Thread by @BobbyBentham

Out now in @nature! Introducing our new method T cell ExTRECT that uses DNA sequencing data to directly quantify T cell fraction and its application in cancer. A quick thread 🧵 nature.com/articles/s4158… (1/20)

First, this work came about quite unexpectedly. One day about 18 months ago we noticed something strange in our data. The tools we use to detect somatic copy number alterations in cancer were very frequently detecting alterations in a very small region of chr 14... (2/20)

Given how common these alterations were, we thought they could be important in cancer, particularly since we often saw amplifications in this region in cases with few other copy number changes. (3/20)

We were completely wrong! The alterations were entirely in a gene called TCRA that encodes the T cell receptor alpha chain, this is uniquely expressed on T cells and is used to recognise antigens and stimulate an immune response... (4/20)

The lung cancer cells we study almost certainly do not have T cell receptors, so why was this region being amplified so frequently? The answer could be found in simple T cell biology (5/20)

To recognise the vast array of potential antigens T cells undergo V(D)J recombination. This creates the diversity of T cell receptors we need but crucially also involves the loss of a section of DNA within the TCRA as T cell Receptor Excision Circles (TRECs). (6/20)

The alterations we were detecting were the loss of TRECs in T cells - not any changes related to cancer cells. Whether we detected amplifications or losses depended on the ratio of T cells in the cancer sample compared to the germline blood sample we compare it to! (7/20)

With the mystery solved we could have stopped there but instead we had the idea that we could use this signal to quantify the fraction of T cells in any whole exome sequencing (WES) sample. In short, we developed T cell Exome TREC Tool (T cell ExTRECT) to do exactly that (8/20)

T cell ExTRECT can be run on any WES sample which has coverage of TCRA (check the capture kit for compatibility), unlike tools such as ASCAT it does not require two WES samples (e.g. a cancer and matched normal) but is run on a single sample (for full details see paper!) (9/20)

T cell ExTRECT provides an insight into the immune microenvironment of WES samples. Often you may not have any orthogonal immune measures, e.g. RNA-seq. We show how it compares to orthogonal measures and use it to investigate data sets previously lacking immune data. (10/20)

We used the TRACERx100 lung cancer cohort that contains both WES and RNA-seq samples to do some basic validation. How does T cell ExTRECT compare to RNA-seq measures? Very well! (11/20)

We saw the strongest correlation with signatures related to T cells, and unlike RNA seq measures T cell ExTRECT outputs a fraction of your sample that come from T cells. (12/20)

We next investigated what determines the T cell fraction in either blood or tumour samples. In TRACERx100 cohort females had a higher blood T cell fraction than males and that there was a significant correlation between the T cell fraction we see in blood vs tumour. (13/20)

In a multi-sample pan-cancer cohort (nature.com/articles/s4158…) we identified different levels of immune heterogeneity by cancer type and subclonal SCNAs as potential drivers of this heterogeneity, with a loss event on 12q24.31-32 being identified as significant. (14/20)

This loss event was linked to SPPL3, a gene recently linked to the HLA class I immune response (sciencedirect.com/science/articl…) suggesting that subclonal loss on 12q24 could be a mechanism of immune evasion. (15/20)

We then investigated the potential of TCRA T cell fraction as a prognostic value for survival, identifying it as significant in the TRACERx100 LUAD cohort when considering the number of immune-cold samples, with more cold samples being associated with worse survival. (16/20)

This is consistent with what we have previously seen in the TRACERx100 data set (nature.com/articles/s4159…) but only using DNA, with no RNA or histopathological images of tumour infiltrating lymphocyte needed! (17/20)

Next we tested if we could predict response to immunotherapy. Using the CPI1000+ data set (cell.com/cell/fulltext/…) we showed that TCRA T cell fraction is just as predictive as CD8A expression and is still predictive in a lung cancer cohort that lacks any expression data.(18/20)

Finally, I would like to thank all my co-authors across the McGranahan and Swanton labs and funders, this was a big team effort! @kevlitchfield @tbkwatkins @LimEmilia @CarlosEcolEvol @crispinhiley @drmaiseb @SwantonLab @NickyMcGranahan @CR_UK @wellcometrust @rosetreesT (19/20)

Also: @MariamJHanjani @CRUKCOLcentre @CRUKLungCentre @uclcancer ! And not forgetting if you want to use T cell ExTRECT you can find it here github.com/McGranahanLab/… (20/20)