Cyclophosphamide/granulocyte colony-stimulating factor causes selective mobilization of bone marrow hematopoietic stem cells into the blood after M phase of the cell cycle

In adult mammals, hematopoietic progenitors and hematopoietic stem cells (HSC) migrate from bone marrow (BM) to the periphery in response to several stimuli, including cytoreductive drugs, such as cyclophosphamide (CY),1,2 and cytokines, such as granulocyte colony-stimulating factor (G-CSF).3-5G-CSF is widely used clinically to mobilize progenitors and HSC for collection and transplantation, often in combination with CY, which augments its effect.1 2 Despite common use, the mechanisms of cytokine mobilization of HSC are poorly understood.

Combined administration of CY and G-CSF is associated with expansion of the HSC pool.6 Following a single dose of CY and subsequent daily doses of G-CSF to mice, a large fraction of HSC in the BM entered the cell cycle, leading to a more than 12-fold expansion in long-term (LT) repopulating HSC in the BM after CY and 2 days of G-CSF treatment. On the day after the dramatic 3-day expansion, the number of BM LT-HSC began to decline, concomitant with a precipitous increase in the number of LT-HSC in the blood and spleen. The rapid expansion of LT-HSC in BM following administration of CY and G-CSF suggested that most of these cells had entered the cell cycle.6 The question of whether all LT-HSC entered the cell cycle after CY/G-CSF, however, or whether a subset of quiescent HSC escaped stimulation, was not addressed.

Intriguingly, despite the fact that HSC in the BM were cycling rapidly at the time of mobilization, HSC released into the blood after treatment with CY/G-CSF were either in the G₀ or G₁ phase of the cell cycle.6 In addition, hematopoietic progenitor and stem cells that appear in the blood following treatment of mice or humans with various mobilizing cytoreductive drugs or cytokines are in either the G₀ or G₁ phase.7-14 At least 3 hypotheses could explain the strong bias in favor of G₀/G₁ HSC in the blood7: (1) HSC may be released from the BM in all phases of the cell cycle, but S/G₂/M-phase HSC might be preferentially cleared, leaving G₀/G₁ cells overrepresented in the blood; (2) alternatively, there could be a preferential release of noncycling (G₀) HSC into the blood, leaving cycling cells behind in the BM; and (3) it is also possible that all HSC enter the cell cycle in response to CY/G-CSF, and that actively cycling cells migrate selectively to blood after M phase and prior to the next S phase, thus placing them in G₀ (if they were to exit the cell cycle on mobilization) or in G₁phase.

We have previously shown that nearly all of the reconstituting activity of mobilized HSC isolated from CY/G-CSF–treated C57BL/Ka-Thy-1.1 mice is contained in 2 phenotypically defined populations: Thy-1.1^loSca-1⁺Lineage⁻(Lin⁻)c-Kit⁺Mac-1⁻cells with mostly long-term self-renewing multipotent progenitor activity (LT-HSC), and Thy-1.1^loSca-1⁺Lin⁻c-Kit⁺Mac-1^locells with mostly transiently self-renewing multipotent progenitor activity.6 In the present study, we focus primarily on LT-HSC. We provide evidence for hypothesis 3, that there is a selective migration of actively cycling postmitotic HSC from BM to blood during mobilization. In addition, we demonstrate directly that the transit time of mobilized peripheral blood (MPB) HSC in the blood is extremely brief, and that mobilized progenitor cells show enhanced migration to the spleen and liver, as opposed to BM, at early time points following transplantation.

Six- to 12-week-old C57BLKa-Thy-1.1 mice were bred and maintained at the Stanford University Laboratory Animal Facility and allowed to freely imbibe acidified water (pH 2.5). β-Actin/enhanced green fluorescent protein (eGFP) transgenic mice were generated using the pCXEGFP vector, generously provided by Dr Jun-ichi Miyazaki. The construction of the pCXEGFP vector and its use in generating eGFP transgenic mice have been described previously.15 16 To generate eGFP transgenic mice, pCXEGFP was linearized withBamHI and SalI and the fragment containing the cytomegalovirus–immediate early (CMV-IE) enhancer, chicken β-actin promoter, eGFP complementary DNA (cDNA), and β-globin polyadenylation sequences was gel purified using the Qiaquick gel extraction kit (Qiagen, Valencia, CA). Purified, linearized DNA was injected into zygotes from crosses of F1 (C57BL/Ka-Thy1.1 × C3H) mice. Founder mice were screened by FACS analysis of peripheral blood leukocytes, and eGFP⁺ animals were back-crossed to C57BL/Ka-Thy1.1 mice. Mice used for stem/progenitor cell isolation had been back-crossed for at least 5 generations.

Cyclophosphamide and G-CSF were administered as previously described6 (Figure 1A).

Bromodeoxyuridine (BrdU) was administered as previously described17 (Figure 1B). Mice treated with CY/G-CSF and BrdU showed changes in HSC frequency, absolute numbers, and tissue distribution (data not shown) characteristic of treatment with CY/G-CSF alone.6 No alterations in peripheral blood counts or in HSC function, frequency, absolute numbers, tissue distribution, or cell cycle status were observed following treatment with BrdU alone.17 

Single-cell suspensions of BM and spleen cells were prepared as previously described.6 In a typical BrdU experiment, BM was pooled from 5 mice, and blood was pooled from the same 5 mice and from 5 to 15 additional mice.

Monoclonal antibodies (mAbs) used in immunofluorescence staining were prepared from hybridomas and included 19XE5 (anti–Thy-1.1), 2B8 (anti–c-Kit), E13 (anti–Sca-1, Ly6A/E). Lineage marker Abs included KT31.1 (anti-CD3), GK1.5 (anti-CD4), 53-7.3 (anti-CD5), 53-6.7 (anti-CD8), Ter119 (antierythrocyte-specific antigen), 6B2 (anti-B220), 8C5 (anti–Gr-1), and M1/70 (anti–Mac-1). The mAbs BU-1 (anti-BrdU, Caltag, South San Francisco, CA) and fluorescein isothiocyanate (FITC)-conjugated antimouse polyclonal Ab (Caltag) were used in BrdU analyses. Biotinylated 3C11 (anti–c-Kit), was sometimes used for positive selection of HSC. 2B8 and 3C11 recognize nonoverlapping epitopes on c-Kit.

The HSC were isolated as described previously.18Sca-1⁺ or c-Kit⁺ cells were sometimes enriched by positive selection using MACS (Miltenyi Biotec, Sunnyvale, CA) streptavidin-conjugated magnetic beads according to the manufacturer's instructions. Figure 1C shows staining profiles and approximate sorting gates used for isolation of CY/G-CSF day +4 BM (i, i′) and blood (ii, ii′) LT-HSC.

Cell cycle analyses were performed as described.17Briefly, double-sorted HSC were stained with Hoechst 33342 (Molecular Probes, Eugene, OR) at 10 μg/mL for 45 minutes at 37°C in Hoechst medium.19 Pyronine Y (PY) was then added to 1.0 μg/mL, and the cells were incubated for an additional 15 minutes, prior to washing and analysis by flow cytometry.

Cells were stained with BrdU and visualized by fluorescence microscopy as previously described.17 To obtain percentages of BrdU⁺ HSC, cell nuclei (stained blue with Hoechst 33342) were identified using the Hoechst filter, then BrdU⁺ (green) nuclei in the same field were counted using the FITC/Texas Red filter (Figure 1D). In each experiment, whole thymocytes sorted from mice not exposed to BrdU were used as negative controls and were always 100% negative (Figure 1D, i, i′, and data not shown). Thymocytes from mice exposed to BrdU were used to control for possible nonspecific staining. Consistent with expectations,20 a significant fraction of thymocytes sorted from mice treated with BrdU alone for 5 days were BrdU⁺ (ii, ii′, and data not shown). LT-HSC isolated from BM (iii, iii′) or blood (iv, iv′) of BrdU/CY/G-CSF–treated animals showed the characteristic BrdU punctate green nuclear staining.21 To rule out the possibility that residual staining from the FITC-conjugated anti–Thy-1.1 could yield false-positive staining, Thy-1.1^lo BM cells were sorted directly onto a slide and observed to have no staining (data not shown).

Nested reverse transcriptase–polymerase chain reaction (RT-PCR) on double-sorted LT-HSC (5 cells/sample) was performed as described.17 Primer sequences are as follows: sense primers, 5′AGA GAC CAT CCC GCT GAC TGC3′; 5′GAA AAG CTG TGC ATT TAC ACC3′. Antisense primers, 5′GCA GGC TGT TCA GCA GCA GAG3′; 5′GCT CAG TCA GGG CAT CAC ACG3′. RT primers, 5′TCC CGC ACG TCT GTA GGG GTG3′.

Red blood cells (RBC) from untreated C57BL/Ka-Thy1.1 mice were isolated by density centrifugation over a Ficoll cushion (Sigma, St Louis, MO). Purified RBC (10⁸) were then labeled with the fluorochrome PKH-26, according to the manufacturer's recommendations (Sigma).

Hematopoietic progenitor cells enriched for HSC (Lin^−/loSca-1⁺c-kit⁺) were isolated from day +4 MPB of β-actin/eGFP transgenic mice. The eGFP⁺ progenitor cells (45 000) and 150 000 PKH-26–labeled erythrocytes were sorted and injected together into the retrorbital sinus of anethetized day +4 CY/G-CSF C57BL/Ka-Thy1.1 recipients. Samples of peripheral blood were collected from the tail vein at the indicated time points (from 30 seconds to 3 hours). After 3 hours, mice were killed and tissues and organs were collected, as indicated, for analysis. Quantitation of PKH-26⁺ RBC or eGFP⁺ progenitors in blood and tissue samples was performed by flow cytometry. Preliminary experiments demonstrated that samples from uninjected control mice did not generate signals in the PKH-26⁺ or eGFP⁺ gates.

The Student t test was used to compute the Pvalues shown in Figure 2B.

To test whether G₀/G₁ HSC appearing in the blood following mobilization derive from noncycling (G₀) BM LT-HSC that are selectively released into the blood, or are instead LT-HSC that after actively cycling in the BM are released into the blood after M phase of the cell cycle, mice were treated simultaneously with CY/G-CSF and with the thymidine analog BrdU, which is incorporated during DNA synthesis (Figure 1B,D). Virtually all (99.1 ± 0.5% SD) Thy-1.1^loSca-1⁺Lin⁻c-Kit⁺Mac-1⁻cells isolated from the BM of day +4 CY/BrdU/G-CSF–treated mice stained positively for BrdU (Table 1). Thy-1.1^loSca-1⁺Lin⁻c-Kit⁺Mac-1⁻cells from MPB of CY/BrdU/G-CSF–treated mice were 98.5% BrdU⁺ (Table 1, experiment 2), and Thy-1.1^loSca-1⁺Lin^−/loc-Kit⁺cells (this population includes the Thy-1.1^loSca-1⁺c-Kit⁺Mac-1^lotransiently self-renewing HSC) were 97.8% BrdU⁺ (Table 1, experiment 3). Thus, nearly all HSC isolated from MPB had divided at least once during BrdU exposure, even though only a small percentage of such cells are typically in the S, G₂, or M phases of the cell cycle at the time of their isolation6 (data not shown). Given that the vast preponderance of HSC are in the BM until day +3, and these are dividing,6 the most likely explanation for the presence of BrdU⁺, G₀/G₁ phase HSC in the blood is that HSC divided in the BM and migrated to the blood prior to their next S phase.

If the MPB and splenic HSC are derived from dividing LT-HSC following CY/G-CSF, then giving these mice a second dose of CY on day +2 of the protocol should eliminate them, because CY preferentially kills dividing cells.22 As reported previously, the initial dose of CY on day −1 did not result in a reduction in the frequency or the absolute number of LT-HSC in the BM on day 0.6 However, as shown in Figure 2, BM, blood, and spleens from mice given the second dose of CY on day +2 contained only a small fraction of LT-HSC on day +4 compared to control mice given the standard regimen. These data suggested that at day +2 of the CY/G-CSF regimen, nearly all of the LT-HSC that could give rise to MPB HSC (or BM or splenic HSC) were cycling.

MPB LT-HSC are found primarily in the G₀/G₁ fraction of the cell cycle,7-14 and appear to migrate into the bloodstream following completion of M phase. To determine whether MPB HSC remain in cycle in the bloodstream, or if entry into the bloodstream is accompanied by exit from the cell cycle (ie, entry into G₀phase), the expression of messenger RNA (mRNA) for cyclin D2, a marker for cycling cells,23 was assayed. In 2 experiments, nested RT-PCR performed on 5 cells per sample revealed that most CY/G-CSF day +4 BM and blood LT-HSC express cyclin D2 mRNA. Data from one of the experiments is shown in Figure 3. In this experiment, 7 of 10 BM and 9 of 10 blood LT-HSC samples were cyclin D2⁺. Little cyclin D2 mRNA from control splenic T cells was detected. In the second experiment (not shown), 12 of 12 BM and 12 of 12 blood samples were positive. Because cyclin D2 mRNA is rapidly degraded following exit from the cell cycle,24 25 these results are consistent with the notion that HSC isolated from blood are in G₁ phase.

To characterize further the cell cycle status of HSC appearing in the periphery following treatment with CY/G-CSF, double-sorted HSC were stained with Hoechst 33342 and PY (see “Materials and methods”), allowing simultaneous assessment of the fraction of cells with more than 2n DNA content (Hoechst 33342), as well as relative levels of total double-stranded RNA (PY26). As cells progress from G₀ to G₁, and from G₁ to S, G₂, and M phases, they accumulate RNA, mostly in the form of double-stranded ribosomal RNA,26 and so it is reasonable to propose that PY high cells have elevated protein synthesis. As shown in Figure 4 and Table2, total RNA levels of G₀/G₁ and S/G₂/M HSC isolated from CY/G-CSF–treated animals (day +3 to day +5) were always higher than RNA levels of corresponding populations from untreated mice in all tissues examined. In 3 experiments, LT-HSC from blood of CY/G-CSF day +4 and +5 animals had 50%, 13%, and 24% (mean = 29% ± 19% SD) higher PY staining than HSC from BM of untreated mice (Table 2), suggesting that MPB LT-HSC are in a more activated state than LT-HSC from BM of untreated mice.

If HSC are selectively mobilized following M phase, but are found in the blood primarily in G₁ phase, then transit of MPB HSC in the blood must be relatively brief (ie, shorter than progression from G₁ to S phase). To test directly the half-life of MPB HSC and progenitor cells in the bloodstream, we purified Lin^−/lockit⁺Sca-1⁺ hematopoietic progenitor cells from MPB of mice transgenically expressing eGFP driven by the β-actin promoter. These mice express high levels of eGFP in all cells,16 including HSC (Figure5A), a fact that can be exploited to track the in vivo migration of transplanted eGFP⁺ cells in a wild-type host. FACS-sorted MPB eGFP⁺ hematopoietic progenitors from day +4 CY/G-CSF–treated animals were injected intravenously into wild-type recipients similarly treated with CY/G-CSF. To control for the injection, normal erythrocytes, fluorescently tagged with the red fluorochrome PKH-26 were coinjected with the eGFP⁺ cells. Whereas PKH-26⁺erythrocytes were easily detectable in the blood of the recipient at the early time points (< 1 minute), eGFP⁺ progenitor cells were not detected (Figure 5B). In fact, although PKH-26⁺RBC were detectable in the bloodstream throughout the experiment, eGFP⁺ progenitors cells were never observed in the blood. Importantly, eGFP⁺ progenitor cells were easily identified in the spleen, liver, and BM when recipient mice were killed 3 hours after intravenous injection (Figure 5C). In addition, the migration pattern of transplanted MPB HSC was consistent with previous studies showing that mobilization initiates migration of BM HSC to the spleen through the blood.6 The eGFP⁺ stem and progenitor cells were found primarily in the liver and spleen, and only a small fraction of cells migrated to BM. The eGFP⁺ cells were not detected in lymph nodes, Peyer patches, or thymus. Taken together, these data demonstrate that the transit time of mobilized stem and progenitor cells in the blood is very brief and suggest that migration of HSC into spleen and liver may contribute to the rapid clearance of HSC from the circulation.

We examined the relationship between LT-HSC expansion in the BM and mobilization from BM to blood. By administering BrdU to mice undergoing a mobilizing regimen of CY/G-CSF and then isolating BM and MPB LT-HSC, the proliferation history of LT-HSC in both compartments was determined. As previously reported, in otherwise untreated mice placed on continuous oral BrdU for 5 days, about 40% of Thy-1.1^lo Sca-1⁺ Lin⁻c-Kit⁺ Mac-1⁻ cells isolated from the BM stained positively for BrdU.17 Similarly, less than 30% of rhodamine 123^lo/Hoechst 33342^lo primitive HSC isolated from the BM of mice on continuous oral BrdU for 5 days stained positively for BrdU.27 In contrast, virtually all HSC isolated from the BM of day +4 CY/G-CSF–treated mice were BrdU⁺, as were nearly all MPB HSC from the same mice. Because appreciable DNA synthesis does not occur while LT-HSC are in the blood, we concluded that BrdU⁺ MPB LT-HSC underwent cell division elsewhere. Because significant numbers of LT-HSC do not appear in the blood until day +3, it is not possible that all of the cells could have cycled in the blood prior to day +4, when the cells were obtained, even if the cell cycle time was short (8-10 hours), and even if all LT-HSC that entered the blood remained there. The most likely place for HSC division to have occurred is the BM. The near total susceptibility of BM LT-HSC from CY/G-CSF–treated mice to a second dose of CY confirms that most or all of these cells were cycling on day +2. This result is consistent with the increased sensitivity to 5-fluorouracil (5-FU) of splenic HSC stimulated to enter the cell cycle with Steel factor,28 and of BM HSC stimulated to cycle with 5-FU,29 as well as with the observation that the HSC compartment is significantly damaged by the administration of alternating doses of cytotoxic agents and G-CSF to mice.30The fact that 99% of MPB LT-HSC had previously cycled in the BM, and that 95% of these cells were in G₀/G₁phase,6 indicates that LT-HSC egress from BM to blood occurs after M phase and before the next S phase. Thus, it is not the case that a quiescent subset of G₀-phase LT-HSC is selectively mobilized in CY/G-CSF–treated mice. Rather, all BM LT-HSC enter the cell cycle, and some of these cycling cells then migrate into the blood, by a process that appears tightly coordinated with the cell cycle.

It remained formally possible that HSC may be released from the BM in all phases of the cell cycle, but S/G₂/M-phase HSC are preferentially cleared, leaving G₀/G₁ cells overrepresented in the blood. If this were true, the earliest HSC appearing in the spleens of mice following mobilization with CY/G-CSF would be expected to have an S/G₂/M fraction as high or higher than that of the BM, because the spleen is a primary destination for circulating HSC in mice.6,11,31-33 However, on the day that significant numbers of HSC abruptly appear in the blood and spleen following CY/G-CSF, only about one third as many splenic HSC as BM HSC are in S/G₂/M.6 Further, if this hypothesis were true, one would expect that MPB HSC from splenectomized mice would have a higher percentage of S/G₂/M cells than MPB HSC from intact mice, which has been shown not to be the case.7Finally, the demonstration that MPB progenitors, the majority of which are in G₁ phase of the cell cycle, are rapidly and uniformly cleared from circulation argues against selective clearance of S/G₂/M-phase cells from the blood.

The association of LT-HSC expansion with mobilization begs the question of whether mobilization follows cell division in all instances in which HSC migration is observed. In normal prenatal development, sequential migrations of pluripotent HSC are thought to be associated with expansion of the HSC pool.34,35 Following treatment of adult animals with some cytokines, however, mobilization can occur very rapidly. Within 30 minutes of administration, interleukin 8 (IL-8) causes mobilization of hematopoietic progenitors and LT-HSC in mice,36 and of hematopoietic progenitors in primates,37 albeit in numbers much lower than at the peak of CY/G-CSF mobilization.6 Thus, it might appear that IL-8 causes mobilization by a different pathway than CY/G-CSF or that it acts by bypassing the early steps of the CY/G-CSF pathway. However, it may be that only BM HSC that have recently undergone cell division are mobilized by IL-8. On the order of 8% of the approximately 80 000 LT-HSC in adult mouse BM6 enter the cell cycle each day.17,27 If 8% of these LT-HSC also complete M phase each day, and if half of these cells are rapidly mobilized by IL-8 treatment, MPB would contain a maximum of about 1.5 LT-HSC/μL blood after IL-8 (assuming a blood volume of 2 mL). This estimate appears consistent with published data on mobilization induced by IL-8.36 Therefore, even though there are clearly differences between mobilization induced by IL-8 and CY/G-CSF, postmitotic, G₁-phase cells may migrate in both cases.

About 60 000 LT-HSC appear in the spleen between days +2 and +3 of the CY/G-CSF treatment protocol (an ∼100-fold increase), even though the number of LT-HSC in the blood at any moment during this interval averages fewer than 10 000.6 At a minimum, it would appear that the equivalent of the entire complement of LT-HSC in the blood would have to be delivered to the spleen 6 times in 24 hours. The percentage of cycling cells in the BM or spleen is insufficient to account for more than a small fraction of the ∼100-fold increase in the number of splenic cells.6 Thus, the most likely explanation is that LT-HSC arrive in the spleen from the BM through the blood at a high rate of flux. This hypothesis was confirmed by directly measuring the clearance of mobilized stem and progenitor cells from MPB. Transferred eGFP⁺ progenitor cells were removed from the circulation of day +4 CY/G-CSF–treated wild-type recipients in less than 1 minute, whereas control PKH-26–labeled RBC were observed in the blood for at least 3 hours after injection, the observation interval. Importantly, the rapid clearance of mobilized eGFP⁺ progenitor cells from the circulation cannot be explained entirely by nonspecific clearance mechanisms, because significant migration of eGFP⁺ cells into the spleen and liver of the mobilized recipients was observed within 3 hours of transfer. Taken together, the above data suggest that MPB HSC are actively cycling G₁-phase cells that migrate rapidly from the blood into specific hematopoietic organs. These data suggest that MPB HSC are in the circulation for only one or 2 passes prior to migration to other sites, whereas RBC are fated to circulate in the bloodstream without homing.

Alternatively, LT-HSC may cycle in the BM, pass through M phase, and enter G₀ phase before or on arriving in the blood. A recent study of human MPB CD34⁺ cells,12 only a fraction of which are HSC,38 supports this view, whereas another group concluded that human MPB CD34⁺ cells were in G₁ phase.13 It has been shown that human CD34⁺Thy-1⁺Lin⁻ cells sorted from MPB were about 12 hours slower to enter S phase when placed in culture under cytokine stimulation than their counterparts isolated from BM of untreated volunteers.11 This was reasonably interpreted to suggest that mobilized HSC were in a relatively quiescent state. In light of the current study, however, some of this difference may be due to the fact that the process of mobilization gives rise to a population of HSC in the blood that is synchronized in early G₁ phase, whereas G₁ phase HSC isolated from BM of untreated mice are distributed throughout the entire G₁ phase, with some cells very close to entry into S phase. Thus, HSC isolated from normal BM may have a head start in relation to entry into S phase. The cyclin D2 and PY data are consistent with the notion that HSC entering the blood are mostly in G₁ phase or are in a state of activation intermediate between the G₀ and G₁ phases. (The fact that MPB HSC are largely cyclin D2⁺ and have increased PY staining also is strong evidence that the BrdU incorporation resulted from authentic S-phase synthesis and not from BrdU-induced unscheduled DNA repair, as these are independent markers of cells in the mitotic cycle.)

One might imagine that an early step in physiologic or drug-induced HSC migration would involve weakening or detachment of specific adhesive interactions between HSC and surrounding stromal cells and ECM elements. There is reason to assume that BM LT-HSC, which express high levels of c-Kit,18 α₄-integrin, and CD44 (D.E.W., S.J.M., I.L.W., unpublished data, October 1996), have interactions of significant avidity with stromal membrane-bound Steel factor, vascular cell adhesion molecule 1 (VCAM-1), and hyaluronate, respectively, as each of these ligands is ubiquitous in the BM.39-43 The α₄-integrin–VCAM-1 interaction is believed to play a role in retention of HSC and progenitors in the BM.44-48 Detachment from substrate can occur in association with cell division.49,50 This could be accomplished either by modulation of expression levels of adhesion molecules such as α₄β₁ integrin, or by alteration of the affinity of receptors such as integrins (reviewed in reference 51) or CD4452 for their ligands. It has long been observed that cultured adherent cells loosen their attachment to the substrate and “round-up” during M phase.49 Similarly, dividing HSC may detach from stromal elements during M phase, and in the process become competent to migrate, given the appropriate secondary signals. This model is consistent with the observed appearance in the blood of postmitotic HSC that have recently cycled.

We thank Libuse Jerabek for superb laboratory management, Veronica Braunstein for antibody production, Koichi Akashi and Dennise Dalma-Weiszhausz for careful review of the manuscript, David Parks for advice on flow cytometry, Stanley Tamaki and Nobuko Uchida for advice on DNA/RNA staining for flow cytometry, Allen Smith for advice on DNA repair, Jos Domen for helpful discussions, and Diana J. Laird for statistical advice.