free HG polysaccharide or a more complex glycan or proteoglycan containing HG. The synthesis of high-MW HG polysaccharides has not yet been described from in vivo or in vitro experiments.
We used recombinant GAUT1:GAUT7 and HG acceptors enriched for homogeneous degrees of polymerization to measure the size of products synthesized by HG:GalATs. We reasoned that if the GAUT1:GAUT7 complex elongates HG acceptors by a distributive mechanism, the chain length of the products synthesized should be dependent on the ratio of donor to acceptor substrates (49,50).
The degree of HG acceptor elongation by GAUT1:GAUT7 in vitro was determined by size-exclusion chromatography, highpercentage PAGE, and MALDI-TOF. Using 1 mM UDP-GalA and varying amounts of DP11 HG acceptor, reactions were performed under two conditions: 10 M HG acceptor (100:1 molar excess of UDP-GalA donor) (Fig. 3, A and B) and 100 M HG acceptor (10:1 molar excess of UDP-GalA donor) (Fig. 3, C and D). DP11 acceptors were rapidly elongated under both assay conditions. The products synthesized were larger at all time points when a 100:1 molar excess of UDP-GalA over HG acceptor was used (Fig. 3, A and B). The high-MW products are

Mechanism of homogalacturonan synthesis
measured to be Ͼ100 kDa, but pectin MW may be overestimated relative to dextran standards due to the potential for aggregation or anomalous behavior in size-exclusion columns (51). As observed in high-percentage polyacrylamide gels in which individual bands corresponding to HG products up to DP30 could be distinguished, the DP11 acceptor was elongated to an estimated DP of 30 -50 (Ͻ10 kDa) when incubated at a 10:1 donor/acceptor ratio (Fig. 3D). Comparison of these two reaction conditions showed that in vitro HG synthesis by GAUT1:GAUT7 matches the results expected for a distributive GT mechanism in which the narrow product distribution has a DP dependent on the available donor/acceptor ratio, as discussed for polysialyltransferases (49). Overnight incubation of GAUT1:GAUT7 with donor/acceptor ratios ranging from 10 to 10,000 further demonstrated that GAUT1:GAUT7 synthesizes high-MW products given available UDP-GalA (Fig. S2).
Labeling of the reducing ends of HG acceptors with the fluorescent tag 2-aminobenzamide (2-AB) has been used to detect HG elongation using high-performance anion-exchange chromatography (40). Fluorescent labeling of oligosaccharides has also been shown to enhance detection of oligosaccharides by MALDI-TOF (24). We therefore used MALDI-TOF to test the pattern of elongation of a DP15 fluorescently labeled HG acceptor in reactions incubated from 5 min up to 4 h. Over time, products up to DP30 were detected (Fig. 3E). The apparent Poisson distribution of HG products formed over time is also typical of a distributive mechanism of synthesis (24). However, it should be noted that, as observed in polyacrylamide gels (Fig.  3D), products larger than DP30 were also synthesized under these reaction conditions, indicating that the apparent Poisson distribution observed by MALDI-TOF was partially due to the loss of higher-MW signals. Longer-chain oligosaccharides may be poorly ionized or may have precipitated when exposed to acidic conditions during the labeling procedure.

Elongation of short-chain HG acceptors (DP < 7) occurs with low efficiency relative to longer-chain acceptors (DP > 11)
Prior assays of HG:GalAT activity from N. tabacum and Petunia axillaris solubilized membranes demonstrated a preference of the enzyme for exogenous HG acceptors of DP Ն 10 -12,  Table S1.

Mechanism of homogalacturonan synthesis
but also indicated that acceptors as short as DP5 can be elongated (38,39). To more clearly define the acceptor specificity of GAUT1: GAUT7, we compared the rates of elongation of homogenous HG acceptors enriched for DPs of 3, 7, 11, and 15. GAUT1:GAUT7 elongates HG acceptors at least as small as DP3 (Fig. 4A). GalA transfer is most rapid using longer-chain acceptors (DP Ն 11). The rate of transfer to DP11 and DP15 acceptors was approximately equal from 5 to 60 min. Activity was detected using shorter acceptors (DP7 and DP3), but the initial rates of synthesis were low relative to the longer-DP acceptors. The highest rates of synthesis for longer-chain acceptors were detected in the initial linear phase of the reaction, as expected for a standard steady-state reaction progress curve. A lag phase was observed during the elongation of DP7 and DP3 acceptors, and longer incubation periods were required to detect above-background levels of activity.
Reaction kinetics were measured using increasing amounts of DP11 (Fig. 4B) and DP7 (Fig. 4C) HG acceptors. The reaction kinetics using the DP11 HG acceptor appear similar to the results observed using the HG acceptor mix (Fig. 2E), with inhibition observed at concentrations above 5 M. Because activity with the DP7 acceptor was near background level when measured at 5 min, a 30-min incubation time was used. Standard hyperbolic Michaelis-Menten kinetics were observed under these conditions. The catalytic efficiency, defined as k cat /K m , of a DP11 acceptor was 45-fold higher than a DP7 acceptor, representing a strong preference for longer-chain acceptors. This measurement may underestimate the difference in catalytic efficiency between DP7 and DP11 acceptors because the low initial rates and the lag phase observed only with short-chain acceptors makes it difficult to calculate true initial rates of synthesis.

Elongation of short-chain HG acceptors and de novo synthesis in the absence of exogenous oligosaccharide acceptors leads to a bimodal product distribution with minimal observable intermediates
As shown in Fig. 4, initial rates of synthesis using short-chain HG acceptors (DP Յ 7) are low relative to longer-chain accep-

Mechanism of homogalacturonan synthesis
tors. To test whether the DP7 acceptor is elongated by a different reaction mechanism distinct from longer-chain acceptors, the degree of DP7 acceptor elongation was measured by size-exclusion chromatography, high-percentage PAGE, and MALDI-TOF. Reactions were carried out using 100 M acceptor (10:1 molar excess of UDP-GalA donor) (Fig. 5, A and B).
Unlike the results observed using a DP11 HG acceptor (Fig.  3), a bimodal distribution of high-MW and short-chain products was synthesized during the elongation of a DP7 acceptor, even under conditions of a low molar ratio of UDP-GalA to acceptor. Polysaccharides of intermediate molecular weights were not observed. When detecting low-MW products, the addition of only 1-2 GalA residues was observed at all time points (Fig. 5B and Fig. S3B). Unlike DP11 elongation, a Poisson distribution of HG oligomers was not observed. Instead, the intensity of high MW products increased with reaction time from 1 to 12 h. A similar bimodal product distribution was observed with donor/acceptor ratios ranging from 10 to 200 (Fig. S4).
Having identified that GAUT1:GAUT7 can elongate shortchain acceptors as small as DP3, the complex was tested for the ability to initiate HG synthesis de novo in the absence of exogenously added acceptors. Prior attempts to identify de novo initiation of HG using detergent-solubilized membrane fractions yielded no measurable activity, leading to the conclusion that the enzyme only functions in the elongation step of HG biosynthesis (1).
Three monoclonal antibodies previously shown to react with HG (CCRC-M38, CCRC-M131, and JIM5) (52) were used in ELISAs to determine whether GAUT1:GAUT7 could synthesize HG de novo. Following incubation with 1 mM UDP-GalA, all three anti-HG antibodies reacted with GAUT1:GAUT7synthesized product in a time-and concentration-dependent manner (Fig. 5C). Control ELISAs confirmed that the anti-HG antibodies detected HG acceptor controls but had no reactivity toward GAUT1:GAUT7 itself or toward UDP-GalA (Fig. 5D). To the best of our knowledge, this is the first evidence that GAUT1:GAUT7 can synthesize HG de novo.
The de novo synthesis product synthesized in overnight reactions was of high MW, similar to the products of overnight reactions with DP7 and DP3 acceptors (Fig. 5E). The de novo synthesis product is of a more uniform MW than the relatively polydisperse product formed following elongation of DP7 acceptors. The polysaccharides synthesized de novo were sensitive to digestion by the HG-specific enzyme, endopolygalacturonase (Fig. S5). These results demonstrate that in addition to elongation of acceptors of all degrees of polymerization, GAUT1:GAUT7 can initiate the synthesis of high-MW HG de novo.

A two-phase model of distributive elongation is favored over a processive model for HG synthesis
During in vitro HG polymerization by GAUT1:GAUT7, small differences in the chain length of HG acceptors appeared to affect the mechanism of elongation. We investigated whether short-and long-chain acceptors are elongated using distinct mechanisms by further defining the effect of acceptor DP on product size distribution. GAUT1:GAUT7 was incubated for 12 h in the presence of a low donor/acceptor ratio (10:1) of intermediately sized HG acceptors ranging from DP7 to DP11 (Fig. 6A). As previously observed, elongation of the DP11 acceptor resulted in products ranging in size from DP11 to a DP of ϳ30 -50. The majority of the starting acceptor was incorporated into larger-sized HG products. In contrast, all  Table S1.

Mechanism of homogalacturonan synthesis
acceptors of DP Յ 10 were elongated relatively poorly, as indicated by the appreciable amount of acceptors remaining unelongated even following a 12-h incubation. Quantitation using fluorescently tagged acceptors indicated that Ͻ3% of the original DP7 acceptor was elongated to high-MW products in overnight reactions (Fig. S7).
Product size was inversely related to the size of the HG acceptor, with high-MW polymeric HG observed following incubation with smaller acceptors. Contrary to the DP10 and DP11 acceptors, elongation of DP7 and DP8 acceptors resulted in a bimodal product distribution, with an intermediate distribution for DP9 (Fig. 6A).
Three observations appeared to support the hypothesis that GAUT1:GAUT7 uses distinct processive and distributive mechanisms, depending on the chain length of the acceptor, and that short-chain acceptors were elongated by a processive  Fig. 3B. C, detection of HG polysaccharides synthesized de novo by ELISA. Reactions (30-l total volume) containing 100 nM GAUT1:GAUT7 and 1 mM UDP-GalA were incubated for 5 min or 12 h. At the indicated time points, aliquots containing a 1-or 0.1-l reaction volume ( 1 ⁄30 or 1 ⁄300 of the total reaction volume) were boiled and spotted onto a 96-well plate. Reaction products were detected by anti-HG antibodies CCRC-M38, CCRC-M131, and JIM5 but showed no reactivity toward anti-xylan antibody CCRC-M149. D, anti-HG antibody ELISA controls. In C and D, error bars represent the S.D. from a total of four replicate measurements from two independent experiments. E, high-percentage polyacrylamide gel of products synthesized de novo by GAUT1:GAUT7 or in the presence of DP3 or DP7 acceptors, stained with a combination of alcian blue/silver. Reactions (30-l total volume) containing 100 M DP7, 100 M DP3, or no acceptor (de novo synthesis) were incubated for 24 h, and then 5 l was removed for separation on a 30% polyacrylamide gel.

Mechanism of homogalacturonan synthesis
mechanism. First, a bimodal product distribution formed over time (Fig. 5, B and E). The archetypal processive model for GT activity proposes that a single acceptor molecule remains tightly bound without dissociating from the enzyme until after many rounds of elongation, leading to the formation of high-MW products with minimal observable intermediates (49,50,53). Second, a lag phase was observed during early points in the progress curve (Fig. 4A). The lag phase has been proposed to be a feature common to processive enzymes due to the low affinity of short-chain acceptors that are not capable of filling a requisite number of "acceptor subsites" within an extended active-site domain (50,54). Third, the amount of acceptors within the starting pool used during the reaction was low, measured at Ͻ3% (Fig. S7). In traditional measures of processivity, if Ͻ10% of the total starting acceptor pool is elongated, then it has been assumed that each acceptor has only associated with the enzyme in a single priming event (53,55). All three of these observations were suggestive of the processive model but did not directly demonstrate the existence of tight enzyme-acceptor binding complexes demanded by that model. Because these results were inconsistent with the rapid elongation of DP Ն 11 acceptors in a nonprocessive manner, we considered that the processive model may not accurately describe the elongation of HG by GAUT1:GAUT7.
To directly test for evidence of a processive elongation mechanism in the presence of low-DP acceptors, a competition assay was designed. GAUT1:GAUT7 was preincubated with UDP-GalA and a DP7 acceptor for 15 min at a 1000-fold excess of acceptor over enzyme. We reasoned that 15 min was a sufficient preincubation time to allow all enzyme molecules to bind to the acceptor and to begin processive synthesis. Following preincu-bation, a DP11 acceptor was added, and incubation was continued for another 5 min (Fig. 6B, lanes 4 and 5). If the DP7 HG served as an acceptor for processive catalysis, the DP11 acceptor would not have been able to compete for binding to GAUT1: GAUT7 or to serve as an acceptor. The results show, however, that the DP11 HG acceptor was able to compete for enzyme binding and was elongated during the 5-min reaction period.
The DP11 acceptor was elongated to the same chain lengths as two control reactions: a standard 5-min reaction with no competing DP7 acceptor added during the preincubation period (Fig. 6B, lanes 6 and 7) and a 5-min reaction in which DP7 and DP11 acceptors were added at the same time (lanes 8 and 9). These results argue against the hypothesis that shortchain acceptors are synthesized by a processive mechanism, in which the growing acceptor should have formed tightly binding complexes with the enzyme, making it unavailable for binding to DP11 acceptors. Furthermore, there was no effect on the size distribution of the products synthesized by elongation of DP11 acceptors in the presence of DP7. This result is consistent with the enzyme having a strong preference for binding to longerchain acceptors and releasing the acceptor following each round of GalA transfer.
The competition assay argues against the hypothetical processive elongation model because the results suggest that GAUT1:GAUT7 and HG acceptors do not form tight enzymeacceptor binding complexes. An alternative, two-phase model proposes that acceptors of all sizes are elongated by a distributive mechanism and that the bimodal product distribution results from large differences in catalytic efficiency between shorter-and longer-chain acceptors. Small acceptors are inefficiently elongated by the enzyme, as evidenced by the Ͼ45-fold  6 and 7), no acceptor was added during the preincubation phase. DP11 acceptor was added after 15 min. For simultaneous incubation reaction controls (lanes 8 and 9), no acceptor was added during the preincubation phase. A mix of 100 M DP7 and the indicated DP11 acceptor was added after 15 min. DP11 standard (lane 10) shows the DP11 acceptor and the presence of background bands due to minor impurities.

Mechanism of homogalacturonan synthesis
difference in catalytic efficiency between DP7 and DP11 acceptors. Acceptors are only rapidly elongated after reaching a critical DP, estimated to be DP11 (Fig. 6A). The relative inefficiency of short-chain acceptors results in a slow rate of synthesis during the early phase of chain elongation. In reactions containing only DP7 acceptors, high-MW products are observed because there is an effective increase in the donor/acceptor ratio for the small number of acceptors that reach the critical DP, causing them to become rapidly elongated.

Synthesis of high-MW HG depends upon GAUT1:GAUT7 complex formation and electrostatic interactions
HG is a negatively charged polymer. It has been shown for other charged polymers, including polysialic acid (49) and DNA (56), that electrostatic substrate-enzyme interactions affect product size distribution and the mechanism by which enzymes maintain contact with their charged substrates. We hypothesized that electrostatic HG-enzyme interactions may be involved in acceptor binding and elongation by GAUT1: GAUT7. Prior evaluation of a homology model of the GAUT1 GT8 domain using LgtC from N. meningitidis as a template identified a patch of positively charged residues near the active site of GAUT1 that could create an extended acceptor binding groove (57). We predicted that incubation with NaCl would disrupt these interactions and limit the ability of GAUT1: GAUT7 to maintain the contacts with the HG acceptor needed for efficient transfer. We tested whether the addition of NaCl affected product formation by GAUT1:GAUT7.
The presence of 100 mM NaCl in reactions containing a 1000:1 ratio of UDP-GalA to DP11 HG acceptor inhibited the synthesis of high-MW HG products, even under long reaction times (Fig. 7A). In reactions containing a 10:1 donor/acceptor ratio and 0, 50, and 100 mM NaCl, elongation of a DP7 acceptor was likewise inhibited with increasing salt concentration (Fig.  7B). Individual bands of intermediate-sized products (DP Ͼ 20) were detectable following reactions containing NaCl, leading to the loss of the bimodal product distribution. This result also argues against DP7 being elongated processively. Short-chain elongation products were visible in both reactions, but the addition of NaCl to the reaction appears to prevent high-MW product formation.
Synthesis of intermediate-sized HG upon elongation of a 10:1 donor/acceptor ratio of DP11 acceptor was only weakly affected by the presence of NaCl (Fig. 7B). For both DP7 and DP11 acceptors, product size was limited to approximately DP30 -50 when GAUT1/GAUT7 was assayed with 100 mM NaCl. We propose that the electrostatic interactions guide the rapid elongation phase by orienting longer-chain acceptors within the active site for GalA transfer to the nonreducing end. The NaCl concentrations used here are within the physiological range and may not otherwise be expected to have such a strong inhibitory effect, as many glycosyltransferases are regularly purified, stored, and assayed under similar NaCl concentrations (32,58).
The possibility that previously unidentified structural domains are required for the synthesis of high-MW HG opens the possibility that GAUT7 also contributes to the function of the complex. The activity of GAUT1 expressed and purified in the absence of GAUT7 was tested. Upon incubation of GAUT1 with DP11 and DP7 acceptors (Fig. 7C), high-MW product was not observed. The size of the products synthesized by GAUT1 was limited to DP of ϳ30 -50, even in reactions containing a large excess of donor (10 M acceptor, 100:1 donor/acceptor ratio). The consistency of this result with the reduced product sizes observed following incubation of GAUT1:GAUT7 with NaCl suggests that GAUT7 also contributes to an acceptorbinding domain/pocket required for high-MW polymerization. Due to poor expression of GAUT1 and GAUT7 as individual enzymes, the nature of the contribution of GAUT7 to HG synthesis remains to be more thoroughly investigated.

Soluble expression of plant cell wall glycosyltransferases using the HEK293F cell system
Plant cell wall matrix polysaccharides are synthesized in the secretory pathway by the coordinated efforts of an estimated 200 glycosyltransferases that generate the backbone and sidechain linkages of pectins and hemicelluloses, including xylans and xyloglucan. A study published in 2009 (16) identified a total of nine plant cell wall GTs that had in vitro activities verified by expression in a heterologous system. In the years since that publication, numerous additional plant cell wall GT in vitro activities have been demonstrated and mapped to individual genes or multigene families (18 -24, 58 -65). Similar to activities assayed from native plant membranes, the use of recombinant enzymes produced in microsomal membranes from sources such as P. pastoris or N. benthamiana may suffer from difficulties including low protein yields and unknown concentrations of the protein of interest (61). Here, the HEK293F cell expression system was used for robust co-expression of a disulfide-linked enzymatically active GT complex in a soluble secreted form. In addition to the expression of the GAUT1: GAUT7 heterocomplex, the HEK293F cell system has been used to express the plant cell wall GTs xylan synthase-1 (Xys1) and fucosyltransferase 1 (Fut1) (24,25).

A model for HG synthesis by GAUT1:GAUT7
The results described here expand upon previous reports of HG synthesis by demonstrating that the GAUT1:GAUT7 complex can synthesize high-MW polysaccharides in vitro using a distributive mechanism. Previously, intact and partially solubilized N. tabacum membranes were shown to incorporate [ 14 C]GalA into high-MW polysaccharides with an apparent mass Ͼ 100 kDa (39,48). It was uncertain whether the large size of the polysaccharide products detected in cellular membrane preparations was due to the initiation of polymeric HG or due to incorporation of [ 14 C]GalA into large endogenous acceptors of an unknown size. We demonstrate here that as long as a sufficient concentration of UDP-GalA is available, recombinant GAUT1:GAUT7 is capable of producing high-MW polysaccharides in vitro.
A hypothetical model of distributive HG elongation is depicted in Fig. 8 as a series of five steps. Binding of UDP-GalA (step 1) is followed by HG acceptor binding (step 2). The binding of longer-chain acceptors is enhanced by structural features

Mechanism of homogalacturonan synthesis
of the GAUT1:GAUT7 complex, including charged interactions within an extended acceptor-binding groove that also appears to require the presence of GAUT7. GalA is transferred to the nonreducing end of the HG acceptor (step 3). The HG acceptor, elongated by a single GalA residue, departs from the active site (step 4), followed by departure of UDP (step 5). The conformational state of the enzyme is then reset for the next round of glycosyl transfer. Numbered GalA units in the acceptor molecule (center images) represent subsites within the proposed extended acceptor binding domain.
The two-phase model is more complete than traditional descriptions of distributive glycosyltransfer mechanisms because it accommodates the observation of distinct, acceptor size-dependent slow and rapid elongation phases. This phenomenon may be common to GTs but would not be observed unless the reaction progress is independently assayed with a wide range of oligosaccharide acceptors. Elongation of short-chain acceptors exhibits several characteristics that have previously been proposed to be common to processive GTs, including a bimodal product distribution, the lack of intermediate-sized products, a lag phase during the early time points of the reaction progress, and large proportions of the starting acceptors remaining unelongated (49,50,53,55,66). However, here we show that HG acceptors of DP Ն 11 are elongated in vitro by a distributive mechanism with greater catalytic efficiency than smaller HG acceptors.
For several polymerases that favor longer acceptors, a model has been proposed in which oligosaccharides must be elongated to a certain minimum length before the transferase exhibits maximum activity. Longer acceptors can fill acceptor-binding subsites within the active site groove (50,54,67). This "acceptor subsites" model is compatible with the slower rates of synthesis that have been observed with DP Յ 7 HG acceptors. Only acceptors longer than the critical DP can efficiently bind to the active site and be rapidly elongated. The nearly identical rates of transfer to DP11 or DP15 HG acceptors support an HG acceptor size of DP11 as being sufficient for maximum activity. The model that we present expands upon the hypothesis presented during the analysis of the processive bacterial galactan polymerase GlfT2 (50) because we argue that acceptor subsites and kinetic lag phases are features that are not necessarily limited to processive polymerases.
Processive polymerization mechanisms may require enzymes that physically constrain and enclose the growing acceptor chain, such as cellulose synthase (68), DNA-binding enzymes (56,69), or, putatively, multi-transmembrane-channel-forming GTs, such as the xyloglucan backbone synthase CSLC4 (70,71). Many type II GTs have a single GT domain and have no obvious structural basis for confining the growing glycan chain within a processive tunnel. Rather than inferring a processive model by induction from end-point product distribution data, the competition assay that we present (Fig. 6) is part of a recent push to find direct assays to test for the formation of enzyme:acceptor complexes that would serve as positive evidence for the archetypal processive model (50). The competition assay showed that longer, more efficient acceptors can compete for binding to the enzyme following a preincubation period sufficient for formation of processive enzyme:acceptor complexes. The results of this assay show that GAUT1:GAUT7, at least in vitro, does not remain tightly bound to the HG polymer during elongation, a key feature that distinguishes distributive from processive polymerases.
A two-phase model with distinct differences in elongation rates between short-chain and longer-chain acceptors has been described previously for the nonprocessive in vitro activity of K92 polysialyltransferase from Escherichia coli, which has a similar critical acceptor chain length, approximately DP10 -12 (72). Heparosan synthase from Pasteurella multocida provides a similar example of a nonprocessive transferase in which a slow initiation phase is observed (73). Two-phase elongation provides an alternative hypothesis for the interpretation of bimodal or polydisperse product distributions that may be observed in studies of GT activity. As an example, levansucrase from Bacillus subtilis was determined to have a processive activity on the basis of a bimodal product distribution, which was lost upon the addition of DP16 acceptors (66). If levansucrase functions similarly to GAUT1:GAUT7, then longer-chain acceptors are more efficiently elongated. The bimodal product distribution would result from large differences in catalytic efficiency and acceptor preference based on chain length. The buildup of inefficient short-chain acceptors and rapid elongation as soon as the acceptor reaches a critical DP can explain non-Poissonian product distributions for GTs that have not been directly shown to be processive.
In vitro polymer initiation and extension in the absence of a glycan primer has been demonstrated for at least four bacterial capsular polysaccharides (74 -77), including two negatively charged polysaccharides: Neisseria meningitidis serogroup X and serogroup A. The de novo synthesis products are both synthesized as high-MW polysaccharides with relatively homogeneous product dispersity compared with acceptor-primed reactions (75,76). For the de novo initiation of HG polysaccharides, a proposed candidate for the primer is a molecule of UDP-GalA, similar to elongation of ␤-1,4-linked GlcNAc oligomers synthesized by hyaluronan synthase that retain UDP at the reducing end (78,79). Alternatively, the primer could be monomeric GalA formed following hydrolysis of UDP-GalA. The large size of the products synthesized de novo, containing several hundred GalA units per reducing end, have prevented the primer for de novo synthesis from being identified in the current study.
Anionic polymers, such as HG, may have electrostatic enzyme-acceptor interactions that aid polymerization by promoting the proper binding and orientation of growing acceptor chains. The proposed positively charged acceptor-binding groove was identified using a homology model of the GT8 domain of GAUT1 (57). Future structural and acceptor binding studies will be necessary to confirm the existence of this extended binding groove. Basic residues located in the acceptor-binding pocket influence processivity and product dispersity in a polysialyltransferase from N. meningitidis (49). Incubation with NaCl disrupts polymer sliding in enzymes that bind to DNA, such as the endonuclease BamHI (56). In the case of the nonprocessive activity of GAUT1:GAUT7, strong inhibition of activity by NaCl may be due to disruption of electrostatic inter-

Mechanism of homogalacturonan synthesis
actions that contribute to the ability of acceptors to bind to the charged acceptor-binding groove.
The model described in this report provides a basis for the synthesis of high MW HG polysaccharides. GAUT1:GAUT7 uses a distributive mechanism in which rapid elongation requires that the acceptor reach an intermediate chain length, approximately DP11. The significance of this acceptor size, including comparisons with other GTs that function using multiphase elongation mechanisms, will require further investigation. The relative inefficiency of elongation of short-chain acceptors may provide a mechanism by which HG polymer size and chain initiation are regulated in vivo. This model is consistent with previous reports that HG:GalAT activity is nonprocessive (37)(38)(39)(40). Whereas the precise role of GAUT7 remains uncertain, GAUT1 appears to be unable to synthesize high-MW products in the absence of GAUT7. Proper folding and secretion of the complex requires co-expression of GAUT1 with GAUT7, but GAUT7 also appears to have a functional role in contributing to HG polymer synthesis. The results reported here provide a comprehensive characterization of the in vitro distributive activity of a heterologously expressed plant cell wall polysaccharide biosynthetic GT complex, with implications for investigations of kinetic mechanisms and processivity of GT heterocomplexes.
Sequences were verified following cloning into entry and expression vectors. Expression plasmids were purified using Purelink HiPure Plasmid Gigaprep kits (Invitrogen). HEK293F suspension culture cells grown to a cell density of 2.5 ϫ 10 6 in Freestyle 293 expression medium (Thermo Fisher Scientific) were used for transfection. Transfection of DNA, either GAUT1⌬167 or GAUT1⌬167 co-transfected with GAUT7⌬43, was done at a total concentration of 3 g/ml total culture volume with polyethyleneimine (9 g/ml). Cells were incubated in a humidified shaking 37°C CO 2 incubator at 150 rpm for 24 h before 1:1 dilution in Freestyle 293 medium supplemented to a final concentration of 2.2 mM valproic acid (Sigma). Medium containing secreted protein was collected after a 6-day total incubation time.
Secreted proteins were purified from suspension culture medium by nickel-affinity purification using HisTrap HP (GE Healthcare) columns connected to an Ä KTA FPLC system (GE Healthcare). Vacuum-filtered culture medium was injected into HisTrap HP columns at 1 ml/min, washed with column buffer, and eluted using a gradient from 20 -300 mM imidazole (20 column volumes). Fractions containing protein were exchanged into a storage buffer containing 50 mM HEPES, pH 7.2, 0.25 mM MnCl 2 , and 20% glycerol using a PD-10 desalting column (GE Healthcare) and concentrated using a 30-kDa MW cutoff Amicon Ultra centrifugal filter unit (Millipore).
The concentration of nickel-affinity purified proteins was determined by UV-visible spectroscopy (Nanodrop) using a 10-cm path length cuvette, and purification was confirmed by SDS-PAGE on a 4 -15% gradient gel (Bio-Rad).

Synthesis of UDP-[ 14 C]GalA and HG oligosaccharide acceptors
UDP-D-[ 14 C]galactopyranosyluronic acid was synthesized enzymatically from UDP-D-[ 14 C]glucopyranosyluronic acid (PerkinElmer Life Sciences) as described (15,80). The batchspecific activity value of 249 mCi/mmol was used to convert cpm readings from scintillation counting to reported pmol values in activity assay figures. Nonradiolabeled UDP-D-galactopyranosyluronic acid was purchased from CarboSource Services.
The HG acceptor mix, enriched for HG oligosaccharides of DP7-23, was generated by partial digestion of polygalacturonic acid with endopolygalacturonase, and the purity was confirmed by HPAEC-PAD as described (39). HG acceptors enriched for homogeneous degrees of polymerization of 7-15 were purified by HPAEC-PAD as described (39). Trigalacturonic acid (DP3) was purchased from Sigma.

Deglycosylation, fusion tag removal, and protein gel electrophoresis
Recombinant TEV protease and PNGase F were expressed as N-terminal His/GFP fusion proteins in E. coli and purified by nickel-affinity chromatography as described (26). Recombinant

Mechanism of homogalacturonan synthesis
proteins were incubated at a 1:10 ratio of TEV and PNGase F relative to the GAUT1:GAUT7 complex overnight at room temperature to cleave the fusion tags and N-glycan structures.
Protein samples (4 g) were mixed with Laemmli sample buffer (81) containing 25 mM DTT to reduce the samples. DTT was omitted in nonreduced samples. Samples were boiled for 10 min and resolved by 4 -15% gradient Tris-glycine SDS-PAGE gel. Electrophoresis was performed in a running buffer containing 25 mM Tris, 192 mM glycine, and 0.1% (w/v) SDS using 150-mV constant voltage.

HG:GalAT activity radiolabeled filter assays
Unless otherwise noted, HG:GalAT activity was measured under standard conditions in 30-l reactions containing 100 nM GAUT1:GAUT7, 5 M UDP-[ 14 C]GalA, 1 mM total UDP-GalA, 10 M HG acceptor, HEPES buffer, pH 7.2, 0.25 mM MnCl 2 , and 0.05% BSA. Reactions were incubated at 30°C, and 5 min was used as a standard time for linear range specific activity comparisons. As indicated, reactions were modified from the standard conditions to include acceptors enriched for a homogeneous degree of polymerization or higher concentrations of acceptors to modify the donor/acceptor ratio or were conducted in the absence of exogenous acceptors for samples labeled "de novo" synthesis.
HG:GalAT activity was measured using a filter assay as described (82) with modifications. Reactions were terminated by the addition of 400 mM NaOH (5 l). Reactions were spotted onto 2 ϫ 2-cm squares of Whatman 3MM chromatography paper coated with cetylpyridinium chloride. Filters were airdried for 5 min prior to three 15-min rounds of washing in a 4-liter bath containing 150 mM NaCl, for a total washing period of 45 min. Filters were air-dried for at least 2 h prior to scintillation counting. Scintillation counting was performed using a PerkinElmer Life Sciences Tri-Carb 2910 TR liquid scintillation counter, 14 C program, 1-min count time/sample. Background cpm was measured in each assay using T0 samples, in which NaOH was added to the reaction mixture prior to the addition of enzyme. Background cpm was subtracted from net cpm readings for all reaction samples, and cpm was converted to total pmol transferred.

Alcian blue-stained PAGE
Reactions were incubated under standard conditions described above, except UDP-[ 14 C]GalA was omitted from the reaction buffer. Reactions were terminated by boiling. Aliquots containing an estimated 200 -500 ng of total polysaccharide were analyzed. PAGE and visualization by a combination of alcian blue and silver nitrate staining was performed as described with modifications (14). Samples were mixed with a loading buffer (final concentration 0.1 M Tris, pH 6.8, 0.01% phenol red, and 10% glycerol), loaded onto a stacking gel (5% acrylamide (Bio-Rad), 0.64 M Tris, pH 6.8), and separated over a 30% acrylamide resolving gel (0.38 M Tris, pH 8.8, 30% acrylamide) at 17.5 mA for 60 min. The gel was stained for 20 min with 0.1% alcian blue (Sigma) in 40% ethanol and washed with at least three changes of water until background staining was eliminated. Silver stain-ing and developer was performed using a silver staining kit (Bio-Rad). Staining was terminated by the addition of 5% acetic acid.

Size-exclusion chromatography of HG:GalAT products
Scaled up reactions with a total volume of 400 l were incubated, using concentrations of reagents consistent with smallscale, nonradioactive, standard condition HG:GalAT assays. At each indicated time point, a 50-l aliquot was removed, and the reaction was stopped by boiling. Denatured enzyme was removed from the sample by centrifugation at 12,000 rpm for 5 min, and the polysaccharide sample was frozen at Ϫ20°C until analysis. For T0 samples, an equivalent aliquot was removed prior to the addition of UDP-GalA to the reaction.
Size-exclusion chromatography (SEC) was performed using a Superose 12 10/300GL column connected to a Dionex system at a flow rate of 0.5 ml/min in 50 mM ammonium formate buffer. The refractive index of polysaccharide products was measured. Dextran standards (270, 150, 50, and 12 kDa) (Sigma) were used as molecular mass standards. The peak retention volume of each dextran standard is indicated by an arrow at the top of each SEC figure. The molecular weight estimations of pectins may be overestimations due to anomalous behavior of pectins compared with dextrans during SEC (51).
For 2-AB-labeled polysaccharides measured by fluorescence detection, reactions were incubated for 12 h, polysaccharide reducing ends were chemically labeled with 2-AB (described below), and reaction products were injected onto the SEC column under the same conditions as above. An RF 2000 fluorescence detector, under high sensitivity (x16) settings, was used for detection. Quantitation of high-MW polysaccharide was performed by fluorescence signal integration of product and acceptor peaks in Chromeleon version 6.80 software (Dionex).

MALDI of 2-AB-labeled HG oligosaccharide products
Following incubation under the indicated reaction conditions and times, HG products were incubated with 0.2 M 2-AB and 1 M sodium cyanoborohydride in 10% acetic acid to chemically label the reducing ends of HG oligosaccharides, as described (24,40). Samples were dialyzed four times against water in a 3500 MW cutoff tubing (VWR Scientific) and recovered by lyophilization. Retention of HG oligosaccharides during dialysis has been described (51).

ELISA of HG:GalAT activity using anti-HG monoclonal antibodies
Monoclonal antibodies directed against plant cell wall polysaccharides were characterized previously (52). Antibodies directed against HG (CCRC-M38, CCRC-M131, and JIM5) and xylan (CCRC-M149) were used in this study. Antibody epitope, immunoreactivity, and supplier information is available at

Mechanism of homogalacturonan synthesis
WallMabDB (http://www.wallmabdb.net). 4 Following incubation under standard reaction conditions unless otherwise noted, aliquots containing either 1 l ( 1 ⁄ 30 of the reaction) or 0.1 l ( 1 ⁄ 300 of the reaction) were diluted to a final volume of 50 l, and reactions were terminated by boiling. ELISAs were performed as described previously (84). The 50-l diluted reaction sample was incubated in a 96-well plate (Costar 3598) and evaporated to dryness overnight. Nonspecific spots on the plate were blocked by incubation for 1 h in 0.1 M Tris-buffered saline (TBS) containing 1% nonfat dry milk (Publix) (200 l). Primary antibodies (50 l), diluted 10-fold from the hybridoma supernatant, were dispensed into each well and incubated for 1 h. All wells were washed with 300 l of wash buffer (0.1 M TBS containing 0.1% nonfat dry milk) for a total of three washes. Secondary antibodies were diluted to 1:5000 in wash buffer. For CCRC series antibodies, anti-mouse (Sigma, A4416), and for JIM series antibodies, anti-rat (Sigma, A9037) secondary antibodies conjugated with horseradish peroxidase (50 l) were incubated for 1 h. Plates were washed with wash buffer for a total of five washes. 3,3Ј,5,5Ј-Tetramethylbenzidine peroxidase substrate (Vector Laboratories) was incubated in each well (50 l) for 20 min, and the reaction was stopped by the addition of 0.25 M sulfuric acid. OD values were measured using a plate reader at 450 nm with background readings at 655 nm subtracted. Boiled enzyme reaction buffer controls contain reaction mixtures incubated under the same conditions, except GAUT1:GAUT7 enzyme was inactivated by boiling prior to the addition to the reaction mixture.

Endopolygalacturonase digestion
Following incubation of the HG:GalAT reaction, the mixture was adjusted to pH 4.2 by the addition of 1 M sodium acetate buffer, pH 4.2 (3 l) and 2 M acetic acid (15 l). The mixture was incubated overnight at 30°C with 20 milliunits of endopolygalacturonase-I (Aspergillus niger, EC 3.2.1.15) (85); 1 unit ϭ 1 mol of reducing sugar produced per min, as determined by a p-hydroxybenzoic acid hydrazide reducing sugars assay (Sigma). The endopolygalacturonase (EPG) reaction was terminated by the addition of 1 M NaOH (30 l). The final reaction mixture was assayed using the filter assay, as described above. Control samples labeled Boiled EPG (Fig. S5) were incubated with EPG that was deactivated by boiling for 1 h prior to use.

Modeling of Michaelis-Menten kinetics
Standard Michaelis-Menten kinetics and substrate inhibition kinetics were modeled using formulas for nonlinear regression analysis using GraphPad Prism version 7 for Windows (GraphPad Software, La Jolla, CA). Kinetic parameters (V max , K m , and K i ) were calculated for each independent experiment, measured using 8 -12 different concentrations of the variable substrate spanning a range of 0.1-10 times K m . Replicate experiments were calculated as separate data sets, using the standard Michaelis-Menten equation (Equation 1) and the substrate inhibition equation (Equation 2).