Main

Stem cells are uniquely defined by their ability to self-renew and differentiate into mature cell types. By definition, stem cells must retain these two properties simultaneously and must make cell fate decisions in a highly controlled manner to accomplish the appropriate level of tissue and organ regeneration for homeostasis, repair and growth1,2. But even though stem cells are defined as single cellular entities, they are rarely analyzed clonally.

Clonal analysis is of particular importance as cells with stem cell properties are not homogenous. Self renewal of human pluripotent stem cells (PSCs) maintained in culture in vitro, for example, depends on specific subpopulations of cells in the culture3,4,5,6. Several stem cell types in vivo, including hematopoietic and neural stem cells, are also functionally heterogeneous, and this is also the case in cancer, as it has been demonstrated that cancer-initiating cells in the same malignancy can have considerable functional and genetic diversity7,8,9,10. Population-based analyses of stem cell behavior often fall short of defining mechanisms that may be unique to these specialized cells. What is more, stem cells do not function in a completely cell-autonomous fashion but reside in and react to complex niches both in vitro and in vivo11,12. Methods to measure and track stem cells as single cells are experimentally complex but are essential for improving our understanding of the molecular and cellular mechanisms that govern stem cell fate decisions. Here we will discuss clonal assays using as examples human PSCs and hematopoetic stem cells (HSCs).

Clonal analysis of stem cells

In contrast to other cell types, stem cells are operationally or functionally defined13. This includes their ability to self-renew but also encompasses their developmental potential, that is, which cell types they can differentiate into. Understanding heterogeneity in populations of stem cells therefore requires assessment of these functional properties in single cells. Methodologically, such single-cell (or clonal) analyses can be broadly divided into two categories: analysis of physically separated single cells versus in situ analysis of the properties of single cells, which retains the in vitro or in vivo microenvironment (Fig. 1).

Figure 1: Studying heterogeneity within stem cell populations.
figure 1

Stem cells are heterogeneous populations that normally reside in complex environments. Individual cells can be physically isolated or marked genetically in situ to allow tracking of individual cell behavior. Single cells are assayed for stem cell properties with in vitro or in vivo assays, permitting retroactive stem cell classification.

Full size image

Human pluripotent stem cells. Physically isolated single human PSCs do not survive well in vitro during serial passage, in contrast to mouse embryonic stem cells (ESCs). Accordingly, researchers in many laboratories have attempted to modify the culture conditions to increase human PSC survival after dispersal into single cells during passage. These modifications include modulation of the Rac-Rho5 or IGF-AKT signaling pathways4, or provision of suitable cellular microenvironments such as mouse embryonic fibroblasts or human ESC-derived fibroblasts3. Using such approaches, it has been possible to determine the frequency of stem cells capable of clonogenic growth (the ability to regenerate cultures starting from a single cell) as a functional measure of stem cell self-renewal at the single-cell level14. However, because culture conditions conducive to self-renewal are necessary for these experiments, the differentiation potential of the same single human PSCs—which can, in principle, be measured through in vitro differentiation to specific lineages but requires different culture conditions—cannot be determined simultaneously.

Human PSCs can also be evaluated in vivo, by transplantation into immunodeficient mice to detect their potential contribution to multilineage teratomas composed of endoderm, ectoderm and mesoderm tissue15,16,17, but it is exceptionally difficult to quantitatively determine stem cell contribution to the different lineages in the absence of tracking methods18,19. As such, teratoma assays can only indicate that the collection of cells transplanted have pluripotent capacity. Single-cell transplantation of human PSCs into mice to form multilineage teratomas has yet to be reported, likely because of poor survival of transplanted single cells or the inability of human PSCs to operate in a cell-autonomous fashion in vivo. Notably, single mouse PSCs have been functionally assessed using tetraploid complementation20,21, but other cell types in the blastocyst are required to support the development of the single stem cells in these assays. Taken together, physical isolation methods are limited at present for a complete description of single human PSCs, and in situ labeling methods are required.

In situ clonal analysis usually involves genetic modification of the target cells to be analyzed, either by introduction of a nonspecific element such as a retrovirus that randomly inserts into the stem cell genome or, as in more classical lineage tracing, by fluorescent cell marking via site-specific recombination. In the case of retroviral marking, the specific viral integration site provides a unique 'tag' for each cell, which can then be used to track the resulting cell clone (the progeny of the original marked cell) after the culture has been assayed for stem cell self-renewal and differentiation in vitro or in vivo (Fig. 1). For example, we have used this approach in human ESC cultures and observed that cells with otherwise identical phenotypes have unique properties of self-renewal and differentiation when interrogated using standard assays19. Furthermore, clonal tracking revealed that the assays used to measure human PSC differentiation in vitro (embryoid body formation) or in vivo (teratoma formation) detected different populations of stem cells19. We note that, for this approach to work, the retrovirus must insert at a single site per cell, or it must be possible to identify cases of multiple integrations per cell. We discuss technical caveats of in situ tracking methods below.

Hematopoetic stem cells. For HSCs and other somatic stem cells that are difficult to maintain in culture, in vivo methods are essential to measure developmental potential. What is more, this approach permits the study of clonal behavior within the natural complexity of the in vivo environment, under the myriad extracellular cues that are difficult if not impossible to mimic in vitro, and also permits the study of cell movement in the whole organism. Clonal tracking has arguably been exploited most exhaustively in the hematopoietic system in part owing to the ease of sampling this liquid tissue over time and with minimal invasiveness. It involves either the in vivo tracking of clones of cells with unique genetic tags or the interrogation of the output of physically isolated single HSCs after transplant.

Elegant in situ tracking experiments first initiated in the late 1960s used whole-animal radiation treatments to induce hematopoietic chromosomal abnormalities in vivo22. Although the unique radiation-induced genetic mark initiated in a given precursor was inherited by all of its descendants whose number, location and phenotype could then be dynamically sampled, this technique was plagued by the possibility that cell function was perturbed by the very radiation needed to induce the clonal mark. Using the principles exemplified by these early methods, however, several groups have developed similar approaches, this time exploiting the naturally occurring mutations, rearrangements and copy-number variations in human leukemias to effectively map clonal diversity, evolution and origin in the disease setting9,10,23,24. As for human PSCs, replication-incompetent retroviruses are also used as a standard method to genetically mark HSCs. This method has important advantages in that retroviruses can be engineered to infect a wide range of cell types with high efficiency, whereas the semi-random insertions are thought for the most part to have minimal effects on cellular behavior25. Using retroviral marking followed by transplantation of cell pools, it has been confirmed that a single cell, marked by a unique proviral insertion site, could contribute to all hematopoietic lineages, providing the first conclusive evidence for a multipotent, long-lived mouse HSC26,27.

To ensure the efficiency of single-cell HSC transplants into irradiated recipients, an important requirement is the simultaneous introduction of a progenitor-enriched yet stem cell–depleted 'carrier' population that is only capable of, and serves primarily to, support a short-lived first wave of hematopoiesis critical in the early rescue of hematopoiesis that prevents the death of the mouse. This population may also have additional roles in support of single-cell homing to and retention in the bone marrow microenvironment28. An additional mouse model (W41 /W41 ), that as a result of defective c-kit kinase provides a competitive advantage to transplanted congenic cells, has also been used as a sensitive recipient of single cells7,29. The advantage to this strain is that it requires only light irradiation and therefore transplantation of helper populations is unnecessary for survival. In either transplant setting, evidence of long-term donor-mediated regeneration of all hematopoietic lineages is indicative that the single donor cell transplanted was an HSC.

Because these assays are technically challenging and require lengthy readout times, much effort has been put into the identification of markers that can be used to prospectively identify functional HSCs. Indeed, single-cell transplants to characterize the behavior of individual HSCs would be futile if there was no method to a priori enrich for these cells. This has culminated in the use of combinations of surface markers that can highly enrich for HSC content, but where at best 50% of the identified subpopulation does not display full stem cell activity30. The current inability to purify HSCs to homogeneity could simply be a reflection of not yet having identified the essential surrogate marker(s). In contrast, it is also possible that stochastic events such as chance exposure to different external cues or random fluctuations in intrinsic signaling could obscure an underlying stem cell identity31.

Despite these qualifications, the identification of faithful and specific markers for HSCs holds the promise of eventually bypassing functional assays. The ideal indicator(s) will be one that not only marks all functional stem cells regardless of the conditions under which they are grown (in vivo versus in vitro) but that is also absent from non-stem cells. Some surface-marker cocktails, for example, have been successfully used to specify HSCs from unmanipulated mice, but when applied to transplanted bone marrow or HSCs cultured for short periods in vitro these markers no longer identify the HSCs32,33. Thus caution should be exercised when using surrogate markers, as they may be relevant only in specific conditions.

Limitations of clonal analysis methods. Generally speaking, both tracking by physical separation and in situ clonal detection (Table 1) are plagued by the fact that, like all experimental methods, they perturb the system under study and may therefore interfere with the behavior of the stem cell in question. They may even select for single stem cells with specific properties that differ from other stem cells in the same tissue34. More specifically, for tracking of retrovirally marked cells, the sensitivity of detection of integration sites is an important issue. In the event of loss of a clone from a sample, it is essential to consider that the detection method has a threshold below which a clone will not be read out. For integration sites that are identified through restriction cutting and Southern hybridization, for instance, the limit of detection is a clone that comprises at least 5% of the overall population35. More recent PCR-based techniques provide much more sensitive means to liberate the unique retroviral insertion sites and have reported limits of detection of 4–10 copies of the integration site36. Several biological phenomena can also explain the apparent disappearance of a clone. This could be a reflection of death or quiescence but could also be due to deletions or translocations of the original retroviral insertion site, possibly as a result of clonal evolution. Newly developed technologies promise to provide ever more high-resolution descriptions of clonality. In one of the latest iterations of retrovirus-induced clonal marking, for example, Gerrits and colleagues have incorporated barcodes into the vector backbone such that each clone is denoted by a unique identifier that can then be resolved through sequence-based detection37. An important benefit of this technique is an increased sensitivity as it allows detection of individual clones that PCR-based approaches may miss. This method is also unbiased and quantitative, providing accurate information on the proportionality of clones. Because it does not rely on restriction sites, it also avoids the problem of differential amplification of fragments with varying lengths or melting temperatures37.

Table 1 Strategies for clonal tracking of stem cells
Full size table

Lineage tracing

A discussion of the clonal analysis of stem cells would be incomplete without a consideration of a lineage tracing in vivo. Lineage tracing techniques were classically developed to study organ or tissue neogenesis during development. With variations, such techniques typically involve the use of a ubiquitous or cell type–specific promoter-driven Cre recombinase to remove a LoxP-flanked stop codon interrupting expression of a gene encoding a marker such as beta galactosidase or a fluorescent protein on an adjacent chromosome38. Temporal control of recombinase activation can be introduced by engineering an inducible Cre recombinase, for instance by fusing it to the estrogen receptor, which allows control of its expression by the estrogen receptor ligand tamoxifen. Because the recombination event results in an irreversible genetic change, it effectively induces an 'ancestral mark' in the targeted parent cell that is then transmitted to all its progeny. As this approach is likely to mark several stem cells and possibly other cells at varying points along the lineage (depending on the promoter driving Cre), it is typically used to study cell populations with common expression profiles rather than the progeny of a single cell. In cases such as the skin and intestine, however, which have limited migration and cell mixing, low-level induction of recombination has permitted lineage tracing of individual stem and progenitor cells39,40.

Lineage tracing has yielded important insight into the origin of specific tissues or lineages. For example, by driving Cre expression with the insulin promoter, generally assumed to only be active in mature pancreatic beta cells, Dor et al.41 argued that the detection only of labeled new beta cells arising during adulthood or after pancreatic injury indicated that these cells are generated by pre-existing beta cells and not from multipotent stem or progenitor cells as had been previously proposed. In the hematopoietic system a similar controversy surrounding the origins of the first definitive mammalian HSCs was addressed by tracing cells in which recombinase activity was under the control of the Runx1 promoter, active early in embryonic life in the yolk sac and never in the embryo proper. With subsequent analysis, Samokhvalov and colleagues demonstrated that adult HSCs derived from hematopoietic progenitors in the yolk sac then migrate to seed the fetal liver and thymus42.

However, these methods depend heavily on the promoter driving Cre. Owing to the promiscuity of gene expression, a stem cell may in fact express a gene that is thought to mark a more differentiated cell type and that is not predictive of eventual cell fate choice. For example, in contradiction to the above-mentioned study, recent work has shown that multipotential pancreatic stem cells express the transcript for insulin, and show the potential to differentiate into and stably replenish insulin-producing pancreatic beta cells43. As such, results from studies using these approaches without additional independent evidence should be interpreted cautiously.

Careful titration of the dose of inducer in the Cre–estrogen receptor system can be used to more sporadically mark cells to study the developmental potential of single cells in vivo, but as the number of marked stem cell clones becomes limiting, many more animals must be analyzed. One potential solution to this has come with the advent of combinatorial marking of cells using an array of fluorescent proteins. The 'Brainbow' mouse relies on Cre-lox technology to induce in the mouse brain an essentially random combination of fluorescent protein expression resulting in brain cells marked by one of 89 different hues44. This combinatorial approach has opened the door to a new type of lineage tracing. Using mathematical modeling and capitalizing on the knowledge of the anatomic location of stem cells in the intestine, Snippert and colleagues have coupled intestinal stem cell–specific, inducible Cre recombinase expression with the multicolor Brainbow technology to track many single stem cell clones, each marked in a unique color45. From these observations, they proposed a model whereby intestinal stem cells, instead of dividing asymmetrically as was widely hypothesized to be the case, actually divide in a symmetric manner with tissue homeostasis being attained by neutral competition between the daughter stem cells. Despite the large array of possible colors using the Brainbow technology, these hues are randomly generated, and it is therefore possible to have two or more clones marked with the same color. This is not so problematic in more static tissues, but in very dynamic or fluid systems such as the hematopoietic system where clones can easily intermix, this technology is still limited in its usefulness for describing single stem cells and will remain so until modifications can be made that ensure a unique identifier is imparted to every marked cell. In this respect, retroviral marking remains the preferred clonal tracking methodology in such systems.

Molecular capture of functional stem cell behavior

A complete understanding of stem cell heterogeneity will require characterization of both molecular and functional properties of single cells. Molecular evaluation includes gene expression, proteome and epigenome analyses of single cells; some of these technologies are now becoming available46,47,48. But there remain obstacles to simultaneous analyses of molecular and functional properties in single cells. Molecular characterization requires destruction of the cell, precluding measurement of its biological potential and hence its definition as a stem cell. Experimental approaches to address this conundrum have varied. One approach that has met some success involves the analysis of mouse HSCs by a method termed 'sibling analysis'. Single quiescent mouse HSCs are prospectively isolated at the highest purity possible using surrogate markers. These cells, called 'clonal starts' are then placed in defined in vitro cultures that permit initial cell division. Of the resulting two siblings, one can be used for molecular investigation and the other for functional analysis to measure HSC properties49. Individual cells collected for molecular study whose sibling demonstrated HSC capacity can then be used as an indirect capture of the HSC molecular signature, for example, its gene expression profile. Of course, these analyses are still burdened by the uncertainty that each sibling, after initial cell division, may not be equal in capacity. It has been reported that, even when seeding clonal starts with cells isolated using highly enriching parameters for HSCs, the frequency of functional equality of siblings is in the 60% range49.

One of the most compelling reasons to study stem cell behavior is for their use in cell replacement therapies in regenerative medicine12,50. In these cases, stem cells would have to be transplanted in vivo, assemble and regenerate tissue, often in the context of damaged tissue, which may be an inhospitable microenvironment and which will most probably need to be understood better to define 'regenerative' stem cells50,51. In vivo imaging of stem cell behavior upon transplantation is likely to be a beneficial approach for this purpose52,53, and the combination of live-cell imaging and single-cell tracking in vitro54,55 will also continue to provide important insights. As our understanding of molecular signatures of stem cells increases, in vivo imaging could incorporate molecular tracers that report on self-renewal or differentiation to study the identity and behavior of a stem cell as it functions.