Proteomics reveals multiple routes to the osteogenic phenotype in mesenchymal stem cells

Kristin P. Bennett, Charles Bergeron, Evrim Acar, Robert F. Klees, Scott Vandenberg, Bülent Yener, George E. PlopperBMC 2007 · DOI

TL;DR

This study uses a novel multi-way tensor analysis of proteomic data to reveal that human mesenchymal stem cells (hMSC) can differentiate into osteoblasts via at least two distinct pathways, one driven by extracellular matrix (ECM) signaling and another by steroid hormone signaling. This work provides a deeper understanding of stem cell differentiation and suggests that ECM-driven differentiation leads to a more osteoblast-like phenotype.

What Problem Does This Paper Address?

Understanding how human mesenchymal stem cells (hMSC) differentiate into specialized cell types, like bone-forming osteoblasts, is crucial for regenerative medicine. While it's known that hMSC can differentiate into osteoblasts, the precise molecular mechanisms and whether there are multiple distinct routes are not fully understood. Previous studies have shown that factors like dexamethasone (a steroid hormone analog) and extracellular matrix (ECM) proteins can induce osteogenic differentiation, but it's unclear if these stimuli lead to the same intermediate states or final phenotypes. This paper aims to dissect these different differentiation pathways by analyzing protein expression profiles under various stimuli.

Methods

The researchers employed a sophisticated analytical approach combining proteomics with gene ontology (GO) analysis and advanced statistical modeling.

  • Proteomics: They used 2D Liquid Chromatography-Mass Spectrometry/Mass Spectrometry (2D LC-MS/MS) to identify and quantify proteins expressed in different cell populations:
  • Naïve hMSC
    • hMSC stimulated with osteogenic stimulant (OS) media
    • hMSC stimulated with ECM proteins (collagen I, vitronectin, laminin-5)
    • Human osteoblasts (hOST)
  • Gene Ontology (GO) Analysis: Identified proteins were annotated with GO terms to understand their biological functions and pathways. This created a structured dataset for analysis.
  • Directed Component Analysis (DCA): A novel two-way analysis method was developed. It projects the data onto a "differentiation axis" that directly connects naïve hMSC to hOST, allowing for the ranking of intermediate states based on their osteoblast-like characteristics.
  • Tucker3 Tensor Analysis: A three-way analysis was used to model the data as a "datacube" with modes for samples, GeneIDs (proteins), and GO categories. This method preserves the full dimensionality of the data, avoiding information loss inherent in two-way methods. It allows for the identification of complex interactions between these three modes.

Datasets

The study generated and analyzed proteomic data from six distinct experimental conditions:

  1. Naïve human mesenchymal stem cells (hMSC)
  2. hMSC stimulated with osteogenic stimulant (OS) media
  3. hMSC stimulated with collagen I (ECM protein)
  4. hMSC stimulated with vitronectin (ECM protein)
  5. Human osteoblasts (hOST)
  6. hMSC stimulated with laminin-5 (ECM protein) - used for validation.

The data was structured into a tensor with dimensions representing samples, GeneIDs (proteins), and Gene Ontology categories.

Key Insights and Ideas

  • Multiple Differentiation Routes: The most striking finding is the clear evidence for at least two distinct pathways for hMSC osteogenic differentiation: one influenced by ECM proteins and another by OS media (containing dexamethasone).
  • ECM vs. OS Phenotypes: ECM-stimulated hMSC appear to adopt an intermediate phenotype that is more osteoblast-like than that of OS-stimulated hMSC, as revealed by their position along the DCA-derived differentiation axis.
  • GO Categories as Biomarkers: Specific GO categories were identified as critical for defining the differentiation axis (e.g., cytokine activity, protein kinase activity) and for distinguishing between the OS and ECM intermediate states (e.g., oxidoreductase activity, amino acid metabolism).
  • Tensor Analysis Power: The multi-way tensor analysis proved essential for capturing the complex relationships between proteins, their functions (GOs), and the experimental conditions, providing a more comprehensive view than traditional two-way methods.

Key Results

  • DCA Differentiation Axis: A directed component analysis identified a "differentiation axis" that explains a significant portion of the variance between hMSC and hOST.
  • ECM-hMSC Closer to Osteoblasts: In the DCA plots, ECM-stimulated hMSC (collagen I, vitronectin, laminin-5) clustered together and were positioned closer to hOST than OS-stimulated hMSC.
  • Distinct Intermediate States: The second DCA component clearly separated ECM-hMSC from OS-hMSC, indicating they represent distinct intermediate phenotypes.
  • Protein Signatures: The study identified specific proteins and GO categories that characterize naïve hMSC, intermediate states, and fully differentiated osteoblasts, as well as those that differentiate the two intermediate pathways. For example, proteins involved in collagen synthesis were more prominent in ECM-stimulated cells, while those related to steroid metabolism were more prevalent in OS-stimulated cells.

Limitations and Future Work

  • Sample Size: The authors acknowledge the limitation of a low sample size (five main experimental conditions for the initial analysis) and used validation with laminin-5 to bolster their findings.
  • Protein Identification Threshold: The study focused on proteins detected above a certain threshold (200 pmol), meaning some lower-abundance proteins might have been missed.
  • Future Directions: The identified GO categories and specific proteins serve as potential biomarkers for understanding and potentially controlling stem cell differentiation. Future work could focus on validating these biomarkers and exploring the precise molecular mechanisms underlying the distinct ECM and OS signaling pathways. The authors suggest that these findings contribute to a better understanding of osteogenesis and could lead to improved treatments for bone-related diseases and tissue regeneration.

Summary updated

Abstract

The differences between the proteomic profiles of OS-stimulated hMSC and ECM-hMSC characterize different transitional phenotypes en route to becoming osteoblasts. This conclusion is arrived at via a three-mode tensor analysis validated using hMSC plated on laminin-5.
BibTeX
@inproceedings{bennett2007proteomics,
  author = {Kristin P. Bennett and Charles Bergeron and Evrim Acar and Robert F. Klees and Scott Vandenberg and Bülent Yener and George E. Plopper},
  title = {{Proteomics reveals multiple routes to the osteogenic phenotype in mesenchymal stem cells}},
  booktitle = {BMC},
  year = {2007},
  doi = {10.1186/1471-2164-8-380},
}
Figure