Cancer Stem Cells, Endothelial Progenitors, and Mesenchymal Stem Cells: “Seed and Soil” Theory Revisited

Abstract

Isolation of putative cancer stem cells (CSCs) in various tumors has generated much excitement among researchers who consider these cells the potential “culprits” behind resistance to conventional therapy. Both cancer and cardiovascular disease are believed to be stem cell disorders involving circulating endothelial progenitors (CEPs) and mesenchymal stem cells (MSCs). CD133 and CD44, markers of CSCs in many tumors, also enrich CEPs and MSCs, respectively. We propose an integrated tumorigenesis model that involves all three interdependent stem cell (CSC, CEP, MSC) compartments by revisiting the “seed and soil” model. Developing therapeutics that can effectively target CSCs and spare normal cardiovascular tissue will remain a challenge. Preliminary laboratory and clinical data on monitoring and targeting colon CSCs, using such a modeling system, are discussed.

More than a century ago, Virchow and Paget indirectly implied the existence of cancer stem cells (CSCs)¹ and a CSC niche,² respectively. The term “cancer stem cells” was coined in the early 1980s,³ but research in the field was hampered by the lack of specific markers, cell-sorting technology, and the extremely low frequency of CSCs. The first isolation of putative CSCs in acute myelogenous leukemia (AML) by Dick and colleagues in 1994⁴ spurred a wave of discoveries of CSCs in acute lymphocytic leukemia (ALL),^5,6 chronic myelogenous leukemia (CML),^7,8 multiple myeloma,⁹ and in cancers of the colon,^10–12 breast,¹³ prostate,^14–16 brain,^17,18 head and neck,¹⁹ retina,²⁰ lung,^21,22 pancreas,²³ melanoma,^24,25 kidney,²⁶ and liver.^27,28

Researchers postulated that CSCs might be the “culprits” capable of evading conventional therapy; thus, eradication of CSCs could lead to complete cure.^29,30 However, CSCs are quiescent or slow cycling,^31,32 they overexpress antiapoptotic proteins,^24,25,33 possess multidrug resistance proteins,^24,25,34,35 and are characteristically similar to normal stem cells, except that their ability to differentiate is impaired.^4–29

In 1997, Asahara et al first identified bone-marrow derived circulating endothelial progenitors (CEPs) vital in postnatal physiologic and pathologic angiogenesis.³⁶ In the same year, Yin et al discovered CD133, a member of the prominin family that identifies and enriches CEPs.³⁷ Two large prospective studies showed that decreased CEP levels independently predict increased cardiovascular death and correlate with all known cardiovascular risk factors.^38,39 Conversely, elevated CEP levels were found in patients with leukemia,⁴⁰ myeloma,^41,42 myelodysplastic syndrome,⁴³ cancers of the lung,⁴⁴ breast,^45,46 liver,⁴⁷ colon and rectum,^48–50 prostate,⁴⁹ kidney,²⁶ and in infantile hemangioma.⁵¹ Furthermore, the presence of elevated levels of CEPs or CD133 mRNA is predictive of poor outcomes and death in certain tumors.^{44,47–49,52}

Circulating endothelial progenitors are known to form the premetastatic niche essential in the earliest step to tumor angiogenesis,⁵³ and it has been shown that targeting CEPs inhibits the development of macrometastases from micrometastases in mouse lung metastasis models.⁵⁴ Moreover, it has been demonstrated that bone marrow-derived stem cells can transdifferentiate into endothelial precursor cells with the potential to acquire mutations and contribute “cancerous” microvascular endothelial cells to tumor vessels.⁵⁵ In 1997, Zohar and others identified mesenchymal stem cells (MSCs) using CD44, a glycoprotein ligand that binds with hyaluronate and E-selectin.^56,57 Variant forms of CD44 are expressed in many human cancers and correlate with cancer outcomes in certain tumors.^58–67 Interestingly, CD133 and CD44 have both been used to isolate CSCs in many tumors.^4–25 The purpose of this review is to integrate these complex, confusing data on cancer stem cells, circulating endothelial progenitors, and mesenchymal stem cells and to discuss strategies directed at these novel cellular targets by revisiting Paget’s “seed and soil” theory² in the context of today’s understanding of CSCs, CEPs, and MSCs.

CANCER STEM CELLS: “THE SEED”

Normal stem cells are defined by two properties; (1) self renewal, and (2) the capacity to differentiate into a wide array of specialized cells in the body. To date, stem cells have been identified in hematopoietic tissue,⁶⁸ lung,⁶⁹ skin,⁷⁰ and colon,⁷¹ and in less regenerative tissues, such as those of the heart,⁷² and central nervous system.⁷³ Epithelial-mesenchymal transition (EMT) dictates the fate of morphologic and functional diversities among these tissues and are regulated by common signaling pathways, such as Wnt/β-catenin, Notch, and bone morphogenetic protein (BMP) for stem cell renewal, maintenance, activation, and differentiation.⁷⁴

CSC Research

Cancer stem cell research has advanced considerably since Virchow first described the similarities between neoplasia and embryonic tissues in 1885.¹ An elegant summary of CSC research authored by Huntly and Gilliland was published in 2005.⁷⁵ Cancer stem cells are rare, with plating efficiency ranging from 1/1,000–5,000 in solid tumors,³ to 1/1,000,000 in leukemia.⁴ Identification of stem cell markers CD133 and CD44 and the use of non-obese diabetic mice with severe combined immunodeficiency disease (NOD/SCID mice) xenografts led to the successful isolation of CSCs in many tumors.^8–24

Leukemia stem cells, characterized by CD34+ and CD38 low/-, comprised 0.1% of the total leukemia population and were capable of inducing all subtypes of AML observed in patients, except for the M3 subtype.⁴ BCR-ABL fusion gene can only be found in the leukemia initiating fraction of ALL and CML in blast crisis with similar phenotypes of CD34+/CD38-.^6–8 Singh et al first isolated CD133+ CSCs from patients with glioblastoma.^17,76 Later, two groups independently identified colon CSCs estimated at a frequency of one per 5.7x10⁴ colon cancer cells.^10,11 One out of 262 CD133-enriched tumor cells contains a colon cancer stem cell, and as few as 100–500 of these enriched cells with either CD133+ or CD44+ tumor cells could recapitulate human tumors in NOD-SCID mice.^{4,9–13,16,17} Tumor spheres generated from CD133+ cells contain self-renewal units that could be passaged in vitro over 1 year.^11,13,14,76 G-protein coupled receptor GPR) 49, a new colon stem cell marker, may also identify colon CSCs.⁷¹ It is important to point out that Ricci-Vitiani reported 2% of CEP cells in the enriched CD133+ colon CSC fraction only using CD45 and CD31 staining. CD133 enriched hematopoietic stem cells express many “stemness” genes that are also mutated or abnormally expressed in leukemia.⁷⁷

The CSC model is being challenged in a report by Kelly and Strauss, who showed that low efficiency tumor grafting may be due to poor grafting between humanmouse interface, as 1/10 mouse leukemia and lymphoma cells can establish xenograft. ⁷⁸ Table 1 summarizes estimated CSC frequency, their surface markers, and major signaling pathways.

Table 1.

Stem cells and their respective markers and pathways

Sites	Frequency in cancer	Surface markers	Target pathways
Colon 10–12	0.7–24.5%	CD133, CD44, GRP49	Wnt, Sonic Hedgehog
Liver27,119,120	<2%	CD133, G	Akt/PKB, BCL-2
Pancreas23	0.2–0.8%	CD44/CD24/ESA/CD133	Wnt?
Kidney26	0.9±0.17%	CD133
Prostate14,15,121	1%	CD133/CD44/α2β1,Sca-1,CXCR2	PTEN/AKT, SDF1/CXCR4
Melanoma24,25	<1%	CD133/ABCG2	Notch4
Lung & Head/neck19,21	10%	CD133	?
Brain17,18,74,122	1–27.5%	CD133	Notch
Myeloid (AML)4–6	0.2%	CD34+/CD38−	Fas, Wnt
Breast13	2–4%	CD44+/CD24−/ESA+	Wnt, Oct4

Niche And Tumor Angiogenesis: Roles of MSCs and CEPs

The differences between normal stem cells and CSCs are also determined by their dependence on the stem cell niche, a specialized microenvironment where stem cells reside, proliferate, and differentiate.^{79– 81} The niche maintains stem cell homeostasis, serves as a shield from tumorigenesis, and provides regulatory signals for tissue regeneration.^79–81 Niche normally generates dominant signals that inhibit stem cell differentiation and restrict normal stem cells from overproliferation for self renewal and multilineage commitment.⁷⁰ Disruption of the niche signal can lead to either loss of stem cells or cancer (Figure 1).⁸¹ One of the best studied models is the intestinal stem cell niche characterized by high turnover rate and homeostasis through intestinal crypts.⁷⁹ However, stem cell and niche is a potent combination, and a single stem cell is able to regenerate an entire crypt or generate additional crypts through crypt fission following radiation damage^82–84 and can also convert into cancer initiation cells.^84,85

Figure 1.

Normal stem cell homeostasis and revised “Seed and Soil” model involving CSCs, CEPs, and MSCs.

In contrast to the belief that endothelial cells of tumor vessels are considered normal diploid cells that do not acquire mutations, a recent study showed that 15% to 85% of microvascular endothelial cells in B-cell lymphomas harbored lymphoma-specific chromosomal translocations with numerically similar chromosomal aberrations, supporting the idea that change in the niche signal also contributes to tumorigenesis by providing stimulatory signals.⁵⁵

Tumor angiogenesis, one of six essential hallmarks of cancer, is a dysregulated modeling process that involves branching of pre-existing vessels, as well as mobilization of CEPs as a result of tumor hypoxia and excess proangiogenic factors.^86–90 CEPs are recruited to form the premetastatic niche conducive to the earliest steps of tumor angiogenesis.^53,54,90 Angiogenesis from CEPs differs from the branching out of pre-existing blood vessels in that CEPs possess higher proliferative capacity than activated endothelial cells capable of expanding a thousandfold in vitro.⁹¹ Compared to as high as 90%CEPs found in xenograft models, CEPs range from 1% to 12% in human cancers, depending on the surface markers.⁹⁰ Given the proliferative and differentiation potential of CEPs, which lose CD133 expression, enumerating CEPs based on CD133 expression in established tumors may underestimate the true CEP contribution to tumor angiogenesis. Bone marrow derived CEPs are mobilized by cytokines secreted by tumors and directed by chemokines to sites of hypoxia and neovascularization.⁹⁰ Stimulation of CEPs promotes tumor angiogenesis and lymphogenesis, while inhibition of CEPs retards tumor angiogenesis and tumor growth.^54,90,92,93

As discussed previously, CEP levels measured by flow cytometry (or its marker CD133 mRNA) were elevated in a variety of cancers and were associated with poor clinical outcomes in cancers of the lung,⁴⁴ breast,⁴⁶ liver,⁴⁷ and colon,⁵² and in solid tumor patients with bone metastases.⁴⁹ It is estimated that 1,400–700,000 circulating tumor cells exist based on the observation that there are 2–1,000 circulating tumor cells per 7 mL of blood.⁹⁴ If 1% of circulating colon cancer cells were CSCs, only 14–7,000 circulating CSCs would be present in a 70 kg patient (or ~ 5 liters of blood), supporting the clinical observation that most patients had only one metastasis to few metastases. Furthermore, any condition in which CSCs could commandeer and mobilize CEPs will also limit the number of metastases. Given its sensitivity, CD133 mRNA assay could have identified CEPs as well as CSCs if the latter are present in peripheral blood. We observed that elevated CD133 mRNA levels ≥ 4.79 predict colon cancer recurrence with an odds ratio of 22.2 (P = .02) compared to 17.2 for patients with stage III/IV vs. stage I/II disease (P = .01). Likewise, CEP levels were also found to correlate with stage in patients with invasive breast cancers⁴⁶ and are emerging as a useful surrogate marker of response in a variety of solid tumors.^{44,47–49,95}

Standard CD44 isoform (CD44s) identifies and enriches CD34-, CD133- MSCs that can differentiate into all stroma components, including fibroblast, bone, and cartilage. ^56,57,96 A variant form of CD44v is found in many epithelial cancers and is the ligand for E- and P-selection and hyaluronate. The prognosis values of CD44 and its soluble forms are controversial.^52–61,97 Recently, Liu and colleagues studied 186 gene expression profiles of CD44+/CD24-/low putative breast cancer stem cells to that of normal breast epithelium and found a significant association between the gene signature in both overall and metastasisfree survival (P < .001, for both) independent of established clinical and pathologic variables. The unique gene signatures could further stratify women with high-risk early breast cancer into good prognosis (10-year metastasis-free survival of 81%) and poor prognosis (10-year metastasis-free survival of 57%, P = .01) groups when combined with National Institutes of Health prognostic criteria. The gene signatures are also associated with prognosis in medulloblastoma (P = .004), lung cancer (P = .03), and prostate cancer (P = .01).⁹⁸

Given the dynamic plasticity and interdependency among all three stem cell compartments (CSC, MSC, and CEP), we hypothesize that “cancer stemness” functions in a complex “seed and soil” organizational hierarchy (Figure 1). Karnoub et al showed that mixing low potency tumor with MSCs dramatically improved the xenograft potential.⁹⁹ One alternative explanation to Kelly and Strauss’s findings would be that mouse leukemia and lymphoma progenitors are proficient in establishing their niche and angiogenesis given mouse origin and tumor types selected. Not all tumor types would need to fit in a CSC model, and a marked increase in tumor stem cell marker in the majority of tumor cells suggests that these are not necessarily stem cells but cells that had acquired stem cells or stem-cell-like characteristics.

TARGETING STEM CELLS IN CANCER THERAPY

Cancer stem cells possess many inherent in vitro resistance mechanisms;^{24,25,33–35} however, most clinically relevant is the observation that demonstrated host-tumor interaction conributes to the evolution of in vivo resistance to chemotherapy.¹⁰⁰ In vivo drug resistance occurs through EMT via MSC activation and angiogenesis through CEP mobilization by cytotoxic chemotherapy. ^92,101,102 Gemcitabine stimulates CD44 expression in pancreatic cancer cell lines,¹⁰³ whereas 5-fluorouracil (5-FU) or oxaliplatin leads to a sixteen- to thirtyfold increase in CD133 expression and a twofold increase in CD44 expression in colon cancer cell lines.¹⁰⁴

In addition to normalization of tumor vasculature,⁸⁷ bevacizumab, a VEGF inhibitor may also affect stem cell vascular niches and disrupt cancer stem cell maintenance, as mice bearing glioblastoma treated with bevacizumab showed depleted vasculature and a dramatic reduction in the number of glioblastoma stem cells without affecting their proliferation.¹⁰⁵

Given that CD44 and CD133 are coexpressed in three stem cell compartments, targeting CD44,¹⁰⁶ or CD133, though logical,^107,108 may lead to potential toxicities in heart, renal tubule,¹⁰⁹ and retina.¹¹⁰ Inhibitors of Wnt signaling,^111,112 Sonic Hedgehog,¹¹³ and AKT28 appears promising in certain tumor models. Use of genetically modified MSCs as a vehicle for targeting CSCs remains experimental at this stage.¹¹⁴ Novel immunotherapy and nanotechnology targeting of CSCs is also being explored. Successful strategies targeting CSCs, however, would require an in-depth understanding of the biology of all three stem cell compartments: CSC, CEP, and MSC. Targeting cancer stem cells using conventional phase I dose escalation may not be relevant or applicable, and well-designed prospective clinical trials are needed to validate surrogate markers, such as CTCs and CEPs, for drug development.

Therapy directed at these stem cell compartments may lead to significant survival benefits in patients with metastatic colorectal cancer. The median overall survival (OS) for patients with metastatic colorectal cancer reached 20 months with chemotherapy.¹¹⁵ Though complete response is rare (4%) and not durable, median OS reached 44 months, comparable to those who underwent surgical resection of metastasis.^116,117

Integrating multimodality therapy with long-term maintenance (XCEL) with capecitabine and celecoxib, a selective COX-2 inhibitor, we reported a 28% complete response rate in patients with unresectable metastatic colorectal cancer.¹¹⁸ Among 19 patients who achieved complete radiographic responses, we observed a paradoxic 3-year relapse-free survival of 70% for those who did not undergo resection or had R1-2 resections vs. 20% for those who had R0 resection.¹¹⁹ Median OS reached 73.3 months in this group of patients whose median survival was projected to be around 20–30 months, based on their clinical risk scores.

Our unpublished preclinical data confirmed findings that 5-FU enriches CD133 expression,¹⁰⁴ and maintenance XCEL may effectively target colon CSCs. Furthermore, celecoxib can also inhibit CEP proliferation by inhibiting AKT and caspase 3 activation. ¹²⁰ Celecoxib can reduce high-risk polyps by 50%, probably through Wnt signaling. ^121,122 The cardiovascular toxicities of celecoxib and bevacizumab may be due to their disturbance in CEP function and levels.^115,121 Interestingly, celecoxib failed to demonstrate any survival advantage when combined with irinotecan-based chemotherapy in patients with metastatic colorectal cancer.¹²³

DISCUSSION

Cancer and cardiovascular disease are believed to be stem cell-related disorders, with opposing CEP levels. Tumorigenesis involves CSCs, CEPs, and MSCs, three distinct, interdependent stem cell compartments essential to the delopment of in vivo resistance to chemotherapy and radiation. Targeting CSC/CEP/MSC compartments may lead to tumor dormancy, a new paradigm in cancer therapy (Figure 1); however, it is important to avoid serious injury to cardiovascular and other normal tissue, given the striking similarities within all three stem cell compartments.^38,39,124 The revised “Seed and Soil” theory invokes new levels of molecular and cellular complexity, which may help researchers to narrow the search for “ideal” targets or pathways within these cellular compartments.