Combining perspectives: Multiscale integration of Stem Cell research

Posted: 12 December 2009 | Dr George Plopper, Associate Professor, Rensselaer Polytechnic Institute and Amanda Lund, Postdoctoral Fellow, Integrative Biosciences Institute (IBI) Faculty of Life Sciences | No comments yet

The promise of stem cell-based therapy is predicated on harnessing the plasticity of stem cell phenotypes to repair or replace damaged tissues. As technologies for detecting, isolating, modifying, and tracking stem cells improve, the very definition of what constitutes a stem cell is now an open question. Addressing this fundamental problem has triggered an explosion […]

The promise of stem cell-based therapy is predicated on harnessing the plasticity of stem cell phenotypes to repair or replace damaged tissues. As technologies for detecting, isolating, modifying, and tracking stem cells improve, the very definition of what constitutes a stem cell is now an open question. Addressing this fundamental problem has triggered an explosion of activity that spans the entire breadth of biological fields, from molecular biology to population biology. While this has clearly increased the gross amount of information concerning stem cells, its net impact is limited by a lack of integrative multiscale models that are readily accessible to researchers from many disciplines. The field of mesenchymal stem cell biology is a good example of the strengths and limitations of this segregative approach. The goal of this brief review is to highlight some of the most promising recent advances in mesenchymal stem cell research, with an emphasis on how data gathered from one level can benefit research across multiple scales.

Multiscale modeling as an integrative tool

The high expectations generated by stem cells and their potential clinical applications have resulted in an alignment of several biological disciplines around a central purpose. While this has been a positive force for streamlining collaboration between clinicians and basic scientists, it has also exposed gaps in our ability to synthesise this information into a coherent whole. In part, this reflects the historical subdivision of biomedical sciences into discrete fields that by necessity establish their own jargon and measures of “success.” While a great deal of effort is committed to achieving these successes, all too often it fails to impact other fields aiming for the same target, stem cell therapy. In many ways, the current situation in stem cell biology resembles that of the mid-1980s in the cancer field, when bench-to-bedside research was more concept than reality1.

Multiscale modeling has emerged as a powerful means to integrate research across many levels, from molecular structure to potential therapeutic interventions2,3. These models also permit the generation and testing of hypotheses, creating an iterative, systems biology loop4. All systems biology approaches focus on defining three characteristics of biological systems: robustness (ability to maintain phenotypic stability in response to perturbation), modularity (clustering of components into functional “teams”), and, most importantly, emergent properties (behaviors unique to the entire system, and not found in any of its constituents)5. At present, all three of these remain largely unknown for most stem cells. To accelerate discovery, systems biology loops learn more quickly than traditional trial-and-error approaches, intelligently shrinking the experimental search space to identify high-priority experiments.

Effective models often result from close collaboration between wet-bench experimentalists and computation experts6, who bring the rigor of predictive modeling strategies from a variety of disciplines, ranging from advertising campaigns to manufacturing processes that ensure proper seasoning of snack foods7. For many biologists, the key to developing such a multiscale model is to convert a medical goal into a design optimisation problem. Four stages are required: (1) initial data is captured to yield a profile, (2) the predictive model, built from profiles, is used to identify more data and hypotheses to be tested, (3) the models are tested directly to assess how close the process is to the goal, then (4) this new data can be used to improve the model and ultimately yield a better product. The challenge is to define the “gold standard” for such optimisation, since each scale of stem cell biology focuses on its own set of features.

Stem cells are difficult to define

The stem cell concept was first proposed by Alexander Maximov in 1906 to describe a cell that is capable of reconstituting hematopoiesis in bone marrow8. His ideas were experimentally confirmed in 19639. Since that time the definition of a stem cell has expanded and evolved, to include several different perspectives. To the clinician, a stem cell has the ability to replicate and differentiate to regenerate a healthy replacement for damaged tissue; to the molecular biologist, a stem cell undergoes asymmetric mitosis to generate two daughter cells with distinct gene and protein expression profiles, such that one retains the ability to complete another round of asymmetric division while the other commits to establishing a mature, fully differentiated phenotype10. For others, stem cells are those cells capable of replacing an entire adult tissue de novo, either by generating the necessary cells themselves or recruiting them from elsewhere in the body. By analogy, defining a stem cell is somewhat like defining a “healthy” diet11.

While any “true” stem cell may possess each of these traits, plus others, there is currently no consensus. Indeed, the plethora of proposed “defining characters” in stem cells has caused some to question whether any single cell in the human body qualifies12. Current classification of stem cells is primarily focused on tissue of origin and/or differentiation potential. At present, the ~20 known human stem cells are grouped into two differentiation classes: pluripotent stem cells possess the ability to differentiate into any adult tissue type, while multipotent cells possess a more limited differentiation potential. Most stem cells are multipotent, capable of generating a handful of cell types that function together in the same tissue. This “all-or-some” boundary is helpful for discriminating between embryonic stem cells and most others, but offers no mechanistic insight as to how “stemness” is generated or controlled. The recent demonstration that fibroblasts and other adult cells can be reprogrammed to resemble pluripotent embryonic stem cells by inducing expression of only three transcription factors13 blurs the boundary between the two categories considerably.

From a design optimisation standpoint, an absolute definition is not necessary; one merely needs to identify a quantitative goal, such as achieving a specific gene expression pattern or specificity of homing to a defined target tissue. In this light, the abundance of “discriminative features” is an advantage, because it provides greater coverage of stem cell phenotype. As an example, we will examine the mesenchymal stem cell field at several levels to identify possible inputs to such a model.

Molecular level: Can we optimise a molecular profile?

Originally described in bone marrow by Friedenstein14, adult human mesenchymal stem cells (hMSC) are an excellent example of both the promise and the confusion regarding stem cell applications. hMSC are multipotent, self-renewing cells that can be isolated from at least 12 different adult tissues15. They differentiate into at least eight cell types representing all three embryonic germ layers, and elicit very little immune response, making them attractive for transplant therapies16. It is still not clear how multipotent these cells are, or even if the same cell types reside in each tissue source.

The initial method for isolating hMSC was expanding the fraction of bone marrow cells that adhere to a glass or tissue culture plastic substrate and subsequently generate a colony forming unit-fibroblast (CFU-F)17. To overcome possible contamination by fibroblasts and other cells, most hMSC are now isolated by sorting for cell surface markers. Stan Grothos and others have been instrumental in defining a handful of cell surface markers for hMSC and isolating cell populations that express some or all of these markers18-20, suggesting that hMSCs may exist in several transitional states even before they commit to a differentiation program.

This heterogeneity is reflected by genome and proteome profiles as well21. Recent studies, including our own, identify hundreds of genes and dozens of proteins as potential markers for undifferentiated hMSC22-26. Indeed, these profiles are even reported to change as a function of time in culture27. To further complicate matters, the signal transduction pathways controlling gene expression and protein turnover are themselves dynamic and linked by complex regulatory mechanisms. Our own work with extracellular matrix-stimulated signaling pathways demonstrates that the same signaling proteins can promote or retard differentiation of hMSC depending on the three dimensional arrangement of the matrix and receptors28. Variations in the mechanical properties of the extracellular environment also have a profound effect of the genomic profile of these cells29.

One of the most pressing questions for defining hMSCs at the molecular level is how specific environmental stimuli, signaling molecules, and gene regulatory networks cooperate to control growth and differentiation of these cells. Multiscale models show great promise in this area. For example, deterministic, probabilistic, and statistical learning models are used to extract information about proteomic networks30. The stochastic nature of proteomic, genomic, and signaling data suggests that other machine learning methods, such as Bayes Networks, can also be used to model proteomic networks such as protein-protein interactions. Doug Lauffenburger’s group at Massachusetts Institute of Technology pioneered the use of systems biology with stem cells31,32, and this has now expanded into a robust research field33-35. One key component of this strategy is to adopt macro-scale approaches to gather large amounts of molecular data. For example, commercial protein phosphorylation arrays capture the signaling behavior of dozens of protein kinases controlling cell growth and differentiation, and can be combined with proteomic and DNA microarrays to capture a comprehensive picture of cellular phenotype in these studies36.

Predictions from such models can be tested experimentally using existing techniques to generate mechanistic relationships between these molecules. This would benefit not only those interested in pinning down the discriminative profile(s) of hMSC, but would provide the necessary gold standards for optimisation of the culture conditions for expanding and inducing differentiation of these cells. This is how data from one level of stem cell research can positively impact others.

Cellular level: Can we define the hMSC niche?

Like the stem cell, the notion of a specific environment that controls its growth and preserves its undifferentiated state, or niche, emerged as a concept before its existence was positively demonstrated37. The molecular makeup and regulatory mechanisms of most stem cell niches, including that of hMSCs, are almost entirely unknown. Extrapolating from what we know about hematopoietic and reproductive stem cell niches in animal models, the hMSC niche is likely defined by three features: 1. cell-cell adhesions to anchor stem cells and signal their self-renewal, 2. physical organisation that provides spatial and regulatory cues and 3. soluble signals that regulate maintenance. These three features are maintained by heterologous cell populations, the extracellular matrix, and paracrine signaling, respectively.

How do we find these features? Some promising approaches include in vivo fluorescence microscopy, in vitro culture of intact 3D tissues, and reconstitution with tissue extracts and engineered scaffolds. Most often, the strategies operate largely independent of the pursuit for the definitive stem cell profile occurring at the molecular level. Without a definitive target cell to characterise, those in search of its niche are handicapped from the outset. Multiscale models have the potential to use input such as composition and 3D orientation of specific extracellular stimuli to both predict expression of marker proteins and genes but also to help predict the composition of the extracellular environment in the cell niche.

This is precisely the approach used by van Leeuwen et al.2 to establish links between molecular signaling networks, mechanical properties of cellular interactions, and tissue-level organisation of the gut epithelium. Virtual microdissections and labeling experiments revealed that the gut epithelial stem cell niche is dynamic and the stem cells contained within it are mobile even as they retain their stemness. Such an elegant approach could likely assist in the definition of many other stem cell niches, including those for hMSCs in a variety of tissues. Most importantly, this strategy takes full advantage of the genetic, molecular, and histological data focused on characterising the maintenance of an entire tissue.

Tissue level: Can we engineer a stem cell niche ex vivo?

For many investigators, the ultimate goal of stem cell research is to harness stem cells’ capacity to reconstitute tissues to treat specific injuries or diseases. In some cases, simply injecting hMSC into decellularised tissue explants (e.g. heart valves38) or damaged tissues (e.g., skeletal39 and cardiac40 muscle) yields promising results. But in most cases, at least partial reconstitution of the proper cellular niche will be required for long term success.

Since the late 1980s engineering design principles have been applied to living systems to create replacement tissues de novo. In its most basic sense, an engineered tissue construct (ETC) is a three-dimensional assembly of one or more cell types suspended in an extracellular scaffold material and fed by soluble molecules, including growth factors, hormones, and nutrients. Once assembled, the ETCs are intended to be implanted as replacements for damaged or diseased tissues. The results thus far have been promising41, and some enjoy widespread use in the clinic42. While stem cell-based ETCs hold great promise due to their potential for self-renewal, creating a suitable niche ex vivo is a major hurdle to widespread stem cell therapy43.

The pace of ex vivo stem cell niche development is hampered by two design problems. First is sorting through the enormous number of possible “ingredients” (cells, scaffolds, biochemicals, stimuli). As an example, some groups have identified as many as nine classes of functional parameters for designing musculoskeletal tissues: differential fiber length, in vivo force and displacement, variations in relative attachment site locations, loading from adjacent structures, fiber interactions, types of insertion, regional variations in material properties, nonparallel fiber orientations, and complex loading within the structure44. Tissue engineers develop “educated guesses” based on prior experience to guide them. Yet this task is too daunting for even the best-informed tissue engineers to solve by sheer brute-force experimentation.

Second, most educated guesses in tissue engineering are based on data from complex measures of tissue performance. These tests are often difficult, expensive, and time consuming. If one wants to assemble a blood vessel, for example, the only reliable ex vivo measures of “functionality” are end-point measures (burst pressure, tensile strength, strain modulus, etc.)45. Increasing the amount of reliable data would be a great help.

Multiscale models such as that developed by van Leeuwen et al. hold great promise for solving these issues. Integrating the wealth of molecular data with performance data and other measures (e.g., multispectral imaging, electrophysiology, etc.46), can reveal correlations that brute force analysis simply cannot detect. Once these constructs enter clinical trials, clinical outcomes add yet another level of input. The ultimate test of our careful deliberation will take place in stem cell-based clinical trials; the recent report of poorly characterised “fetal neural stem cells” causing a brain tumour after being transplanted in a 13 year old ataxia telangiectasia patient47 underscores the danger of moving forward without due diligence.


  1. Siminovitch L: Advances in cancer research: bench to bedside. J Thorac Cardiovasc Surg, 100: 874-878(1990).
  2. van L, I, Mirams GR, Walter A, Fletcher A, Murray P, Osborne J et al.: An integrative computational model for intestinal tissue renewal. Cell Prolif (2009) In Press.
  3. Engler AJ, Humbert PO, Wehrle-Haller B, Weaver VM: Multiscale modeling of form and function. Science, 324: 208-212 (2009).
  4. Schadt EE, Zhang B, Zhu J: Advances in systems biology are enhancing our understanding of disease and moving us closer to novel disease treatments. Genetica, 136: 259-269 (2009).
  5. Aderem A: Systems biology: its practice and challenges. Cell, 121: 511-513 (2005).
  6. Liu ET: Systems biology, integrative biology, predictive biology. Cell,121: 505-506 (2005).
  7. Yu H, MacGregor JF: Multivariate image analysis and regression for prediction of coating content and distribution in the production of snack foods. Chemometrics and Intelligent Laboratory Systems, 67: 125-144 (2003).
  8. Maximow AA: U¨ber experimentelle Erzeugung von Knochenmarks-Gewebe. Anatomischer Anzeiger, 28: 24-38 (1906).
  9. Becker AJ, McCulloch EA, Till JE: Cytological demonstration of the clonal nature of spleen colonies derived from transplanted mouse marrow cells. Nature, 197: 452-454 (1963).
  10. Orford KW, Scadden DT: Deconstructing stem cell self-renewal: genetic insights into cell-cycle regulation. Nat Rev Genet, 9: 115-128 (2008).
  11. Drewnowski A, Maillot M, Darmon N: Should nutrient profiles be based on 100 g, 100 kcal or serving size? Eur J Clin Nutr, 63: 898-904 (2009).
  12. Parker GC, Anastassova-Kristeva M, Eisenberg LM, Rao MS, Williams MA, Sanberg PR et al.: Stem cells: shibboleths of development, part II: Toward a functional definition. Stem Cells Dev, 14: 463-469 (2005).
  13. Kaji K, Norrby K, Paca A, Mileikovsky M, Mohseni P, Woltjen K: Virus-free induction of pluripotency and subsequent excision of reprogramming factors. Nature, 458: 771-775 (2009).
  14. Friedenstein AJ: Precursor cells of mechanocytes. Int Rev Cytol, 47:327-59: 327-359, (1976).
  15. da Silva ML, Chagastelles PC, Nardi NB: Mesenchymal stem cells reside in virtually all post-natal organs and tissues. J Cell Sci, 119: 2204-2213 (2006).
  16. Garcia-Castro J, Trigueros C, Madrenas J, Perez-Simon JA, Rodriguez R, Menendez P: Mesenchymal stem cells and their use as cell replacement therapy and disease modelling tool. J Cell Mol Med, 12: 2552-2565 (2008).
  17. Friedenstein AJ, Chailakhjan RK, Lalykina KS: The development of fibroblast colonies in monolayer cultures of guinea-pig bone marrow and spleen cells. Cell Tissue Kinet, 3: 393-403 (1970).
  18. Gronthos S, Zannettino AC: A method to isolate and purify human bone marrow stromal stem cells. Methods Mol Biol, 449: 45-57 (2008).
  19. Battula VL, Treml S, Bareiss PM, Gieseke F, Roelofs H, de ZP et al.: Isolation of functionally distinct mesenchymal stem cell subsets using antibodies against CD56, CD271, and mesenchymal stem cell antigen-1. Haematologica, 94: 173-184 (2009).
  20. Majore I, Moretti P, Hass R, Kasper C: Identification of subpopulations in mesenchymal stem cell-like cultures from human umbilical cord. Cell Commun Signal, 7: 6 (2009).
  21. Phinney DG: Biochemical heterogeneity of mesenchymal stem cell populations: clues to their therapeutic efficacy. Cell Cycle, 6: 2884-2889 (2007).
  22. Klees RF, Salasznyk RM, Vandenberg S, Bennett K, Plopper GE: Laminin-5 activates extracellular matrix production and osteogenic gene focusing in human mesenchymal stem cells. Matrix Biol, 26: 106-114 (2007).
  23. Song L, Webb NE, Song Y, Tuan RS: Identification and functional analysis of candidate genes regulating mesenchymal stem cell self-renewal and multipotency. Stem Cells, 24: 1707-1718 (2006).
  24. Mareddy S, Broadbent J, Crawford R, Xiao Y: Proteomic profiling of distinct clonal populations of bone marrow mesenchymal stem cells. J Cell Biochem, 106: 776-786 (2009).
  25. Menicanin D, Bartold PM, Zannettino AC, Gronthos S: Genomic profiling of mesenchymal stem cells. Stem Cell Rev Rep, 5: 36-50 (2009).
  26. Salasznyk RM, Klees RF, Vandenberg S, Bennett K, Plopper GE: Gene focusing as a basis for controlling stem cell differentiation. Stem Cells Dev, 14: 608-620 (2005).
  27. Lee KA, Shim W, Paik MJ, Lee SC, Shin JY, Ahn YH et al.: Analysis of changes in the viability and gene expression profiles of human mesenchymal stromal cells over time. Cytotherapy, 1-10 (2009).
  28. Lund AW, Stegemann JP, Plopper GE: Inhibition of ERK promotes collagen gel compaction and fibrillogenesis to amplify the osteogenesis of human mesenchymal stem cells in three-dimensional collagen I culture. Stem Cells Dev, 18: 331-341 (2009).
  29. Engler AJ, Sen S, Sweeney HL, Discher DE: Matrix elasticity directs stem cell lineage specification. Cell, 126: 677-689 (2006).
  30. Janes KA, Lauffenburger DA: A biological approach to computational models of proteomic networks. Curr Opin Chem Biol, 10: 73-80 (2006).
  31. Prudhomme W, Daley GQ, Zandstra P, Lauffenburger DA: Multivariate proteomic analysis of murine embryonic stem cell self-renewal versus differentiation signaling. Proc Natl Acad Sci U S A, 101: 2900-2905 (2004).
  32. Woolf PJ, Prudhomme W, Daheron L, Daley GQ, Lauffenburger DA: Bayesian analysis of signaling networks governing embryonic stem cell fate decisions. Bioinformatics, 21: 741-753 (2005).
  33. Halley JD, Burden FR, Winkler DA: Stem cell decision making and critical-like exploratory networks. Stem Cell Res (2009). In Press.
  34. Foster DV, Foster JG, Huang S, Kauffman SA: A model of sequential branching in hierarchical cell fate determination. J Theor Biol (2009). In Press.
  35. Murali TM, Rivera CG: Network legos: building blocks of cellular wiring diagrams. J Comput Biol, 15: 829-844. (2008)
  36. Albeck JG, MacBeath G, White FM, Sorger PK, Lauffenburger DA, Gaudet S: Collecting and organizing systematic sets of protein data. Nat Rev Mol Cell Biol, 7: 803-812 (2006).
  37. Schofield R: The relationship between the spleen colony-forming cell and the haemopoietic stem cell. Blood Cells, 4: 7-25 (1978).
  38. Iop L, Renier V, Naso F, Piccoli M, Bonetti A, Gandaglia A et al.: The influence of heart valve leaflet matrix characteristics on the interaction between human mesenchymal stem cells and decellularized scaffolds. Biomaterials, 30: 4104-4116 (2009).
  39. Bhagavati S: Stem cell based therapy for skeletal muscle diseases. Curr Stem Cell Res Ther, 3: 219-228 (2008).
  40. Katritsis D: Cellular replacement therapy for arrhythmia treatment: early clinical experience. J Interv Card Electrophysiol, 22: 99-105 (2008).
  41. Malchesky PS: Artificial Organs 2008: a year in review. Artif Organs, 33: 273-295 (2009).
  42. Shieh SJ, Vacanti JP: State-of-the-art tissue engineering: from tissue engineering to organ building. Surgery, 137: 1-7 (2005).
  43. Toyoda M, Takahashi H, Umezawa A: Ways for a mesenchymal stem cell to live on its own: maintaining an undifferentiated state ex vivo. Int J Hematol, 86: 1-4 (2007).
  44. Butler DL, Shearn JT, Juncosa N, Dressler MR, Hunter SA: Functional tissue engineering parameters toward designing repair and replacement strategies. Clin Orthop Relat Res, S190-S199 (2004).
  45. Nerem RM: Tissue engineering of the vascular system. Vox Sang, 87 Suppl 2: 158-160 (2004).
  46. Roysam B, Shain W, Ascoli GA: The central role of neuroinformatics in the National Academy of Engineering’s grandest challenge: reverse engineer the brain. Neuroinformatics, 7: 1-5 (2009).
  47. Amariglio N, Hirshberg A, Scheithauer BW, Cohen Y, Loewenthal R, Trakhtenbrot L et al.: Donor-derived brain tumor following neural stem cell transplantation in an ataxia telangiectasia patient. PLoS Med, ­6: e1000029 (2009).