Vaccines Incorporating Toll-Like Receptor Ligands

Published on: 
BioPharm International, BioPharm International-10-02-2008, Volume 2008 Supplement, Issue 6

This article discusses the production process of the major influenza antigen, hemagglutinin (HA), by rDNA methods in E. coli.


This article discusses the production process of the major influenza antigen, hemagglutinin (HA), by rDNA methods in E. coli. Fusing the gene for HA to the gene for flagellin, a Toll-like receptor (TLR) ligand, yields a bi-functional protein. The HA moiety contains the structures recognized by the immune system as it generates neutralizing antibody, and the flagellin targets the HA antigen into the appropriate compartment of an antigen presenting cell. Having the antigen and the TLR ligand physically connected drives a robust antibody and cellular immune response. Producing this protein in E. coli provides the additional benefit of high yield per culture volume and global portability. These vaccines could help immunize the global population rapidly during an influenza pandemic.

Next to clean water and sanitation, vaccines have had the greatest impact on public health. Early vaccines were based on animal pathogens related to human pathogens (cowpox versus smallpox, Mycobacterium bovis versus Mycobacterium tuberculosis), killed pathogens (pertussis, poliovirus, typhoid fever), attenuated viruses (yellow fever, measles), or components of pathogens (tetanus toxoid, diphtheria toxoid).1 Although these vaccines were considered to be quite effective in protecting against disease, their use was limited, in some cases, because of poor tolerability. In addition, not all infectious diseases could be addressed by these limited vaccine approaches.


The advent of molecular biology in the late 1970s brought the promise of vaccines made of highly purified, well-characterized subunits of pathogens, which could be produced by cloning and expressing pathogen genes using recombinant DNA techniques. There have been two spectacular successes in this realm: the hepatitis B vaccines licensed in the mid 1980s, and the human papillomavirus vaccines licensed in the past year. These two successful vaccines are based on virus-like particles, which are made up of self-assembling virus capsid proteins. These virus-like particles display hundreds of arrayed viral epitopes on their surface, which allows the particles to be picked up, processed, and presented by the antigen-presenting cells (APCs) of the immune system in a very efficient manner.1

Despite these successes, the fact is that the vast majority of viral antigens do not self assemble into regular arrays and thus are not presented efficiently by APCs. For the past two decades, researchers have been cloning, expressing, and purifying proteins from pathogens and testing them as vaccines. Overall, these monomeric proteins have been found to be poorly immunogenic. The collective experience has shown that while natural infection generally results in a robust and durable immune response to a variety of components of the pathogen, vaccination with the same components as purified proteins does not. Clearly, something was missing from vaccine candidates based solely on purified recombinant viral proteins.

The Role of Adjuvants

For years, researchers have known that one can improve the immunogenicity of an antigen by adding an adjuvant. Freund's Complete Adjuvant (FCA), has been the benchmark for laboratory work. FCA, an oily emulsion containing killed mycobacteria, can raise both robust antibody and cytotoxic T-cell immune responses in vaccinated animals. But this comes at the expense of a high frequency of sterile abcesses, making FCA unsuitable for human use. The adjuvant that is extensively used in man is alum, a generic term for a range of particulate forms of aluminum sulfate mixed with sodium hydroxide and varying amounts of phosphate. Aluminum-containing adjuvants work both as depots—most antigens can be made to stick to it—and as irritants, drawing cells of the immune system to the site of injection. Antigens formulated on aluminum-containing adjuvants tend to elicit good antibody responses, but not cytotoxic T-cell responses. However, for most naturally caused infectious diseases, both antibody and cytotoxic T-cell responses are required to control the disease and protect against future infection. This suggests that naturally infecting pathogens must carry some kind of FCA-like adjuvant activity.

The Advent of Toll-like Receptors

In the mid-1980s, the Toll protein of the fruit fly, Drosophila melanogaster, was first described.2 In Drosophila, Toll controls, among other things, the release of peptides that are active against infectious agents, a major line of defense.3 In subsequent years, a related set of Toll-like receptors (TLRs) were found on or in mammalian APCs, including dendritic cells, macrophages, endothelial cells, and some plasmacytoid cells.4 Together with neutrophils, eosinophils, basophils, and mast cells, these constitute the "innate immune system." These cells mount the first attack against incoming pathogens, resulting in an inflammatory response that is the necessary predecessor of the T-cell and B-cell "adaptive immune response."


In 1997, Medzhitov, Janeway, and others demonstrated that TLRs are the mediators of the innate immune response.5 TLRs collectively recognize families of structures that are unique to pathogens and not part of mammalian biology. These structures are collectively termed pathogen-associated molecular patterns (PAMPs), and they provide the primary signal to the mammalian host that a pathogenic insult has occurred.5 Figure 1 illustrates the known TLRs and their ligands, or PAMPs.6 Each TLR binds a specific class of pathogen-related molecules. For example, TLR4 binds to lipopolysaccharides (LPS) found on the surface of bacteria but not in the mammalian host.1 TLR3 binds to double-stranded RNA (dsRNA), a non-mammalian nucleic acid conformation that is the genome structure or replication intermediate of many viruses.1 TLR5 binds to flagellin, a protein that polymerizes to form the bacterial flagella.1 Mammalian TLRs are expressed either on the surface or in an internal compartment inside of an APC. In general, cell surface TLRs tend to recognize bacterial components (e.g., lipoproteins, lipopolysaccharides, flagellin) whereas internal TLRs tend to recognize nucleic acids and analogs (e.g., dsRNA, CpG oligonucleotides, or nucleotide derivatives).

Figure 1. The known Toll-like receptors (TLRs) and their ligands. Pathogen-associated molecules are recognized by TLRs.

PAMP and TLR Binding

Binding of a PAMP to a TLR results in a cascade of events inside an APC (Figure 2, unpublished data). First, the PAMP-pathogen complex binds to a TLR and is internalized to an endocytic vesicle. The vesicle now contains an activated TLR that sends a series of signals to the nucleus. These signals drive the fusion of the vesicle with a lysosome where the contents are degraded into peptides for presentation in the groove of the MHC class I or class II complex.5 TLR signaling triggers expression of genes for cytokines, which stimulate neighboring T cells as well as the expression of co-stimulatory molecules (CD80/86) that bind to the T cell and promote its activation.1 This triad of peptide processing and presentation, cytokine secretion, and T-cell co-stimulation satisfies the criteria for activating a T cell and initiating a productive immune response.

Figure 2. Toll-like receptors' mechanism of action: the stimulation of Toll-like receptors activates pathways that 1) promote antigen processing and presentation and 2) provide the second signal to T cells, informing them that the antigens are proper subjects for vigorous immune response.

How does this relate to vaccines? First, it explains why FCA works so well. Mycobacteria, the active ingredient of FCA, are also pathogens, and therefore, contain numerous PAMPs. More importantly, it suggests that using a TLR as a portal of entry for an antigen into an APC might result in more efficient antigen processing and presentation. This, in turn, should elicit a more robust immune response.

Multiple PAMP–TLR pairs are available for use as a portal of entry (Figure 1).6 However, only one of these pairs, flagellin-TLR5, involves a purely protein ligand and a cell- surface TLR. Coupling the DNA sequence for flagellin in frame with the DNA sequence for a protein antigen and expressing this chimeric assembly in E. coli yields a protein that contains the cell entry and immunostimulatory activity of both a TLR ligand and an antigen in a single molecule. This ensures that both components enter not only the same cell but also the same endocytic vesicle within that cell. Thus, the target antigen is present within an endosome that contains an activated TLR, ensuring that both components are trafficked through the same endosome-lysosome-processing and presentation cycle in close proximity. Elegant confocal microscopy studies using other PAMPs, including LPS, have shown that this is the case at the cellular level.4

Co-Delivering Antigens and TLR Ligands to APCs

We have built our platform technology on the concept of co-delivering vaccine antigens and TLR ligands to APCs. Using this technology, we have demonstrated that coupling flagellin to a model antigen, ovalbumin, can elicit a rapid IgG antibody response in mice one week after a single dose; ovalbumin alone on aluminum hydroxide takes three weeks to reach the same level of antibody.7 The same flagellin–ovalbumin fusion elicits a classical T-helper 1 antibody and CD8+ T-cell response, while ovalbumin alone, or mixed with flagellin, yields a classical T-helper 2 response.1,7 Similar results have been obtained by a fusion of flagellin to antigens from Listeria monocytogenes.1

The same strategy has recently been applied for developing a West Nile virus vaccine that is based on the conformation-dependent structure of a subunit of the envelope of the virus, EIII. EIII contains a disulfide bond that is essential for proper folding of the domain and presentation of epitopes that are recognized by virus-neutralizing antibodies.8 Genetic fusion of the sequence for EIII to the sequence for flagellin yields a protein that retains TLR5 activity and properly folds the EIII domain such that it is recognized by neutralizing monoclonal antibodies and induces protective immune response in mice.8 The ability to express such a conformation-dependent protein in E. coli suggested that this strategy could be applied to other vaccine targets.7

Application in M2e

This technology has also been applied to both seasonal and pandemic influenza vaccines. Two distinct, but complementary, vaccines are currently in development. The first is based on the highly conserved ectodomain of the viral M2 ion channel protein (M2e). M2e is a 24 amino acid extension of M2 protruding through the membrane of the virus and the infected cell. Antibodies raised against the M2e sequence have been shown to reduce the rate of spread of influenza virus in culture and to protect mice from a lethal challenge of virus. The M2e sequence is highly conserved across the H1, H2, and H3 serotype viruses that have infected humans over the past 90 years.7,9 It is also well-conserved in the avian serotypes, suggesting that a cocktail of three or four M2e vaccines could provide broad protection against any influenza virus. A significant advantage of such an approach would be that the vaccine would not have to change on an annual basis, unlike the current influenza vaccine strategies.

Figure 3. Influenza M2e vaccine: mice vaccinated with 0.3 or 3.0 μg of globular head fusion protein are protected 100% from death following a challenge with virus that kills 90% of naïve mice

M2e as a peptide is poorly-immunogenic and ineffective, even when delivered at high doses on aluminum-containing adjuvants. Four copies of M2e head-to-tail mixed with flagellin, rather than fused, are also ineffective. However, fusing four head-to-tail copies of the human consensus M2e sequence to the C-terminus of flagellin creates a highly potent M2e vaccine. Doses of 0.3 μg provide 90–100% protection against a challenge dose of virus that kills 90% of naïve mice (Figure 4, unpublished data),10 reinforcing the value of coupling a TLR ligand to an antigen. The first study of M2e in man, funded by the Gates Foundation, has been completed and it was found that the vaccine is highly immunogenic at low doses.

Figure 4a. Influenza STF2.1HA vaccine efficacy: 0.3 μg provide 90–100% protection against a challenge dose of virus that kills 90% of naïve mice

Application in Hemagglutinin

Another application of this technique has been in the major antigen of the virus, hemagglutinin (HA), one of the two predominant proteins on the viral surface. HA binds to sialic acid sugars on the surface of cells, and therefore, is a prime target for the immune system. Virus-neutralizing antibody raised by natural infection is known to be directed to the area of HA around the sialic acid binding site, and can thus prevent binding and block infection.11 Viruses that have mutations in the HA gene in the region of the sialic acid binding site have an infectious advantage over their neighbors, and thus can dominate the population (a primary reason the influenza vaccine needs to be updated every year).

Figure 4b and 4c. Low doses of PR8 STF2.HA subunit protect mice against challenge with homologous virus

Current inactivated influenza vaccines are made by growing the virus in embryonated eggs. The allantoic fluid is harvested, and the virus in it is killed with formalin. The killed virus is enriched by centrifugation and then disrupted with detergent, and is generally followed with a second formalin kill step. The result is the bulk vaccine, of which HA comprises up to 20% of the protein component.12

We have discovered a means of producing just the "globular head" of HA, the segment of the protein that contains all of the structures recognized by neutralizing antibodies. This globular head can be produced as a fusion protein at the C-terminus of flagellin in E. coli, and can be purified to endotoxin-free homogeneity. Although the globular head moiety is a complex structure with disulfide bonds that must be configured and folded correctly for optimum immunogenicity, we have found that the purified protein reacts strongly with serum from convalescent animals as well as with monoclonal antibodies directed at the neutralizing epitopes, suggesting that it is appropriately structured.

In addition, vaccinating mice with two doses of the HA globular head fusion can raise antibodies that recognize native HA on infected cells and also have the ability to block the virus's ability to agglutinate red blood cells, a classic assay for quality of HA antigens. Mice vaccinated with 0.3 or 3.0 μg of globular head fusion protein are protected 100% from death following a challenge with virus that kills 90% of naïve mice (Figure 3).7 This vaccine is also highly immunogenic in rabbits and ferrets (unpublished data). As seen with other proteins, the globular head alone, or mixed with flagellin, is not immunogenic. In the last 12 months, we have made a variety of HA vaccines with different serotypes, including H1 PR8, H1 New Caledonia, H1 Solomon Islands, H3 Wisconsin, H5 Vietnam, H5 Indonesia, H5 Anhui, B Malaysia, B Lee, and B Shanghai. All are well-expressed in E. coli (20–25% of total cell protein on induction) and the first clinical trial is underway. A new vaccine seed strain may be made and initial protein produced in a month, providing a rapid response to an emerging virus.

The speed, simplicity, and efficiency of production of these vaccines in E. coli provides an attractive alternative to the present egg-derived influenza vaccine. A typical protein yield is >2 g/L of fermentation broth, and this may be optimized further. Compared to the average of about 6–8 mg of HA recovered per liter of allantoic fluid of infected eggs, and 0.6 to 6 mg of HA per liter from mammalian cell culture, an E. coli-based vaccine approach is highly attractive (unpublished data).

Pandemic Preparedness

The US Department of Health and Human Services has stated that its goal is to have within the US the capacity to make enough monovalent pandemic influenza vaccine to give two doses to the entire US population, or 600 million doses. Current egg-based systems are unable to meet this demand, and even with recent enormous investments in expanded egg facilities and new cell-culture efforts, there will be a shortfall. As described here, an E. coli-based system could produce the required vaccine in one 5,000-L fermenter cycle. Moreover, because the methodologies are standard, the production process could easily be transferred to other countries or regions. Given the projected worldwide impact of a pandemic, local or regional manufacture of this vaccine may be able to provide countries or regions without a current flu vaccine, the means to be self-sufficient. With this kind of production capability, one can contemplate vaccinating the global population effectively and rapidly, if the necessary public health infrastructure could be developed.

Alan Shaw, PhD, is the president and CEO of VaxInnate, Cranbury, NJ, 609.860.2260,


1. Huleatt JW, Jacobs AR, Tang J, Desai P, Kopp EB, Huang Y, et al. Vaccination with recombinant fusion proteins incorporating Toll-like receptor ligands induces rapid cellular and humoral immunity. Vaccine. 2007;25(4):763–75.

2. Anderson KV, Bokla L, Nusslein-Volhard C. Establishment of dorsal-ventral polarity in the Drosophila embryo: the induction of polarity by the Toll gene product. Cell. 1985;42(3):791–8.

3. Bettencourt R, Tanji T, Yagi Y, Ip YT. Toll and Toll-9 in Drosophila innate immune response. J Endotoxin Res. 2004;10(4):261–8.

4. Blander JM, Medzhitov R. Toll-dependent selection of microbial antigens for presentation by dendritic cells. Nature. 2006;440(7085):808–12.

5. Medzhitov R, Janeway CA, Jr. Innate immunity: impact on the adaptive immune response. Curr Opin Immunol. 1997;9(1):4–9.

6. Kawai T, Akira S. Pathogen recognition with Toll-like receptors. Clin Op Immunol. 2005;17:338–344.

7. Huleatt JW, et al. Potent immunogenicity and efficacy of a universal influenza vaccine candidate comprising a recombinant fusion protein linking influenza M2e to the TLR5 ligand flagellin. Vaccine. 2008;26:201–214.

8. McDonald WF, Huleatt JW, Foellmer HG, Hewitt D, Tang J, Desai P, Price A, Jacobs A, Takahashi VN, Huang Y, Nakaar V, Alexopoulou L, Fikrig E, Powell TJ. A West Nile virus recombinant protein vaccine that co-activates innate and adaptive immunity. J Infect Dis. 2007;195:1607–17.

9. Zebedee SL, Lamb RA. Growth Restriction of Influenza A Virus by M2e Protein Antibody is Genetically linked to the M1 Protein. Proc Natl Acad Sci USA. 1989 Feb;86(3):1061–5.

10. Song L, Nakaar V, et al. Efficacious recombinant influenza vaccines produced by high yield bacterial expression: a solution to global pandemic and seasonal needs. PLoS ONE. 2008;3:e2257.

11. Yewdell JW, Caton AJ, Gerhard W. Selection of influenza A virus adsorptive mutants by growth in the presence of a mixture of monoclonal antihemagglutinin antibodies. J Virol. 1986;57:623–628.

12. Matthews JT. Egg-based production of influenza vaccine: 30 years of commercial experience. National Academy of Engineering. 2006;36.