The aim of this post is to improve the accessibility of the talk given at the Wellcome Single Cell Conference 2022 by providing with the notes used for the oral presentation.
An improved iteration of this talk was given at the Cancer Institute 2022 conference as part of the Recent Discoveries series.
The work presented in this talk eventually developed into that shown in Cardoso Rodriguez & Qin et al. 2023.
A PDF copy of the slides can be downloaded from here
Introduction (1/10 & 2/10)
Intrinsic and Extrinsic Cues as Cell Fate Regulators
It is understood that both intrinsic and extrinsic cues are crucial in determining cell fate.
- As Intrinsic cues we might have mutations, a cell’s epigenetic state, and transcriptional networks.
- As for extrinsic cues, we might have mechanobiological signals, metabolites or growth factors.
Now, in a cancerous cell we know that intrinsic cues such as oncogenic mutations are key when controlling cell fate, but so are the extrinsic cues coming from the tumour microenvironment.
Here at the tape lab and in this project in particular are interested in understanding how do oncogenes and the TME regulate cell fate in the context of Colorectal Cancer.
Single-cell Technologies to Study CRC Organoids
CRC is a very common cancer, claiming 800 thousand worldwide deaths per year Much like other cancers it presents as a collection of malignant and non-transformed cells, resulting in a complex heterocellular setting where colonic epithelia closely interacts with the surrounding tumour microenvironment, consisting among others of stromal fibroblasts and myeloid macrophages. In fact, TME is key factor in CRC, with increased stromal and myeloid components correlating with worse prognosis.
In order to study this disease we model it using organoids, self assembling 3D structures that we can grow in vitro in heterocellular settings that can mimic the TME.
Such a heterogenous system requires of single-cell technologies to be properly characterised. Mass cytometry is easily scalable and ideal to study intra-cellular communications via PTMs. On the other hand, scRNA-seq can be used to describe inter-cellular communication by looking at the expression of ligands and receptors between cells.
Experimental setup and methods (3/10)
Multivariate scRNA-seq of CRC-TME Organoids
To explore how intrinsic oncogenes and extrinsic stromal cues regulate colonic epithelial fate we performed a multivariate scRNA-seq analysis of organoids.
In it we had organoids ranging from WT to those progressively harbouring mutational hallmarks of CRC, either in a monoculture or co-cultured with colonic fibroblasts and/or macrophages. Furthermore we also included a WT monoculture grown with exogenous WNT3A, EGF, Nogging and R-spondin1 as a niche factor control.
Analysed using scRNA-seq we were able to identify the three cell types in the system. We then focused on the epithelial compartment, with the main steps of the analysis shown here; including velocity, differential abundance, and cell-cell communication analysis.
Results (4/10 - 9/10)
scRNA-seq Captures Colonic Epithelia Differentiation
Now with a subset of just of the epithelial cells from the system (around 30k cells across 17 conditions), we were able to capture colonic epithelia differentiation. We identified clusters associated with the main colonic epithelia cell types;
including a stem, secretory and absorptive populations.
The overall dynamics of the system were also recovered, with SCENT’s CCAT entropy based metric indicating greater pluripotency estimates for the colonic stem cell clusters, as well as RNA velocity stream embeddings going towards the differentiated states. Now, it is important to note that instead of a monolithic colonic stem cells population, we identified a heterogenous compartment with multiple states
Oncogenes and Fibroblasts Differentially Regulate CSC
We then investigated when these multiple colonic stem cell states were being regulated using differential abundance analysis. Comparing our control WT monocultures with CRC organoids we saw a significant enrichment of cellular neighbourhoods associated with the proliferative Colonic stem cell population, shown in this PHATE embedding in red. Note also how the more differentiated late enterocyte cells are on the other hand being lost.
On the other hand, we noticed how the addition of fibroblasts in the co-culture induced a clear enrichment of the revival CSC in the WT organoids.
These 2 distinct and differentially regulated CSC populations where characterised using Differential Expression.
As shown in this curated heatmap, proCSC are characterised by being particularly hyperproliferative and expressing Lgr5 and Birc5. RevCSC are quite distinct from the other colonic stem cells, being less proliferative and expressing Clusterin; their unique transcriptional signature matching that of the revival stem cell population in the intestine.
Interestingly, while the CSC and proCSC clusters consist mostly of AK and AKP cells, the revCSC cluster is dominated by WT cells.
Oncogenic Entrapment of the Colon Epithelia
That is because it appears that the oncogenic mutations are trapping the colonic epithelia in an oncogenic state.
First we have a summary of cluster abundance changes across all conditions. It clearly shows how the first 2 colonic stem cell clusters are being upregulated by the oncogenes, and how the revCSC are being induced in WT and ApcKD fibroblast co-cultures.
However, there’s a clear loss of fibroblast regulation in AK and AKP organoids. Furthermore it seems like WENR is inducing all CSC states in wt organoids.
We also used RNA velocity lengths to compare differentiation rates across organoids. We compared side-by-side our WT controls, WT+fibroblast cocultures, AK organoids and the WENR enriched condition,
As you can see here the AK organoids showed the lowest velocity lengths (which was consistent with AKP organoids too).
To explore the mechanisms behind this oncogenic entrapment we used cell-cell communication analysis of the fibroblast co-cultures across the different genotypes. With these plots we can compare the quantity of incoming and outgoing signals as detected by CellChat, with the arrows showing aggregate interaction strengths from the fibroblasts to the epithelia clusters.
As we had expected, the fibroblasts appear to be predominantly characterised by outgoing signalling consistently across all genotypes. However, while the WT epithelia shows a high score for incoming signals it quickly dropped following the accumulation of oncogenic mutations.
These results suggest that oncogenic mutations erode fibroblast to epithelia communication, with the CRC organoids acting as “bad listeners”.
Exogenous Ligands Model CSC Regulation
(Credit to Dr Xiao Qin)
Now if you remember we had seen that WENR promoted all CSC states, so to explore how exogenous ligands model CSC regulation we used mass cytometry to analyse organoid monocultures across 60 different conditions with different genotypes and ligands.
In this condition-based PHATE embedding we can see how:
- We can recapitulate the revCSC cells characterised by Clusterin expression. These cells seem to be associated with exogenous wnt3a and low MAPK and b-cat signalling.
- We can also see the Birc5 positive proCSC , which were characterised by high MAPK and b-cat signalling.
Now using a Scaffold map centred around key genotypes we can see how oncogenes and the exogenous ligands navigate a shared regulatory landscape. If for example, we take a WT organoid and add exogenous WNT3A we will be promoting the revCSC. On the other hand, if we take an organoid with just oncogenic Kras (not sufficient on its own to trap the epithelia in a proCSC-like state) and add r-spondin we shift cell fate towards the proCSC.
Summary (10/10)
We have identified 2 distinct and differentially regulated CSC states: Intrinsic oncogenic cues seem to induce a hyper proliferative CSC population Whereas extrinsic stromal cues induce the revCSCs
Oncogenic mutations are trapping the colonic epithelia in the hyperproliferative state. This seems to be due to the erosion of stromal regulation in CRC organoids. And is also suggested by the lowered differentiation rates in them
Finally, it appears that both intrinsic and extrinsic cues navigate a shared regulatory landscape Regulated by EGF and WNT signalling, up to the point until APC loss and oncogenic KRAS trap the organoids, turning them refractory towards the external cues