MCF-7 cells (ZQ0071) and HUVECs (ZQ1099) were acquired from Shanghai Zhongqiao Xinzhou Biotechnology Co., Ltd. (Shanghai, China). HMFs (MZ-2683) were obtained from Ningbo Mingzhou Biotechnology Co., Ltd. (Ningbo, China). Cell lines were authenticated by the Genomics Unit, using ShortTandem Repeat profiling (AmpFLSTR® ldentifler® Plus PCR Amplification Kit, Waltham, MA, USA). Mycoplasma test was perfomed in all cell lines every other week using the lycoAlertMycoplasma Detection Kit (LONZA, Basel, Switzerland), The test results were negative. MCF-7 cells, HMFs, and HUVECs were mixed at a ratio of 1:1:1 [12] and then cultured in a previously prepared Matrigel matrix (356237, Corning Co., Shanghai, China) to construct a 3D co-culture model. In the MCF-7 cell 3D culture, only MCF-7 cells were added in the Matrigel matrix. For preparation, the Matrigel matrix was removed from –80 °C and placed in a 4 °C refrigerator for 2 h. Cells were collected and re-suspended in the Dulbecco’s Modified Eagle Medium-high (DMEM-H) medium supplemented with 15% fetal bovine serum (FBS) and mixed with the Matrigel matrix together at a final concentration of 4.5 mg/mL for Matrigel and 6 $\times$ 10⁶/mL for cells. In total, 24 µL of 2.5 mmol cisplatin was added to each 2-mL medium for treatment. The final concentration was 1.25 mmol/mL. The experiment was performed on four groups, which were as follows: MCF-7 cells in the 3D culture, MCF-7 cells in the 3D culture treated with cisplatin, 3D co-culture model, and the 3D co-culture model with cisplatin. In each group, cells were harvested 2 days after cisplatin treatment and were subjected to scRNA-seq.

Using the Single Cell 3′ Library and Gel Bead Kit V3 (10$\times$ Genomics, 1000075, Pleasanton, CA, USA) and the Chrome Single Cell B Chip Kit (10$\times$ Genomics, 1000074), cell suspensions (300–600 live cells per microliter as measured using Count Star) were loaded onto a chromium single-cell controller (10$\times$ Genomics). Single-cell gel beads were generated in the solution according to the BD Rhapsody™ Single-Cell Analysis System manual. Thus, single cells were suspended in phosphate-buffered saline containing 0.04% bovine serum albumin. The captured cells were cleaved, and the released RNA was barcoded by performing reverse transcription in a single GEM. Reverse transcription was performed on the S1000TM Touch Thermal Cycler (Bio Rad, Hercules, CA, USA) at 53 °C for 45 min and then at 85 °C for 5 min. Next, it was held at 4 °C. Finally, cDNA was generated. The final library was quantified using the Qubit fluorescence assay with the Qubit dsDNA HS kit (Thermo Fisher, Carlsbad, CA, USA). The library was sequenced by Novogene Biotech Co., Ltd. (Beijing, China) on the NovaSeq 6000 in the paired-end mode.

Cells were revived and processed using BD Rhapsody single cell analysis system the BD Rhapsody Express Single-Cell Analysis System (BD Biosciences, Franklin Lake, NJ, USA), according to the manufacturer’s instructions. Briefly, an average of 20,000–30,000 pooled cells was loaded in each cartridge on the BD Rhapsody express single cell analysis system for single cell capture, and the library was prepared using BD™ Single-Cell Multiplexing Kit-Human and BD Rhapsody Single Cell Multiplexing Kit as per manufacturer’s guide (Doc ID: 214419 Rev. 2.0). Final libraries were quantified using Qubit and Agilent TapeStation and pooled to achieve a final concentration of 5 nM. The libraries were sequenced using NovaSeq 6000 by Novogene Biotech Co., Ltd. (Beijing, China).

To control quality, cells expressing 300 or $>$6000 genes were excluded. To eliminate dying cells or low-quality cells with extensive mitochondrial contamination, cells with mitochondrial genes accounting for $>$30% of the total genes were discarded. Subsequently, t-distributed random neighborhood embedding (t-SNE) was applied to the 2D representation of data structures. After clustering, the Bausingler (version 1.2.4) package was used to identify the types of co-cultured cells. After obtaining the differentially expressed gene information of each cell group, functional analysis such as Gene Ontology (GO) pathway enrichment analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) set analysis, and functional prediction were performed using data from the CancerSEA database (http://biocc.hrbmu.edu.cn/CancerSEA/), which is a multi-functional website designed to comprehensively explore the different functional states of cancer cells at the single-cell level. R package (version 3.6.2) was used to process the raw scRNA-seq data, and Seurat package (version 3.1.4) was utilized to analyze the gene expression data.

Total RNA was extracted from MCF-7 cells in 3D culture group and 3D co-culture model group (see 2.1) using TRIzol (Invitrogen, Carlsbad, CA, USA). Then, RNA was aliquoted to synthesize cDNA using the Reverse Transcription System (Promega, Madison, WI, USA). Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) products were amplified using a SYBR green PCR Master Mix kit (Qiagen, Germantown, PA, USA) according to the manufacturer’s instructions on the ABI Prism 7500 Detection System (Applied Biosystems, Carlsbad, CA, USA). Data were analyzed using the 2^{-Δ⁢Δ⁢Ct} method, and GAPDH was regarded as an internal control [13]. The following q-PCR primer sequences were used: S100P forward, 5^′-GCACCATGACGGAACTAGAGACA‑3^′, and reverse, 5^′-CAGGTCCTTGAGCAATTTATCCAC‑3^′; HSPB1 forward, 5^′-CTGACGGTCAAGACCAAGGATG-3^′, and reverse, 5^′-GTGTATTTCCGCGTGAAGCACC-3^′; COX6C forward, 5^′-GTAGCATTCGTGCTATCCCTGG-3^′, and reverse, 5^′-GATACCAGCCTTCCTCATCTCC-3^′; CRABP2 forward, 5^′-TTGAGGAGCAGACTGTGGATGG-3^′, and reverse, 5^′-GTTCTCTGGTCCACGAGGTCTT-3^′; SOD2 forward, 5^′-CTGGACAAACCTCAGCCCTAAC-3^′, and reverse, 5^′-AACCTGAGCCTTGGACACCAAC-3^′; GAPDH forward, 5^′-GCACCGTCAAGGCTGAGAAC‑3^′ and reverse, 5^′-TGGTGAAGACGCCAGTGGA‑3^′. Analysis between two groups was performed using an unpaired Student’s t-test.

Two types of 3D culture were established. One was constructed for individual MCF-7 cell 3D culture. Under a microscope, it appeared messy, and it had a rough edge (Fig. 1A–C). The other was established for MCF-7 cells co-cultured with HMFs and HUVECs. The cell masses presented with a smoother and more regular edge (Fig. 1D–F).

Fig. 1.

Cell morphology in the 3D culture. (A–C) Individual MCF-7 cells in the 3D culture produced irregular cell spheres. (D–F) Three-dimensional co-culture model for MCF-7 cells, HMFs, and HUVECs formed 3D structures with smooth edges. Scale bar value: 50 µm. MCF, michigan Cancer Foundation; HMFs, human mammary fibroblasts; HUVECs, human umbilical vein endothelial cells.

After controlling the quality of single-cell analysis (Fig. 2A), dead or injured cells were eliminated. Eventually, we obtained 2041 cells in MCF-7 cells in the 3D culture, 1043 cells in MCF-7 cells in the 3D culture treated with cisplatin, 2421 cells in the 3D co-culture model, and 4667 cells in the 3D co-culture model with cisplatin. The cells of the 3D co-culture model (from the 3D co-culture model and 3D co-culture model with cisplatin) were clustered into six according to the hypervariable gene using the standard data analysis tool K-means in Seurat package (Fig. 2B). Next, the cells were visualized in the t-SNE map. The heatmap presented the expressions of the top 5 differentially expressed genes in each of the six cluster (Fig. 2C, Table 1). The differentially expressed genes were analyzed using the Kruskal–Wallis test in R package. The cell type was identified with marker genes and the characteristic genes of each cell type. The HUVECs, HMFs, and MCF-7 cells of the 3D co-culture model, which was shown in the t-SNE map, were separated (Fig. 3A). Each cell type was featured individually in the sole t-SNE map (Fig. 3B–D). The heatmap depicted the expressions of the top 10 differentially expressed genes in HUVECs, MCF-7 cells, and HMFs (Fig. 2D, Table 2). Fig. 3E presents the characteristic genes of HUVECs including CCL20, MT2A, SOD2, MT1X, and CEBPB. In addition, the relative expressions of C15ORF48, PLAU, PDZK1IP1, and ARL4C in the three cell types were examined, as these genes had relatively high expression levels in HUVECs (Fig. 3E). Similarly, Fig. 3F depicts the characteristic genes of HMFs such as MMP2, SPARC, and COL1A1. In addition, the relative expressions of COL6A1, DCN, COL6A3, HLA-C, LUM, and IGFBP5 in the three cell types were examined, as these genes had a relatively high expression levels in HMFs (Fig. 3F). Further, Fig. 3G presents the characteristic genes of MCF-7 cells including TXNIP, COX6C, CRABP2, CCN5, CLDN4, TFF1, KRT19, and HLA-A. Fig. 3G shows the relative expressions of AL450405.1 in the three cell types, as it had a relatively high expression level in MCF-7 cells.

Fig. 2.

Cell cluster and single-cell transcriptome sequencing atlas. (A) Quality control for scRNA-seq. Violin plot of nFeature_RNA presented the number of genes captured in a sample cell. Violin plot of nCount_RNA presented the number of the total number of genes expressed in the cells. Violin plot of percent.mt presented the mitochondrial gene expression ratio. Violin plot of the erythrocyte gene expression ratio. (B) Cells of the 3D co-culture model were clustered into six. (C) The heatmap showed the expressions of the top 10 differentially expressed genes in each of the six clusters. (D) The heatmap showed the expressions of the top 10 differentially expressed genes in HUVECs, MCF-7 cells, and HMFs. scRNA-seq, single-cell RNA sequencing; tSNE, t-distributed Stochastic Neighbor Embedding; HUVECs, human umbilical vein endothelial cells; HMFs, human mammary fibroblasts.

Fig. 3.

Identification of cell types. (A) The map of the identification of all cells. (B–D) The map of distribution of HUVECs, HMFs, and MCF-7 cells. (E) The relative expressions of the characteristic genes of HUVECs. (F) The relative expressions of the characteristic genes of HMFs. (G) The relative expressions of the characteristic genes of MCF-7 cells.

To further assess the 3D co-culture model, the function differentialGeneTest in R package was used to calculate the differentially expressed genes in each pairwise comparison. As shown in Fig. 4A, the heatmap presented the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture and MCF-7 cells in the 3D co-culture model. Further, as shown in Fig. 4B, the heatmap depicted the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture with cisplatin and MCF-7 cells in the 3D co-culture model with cisplatin. According to the two figures, they showed plenty of similarities. With or without cisplatin, the expressions of SOD2, CEBPB, PLAU, SFN, IGFBP7, MT2A, and VIM were relatively high in MCF-7 cells in the 3D culture but relatively low in MCF-7 cells in the 3D co-culture model. Meanwhile, the expressions of S100P, ALDOC, HSPA1B, HSPB1, COX6C, CRABP2, TXNIP, and AQP3 were relatively low in MCF-7 cells in the 3D culture but relatively high in MCF-7 cells in the 3D co-culture model. RT-qPCR results demonstrated that SOD2 mRNA level was downregulated, while S100P, HSPB1, COX6C and CRABP2 mRNA expressions were upregulated in the 3D co-culture model in comparison to in the 3D culture alone (Supplementary Fig. 1A–E), which consistent to above analysis. As shown Fig. 4C, the heatmap depicted the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture and MCF-7 cells in the 3D culture with cisplatin. Further, as depicted in Fig. 4D, the heatmap presented with the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D co-culture model and MCF-7 cells in the 3D co-culture model with cisplatin. As shown in the two figures, regardless if the MCF-7 cells were co-cultured with other cells or not, the ZFP36 expressions were partially low in cells without cisplatin but relatively high in cells with cisplatin.

Fig. 4.

Analysis of differentially expressed genes and GO pathway enrichment based on single-cell RNA sequencing in MCF-7 cells. (A) The heatmap showed the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture and MCF-7 cells in the 3D co-culture model. (B) The heatmap showed the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture with cisplatin and MCF-7 cells in the 3D co-culture model with cisplatin. (C) The heatmap showed the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D culture and MCF-7 cells in the 3D culture with cisplatin. (D) The heatmap showed the expressions of the top 20 differentially expressed genes in MCF-7 cells in the 3D co-culture model and MCF-7 cells in the 3D. (E) GO pathway enrichment between MCF-7 cells in the 3D culture and MCF-7 cells in the 3D co-culture model. (F) GO pathway enrichment between MCF-7 cells in the 3D culture with cisplatin and MCF-7 cells in the 3D co-culture model with cisplatin. (G) Expression level of each differentially expressed gene in each breast cancer cell from the CancerSEA website database. (H) Functional correlation analysis was performed for differentially expressed genes of MCF-7 cells in 3D co-culture models and MCF-7 cells in 3D co-culture models of cisplatin from 4 different datasets of CancerSEA website. GO, Gene Ontology; TME, tumor microenvironment. FDR, false discovery rate

The bubble graph displayed the top 30 differential GO pathway enrichment between MCF-7 cells in the 3D culture and MCF-7 cells in the 3D co-culture model (Fig. 4E), including the pathways of focal adhesion, cell-substrate adherens junction, and cell-substrate junction. The bubble graph showed the top 30 differential GO pathway enrichment between MCF-7 cells in the 3D culture with cisplatin and MCF-7 cells in the 3D co-culture model with cisplatin (Fig. 4F), including focal adhesion, cell-substrate adherens junction, and cell-substrate junction.

The differentially expressed genes between MCF-7 cells in the 3D co-culture model and MCF-7 cells in the 3D co-culture model with cisplatin were uploaded to further understand the influence of the 3D co-culture model treated with cisplatin. The map explained the expression level of each differentially expressed gene in each breast cancer cell from the CancerSEA website database (Fig. 4G). Moreover, the heatmap performed the functional correlation analysis of differentially expressed genes between MCF-7 cells in the 3D co-culture model and MCF-7 cells in the 3D co-culture model with cisplatin from four different datasets in the CancerSEA website, which include mechanisms such as apoptosis, DNA damage, hypoxia, and metastasis (Fig. 4H).