Microenvironmentally-driven Plasticity of CD44 isoform expression determines Engraftment and Stem-like Phenotype in CRC cell lines

Abstract

Theranostic biomarkers for putative cancer stem-like cells (CSC) in colorectal cancer (CRC) are of particular interest in translational research to develop patient-individualized treatment strategies. Surface proteins still under debate are CD44 and CD133. The structural and functional diversity of these antigens, as well as their plasticity, has only just begun to be understood. Our study aimed to gain novel insight into the plasticity of CD133/CD44, thereby proving the hypothesis of marker-associated tumorigenic and non-tumorigenic phenotypes to be environmentally driven. Methods: CD133/CD44 profiles of 20 CRC cell lines were monitored; three models with distinct surface patterns in vitro were systematically examined. CD133/CD44 subpopulations were isolated by FACS and analyzed upon in vitro growth and/or in limiting dilution engraftment studies. The experimental setup included biomarker analyses on the protein (flow cytometry, Western blotting, immunofluorescence) and mRNA levels (RT-/qPCR) as well as CD44 gene sequencing. Results: In general, we found that (i) the in vitro CD133/CD44 pattern never determined engraftment and (ii) the CD133/CD44 population distributions harmonized under in vivo conditions. The LS1034 cell line appeared as a unique model due to its de novo in vivo presentation of CD44. CD44v8-10 was identified as main transcript, which was stronger expressed in primary human CRC than in normal colon tissues. Biomarker pattern of LS1034 cells in vivo reflected secondary engraftment: the tumorigenic potential was highest in CD133⁺/CD44⁺, intermediate in CD133⁺/CD44^- and entirely lost in CD133^-/CD44^- subfractions. Both CD44⁺ and CD44^- LS1034 cells gave rise to tumorigenic and non-tumorigenic progeny and were convertible - but only as long as they expressed CD133 in vivo. The highly tumorigenic CD133⁺/CD44(v8-10)⁺ LS1034 cells were localized in well-oxygenated perivascular but not hypoxic regions. From a multitude of putative modulators, only the direct interaction with stromal fibroblasts triggered an essential, in vivo-like enhancement of CD44v8-10 presentation in vitro. Conclusion: Environmental conditions modulate CD133/CD44 phenotypes and tumorigenic potential of CRC subpopulations. The identification of fibroblasts as drivers of cancer-specific CD44 expression profile and plasticity sheds light on the limitation of per se dynamic surface antigens as biomarkers. It can also explain the location of putative CD133/CD44-positive CRC CSC in the perivascular niche, which is likely to comprise cancer-associated fibroblasts. The LS1034 in vitro/in vivo model is a valuable tool to unravel the mechanism of stromal-induced CD44v8-10 expression and identify further therapeutically relevant, mutual interrelations between microenvironment and tumorigenic phenotype.

Keywords: CD44/CD44v8-10; colorectal cancer; fibroblasts; limiting dilution engraftment; microenvironment.

Figure 1

CRC cell lines show heterogeneous CD133/CD44 surface expression pattern not reflecting engraftment. (A) Representative flow cytometric dot blot diagrams and histograms showing CD133-PE and CD44-APC surface pattern in exponentially grown SW480, SW620, and LS1034 cells kept under identical 2-D in vitro conditions. Antibodies and staining details are given in Table S1; the CD133 (AC133) fluorescence signal was enhanced by a two-step FASER series as previously described ,; (B) Engraftment rates of SW480, SW620, and LS1034 cells in NMRI nu/nu mice upon s.c. injection of defined single cell suspensions derived from exponentially grown monolayer cultures applied in limiting dilution approaches with 10 - 10,000 cells per mouse and injection site, respectively, n.d. - not determined; (C) Representative areas of H&E-stained 10 µm frozen sections from SW480, SW620, and LS1034 xenografts.

Figure 2

CD133/CD44 surface expression in SW620 monolayer cells does not determine engraftment in vivo; CD133/CD44 pattern and population distributions in xenografts harmonize independent of the injected subpopulation. (A) Representative flow cytometric dot blot diagrams of CD133-PE and CD44-APC surface pattern in SW620 cells before (cf. Figure 1) and after sorting for subcutaneous injection at limiting dilution (sort layout for subpopulations 1-4); bottom dot blots document representative CD133/CD44 pattern in cell suspensions derived from xenografts originated from the respective FACSorted subpopulation; (B) Engraftment rates of SW620 cell populations separated by FACS according to their in vitro CD133/CD44 surface pattern; control = stained cells processed (“run-through-sorter”) according to the subpopulations; (C) Tumor control as function of time after subcutaneous injections of 10 SW620 cells with different CD133/CD44 surface pattern (from B); (D) Normalized volume growth kinetics of xenografts derived from 100 SW620 cells with different CD133/CD44 surface pattern; (E) Distribution and proportion, respectively, of CD133⁺, CD44⁺, and CD133⁺/CD44⁺ SW620 cells (+SD) in xenografts originated from the different FACSorted in vitro SW620 subpopulations.

Figure 3

CD133 pattern in LS1034 cells in vitro does not determine engraftment and in vivo behavior; CD133/CD44 cell surface profiles and distributions in LS1034 xenografts are consistent, with CD44 been newly expressed on a CD133⁺ subpopulation. (A) Representative flow cytometric dot blot diagrams of CD133-PE and CD44-APC surface pattern in LS1034 cells before sorting (cf. Figure 1) and CD133 histograms of CD133^- (1) and CD133⁺ (2) subfractions after FACSorting before s.c. injection; bottom dot blots document representative CD133/CD44 pattern in cell suspensions prepared from xenografts originated from the respective injected subfractions; (B) Engraftment rates of LS1034 cell subfractions separated by FACS according to their in vitro CD133 surface expression; control = “run-through-sorter” (cf. Figure 2). (C) Tumor control as function of time after s.c. injection of 100 CD133^- or CD133⁺ LS1034 cells (according to B). (D) Normalized volume growth kinetics of xenografts derived from 500 CD133^- or CD133⁺ LS1034 cells; (E) Distribution and proportion, respectively, of CD133⁺, CD44⁺, and CD133/CD44 single or double negative and positive LS1034 cells (+SD) in xenografts originated from either CD133⁺ or CD133^- FACSorted in vitro cells.

Figure 4

CD133/CD44 pattern in LS1034 primary xenografts correlates with secondary engraftment. (A) Representative flow cytometric dot blot diagrams of CD133-PE and CD44-APC surface pattern in suspensions of LS1034 xenograft cells before (cf. Figure 3) and after FACSorting and following secondary engraftment; (B) Engraftment rates of LS1034 xenograft cell populations separated by FACS according to their in vivo CD133/CD44 surface expression; control = “run-through-sorter” (cf. Figure 2). (C) Tumor control as function of time after s.c. injection of 2,500 FACSorted LS1034 cells originated from xenografts; *** p<0.001; (D) Normalized volume growth kinetics of secondary xenografts derived from 10,000 CD133⁺/CD44^- or CD133⁺/CD44⁺ LS1034 in vivo cells; (E) Distribution and proportion, respectively, of CD133⁺, CD44⁺ and CD133/CD44 single or double negative and positive LS1034 cells (+SD) in secondary xenografts originated from CD133⁺/CD44^- or CD133⁺/CD44⁺ cell populations isolated from primary LS1034 xenografts.

Figure 5

CD44⁺ tumor cells are located in well-oxygenized but not in a proposed hypoxic cancer stem cell niche in LS1034 xenografts (see also Figure S4 A, B). Median frozen section (10 µm) of an LS1034 xenografts co-stained for CD44, CD31 (endothelial cells), pimonidazole accumulation (hypoxia), and DAPI (nuclei) and imaged with a magnification of 200x; whole tumor section (stitched from >1,000 single images - top) and a selected region at higher magnification are displayed as four-channel overlays, while single channel images of the respective region are documented on the bottom.

Figure 6

CD44 enhancement in LS1034 xenografts is mainly due to up-regulation of CD44v8-10 expression. (A) Upper panel: section of a representative Western blot according to Figure S5A (standard handling and illumination times, reducing conditions) to visualize the CD44 pattern in LS1034, SW620, and SW480 cells (40 µg protein loaded per lane). Lower panel: Western blot performed with similar protocol showing human-specific CD44 in protein lysates of four different LS1034 xenografts, three individual LS1034 monolayer culture samples, and lysates from cell mixtures containing 75%, 90%, or 100% mouse fibroblasts (= 25%, 10%, and 0% LS1034 cells); 50 µg protein loaded per lane (PC - positive control = 25 µg of an HT29 cell culture lysate); β-actin or α-tubulin were detected as protein loading controls (non-species specific), while HLA-B (MHC-I+HLA-B directed antibody) was displayed as human-specific control; (B) Representative flow cytometric histogram of CD44 labeling in permeabilized LS1034 cells (cf. Figure 1); SW480 cells served as positive control; (C) Scheme of human CD44 gene structure, specific primer design, and representative RT-PCR analysis of (D) total, (E) isoform-specific, and (F) v9 exon-specific CD44 mRNA expression in LS1034 cells in vitro and in xenografts; human β-actin (ACTB) and B2M mRNA levels were detected as reference. Note: Xenografts used for protein (-P) and RNA (-R) extraction were not identical as indicated by the respective name affix.

Figure 7

CD44v8-10 upregulation in LS1034 xenografts associates with transcriptional EMT markers, can be triggered in vitro by interaction with stromal cells, and reflects the high expression level in primary colon adenocarcinoma tissues. (A) Fold of gene expression of CD44, CD133 as well as various EMT specific transcription factors and biomarkers in LS1034 xenografts versus cultured cells measured by q-PCR. Data normalized to ACTB gene expression are shown as means (±SD); * p<0.05; ** p≤0.01; *** p≤0.001; (B) Box plot comparing the CD44 tv4 (st) and CD44 tv3 (v8-10) transcript-specific expression (% of isoform) in colon adenocarcinoma TCGA (n=331) and primary normal colon epithelium GTEX (n=308) data (Welch's t-test - normal vs tumor tissue: CD44 tv4, p=3.952e-60, t=-18.87; CD44 tv3, p=7.748e-156, t=-36.12). Data were generated via the UCSC Xena platform (https://xenabrowser.net). (C) Representative flow cytometric dot blot diagrams and histograms documenting the CD44 surface presentation in membrane-intact LS1034 cells upon expose to IL-6 (10 ng/mL; mono-culture) or when co-cultured with human umbilical vein endothelial cells (HUVEC - EC), normal skin fibroblasts (VF2) or colon adenocarcinoma-derived fibroblasts (CF). LS1034 cells and CD44 highly positive fibroblasts were discriminated by CD326 labeling. Three examples of LS1034/VF2 co-cultures are documented to demonstrate the impact (i) of a higher fibroblast concentration (factor 1.7 -1.8) and (ii) of direct cell-cell versus paracrine interactions on the CD44 signal in LS1034. All acquired data are summarized in Table 1. (D) RT-PCR analysis of total and exon-specific (tv3) CD44 mRNA expression in LS1034 in vitro cells grown in direct contact with fibroblasts. Cells were sorted according to their CD326/CD44 expression pattern: CD326^-/CD44⁺ = VF2 fibroblasts, CD326⁺/CD44^- and CD326⁺/CD44⁺ = LS1034 cell fractions. LS1034 mono-culture and xenograft samples were analyzed in parallel as negative and positive controls. Primers A/B and C/D according to Figure 6D/F were applied and human β-actin (ACTB) mRNA served as reference. (E) Representative flow cytometric dot blot diagrams of CD44 (total) versus CD44v9 fluorescence signals on the surface of membrane-intact (PI-negative) VF2 fibroblasts (CD326^-) and LS1034 cells (CD326⁺) after 4 days of co-culturing. The cell type-specific isotype controls are shown as overlay (grey).