With the exception of a few hereditary cancers, there is currently no accurate method to predict whether someone is going to get cancer.
Dr Andrew Teschendorff from the Shanghai Institute of Nutrition and Health, in collaboration with Dr Chen Wu from the Chinese Academy of Medical Sciences, has created a computational method called CancerStemID that could help calculate a patient’s risk of cancer by analysing a vast amount of RNA data from precancerous cells.
Read the original research: doi.org/10.1158/0008-5472.can-22-0668
With the exception of a few hereditary cancers, there is currently no accurate method to predict whether someone is going to get cancer.
Dr Andrew Teschendorff from the Shanghai Institute of Nutrition and Health, in collaboration with Dr Chen Wu from the Chinese Academy of Medical Sciences, has created a computational method called CancerStemID that could help calculate a patient’s risk of cancer by analysing a vast amount of RNA data from precancerous cells.
Read the original research: doi.org/10.1158/0008-5472.can-22-0668
Hello and welcome to Research Pod! Thank you for listening and joining us today.
In this episode, we look at the work of Dr Andrew Teschendorff who has created a method for accurately identifying and measuring the different states of precancerous cells by analysing their RNA sequences using computational analysis.
With the exception of a few hereditary cancers, there is currently no accurate method to predict whether someone is going to get cancer. Creating a tool that would help clinicians identify patients that will develop malignant tumours would radically transform this medical field by making prevention or timely detection of cancer possible. Working towards this goal requires a broad understanding of the biology of cells and, more specifically, being able to identify which, and exactly how, cells are going to transform into cancer cells.
Dr Andrew Teschendorff from the Shanghai Institute of Nutrition and Health, working in collaboration with Dr Chen Wu from the Chinese Academy of Medical Sciences, has managed to create a computational method called CancerStemID that could help calculate a patient’s risk of cancer by analysing a vast amount of RNA data from precancerous cells.
All the cells in our bodies come from one single initial cell: the zygote or fertilised egg. The zygote gets divided into two new cells and the cell division keeps going until the whole organism is formed. However, to form a complex and fully functional organism, the cells have to follow a process called differentiation, which is the development of specialised features that enable cells to fulfil specific roles. The cells are then called ‘differentiated’, as opposed to the original cells they come from, which are called stem cells. The differentiated cells fall into categories based on their function and appearance, such as endothelial cells that line the human gastrointestinal tract or red blood cells that carry oxygen to our tissues.
Carcinogenesis, the process of healthy cells transforming into cancer cells, is characterised by an abnormal cell division. The model of carcinogenesis most supported by scientists suggests that cancer cells arise from dedifferentiated cells. That is, cells that have essentially lost their specialisation. This theory is supported by the fact that such dedifferentiated cells are almost always found in tissues surrounding malignant tumours and are usually considered as precancerous markers in pathology examinations and in everyday clinical practice.
The precancerous dedifferentiated cells have been shown to have special properties, such as the silencing or reduced expression of specific factors involved in the process of differentiation and, more specifically, in the first step of copying a gene’s DNA, called transcription. These tissue-specific transcription factors, or TFs, are essential for a smooth and normal cell differentiation process and, therefore, the silencing of these TFs can lead to abnormalities including the development of precancerous and cancer cells. It is not clear exactly how this starts, but it is most likely related to heritable changes of cell function that don’t involve changes of the actual cell’s DNA sequence, called epigenetic changes. With epigenetics the functional state of a cell can be fine-tuned by neighbouring cells through direct physical interaction or through signalling molecules secreted by one cell that can act on remote cells.
Single-cell RNA sequencing is a robust new technology that allows the identification, measurement and analysis of the RNA molecules of individual cells. This groundbreaking method can help with the discovery and identification of different cell types, as well as with understanding their various differentiation states including dedifferentiated cells. Since dedifferentiated cells are a marker of cancer risk or progression, by accurately identifying and quantifying them single-cell technology could help calculate the risk of cancer.
Several methods are currently being studied for measuring cell differentiation using single-cell RNA sequencing, but none of them have yet shown to be accurate enough to be used for cancer risk prediction. The lack of a specialised and accurate tool for predicting cancer has motivated Dr Andrew Teschendorff and his team to develop a novel computational method called CancerStemID that analyses RNA data to accurately identify dedifferentiated cells based on their TF activity patterns. A prerequisite for the effective use of this method is the identification of tissue-specific TFs and the target genes they activate, to allow more accurate and sensitive quantification of a TF’s activity. Importantly, this computational technique can be applied to any tissue-type and results in a tissue-specific regulatory network composed of the tissue-specific TFs and their target genes.
The researchers decided to study the precancerous cell states and cancer risk of oesophageal squamous cell cancer, a type of cancer stemming from the epithelial cells lining the inside of the gullet or oesophagus. The team started by obtaining tissue samples from 14 patients that had their surgery at the Linzhou Oesophageal Cancer Hospital in Henan, China for oesophageal squamous cell cancer. More specifically, tissue samples were taken from the tumours themselves, the area of tissue immediately around the tumours, while healthy tissues and blood samples were also collected from the respective patients. The type of the tissue samples was confirmed in the laboratory by the hospital pathologists resulting in a total of 49 samples, including: 8 of normal oesophageal tissue, 10 of inflammatory non-cancerous tissue, 17 of low and high grade precancerous tissues, and 14 invasive cancers. The patients’ medical records were also reviewed by the team for the collection of demographic and clinical data, such as age, gender and social history including the use of tobacco and alcohol.
All the collected tissue samples were next stained, cut and prepared for single-cell RNA sequencing. The sequencing data were then analysed and cells were categorised into clusters based on the expression of cell-type specific relevant genes called markers. Focusing on the epithelial cells, the authors then applied the CancerStemID algorithm to build an oesophageal-epithelial specific regulatory network. From this network, the dedifferentiation state of each epithelial cell was calculated by computing the activity of oesophageal-specific TFs. Finally, their dedifferentiation states were compared across disease stages, including normal, precancerous and cancer stages.
Analysing the data revealed that the TF activity levels of the epithelial cells decreased as cells progress towards cancer. In other words, the more the cell progresses towards malignancy, the higher the number of tissue-specific TFs that display reduced activity in that cell – a result consistent with the team’s original hypothesis. Importantly, just focusing on the cells from the normal and precancerous states, the authors were able to identify those precancerous cells that are more likely to turn cancerous.
Single-cell data is notoriously noisy which can result in artefactual findings. Thus, the authors sought to replicate the above findings in an independent human cohort of oesophageal cancer and normal tissue, as well as in a mouse model that mimics oesophageal cancer development in humans. Strikingly, in both datasets, similar findings were obtained, further demonstrating that low TF activity is a robust marker of dedifferentiation that can help identify precancerous cells that are most likely to turn cancerous.
The team therefore found that the less differentiated the cell, the closer to becoming malignant it was. However, this doesn’t mean that the precancerous cells are gradually becoming healthy stem cells. In fact, the precancerous cells were found to be even less differentiated than the actual healthy stem cells, which suggests that during carcinogenesis a form of reprogramming of the actual stem cells seems to be taking place. Data generated by the team further indicates that this reprogramming, characterised by a gradual and irreversible inactivation of the tissue-specific TFs, is likely to be epigenetic in origin, and mediated by an abnormal covalent modification of DNA known as DNA methylation. It is thought that these DNA methylation changes may be caused by mild increases in cell-proliferation and exposures to cancer risk factors like smoking, inflammation and obesity.
The authors believe that this pattern of an epigenetically induced gradual reprogramming of tissue-specific TFs is a cancer-hallmark, since the reduced TF activity is also seen in many other cancer-types, including lung, colon, kidney, bladder and stomach cancer.
The findings confirm that the CancerStemID algorithm developed by the researchers is a reliable computational method that analyses the diversity of precancerous cells by identifying a subgroup of dedifferentiated cells with a high number of silenced TFs, demonstrating that their gradual silencing correlates with cancer progression. This groundbreaking technique could become the base for the development of a much-needed early cancer detection and risk prediction system, as well as a very useful research tool for assessing the effectiveness of cancer prevention trials.
That’s all for this episode – thanks for listening. Links to the original research can be found in the shownotes for this episode. And, as always, stay subscribed to Research Pod for more of the latest science!
See you again soon.