Proc. Nati. Acad. Sci. USAVol. 88, pp. 2788-2792, April 1991Developmental Biology

Clonal analysis of hematopoietic stem-cell differentiation in vivo(self-renewal/stem-cell purification/bone marrow transplantation)

LAURIE G. SMITH*, IRVING L. WEISSMAN*, AND SHELLY HEIMFELDtt*Departments of Pathology and Developmental Biology, Howard Hughes Medical Institute, Stanford University School of Medicine, Stanford, CA 94305; andtSyStemix, 3400 West Bayshore Road, Palo Alto, CA 94303

Contributed by Irving L. Weissman, November 21, 1990

ABSTRACT Previous work has shown that the 0.02-0.05% of adult mouse bone marrow cells that bear the cellsurface phenotype Thy 1IoLin-Sca-l+ are enriched 1000- to2000-fold for hematopoietic stem-cell activity in a variety ofassays. When 50-100 cells of this phenotype are iijected into anirradiated animal, they can permanently repopulate the entirehematopoietic system. In the present study, limiting-dilutionand single-cell experiments were used to address the questionof how individual Thy-11OLin-Sca-1+ stem cells contribute torepopulation of the hematopoietic system following irradiation.We calculated that 1 of 13 Thy-loLin-Sca-l+ cells formed aclone comprising >1% of peripheral white blood cells 3-7weeks after injection. The majority of these clones includedboth lymphoid and myeloid lineages. Approximately one-thirdof the clones continued to produce new blood cells for 9 weeksor more, but the remainder disappeared earlier, includingmany that were multilineage. Thus, while the majority ofThy-1IOLin-Sca-l+ bone marrow cells whose progeny aredetected in the in vivo repopulation assay are pluripotential,only a subset undergo long-term self-renewal in vivo. Repop-ulation appears to be oligoclonal when limiting numbers ofThy-1IOLin-Sca-l+ cells are injected. However, the number ofclones contributing to hematopoiesis increases in proportion tothe number of Thy-1IOLin-Sca-l+ cells i 'ected, bringing intoquestion the notion that steady-state hematopoiesis in normalindividuals is oligoclonal.

All cell lineages of the blood system are derived frompluripotential hematopoietic stem cells (PHSCs), a rare classof cells that reside in the bone marrow of adult vertebrates(1). Blood cell lineages are transient, with half-lives rangingfrom hours (e.g., granulocytes) to months (e.g., lympho-cytes), and must therefore be generated continuouslythroughout the life ofthe animal (2). How PHSCs individuallycontribute to this overall outcome is not well understood.Because of their low frequency in the bone marrow (10'- to10-5, depending on the method used to assay PHSCs; refs.3-6), these cells have been difficult to study directly. Mostwork on this problem has relied on the ability to study cloneselaborated from marked cells of unknown identity in bonemarrow or fetal liver (e.g., refs. 7-10). Previous work hasdemonstrated that adult mouse bone marrow cells (BMCs)expressing Sca-1 (Ly-6A) and low levels ofThy-1 but lackingexpression of several other hematopoietic lineage markers(Thy-1loLin-Sca-1+) are enriched 1000- to 2000-fold com-pared with unseparated BMCs for PHSC activity in a varietyofassays (11-13). As few as 30 Thy-1VoLin-Sca-1+ BMCs canrescue 50% of lethally irradiated mice and repopulate thehematopoietic system of the host. To characterize Thy-1ioLin-Sca-1+ stem cells more precisely, and to study the clonalorganization of hematopoietic repopulation by this definedcell population, we have assessed the developmental poten-

tial of single Thy-11oLin-Sca-1+ cells in vivo by two methods.We developed a limiting-dilution system that allows us totrace the progeny of individual Thy-1loLin-Sca-l+ cells withhigh probability. In addition we injected single Thy-1IoLin-Sca-1l cells [isolated by fluorescence-activated cellsorter (FACS)] into irradiated mice and observed the clonesthey produced. Our results indicate that individual Thy-110Lin-Sca-1l cells produce multilineage clones of variable sizeand life-span. The number of clones contributing to he-matopoiesis in any one irradiated host appears to depend onthe number of Thy-11oLin-Sca-1+ cells injected, at leastwithin the first 1-2 months after irradiation. These results arediscussed in relation to previous studies on the clonal orga-nization of hematopoiesis.

MATERIALS AND METHODSMouse Strains. Donor and host mice were either C57BL or

congenic with this strain and were bred and maintained in themouse facilities at Stanford University or SyStemix. Donor/host strain combinations are summarized in Table 1.BMC Separation and Construction of Radiation Chimeras.

Thy-11oLin-Sca-l+ cells were isolated from bone marrow of4- to 6-week old mice as described (12), with some modifi-cations (14). Monoclonal antibodies used for FACS sortingand analysis have been described (11, 12, 14) and are listedin Table 2. Cell separation was performed pn a dual-laserFACS (Becton Dickinson), modified as described (15) andmade available through the FACS shared-users group atStanford University, or on a dual-laser FACStar Plus atSyStemix. Populations of Thy-1l°Lin-Sca-1+ BMCs are rou-tinely prepared with 80-90% purity by this method. Themajor contaminants are Thy-1-Lin- and their progeny arenot detected at high frequency in any of several assays forhematopoietic stem-cell activity (refs. 11 and 16; unpublishedobservations). Before injection, sorted populations werecounted and mixed in appropriate numbers. To that mixturewere added irradiated (2000 rads, 137Cs source; 1 rad = 0.01Gy) thymocytes (5 x 106 per ml) as carrier cells from donorsnot carrying the Ly-5 allele of interest for that experiment.The final mixture was injected (200 ,ul per mouse) intrave-nously into the retroorbital plexus of etherized mice that hadbeen lethally irradiated earlier the same day (2 x 450 rads ofx-rays, 4 hr apart).

Single Thy-1IoLin-Sca-1+ cells were sorted directly intoseparate wells of a 96-well plate containing 100 ,ul of buffer byan automatic cell deposition unit. Calibration of this cloningdevice with fluoresceinated beads and hybridoma cells hasshown that

Proc. Natl. Acad. Sci. USA 88 (1991) 2789

Table 1. Donor/host strain combinations

Limiting dilution Single-cell and competitiveExperiments 1 and 2 Experiments 3 and 4 repopulation experiments

Donor 1 C57BL/Ka-Thy-1.1(Ly-5.2) C57BL/Ka-Thy-1.1(Ly-5.2) F1*Donor 2 C57BL/6-Ly-5.1-Pep3b F1* C57BL/Ka-Thy-1.1(Ly-5.2)Host C57BL/6-Ly-5.1-Pep3b C57BL/6-Ly-5.1-Pep3b C57BL/Ka(Ly-5.2)Phenotype markingdonor 1 progeny Ly-5.2' Ly-5.2'Ly-5.1- Ly-5. 1 +

*F, = (C57BL/Ka-Thy-1.1 x C57BL/6-Ly-5.1-Pep3b)F1.were then injected as described, using separate syringes toprevent cross-contamination between wells. About 30% ofthe total volume was left behind in the syringe as "deadvolume." Thus we estimate that about 50%o of mice injectedreceived no cell.

Analysis of Repopulation. Peripheral blood samples werecollected, erythrocyte-depleted, and stained as described(17), except that cell suspensions were stained first withbiotin-conjugated anti-Ly-5.1 or anti-Ly-5.2 and subse-quently with Texas Red-avidin mixed with directly fluores-cein isothiocyanate (FITC)- and allophycocyanin-conjugatedreagents as follows: FITC-anti-B220, or FITC-anti-Mac-1plus FITC-anti-GR-1, or FITC-anti-CD8 plus allophycocya-nin-anti-CD4, or FITC-anti-Thy-1.1 plus FITC-anti-Thy-1.2.Dual-laser FACS analysis was performed as described ear-lier.

Statistical Methods. Limiting-dilution experiments wereanalyzed using the binomial equation (18)

P(r) = [n!lr!(n - r)!] prq(n-r).

P(r) is the probability that r clones will form from n markedThy-1IoLin-Sca-1+ cells, of which each has a probability p offorming a clone and a probability q of not forming a clone (p= 1 - q). The probability that a mouse will be negative (i.e.,r = 0) reduces to

log P(0) = n log q.

Thus, if q is fixed, then a linear relationship (with the slopeof log q) will exist between the logarithm of the proportion ofnegative mice in a group and the number of cells injected.Values of P(0) and n can thus be used to calculate p and q foreach experimental group. The value of 1/13 for p is aweighted mean of the values obtained for each experimentalgroup plotted in Fig. 1. The probability that positive mice inany one group will be repopulated by one and only onemarked clone can be calculated from the binomial equationby using the calculated values for p and q (1/13 and 12/13,respectively). To do this, the probability P(1) (i.e., r = 1) isdivided by the sum P(1) + P(2) + P(3) + P(4) + P(5) (i.e., r. 1). If marked and unmarked Thy-11OLin-Sca-1+ cells haveequal probabilities of forming clones, then the mean of thetotal number of clones (marked and unmarked) contributingto hematopoiesis is p x n, where n now represents the totalnumber of cells injected, i.e., 8 for 105 cells injected, 8.5 for110 cells injected, and 9.2 for 120 cells injected.

Table 2. Monoclonal antibodies

Hybridoma Specificity Hybridoma Specificity19XE5 Thy-1.1 RB6-8C5 GR-153.2.1 Thy-1.2 53.6.7 CD8A20.1 Ly-5.1 GK1.5 CD4ALI-4A2 Ly-5.2 E13-161-7 Sca-1RA3-6B2 B220 145-2C11 CD3M1/70 Mac-1

Data shown in Fig. 3 B and C were analyzed by theKruskall-Wallis analysis of variance of ranks (18).

RESULTSLimiting-dilution experiments were performed initially todetermine the frequency of FACS-sorted Thy-1l°Lin-Sca-1+cells that can contribute to hematopoiesis when injected intoa lethally irradiated host. Limiting numbers of Thy-11o-Lin-Sca-1+ cells (5-20 cells) were injected into lethallyirradiated host mice together with a minimum dose of stemcells necessary for near 100% host survival: either 105 un-separated BMCs (Exps. 1 and 2) or 100 Thy-loLin-Sca-l+cells (Exps. 3 and 4) from a different donor. To mark progenyof the limiting number of Thy-1IoLin-Sca-1+ cells, we madeuse of congenic mouse strains carrying different alleles ofLy-5 on the C57BL background. Ly-5 is a polymorphic cellsurface antigen found on all mature leukocytes (19). Mono-clonal antibodies specific for Ly-5.1 and Ly-5.2 allowed us todistinguish progeny of the 5-20 Thy-1loLin-Sca-1+ cells fromthe progeny of the other stem cells and the host. Donor/hostcombinations are summarized in Table 1. Peripheral blood ofrecipient animals was screened 3-4 and 6-7 weeks later forcells carrying the Ly-5 marker, and simultaneously stainedwith antibodies that define mature T lymphocytes, B lym-phocytes, and myeloid lineages.An approximately linear relationship was observed be-

tween the number of marked Thy-loLin-Sca-l+ cells in-jected and the proportion (logarithmic scale) of mice thatwere negative for marked cells in peripheral blood 3-7 weekslater. Pooled results of four independent experiments areshown in Fig. 1. We analyzed these data with the binomialequation, assuming that there is a probability, p, that aninjected Thy-1IoLin-Sca-1+ cell will form a clone; p will beinfluenced by factors such as heterogeneity in the clonogenicpotential ofThy-11oLin-Sca-l+ cells and seeding efficiency toappropriate sites in the bone marrow. We calculated that 1 of13 sorted Thy-11OLin-Sca-l+ cells forms a clone comprising>1% of circulating white blood cells 3-7 weeks after injectioninto irradiated hosts. The probability that marked cells in the

U

S

CZUto:

No. of cells injected0 10 20

80

60 E50 E40 E

30

20

101

30

FIG. 1. Limiting-dilution analysis of Ly-5-marked Thy-1IoLin-Sca-1+ cells. The proportion of mice that were negative forLy-5-marked repopulating cells is plotted against the number ofmarked Thy-11OLin-Sca-1+ cells injected. Total numbers of micecontributing to each data point are as follows: 36 for 5 cells injected,14 for 10 cells injected, and 5 for 20 cells injected.

Developmental Biology: Smith et al.

Dow

nloa

ded

by g

uest

on

June

28,

202

1

2790 Developmental Biology: Smith et al.

Table 3. Data for positive mice injected with one or five Ly-5-marked Thy-11OLin-Sca-l+ cells

Last timeat which GM

Peak size, lineages wereMouse Lineage(s) % detected, weeks

Mice injected with five cellsExp. 1

1359

Exp. 2134

Exp. 345

Exp. 4

T, B, GMB, GMT, B, GMT, B, GM

BBT, B, GM

T, B, GMT, B, GM

13

2911

661235

22

11 38

108

2 T, GM 311 T, B, GM 2914 T, B, GM 6516 GM 2

Mice injected with single cellsAS T, B 1H2 B, GM 142 B, GM 163 T, B, GM 541 T, B, GM 1111 T, B, GM 26C7 T, B, GM 57

1111

4694

55766

GM, granulocyte/monocyte.

peripheral blood of positive mice are clonally derived from asingle Thy-1lOLin-Sca-l+ cell can also be determined: 77%for the group injected with 5 cells. In addition, we estimatedthat the mean number of clones contributing to repopulationofthe hematopoietic system 3-7 weeks after irradiation when105 Thy-11OLin-Sca-1+ cells (marked plus unmarked) wereinjected was 8.A modified experiment was also carried out, in which

single Thy-11OLin-Sca-1+ cells were injected instead of 5-20.In this case, we could be certain that marked progenyrepresented a clone derived from a single Thy-11OLin-Sca-l+cell. Single Thy-11OLin-Sca-l+ cells were isolated by meansof a cloning device on the FACS (see Materials and Meth-ods), and each cell was coinjected with 100 Thy-1loLin-Sca-1l cells from congenic donors into a C57BL/Kahost. Of 280 recipients screened in three independent exper-iments, we recovered 7 "positive" mice with detectable(>1%) progeny from the single Thy-11OLin-Sca-1+ cell intheir peripheral blood. This frequency, 1/40, is approxi-mately one-third the frequency of 1/13 observed in thelimiting-dilution experiments, probably due to (i) a slightlylower frequency of stem-cell activity in preparations ofThy-1l°Lin-Sca-1+ cells from the donors used for these

Proc. Natl. Acad. Sci. USA 88 (1991)

experiments (data not shown) and (it) lower efficienciesassociated with single-cell sorting into 96-well plates and lossof cells in the modified injection procedure (see Materialsand Methods).Table 3 summarizes data on the composition and life-span

ofthe clones obtained by single-cell injection and the putativeclones obtained after injection of five cells. The two types ofexperiments gave comparable spectra of repopulationevents, supporting the conclusion that the repopulation ob-served in most of the positive "five-cell" mice was indeedclonal. The clones varied widely in size. The peak clone size,usually reached at 4-6 weeks, varied from 65% of peripheralwhite blood cells for the largest clone to 1% for the smallest,which was the threshold of detection. Despite their variablesizes, the majority of the clones included all lineages welooked for (B and T lymphocytes, granulocytes, and mono-cytes), and all but three included at least two lineages. Anexample of a multilineage clone is shown in Fig. 2. Lymphnode cells from single-cell mouse C7 were analyzed 8 weekspostinjection. The contour plots demonstrate that C7 lymphnodes contained donor cells that were B220' (B cells), GR-1+and/or Mac-1+ (granulocytes and monocytes), and CD4+ orCD8+ (T cells). The only clones that appeared to lack somelineages were very small, so it is possible that the "missing"lineages fell below the limit of detection but were neverthe-less present. Thus we conclude that most, if not all, Thy-1loLin-Sca-1+ cells whose progeny are detected in this assayare capable of both myeloid and lymphoid differentiation.The clones also exhibited variable life-spans. In assessing

clone life-span, we concentrated on myeloid lineages becausethey are shorter-lived than lymphocytes, circulating on av-erage 25%, also exhibited short life-spans (

Proc. Natl. Acad. Sci. USA 88 (1991) 2791

V

u

:

100 300 1000 1020 3 50100 of +Buc C50 .~~~~~~U 40 +

30 co 10. cE ~~~~~~20*+ +20 t * *

103001000 0 ~ ~ ~~~101003W low ~ 100300 1000 100300 1000

No. of unmarked cells injected with 25 marked cells

FIG. 3. Competitive repopulation experiments. (A) The proportion of mice that were negative for Ly-5-marked repopulating cells is plottedagainst the number ofunmarked Thy-loLin-Sca-l+ cells coinjected with 25 marked Thy-1IOLin-Sca-1+ cells. Total numbers ofmice contributingto each data point are as follows: 34 for 1000 unmarked cells injected, 53 for 300 unmarked cells injected, and 42 for 100 unmarked cells injected.(B and C) The proportion of marked cells in peripheral blood for each positive mouse in one experiment (4-week and 8-week time points,respectively) is shown as a function of the number of unmarked Thy-loLin-Sca-l+ cells injected; each symbol represents a single mouse.

eloid cells by 8 weeks, showed no significant repopulation ofany secondary hosts. We conclude that while the majority ofThy-lILin-Sca-1+ BMCs whose progeny can be detected inthis in vivo repopulation assay are pluripotential, only asubset undergo extensive self-renewal.The conclusions of the experiments presented so far are

largely consistent with work by others demonstrating that thehematopoietic system can be repopulated by small numbersof stem cells after irradiation (7-10). Oligoclonal repopulationhas usually been interpreted in terms ofa selective process inwhich a few stem cells are active at a time, while others areheld in reserve. However, it is possible that oligoclonalrepopulation simply results from the engraftment of a limitednumber of stem cells. To address this possibility, we testedthe effect of stem-cell dose on the number of clones contrib-uting to hematopoiesis and on their sizes. If the total numberof clones contributing to hematopoiesis is relatively fixed,then increasing the size of the available stem-cell pool shouldreduce the probability that any individual stem cell will giverise to mature progeny. For a given limiting number ofmarked stem cells in the pool, this would result in a lowerfrequency of positive animals as the number of unmarkedcompetitors increases, while individual clone sizes shouldremain unchanged. Alternatively, if the number of contrib-uting clones increases in proportion to the stem-cell dose,then the frequency should not change, but clone sizes shouldgenerally decrease. These predictions were directly tested.Competitive repopulation experiments were carried out inwhich the number of marked Thy-11oLin-Sca-l+ cells washeld constant at 25 cells while the number of unmarkedThy-1IoLin-Sca-1+ cells was varied from 100 to 1000. Thedose of 25 donor cells was chosen because this numberreliably gave 60-70% positive hosts when mixed with 100other stem cells, and thus would allow measurement of anydecreases in frequency in the presence of additional un-marked stem cells, as well as an analysis of the distributionof clone sizes. At this dose =50% of the positive animals areexpected to have more than one marked clone contributing torepopulation. Pooled results of three independent experi-ments (Fig. 3A) demonstrate that increasing the total cell dosefrom 125 to 1025 does not alter the proportion of hosts inwhich one or more of the 25 marked Thy-11OLin-Sca-1+ cellshave contributed to hematopoiesis. This indicates that thenumber of stem cells initially contributing to repopulation ofthe hematopoietic system after irradiation is not fixed butincreases with the stem-cell dose. Assuming that marked andunmarked Thy-1l°Lin-Sca-1+ cells have equal probabilitiesofforming clones, we can estimate the mean numbers of stemcells contributing to hematopoiesis at each cell dose: 10 at125, 25 at 325, and 79 at 1025. The effect of cell dose on clonesize is shown for one of the experiments in Fig. 3B (at the4-week time point) and Fig. 3C (8 weeks). While the propor-tion of marked cells in positive mice for each cell dose isbroadly distributed, the three distributions are significantly

different at both time points (P < 0.05), showing an overalltrend toward smaller clones at higher cell doses.

DISCUSSIONLimiting-dilution and single-cell experiments were done tostudy the clonal organization of hematopoietic repopulationby Thy-11oLin-Sca-1+ cells from adult mouse bone marrow.The binomial equation was used to calculate three parame-ters. (i) The frequency of Thy-11oLin-Sca-1+ cells formingclones large enough to be measured in our assay (>1% ofperipheral white blood cells 3-7 weeks after injection) is1/13. (ii) The probability that marked progeny obtainedfollowing injection of five Ly-5-marked Thy-11oLin-Sca-1+cells represent a clone is 77%. (iii) The mean number ofclones (marked and unmarked) initially contributing to he-matopoiesis when 105 Thy-1IoLin-Sca-1+ cells are injected is8. While the Poisson equation is commonly used to analyzedata from limiting dilution experiments, we have used thebinomial equation because it is more appropriate in thissituation, where the number of cells being tested is small(5-20 cells) and the frequency of a positive response is high(1/13). However, application of either the binomial or thePoisson equation requires the unverified assumption thatindividual Thy-1loLin-Sca-1+ BMCs have a fixed probabilityof producing a measurable clone that is independent of whatother stem cells do, marked or unmarked. While we cannotprove that this assumption is correct, our finding that thenumber of clones initially contributing to hematopoiesis afterirradiation increases in direct proportion to the number ofThy-11OLin-Sca-1+ cells injected supports the assumptionthat the probability of any one Thy-11oLin-Sca-1+ cell con-tributing is fixed, -1/13, regardless of the cell dose. That therepopulating clones obtained after injection of single Thy-110Lin-Sca-1+ cells showed the same range of properties (Table3) further supports the conclusion that we were indeedobserving clonal repopulation in most positive mice injectedwith five cells.The existence of rare stem cells capable of giving rise to

both lymphoid and myeloid lineages has been known forsome time (20-24). In this report we have demonstrated thatthe majority of Thy-11OLin-Sca-1+ cells whose progeny aredetected in the in vivo repopulation assay described have thisability. Other protocols have been defined for the enrichmentof stem cells from adult bone marrow or fetal liver, using avariety of assays for stem-cell activity, including the abilityto form discrete myeloerythroid colonies in the spleen ofirradiated recipients, ability to protect mice from lethalirradiation, and ability to repopulate hematopoietic tissuesafter lethal irradiation (25-30). The variety of assays usedcreates some difficulty in comparing stem-cell-enriched pop-ulations prepared by different methods, but they are likely tobe overlapping populations (reviewed in ref. 31). In somecases, the capacity ofother stem-cell-enriched populations to

Developmental Biology: Smith et al.

Dow

nloa

ded

by g

uest

on

June

28,

202

1

2792 Developmental Biology: Smith et al.

contribute to lymphoid lineages has been tested (28-30), andin two cases, clonal markers revealed that single cells in thesestem-cell-enriched populations had contributed to both lym-phoid and myeloid lineages (29, 30). However, in neither casewas it possible to determine the frequency of pluripotentialcells in the stem-cell-enriched population. To our knowledge,no other groups have demonstrated that single cells ofdefined phenotype can undergo myeloid and lymphoid dif-ferentiation at high frequency.The Thy-11oLin-Sca-l+ cells formed clones of varying size

and life-span (Table 3). Clonal life-span was not tightlycorrelated with clone size. It is perhaps surprising that manymultilineage clones were nevertheless very short-lived. How-ever, in this respect and others, our results are largely inagreement with recent studies on the clonal organization ofhematopoiesis, in which unique retroviral integration siteswere used to mark clones of PHSC progeny (9, 10, 32). Inthose studies, the number of repopulating clones was variablebut relatively small (on the order of a few to a dozen). It wasalso found that the clones varied in size and life-span, somedying out within the first few months after irradiationwhereas others persisted for many months. The spectrum ofclones derived from Thy-1loLin-Sca-1+ cells appears to becomparable to those marked by retroviral integration sites,and thus the bone marrow stem-cell population marked byretrovirus in those studies (9, 10, 32) is likely to be at leastoverlapping with, if not identical to, the Thy-11oLin-Sca-l+population.While 1 in 13 Thy-1lOLin-Sca-1+ cells produced a detect-

able clone, only about one-third of these could be classifiedas both multilineage and long-term (producing myeloid cellsfor .9 weeks; Table 3). Thus, 1 in 39 Thy-11OLin-Sca-1+ cellswas able to generate a long-term, multilineage clone in ourassay system. Since Thy-11OLin-Sca-1+ cells are present at afrequency of 0.02-0.05% in adult bone marrow, this corre-sponds to a frequency of 0.5-1.3 per 105 adult BMCs. Thisfrequency matches well with estimates ofPHSC frequency inadult mouse bone marrow based on calculation of the numberof clones contributing to long-term hematopoiesis after in-jection of irradiated recipients with graded numbers of adultBMCs: 1-3 per 105 (3, 4). Estimates ofPHSC frequency basedon other assays, such as the ability to protect mice from lethaldoses of irradiation (5) or to "cure" genetically anemicWIWI mice (6), have led to higher estimates of the frequencyof stem cells in adult bone marrow, but these assays may notbe exclusive for PHSCs with long-term repopulating ability.The formation of multilineage long-term clones by 1 in 39Thy-11OLin-Sca-1+ cells is thus consistent with the hypoth-esis that all true PHSCs in normal adult bone marrow belongto this population.The number of clones initially contributing to hematopoie-

sis after irradiation increased in direct proportion to thenumber of Thy-1loLin-Sca-1+ cells injected (Fig. 3). Thisfinding is in complete agreement with the conclusions ofprevious studies in which the number of clones contributingto hematopoiesis after irradiation was found to increase indirect proportion to the number of BMCs injected (3, 4, 33,34). Thus, in this respect Thy-1lOLin-Sca-1+ cells behave ina manner previously described for repopulating stem cells inunseparated bone marrow. This finding argues stronglyagainst the possibility that repopulation of the hematopoieticsystem following irradiation results from the selection of arelatively fixed, small number of stem cells from the totalavailable pool, regardless ofhow many are available. Indeed,this finding raises the intriguing question of how the numberof stem cells contributing to hematopoiesis is determined.

Libuska Jerabek for tireless moral and technical support, and mem-bers of the laboratory for comments on the manuscript. L.G.S. is aBurroughs Wellcome Fund Fellow of the Life Sciences ResearchFoundation. S.H. is a Leukemia Society Fellow. This research wassupported by National Institutes of Health Grant CA42551 and theHoward Hughes Medical Institute.

1. Till, J. E. & McCulloch, E. A. (1980) Biochim. Biophys. Acta605, 431-459.

2. Boggs, D. R. & Winkelstein, A. (1983) White Cell Manual(Davis, Philadelphia), 4th Ed.

3. Micklem, H. S., Lennon, J. E., Ansell, J. D. & Gray, R. A.(1987) Exp. Hematol. 15, 241-257.

4. Harrison, D. E., Astle, C. M. & Lerner, C. (1988) Proc. Natl.Acad. Sci. USA 85, 822-826.

5. Micklem, H. S. & Loutit, J. F. (1966) Tissue Grafting andRadiation (Academic, New York).

6. Boggs, D. R., Boggs, S. S., Saxe, D. F. & Gress, L. A. (1982)J. Clin. Invest. 70, 242-253.

7. Micklem, H. S., Ansell, J. D., Wayman, J. E. & Forrester, L.(1983) in Progress in Immunology, eds. Yamamura, Y. & Tada,T. (Academic, Tokyo), Vol. 5, pp. 633-644.

8. Mintz, B., Anthony, K. & Litwin, S. (1984) Proc. Natl. Acad.Sci. USA 81, 7835-7839.

9. Jordan, C. T. & Lemischka, I. R. (1990) Genes Dev. 4, 220-232.

10. Keller, G. & Snodgrass, R. (1990) J. Exp. Med. 171, 1407-1418.11. Muller-Sieburg, C. E., Whitlock, C. A. & Weissman, I. L.

(1986) Cell 44, 653-662.12. Spangrude, G. J., Heimfeld, S. & Weissman, I. L. (1988)

Science 241, 58-62.13. van de Rijn, M., Heimfeld, S. & Weissman, I. L. (1989) Proc.

Natl. Acad. Sci. USA 86, 4634-4638.14. Heimfeld, S., Spangrude, G. J., Smith, L. G. & Weissman,

I. L. (1991) J. Exp. Med., in press.15. Parks, D. R., Lanier, L. L. & Herzenberg, L. A. (1984) in

Handbook of Experimental Immunology, eds. Weir, D. M.,Herzenberg, L. A. H., Blackwell, C. C. & Herzenberg, L. A.(Blackwell, Oxford), 4th Ed., pp. 29.1-29.21.

16. Muller-Sieburg, C. E., Townsend, K., Weissman, I. L. &Rennick, D. (1988) J. Exp. Med. 167, 1825-1835.

17. Spangrude, G. J. & Weissman, I. L. (1988) J. Immunol. 141,1877-1890.

18. Campbell, R. C. (1989) Statistics for Biologists (CambridgeUniv. Press, Cambridge, U.K.), 3rd Ed.

19. Scheid, M. P. & Triglia, D. (1979) Immunogenetics 9, 423-433.20. Wu, A. M., Till, J. E., Simonovitch, L. & McCulloch, E. A.

(1968) J. Exp. Med. 127, 455-464.21. Abramson, S., Miller, R. G. & Phillips, R. A. (1977) J. Exp.

Med. 145, 1567-1579.22. Keller, G., Paige, C., Gilboa, E. & Wagner, E. F. (1985) Nature

(London) 318, 149-154.23. Dick, J. E., Magli, M. C., Huszar, D., Phillips, R. A. &

Bernstein, A. (1985) Cell 42, 71-79.24. Lemischka, I. R., Raulet, D. H. & Mulligan, R. C. (1986) Cell

45, 917-927.25. Visser, J. W. M., Bauman, J. G. J., Mulder, A. H., Eliason,

J. F. & de Leeuw, A. M. (1984) J. Exp. Med. 59, 1576-1590.26. Lord, B. I. & Spooncer, E. (1986) Lymphokine Res. 5, 59-72.27. Ploemacher, R. E. & Brons, N. H. C. (1988) Exp. Hematol. 16,

21-26.28. Mulder, A. H., Visser, J. W. M. & van den Engh, G. J. (1985)

Exp. Hematol. 13, 768-775.29. Szilvassy, S. J., Fraser, C. C., Eaves, C. J., Lansdorp, P. M.,

Eaves, A. C. & Humphries, R. K. (1989) Proc. Nati. Acad. Sci.USA 86, 8798-8802.

30. Jordan, C. T., McKearn, J. P. & Lemischka, I. R. (1990) Cell61, 953-963.

31. Spangrude, G. J. (1989) Immunol. Today 10, 344-350.32. Capel, B., Hawley, R. G. & Mintz, B. (1990) Blood 75, 2267-

2270.33. Micklem, H. S., Ford, C. E., Evans, E. P. & Ogden, D. A.

(1975) Cell Tissue Kinet. 8, 219-232.34. Wallis, V. J., Leuchars, E., Chwalinski, S. & Davies, A. J. S.

We thank Moncrieff H. Smith, Jr., for consultations on statistics,

Proc. Natl. Acad. Sci. USA 88 (1991)

(1975) Transplantation 19, 2-12.

Dow

nloa

ded

by g

uest

on

June

28,

202

1