Distinct populations of crypt-associated fibroblasts act as signaling hubs to control colon homeostasis

Despite recent progress in recognizing the importance of mesenchymal cells for the homeostasis of the intestinal system, the current picture of how these cells communicate with the associated epithelial layer remains unclear. To describe the relevant cell populations in an unbiased manner, we carried out a single-cell transcriptome analysis of the adult murine colon, producing a high-quality atlas of matched colonic epithelium and mesenchyme. We identify two crypt-associated colonic fibroblast populations that are demarcated by different strengths of platelet-derived growth factor receptor A (Pdgfra) expression. Crypt-bottom fibroblasts (CBFs), close to the intestinal stem cells, express low levels of Pdgfra and secrete canonical Wnt ligands, Wnt potentiators, and bone morphogenetic protein (Bmp) inhibitors. Crypt-top fibroblasts (CTFs) exhibit high Pdgfra levels and secrete noncanonical Wnts and Bmp ligands. While the Pdgfralow cells maintain intestinal stem cell proliferation, the Pdgfrahigh cells induce differentiation of the epithelial cells. Our findings enhance our understanding of the crosstalk between various colonic epithelial cells and their associated mesenchymal signaling hubs along the crypt axis—placing differential Pdgfra expression levels in the spotlight of intestinal fibroblast identity.


The introduction mentions that most other studies use specific cell isolation protocols. It is not clear what the authors refer to, given that they themselves of course also use an enzymatic digestion and cell isolation protocol (i.e. FACS sort for EPCAM+ and EPCAM-cells). I think I get what they mean (theirs is 'just' a sort for epithelial versus non-epithelial), rather than a specific enrichment for a subpopulation, but perhaps they can clarify this a bit.
We have adjusted the text to clarify this point. While indeed we use protocols specifically aimed at isolating both epithelial and non-epithelial cells in our study, we sort epithelial (EPCAM+, CD45-) and non-epithelial (EPCAM-, CD45-) cells originating from both isolations (i.e. epithelial and non-epithelial cells isolated from the epithelial fraction after, and remaining epithelial and non-epithelial cells isolated from the fraction after the majority of epithelial cells was removed). Therefore, we make sure not to lose e.g. non-epithelial cells that closely interact with epithelial cells or vice versa, that were hitherto potentially lost when looking for example only at epithelial cells. Such a "pooling approach" is one of the benefits of our study, since it was not done so far. (The respective sections were changed in the manuscript, both in the introduction and results part, to describe it more thoroughly) 2. In Figure 1B, does this refer to the input prior to sorting or is this purity after sorting? Figure 1B refers to the gating strategy chosen prior to sorting (information clarifying it was added to manuscript).

Related to the sorting: Can the authors comment on and/or quantify the ratio of these different cell populations (CBF1/2 vs CTF vs epithelial cells)?
The following numbers are based on the overall sorting numbers and ratios observed in the single-cell data and refers to the isolated cells: 91.2% are epithelial cells and 8.8% are mesenchymal cells. This information has been added to the manuscript. However, the majority of the lamina propria smooth muscle cells, are not dissociated by our enzymatic treatment and thus underrepresented.
CTFs make up 1.6% and CBFs1 and CBFs2 make up 2.4% and 1.6% of the isolated cells, respectively.

Figure 2 and throughout: Can the authors speculate on the difference between the CBF1 and CBF2 population? Are there any discriminatory marks that separate these two otherwise functionally overlapping populations? Based on 2G this seems a logical step to include in this manuscript and something the authors should be able to do with the tools and analysis pipeline at hand, or is this something the authors plan to follow up on later? Alternatively, could it be some sort of experimental artifact? A list of differentially expressed genes in these populations might be informative.
This is a very important comment. The paragraph concerning the difference between CBFs1 and CBFs2 has been added to the manuscript.

From the legend of Figure 3A-C it is not clear that we are looking at cryosections, but I imagine that this is what is depicted, based on the methods?
The Figure 3A-C depict cryosections of fixed frozen tissue. The legend has been updated accordingly.

Related to 5: Has the GFP signal been corrected for DAPI? The authors describe in the methods how they performed the GFP quantification, but this seems to related to masking of nuclei to include. Ideally, the GFP signal is corrected for the DAPI intensity to make up for depth/sectioning/processing/imaging differences at different locations on the slide. If it is not done or cannot be done, this should be noted in the legends and/or methods.
While we did not correct the GFP signal for DAPI, as described in the methods, we excluded all the nuclei with only partially sectioned/dim nuclei (low quality) and made sure that only nuclei that were fully capture in the Z-stack were considered for the final analysis. To clarify this, we added more detail to methods section.

In this work, Brugger et al. perform single cell RNAseq experiments to characterize the mesenchymal cell types constituting the colonic crypt niche. They uncover three distinct Pdgfra+ fibroblast populations, expressing distinct antagonistic morphogens. More specifically, crypt bottom fibroblasts, which express canonical Wnt ligands, Rspo3 and Bmp inhibitors, and crypt top fibroblasts, expressing non-canonical Wnt ligands and Bmp ligands. The study is well performed and will be an important resource for biologists interested in stem cell biology. The following points should be addressed in a revision: -The supplementary figures nicely show the expression of ligands of major signaling pathways among the three fibroblast populations, however the paper would be strengthened by a more comprehensive analysis of ligand-receptor interactions, most importantly between the fibroblasts and epithelial compartments. Specifically, what are the main elevated epithelial ligands expressed in tip/crypt enterocytes with matching elevated tip/crypt fibroblast receptors and vice versa? There are several examples, e.g. Pdgfra in the crypt tip enterocytes and Pdgfra in the CTF, Rspo3 in CBF and matching epithelial receptor Lgr5 at the crypt bottom. Such analysis can be performed by tools such as https://www.cellphonedb.org/ or NicheNet (https://www.nature.com/articles/s41592-019-0667-5). Similarly, are there interactions between the fibroblast populations and pericytes/endothelial clusters?
Following this valuable suggestion, we ran the CellphoneDB algorithm on our data. We were able to confirm known interactions, such as Wnt2 from CBFs1 interacting to Fzd3 in epithelial cells or Wnt5a from CTFs signaling to Ror1 and Ror2 in epithelial cells. In addition, we discovered potentially interesting new interactions, such as Igf1 from CBFs2 signaling to Igf1r in epithelial cells or Fgf2 from CBFs1 and CBFs2 signaling to Fgfr2 and Cd44 in epithelial cells. We have provided this new data as a new S3 Fig. However, given the current (lack of) understanding of colon crypt zonation we cannot extract the information to answer what are possible ligand/receptor interactions between, for example, crypt tip enterocytes and fibroblasts. We cannot simply assume that the same markers that were found in villus zonation of enterocytes in the small intestine are accordingly distributed in colonic enterocytes in the same way. Once data on colon crypt zonation become available our dataset could be interrogated to investigate this interesting question.

-Can the authors comment about the potential spatial differences between CBF1 and CBF2? Is one population closer to the epithelium? Do any of these populations overlap trophocytes (https://www.nature.com/articles/s41556-020-0567-z)?
In contrast to the situation in the small intestine, in the colon all Pdgfra low cells -CBFs1 and CBFs2 -were located in close proximity to the crypt/stem cell compartment: 1-3 cell diameters (Fig 3A). Looking in more detail into the different spatial localization of CBFs1 and CBFs2 is hampered by the fact that most marker genes/differentially expressed genes do not show clear black/white patterns of expression within the colonic stromal cell populations. As a result, the selection and analysis of potential antibody/smFISH staining proves difficult and is uncertain to yield further insight into the subject. Based on a comparison of the scRNAseq data, CBFs2 show stronger overlap to the trophocytes described by McCarthy and colleagues for the small intestine. For example, high expression of the marker gene Cd81 and the Bmp antagonist Gremlin1 are shared characteristics. These observations have been added to the manuscript when discussing the difference between CTFs1 and CTFs2, and corresponding data added to S3 Fig.   -

Do CTFs express Lgr5, as has been shown for villus tip telocytes (VTTs, ref 20)? The authors should comment on the overlap and differences in gene expression between CTFs and VTTs, which express many of the discussed CTF markers, such as the non-canonical Wnt5a, Bmp ligands and F3.
Indeed, a low number of cells in the CTFs cluster express Lgr5.
Moreover, many of the factors that are expressed in VTTs are shared with CTFs, such as Wnt5a and Fn3, and high expression of Bmp2 and Bmp4. However, there are differences; several interesting genes expressed in VTTs have a different pattern: the ligand Egf is not expressed in CTFs and microfibrillar genes, such as Mfap5, Emilin2 and Fbn1, are more highly expressed in CBFs1 and CBFs2 in the colon.

-It would be informative to add immunofluorescence for PDGFRA protein, which labels the entire cell bodies. Telocytes have a unique morphology with elongated processes that 'wrap around' the crypts, as well as tiny 'telopode' protrusions into the lamina propia, such staining for PDGFRA would reveal such morphologies in the colon.
The antibody staining for PDGFRA protein can be found in Figure 3A. However it does not label the entire cell bodies, probably corresponding to the distribution of the protein. More information about the morphology of Pdgfra positive cells can be learned from pictures of Pdgfra-CreERT2; LSL-tdTomato mice -tdTomato labels the entire cell. This data has been added to S4 Fig.   -

The expression of both the canonical Wnt ligands and antagonists is intriguing, do the same CBF cells express both ligands and antagonists, or are they expressed in different CBF subsets?
This question is not so easy to answer due to general issues with dropout rate in single-cell RNA sequencing and the resulting uncertainty whether low expressed genes, such as Wnt ligands are possibly expressed in all cells of the clusters and simply not captured or whether they are really only expressed in a subset of the cells. However, we queried the co-expression the canonical Wnt ligand Wnt2b and Wnt signaling inhibitors (Sfrp1, Sfrp2, Dkk3) and found rare cells that co-express both (yellow cells in graph).