High marrow seeding efficiency of human lymphomyeloid repopulating cells in irradiated NOD/SCID mice

During the lifetime of a person, both lymphoid and myeloid cells are continuously generated from a small population of multipotent hematopoietic stem cells.1-3 However, their low frequency and phenotypic heterogeneity4,5 has made them a difficult human cell population to study directly. On the other hand, they can engraft the marrow of sublethally irradiated immunodeficient mice for prolonged periods, even after the intravenous injection of limiting numbers of cells. This, in turn, has allowed their quantification using the in vivo competitive repopulating unit (CRU) assay,4,6 first developed for murine stem cell enumeration.7 All repopulation assays likely underestimate hematopoietic stem cell frequencies for 2 reasons. First, only a fraction of the injected stem cells would be expected to seed into the extravascular space of the bone marrow (even when syngeneic or autologous myelo-ablated hosts are used), and, second, not all might be stimulated to produce measurable levels of lymphoid and myeloid progeny. Although the marrow-seeding efficiency of murine repopulating stem cells in vivo has not yet been measured, the high purities of such cells that can be obtained (greater than 20%)8,9 indicate that their in vivo seeding fraction must be at least as high. In support of this is the demonstration that approximately 20% of intravenously injected murine week 5 cobblestone area-forming cells (CAFC–week 5, a closely related cell type detectable in vitro), are found 24 hours later in the marrow of syngeneic recipients.10 

The purpose of the current experiments was to obtain a direct measure of the marrow-seeding efficiency of human repopulating stem cells in sublethally irradiated NOD/SCID mice. This involved measuring the total number of human stem cells initially injected into a group of primary test mice (by limiting-dilution analysis of another set of primary mice injected with much smaller numbers of the same human cells) and comparing this value with the number of human stem cells in the marrow of the test mice killed 1 day later (by limiting-dilution analysis of another set of secondary mice).

Cord blood (CB) cells were collected from healthy, full-term infants delivered through cesarean section and were placed in tubes containing heparin. Human fetal livers (FL) were removed from 14-to 21-week-old aborted fetuses, using foot-length measurement as a determinant of age, and single-cell suspensions were obtained by first mincing the livers into small fragments and then dissociating these with dispase. For both types of cell samples, approved institutional procedures for obtaining informed consent were observed. Low-density (less than 1.077 g/mL) previously cryopreserved cells, pooled from several CB or FL samples, were washed twice in Iscove medium plus 10% fetal calf serum (StemCell Technologies, Vancouver, BC, Canada) and resuspended either in phosphate-buffered saline for injection into mice or in Iscove medium for colony-forming cell assays.

CRU assays were performed, and values were calculated as previously reported.11 Briefly,NOD/LtSz-scid/scid (NOD/SCID) mice given 350 cGy and injected with graded numbers of test cells were killed 6 to 8 weeks later and analyzed individually for the presence of human lymphoid (CD34⁻CD19/20⁺) and myeloid (CD45/71⁺CD15/66b⁺) cells (5 or more of each) per 2 × 10⁴ propidium iodide [PI] events) in suspensions prepared from the marrow of femurs and tibias of each mouse. Human CRU frequencies were then calculated from the proportions of negative mice (mice not considered positive for both human lymphoid and myeloid cells) with L-calc software (StemCell), which uses Poisson statistics and the method of maximum likelihood. Human CRU numbers per total primary mouse bone marrow were calculated by multiplying the determined frequency by the total number of cells recovered from 2 femurs and 2 tibias by 4, on the assumption that 2 femurs and 2 tibias comprise 25% of the total marrow.12 

Table 1 shows the CRU frequencies determined for the low-density human CB and FL cells used in the current studies. These values, which are similar to previously published values,4,6,11 were used to calculate the number of CRU injected at the same time into other recipients of much larger numbers of the same cells. The latter mice were killed 1 day later, and the number of human CRU detectable in their marrow was assessed by a second series of limiting-dilution assays in NOD/SCID mice. These numbers and the derived marrow seeding efficiencies for human CB and FL CRU in irradiated NOD/SCID mice are summarized in Table2. The values obtained for human CRU from both sources were similar (7% for CB and 5% for FL) and also remarkably close to the value of 4% to 5% reported recently by Van Hennik et al13 for human CB CAFC–week 6 in NOD/SCID mice given 9 cGy. The fraction of human CFC that could be recovered from the marrow of the CB-injected mice (1 experiment) and FL-injected mice (3 experiments) in the current studies (assessed using conditions that are nonpermissive for murine CFC14) was lower (approximately 2%, data not shown). The corresponding value reported by Van Hennik et al13 was approximately 4%.

These findings bear out the prediction that human CRU and CAFC/LTC-IC–week 6 would have similar seeding efficiencies in the NOD/SCID xenotransplant model based on a growing body of evidence that the human CRU and CAFC/LTC-IC–week 6 assays detect highly overlapping populations.4 15-17 However, CRU frequencies would also be expected to be at least 10 to 20 times lower because of the in vivo parameters that specifically reduce their efficiency of detection. Thus, in the absence of rigorous quantitative studies, LTC-IC/CAFC–week 6 may be obvious in populations in which CRU appear undetectable.

The finding that human CB and FL CRU seed to the marrow of NOD/SCID mice with approximately equal efficiency suggests that the molecular mechanisms regulating this property are similar for these 2 sources of human CRU. This does not preclude the possibility that such mechanisms are subject to variation or that the seeding efficiency and developmental potential of stem cells may vary independently of each other. Interesting examples of populations that show large differences in repopulating activity that cannot be readily explained by changes in cell numbers include stem cells progressing from G₀ to G₁ and from G₁ to S/G₂/M.17-20 In addition, brief exposure of murine cells to IL-3 or to IL-3 in combination with SF and IL-12 was found to decrease the in vivo seeding efficiency of CAFC–week 6.10 Conversely, others21-24 have reported enhanced repopulating activity of marrow cells pretreated with these or other cytokines.

A 5% to 7% seeding efficiency of CRU in NOD/SCID mice means that their previously reported frequencies represent underestimates by a factor of 14 to 20. This also impacts on previous measurements of human stem cell amplification in NOD/SCID mice (up to 5-fold over input levels at 4 to 6 weeks after transplantation11 25). These would now be seen to represent net increases up to 100-fold and to indicate an ability of human stem cell self-renewal to be potently stimulated in vivo by broadly species cross-reactive mechanisms.

We thank Yvonne Yang for expert secretarial assistance in preparing the manuscript.