Division of Experimental Hematology, Department of Hematology/Oncology, St. Jude Children's Research Hospital, Memphis, Tennessee, USA
Division of Experimental Hematology, Department of Hematology/Oncology, St. Jude Children's Research Hospital, 332 N. Lauderdale Street, Memphis, Tennessee 38105, USA. Telephone: 901-495-2727; Fax: 901-495-2176
Stem cells from a variety of tissues can be identified by a side population (SP) phenotype based on Hoechst 33342 dye efflux. The Abcg2 transporter is expressed in hematopoietic stem cells (HSCs) and confers this dye efflux activity. To further explore the relationship among Abcg2 expression, the SP phenotype, and HSC activity, we have generated mice in which a green fluorescent protein (GFP) reporter gene was inserted into the Abcg2 locus. In these mice, the majority of bone marrow (BM) cells that expressed the Abcg2/ GFP allele were Ter119+ erythroid cells. The Abcg2/GFP allele was also expressed in approximately 10% of lineage-negative (Lin−) and in 91% of SP cells using stringent conditions for the SP assay. Flow cytometric sorting was used to isolate various Abcg2/GFP+ BM cell populations that were then tested for HSC activity in transplant assays. There was significant enrichment for HSCs in sorted Lin−/ GFP+ cells, with a calculated HSC frequency of approximately one in 75. There was no HSC activity detected in Lin−/GFP+ cells. Altogether, these results show that Abcg2 is expressed on essentially all murine BM HSCs and can be used as a prospective marker for HSC enrichment.
Bone marrow (BM) cells with the side population (SP) phenotype are enriched for hematopoietic stem cell (HSC) activity and can be identified based on their capacity to efflux the fluorescent dye Hoechst 33342 . This dye efflux has subsequently been attributed to expression of Abcg2, an ATP-binding cassette (ABC) transporter [2–4]. ABCG2/Abcg2 was originally identified as a drug resistance gene in human cancer cell lines [5, 6], but it is also expressed in a variety of normal tissues . Cellular detoxification is an important physiologic role of Abcg2 [8, 9], suggesting that Abcg2 expression in HSCs may reflect a role in protecting the stem cell genome from chronic exposure to environmental toxins. It has also been suggested that Abcg2 could provide a novel marker for HSC purification; however, prospective isolation of HSCs by expression of Abcg2 has not yet been reported. The ability to isolate HSCs based on expression of Abcg2 expression would be useful in circumventing the toxic effects of exposure to Hoechst dye .
The goal of this study was to test whether direct isolation of cells expressing Abcg2 would be useful for the isolation and enrichment of HSCs. Furthermore, we wanted to better define the relationship between Abcg2 expression and the SP phenotype in hematopoietic cells. All currently available monoclonal antibodies against murine Abcg2 bind exclusively to intracellular epitopes and therefore cannot be used to sort living cells. To overcome this limitation, we have generated a knockin mouse model in which an internal ribosomal entry site (IRES)-green fluorescent protein (GFP) expression cassette was inserted immediately downstream of the stop codon of the Abcg2 gene. This allelic modification results in expression of the GFP reporter gene under control of the Abcg2 transcriptional regulatory elements, with coexpression of a functional wild-type (WT) Abcg2 protein. This model has allowed us to characterize Abcg2 expression within various hematopoietic cell subsets, to more precisely define the relationship between Abcg2 expression and the SP phenotype, and to isolate cells that express the bicistronic Abcg2/GFP mRNA for testing in murine transplant assays.
Materials and Methods
The 5′ arm of the targeting construct was generated by polymerase chain reaction (PCR) amplification of a 2.0-kb fragment of the murine Abcg2 gene up to and including the stop codon. This fragment was subcloned into the pKONTKV1901 vector (Stratagene, La Jolla, CA, http://www.stratagene.com). Next, a loxP-flanked NeoR gene linked to an IRES-enhanced GFP (EGFP) cassette that contained a polyadenylation signal from bovine growth hormone  was introduced downstream to the 5′ arm. A 2.1-kb 3′ fragment immediately downstream of Abcg2 stop codon was amplified by PCR and then subcloned downstream to the IRES-EGFP expression cassette to achieve the final targeting construct. Purified DNA was linearized by digestion with Pvu I and electroporated into 129/SvJ-derived PrmCre embryonic stem cells (ESCs), which express Cre recombinase under control of the germline-specific protamine promoter . Resistant ESCs were selected in the presence of 350 μg/ml G418 and 2 μM ganciclovir (Invitrogen, San Diego, http://www.invitrogen.com). Genomic DNAs from selected clones were digested with SspI (Promega, Madison, WI, http://www.promega.com) and analyzed by Southern blot using a 750-bp probe 5′ to the targeting vector. Two correctly targeted ESC clones were identified and injected into C57BL/6 blastocysts to generate chimeric mice. Male chimeras were bred to female C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME, http://www.jax.org), and germline transmission of the targeted allele in the offspring was confirmed by Southern blot. All experiments with mice were done according to a protocol approved by the Institutional Animal Care and Use Committee.
Slides of 5- to 6-μM sections, cut from formalin-fixed, paraffin-embedded tissues, were baked at 60°C for 30 minutes to ensure tissue adherence to the slide and were then deparaffinized. Heat-induced epitope retrieval was performed using a Black & Decker Steamer (Hunt Valley, MD, http://www.blackanddecker.com) in citrate buffer (pH 6.0, Zymed; Invitrogen) at more than 96°C for 30 minutes. Slides were allowed to cool for 30 minutes in the covered steamer and were then placed in Tris-buffered saline/Tween 20 (TBST) buffer (Dako, Carpinteria, CA, http://www.dako.com) for 10 minutes prior to assay.
Immunohistochemistry assays were performed on a Dako autostainer at room temperature. TBST buffer was used to rinse the slides between incubation steps. Endogenous peroxidase activity was blocked by incubation with 3% H2O2 (Humco, Texarkana, TX, http://www.humco.com) for 5 minutes. Non-specific protein binding was blocked by incubation with Superblock (Pierce, Rockford, IL, http://www.piercenet.com) for 30 minutes, avidin block (Dako) for 10 minutes, biotin block (Dako) for 10 minutes, and 10% goat serum for 30 minutes. Slides were incubated with rabbit anti-GFP (Molecular Probes, Eugene, OR, http://probes.invitrogen.com) (1:200 dilution for 30 minutes or 1:400 for 60 minutes) or with concentration-matched rabbit immunoglobulins (Dako). Slides then were incubated for 10 minutes with biotinylated goat anti-rabbit antibody (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com), for 10 minutes with streptavidin conjugated to horseradish peroxidase (Dako), and for 5 minutes with DAB (3,3′ diaminobenzidine tetra hydrochloride; Dako). The slides were analyzed by conventional light microscopy using a Nikon E800 digital microscope (Melville, NY, http://www.nikonusa.com). The images were captured with a Nikon DXM 1200 digital camera and analyzed using ACT-1 software (Nikon).
Hoechst 33342 Staining of BM Cells
Hoechst staining of BM cells for SP cell analysis was performed as previously described . In brief, cells were counted and resuspended at 106 cells per ml in prewarmed Dulbecco's modified Eagle's medium-positive (Gibco, Grand Island, NY, http://www.invitrogen.com) containing 2% fetal bovine serum (FBS; HyClone, Logan, UT, http://www.hyclone.com) and 1 mM HEPES (HyClone) and incubated with 2.5–4.5 μg/ml Hoechst 33342 (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) for 90 minutes at 37°C. Subsequently, Hoechst-stained cells were pelleted and then resuspended in 100 μl of ice-cold Hanks' balanced saline solution (HBSS)+ (Gibco), 2% FBS, and 1 mM HEPES. The cells were then incubated for 20 minutes on ice with biotinylated mouse lineage panel antibodies comprised of anti-Gr-1, anti-CD11b/Mac-1, anti-CD3e, anti-CD4, anti-B220, anti-Ter-119, and anti-NK1.1, together with streptavidin/allophycocyanin (APC)-conjugated c-Kit antibody (2B8) and phycoerythrin (PE)-conjugated Sca1 antibody (E13-161.7), all from BD Pharmingen (San Jose, CA, http://www.bdbiosciences.com/pharmingen). The cells were then washed and resuspended in 100 μl of ice-cold HBSS+ with 20 μl of streptavidin-APC (BD Pharmingen) for 20 minutes on ice. After the final round of staining and washing, the cells were resuspended in 0.5 ml of HBSS+ containing propidium iodide (PI) and maintained on ice until analyzed in a FACS Vantage/SE/ DiVa cell sorter (BD Biosciences, San Diego, http://www.bdbiosciences.com). The SP gate was defined based on gates that were established in WT BM cells and analyzed concurrently, so that 0.05%–0.1% of cells were within SP gate. WT cells were also stained with an immunoglobulin isotype control to define the gates for antibody staining.
Flow Cytometric Analysis and Cell Sorting
BM cells were harvested from the tibias and femurs of Abcg2GFP/GFP mice and enriched for Lin− cells by incubation with a lineage cocktail of PE-conjugated antibodies. Subsequently, cells were washed and incubated with 80 μl of anti-PE magnetic beads per 107 cells for 20 minutes at 4°C, followed by negative selection using the MACS (magnetic activated cell sorting) cell separation system according to the manufacturer's instructions (Miltenyi Biotec, Auburn, CA, http://www.miltenyibiotec.com). Lin− cells were then either directly sorted based on GFP expression or stained with Hoechst and sorted afterward. Cells were stained with PI prior to sorting. Cell sorting was performed on BD Biosciences FACS-Aria flow cytometer. Cell-sorting gates for GFP and PE antibodies were defined using BM cells from WT mice and appropriate isotype control antibodies.
Competitive Repopulation Assay
Different numbers of sorted BM cells expressing the CD45.2 from Abcg2GFP/GFP knockin mice were mixed with 105 unfractionated BM cells from female C57/B6.SJL (CD45.1) recipient mice (The Jackson Laboratory). Cells were then injected into lethally irradiated (1,100 cGy) recipients in 0.5 ml of phosphate-buffered saline through tail-vein injection. For analysis of peripheral blood cells, red blood cells were lysed using FACS Lyse/Wash Assistant (BD Biosciences) and the remaining cells were stained with perCp-Cy-5.5-conjugated CD45.2 antibody (BD Pharmingen) in combination with the PE-conjugated anti-CD3 and anti-B220 antibodies and APC-conjugated anti-GR-1 and Mac-1 lineage antibodies. Cells were analyzed using the BD Biosciences LSR II flow cytometer.
Construction of the Abcg2-IRES-GFP Allele
We generated the Abcg2/GFP allele by inserting an IRES-GFP cassette immediately downstream of the Abcg2 stop codon and upstream of the endogenous polyadenylation signal sequence (Fig. 1A). Homologous recombination was achieved in PrmCre ESCs that express Cre recombinase under control of the protamine promoter, thus allowing Cre-mediated deletion of the NeoR gene from the male germline . Targeted ESC clones were screened for correct integration by Southern blot, using an outside intronic probe located between exons 14 and 15. Two correctly targeted clones were injected into C57BL/6J blastocysts and implanted into foster mothers. Three chimeric male mice exhibited germline transmission and served as founders for the colony. All of the offspring from these founders that inherited the targeted allele showed NeoR deletion by Southern blot. Homozygous mice (Abcg2GFP/GFP) were identified and were born in the expected mendelian ratio. Analysis of peripheral blood cells showed normal numbers and proportions of mature cells.
Tissue-Specific Expression of the Abcg2/GFP Allele
Prior studies examining the expression pattern of Abcg2 in various tissues have shown high levels of expression in the renal proximal tubules, intestinal epithelial cells, vascular endothelium, as well as the liver and placenta [8, 7]. To verify that tissue-specific expression was maintained with the recombinant Abcg2 allele, organs from Abcg2GFP/GFP mice were stained with an anti-GFP antibody and analyzed by immunohistochemistry. GFP protein expression was detected in the proximal tubules of the kidney and in the epithelial cells of the small intestine (Fig. 1B). Venous endothelium was also stained with the anti-GFP antibody (data not shown). All together, these results demonstrate that expression of the Abcg2GFP/GFP allele correlated with the known tissue-specific expression pattern.
Expression of the Abcg2/GFP Allele in BM SP Cells
Flow cytometric analysis of total nucleated BM cells revealed that approximately 10% of all cells expressed the Abcg2/GFP allele (Fig. 2A). Further analysis revealed that approximately 80% of all GFP+ cells were Ter119+ erythroid cells (Fig. 2B). This finding is consistent with prior reports showing that erythroblasts express Abcg2 mRNA  and protein . Relatively low proportions of GFP+ cells expressed markers for T or B lymphocytes, macrophages, granulocytes, or natural killer cells (Fig. 2B). These results demonstrate that the majority of cells in the BM that express the Abcg2/GFP allele are maturing erythroid cells.
We next examined the relationship between the Abcg2/GFP allele expression and the SP phenotype. Because our prior studies have shown that the SP assay becomes more specific for identifying cells that express Abcg2 as the concentration of Hoechst dye is increased , we performed the SP assay using three different dye concentrations. The overall percentage of SP cells decreased from 1.3% at 2.5 μg/ml Hoechst to 0.06% at 4.5 μg/ml (Fig. 3A). This result likely reflects the exclusion of cells with lesser degrees of transporter activity when the assay is performed at higher dye concentrations. As expected, increased dye concentration was also associated with higher level of GFP expression in the SP cells (Fig. 3B). These results demonstrate that almost all BM SP cells expressed the Abcg2/GFP allele when the SP assay was performed using stringent conditions (i.e., high concentrations of Hoechst dye).
It has previously been shown that there is a higher concentration of HSCs in the “distal tip” of the SP region compared with the more proximal regions in the SP “tail” [1, 14]. If Abcg2 were a marker for HSCs, it would be predicted that the concentration of cells that expressed GFP would be greatest in the distal tip of the SP region. This indeed proved to be the case. Flow cytometry was used to identify SP cells present in the Lin−, c-Kit+, Sca1+ (KSL) BM cell population. Analysis of GFP expression in this KSL-gated SP population showed an increasing proportion of GFP+ cells in sequential gates progressing down the SP tail (supplemental online Fig. 1). In the most distal region (R3), 98% of cells expressed the GFP marker compared with 49% in the proximal region (R1). This result is consistent with our previous finding in Abcg2−/− mice that SP cells appeared only in the proximal region of the SP gate and were completely absent in the distal tip of the SP region .
Expression of the Abcg2/GFP Allele in Lin− BM Cells
To further analyze expression of Abcg2 in immature hematopoietic cells, GFP expression was analyzed in Lin− BM cells. BM cells were stained with antibodies directed against T and B lymphocytes, erythroid cells, natural killer cells, granulocytes, and macrophages. Approximately 4% of the BM cells lacked expression of any these markers. Within this Lin− population, approximately 14% of the cells expressed the GFP reporter gene (Fig. 4). Within the total Lin− /GFP+ population, approximately 11% displayed the SP phenotype. When the stringency of the GFP gating was increased to include only the most GFP-bright cells, the number of SP cells within the GFP+ population also increased to as high as 58% (data not shown). In contrast, only 0.3% of Lin− /GFP− cells displayed the SP phenotype, and these cells occurred predominantly in the proximal shoulder region (Fig. 4). Therefore, although Lin− /GFP+ cells are significantly enriched for SP cells, a significant number of Lin− /GFP+ cells lack the SP phenotype. These results demonstrate that cellular factors other than expression of Abcg2 are required for expression of the SP phenotype.
Lin− /GFP+ BM Cells Are Highly Enriched for HSC Activity
To functionally characterize cells expressing the Abcg2/GFP allele, cell sorting was used to isolate Lin−, GFP+, and GFP− populations (Fig. 5A). BM cells were first enriched for Lin− cells using an immunomagnetic column. Further purification was achieved using flow cytometry to gate on Lin− cells that either lacked GFP expression (R2) or expressed the highest levels of GFP (R1). Morphologic analysis of the Lin−/GFP+ cells showed a relatively uniform population of undifferentiated cells with a high nuclear to cytoplasmic ratio, an open chromatin configuration, and numerous prominent nucleoli (Fig. 5B). In contrast, the Lin− GFP− cells showed heterogeneous morphology with varying degrees of myeloid and lymphoid differentiation. Approximately 55% of the Lin− /GFP+ cells expressed the HSC-associated KSL phenotype, whereas only 4% of cells in the Lin− /GFP+ populations expressed the KSL phenotype (Fig. 5C). These results suggest that the Lin−/GFP+ cells were significantly enriched for repopulating HSCs.
To determine the HSC frequencies in these populations, limiting dilution transplant assays were performed. Varying numbers of sorted cells (CD45.2 background) were mixed with 1 × 105 BM cells from congenic CD45.1 mice, and this mixture was injected into lethally irradiated CD45.1 recipient mice. The sorted GFP+ population comprised 15% of the total Lin− population, and the GFP− population 38% (Fig. 5A). Transplanted mice were analyzed 16 weeks after transplant for reconstitution with sorted cells. Reconstitution was defined as at least 5% CD45.2+ peripheral blood cells in both myeloid (Gr-1+ and/or Mac-1+) and lymphoid (B220+ or CD3+) lineages. Sorted cells were also analyzed for day-12 colony-forming units-spleen (CFU-S). Lin−/GFP+ cells gave multiple large colonies when mice were injected with 500 cells, whereas no CFU-S activity was noted from Lin−/GFP− cells at the same cell dose (data not shown).
In the group of recipients receiving only 10 Lin−/GFP+ cells, significant reconstitution was noted in one of seven mice (Table 1). In mice transplanted with 60 Lin−/GFP+ cells, 50% of the recipients had high-level reconstitution. Nearly all mice were repopulated with doses of Lin−/GFP+ cells ranging from 125 to 1,000 cells per recipient (Table 1). In all cases, reconstitution from the sorted cells was robust, with CD45.2 cells comprising 20%–50% of all peripheral blood cells. The calculated repopulating cell frequency in the Lin−/GFP+ population, based on the Poisson probability distribution, is one HSC per 75 cells. In contrast, no animals were reconstituted with Lin−/ GFP− cells at doses ranging from 1,000 to 60,000 cells per recipient. Altogether, these results show that the Lin−/GFP+ cell fraction was significantly enriched for repopulating cell activity and that there was little to no repopulating activity in the Lin− cells that lacked GFP expression. Thus, the majority, if not all, of BM HSCs expressed the Abcg2/GFP allele.
Several of the mice transplanted with 500 Lin−/GFP+ cells were killed at 20 weeks after transplantation, and BM, thymus, and spleen cells were analyzed for reconstitution with CD45.2 cells. Approximately 26%–41% of BM erythrocytes, granulocytes, and B cells, as well as thymocytes were donor-derived CD45.2+ cells (supplemental online Fig. 2). Similar levels of reconstitution were also noted in T and B lymphocytes and myeloid cells in the spleen, as well as in natural killer cells (data not shown). Secondary transplant experiments confirmed that the original sorted Lin−/GFP+ cells were able to reconstitute secondary recipients (supplemental online Fig. 4).
Functional Analysis of SP Versus Non-SP BM Cells That Express the Abcg2/GFP Allele
Our phenotypic analysis indicated that cells expressing the Abcg2/GFP allele were present in both the SP and non-SP populations, raising the question as to whether the GFP+, non-SP cells would contain repopulating cells. To address this question, we isolated both the SP and non-SP cell subpopulations within the Lin−/GFP+ population and performed quantitative repopulation assays. In the group transplanted with Lin−/ SP/GFP+ cells, 100% of recipient mice were reconstituted when 100 cells were transplanted. In contrast, significantly lower (five- to 10-fold) repopulating activity was detected in the Lin−/SP/GFP− subpopulation (Table 2). When we re-examined the GFP sorting gates used in this experiment, we noted that GFP-dim cells were present in the “GFP−” gate. We repeated a second experiment with revised gating for GFP− cells and found that the Lin−/SP/GFP− population was relatively depleted of c-Kit+Sca1+ cells (Table 2; supplemental online Fig. 4). Taken together, these results indicate that the number of HSCs in the GFP− SP cells is very low. When Lin−/non-SP/ GFP+ cells were used, repopulating cells were present but at a lower frequency than seen in the Lin−/SP/GFP+ population. No repopulating activity was detected using Lin−/non-SP/ GFP− cells.
Flow cytometry analysis of these populations showed expression of the c-Kit+Sca1+ phenotype in approximately 78% of the Lin−/SP/GFP+ cells, 9.5% of the Lin−/SP/GFP−, 5% of the Lin−/non-SP/GFP+ cells, and less than 1% of the Lin−/ NON-SP/GFP− cells (Table 2; supplemental online Fig. 4). This phenotypic analysis correlated well with the repopulating cell frequency seen in the transplant assays. Altogether, these results demonstrate that whereas nearly all repopulating cells express that Abcg2/GFP allele, a minority of these GFP+ HSCs do not display the SP phenotype.
Following the initial description of the SP phenotype as a new marker for HSCs [1, 15], SP cells have been identified as a candidate stem cell population in other adult organ systems [16–20] as well as in tumors [21–23]. The usefulness of the SP assay for isolating stem cells from these various sources is tempered by several limitations. Hoechst 33342 is a DNA intercalating agent that results in significant cellular toxicity and therefore is not suitable for clinical HSC isolation. Second, the SP assay is sensitive to changes in cell metabolism which occur in vitro and therefore can be problematic when attempting to obtain reproducible quantitative results . Identification of the role of the Abcg2 transporter in the SP phenotype has offered a potential cell surface marker for SP stem cells. The results reported here provide proof of principle that Abcg2 can indeed be used as a prospective marker for murine HSCs. Furthermore, the Abcg2/GFP knockin mouse model can now be used to test the hypothesis that Abcg2 will be a marker for stem cells in other organs and in cancer systems. For instance, we have found that skeletal muscle contains a population of CD45−/GFP+ cells (supplemental online Fig. 3) that are correlated with the previously identified SP muscle stem cell phenotype [16, 25].
We now report several new findings regarding the relationship among Abcg2 expression, SP cells, and HSC activity. First, the majority of Abcg2/GFP+ cells in the BM were Ter119+ erythroblasts rather than HSCs. Abcg2 expression is highly induced during erythroid development, resulting in significant levels of Abcg2 expression in mature enucleated red blood cells . The physiologic role of Abcg2 in erythrocytes involves transport of substrates involved in heme biosynthesis. Abcg2-null mice have an elevated level of protoporphyrin IX in erythrocytes  because of the fact that protoporphyrin IX is a direct substrate of Abcg2 . We show here that Abcg2 is not expressed in other mature hematopoietic lineages in the BM.
Within the immature Lin− population of BM cells, expression of Abcg2 can be used to isolate an enriched HSC population. Using a simple two-color flow cytometry assay to isolate Lin−, Abcg2/GFP+ cells, we show that this population was highly enriched for cells with the KSL phenotype. More importantly, the functional HSC frequency in this population was one in 75 cells, a significant enrichment compared with the Lin− cells, in which the HSC frequency is approximately one in 2,500. Furthermore, no repopulating cells were present in as many as 60,000 Lin−, GFP− BM cells. This result indicates that most, if not all, of the HSCs in the BM express Abcg2. Therefore, we find that Abcg2 expression is not associated with a specific subset of HSCs in adult BM but is a common feature of all HSCs.
We have previously shown that Abcg2 mRNA is expressed within both the SP and non-SP populations of murine ESCs  and now extend these results to show that Abcg2 expression is also present in Lin− BM cells that reside both within and outside of the SP fraction. These observations are consistent with a recent report showing that a number of murine HSCs lack the SP phenotype in a subpopulation defined by expression of endogolin and the Sca1 antigen . Therefore, Abcg2 expression in Lin− cells is more specific for HSC activity than the SP phenotype alone.
An important remaining question is whether this approach can be used to isolate human HSCs and, if so, whether this population will be useful for clinical transplantation approaches. We have previously shown that CD34+ and AC133+ human hematopoietic cells express ABCG2 mRNA . In addition, we have generated an antibody (5D3) that recognizes an external epitope of human ABCG2 and can be used to isolate a variety of cell lines that express ABCG2 on the cell surface [21, 27, 28]. This antibody can be used to identify ABCG2 expression on human erythrocytes  and on human BM cells , although the sensitivity for detection of ABCG2 expression in primary human hematopoietic cells is relatively low (B.P.S., unpublished data). Further work will be needed to determine the utility of this approach for human HSC isolation.
Table Table 1.. Reconstitution with sorted bone marrow cells from Abcg2GFP/GFP knockin mice
Table Table 2.. Long-term competitive reconstitution by SP and non-SP subsets sorted based on GFP expression
The authors indicate no potential conflicts of interest.
We thank Dr. R. Ashmun and the Flow Cytometry Core for invaluable assistance in cell-sorting experiments, Dr. G. Grosveld and the transgenic animal core for invaluable assistance in generating the knockin mice, Drs. S. O'Gorman and R. Locksley for the generous gift of the PrmCre ESCs and reporter fragment, respectively, and Dorothy Bush for assistance with the immunohistochemistry assay. This work was supported by grants from the National Institutes of Health (R01 HL67366 to B.P.S.), Cancer Center Support Grant P30 CA 21765, The Assisi Foundation of Memphis, and the American Lebanese Syrian Associated Charities.