Mosaic Masterpieces


A few years ago in one of these pieces I exhorted you to visit the Villa Romana del Casale in the middle of Sicily. What? You still haven’t been?! Shame on you!! It’s even more of a good idea now than it was then.

As we explained in Molecular Mosaics, the villa houses the most extensive set of mosaics anywhere in the world and we were prompted to draw a parallel between their complexity and that of cancer by one of several pieces of research that took small bits of primary and secondary tumours, hit them with the power of DNA sequencing and showed that every region was different. It’s turned out that each individual tumour cell has sequence differences from even its nearest neighbours. So if you look at DNA sequence in cells across a tumour it would indeed be a mosaic.

Beautiful mosaics but …

Wonderful though DNA is, whilst it contains all the information required for life it doesn’t show you how to use it. It’s a bit like an architect’s plan for a new building: immensely detailed as to shape, where the doors and windows are etc., but with no information about the different bits — which is steel or glass, concrete or wood — yet alone how they’re put together to make the final edifice.

DNA is indeed a ‘blueprint’ as it’s so often described but to find out how cells behave we need to know which proteins are being produced at any time by the machinery of the cell as it translates the sequence of DNA bases (A, C, G & T) into the molecules that actually make things happen.

Detecting a killer

So it’s the proteins in a cell that control what the cell can do and indeed define what type of cell it is and it’s worth a moment to look at how modern technology can enable us to pick out individual cells from the 200 or so different types that make up a human being. In principle it’s fairly easy: add antibodies (proteins made by white blood cells that stick to other proteins — antigens) to your sample, chosen to recognise proteins known to be present on cell surfaces. If the antibody has an attached ‘flag’ — e.g., a fluorescent molecule (reviewed in I Know What I Like) you can pick out your cells with the aid of a microscope:


Killer T cells in action. The image shows a group of killer T cells (a sub-group of lymphocytes that destroy target cells identified as abnormal, including virus-infected cells and tumour cells) surrounding a cancer cell. DNA is stained blue. The T cells are green (via an antibody that attaches to a surface protein (CD8) that defines this type of cell) and red (small sacs within the cells that deliver ‘the kiss of death’ to kill the cancer cell).

The bigger picture

That’s OK for one type of cell but how can you look at a piece of tissue and ask: “What is the range of cell types present?”  That’s the question Leeat Keren, Michael Angelo and colleagues at Stanford University posed in the context of breast tumours and the stunning answers are to be found in their recent paper in the journal Cell. The methods they used are formidable but, armed with what we’ve just said about detecting one type of lymphocyte (killer T cells), we can break them down into simple steps from which emerge astonishingly complex patterns.

How’s it done?

They used a panel of antibodies that could pick out 36 different cell types and added them to slices of tumours. Rather than using fluorescent labels their antibodies carried ‘mass tags’. When a beam of charged particles is fired at the sample (so that it rasters across the target) cells are nebulized into single-cell droplets. The location of fragments released carrying the tags can be pinpointed and they are analysed by mass spectrometry to identify the antibody. From this comes an image of protein expression (i.e. cell type), each being given a false colour (or pseudo colour) to show up the distribution.

Cellular architecture of tumour samples. Each panel is a section of a breast tumour: each is from a different patient. Individual cell types picked out as described above. From Keren et al. 2018.

Not just a pretty picture

You might think this is just the tile pattern you’ve been after for your kitchen — not least because the most striking feature is that no two pieces are the same. The biology behind this variation is that across patients there were large differences in both the variety (type) and number of immune cells, notwithstanding the fact that all had the same type of tumour — triple-negative breast cancer.

However, even from this extraordinary variation it was possible to tease out some trends. Thus, for example, some tumours had high numbers of macrophages (25%) but low numbers of killer cells (1%), with B cells (11%), CD4+ (15%), CD8+ (19%) and regulatory T cells (1%) falling in between. An important point is that these trends relate to patient survival. The variation in patterns also hints at how different cells of the immune system are drawn to the site of a tumour and hence how they might cooperate in mounting a defence against cancer progression. Another notable finding was that proteins known to be important in controlling the immune response (e.g., PD-L1) can appear in different cell types. For example, in tumour cells themselves in some patients but in immune cells in others.

It’s amazing science, complete with pretty pictures — but it should help in categorizing patients and in the rational design of therapies.


Keren, L. et al. (2018). A Structured Tumor-Immune Microenvironment in Triple Negative Breast Cancer Revealed by Multiplexed Ion Beam Imaging. Cell 174, P1373-P1387.E19.