Tuberculosis (TB) is still a serious public health concern worldwide, with World Health Organization data suggesting an estimated 10.7 million people developed TB in 2024, with 1.23 million deaths, making TB a leading cause of death from a single infectious agent. Treatment of TB is challenging and requires a 6-month course of multiple antibiotics. This can make compliance difficult, which increases the risk of the development of antibiotic resistance. Identifying exposure or infection as early as possible could make treatment more effective. CLI chatted with Professor Nophar Geifman (University of Surrey, UK) to learn more her team’s work that has identified different plasma protein signatures in cohorts of people before and after TB diagnosis, which could ultimately aid identification of early infection or even people who are going to develop an infection.

Can you give us a brief introduction to tuberculosis, please?

Let me start by saying that I’m not an expert in tuberculosis (TB) but, rather, my background is in health and biomedical informatics. We came to this project from a data analysis perspective – we were looking to study a contagious disease that has an acute onset perhaps, but also that has a chronic element and TB was exactly that example for us.

Broady, TB, is a contagious bacterial infection caused by Mycobacterium tuberculosis. It primarily attacks the lungs and it causes a range of different symptoms, including – and probably most well known – a cough, which can be persistent and also can cause the coughing up of blood and mucus. Also, there are other symptoms that are perhaps a bit generic, such as fatigue, loss of appetite, and then loss of weight that comes with that. Also pain that comes with the coughing as well as a fever. That’s usually how TB presents. There are, however, also other rarer forms of the disease, when the infection actually goes to other parts of the body. One example would be when it attacks the lymph nodes and then you get symptoms that are associated with those particular areas of the physiology, so one gets swollen glands as well.

TB is an interesting type of infection because it can be split into two kinds: active TB and then the latent type. Active TB, as the name suggests, is when you see the symptoms and the bacteria is in the body and is multiplying. That’s also when the patients are infectious, and so can spread the disease to other people. Again, that’s the typical disease manifestation. In the latent version, the bacteria become dormant – so the infection is there, but it’s not actually causing an illness. The patient won’t have any symptoms, they usually can’t infect other people, and it’s almost as if they don’t have the disease and it’s very hard to know whether they do or not.

We don’t see a lot of it in the UK, but it is one of the highest prevalence infectious diseases worldwide. Most cases are in Southeast Asia, the Western Pacific, and Africa. In the UK, we get small pockets of TB in certain geographical and socioeconomic groups.

Sputum smear positive for Mycobacterium tuberculosis (small red rods)
Acid-fast stain analysed by microscope (×1000).

How is TB usually diagnosed and what are the limitations to the method?

Diagnosis is quite interesting; there are a couple of different methods. There are the approaches that are followed in clinic nowadays, and then there are more novel approaches.

Generally, if a patient presents with an active disease, a chest X-ray is the best way to confirm the presence of a chest infection. However, a mucus sample can also be taken and analysed for the presence of the bacterium, M. tuberculosis. That process works well for active TB but doesn’t capture latent TB.

With latent TB, it’s a bit more complex. The Mantoux skin test involves the clinician injecting a small sample bacterial protein under the skin. If the patient reacts, it causes a little, raised-bump rash in that skin area. That positive response just indicates that that patient has antibodies to the bacterium in their system, which simply means that at some point they were exposed to the bacterium. However, it doesn’t differentiate between whether they currently have TB or ever had TB, but it means that there was some sort of exposure. The downside of that, for example, is that anybody who has been vaccinated against tuberculosis would test positive, even though they have never had the disease.

So, it’s not like you can do a very straightforward blood test and say that patient has TB now or has TB in its latent form. It is easier to find active TB, but a lot harder when it’s the latent form.

Importantly, and especially in the countries that are most affected, rural and lower income countries don’t always have the facilities to do an X-ray on each patient and blood tests for the presence of the bacterium – a hospital is needed.

What defines the search for biomarkers and what are the challenges of this search?

There are lots of issues around the diagnosis, which in part fed into the work that we did, although we never set out to find a new diagnostic for tuberculosis, our work was not powered to do that at all. There are a couple of suggested biomarkers for TB, although none of them are being comprehensively used. One example is lipoarabinomannan (LAM), which is a heat-stable glycolipid found in the outer cell wall of M. tuberculosis and can be found in a patient’s urine. So it’s a very non-invasive approach and is very specific but is very limited in terms of sensitivity, meaning it misses real cases of disease.

A good biomarker must be altered between the disease state and the healthy control. It has to be sensitive enough to detect all the true positives – all the people with the disease, without missing any of them. It also has to be specific to that disease. This is where biomarkers generally tend to fail – if they have good sensitivity, then they tend to have lower specificity and also detect other forms of infection or inflammation. Additionally, ideally, we want sample collection to be non-invasive, or minimally invasive, such as blood or urine samples. There is a lot of work being done now on finger-prick technology, where a drop of blood is collected at home in a little plastic device and then sent off to the lab for analysis. Another criterion for a good biomarker is that we have to have the capability to measure it accurately and reliably. Finally, an often overlooked but important criterion is that the analysis just has to be cost effective.

Transmission of tuberculosis
Transmission occurs through the air when a person with active TB disease of the lungs or throat coughs, sneezes, speaks, sings, or laughs, releasing tiny microscopic droplets containing the Mycobacterium tuberculosis bacteria.

How can the search for biomarkers be improved?

The challenge, especially in TB, is getting access to the right kind of data, getting sufficient number of participants or individuals who have the disease or have had the disease and then making sure we’ve run the correct assays on those individuals. That’s where the UK Biobank provides an amazing resource. This large-scale biomedical database and research resource contains in-depth genetic, lifestyle and health information from half a million UK participants. The limitations are that the participants tend to be mostly from a higher socio-economic background and of white ethnicity; as such, relatively few had ever been diagnosed with TB – reflective of the incidence rate in the UK. However, we were able to analyse plasma proteomic data (on nearly 3000 proteins) from the Biobank to investigate for host protein signatures associated with TB. In study, we had plasma profiles from 23 individuals with a prior diagnosis of TB, 19 individuals who were diagnosed after sample collection and 210 matched controls with no history of TB. Admittedly, our cohort’s quite small, but it was still large enough that we were able to run an analysis and find something useful. As we clarified before, we weren’t really looking to be able to walk away from that project saying, here’s a definitive biomarker that can be used. But, rather, it was to take a step back and say, if we looked at blood and we measured all the proteins without any sort of pre-assumption about what we might see as elevated or altered, can we answer the following questions.
• Can we identify some sort of signature that associates with patients who were diagnosed with TB?
• Can we see some sort of lasting effect of having TB in individuals’ blood?
• Can we detect something that indicates that there is perhaps a higher risk of those of individuals developing TB in the future?

So, once we had identified our cohort within this huge biobank, Natalia [Koziar, first author on our paper] pulled all the data together and did a differential analysis, which is straightforward. We analysed the expression levels of all the proteins and compared them between our case group and our control group. Then for those proteins that showed a significant difference, we gave a statistical measure to say are these proteins a different between the controls and our cases? And if they are, is it significant statistically or is it something that we could possibly see by chance? And for those that are statistically significant, we then looked at what these proteins are, and did it make sense that they’re related to the condition at all? We linked each protein (via its gene) back to the biological pathways that they’re involved in and we found a really strong representation of infection and inflammation, as well as other things, in these altered proteins. One of the aspects we saw was host pathogen recovery, with macrophage involvement, which is not surprising as TB can live in those cells. That gave us a good indication that we were on the right track.

In the cohort who had had TB prior to having their sample taken, we saw a lasting manifestation or “signal” of the disease in the blood. This is also interesting, although more is known in this field because we know that macrophages are involved and there’s going to be that effect that lasts a bit longer. However, we did look at over quite an extensive period of time, and we did do a sensitivity analysis where we only looked at the five-year period, which showed that there were a couple of proteins that were still significantly associated, which demonstrates the chronic nature of the condition. This isn’t your typical flu, for example, where you get it, you’re ill, you get better. Hopefully there’s no indication other than perhaps a few antibodies that are still around that you’ve ever had the flu. With TB, however, it seems to have a more lasting effect that can still be detected in the blood – one of the most important things for us is that can we actually see this in the proteome, making it an interesting tool to use not only for biomarker development but also mechanistic discovery in disease.

As mentioned above, one of our cohorts contained samples that were taken before a subsequent TB diagnosis. In that group, it’s extremely possible, that they may have had a latent form of the disease when the samples were taken, and we have no way of excluding that possibility. However, what we found was a handful of proteins that seems to be associated with that group but not with the other groups, which in itself is quite interesting and perhaps not something that you necessarily expect.

What future developments do you envisage in this field?

If I were to imagine the best future, it would be that a good, definitive biomarker (panel) has been developed for TB that can be rolled out in a lateral flow test, like we had for COVID-19. Once these are fully developed, these are cheap to manufacture. This has huge potential for low- and middle-income countries because people at high risk of infection could test themselves frequently (daily, even) for exposure and then immediately seek treatment, which is much more effective if the disease is caught early, rather than suffer the more severe consequences that some people do indeed suffer from. The realist in me knows that this is unlikely, as TB is, sadly, a very underfunded disease.

The interviewee

Professor Nophar Geifman M Med Sc, PhD, FHEA
Professor of Health and Biomedical Informatics, and Director of Informatics, the Veterinary Health Innovation Engine (vHIVE)

School of Health Sciences, Faculty of Health and Medical Sciences, University of Surrey, Guildford, UK

Email: n.geifman@surrey.ac.uk

 Bibliography
1. Schildkraut JA, Köhler N, Lange C, Duarte R, Gillespie SH. Advances in tuberculosis biomarkers: unravelling risk factors, active disease and treatment success. Breathe (Sheff). 2024;20(3):240003. PMID: 39660087  (https://publications.ersnet.org/content/breathe/20/3/240003).
2. Gadd DA, Hillary RF, Kuncheva Z, Mangelis T, Cheng Y et al. Blood protein assessment of leading incident diseases and mortality in the UK Biobank. Nat Aging. 2024;4(7):939-948. PMID: 38987645 (https://www.nature.com/articles/s43587-024-00655-7).
3. Koziar N, Whetton AD, Geifman N. A plasma-based protein signature association with all-cause mortality. PLoS One. 2025;20(11):e0336845. PMID: 41270070 (https://shorturl.at/Sp1T0).
4. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203–209. PMID: 30305743 (https://www.nature.com/articles/s41586-018-0579-z).

For further information, see:

Koziar N, Whetton AD, Geifman N. Plasma protein signatures altered before and after tuberculosis diagnosis in a population-based cohort. Clin Proteomics. 2026;23(1):28. PMID: 42251258 (https://doi.org/10.1186/s12014-026-09602-7).