Decoding breast cancer tissue-stroma interactions using species-specific sequencing

Abstract

Introduction: Decoding transcriptional effects of experimental tissue-tissue or cell-cell interactions is important; for example, to better understand tumor-stroma interactions after transplantation of human cells into mouse (xenografting). Transcriptome analysis of intermixed human and mouse cells has, however, frequently relied on the need to separate the two cell populations prior to transcriptome analysis, which introduces confounding effects on gene expression.

Methods: To circumvent this problem, we here describe a bioinformatics-based, genome-wide transcriptome analysis technique, which allows the human and mouse transcriptomes to be decoded from a mixed mouse and human cell population. The technique is based on a bioinformatic separation of the mouse and human transcriptomes from the initial mixed-species transcriptome resulting from sequencing an excised tumor/stroma specimen without prior cell sorting.

Results: Under stringent separation criteria, i.e., with a read misassignment frequency of 0.2 %, we show that 99 % of the genes can successfully be assigned to be of mouse or human origin, both in silico, in cultured cells and in vivo. We use a new species-specific sequencing technology-referred to as S(3) ("S-cube")-to provide new insights into the Notch downstream response following Notch ligand-stimulation and to explore transcriptional changes following transplantation of two different breast cancer cell lines (luminal MCF7 and basal-type MDA-MB-231) into mammary fat pad tissue in mice of different immunological status. We find that MCF7 and MDA-MB-231 respond differently to fat pad xenografting and the stromal response to transplantation of MCF7 and MDA-MB-231 cells was also distinct.

Conclusions: In conclusion, the data show that the S(3) technology allows for faithful recording of transcriptomic changes when human and mouse cells are intermixed and that it can be applied to address a broad spectrum of research questions.

Fig. 1

Species-specific sequencing—separation of mouse and human transcriptomes in silico. a Schematic flowchart of the principle steps in the S³ technology. A specimen of mixed human and mouse cells (for example, from a tumor–stroma xenograft experiment) is subjected to RNA-seq. The mixed transcriptome is bioinformatically separated into human and mouse transcriptomes, discarding transcripts with a defined maximum number of mismatches. b Fraction of reads that are species-ambiguous; i.e., cannot be assigned only to one species and therefore discarded. c Fraction of misassigned reads; i.e., species-specific reads that align to mouse genes for human samples or to human genes for mouse samples. Data in b and c are from three human (79 bp reads, in blue) and three mouse (51 bp reads, in pink) samples. d Percentage of expressed genes with full (orange) or partial (purple) loss of reads after S³ in the three human (H1–3) and mouse (M1–3) samples. e Scatterplot and Spearman correlation between a mouse sample (M1) and mouse expression values from S³ applied to a mix of human (H1) and mouse (M1) samples. f Spearman correlations for the human (blue) and mouse (pink) S³ components of in silico mixes of H1 + M1, H2 + M2 and H3 + M3. g The number of reads assigned by S³ as human, mouse or rat for three rat samples, normalized by the number of rat reads

Fig. 2

Analysis of ligand-induced Notch signaling using S³ technology. a Schematic depiction of the co-culture system used to analyze the Notch downstream response. The human MDA-MB-231 cells express robust levels of the Notch1 receptor and are co-cultured with mouse 3T3-L1 cells, which in some experiments are transfected with the Delta-like 4 (DLL4) ligand. b Analysis of 12xCSL-Luc activity for various combinations of co-culture of 3T3-L1 and MDA-MB-231 cells, where the latter are transfected with the Notch reporter 12xCSL-Luc. Note the increase in reporter activity where 3T3-L1 cells transfected with the DLL4 ligand are co-cultured with MDA-MB-231 cells, and that this increase is abrogated by the addition of N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT). Relative luciferase units (RLU) were normalized to beta-galactosidase values before fold change analysis. *p < 0.05, **p < 0.01 (Student’s t-test). Replicates per treatment group (n = 3) are from one culture split prior to transfection and measurement. c Fold change of expression levels (RPKM) for four genes (GPR1, MTHFS, SGK3 and NME2) from MDA-MB-231 cells co-cultured with 3T3-L1 cells transfected with DLL4 or green fluorescent protein (GFP) in the presence (+DAPT) or absence (-DAPT) of DAPT, as indicated. d Principal component analysis (PCA) of the genome-wide transcriptomes in MDA-MB-231 cells in response to DLL4 ligand-stimulation and DAPT treatment, as described in the figure. e Expression levels of human DLL4 in the co-cultures of MDA-MB-231 and 3T3-L1 cells, as described. Note the high level of DLL4 expression in cells transfected with a human DLL4 plasmid (the two bars to the right, light green)

Fig. 3

Analysis of two different modes of Notch ligand presentation. a Schematic depiction of activation of Notch by immobilized ligand (Fc-DLL4) or with Fc as control. b Analysis of 12xCSL-Luc activity in MDA-MB-231 cells cultured on immobilized Fc-DLL4 or Fc alone as control, and in the presence or absence of N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT), as indicated. Note the increase in reporter activity when cells are cultured on Delta-like 4 (DLL4) and that this activity is abrogated by the addition of DAPT. Relative luciferase units (RLU) were normalized to beta-galactosidase values before fold change analysis. *p < 0.05, **p < 0.01 (Student’s t-test). Replicates per treatment group (n = 3) are from one culture split prior to transfection and measurement. c Fold change of expression levels (RPKM) of four genes (P2RY11, MOB4, FAM183A and PRSS22) in the MDA-MB-231 cells in response to DLL4 ligand-stimulation and DAPT treatment, as described in the figure. d Principal component analysis (PCA) of the genome-wide transcriptomes in MDA-MB-231 cells in response to DLL4 ligand-stimulation and DAPT treatment, as indicated. e Comparison of Notch response signatures derived by DLL4 presented from co-cultured cells (left) or immobilized DLL4 (right). In the upper left circle are the 164 genes that are >2-fold upregulated in MDA-MB-231 cells by DLL4 on 3T3-L1 cells and the lower left circle denotes the 164 genes that are downregulated by DAPT. The overlap between these two categories (63 genes) are genes that are both upregulated by DLL4 and downregulated by DAPT, i.e. the Notch signature. For the immobilized ligand there are 76 genes that are upregulated by immobilized DLL4, and 84 genes are downregulated by DAPT. The overlap is 29 genes, which are both upregulated by immobilized DLL4 and downregulated by DAPT. Comparison of the “co-culture” and “immobilized” signatures identifies only one gene in common (the middle section of the figure). *p < 0.05 (Fisher’s exact test). f Fold change of expression levels (RPKM) from four well-established Notch target genes (NRARP, HES4, HES1 and SNAI1) in the MDA-MB-231 cells in response to DLL4 ligand-stimulation by co-culture (upper row) or DLL4 immobilized ligand (immob. lig., lower row), respectively, and in the presence or absence of DAPT as indicated. GFP green fluorescent protein

Fig. 4

Transcriptional consequences of xenografting MCF7 and MDA-MB-231 cells into mammary fat pads in mice. a,b Venn diagrams comparing MCF7 and MDA-MB-231 transcriptomes in vitro and in vivo. a The overlap (324 genes) between the 2142 genes that are upregulated in MCF7 tumors vs. in vitro culturing (left) and the 1456 genes that are upregulated in MDA-MB-231 tumors vs. in vitro culturing (right). b Similar analysis as in a, but for genes downregulated in MCF7 and MDA-MB-231 tumors as compared to in vitro culturing. c Principal component analysis (PCA) of untransplanted mammary gland, MCF7 tumor stroma (top to bottom: 3R, 1L, 3L) and MDA-MB-231 tumor stroma (top to bottom: 1, 32, 13, 22). d The number of genes upregulated (fold change (FC) >2) in MCF7 tumor stroma compared to MDA-MB-231 tumor stroma, as a subset of the number of genes upregulated in MCF7 or MDA-MB-231 tumor stroma compared to untransplanted mammary gland. e The number of genes downregulated (FC >2) in MCF7 tumor stroma compared to MDA-MB-231 tumor stroma, as a subset of the number of genes downregulated in MCF7 or MDA-MB-231 tumor stroma compared to untransplanted mammary gland. *p < 0.05 (Fisher’s exact test)