Combined effects of glycan chain length and linkage type on the immunogenicity of glycoconjugate vaccines

The development and use of antibacterial glycoconjugate vaccines have significantly reduced the occurrence of potentially fatal childhood and adult diseases such as bacteremia, bacterial meningitis, and pneumonia. In these vaccines, the covalent linkage of bacterial glycans to carrier proteins augments the immunogenicity of saccharide antigens by triggering T cell-dependent B cell responses, leading to high-affinity antibodies and durable protection. Licensed glycoconjugate vaccines either contain long-chain bacterial polysaccharides, medium-sized oligosaccharides, or short synthetic glycans. Here, we discuss factors that affect the glycan chain length in vaccines and review the available literature discussing the impact of glycan chain length on vaccine efficacy. Furthermore, we evaluate the available clinical data on licensed glycoconjugate vaccine preparations with varying chain lengths against two bacterial pathogens, Haemophilus influenzae type b and Neisseria meningitidis group C, regarding a possible correlation of glycan chain length with their efficacy. We find that long-chain glycans cross-linked to carrier proteins and medium-sized oligosaccharides end-linked to carriers both achieve high immunogenicity and efficacy. However, end-linked glycoconjugates that contain long untethered stretches of native glycan chains may induce hyporesponsiveness by T cell-independent activation of B cells, while cross-linked medium-sized oligosaccharides may suffer from suboptimal saccharide epitope accessibility.


INTRODUCTION
Bacterial polysaccharide (PS) molecules are versatile compounds that include lipopolysaccharides with their O antigens, capsular PSs, and exopolysaccharides. They play an important role as virulence factors of bacterial pathogens 1 and protect the bacterium against host immune responses in at least three ways. First, surface PSs shield bacterial structures from the destructive activity of the alternative complement pathway of the innate immune system 2,3 . Second, capsular PSs also obstruct the classical complement pathway since antibodies cannot easily reach subcapsular bacterial surface proteins, thwarting subsequent complement deposition. Third, adaptive immunity to the bacterial PSs remains inefficient since saccharide structures are poorly bound by the antigen-presenting cells' major histocompatibility complex (MHC), which leads to minimal T-helper cell-dependent antibody responses 4 . The latter probably explains why many hostdependent bacteria have developed surface carbohydrate structures during their evolution.
Bacterial PSs also contribute to the ability of the microorganism to colonize specific ecological niches, resist antibiotics, and fight bacteriophages 5 . Their omnipresence in many bacterial pathogens has long generated interest in these compounds for vaccine development. The glycan chains of bacterial PSs form epitopes or antigenic determinants that can be recognized by B cell receptors (BCRs). These epitopes can be linear or conformational in nature and may consist of a minimum span of identical monosaccharides (in homopolymeric PS chains) or of various different monosaccharides including side chains (in heteropolymeric PS chains, Fig. 1). By the 1970s and 1980s, bacterial capsular PS vaccines became part of routine vaccinations, and licensed vaccines for adults based on purified PSs had been established for disease caused by Neisseria meningitidis, Haemophilus influenzae type b (Hib), Salmonella enterica serovar Typhi, and Streptococcus pneumoniae, among others 6,7 . Immunogenicity of those plain (i.e., unconjugated) PS vaccines is dependent on the capacity of long-chain PSs to cross-link surface immunoglobulin receptors on B cells [8][9][10] , leading to an antibody response with typical T cellindependent characteristics.
By themselves, bacterial PS generally induce only a relatively short-lived T cell-independent immune response and largely fail to activate maturation of memory B cells 11 and long-lived plasma cells 12 . Consequently, they primarily induce low-affinity antibody responses and do not efficiently elicit long-term boostable immunological memory 9,13,14 . Because of their T cellindependent nature, plain PS vaccines such as those against Hib do not induce antibody responses and are therefore not effective in infants under 2 years of age [15][16][17] , with very few exceptions (such as the meningococcal serogroup A PS 18 ). Fortuitously, conjugation of glycans to a carrier protein, a principle first investigated almost 100 years ago with pneumococcal PSs 19 , can convert these compounds to T cell-dependent antigens 20 that induce highaffinity antibodies, antibody isotype switching, and a long-lasting memory immune response, making them efficacious in infants 21,22 . Coupling of a glycan to a carrier protein allows binding of the conjugate to saccharide-specific B cells, internalization of the conjugate, and presentation of carrier protein-derived peptide epitopes by MHC to T-helper cells. This eventually leads to clonal activation of the cognate B cells and their differentiation into plasma cells producing antibodies, with isotype switching to IgG and affinity maturation, and into memory B cells 21,22 .
All currently licensed glycoconjugate vaccines contain one of five carrier proteins. These are tetanus toxoid (TT), diphtheria toxoid (DT), a genetically modified cross-reacting material of the diphtheria toxin termed CRM 197 , meningococcal outer membrane protein complex OMPC, or H. influenzae protein D HiD 23 . However, a variety of alternative carrier proteins are also actively explored, such as modified versions of exotoxin A from Pseudomonas 1 Bacterial Vaccines Discovery and Early Development, Janssen Vaccines and Prevention B.V., Leiden, Netherlands. ✉ email: mbeurret@its.jnj.com aeruginosa 24,25 , and the S. pneumoniae ABC transporter protein PiuA 26 .
Among the five carriers used in licensed vaccines, CRM 197 and TT are by far the most used carriers. DT is often less potent, 27 probably due to epitope inactivation during the chemical detoxification process, and both OMPC and HiD are rarely employed. The carriers are covalently bound to glycan molecules, either in end-linked configurations after activation of sugar moieties at glycan ends or as cross-linked substances following random activation of PS chains (Fig. 2).
Specific physicochemical properties of the glycoconjugate affect vaccine efficacy, and key parameters, such as the molecular size of the PS used for conjugation and the extent of the crosslinking, can have a profound impact on vaccine immunogenicity 28 . The number of attachment sites per carrier protein, e.g., needs to be sufficiently high to ensure efficient conjugation, limiting the level of residual free PS molecules. However, the degree of derivatization of carrier proteins requires careful optimization to avoid unwanted destruction of critical MHC epitopes within the protein sequence, which would abolish its T cell-stimulating activity. Selection of these key parameters as well as the choice of carrier protein and the activation and conjugation chemistries are influenced by the practicalities of the purification processes during vaccine production, where viscosity and molecular mass ratios of bound to unbound glycan may impact purity and yield of the final product 6 . As a consequence, different process chemistries and conjugation methods have been employed to prepare those conjugates. These chemistries include activation procedures to guide protein-glycan linkage sites and their numbers per molecule, as well as the introduction of spacer molecules between the two glycoconjugate components.
One of the primary quality attributes of microbial PSs obtained in vaccine production is the number of glycan repeating units per PS chain termed the degree of polymerization or chain length. This chain length is heavily affected by the production process. Often, the PSs are subjected to a size reduction from very large molecules (>500 kDa) to a heterogeneous mix of PSs of about 100-300 kDa prior to activation and chemical linkage to the specific carrier protein 29,30 . Size reduction of PS molecules improves subsequent conjugation efficiency in vaccine manufacturing.
The level of immunogenicity of plain bacterial PSs is correlated to their size and, therefore, their chain length 31,32 . Consequently, large PS size determination became a critical quality measurement to guarantee the immunogenicity of plain PS vaccines, probably explained by the need to cross-link BCRs to elicit T cellindependent antibody responses. But is this association also true for glycoconjugate vaccines? Unlike plain PS vaccines, induction of antibody mechanisms by glycoconjugates does not depend on cross-linking of BCRs. Instead, glycoconjugates are internalized by antigen-presenting cells, leading to T-helper cell responses triggered by carrier protein-derived peptides. Does a dependency on glycan chain length persist despite this difference? The following survey of the existing body of literature regarding the effect of chain length disparities on glycoconjugate vaccine efficacy was undertaken to clarify this question.

FACTORS DETERMINING THE GLYCAN CHAIN LENGTH OF VACCINE PREPARATIONS
The size of the glycan chain in vaccine preparations is one important aspect of its structure. It is determined in various ways, utilizing terminology outlined in Table 1.
Two major aspects govern the length of the glycan chain in glycoprotein vaccine preparations-the underlying microbial genetics of the bacterial strain used to produce the glycan and the vaccine production process. These aspects are briefly discussed below.

Microbial genetics
Depending on their genetic repertoire, bacterial strains can produce PSs of different chain lengths. To date, several of the bacterial regulators that govern the size of microbial glycans are known. Importantly, these can be genetically altered and swapped across species boundaries to generate glycans of altered lengths [33][34][35] . Fig. 1 Key types of antigenic determinants (epitopes) of bacterial polysaccharides. In homopolymeric polysaccharides, identical monosaccharide residues repeat as a polymer. In heteropolymeric (or complex) PS, the repeat unit (RU, in square brackets) consists of diverse monosaccharide residues and may include branching side chains. The minimal required unit for antigenicity in linear epitopes consists of about 6-7 contiguous monosaccharides 121 . Linear epitopes may include the terminal monosaccharide residues of glycan chains. Conformational epitopes are formed by residues that are in close spatial proximity but dispersed across their primary sequence 122 . These require a sequence of residues sufficiently long to build or mimic the spatial conformations necessary for interaction with the antibody. As an example, the homopolymeric conformational epitope of the Neisseria meningitidis group B capsular PS antigen contains ten residues, although only the inner six residues interact with its cognate antibody 123,124 . Glycan fragments that exceed the length of the identified antigenic epitope are usually used as immunogens in glycoconjugate vaccines 125 .
Bacterial capsular PSs are, by and large, assembled either by Wzy-dependent polymerization or via the ABC transporter pathway. In Wzy-dependent polymerization, the PS is assembled in a stepwise manner in the bacterial periplasm from individual repeating units that are attached to a specific lipid carrier, undecaprenol diphosphate [36][37][38][39] . The so-called PS copolymerase proteins regulate the length of the polymer 40 . The best-studied of these enzymes, Wzz, has been known for a decade to affect the chain length of the O-antigen polymer in Escherichia coli lipopolysaccharide (LPS) 41 . In the ABC transporter pathway, however, the full-length PS is manufactured in the cytoplasm on a reducing terminal phosphatidylglycerol lipid bound to an oligosaccharide (OS) of β-linked KDO residues 36,39,42 . The chain length of PSs manufactured by this pathway is thought to be regulated in a geometry model by two interacting proteins, an extension enzyme called WbdA and a termination enzyme named WbdD. Overexpression of WbdD or diminished WbdA expression reduces chain length 43,44 . However, variations of this system have also been identified. As an example, polymerization, termination, and chain length regulation are performed by domains of a single protein, WbbB, in Klebsiella pneumoniae O12 45 . Similarly, S. pneumoniae serotype 3 uses a single membrane-bound synthase to produce its long-chain capsular PS. During polymerization, the nascent glycan remains tightly associated with the synthase until its release 46 , and enzyme catalytic phases may be affected by temperature and concentrations of the saccharide units 47 .
The vaccine production process Standard glycoconjugate vaccine production. Traditionally, glycoconjugate vaccines are produced by extraction of PSs from bacterial fermentation, followed by a multitude of purification and fragmentation steps, chemical activation, and subsequent conjugation to the carrier protein 48 . The carrier protein itself is similarly harvested from bacteria and subsequently purified. After chemical activation, it can be conjugated to the glycan via its naturally existing functional groups or on chemically introduced linker molecules.
There are a number of critical junctures in a production process that may affect the final size of the glycan in a vaccine formulation (Fig. 3). The chain length in a glycoconjugate vaccine manufactured by the traditional production process is determined not only by the genetic makeup of the microorganism but also by the fermentation conditions employed, including the carbohydrate source and the carbon/nitrogen ratio in the growth medium, the pH, the temperature, oxygen tension, and the levels of amino acids, vitamins, phosphate, and minerals 49 . Moreover, glycan isolation and purification conditions have a profound impact on the glycan chain length. Consequently, the molecular weight of meningococcal and pneumococcal capsular PSs, e.g., after industrial fermentation can range from approximately 100 to roughly 1000 kDa.
Often, the harvested PSs is subsequently chemically or mechanically fragmented 50 and size-fractionated to alleviate glycan heterogeneity in the vaccine product 51,52 , to improve yields and consistency during purification processes, and to optimize the conjugation process. PSs are typically fragmented to about 100-300 kDa in cross-linking chemistries (usually still referred to as PSs) and to 5-20 kDa (typically termed OSs) in end-linking chemistry approaches 6 . The polysaccharide or OS is then chemically activated prior to conjugation, either randomly at hydroxyl groups or localized at the glycan end(s). Cross-linking or end-linking conjugation methods are subsequently used to achieve either random linkage along the glycan chain, or a selective attachment at the terminal end(s) of the sugar moiety, with or without spacer molecules 53 (Fig. 2).
Bioconjugation. In an alternative technology called in vivo production of glycoproteins (aka protein glycan-coupling technology, or bioconjugation), the glycoconjugate is produced in a single step inside a microbe, where the glycan and the carrier protein are expressed and coupled inside the bacterial cell 54,55 , most often E. coli. Bioconjugation omits several necessary chemical reaction and purification steps that may hamper the yield, purity, and homogeneity of vaccine preparations produced by the traditional process. However, bioconjugation requires deliberate and careful genetic manipulation (bioengineering) of the host bacterium to introduce production of the glycan of choice (as a lipid-linked OS), the selected carrier protein, and protein glycosylation capability, often via dedicated plasmids 56 .
Bioconjugates are basically OS-based end-linked chemical glycoconjugates. Bacterial protein glycosylation is either oligosaccharyltransferase (OTase)-dependent (which can be N-linked, where glycans get attached to the amide nitrogen of asparagine residues, or O-linked, where glycans bond with the hydroxyl oxygen of serine or threonine residues) or OTase-independent 56 . In the former, glycans are moved en bloc from preassembled lipid-bound precursors to proteins in the periplasm. In the latter, glycosyltransferases move monosaccharides from nucleotide-activated precursors to sequentially form glycoproteins in the cytoplasm 56,57 . OTasedependent strategies are (currently) much more prevalently applied, often using the N-linked OTase PglB from Campylobacter jejuni with a remarkable capacity to transfer a number of different PSs. However, several O-linked OTases are also actively investigated for use in bioconjugation processes, including pilin-specific OTases PglL, and PglS, the latter of which can transfer glycans with glucose at the reducing end 56,58 . In modern bioconjugation techniques, the carrier protein is also often genetically modified to include glycosylation sites in specific sequence patterns.
Sophisticated strategies have been developed that can control the length of the glycan chain of the glycoconjugate produced in these bacterial strains. Genetic manipulation of the chain length controller Wzz is an established strategy to increase the chain length of PS residues. Its overexpression can increase the production of a defined, long-sized PS population and decrease the production of short and very long saccharides 33 .
Other technologies. The glycan of a conjugate vaccine can also be synthetically (chemically) produced, such as in the anti-Hib vaccine Quimi-Hib 59,60 , where the repeat unit (RU) octamer of ribosylribitol phosphate (RRP) 61 is end-linked to the carrier protein, TT. Unfortunately, the cost-effective manufacture of the necessary quantities of more complex synthetic OSs for standard vaccine production has remained a challenge. Enzymatic production of glycans is another alternative 52,62 , but has not permeated vaccine production pipelines, to date. These technologies can also be merged, as reported for the production of immunogenic meningococcal group C polysialic acid-TT glycoconjugates 63 .
Other interesting alternative technologies currently considered for glycan vaccine purposes use outer membrane vesicles from engineered bacteria, the so-called generalized modules for membrane antigens (GMMAs), or high-affinity noncovalently linked avidin-biotin PS/protein complexes, creating macromolecular multiple antigen-presenting systems 48,64,65 . Furthermore, liposomes and gold nanoparticles are currently investigated as novel delivery vehicles for glycoconjugates, with encouraging results 52 .

The impact of glycan chain length on vaccine immunogenicity
The minimal OS size to support immunogenicity. Efforts to determine the minimal glycan length to elicit immune responses are not only important from a cost perspective but are also useful in future research efforts to standardize and simplify vaccine production processes. OSs as short as tetramers have been shown to elicit effective immune responses, as components of an experimental N. meningitidis serogroup C (MenC) glycoconjugate 66 68 .
In tests using OSs representing the Shigella flexneri serotype 2a O antigen (OAg), the highest anti-LPS 2a IgG titer was generated by a glycoconjugate linked to TT that contained a chemically synthesized pentadecasaccharide that consisted of three RUs 69 . Subsequent studies showed that a minimum segment of nine sugar units is required to present the complete epitope to one of the protective antibodies 70 . Similarly, decasaccharide glycoconjugates elicited higher titers of anti-LPS IgG against the O-specific monomeric PS of Vibrio cholerae O1 serotype Ogawa than octamers or hexamers 71 . Consequently, the minimal oligo size capable of eliciting protective immune responses may depend, in part, on the RUs of the target PS and the steric characteristics that govern the size and structure of epitopes recognized by protective antibodies. A thorough understanding of these steric characteristics and their role in the interaction of the antigen with the host's antibodies may help in future rational designs of synthetic glycans as components of novel glycoconjugate vaccines 7 .
Importantly, the minimal antigenic glycan epitope must be freely accessible to dock to the BCR's immunoglobulin-binding cavity. In glycoconjugate vaccines, this can be achieved by the introduction of a spacer molecule between the antigenic epitope (glycan) and the carrier protein. These spacers enhance the exposure of antigenic epitopes on the surface of glycoconjugate vaccines and alleviate steric hindrances. They are (usually) small molecules with reactive groups on either end, which can be identical (such as in adipic acid, cystamine, or dithiobis(succinimidyl propionate)) or different (e.g., N-hydroxysuccinimidyl-3maleinimidopropionate) 72 . However, many successful conjugates (both cross-linked long-chain PS and medium-sized end-linked OS formulations) do not employ spacers.
In addition, the immunogenicity profile of glycan epitopes that are composed of mostly homopolymers of monosaccharides with structural similarity to mammalian glycans is likely to differ from that of more complex glycan epitopes that contain unique  pathogen-specific monosaccharides, many of which exist 73 . Therefore, the minimal glycan chain length to elicit a robust antibody response may depend on the exact serological characteristics of the PS.
The complex relationship between glycan length and immunogenicity. While glycan length undoubtedly impacts the immunogenicity of unconjugated PS vaccines, its impact on the performance of glycoconjugate vaccines seemed less clear, although early studies supported this viewpoint 74,75 . The complex relationship between these two parameters in glycoconjugates may be brought into better focus when separating end-linked constructs from cross-linked formulations.
End-linked constructs. An early clinical trial in the 1980s investigated the immunogenicity of two end-linked glycoconjugates against Hib, where one contained glycans of an average chain length of 8 RU and the other an average chain length of 20 RU 15 . While adults did not produce different titers of specific antibodies in response to vaccination with these different formulations, infants produced much more anti-polyribosylribitol phosphate (PRP) antibodies when the 20 RU formulation was used. Importantly, both of these conjugates were made without spacer molecules, which may have affected the accessibility of the glycan epitope for binding. Perhaps affecting interpretability even further, these constructs were end-linked on both ends of most glycan chains, presumably leading to a low-to-intermediate level of cross-linking in the final product. The ostensible linear dependency of immunogenicity on glycan chain length was not corroborated in subsequent investigations on other end-linked formulations. Studies using glycoconjugate constructs of MenC PS fractions of different molecular weights (2-4, 4-10, and 10-50 kDa) conjugated to the carrier protein P64k did not identify differences in the immune response in BALB/c mice 76 . Moreover, a study using group B streptococcal (GBS) PSs conjugated with various end-linking chemistries to TT determined that immunogenicity of the preparations, as measured by IgG and opsonophagocytic titers in mice, increased with decreasing glycan size for GBS type II conjugates, whereas immunogenicity remained unaffected by glycan size for type III conjugates 77 .
A systematic investigation of the effect of chain lengths on the immunogenicity of primarily end-linked OS fragments of S. pneumoniae types 3, 6A, 18C, 19F, and 23F found little variation in antibody and opsonophagocytic titers generated by different conjugates that ranged in glycan molecular weights from 3 to 100 kDa 78 . However, a notable statistically relevant correlation was found for type 19F, where the shortest glycan fragments (about 8 kDa) generated by far the highest antibody titers in immunized rabbits, whereas native-sized conjugated glycan chains (100 kDa) displayed very poor immunogenicity, comparable with unconjugated native PS. The same study also confirmed that cross-linked native-length glycans (100 kDa) resulted in similar immunogenicity and opsonophagocytic killing as the end-linked construct produced with the shortest OS of S. pneumoniae type 6A (8 kDa).
Similarly, a study using end-linked conjugates of various sizes (10, 50, and 100 kDa) of capsular Hib PS to TT found that the shortest-length glycan achieved the highest immunogenicity of these preparations (as measured by IgG titers in rats), although statistical significance was not always achieved 79 . Finally, when testing several end-linked preparations of S. enterica serovar Typhimurium O-antigen (OAg)-CRM 197 conjugates, those that were made with OAg of high molecular weight were considerably less immunogenic in mice than their counterparts of lower molecular weight, or those where the different HPLC-separated OAg preparations were mixed 80 .
A possible explanation for these observed phenomena was identified in a study that investigated end-linked glycoconjugate preparations of the S. enterica sv Typhi Vi antigen, coupled to CRM 197 , in mice. These constructs ranged in average glycan size from 9.5 kDa (about 37 RU) to 165 kDa (about 636 RU). The authors found that the conjugates with the longest glycan chains elicited a T cell-independent response (likely by cross-linking BCRs and, surprisingly, induced late apoptosis of Vi-specific B cells in spleen and early depletion of Vi-specific B cells in the bone marrow 81 . These detrimental effects were not observed in shorter-chain endlinked constructs, which induced a more prolonged proliferation of Vi-specific B cells in the spleen compared to their long-chain counterparts. The latter also generated a faster decline of Vispecific IgG antibodies in mice than their smaller equivalents 81 .
Overall, these studies strongly suggest that medium-sized OS constructs result in better immunogenicity of end-linked glycoconjugates than long-chain PS formulations.
Cross-linked constructs. What about the effects of different glycan sizes on the performance of formulations that had been obtained using cross-linking chemistry? For group B type III streptococci, cross-linked glycoconjugate vaccine preparations of differing molecular glycan mass (approximate molecular weights of 349, 105, and 38 kDa) conjugated to TT produced different levels of specific IgG antibodies in mice. Titers largely correlated with saccharide length, although complete protection against a subsequent bacterial challenge was afforded by the two bigger saccharide fractions equally well 28 . A comparison of pentavalent pneumococcal cross-linked conjugate PS or OS vaccine preparations conjugated to CRM 197 in infants showed that the PS formulations consistently produced higher anticapsular antibody titers, although not always in a statistically significant manner 82 . When comparing antibody titers generated by repeated vaccination of mice with pneumococcal type 4 PS-and OS-based crosslinked TT conjugates, the PS formulation resulted in significantly higher IgM and IgG titers compared with their OS counterparts 83 . Infants aged between 18 and 30 months generated higher antibody titers against S. pneumoniae types 6A and 23F when immunized with cross-linked glycoconjugates of higher chain length (3-5 RU versus 10-20 RU versus native PS) 84 .
When investigating full-length or fragmented Vi antigen crosslinked glycoconjugates bound to different carrier proteins (CRM 197 , DT, and TT), full-size fragments elicited higher amounts of anti-Vi IgG antibodies in mice after the first immunizationhowever, this advantage became statistically insignificant after the second immunization 85 . A study on glycoconjugates of the Francisella tularensis OAg of different sizes coupled to TT showed that cross-linked vaccine preparations from a genetically modified bacterium resulting in higher molecular weight (around 220 kDa) elicited a more protective immune response than low and native molecular weight preparations (25 and 80 kDa, respectively). Remarkably, the higher efficacy of the 220 kDa preparation was not due to superior antibody titers, but likely because of the antibodies' improved affinity strength 35 .
These results all suggest that, unlike in end-linked glycoconjugates, in vaccines obtained with cross-linking conjugation chemistry, glycans of higher molecular weight may elicit a higher antibody response and/or improved protection. In cross-linked glycoconjugates, long-chain PS molecules may ensure optimal epitope density, with long glycan stretches that activate PSspecific BCRs. A carefully designed cross-linked long-chain conjugate vaccine avoids overly long glycan stretches. Such overly long chains may promote T cell-independent activation of B cells and subsequent hyporesponsiveness; the inability to mount an immune response after booster vaccination of at least the same magnitude as the response induced after primary vaccination 86 .
The most important parameter of a successful vaccine preparation remains its efficacy, i.e., whether it is able to prevent disease. Therefore, the next two chapters will briefly focus on the chain lengths of licensed glycoconjugate vaccines and investigate whether their efficacy correlates with glycan size. A second glycoconjugate of RRP OS molecules linked to CRM 197 was later produced with alternative chemistries, Vaxem-Hib. Finally, a vaccine consisting of chemically synthesized 8 RU of RRP end-linked to TT was also licensed, Quimi-Hib 60 (Table 2). Consequently, licensed monovalent glycoconjugate vaccines against H. influenzae include long-chain PSs, medium-sized OSs, and short-chain octamers.
Initial studies of the four primary licensed Hib vaccine constructs (PRP-D, PRP-T, and PRP-OMP, and HbOC) showed that all constructs but PRP-D generated mean serum antibody levels against PRP above 1 µg/mL 88,89 . The failure of PRP-D was likely related to its weaker carrier protein DT 27 . Notably, in a study of almost 500 American children, there were no statistically significant differences in the geometric mean antibody concentration after three doses of PRP-T or HbOC, with 95 and 91% of infants producing >1.0 µg/mL of antibody 90 . In addition to antibody measurements, efficacy data, the ultimate parameter for vaccine success, had been determined for the four constructs. Post dose 3, PRP-D was found to reach 90-100% efficacy when tested in thousands of Finnish children 91,92 , but only about 35% in Alaskan native infants 93 . For the OS-based formulation HbOC, efficacies had been determined to be 100% in two large studies with thousands of children 91,94 , while a smaller case-control trial in the United States determined 94.4% efficacy of HbOC after dose 3. For PRP-OMP, the efficacy of about 93% after dose 2 had been determined in a large trial involving Navajo and Apache native children 95 .
Subsequent studies investigated serum bactericidal antibody (SBA) titers of newer vaccine formulations and found that the endlinked OS-CRM 197 formulation Vaxem-Hib usually outperforms the cross-linked-TT PS-containing glycoconjugate ActHIB, although both generated high titers suggesting long-term seroprotection [96][97][98] . In addition, the reported antibody concentrations raised by Quimi-Hib, the synthetic 8-mer vaccine, matched that of Vaxem-Hib 12 months after the first immunization series 60 .
In conclusion, efficacious Hib glycoconjugate vaccines were obtained by both medium-sized OS-based end-linked constructs   102 , and NeisVac-C (MenC-TT, primarily end-linked with a moderate level of crosslinking, 20-47 RU) 103 . When directly comparing the performance of these three vaccine formulations, both MenC-CRM 197 vaccines were inferior to NeisVac-C in many studies [104][105][106][107] , and a hierarchy of NeisVac-C > Menjugate > Meningitec was usually observed. However, instead of glycan chain length, these differences in performance are perhaps associated with the fact that NeisVac-C used de-O-acetylated MenC OSs and TT as a carrier protein. Nevertheless, even Meningitec performed relatively well in absolute terms-a study evaluating its immunogenicity in over 100 British infants found that 98% of all participants achieved protective rabbit SBA titers of ≥1:8 post dose 3, with all of them achieving IgG titers of ≥2 µg/mL 108 .
Menactra, a meningococcal tetravalent ACWY conjugate vaccine, contains MenC glycans of roughly 20 kDa 109 , approximately 50 RU, end-linked to DT 110 . In trials encompassing hundreds of USbased infants, Menactra generated protective human SBA titers of ≥1:8 against MenC 30 days post dose 2 for 98.9% of participants 111 . The tetravalent ACWY end-linked mid-sized OSbased vaccine Menveo achieved the same level of human SBA protective titers ≥1:8 (82%) after a single dose as Menactra in a phase III study involving over 1000 adolescents 112 . In addition, phase III trials of the most recent addition to the arsenal of endlinked multivalent size-reduced MenC glycoconjugate vaccines, MenQuadFi, confirmed its non-inferiority compared to Menactra and Nimenrix and suggested a superior seroresponse against MenC, with protective human SBA titers 30 days after immunization in >99% of naive or preimmunized toddlers 12-23 months of age 113 , and >88% in adolescents and adults 114 .
There is only one MenC glycoconjugate construct that uses native PS, skipping a targeted glycan sizing step entirely. This construct is used in several cross-linked polyvalent vaccine formulations (MenHibrix, Nimenrix, and Menitorix). Immunogenicity experiments comparing MenHibrix with the OS-based end-linked MenC-CRM 197 monovalent vaccine (Menjugate) in hundreds of healthy infants suggested no significant difference in rabbit SBA With T-helper cell support, cognate B cells mature to memory B cells and produce glycan-specific antibodies. However, excessively long glycan stretches (construct d) may act similar to unconjugated PS and cross-link BCRs, generating mainly T cell-independent immune responses resulting in hyporesponsiveness. Without T-helper cell support, B cells become temporarily activated but proceed to undergo apoptosis, impeding the production of long-lasting memory B cells and long-lived plasma cells 81 .
titers after the third dose 115 . Similarly, Nimenrix showed favorable immunogenicity profiles after single and double injections, with 100% of more than 200 participating infants obtaining rabbit SBA levels ≥1:8 against MenC 1 month after the dose, and 90.8% showing that level 3 years post dose 116 .
In summary, end-linking technologies with medium-sized glycans as well as cross-linking chemistries with long-chain glycans lead to efficacious MenC conjugates in licensed vaccines with no obvious correlating differences in vaccine immunogenicity. This observation mirrors the data collected for Hib glycoconjugates, above.

CONCLUSIONS
A thorough review of the literature suggested that cross-linked glycoconjugates may require long-chain PSs for optimal immunogenicity, whereas end-linked constructs may benefit from medium-sized OSs. Either strategy can achieve an optimal density of native saccharide epitopes, allowing strong immune responses. Short chains of too few saccharide molecules may not contain an adequate density of accessible epitopes, while saccharide chains that are too long and formulations containing an excess of unconjugated PS molecules may carry the risk of promoting T cellindependent activation of B cells 81 by BCR cross-linking (Fig. 4). The model relies on the stimulation of either carrier-based peptide T-helper responses (the classical paradigm for conjugate vaccines) or carbohydrate-specific CD4 + T cell clones (Tcarbs) to produce the necessary cytokines for B cell activation 117 , a concept we eagerly await to be consolidated by future research.
Other factors determining vaccine efficacy may lie in the specific conformational features of both the carrier and the glycan components, and in specific features of the host, such as their metabolic state and the degree of senescence or maturity of their immune response system. Chain density per carrier protein molecule has also long been suspected to be of vital importance 15,78,118,119 and is heavily influenced by the process chemistries applied during vaccine preparation. Peak glycoconjugate performance may require the anchoring of minimal stretches of contiguous intact glycan epitopes to carrier proteins in a ratio that optimizes epitope loading for each protein molecule. At the optimal ratio, the carrier/glycan hybrid molecules maximally activate T helper cells.
In our survey, the clinical data on licensed glycoprotein vaccines of various glycan chain lengths do not indicate a relevant difference in protection, the ultimate goal of vaccination. We conclude that saccharide stretches in glycoconjugate vaccines need to be sufficient but not overly long for optimal vaccine performance. This crystallized paradigm can be delivered either via end-linked OS conjugates or via cross-linked PSs. This implies that successful glycoconjugate vaccines do not require long-chain lengths-an important realization that supports the future development of vaccines based on synthetic OSs that are linked to carriers via spacer molecules. Other technologies, such as GMMA and affinity non-covalent coupling, may also flourish 65,120 . The search for optimal glycan chain lengths (or, more precisely, the optimal number of contiguous epitopes between carrier proteins of glycoconjugates) will likely continue and be driven by efforts to minimize the size of each molecule without a loss (and possibly with gain) of function. In short term, though, bioconjugation using medium-sized OSs and direct enzymatic end-linking may offer the best balance between efficiency, simplicity, and optimal immunogenicity of glycoconjugate vaccines.