Efficient solid-phase synthesis of meningococcal capsular oligosaccharides enables simple and fast chemoenzymatic vaccine production

Neisseria meningitidis serogroups A and X are among the leading causes of bacterial meningitis in the African meningitis belt. Glycoconjugate vaccines, consisting of an antigenic carrier protein coupled to the capsular polysaccharide of the bacterial pathogen, are the most effective strategy for prevention of meningococcal disease. However, the distribution of effective glycoconjugate vaccines in this region is limited by the high cost of cultivating pathogens and purification of their capsular polysaccharides. Moreover, chemical approaches to synthesize oligosaccharide antigens have proven challenging. In the current study, we present a chemoenzymatic approach for generating tailored oligosaccharide fractions ready for activation and coupling to the carrier protein. In a first step, the elongation modes of recombinant capsular polymerases from Neisseria meningitidis serogroups A (CsaB) and X (CsxA) were characterized. We observed that CsaB is a distributive enzyme, and CsxA is a processive enzyme. Sequence comparison of these two stealth family proteins revealed a C-terminal extension in CsxA, which conferred processivity because of the existence of a second product-binding site. Deletion of the C-terminal domain converted CsxA into a distributive enzyme, allowing facile control of product length by adjusting the ratio of donor to acceptor sugars. Solid-phase fixation of the engineered capsular polymerases enabled rapid production of capsular polysaccharides with high yield and purity. In summary, the tools developed here provide critical steps toward reducing the cost of conjugate vaccine production, which will increase access in regions with the greatest need. Our work also facilitates efforts to study the relationship between oligosaccharide size and antigenicity.

The prevention of disease and death through vaccination is commonly regarded as one of the major public health achievements of the 20th century and necessary to improve conditions for economic development, particularly in low-income countries (1)(2)(3). Historically developed to protect against bacterial pathogens, glycoconjugate vaccines (i.e. protein carriers with covalently attached pathogen-specific carbohydrates (4)) are attracting growing attention also as vaccines against protozoa, helminths, viruses, fungi, and even cancer cells (for review see Ref. 5). Indeed, glycoconjugate vaccines are the gold standard to protect against infections caused by encapsulated bacteria like Haemophilus influenzae type b, Streptococcus pneumoniae, and Neisseria meningitidis (4). However, the high complexity and cost of glycoconjugate vaccine production is a barrier to achieving the high vaccination coverage necessary to install herd immunity (3, 6 -8). A major obstacle in the production process is the isolation of capsular polysaccharides (CPSs) 2 from largescale fermentation of bacterial pathogens. This not only is a considerable biohazard but also presents challenges with respect to batch-to-batch reproducibility and a complex downstream purification (8,9).
Consequently, research in the private sector and academia has focused on developing alternative vaccine production procedures with the aim to completely omit pathogen culture (8, 10 -16). A breakthrough in this regard has been the successful marketing of a chemically synthesized H. influenzae type b vaccine (15). However, the pure synthetic chemistry approach has not been readily translated to other bacterial pathogens (8).
For several years we have focused our research efforts on establishing safe and effective protocols for the synthesis of urgently needed vaccines. To that end, we exploit the recombinant capsule polymerases (CPs), the key enzymes of CPS synthesis (16 -23). In a proof-of-concept study (16), a fully synthetic NmX vaccine has been demonstrated to induce protective antibodies similar to the benchmark vaccine (24). In this study, we develop a streamlined protocol, avoiding steps that present a biohazard and simplifying downstream processing. With the recombinant capsule polymerases from NmX (CsxA) and NmA (CsaB), we describe here a protocol for the tailored synthesis of oligosaccharides of uniform size (avDP15, average degree of polymerization 15) ready for activation and coupling with the carrier protein. Crucial to this procedure was the use of optimized enzymes that enabled solid-phase coupling and control of product length.

Product profiles of recombinant CsxA and CsaB
Oligosaccharides have been shown to induce immune responses superior to long polysaccharides (25). Accordingly, an oligosaccharide fraction comprising DP10 -DP60 (see Fig. 1, C and D, marker) is routinely used in glycoconjugate vaccines and prepared from long CPS in two consecutive steps: 1) acidic hydrolysis and 2) sizing (16,24,26). With the aim to omit these steps and exploit the enzymatic in vitro synthesis to immediately provide oligosaccharides of the required size (avDP15), enzymes that allow direct and facile control of product length are necessary.
The primary feature determining the product-length distribution of CPs is the extent of their interaction with the growing polysaccharide chain (17,27). Processive enzymes catalyze multiple transfers without intermediate dissociation of the enzyme-product complex (17,28,29). Their products are nonuniform. In contrast, enzymes with distributive elongation mode dissociate after each transfer and thus synthesize uniform oligosaccharides ideally of Poisson distribution (17,30). Under in vitro conditions, product-length control can be imposed by the donor to acceptor ratio (d/a) (17,30). To obtain insight into the elongation mechanisms of CsaB and CsxA, we used the previously described fusion constructs ⌬N69-CsaB co -His 6 (22) and MBP-CsxA-His 6 (21) (for simplicity these constructs are henceforth referred to as CsaB and CsxA).
Assuming that a length bias may be introduced by priming acceptors of varying length, we first tested the polymerases with primers of single DP (DP2-DP7), generated through acid hydrolysis of long CPS and subsequent separation by anion-ex-change chromatography (AEC) (Fig. 1, A and B). Polymerase reactions were performed overnight with catalytic amounts of enzymes (50 nM) and an excess (2 and 5 mM for CsaB and CsxA, respectively) of the donor sugar UDP-GlcNAc (testing of CsaB was carried out in the presence of the epimerase CsaA, needed to produce UDP-ManNAc from UDP-GlcNAc (22)). To visualize products, we used Alcian blue/silver-stained PAGE (Fig. 1, C and D). Although no (Fig. 1C) or only negligible amounts of polymer ( Fig. 1D) were produced in the absence of primers or presence of free GlcNAc or GlcNAc-1P (Fig. 1D), each priming DP (DP2-DP7) was efficiently elongated. Based on these results we decided to use mixtures comprising DP2-DP7 of the respective CPS (henceforth referred to as primCPSA and primCPSX) in experiments aimed to determine the elongation modes of CsxA and CsaB.
In overnight reactions, the polymerases were incubated with priming acceptor (a) and donor sugar (d; UDP-GlcNAc) in increasing ratios (d/a ratio). Products were analyzed by Alcian blue/silver-stained PAGE (Fig. 1, E and F). Unexpectedly, products catalyzed by ⌬N69-CsaB co -His 6 grew in size with the increasing d/a ratio, indicating a distributive mode of elongation (Fig. 1E). In stark contrast, products of CsxA catalyzed reactions were highly dispersed, displaying the typical product profile observed with processive enzymes (Fig. 1F).

C-terminal truncation converts CsxA into an enzyme with distributive elongation mode
Both CsaB and CsxA are members of the so called stealth protein family, which comprises a number of bacterial virulence factors (31). Characteristic for stealth proteins (all are

Tailored synthesis of neisserial capsular oligosaccharides
hexose-1-phosphate transferases) are four conserved regions (CR1-CR4; Fig. 2A). Alignment of the primary sequences of CsaB and CsxA revealed a C-terminal extension of 98 amino acids in CsxA. We hypothesized that this C-terminal extension may confer processivity to the enzyme. To test this hypothesis, the truncation mutant ⌬C98-CsxA was constructed. Moreover, because the N-terminal segment preceding CR1 did not show any similarity to the respective domain in CsaB, the mutant ⌬N58-CsxA was additionally made. When expressed as recombinant proteins, both truncations were soluble and could be purified via the His 6 tag (Fig. 2B) at levels equal to (⌬C98-CsxA, 13 mg/liter culture) or significantly higher (⌬N58-CsxB , 70 mg/liter culture) than wild type. Polymerase activity was tested using a quantifiable spectrophotometric assay (22) with prim-CPSX as acceptor (Fig. 2C). Quite unexpectedly, both truncations showed drastically increased activity (ϳ4-fold ⌬N58-CsxA and ϳ6-fold ⌬C98-CsxA).
To test whether the truncations influenced the elongation mode used by CsxA, the assay described in Fig. 1F was repeated with the two truncated constructs, by increasing the d/a ratio stepwise from 20 to 400. After overnight reaction, the products were separated by gel electrophoresis and stained by Alcian blue/silver (Fig. 3, A and B). The release of the C-terminal 98 amino acids (⌬C98-CsxA) was shown to be sufficient to transform CsxA into a distributive enzyme (Fig. 3A). In contrast, truncation of the N-terminal 58 amino acids had no considerable impact on the elongation mode (Fig. 3B). To confirm these data, we additionally monitored the product profiles of CsxA truncations in a time-course experiment. As an analytical system, we used HPLC-based AEC (22). To exclude any bias at acceptor site, the time-course experiment was carried out with DP5 as priming substrate. After reaction start (time point 0Ј), samples were taken as indicated (Fig. 3, C and D). In perfect agreement with the PAGE results ( Fig. 3B), ⌬N58-CsxA synthesized short oligosaccharides in the early reaction phase (Fig.  3D). After 20 min, a second well-separated peak representing long polysaccharides appeared. Because chains of DP Ͼ 18 were only present in minor quantities, this profile allowed to conclude that chains of DP Ͼ 18 are necessary to start processive elongation. Conversely, the products of ⌬C98-CsxA were observed to grow continuously throughout the reaction, confirming the distributive mode of elongation of this enzyme (Fig. 3C).
Two models exist in the literature on the basis of which the prominence of the C terminus for CsxA processivity could be explained: 1) Similar to enzymes acting on DNA (e.g. helicases; reviewed in Ref. 29), the C terminus might assist in the formation of an oligomeric structure that encloses the nascent chain as soon as an intermediate chain length is reached (Fig. 4A); and 2) in analogy to other enzymes producing carbohydrate polymers, the C terminus might harbor an extended binding site with increasing affinity for the growing polymer chain ( Fig. 4B) (17,27). Because we knew from earlier work that wild-type CsxA elutes as an oligomer (depending on the salt concentration with an apparent molecular mass of an octamer or tritetramer) from size-exclusion chromatography (SEC) columns (21), we first asked whether depletion of the C terminus altered the oligomerization state. As shown in Fig. 4, ⌬C98-CsxA, similar to the wild type, migrated as an octamer in low salt buffer ( Fig. 4C) and formed assemblies Ͼtrimer in high salt buffer (Fig.  4D). Rather unexpectedly, the processive ⌬N58-CsxA, used as a control in these experiments, eluted as a dimer in both high and low salt buffer (Fig. 4, C and D). These data revealed that the N terminus determines the oligomeric state of CsxA and in addition showed that processivity in CsxA is independent of the oligomeric state. Consequently, we evaluated the second possibility, the existence of a second product-binding site in the C-terminal domain. For this purpose, the construct ⌬N387-CsxA-His 6 , which comprises only the C-terminal 99 amino acids, was cloned and expressed. As a simple binding assay, we studied the sensitivity of the purified protein against proteinase K in the presence and absence of CPSX. The Coomassie-stained SDS-PAGE used to monitor the time-dependent proteolysis is shown in Fig. 4E. In the presence of CPSX, proteolytic degradation of the C-terminal domain was significantly delayed. Although compared with the starting material, a size reduction became visible shortly after proteinase K contact (5 min), this first degradation product remained visible for 300 min when CPSX was present. In the absence of CPSX as well as in the presence of CPSA (Fig. 4F), this band was drastically reduced already after 30 min and completely disappeared after 90 min. Together, these results provide strong arguments for a specific binding of CPSX to the C-terminal domain and thus for the existence of a second product-binding site in CsxA.
Finally, we asked whether also the double truncated CsxA can be produced and cloned the mutant ⌬N58⌬C98-CsxA. If compared with CsxA and ⌬C98-CsxA, ⌬N58⌬C98-CsxA could be purified to homogeneity, almost completely lacking degra- A, CsxA and CsaB were aligned according to their conserved stealth protein family regions (CR1-CR4). The first and last amino acid, as well as the sites of truncation introduced into the CsxA sequence, are indicated. B, Coomassiestained SDS-PAGE presenting the purified CsxA constructs. C, activity of the purified CsxA constructs in the absence (Ϫ) and presence (ϩ) of primCPSX as determined by use of a spectrophotometric assay. Wild-type activity was set to 100%.

Tailored synthesis of neisserial capsular oligosaccharides
dation products (compare Fig. 2B and Fig. 5A). The purification yield (27 mg/liter culture) was intermediate between ⌬N58-CsxA (70 mg/liter culture) and ⌬C98-CsxA (13 mg/liter culture). In SEC, the construct appeared to be mono-to dimeric, independent of the salt concentration (Fig. 5, B and C). The product profile produced at a d/a ratio of 20 was even narrower than the one produced by ⌬C98-CsxA (Fig. 5D) and more closely resembled the dispersity of the fraction used for vaccine production (marker). Finally, a time-course experiment was conducted to examine the elongation mode of the double trun-cated enzyme. With DP5 as a primer, ⌬N58⌬C98-CsxA generated a narrow Poissonian product distribution at all sampled time points (Fig. 5E). Taken together, this protein perfectly suited the requirements of vaccine production.

On-column synthesis of length controlled CPSX oligosaccharides
Solid-phase fixation of enzymes in biotechnological processes is of major value, because this measure supports protein longevity, reduces the need for protein production because

Tailored synthesis of neisserial capsular oligosaccharides
enzyme-loaded matrices are reusable, and facilitates downstream processes (32). In a pilot experiment, ⌬N58⌬C98-CsxA was loaded onto a HisTrap column (⌬N58⌬C98-CsxA is Nterminally MBP-tagged and C-terminally His-tagged). After extensive washing, a reaction mixture containing primCPSX and UDP-GlcNAc in the ratio 1:20 was continuously pumped through the column with a peristaltic pump (Fig. 6A). The reaction progress was monitored every 20 min in an HPLC-based assay that allowed the parallel detection of UDP-GlcNAc and UMP at 280 nm (Fig. 6B). After 100 min, only UMP was detected and indicated complete consumption of UDP-GlcNAc. Size and dispersity of oligosaccharides was analyzed by Alcian blue/silver-stained PAGE (Fig. 6C) and HPLC-AEC (Fig. 6D). Both analyses demonstrated that the product disper-sity was identical to the oligosaccharide marker (avDP15), which was obtained by the conventional vaccine production protocols (24) (Fig. 6, C and D). To ensure that oligosaccharide fractions are not contaminated with enzyme that may bleed off the HisTrap matrix, this column was connected with an amylose column, an affinity matrix for MBP (Fig. 6A). The amylose column was eluted after 100 min of production time, but no protein was detected in the eluate (data not shown). Similarly, we were unable to detect the enzyme in the reservoir from which samples were taken every 20 min (Fig. 6E). In contrast, when imidazole buffer was added to the HisTrap column at the end of the reaction, a sharp enzyme band identical to purified ⌬N58⌬C98-CsxA was visible (compare Figs. 5A and 6E). Taken together, these experiments for the first time describe a simple,  (21), the protein formed an apparent dimer after removal of the N-terminal 58 amino acids. E and F, proteinase K sensitivity of the isolated C-terminal domain (⌬N387-CsxA) was assayed in the presence and absence of CPSX (E) and CPSA (F). The delayed degradation of the protein in the presence of CPSX but not CPSA argues for a specific binding of the first. As control, the purified ⌬N387-CsxA and proteinase K (PK) were loaded onto the gel.

Tailored synthesis of neisserial capsular oligosaccharides
fast, and efficient protocol for the synthesis of oligosaccharides ready for activation and protein coupling.

Length controlled CPSA oligosaccharides by use of a CsaB-CsaA fusion enzyme
Although CsaB was demonstrated to exhibit a distributive elongation mode (Fig. 1E), the on-column synthesis of CPSA is challenging, because the donor sugar UDP-ManNAc is unstable and not commercially available. Thus, in-process production of UDP-ManNAc from UDP-GlcNAc by use of the epimerase CsaA is necessary (22) and was planned in this study. From previous work with the soluble enzymes in a one-pot reaction, we knew that CPSA synthesis was efficient as long as the concentration of CsaB was equal or higher than the concentration of CsaA (22). Therefore, we assumed that the generation of a fusion protein would best comply with the requirements for the on-column synthesis and also reduce efforts with respect to protein purification and downstream processing. Guided by our previous observation that a free N terminus increases CsaB expression levels (22), the construct CsaB-CsaA was cloned with C-terminal His 6 tag and purified at a yield of 30 mg/liter expression culture. To confirm activity, CsaB-CsaA was incubated with primCPSA and UDP-GlcNAc in increasing d/a ratios (Fig. 7A). This test series confirmed that fusion with CsaA did not alter the elongation mode of CsaB.
To obtain proof-of-concept that also CPSA synthesis can be carried out on a solid phase, the fusion protein was loaded onto a HisTrap column as described for CsxA. However, because we had learned in the previous experiment that the reaction slowed down with time because of continuous dilution of substrates in the reservoir (Fig. 6, A and B), we modified the experimental setup and loaded the entire reaction mixture (10 ml), containing a d/a ratio of 20, in one step. Samples were taken for HPLC-AEC, PAGE, and Western blot analysis, before the collected fractions (2 ml) were pooled and loaded for a second time. The HPLC-AEC analysis of the first and second load showed that 78 and 89% of UDP-GlcNAc was consumed, respectively (Fig. 7B). Product size evaluation was carried out with Alcian blue/silverstained PAGE and, as in the case of CPSX, demonstrated a dispersity highly similar to the marker material (Fig. 7C). As before, we did not see any contamination of the oligosaccharide fractions with protein but could elute the enzyme from HisTrap by adding imidazole buffer (Fig. 7D). However, it is of note that a second His 6 -tagged band migrating with the theoretical molecular mass of CsaA-His 6 was present in both the soluble lysate and (even enriched) the eluate of the HisTrap column. This second band by all likelihood indicates partial degradation of the fusion protein CsaB-CsaA during expression. Thus, future work will aim to stabilize the fusion protein. This proofof-concept study clearly highlights the advantage of enzyme catalyzed vaccine production protocols.

Discussion
For the first time, we describe in this study the comprehensive characterization of the elongation mechanisms of two polymerizing N-acetylhexosamine-1-phosphate transferases (CsaB and CsxA) of the stealth protein family and their use as biotechnological tools for the production of tailored CPS fragments of the meningococcal serogroups A and X. The decision to study these two enzymes was made on the basis that NmA and NmX represent considerable threats to public health (33). Moreover, no licensed vaccine exists for NmX, and, despite the successful introduction of the anti-NmA vaccine Men-AfriVac TM (34, 35), stockpiling of vaccines and/or vaccine components as requested by the International Coordinating Group on Vaccine Provision for Epidemic Meningitis is particularly difficult in the case of the short-lived CPSA (7,36). Last, but certainly not least, current protocols are not suited to promote decentralized production of vaccines to allow immediate responses to unforeseeable needs foremost occurring in developing countries (6, 7).

Tailored synthesis of neisserial capsular oligosaccharides
Synthetic chemistry promises an alternative source of CPS fragments, which avoids the need for large-scale cultivation of pathogens (8, 11-13, 15, 37). Compared with the use of enzymes, chemical processes require time-intensive multistep procedures to build the complex CPS fragments. With regard to meningococcal vaccines against serogroups A and X, the longest compounds obtained by chemical synthesis are the CPSX and CPSA tetramers (10,11). Mouse immunization experiments were performed with a protein coupled CPSX trimer (12) and a trimer of a CPSA carba-analogue (13). Although both antigens induced a specific antibody response, it was low compared with the natural polymer with the average DP of 15 used as benchmark (24,26).
In the current proof-of-concept study, we produced 8 mg of CPSA of avDP15 and 70 mg of CPSX of avDP15 (see "Experimental procedures") in less than 2 h using standard laboratory equipment (E. coli safety strain for expression, 1-ml commercial HisTrap columns, simple peristaltic pump) and without any further investment in process optimization. The straightforward protocol and avoidance of biohazard shall facilitate the adaptation to low-tech environments.
Covalent chemical coupling of carbohydrate antigens to protein carriers generates glycoconjugate vaccines that elicit T-cell help for B cells recognizing the carbohydrate epitope and, consequently, isotype switching from IgM to IgG (38). A recent study reported that the carbohydrate component of a group B Streptococcus glycoconjugate vaccine is processed by antigen presenting cells to 10 kDa (Ϸ50 hexoses) in size (38,39). On the other hand, small synthetic oligosaccharides (DP 3-4) have been shown to elicit functional antibody responses, although at a low level, if conjugated to a carrier protein (12,38). Our technology provides a valuable tool for the generation of tailored oligosaccharide fractions that can be used to investigate the dependences of antigen size and immunogenicity in vaccine development studies.
Large CPS binding sites have been postulated for several CPs from encapsulated bacteria (both Gram-negative and Grampositive strains). For example, the polysialyltranferase from NmB was shown to accommodate chains of at least 20 residues (17), whereas the extended binding site in the E. coli K92 enzyme may bind up to 10 -12 residues (40). The CP from S. pneumoniae was shown to switch to processive elongation with acceptors greater than an octasaccharide (27). However, none of the enzymes studied to date have been shown to harbor a distinct processivity-conferring polysaccharide binding domain. Of note in this context, a blast search with the C-terminal 98 amino acids of CsxA did not identify any homologous sequences.
Hypothesizing that in ⌬N69-CsaB co -His 6 a processivity mediating element was cut off, we also analyzed reaction products formed by full-length CsaB, but no processive elongation was seen (data not shown). A potential explanation for the distributive elongation mode of CsaB could be that in vivo the protein interacts with additional enzymes (e.g. the epimerase CsaA and the O-acetyltransferase CsaC) (22), which confer processivity. Moreover, distributive chain elongation by CsaB may be required for CsaC to generate uniform O-acetylation patterns. Further research is needed to answer these open questions.
When analyzed with the help of bioinformatics tools like globplot (41) and PHYRE2 (42), the N-terminal 58 amino acids of CsxA were predicted to be disordered. Removal as in construct ⌬N58-CsxA was therefore expected to improve expression levels and protein stability. Indeed, the observed dispensability of the N terminus of the recombinant CsxA and CsaB

Tailored synthesis of neisserial capsular oligosaccharides
prompts us to speculate that these protein domains fulfill functions that are needed only in the in vivo situation (22).
More research is also needed to produce a stable CsaB-CsaA fusion construct and will be carried out in the same way as our previous work with protease-resistant linkers (21,43). Still, it is noteworthy that the partial breakdown of the fusion protein did not negatively influence the on-column production. Indeed, it is conceivable that higher CsaA concentrations established by column retention of the His 6 -tagged protein were of advantage. Generally, there is significant room to further improve the oncolumn procedure. One important measure for instance would be covalent enzyme coupling to facilitate the reuse of matrices (32).
Considering that chemists can synthesize native CPSA and CPSX fragments containing the necessary chemical functionality for downstream protein coupling (10 -12), it is an interesting idea to combine the advantages of the here-described enzyme catalyzed CPS synthesis with chemically optimized primers into a platform that advances the fast and reliable provision of oligosaccharides ready for protein coupling.

Expression, purification, and activity testing of recombinant CsaB and CsxA constructs
Expression and purification of all CsaB-and CsxA-based sequences was performed as described in Refs. 22 and 21, respectively. The spectrophotometric assay published for CsaB (22) was adapted for CsxA, resulting in the alterations as fol-lows: 1) no CsaA (epimerase) was used in the reaction mixture, 2) the reaction was performed using 54 nM of the respective CsxA construct in the presence of 3) 37 ng of CPSX oligosaccharides of avDP18.6.

SDS-PAGE and immunoblotting
SDS-PAGE and immunoblotting were performed as described before (21,22).

Purification of CPSA and CPSX oligosaccharides
The marker, previously described as oligosaccharides of avDP15 (DP10 -DP60) (16,21,22,24), was obtained from the vaccine manufacturer GlaxoSmithKline (Siena, Italy). CPSA and CPSX oligosaccharide mixtures (primCPS) (16,21,22) were dephosphorylated using either acid phosphatase (Sigma) or calf intestinal alkaline phosphatase (New England Biolabs) according to the manufacturer's guidelines. AEC was performed on an ÄKTA FPLC (GE Healthcare) equipped with a MonoQ HR 5/5 column (GE Healthcare) at a flow rate of 1 ml/min. H 2 O and 1 M NaCl were used as mobile phases M 1 and M 2 , respectively. Samples were separated using a combination of linear gradients (0 -5% over 1 ml, 5-20% over 10 ml, and 20 -30% over 20 ml). Acetate and GlcNAc-1P were used to calibrate the column. The amount of oligosaccharide in each fraction was estimated from the peak area obtained at 214 nm under the assumption that each residue contributes equally to the absorbance of the respective oligomer. Fractions were dialyzed against water (ZelluTrans, Roth, 1-kDa molecular mass cutoff) and freeze-dried, and equal concentrations were adjusted with H 2 O. The identity and DP of the oligosaccharide fractions was confirmed by HPLC-PAD (pulsed amperometric detection) using established protocols (36). For the time-dependent analysis, the reaction volume was up-scaled to 700 l, the UDP-GlcNAc concentration was increased to 10 mM, and DP5 was used as primer (d/a ratio of 500). Aliquots were snap-frozen after the indicated time points and heat-inactivated for 3-5 min at 98°C prior to analysis. HPLC-AEC and high percentage PAGE followed by Alcian blue/silver staining was performed as described before (22).

Analytical size-exclusion chromatography
Roughly 300 g of recombinant protein was analyzed at 280 nm on an ÄKTA FPLC (GE Healthcare) equipped with a Superdex 10/300 GL column (GE Healthcare) using 50 mM Tris, pH 8.0, as elution buffer supplemented with NaCl as indicated in Fig. 4. For the analysis of CsxA constructs, 1 mM of DTT was added. The column was calibrated using the gel filtration markers kit for protein molecular masses 29,000 -700,000 Da (Sigma) according to the manufacturer's guidelines.

Tailored synthesis of neisserial capsular oligosaccharides Limited proteolysis
Samples containing 22 g of ⌬N387-csxA-His 6 , 2.2 g of proteinase K, and if present, 0.4 mg of CPSX or CPSA in a volume of 100 l of PBS were incubated at room temperature. Aliquots were taken at the indicated time points, frozen at Ϫ20°C, and analyzed by SDS-PAGE as previously described (21,22).