Absence of Evidence Is Not Evidence of Absence
Pitfalls of Cre Knock-Ins in the c-Kit Locus
c-Kit+ Cells Minimally Contribute Cardiomyocytes to the Heart van Berlo et al Nature. 2014;509:337–341.
A recent article by Jeffrey Molkentin’s group using mice with Cre/lox knockin in the c-kit locus concludes that, contrary to other reports, c-kit+ cells minimally contribute cardiomyocytes to the heart. We think that the interpretation of the results of the author is problematic because critical data to support their conclusions have not been provided.
Recently, van Berlo et al1 published an article in Nature entitled c-kit+ Cells Minimally Contribute Cardiomyocytes to the Heart, in which they assessed the contribution of c-kit–expressing cell lineages to the formation of cardiomyocytes and microvasculature in the myocardium during development, aging, and after myocardial infarction. The authors knocked in the c-kit locus, a constitutive Cre-IRES-nGFP (cyclic recombination enzyme linked to an internal ribosome entry site driving a nuclear green fluorescent protein), in one line of mice and a tamoxifen-inducible mER (membrane estrogen receptor)-Cre-mER in another line. Both gene knockins placed the CRE sequence in frame with the ATG start codon in exon 1 of the Kit gene. These mice were cross-bred with the ROSA26 reporter (R26R) mutated mice carrying a floxed (and Cre dependent) reporter gene (either green fluorescent protein or mT/mG construct). On the basis of a limited set of experiments, the authors concluded that although cardiac c-kit+ cells contribute cardiomyocytes, they do so minimally (from ≈0.001% to ≈0.02%) and, therefore, it was considered to be functionally insignificant. By contrast, c-kit+ cells amply generated cardiac endothelial cells. From these data, the authors concluded that the use of c-kit+ cells in myocardial repair and regeneration protocols should be reconsidered.
Interestingly, this article comes out at a particularly propitious time to challenge the regenerative properties of the c-kit+ cardiac progenitor cells. During the past several months, a flurry of negative reports in the scientific and lay press has predicted the demise of stem cell–based regenerative medicine in general and myocardial regeneration/repair in particular. Although this new critical assessment might be justified for a field used to overblown and to uncritical claims about its potential, it has created confusion among the lay population at large and the research community in particular. Curiously, the results of van Berlo et al1 directly contradict 2 articles2,3 published 10 years apart and many other articles in between by us and other groups. Multiple experimental approaches, including cell fate tracking, genetic tagging, and cell transplantation, have shown that the adult heart harbors bona fide tissue-specific endogenous resident cardiac stem cells (eCSCs) that among their surface receptors, express the stem cell factor tyrosine kinase receptor, c-kit. These cells are multipotent, clonogenic, and self-renewing. The progeny of a single-cell forms multipotent spheres (cardiospheres) and differentiate in vitro and in vivo into cardiomyocytes, vascular smooth muscle, and endothelial cells. Furthermore, the adult myocardium is dependent on the regenerative capacity of the eCSCs to repair diffuse myocardial damage anatomically and functionally.3 If the eCSC response to this diffuse damage is ablated, the spontaneous endogenous heart repair is blunted with animals going into heart failure. When functional eCSCs are repopulated through exogenous administration of clonogenic eCSCs, anatomic damage and functional cardiac impairment are completely and permanently reverted. If these regenerated cells are selectively killed, the recovery is reversed and the animals return to heart failure with a significant increase in cardiac death. Taken together, these data demonstrate that c-kit–expressing eCSCs are necessary and sufficient for myocardial regeneration and repair in the adult heart.3
In short, although both groups agree that myocardial c-kit+ cells can generate cardiomyocytes and vascular cells in the adult, the role assigned to the c-kit+ cells in adult cardiac homeostasis and repair is diametrically opposed. Although the conclusions of each group are supported by apparently solid experimental data, the conclusions reached are incompatible; both cannot be correct.
In science, positive results, unless their methodology and interpretation cannot challenged, usually trump negative ones. These 2 conflicting reports, in addition to technical and methodological differences and questions arising about the generation of the data, also pose important questions about myocardial biology and the prospects of developing effective regeneration/repair protocols: Is the adult myocardium a self-renewing tissue? If so, what is the physiological significance of the self-renewal? Are the c-kit+ cells the regenerative/repair agents? If not, which are these regenerative agents? Clarification of the reason(s) for the reported discrepancies is of high priority, not only for cardiac biology but also for the future of the field of myocardial repair and regeneration.
Discrepant and even contradictory results on a given topic are nothing new in science, even when the experiments are properly planned and the results are accurately reported. In most cases, the discrepancies are often because of a misinterpretation of the power of the experimental system(s) used, which often leads to the over- and misinterpretation of the results. The latter seems to be the main issue with results of the article by van Berlo et al.1
Not All Cre/LoxP Cell-Tracking Systems Provide Reliable Answers
One perennial conundrum in cell biology has been the search for an efficient and accurate method of tracking the cell lineage origin, as well as the fate of a given cell type, both during development and in adulthood. The development of the Cre/LoxP system of constitutive and conditional gene targeting together with trans- or knocked in genes seemed the ideal procedure for fate mapping of specific cells. Inducible or regulated Cre-expressing mouse lines are increasingly popular because they significantly improve temporal and spatial targeting of genes.4 The most widely used inducible Cre lines use a Cre fused to a mutant ligand-binding domain of the estrogen receptor (Cre-ER, mER-Cre-mER, or Cre-ERT2), where Cre-ER, mER-Cre-mER, or Cre-ERT2 activation depends on the selective estrogen receptor modulator tamoxifen.4
Although the constitutive and conditional Cre/loxP systems are both extremely powerful, they have several drawbacks that need to be properly controlled to avoid erroneous interpretation of observed results.4 As the work of van Berlo et al1 is exclusively based on the efficiency and fidelity of the Cre knockin in the c-kit locus to track c-kit+ cell lineage fate, one must consider the main shortcomings of this recombination system that are summarized in Table 1.
Therefore, considering these shortcomings, the conclusions reached with the Cre/lox system are only as reliable as the specific controls performed to ensure that the system performance is appropriate to answer the questions addressed by the experiment unambiguously. Table 2 summarizes the 3 controls that are particularly important. Unless these conditions are fulfilled, the results of a Cre/lox experiment are uninterpretable. This is even more so when the results are negative because there is low or absent identifiable marked progeny of the target cells, as is the case of the results of van Berlo et al.1 These caveats are particularly important in light of the complex and yet not fully elucidated c-kit gene regulation.6 In fact, all the c-kit transgenic or knockins produced to track c-kit–expressing cells are far from closely recapitulating c-kit expression and function either in development or in adult life.6,7
Does the c-kit/Cre Knockin Recombine the Reporter Gene in the c-kit+ eCSCs?
It is evident that both the constitutive and the regulated Cre c-kit gene mutated mice lines used by van Berlo et al1 are c-kit hypomorphs, most likely hemizygous null, which are equivalent to the W/+ mutant mice.8 Although the typical signs of a c-kit hemizygous null allele (white spots on the belly, paws, and tip of the tail) are not clearly reported in the article, this conclusion is confirmed by the fact that the homozygous animals are not viable at birth and mostly die in utero. Whether the level of Cre expression from the mutated allele is sufficient to produce recombination in all or most of the c-kit–expressing cells is unknown, except that there is a recombination in cells/tissues with high endogenous c-kit expression.
The authors provided data showing recombination in several cell types, including myocardial cells, but not in the cells relevant to the results (ie, the c-kit+ eCSCs). The reported recombination in total c-kit+ bone marrow cells and in myocyte-depleted c-kit+ cardiac cells is not relevant to the efficacy of the Cre/lox set-up because it is not specified which are the c-kit+ cell types recombined and which are not. In the constitutive Cre-IRES-nGFPxR26R-eGFP mice, if the recombination was spatially and temporally correct, the majority (if not all) of blood cells in the adult should be labeled because c-kit is expressed in all the embryonic cells with hematopoietic activity. Yet this piece of evidence is not provided. Furthermore, bone marrow and myocardial c-kit+ cells are complex cell mixtures with <10% representing the hematopoietic stem cells (HSCs) and eCSCs, respectively.9–11 Labeling even the majority of bone marrow c-kit+ cells might not be sufficient to follow blood cell lineage origin and specification. For that, it is necessary to tag the long-term repopulating Sca-1+/c-kit+ cells. The same applies to the myocardium where the most abundant c-kit+ cells are not the eCSCs but mast and endothelial cells and their precursors.10 If cardiac mast cells (c-kit+CD45+) or endothelial progenitor cells (c-kit+CD34+/CD31+) are labeled, then only their differentiation into inflammatory and endothelial cells or their fusion can be quantitatively assessed (which indeed van Berlo et al1 appropriately measured) but not the fate of the c-kit+ eCSCs (c-kit+CD45−CD34−CD31−). Unfortunately, no data are presented about the recombination efficiency in cells with intermediate/low c-kit expression, such as HSCs and primordial germ cells. Remarkably, even less is known about the recombination rate in the c-kit+ eCSCs because the authors did not isolate, quantify, or in any way analyze these cells. In fact, the recombination efficiency in the c-kit+ eCSCs is an indispensable piece of information and the main driver for the conclusions derived from the results are reported but it is never specifically assessed in the article itself or in the additional information. The authors cultured total nonmyocytes from the hearts of young adult constitutive Kit+/Cre//R26R-eGFP mice in the presence of dexamethasone showing that eGFP+, Kit-Cre allele-expressing cardiac cells induce expression of the cardiac markers GATA4, α-actinin and troponin T (yet fail to differentiate into functionally competent cardiomyocytes). However, the latter is not evidence that eGFP+ Kit-Cre+ cells are eCSCs with myogenic progenitor properties, as erroneously suggested by the authors. Indeed, true eCSCs fully differentiate into beating myocytes in vitro and in vivo.2,3,11 Moreover, the fact that eGFP+ Kit-Cre+ cells in vitro can switch on contractile protein genes is in line with the report that endothelial cells can acquire cardiomyogenic phenotypes in vitro and in vivo.12 To substantiate that the cells analyzed in vitro were true cardiac stem/progenitor cells, the authors should have rigorously tested the eGFP+c-kit-Cre+ cells for prototypical stem cell properties, including clonogenicity, multipotency, self-renewal, and in vivo regenerative potential.
Therefore, the first question that should have been, but was not answered, is whether in either the constitutive or the conditional mice was there enough Cre produced to recombine a high/sufficient percentage of the c-kit+ eCSCs to track their fate. Without knowing the actual number/fraction of c-kit+ eCSCs labeled by the mouse genetic strategy, it is impossible to measure their contribution to the formation of cardiomyocytes or to any other cell type quantitatively.
The significant weakness of the Cre (and mER-Cre-mER in particular) knockin in the c-kit gene and its failure to faithfully recapitulate the temporal and tissue-specific pattern of c-kit gene expression have been elegantly documented in 2 previous articles where a similar (actually better as a Cre-ERT2 instead of a mER-Cre-mER construct was knocked in) conditional Cre knockin within the c-kit gene was used.13,14 Remarkably, using similar strategies as van Berlo et al1 for Cre-dependent recombination by tamoxifen, this mouse shows that this protocol recombines <3% of the Sca-1+/c-kit+ HSCs. Even the efficiency of recombination in mast cells, despite expressing c-kit (and Cre) at many fold higher level than the HSCs,14 shows heterogeneous rates of recombination from tissue to tissue. Thus, this mouse model, and by extension all those that depend on the strength of the c-kit promoter to drive Cre, is not adequate for tracking the cell fates of adult multipotent stem and progenitors cells expressing intermediate or low levels of c-kit and where the indispensable function of the receptor allows only hemizygosis of the knocked in Cre allele. Furthermore, in the articles by Klein et al13 and Heger et al,14 the undisputable dependence of recombination rate on the level of Cre expression was used to ablate the mast cells and their precursors selectively (or intestinal Cajal cells) without affecting other c-kit–expressing cells (low/intermediate), including the HSCs. These results further emphasize the need to demonstrate a substantial and reproducible level of recombination in the target cells (in this case the eCSCs) in response to the activating protocol used.
c-kit+ eCSCs, by definition, express c-kit, which is used for their isolation. However, the level of c-kit expression by the c-kit+ eCSCs in several strains of mice tested is significantly lower than in the HSCs (Torella et al, unpublished data, 2014). As the c-kit CreT2 knockin (with its higher affinity for tamoxifen and higher recombination efficiency) recombines <10% of the HSCs (even after a longer pulse with tamoxifen-added chow),14 it is likely that an even lower fraction, if any, of c-kit+ eCSCs are recombined in the Cre/lox strategy of Molkentin’s group. Independent of the correctness of the above conclusion, it is indisputable that the efficiency of recombination in the c-kit+ eCSCs, which is the main subject of the article, should have been determined before attempting to evaluate their contribution to the myocardium using a model that is completely dependent on the efficiency of recombination to turn on the expression of the reporter gene. Without these data, their results are inconclusive and qualitative at best.
The authors in the supplementary discussion section reported “as a minor concern that the levels of new cardiomyocyte formation from the c-kit+ lineage reported …. may be under representative due to replacement of one functional Kit allele. Thus, there could be less c-kit+ ‘progenitor-like’ activity in the hearts of these mice due to only a 50% dosage of c-kit protein.” Also, they acknowledged, “another potential issue is if c-kit+ cells present in the heart express the Kit locus (and Cre) at levels below the threshold of the Cre recombinase-based system. If this were the case it would again under-represent the total number of potential cardiomyocytes as being labeled from the c-kit+ lineage. However, Cre-based lineage tracing is a widely accepted standard for relevant gene expression, so the opposite argument could easily be made that if expression is below the Cre-threshold, it may not be physiologically meaningful Kit allele expression, and hence these cells are really not c-kit”. Clearly, these caveats should have been experimentally addressed particularly given the significance of the data discussed above and the fact that many bona fide c-kit+ cells do not recombine.13,14
There are 2 additional issues that might also affect the interpretation of the results of van Berlo et al.1 First, the inner cell mass cells of the blastocyst (ie, the embryonic stem cells) express intermediate but significant levels of c-kit.7,15,16 The level of c-kit expression is even higher in the primordial germ cells.7,16 Therefore, if the Cre knockins in the c-kit locus had recapitulated normal temporal and spatial expression of the gene, as it is assumed in the article, all or most of the cells in the constitutive Cre expression mice should have been labeled along with their progeny. The fact that the constitutive knock-in does not recombine the reporter gene in all the descendants of the ICM (the whole embryo) should have alerted the authors that something was amiss. In addition, van Berlo et al should have tested whether the recombined mice produced labelled progeny, as it would be expected if the knock-in mirrors the expression of the c-kit gene in the primordial germ cells. In addition, although it is still unknown whether myocardial development and myocytes in particular are dependent on an embryonic cardiac progenitor expressing c-kit, the absence of a significant recombination in the embryonic stem cells by the Cre-IRES-nGFP knockin directly questions the quantitation of labeled cardiomyocytes at birth and at 4 to 12 postnatal weeks in the constitutive c-kitCre/+//R26R-eGFP mice in the article by Van Berlo et al.1 Second, the authors overlooked the potential effect of producing a c-kit null allele in the targeted cells because targeting Cre to c-kit exon 1 alters normal c-kit receptor activity.16 This fact was used to identify which cells depend on physiological c-kit activity for their proper function, as is the case for most stem cells.14 Therefore, it remains to be determined whether the null allele has altered the growth and differentiation properties of the c-kit+ eCSCs (the W mutant mice have a cardiac phenotype)8 in response to myocardial damage and these cells have been outcompeted by others not expressing the Cre knockin (c-kit negative eCSCs).
In conclusion, although the data presented by van Berlo et al1 have confirmed that c-kit+ cells fulfill the criteria of true cardiac stem/progenitor cells in the development and adult life because some gave rise to cardiomyocytes, smooth muscle, and endothelial cells, we think that until the significant issues left unanswered by the report—particularly the efficiency of Cre recombinase in the models used—are satisfactorily addressed, their main conclusion that c-kit+ cells minimally contribute cardiomyocytes to the heart should be taken with extreme caution because it has not been tested by the experiments performed and certainly not proven by the data presented. Furthermore, it must be underlined that the used cre-lox systems by their nature, when working properly, would label any cell expressing c-kit at any time during development and postnatal life or during the tamoxifen treatment. Therefore, this system cannot distinguish between a single-cell population with multipotent differentiation potential from several unipotent ones. This can be only be ascertained by single-cell cloning and differentiation of the putative stem/progenitor cell in vitro and its transplantation in vivo. The multipotent and regenerative properties of c-kit+ eCSCs have been reproducibly and robustly ascertained,2,3 and these results are not challenged by the results of van Berlo et al.1 Accordingly, whether the autologous and allogenic use of cardiac c-kit+ CSCs in clinical protocols for cardiac repair is or it is not appropriate cannot be determined from the data presented by van Berlo et al.1 What is needed to make this determination are additional experimental data, using genetically labeled eCSCs in clinically relevant animal models, which allow determining the fate of the transplanted eCSCs, their mode of action (direct participation or paracrine action), and quantification of their participation in the putative regenerative process. Most of these data have already been obtained. Therefore, the proposal for a moratorium on the ongoing phase I/II clinical trials with c-kit+ eCSCs is premature.
Sources of Funding
This work was supported by grants from CARE-MI FP7 (Health-F5-2010-242038), Endostem FP7 (Health F5-2010-241440) large scale collaborative projects, FIRB-Futuro-in-Ricerca (RBFR081CCS, RBFR12I3KA), and Italian Ministry of Health (GR-2008-1142673, GR-2010-2318945).
The opinions expressed in this Commentary are not necessarily those of the editors or of the American Heart Association.
Commentaries serve as a forum in which experts highlight and discuss articles (published here and elsewhere) that the editors of Circulation Research feel are of particular significance to cardiovascular medicine.
Commentaries are edited by Aruni Bhatnagar and Ali J. Marian.
- © 2014 American Heart Association, Inc.
- Berrozpe G,
- Agosti V,
- Tucker C,
- Blanpain C,
- Manova K,
- Besmer P
- Cairns LA,
- Moroni E,
- Levantini E,
- Giorgetti A,
- Klinger FG,
- Ronzoni S,
- Tatangelo L,
- Tiveron C,
- De Felici M,
- Dolci S,
- Magli MC,
- Giglioni B,
- Ottolenghi S
- Kohli L,
- Passegué E
- Condorelli G,
- Borello U,
- De Angelis L,
- et al
- Klein S,
- Seidler B,
- Kettenberger A,
- Sibaev A,
- Rohn M,
- Feil R,
- Allescher HD,
- Vanderwinden JM,
- Hofmann F,
- Schemann M,
- Rad R,
- Storr MA,
- Schmid RM,
- Schneider G,
- Saur D
- Bernex F,
- De Sepulveda P,
- Kress C,
- Elbaz C,
- Delouis C,
- Panthier JJ