Recent advances in genomics and reverse vaccinology have identified promising protein targets for vaccines 
. In many cases, suitable candidate antigens for Gram-negative bacterial vaccines are outer membrane proteins and these pose particular challenges in their expression and purification and in serotype variability. An ideal delivery system especially for bacterial vaccines for developing countries will encompass multiple antigens and enable vaccines to be rapidly tailored to local and changing antigenic serotypes. Ideally, it will also be inexpensive to manufacture. We propose a platform for rapid development and delivery of vaccines against Gram-negative bacteria. The approach is based on the production of outer membrane particles we have named GMMA by genetically modified bacteria. Using genetic manipulation, it is possible to increase their yield, to remove immunodominant structures, to overexpress certain antigens, and to reduce the endotoxic activity 
. GMMA could potentially be a safe, effective and low cost vaccine but need a practical way of manufacture at scale.
53G was chosen for a first approach to develop a scalable process and a null mutation of the tolR
gene was introduced to overproduce GMMA as previously described for E. coli
. To verify that the process is applicable to produce GMMA harboring modified lipid A, which would be more suitable for use as vaccine, and/or lacking the O antigen of the LPS we grew high density cultures of S. sonnei
ΔgalU, S. sonnei
(cured of the virulence plasmid pSS), and S. sonnei
in a 5 L fermenter in complex (HTMC) or chemically defined medium. Chemically defined medium was used to avoid contamination from proteins present in complex media and to have the possibility to regulate iron concentration.
Bacteria were removed from the culture supernatant by a tangential flow filtration step using a 0.2 µm membrane. A second tangential flow filtration step with a 0.1 µm membrane was used to concentrate GMMA and to remove soluble proteins. This choice of appropriate molecular weight membranes allowed the purification of GMMA in an easy, efficient, and scalable process. After purification, approximately 90% of all protein was consistently GMMA-associated with reproducible yields of more than 100 mg of GMMA-associated protein per liter fermentation volume from OD 30–45 cultures of S. sonnei ΔtolR ΔgalU. The integrity of GMMA obtained by this process was confirmed using electron microscopy. The purity and yield can likely be increased as indicated by fermentations with S. sonnei –pSS ΔtolR ΔmsbB to densities of 65 and 80. Furthermore, first results obtained by quantitative amino acid analysis of different types of GMMA indicated an at least two-fold higher protein amount in the GMMA preparations than determined by the Bradford assay used in this study (data not shown). Still, assuming an average yield of 100 mg/L fermentation and a dosage of 25 µg as used for the MeNZB outer membrane vesicle meningococcal vaccine, at least 400,000 doses could be obtained from a 100 L fermenter.
A proteomic approach confirmed that Shigella sonnei
-derived GMMA are composed mostly of outer membrane and periplasmic components. They conserve lipophilic polypeptides. Only a small number of cytoplasmic components and one inner membrane protein were predicted. Thus, the proteomic analysis of GMMA obtained from an OD 45 culture revealed a similar composition as previously seen in proteomic analyses of outer membrane particles that were obtained from cultures at early logarithmic phase to avoid impurities by cytoplasmic proteins 
In accordance with previous reports 
GMMA were highly immunogenic in mice with titers around 1
100,000 after administration of 2 µg of GMMA with and without adjuvant. A 10-fold lower dosage of GMMA (without adjuvant) resulted in only a 3-fold reduction and still very high antibody titers suggesting that low amounts of GMMA might be sufficient for vaccination. GMMA from the msbB
mutant S. sonnei
strain did not show a difference in immunogenicity which was expected due to a recent report that the resulting lipid A modification does not affect LPS recognition in mice 
. Immunoblots confirmed that antibodies to proteins, including outer membrane proteins OmpA, OmpX, and YaeT, strongly contributed to the reactivity of the sera. Interestingly, the outer membrane protein OmpC which represents about 20% of protein in GMMA was not detected by sera raised against GMMA. Previously, an immunoproteomic analysis of isolated outer membrane proteins of Shigella flexneri
also failed to detect OmpC as immunogenic protein. This could suggest that either OmpC is not immunogenic or that epitopes potentially recognized by antibodies are not maintained after SDS-PAGE. This might also apply to other membrane proteins that were not found by the Western blot analysis even though not all reactive proteins could be identified.
mutant strain of Shigella
lacking the genes msbB1
was generated to investigate if the production process was applicable to GMMA with modified lipid A. A previous report 
had shown that these deletions result in the synthesis of a penta-acylated lipid A instead of a hexa-acylated lipid A in Shigella
. While the S. sonnei
mutant grows in rich media at 37°C temperature, its growth is impaired in the chemically defined medium developed for fermentation at 37°C but shows a normal growth in this medium at 30°C. Previously, a Shigella flexneri
mutant and an E. coli msbB
mutant in the K-12 background were reported not to show any growth defects 
. In contrast, an msbB
mutant of the clinical isolate E. coli
H16 formed filaments when grown at 37°C but not at 30°C or when functionally complemented by the cloned msbB
. The S. sonnei
mutant strain used in this study does not form filaments. The reason for the slower growth at 37°C, especially in defined medium, is not clear and could be a result of the background of the strain, the combination of the tolR
mutation, or a suboptimal composition of the defined medium that can likely be optimized. Importantly, a comparison of the protein pattern of GMMA generated from S. sonnei
at 37°C and 30°C showed only minor differences in the protein profile visible by SDS-PAGE indicating that the change in temperature does not have major effects on GMMA composition.
In summary, we have identified an easy process to produce large quantities of GMMA from high density culture. GMMA purified from fermentation are extremely pure particles composed almost exclusively of outer membrane and periplasmic components. The simplicity and high yield of the process support its applicability for large scale manufacturing. We have also shown that this process can be used with strains genetically modified to reduce reactogenicity or to remove immunodominant antigens, e.g. the O antigen. While this work focused on Shigella sonnei, we believe that this technology is an innovative platform for efficient vaccine manufacturing for Gram-negative bacteria.