Membrane RRM2-positive cells represent a malignant population with cancer stem cell features in intrahepatic cholangiocarcinoma

Abstract

Background: Intrahepatic cholangiocarcinoma (iCCA) is one of the most lethal malignancies and highly heterogeneous. We thus aimed to identify and characterize iCCA cell subpopulations with severe malignant features.

Methods: Transcriptomic datasets from three independent iCCA cohorts (iCCA cohorts 1-3, n = 382) and formalin-fixed and paraffin-embedded tissues from iCCA cohort 4 (n = 31) were used. An unbiased global screening strategy was established, including the transcriptome analysis with the activated malignancy/stemness (MS) signature in iCCA cohorts 1-3 and the mass spectrometry analysis of the sorted stemness reporter-positive iCCA cells. A group of cellular assays and subcutaneous tumor xenograft assay were performed to investigate functional roles of the candidate. Immunohistochemistry was performed in iCCA cohort 4 to examine the expression and localization of the candidate. Molecular and biochemical assays were used to evaluate the membrane localization and functional protein domains of the candidate. Cell sorting was performed and the corresponding cellular molecular assays were utilized to examine cancer stem cell features of the sorted cells.

Results: The unbiased global screening identified RRM2 as the top candidate, with a significantly higher level in iCCA patients with the MS signature activation and in iCCA cells positive for the stemness reporter. Consistently, silencing RRM2 significantly suppressed iCCA malignancy phenotypes both in vitro and in vivo. Moreover, immunohistochemistry in tumor tissues of iCCA patients revealed an unreported cell membrane localization of RRM2, in contrast to its usual cytoplasmic localization. RRM2 cell membrane localization was then confirmed in iCCA cells via immunofluorescence with or without cell membrane permeabilization, cell fractionation assay and cell surface biotinylation assay. Meanwhile, an unclassical signal peptide and a transmembrane domain of RRM2 were revealed experimentally. They were essential for RRM2 trafficking to cell membrane via the conventional endoplasmic reticulum (ER)-Golgi secretory pathway. Furthermore, the membrane RRM2-positive iCCA cells were successfully sorted. These cells possessed significant cancer stem cell malignant features including cell differentiation ability, self-renewal ability, tumor initiation ability, and stemness/malignancy gene signatures. Patients with membrane RRM2-positive iCCA cells had poor prognosis.

Conclusions: RRM2 had an alternative cell membrane localization. The membrane RRM2-positive iCCA cells represented a malignant subpopulation with cancer stem cell features.

Keywords: Cell membrane localization; Intrahepatic cholangiocarcinoma; RRM2; Tumor heterogeneity.

Fig. 1

RRM2 was highly expressed in a stem cell-like malignant population of iCCA. A Screening strategy to identify the key candidates which were highly expressed in iCCA subpopulations with malignant and stemness (MS) features. B The information of iCCA cohorts 1–3 and the hierarchical clustering analysis based on the activation status of the MS signature. iCCA patients were divided into MS^High, MS^Mid and MS^Low groups. Samples and genes are displayed as columns and rows, respectively. Hazard ratio of overall survival of MS^High group vs. MS^Low group were shown and log-rank test was performed. C The construction of OS4-GFP reporter was shown in the left. The GFP fluorescence level of GFP⁺ RBE and GFP⁻ RBE cells was analyzed by flow cytometry. Percentage of GFP⁺ cells at day 0 and at the day 5 after cell sorting was quantified. D OCT4, NANOG and SOX2 levels in OS4-GFP⁺ RBE cells and OS4-GFP⁻ RBE cells. Colony formation and cell migration were performed and compared between OS4-GFP⁺ RBE cells and OS4-GFP⁻ RBE cells. E Venn diagram of molecules with significantly higher expression in MS^high group vs. MS^low group in each cohort and in OS4-GFP⁺ vs. OS4-GFP⁻ RBE cells (fold > 2). F Relative expression level of RRM2 in MS^High group and MS^Low group of iCCA cohorts 1–3 and RRM2 protein level in OS4-GFP⁺ RBE cells and OS4-GFP⁻ RBE cells. G Relative expression level of RRM2 in iCCA tumor tissues and non-tumor tissues in iCCA cohorts 1–2. H Percentage of high/medium IHC staining of RRM2 in 21 types of cancers and number of cases with different RRM2 IHC staining in iCCA tumors, normal bile ducts and hepatocytes. The information was collected from the Human Protein Atlas. (C, D, F, G) The Student’s t-test was used

Fig. 2

Silencing RRM2 significantly suppressed malignancy and stemness features of iCCA cells. A Silencing efficiency of RRM2 siRNAs in RBE and HUCCT1 cells was evaluated by qRT-PCR and western blot. B-D Cell viability (B), colony formation (C) and cell migration (D) were performed in RBE and HUCCT1 cells transfected with siCtrl or siRRM2. E Cell viability was examined in RBE and HUCCT1 cells transfected with siCtrl or siRRM2 followed by Doxorubicin treatment. F Spheroid formation assay was performed in HUCCT1 cells transfected with siCtrl or siRRM2. Spheroids with diameter ≥ 50 μm were counted. G, H Tumorigenicity assay with 1 × 10⁴ HUCCT1 cells transfected with siCtrl and siRRM2 with or without Matrigel in male BALB/c nude mice. Five mice for each group (siCtrl, siRRM2#1 and siRRM2#2) were used. Images of tumors derived from nude mice were shown. Tumor occurrence rate and tumor volume were compared. B, E, G, H Two-way ANOVA was used. C, D, F The Student’s t-test was used

Fig. 3

RRM2 presented cell membrane localization in iCCA tumors. A Representative images of RRM2 IHC staining in iCCA tumors, normal bile duct and hepatocytes of iCCA cohort 4. B IHC staining scores for RRM2 membrane localization and cytoplasm localization were measured in iCCA tumors, normal bile duct and hepatocytes, the Student’s t-test was used. C Kaplan–Meier analysis of overall survival in cohort 4 based on RRM2 membrane staining scores or cytosol staining scores. Log-rank test was performed. Mem, membrane; Cyto, cytoplasm. D Confocal microscopy images of endogenous RRM2 in RBE and HUCCT1 cells with or without permeabilization by Triton X-100. Yellow arrows indicate cell membrane localization of RRM2. E Cell fractionation assay by ultra-highspeed centrifugation was performed in RBE and HUCCT1 cells with or without Na₂CO₃ treatment. F Cell surface biotinylation assay was performed in RBE and HUCCT1 cells. G The predicted protein structure of RRM2 according to multiple prediction online tools. SP, signal peptide; TM, transmembrane domain. H The diagram depicting RRM2-Flag vector and cell surface biotinylation assay with RBE cells transfected with RRM2-Flag

Fig. 4

The unclassical signal peptide and the potential transmembrane domain were examined and essential for RRM2 cell membrane localization. A The flow chart of IP/Mass Spec to assess the undersized RRM2. The predicted peptides (up, grey) and Mass Spec detected peptides (down, dark red) of RRM2 were shown. Yellow lines represent “K” or “R” residues of RRM2. B Construction of RRM2 vectors, and the expression of Flag-RRM2 and RRM2-Flag in RBE and HUCCT1 cells. C Schematic diagram of a group of RRM2 truncations and mutations related to its signal peptide. D The expression of RRM2 with different truncations and mutations of SP region was detected by western blot using anti-Flag. E Confocal microscopy images of exogenous RRM2 detected by anti-Flag in RBE and HUCCT1 cells. Yellow arrows indicate the cell membrane localization of RRM2. Percentage of RRM2 cell membrane localization was measured (mean ± SD). F Schematic diagram of RRM2 vectors with truncation and mutation of the predicted TM region. G The expression of RRM2 with 223-246aa truncation or mutations of hydrophobic amino acids in 223-246aa region was detected by anti-Flag in RBE cells. H Confocal microscopy images of exogenous RRM2 detected by anti-Flag in RBE cells. Yellow arrows indicate the cell membrane localization of RRM2. Percentage of RRM2 cell membrane localization was measured (mean ± SD). (E, H) The Student’s t-test was used

Fig. 5

RRM2 trafficked to cell membrane via classical ER-Golgi pathway and membrane RRM2 positive cells could be enriched via cell sorting. A Confocal microscopy images of endogenous RRM2 in RBE and HUCCT1 cells with or without permeabilization detected by a RRM2 N-terminus antibody and a RRM2 C-terminus antibody. B Membrane RRM2 staining in RBE and HUCCT1 cells was determined by flow cytometry via using the antibody which recognized 1-111aa of RRM2, with or without BFA treatment. Grey line, no staining; red line, DSMO treatment; blue line, BFA treatment. C Confocal microscopy images of endogenous RRM2 and BiP and their co-localization in RBE and HUCCT1 cells. D Flow cytometry analysis of parental iCCA cells and the sorted membrane RRM2⁺ cells (the top 5% of cells from the RRM2 staining) and the sorted membrane RRM2⁻ cells from parental RBE and HUCCT1 cells. The antibody recognizing 1-111aa of RRM2 was used. E IF assay was performed with RRM2 antibody for the sorted RRM2⁺ and RRM2⁻ cells at day 1 after cell soring. Confocal microscopy images of endogenous RRM2 were shown

Fig. 6

Membrane RRM2⁺ cells were a malignant population with CSC phenotypic features. A RRM2⁺ cells and RRM2⁻ cells were sorted by FACS and analyzed by flow cytometry in RBE and HUCCT1 cells at day 0 and day 7 after cell sorting. B Spheroid formation assay in RRM2⁺ HUCCT1 and RRM2⁻ HUCCT1 cells. Spheroids with diameter ≥ 50 μm were counted. C, D Colony formation and cell migration assays in RRM2⁺ and RRM2⁻ iCCA cells. E Tumorigenicity assay was performed with RRM2⁺ HUCCT1 and RRM2⁻ HUCCT1 cells in male BALB/c nude mice. Four mice for each group were used. Tumor occurrence rate and tumor volume were compared. The representative images were shown. (B-D) The Student’s t-test was used. (E) Two-way ANOVA was used

Fig. 7

Membrane RRM2⁺ cells possessed the malignant and stemness molecular features. A RNA sequencing of the sorted RRM2⁺ HUCCT1 and RRM2⁻ HUCCT1 cells was performed and the differentially expressed genes between two groups were identified. These genes were termed as membrane RRM2 signature. Hierarchical clustering analysis of the sorted RRM2⁺ and RRM2⁻ HUCCT1 cells was performed with this signature. B Clustering analysis with the membrane RRM2 signature classified iCCA patients into membrane RRM2-positive (mRRM2⁺)-like group and membrane RRM2-negative (mRRM2⁻) -like group. Hazard ratio of overall survival of mRRM2⁺ -like group and mRRM2⁻ -like group in iCCA cohorts 1–3 was shown. Log-rank test was performed. C In iCCA cohorts 1–3, MS signature activation status in mRRM2⁺-like group and mRRM2⁻ -like group. Chi-squared test was performed. D The top 20 signatures enriched in mRRM2⁺-like group analyzed by GSEA in iCCA cohorts 1–3. The number of different types of signatures was shown. E Protein expression of OCT4, SOX2, NANOG, β-catenin and RRM2 in the sorted RRM2⁺ and RRM2⁻ iCCA cells. F Expression of β-catenin direct target genes was detected by qRT-PCR in the sorted RRM2⁺ iCCA and RRM2⁻ iCCA cells. G β-catenin protein level in cytoplasm fraction and nuclear fraction in the sorted RRM2⁺ and RRM2⁻ iCCA cells