During carcinoma progression, mesenchymal stromal cells become recruited to tumors and contribute to the pool of cancer-associated fibroblasts (CAFs). The heterogeneity of CAF populations, changing during disease progression, has been realized1. However, CAF sub-populations have diverse and incompletely understood effects on disease progression and resistance to therapy. Here, by analyzing public databases of human single-cell RNAseq (scRNA-seq) data, we have identified a CAF progenitor population marked by expression of genes APOD, DCN, LUM, typically accompanied by expression of CFD, CXCL14, PTGDS, MGP, SERPINF1, and DPT. We show that these cells are prominent in the following two settings:
- Naturally occurring in cancer-free individuals as (a) adipose stromal cells (ASCs) identified in ref.2 as adipocyte progenitors enriched in the stromal vascular fraction (SVF) of adipose tissue, as well as (b) fibro-adipogenic progenitors (FAPs) in skeletal muscle3. We refer to the corresponding gene co-expression signature as “the ASC/FAP signature.”
- Enriched in the tumor microenvironment of invasive and chemoresistant carcinomas of various types.
Fig. 1 demonstrates the ASC/FAP signature from the remarkably similar lists of the top-ranked differentially expressed genes (DEGs) in clusters identified in ref.2 (Fig. 1a) and ref.3 (Fig. 1b). Each of the nine highlighted genes appears in both lists (P = 10-26 by hypergeometric test). The presence of the same signature in both ASCs and FAPs is consistent with the finding that adipose tissue is a source of FAP-like cells serving as a source of fibroblasts recruited to skeletal muscle undergoing remodeling4.
To demonstrate that the ASC/FAP population exists in abundance in each of the 25 adipose tissue samples presented in ref.2, we analyzed each of those samples derived from 14 cancer-free individuals. We used the attractor algorithm5 (Methods), designed to converge to a ranked list of genes identifying the core of co-expression characterizing cell populations. We independently generated the lists of top-ranked genes for each sample. Supplementary Table 1 shows that all genes mentioned above are consistently top ranked, while Supplementary Fig. 1 demonstrates the abundance of the ASC/FAP population in the SVF.
Fig. 2a shows lists of top-ranked DEGs from clusters identified in bladder6, ovarian7 and breast8 cancer. Fig. 2b shows the twelve top DEGs of a population found to have a three-fold enrichment in chemo-resistant samples of pancreatic cancer, referred to as constituting a “chemo-resistance signature” in commentaries to ref.9. The strong enrichment of the ASC/FAP cell population in multiple cancer types is consistent with the recruitment of this particular progenitor population in aggressive and chemo-resistant cancers. Consistently, mouse lineage tracing and transplantation studies have indicated that ASCs can be recruited by carcinomas and promote cancer progression10,11.
The recruited ASC/FAP cells are typically misrepresented as inflammatory CAFs (“iCAFs”) because they are often included in nonhomogeneous computationally derived clusters that also contain such cells. The iCAFs have been defined as fibroblasts expressing IL6 and additional inflammatory mediators12, and the expression of IL6 has remained a requirement in a Consensus Statement of experts13. IL6 expression also can be induced in a subset of cells in such clusters. However, the ASC/FAP population in its original non-inflammatory status should not be confused with IL6+ iCAFs. For example, in Fig. 2c, taken from ref.14, cluster 1 is marked by genes C7, CFD and PTGDS, all three of which are among the twelve in the chemo-resistance ASC/FAP signature of Fig. 2b. The distinct cluster 2 is marked by IL6, HAS1 and CCL2, all among the iCAF signature genes defined previously15, hence identifying it as the true iCAF population. Consistently, it was recently suggested16 that there are “two separate populations of iCAFs: one IL6 positive and another IL6 negative.” In fact, the IL6-negative cluster largely contains the APOD+ ASC/FAP population and should not be characterized as iCAF.
Furthermore, single-cell analysis has revealed17 that, in aggressive cancers, cells with the ASC/FAP signature convert to a particular type of CAFs expressing COL11A1, THBS2 and INHBA. This CAF signature was first discovered in ref.18 and includes additional genes such as POSTN, COL10A1 and MMP11. Fig. 2d shows an example consistent with the transition, taken from ref.19, in which cluster C0 expresses APOD, DCN, and LUM, while cluster C3 adjacent to it expresses COL11A1, THBS2, and INHBA. Furthermore, the presence of the COL11A1+ cluster 0, adjacent to C7+CFD+PTGDS+ cluster 1 in Fig. 2c is again consistent with the transition, as is the presence of gene POSTN together with APOD, CFD and CXCL14 in the same “poor prognosis” cluster (CAF_0)20 in gastric cancer. COL11A1 is also identified in ref.16 as the collagen marker most strongly associated with poor prognosis.
In summary, this Brief Communication draws attention to the APOD+DCN+LUM+ gene signature as representing an important unrecognized population in cancer derived from ASCs/FAPs recruited by carcinomas. Their differentiation into COL11A1+ CAFs, accompanying transition to metastasis, may underlie a mechanism that accounts for the role of adipose tissue in cancer aggressiveness21. Further research is needed aimed at developing appropriate therapeutics targeting the underlying mechanisms.