Systems immunology is the key to solving sepsis

Sepsis causes about 20% of deaths worldwide and most deaths from severe COVID-19. A dysfunctional immune response to infection leading to multi-organ failure, nearly a quarter of patients die even with intensive hospital treatment. It is critically difficult to diagnose, and we currently have no specific treatments that target it effectively. Doctors’ only options are supportive care and trying to identify and treat the underlying infection.

In their Frontiers in Science article, Hancock et al discuss how systems immunology could solve this medical dilemma. Systems immunology uses computational biology to analyze large amounts of data and give a detailed picture of the effect of sepsis on the whole immune system. It offers potentially crucial leads for diagnostics and treatments.

These tools could lead to diagnostics and specific treatments to lower the death toll from sepsis whatever its cause, including common infections and future pandemics.

This explainer summarizes the article’s main points.

What is sepsis?

Sepsis develops when the body’s immune system fails to cure an infection and becomes dysfunctional, leading to multi-organ failure and death. It is most dangerous to patients who are already vulnerable, and currently, there are no specific treatments. Patients in hospital receive antimicrobial treatment that targets the presumptive infection and supportive care.

Why is sepsis so hard to treat?

Sepsis is hard to treat for two reasons:

• it’s hard to diagnose in its early stages

• it presents and progresses differently in different patients, even varying in individual patients over time.

Early symptoms of sepsis, including changes to a patient’s temperature, heart rate, and breathing, could relate to many other diagnoses. Attempts at providing definitions of early sepsis to guide diagnoses and clinical care have thus far been unsuccessful. The authors also explain that many different factors are thought to influence the severity of sepsis, which makes it difficult to identify patients whose disease will progress faster or who are in more danger of organ failure.

The only way to determine which infection triggered sepsis is by identifying the infectious agent, for example by culturing bacteria sampled from a patient. However, sepsis is not always caused by bacterial infection, and it may be that the initial infection can clear from a patient’s system while sepsis continues. In one analysis, bacteria could not be cultured from approximately half of sepsis cases.

Combined, these factors delay diagnosis and treatment, and complicate research into effective treatments. We now understand that patients fit into groups (called endotypes) with different underlying causes of immune dysfunction. Ideally if we could define these underlying causes we would be able to treat individual patients so as to increase the effectiveness of their immune systems.

Does sepsis have long-term consequences?

Yes. Many patients who survive sepsis experience post-sepsis syndrome, which significantly affects health and quality of life. Up to 40% of patients have lasting physical, cognitive, or physiological symptoms. Patients are also vulnerable to new infections: more than 70% will be readmitted to hospital in the following year. This may be related to unresolved immune dysregulation, which could be driven by epigenetic changes—induced alterations in the way a person’s genes are expressed.

These complications, plus the initial episodes of sepsis, impose a burden of healthcare and socioeconomic costs which could be equivalent to the cost of strokes or coronary heart disease.

Since post-sepsis syndrome shares symptoms with long COVID-19, the authors express hope that increased research into long COVID-19 will also benefit sepsis survivors.

What is systems immunology?

Systems immunology considers the immune system as a highly integrated process, enmeshed with many other bodily systems. It uses very large amounts of varying types of data, collectively known as omics data, to gain a holistic understanding of how the immune system works. Omics data can include:

• genomics

• transcriptomics

• proteomics

• metabolomics

• epigenomics

Systems immunology uses different forms of computer analysis to identify and investigate patterns in these data, and how different factors interact. This includes machine learning, network biology analysis of interactions between individual components such as proteins, and clustering analysis, which groups patients according to shared characteristics. Furthermore, by taking omics samples from the same patients at different points in their disease it is possible to understand changes that take place over time.

These methods have led to important insights that could drive improvements in diagnostics and therapeutics. For instance, systems immunology research has indicated that there are several subtypes of sepsis, called ‘endotypes’, driven by different forms of immune dysregulation. Whether this immune dysregulation is fully resolved is linked to patients’ survival and recovery.

How can artificial intelligence help decode sepsis?

Supervised and unsupervised machine learning, both forms of artificial intelligence, can help provide insights into sepsis. Both approaches analyze volumes of data too large and complex for humans to manage.

• Supervised machine learning uses datasets that have been labeled—for instance, “septic” or “healthy”—to train an algorithm. This algorithm can then be supplied with novel data to indicate whether that data reflects a healthy patient or one with sepsis. Some algorithms also predict outcomes for patients.

• Unsupervised machine learning uses algorithms to classify unlabeled datasets. This is used to spot patterns and generate new hypotheses for experimental validation.

These models provide hypotheses and ideas we can investigate rather than solutions. Additionally, research into other complex, multifactorial conditions shows that models for predicting clinical outcomes which have been developed using machine learning can be very context-dependent and might not generalize well. However, careful model development that takes into account factors like the heterogeneity of sepsis can address this.

How could treating sepsis prevent future pandemic deaths?

Sepsis can be caused by any severe infection: most deaths associated with COVID-19 were the result of sepsis. This is thought to have been the case in historical pandemics as well, such as the 1919 influenza pandemic. By identifying early diagnostics and suitable treatments for different types of sepsis, we could develop pathogen-agnostic methods to treat the sickest patients in future pandemics.

What are the next steps for improving sepsis diagnosis and treatment?

Systems immunology offers promising paths towards diagnosing and treating sepsis, but it is expensive and requires specialized expertise. To overcome these challenges, the authors say we need an aggressive, fundable program which tackles:

• increasing investment

• increasing the volume of omics data to enhance systems immunology approaches

• development of precision medicine approaches

• identifying suitable animal or tissue models for research.

They also suggest that stakeholders might work with existing campaign groups, such as Surviving Sepsis or the Global Sepsis Alliance, to focus efforts and research to secure diagnostics and treatments.