To help manage the escalating burden of antimicrobial resistance, alternative treatment options for bacterial infections are becoming increasingly important. One option is to use therapeutic monoclonal antibodies targeting bacterial surface antigens. However, very few anti-bacterial antibodies have been found to be effective clinically¹. This could be because bacterial surface-exposed antigens tend to be highly diverse, and in addition, bacteria are often able to mask potentially conserved immunogenic surface components by carbohydrate structures such as capsules. There is a therefore a need to screen candidate antibodies against large collections of representative clinical bacterial isolates to determine the suitability of the antigenic target as well as the functional efficacy of the antibody.

Bacterial high-throughput phenotyping

Our team has access to large global and local collections of whole genome sequenced phylogenetically organised bacteria. Recently, we have been establishing a system for linking genomic information with high-throughput bacterial phenotyping. To do this we have been using an Opera Phenix high-content confocal microscope from Perkin Elmer, and optimised high-resolution bacterial imaging in combination with automated image analysis. We currently use this to model the behaviour of major nosocomial bacterial pathogens under antimicrobial pressure, as well as screening therapeutic monoclonal antibodies against multi-drug resistant bacteria.

Linking antibody-induced bacterial phenotypes with antibody function

The idea to image antibody binding came about during discussions with our collaborators at Kymab Ltd, who were starting to set up their bacterial antibody pipelines at the time. Together, we applied to the MRC Proximity to Discovery: Industry Engagement Fund Biomedical Research Exchange Programme in 2018 which got funded to set up a small proof-of-principle study to see if we could use our high-content imaging platform to explore monoclonal antibody binding at a single bacterial cell level.

We chose Escherichia coli ST131 O25b as an example since a monoclonal antibody (3E9-11) targeting the O25b O-antigen was previously shown to have promising anti-bacterial activity². E. coli ST131 O25b is a dominant multi-drug resistant clonal linage of E. coli and the major cause of Extraintestinal Pathogenic E. coli (ExPEC) infections worldwide³. Additonally, we had access to a panel of whole genome sequenced clinical isolates from Addenbrookes Hospital in Cambridge.

Antibody binding analysis using high-content imaging: bacteria are incubated with antibody in 96-well plates, then stained using DAPI and Alexa Fluor-647 secondary antibody and imaged using an Opera Phenix microscope. Different binding phenotypes can be classified using image analysis. (schematic was created using icons from BioRender.com) 

Using a variant of 3E9-11, we initially screened our panel of closely related E. coli ST131 O25b with the aim to investigate what proportion of isolates the antibody would bind to and determine if we could estimate the binding affinity based on staining intensities. This worked really well, and interestingly we found distinct binding phenotypes within our data set. These were classified as no binding, weak binding, strong binding and strong agglutinating binding. We followed up on the different phenotypes and found that no binding or weak binding was associated, in many cases, with identifiable mutations or insertion sequences within the lipopolysaccharide core or O-antigen biosynthesis gene clusters. Further experiments revealed that the agglutinating binding phenotype was associated with enhanced phagocytosis and complement susceptibly, which appeared to be linked with less surface matrix surrounding the outer membrane. However, the majority of the isolates did not display an agglutinating binding phenotype, and most were resistant to complement even in the presence of the antibody. Alarmingly, a significant proportion of the remaining complement susceptible isolates were protected from complement killing in higher antibody concentrations. Our data highlights the importance of screening candidate therapeutic monoclonal antibodies against large panels of clinically relevant isolates as even genetically closely related isolates from the same clonal linage can differ significantly phenotypically.

What started off as a small proof-of-principle side project led to a scalable, robust screening method, linking antibody binding phenotype to functional efficacy. We are currently using this approach to screen novel monoclonal antibodies against panels of nosocomial bacterial pathogens in collaboration with Kymab Ltd.

Mailis Maes is a 2^nd year PhD student with Gordon Dougan at the University of Cambridge studying the phylogenetics of Salmonella Typhi in Latin America, as well as the behaviour of Salmonella Typhi within macrophages and organoids. 

Josefin Bartholdson Scott is a Research Associate in Stephen Baker's lab at the University of Cambridge currently focussing on Salmonella Typhi and Salmonella Paratyphi A vaccine and therapeutic antibody targets.

1. McConnell, M. J. Where are we with monoclonal antibodies for multidrug-resistant infections? Drug Discov. Today 24, 1132–1138 (2019)

2. Szijártó, V. et al. Bactericidal monoclonal antibodies specific to the lipopolysaccharide O antigen from multidrug-resistant Escherichia coli clone ST131-O25b:H4 elicit protection in mice. Antimicrob. Agents Chemother. 59, 3109–3116 (2015)

3. Petty, N. K. et al. Global dissemination of a multidrug resistant Escherichia coli clone. Natl. Acad. Sci. U. S. A.111, 5694–5699 (2014)