Reversible adaptive plasticity: a mechanism for neuroblastoma cell heterogeneity and chemo-resistance

Abstract

We describe a novel form of tumor cell plasticity characterized by reversible adaptive plasticity in murine and human neuroblastoma. Two cellular phenotypes were defined by their ability to exhibit adhered, anchorage dependent (AD) or sphere forming, anchorage independent (AI) growth. The tumor cells could transition back and forth between the two phenotypes and the transition was dependent on the culture conditions. Both cell phenotypes exhibited stem-like features such as expression of nestin, self-renewal capacity, and mesenchymal differentiation potential. The AI tumorspheres were found to be more resistant to chemotherapy and proliferated slower in vitro compared to the AD cells. Identification of specific molecular markers like MAP2, β-catenin, and PDGFRβ enabled us to characterize and observe both phenotypes in established mouse tumors. Irrespective of the phenotype originally implanted in mice, tumors grown in vivo show phenotypic heterogeneity in molecular marker signatures and are indistinguishable in growth or histologic appearance. Similar molecular marker heterogeneity was demonstrated in primary human tumor specimens. Chemotherapy or growth factor receptor inhibition slowed tumor growth in mice and promoted initial loss of AD or AI heterogeneity, respectively. Simultaneous targeting of both phenotypes led to further tumor growth delay with emergence of new unique phenotypes. Our results demonstrate that neuroblastoma cells are plastic, dynamic, and may optimize their ability to survive by changing their phenotype. Phenotypic switching appears to be an adaptive mechanism to unfavorable selection pressure and could explain the phenotypic and functional heterogeneity of neuroblastoma.

Keywords: neuroblastoma; phenotypic switching; tumor cell adaptation; tumor cell plasticity; tumor heterogeneity.

Figure 1

Reversible adaptive plasticity of neuroblastoma cell lines. (A) Neuro2a, IMR-32, and SK-N-SH cells cultured in DMEM + 10% FBS (D10) or Neurocult complete (NC) medium formed a monolayer of anchorage dependent (AD) adhered cells or anchorage independent (AI) tumorspheres, respectively. (B) Dissociated AI cells, originally derived from AD cells readily reversed their phenotype and formed AD monolayers when cultured in the D10 medium. Images acquired with 20× objective.

Figure 2

Cell cycle analysis of AD and AI phenotypes of neuroblastoma cell lines. (A) Cell cycle analysis of Neuro2a, SK-N-SH, and IMR-32 cells using propidium iodide. (B) Graphical representation of the cell cycle analysis reveals that AI phenotype of Neuro2a and SK-N-SH cells has significantly fewer numbers of cells in S-phase compared to their AD phenotype. Data points expressed as mean ± S.D. (n = 3). *p < 0.01 by Student’s t-test.

Figure 3

Effect of doxorubicin on metabolic activity and apoptosis of neuroblastoma cell lines. MTT assay of the untreated and doxorubicin treated Neuro2a (A), SK-N-SH (B), and IMR-32 (C) cells of either AD or AI phenotype revealed slower metabolic activity of the AI cells. Doxorubicin lowered the metabolic activity of both the phenotypes of all cell lines. (D) Graphical representation of the apoptosis assay at 24, 48, and 72 h following doxorubicin (Dx; 0, 0.01, and 0.1 μg/ml) treatment shows that the AI cells are more resistant to doxorubicin than the AD cells. The change in viability between the treated and untreated groups of AI cells is less for all time points and drug doses. Data points expressed as mean ± S.D. (n = 3).

Figure 4

Stem cell-like properties of AD and AI phenotypes of mouse neuroblastoma cell line. (A) Dissociated AI or AD phenotypes of Neuro2a cells reseeded at limited dilution in NC or D10 media exhibited self-renewal capability over time by reforming tumorspheres or growing as adhered cells, respectively. (B) Graphical representation of the percent self-renewal capacity of the AD and AI cells as measured by the percent of wells (seeded at 1 cell/well) growing adherent cells or tumorspheres, respectively. (C) Neuro2a cell-derived AI tumorspheres and AD adhered cells were allowed to differentiate followed by immunostaining for stem cells (nestin), neurons (Tuj1), oligodendrocytes (O4), and GFAP (astrocytes). (D) Graphical representation of the percent immunopositive cells revealed multipotency and ubiquitous expression of nestin for both phenotypes. Data points expressed as mean ± S.D. (n = 3). (E) Western blot analysis supports the abundant expression of nestin by both AD and AI phenotypes of Neuro2a cells. (F) Flow cytometric analysis supports that nestin is a ubiquitous marker of both tumor cell phenotypes. MFI, mean fluorescence intensity. (G) Western blot analysis showing complete absence of E-cadherin and abundance of vimentin and SNAIL proteins in both AD and AI phenotypes of neuro2a cells. Scale bar, 100 μm.

Figure 5

Stem cell-like properties of AD and AI phenotypes of human neuroblastoma cell lines. (A) SK-N-SH and IMR-32 cell-derived AI tumorspheres and AD adhered cells were allowed to differentiate followed by immunostaining for stem cells (nestin), neurons (Tuj1), oligodendrocytes (O4), and GFAP (astrocytes). Both the phenotypes of each cell type differentiated into neurons and oligodendrocytes but no astrocyte was detected. (B) Graphical representation of the percent immunopositive cells revealed multipotency and abundant expression of nestin for both the phenotypes. Data points expressed as mean ± S.D. (n = 3). (C) Western blot analysis supports the abundant expression of nestin by both AD and AI phenotypes. (D) Western blot analysis showing absence of E-cadherin and abundance of vimentin and SNAIL proteins in both AD and AI phenotypes of human neuroblastoma cell lines.

Figure 6

Molecular markers differentiate AD and AI phenotypes of neuroblastoma in vitro. (A) Western blot analysis of proteins that play vital role in neuroblastoma homeostasis revealed differences in expression of PDGFRβ, MAP2, Dcx, NCAM, survivin, and β-catenin between the AD and AI phenotypes. (B) The table represents relative protein expression of AD and AI cell types as determined by Western blot analysis. (C) Immunofluoresence analysis in vitro demonstrated that MAP2 is exclusively expressed by the AD cells whereas β-catenin and PDGFR-β is overexpressed in AI cells compared to the AD cells. (D) Flow cytometric analysis confirmed that MAP2 is expressed by the AD cells only, whereas β-catenin and PDGFRβ are overexpressed in AI cells. Scale bar, 50 μm. MFI, mean fluorescence intensity.

Figure 7

Neuroblastoma tumor cell heterogeneity in mouse model. (A) Flow cytometric phenotyping analysis on mouse neuroblastoma tumors of 10 mm diameter showed no remarkable difference in the expression of MAP2, β-catenin, and PDGFRβ in the two forms of tumor. MFI, mean fluorescence intensity. (B) Western blot analysis on mouse neuroblastoma tumors (5, 10, and 15 mm in diameters) revealed the heterogeneity of MAP2, PDGFRβ, and β-catenin expression in either forms of tumor suggesting that phenotypic transitions occur in vivo. (C) Immunofluorescence staining on frozen sections from mouse neuroblastoma tumors (grown from either AD or AI cells) of 10 mm diameter supports the flow cytometric and Western blot analysis of tumor cell heterogeneity. (D) H&E staining of tumors reveals that both AD and AI forms of Neuro2a cells gave rise to histopathologically similar tumors in mice. (E) In vivo tumorigenic potential of the AD and AI phenotypes of Neuro2a cells were compared by inoculating the mice with either form of cell phenotype and measuring the tumor over time. Both cell types gave rise to very large tumors and the growth rates were indistinguishable. Data points expressed as mean ± S.D. (n = 12). Scale bar, 50 μm.

Figure 8

Cell heterogeneity in human primary neuroblastoma tumors. Immunofluorescence staining with MAP2, β-catenin, and PDGFRβ on frozen sections of three human primary neuroblastoma specimens revealed scattered areas with differential MAP2, PDGFRβ, and β-catenin staining indicating similar heterogeneity of cells as seen in mouse tumors. Scale bar, 50 μm.

Figure 9

Effect of chemotherapy on tumor growth. (A) Doxorubicin treatment slowed down tumor growth in mice inoculated with AD) and AI forms of Neuro2a cells. Data points expressed as mean ± S.D. (n = 5). (B) Combined treatment with doxorubicin, metformin, Erlotinib (EGFR blocker), and PD173074 (FGFR blocker) delayed the initial onset of tumor in mice challenged with AD or AI phenotypes of Neuro2a. Data points expressed as mean ± S.D. (n = 6). These tumors were harvested at a size of 5 mm diameter.

Figure 10

Chemotherapy alters heterogeneity of cells in small tumors. (A) Flow cytometric analysis on mouse neuroblastoma tumors of 5 mm diameter showed no remarkable difference in the expression of MAP2, β-catenin, and PDGFRβ in the two forms of tumor (grown from either AD or AI cells). (B) Doxorubicin treatment had no effect on the cellular heterogeneity on tumors of 5 mm diameter grown from AD cells; whereas the AI-tumors of similar size lost their heterogeneity, remained in their native form, and did not transition during doxorubicin treatment. (C) Combination of doxorubicin and metformin prevented the AD phenotype from establishing itself in the early stage (5 mm diameter) of tumor formation as defined by the absence of MAP2+ cells and the cells remained in the AI form. (D) Combined dose of Erlotinib (EGFR inhibitor) and PD173074 (FGFR inhibitor) prevented the AI phenotype from establishing itself in the early stage (5 mm diameter) of tumor formation as defined by the presence of MAP2+ cells and the cells remained in the AD form. (E) Combination of chemotherapy and growth factor receptor inhibitors resulted in the emergence of newer phenotypes as evident from the presence of unique/unfamiliar populations of MAP2+ and PDGFRβ+ cells. MFI, mean fluorescence intensity. AD, AI.

Figure 11

Large doxorubicin and doxorubicin/metformin double-resistant tumors resume phenotypic heterogeneity. Flow cytometric analysis on large (10 mm diameter) mouse neuroblastoma tumors treated with doxorubicin (A) and doxorubicin/metformin (B) showed no remarkable difference in the expression of MAP2, β-catenin, and PDGFRβ in the two forms of tumor (grown from either AD or AI cells). Apparently either treatment displayed both AD (MAP2+) and AI (MAP2−) phenotypes.