Development of robust analytical pipelines for the analysis of microbial community data from clinical samples

Cystic fibrosis is an inherited disease which affects around 9,000 people in the UK. It is a recessive disorder, meaning that both parents have to carry a faulty copy of a gene for a child to be affected. Approximately 1 in 25 people carry this copy. Cystic fibrosis dramatically shortens the life of those affected and almost half do not live beyond their 30s. Cystic fibrosis has a negative effect on many parts of the body, but it particularly affects the lungs. The result of the defective gene is that thick secretions cannot be effectively cleared from the lungs, resulting in the airways becoming congested and damaged.

This congestion results in frequent and severe exacerbations, sometimes requiring admission to hospital and treatment. We think exacerbations are often a result of infections, which may be caused by viruses and bacteria. One of the most common bacteria found is Pseudomonas aeruginosa. Microbiologists diagnose Pseudomonas by putting samples of sample onto plates and identifying the bacteria that grow in visible colonies. Often Pseudomonas is treated by antibiotics. Other bacteria may be found including Streptococcus and Staphylococcus.

However when looking for causes of exacerbations we are limited to finding pathogens that we know about, and which grow on the types of culture plates we use. It may be that other bacteria are present that we can't see because they do not grow easily. Sometimes those bacteria may be a cause. However, in a similar way to the human gut it may be that there are "good" and "bad" bacteria. It is known that certain types of bacteria can prevent other types infecting and therefore they may help protect against exacerbations.

A new technology utilises the idea of molecular barcodes which identify bacterial species from fragments of DNA in their cell. New instruments termed high-throughput sequencers permit these barcodes to be read from many samples easily and cheaply. This technology gives us a "parts list" of the bacterial species in a particular sample, and a rough idea of how frequently they occur. By reading this parts lists from patients with cystic fibrosis - when they are well, when they are very sick and when they are recovering, we may be able to tell the relative contribution of these unseen bacteria to the condition. For example a particular species increasing or reducing in abundance may be associated with recovery, giving us a potential therapeutic target.

We are also interesting in seeing how patients end up being colonised with particular, commonly seen bacteria. For example in our local patients, about 30% have a particular type of Pseudomonas infection called Midlands 1. However, very little is known about how it is that so many patients end up being infected by the same strain. We are now able to sequence all the DNA in a bacterial cell (the genome) which gives very high resolution view of how it has evolved. By comparing genomes of strains from different patients, we can help determine whether patients are infecting each other with the same strain, or whether the strains are quite different and come from many different sources. We can also see how the Pseudomonas evolves whilst it is in a patients lungs. Previous studies have shown that the Pseudomonas adapts to the specific environment, which may give us clues as to why people with the same cystic fibrosis mutation have different courses and end up with more hospital admissions or exacerbations than others.

The technology we are using is very new, and there are a number of difficulties with it before it can be a routine clinical test. One problem is that the machines generate plenty of sequencing "noise" which may look like species are present that aren't. I want to develop bioinformatics methods that try and increase the reliability of these techniques and generate information that would be useful for clinicians. We expect these techniques to enter the clinic within the next five years.

Cystic fibrosis (CF) affects around 9,000 people in the UK and is one of the most common inherited causes of premature death in adults. CF is inherited recessively from mutations found in the gene encoding for the CF transmembrane conductance regulator (CFTR), an epithelial chloride ion channel. In the respiratory tract, defective chloride ion transport leads to dehydration of the mucosal surface and consequent failure of the mucociliary escalator and plugging of the airways by hyperviscous mucus. Although, without treatment, premature death is inevitable, in developed countries management of the condition has resulted in a dramatically increased life expectancy with most patients now surviving into adult life. However, still only half of CF patients survive into their 40s.

During this fellowship, I aim to answer several key questions relating to chronic lung infection in cystic fibrosis, particularly:
- Is there a typical CF microbiota, and what is it like?
- How does the CF microbiota vary according to disease severity?
- Why has the Pseudomonas aeruginosa Midlands-1 clone established itself in our local cohort, and what are the routes of transmission?
- How does Pseudomonas and other species evolve to adapt to the host during colonisation?

Answering these questions requires the concurrent addressing of important technical questions, specifically:
- Which benchtop instrument (Ion Torrent and Illumina MiSeq) performs best for 16S phylogenetic surveys?
- How well does whole-genome shotgun metagenomics perform compared to 16S analysis for phylogenetic surveys?
- What other information does whole-genome shotgun metagenomics provide compared to 16S?
- Are these methods statistically robust enough to enter clinical practice?
- Can 16S and metagenomics data be presented in a way useful for clinicians?

The methods employed will include use of 16S amplicon sequencing, whole-genome shotgun sequencing and whole-genome metagenomics (see case for support).

This work will have a direct impact to patients, and in particular cystic fibrosis sufferers and their carers who will benefit from an enhanced understanding of the polymicrobial nature of CF lung infection and bronchiectasis. These findings will be useful for those looking at ways of developing novel therapeutics targeting the polymicrobial communities, for example probiotics and bacteriophage therapy. The community will be engaged through the CF patient representative, the CF patient group and the Heartlands CF newsletter and kept abreast of the progress of the study, as well as on the Heartlands CF website (www.heartlandscf.org). An executive summary plus the anonymised results of the final report will be sent to GPs of the subjects involved in the study and published on the CF website and publicised via the CF newsletter. The final report will also be offered for circulation to relevant organizations interested in cystic fibrosis, for example the Cystic Fibrosis Trust and Cystic Fibrosis Medicine.

The economic and social impact of this research will stem from the ability to make more timely and accurate diagnosis. A better understanding of the effects of antimicrobial treatments may lead to the development of more personalised treatments, and potentially a change in the way we treat cystic fibrosis exacerbations. The result of improved cystic fibrosis care with the aim of reducing exacerbations will improve the quality of life of both patients and their carers.

The work on metagenomic data analysis will have an impact on those studying microbiomics of disease, and more broadly the whole field of microbial ecology, due to the novel methods developed during this fellowship. The trials of metagenomics using benchtop sequencers will bring the knowledge required to start using these cutting-edge techniques in the clinical arena, complementing existing conventional microbiological diagnostics and providing a more rich picture of the microbial ecology of the respiratory tract. These techniques can be applied to the field as a whole.