granulocyte colony-stimulating factor in vivo

Proc. Nati. Acad. Sci. USA Vol. 86, pp. 9499-9503, December 1989 Medical Sciences

The kinetics of human granulopoiesis following treatment with granulocyte colony-stimulating factor in vivo B. I. LORD*, M. H. BRONCHUDt, S. OWENSt, J. CHANG§, A. HOWELLt, L. SOUZA¶, AND T. M. DEXTER* *Department of Experimental Haematology, Paterson Institute for Cancer Research, tCancer Research Campaign Department of Medical Oncology, tRegional Department of Medical Physics and Bioengineering, and §Department of Haematology, Christie Hospital and Holt Radium Institute, Manchester M20 9BX, United Kingdom; and 1Amgen Inc., Thousand Oaks, CA 91320

Communicated by Eugene P. Cronkite, July 24, 1989

leneamineoxime (9mTc-HMPAO) (8) and observing their fate with a y camera.

ABSTRACT Cell proliferation in the bone marrow and blood of two patients with metastatic breast cancer who were treated with granulocyte colony-stimulating factor was studied by using [3H]thymidine labeling and autoradiography. Additionally, the fate of neutrophils labeled with "Tc-hexamethylpropyleneamineoxime was observed following granulocyte colony-stimulating factor infusion. Proliferation increased in all stages of granulopoiesis, but a significant amount of the increased production stemmed from a greater input to the myeloblast compartment. Changes in the myelogram combined with the increased labeling indicated a faster throughput of cells, which resulted in labeled cells appearing in the circulation within 1 day compared to the normal 4 or 5 days. The ""'Tc studies demonstrated no sequestration of circulating neutrophils by spleen, lungs, or liver. The halflife of the circulating neutrophils was not significantly changed, and calculations from the flow of labeled cells to the peripheral blood indicated an increase of 3.2 extra amplification divisions during neutrophil development. The dramatic neutrophil response to granulocyte colony-stimulating factor can therefore be accommodated by a relatively modest increase in granulopoietic activity.

MATERIALS AND METHODS Patients and Therapy. Two patients with metastatic breast cancer, age 34 (B.M.) and 41 (R.L.), were studied. Both had received previous radiotherapy (but no chemotherapy) following mastectomy. They were initially treated with rhGCSF (Amgen) continuously infused as previously described (5) for 3 and 5 days, respectively. On the seventh day after starting rhG-CSF therapy, they received doxorubicin (125 mg m-2) followed 24 hr later by further rhG-CSF for 11 days to reduce the period of neutropenia. Both patients received a "priming rhG-CSF bolus" of 2 ,ug/kg when the infusion was started to compensate for the "dead space" of the central venous line. The infusion pump (CADD-1, Pharmacia) was programmed to give 5 ,pg/kg per day of G-CSF to the first patient and 10 ,ug/kg per day to the second patient. Peripheral blood counts were taken before the infusion of growth factor and at 2, 3, 5, 10, 15, 24, and 50 min after the start of the infusion. Both patients gave informed consent. The study protocol was approved by the South Manchester ethical committee. [3H]Thymidine Labeling Studies. Patient I (B.M.). Two hours before starting the initial infusion of rhG-CSF, a bone marrow sample was obtained from the posterior iliac crest. After removal of red cells by methylcellulose sedimentation, the marrow cells were washed and resuspended in Iscove's medium at 5 x 106 cells per ml. Two-tenths milliliter of [3H]thymidine (18.5 kBq; specific activity, 250 GBq/mmol) in isotonic medium was added to the suspension and incubated for 30 min at 37°C. Cytospin preparations were then made and fixed in absolute methanol. Forty-eight hours after commencing the infusion of rhG-CSF, a second bone marrow sample was taken and similarly labeled. At 71 hr, 555 MBq of [3H]thymidine (9.0 kBq/g) in 15 ml of isotonic medium was injected intravenously by means of the central line. One hour later, a further marrow sample was taken and cytospin preparations were made for autoradiography. At this stage, the rhG-CSF infusion was stopped, and venous blood samples were subsequently taken at 24-hr intervals. Erythrocytes were removed over Ficoll at p = 1.077 g ml-1, and cytospins of the peripheral leukocytes were prepared for autoradiog-

The recent availability of recombinant human granulocyte colony-stimulating factor (rhG-CSF) (1) has stimulated considerable interest in its potential applications for promoting hemopoietic regeneration following bone marrow damage by cytoreductive agents. Stimulation, in vivo, of granulopoiesis by rhG-CSF was demonstrated in mice (2) and in humans (3-7). In patients, 2 or 3 days of continuous infusion or repeated injections of rhG-CSF resulted in a 10-fold increase in peripheral neutrophil levels, which were maintained for the duration of infusion and, following chemotherapy, significantly shortened the duration of the subsequent neutropenia (3, 5-7). These neutrophils demonstrated normal function and competence (4). Only a small, nonsignificant increase in granulocyte/macrophage colony-forming cell (GM-CFC) cycling and in the GM-CFC/burst-forming unit-erythroid ratio was observed, suggesting that most of the expansion in neutrophil production probably arises after the GM-CFC stage (4). Our intention was to investigate the changes in kinetics induced by G-CSF and required to maintain this elevated neutrophilia. We therefore treated two patients with rhG-CSF and, by means of autoradiography, studied the kinetics of marrow cell proliferation and efflux into the peripheral blood following labeling in vivo with tritiumlabeled thymidine. Since rhG-CSF in vivo initially produces an early fall in peripheral neutrophils followed by rapid influx of mature neutrophils into the circulatory pool (4), we also studied the early effects of this growth factor on circulating neutrophils by labeling them with 99mTc-hexamethylpropy-

raphy. Patient 2 (R.L.). Bone marrow and blood cell preparations were made following the same labeling sequence in relation to G-CSF administration as described for patient 1. In this case, however, the initial G-CSF infusion was continued for an additional 2 days after labeling in vivo with [3H]thymidine. Abbreviations: G-CSF, granulocyte colony-stimulating factor; rhGCSF, recombinant human G-CSF; GM-CFC, granulocyte/ macrophage colony-forming cells; MB, myeloblasts; MMC, metamyelocyte/band cells; PMC, promyelocytes; MC, myelocytes; HMPAO, hexamethylpropyleneamineoxime.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. 9499

9500

Medical Sciences: Lord et al.

Autoradiography. Autoradiographs were prepared by using Ilford K5 liquid emulsion. They were dried and stored at 40C for 5 weeks (3 weeks for patient 2). The slides were then developed, fixed, and stained with May Grunwald/Giemsa stains. Differential cell counts, percent labeled cells, and grain counts over granulocytic cells were determined from the autoradiographs; at least 500 cells were counted on each slide. Under these conditions, mean overall grain counts were of the order of 75 per cell, with few cells having less than 10 grains. A threshold labeling criterion was therefore set at 5 grains. 99mTc Labeling of Leukocytes. Leukocytes were isolated under aseptic conditions from a 100-ml sample of venous blood of patient 2 (R.L.) and labeled with 99mTc-HMPAO (8). This lipophilic compound is retained specifically inside neutrophils. Erythrocytes were separated by sedimentation in 6% dextran 70/5% dextrose for 1 hr. The plasma and leukocyte fraction was then centrifuged at 150 x g for 5 min to isolate the leukocytes from the platelet-rich plasma. This was removed and the cells were incubated with 99mTc-HMPAO. A differential count was obtained before and after labeling. A separate 25-ml sample of venous blood was centrifuged at 2000 x g for 10 min to obtain platelet-poor plasma for washing the white cells and reconstituting the injection. Elution of the radiolabel was checked in vitro at 370C for up to 24 hr. The percentage of the activity remaining bound to the cells was calculated at various times up to 24 hr. Fifty-microliter aliquots of the resuspended cells were also taken at each time point, and viable cells were counted on a hemocytometer. Imaging. After resuspension in 5 ml of platelet-poor plasma, the leukocytes (---2.6 x 108 neutrophils) labeled with 125 MBq of 99mTc were injected into an antecubital vein by using a 19-gauge cannula. Imaging was performed by using a large field-of-view y camera and a computer system (Digicamera Series 480, Scintronix, Livingston, Scotland) with the patient supine and the camera positioned posteriorly to view the lungs, liver, and spleen. Each time labeled cells were administered, an initial dynamic study was acquired consisting of 30 frames at 60-sec intervals. The activity versus time curve, obtained from a region of interest over the right lung, was used to calculate the 50o washout time. For the first study, carried out before and during the first administration of rhG-CSF, a second dynamic acquisition was performed 2 hr after injection of the cells. It consisted of 45 frames at 60-sec intervals and started 10 min before the commencement of rhG-CSF treatment. For the second study, performed 18 hr following the second cycle of doxorubicin, the second dynamic study was commenced 9 hr after administration of the labeled cells. For each of the delayed dynamics, regions of interest drawn for the lungs, mediastinum, liver, and spleen were used to monitor activity. Activity versus time curves, corrected for decay of 99"Tc and normalized to 100%, were analyzed by the cumulative sum test (9) to detect any significant deviations from the predicted values.

RESULTS Patient 1. Bone marrow. Differential labeling data are shown in Fig. 1; the proportion of each cell type was corrected for a net 20% increase in marrow cellularity as previously reported (4). Prior to rhG-CSF infusion, 18% of all proliferative granulocytic cells were labeled (in vitro labeling). This comprised 44% labeling of the myeloblasts (MB) and promyelocytes (PMC), 17% of myelocytes (MC), and 3% of metamyelocyte/band (MMC) cells. After 48 and 71 hr of continuous rhG-CSF infusion, the labeling index (in vitro and in vivo, respectively) had increased in the MC and MMC cell compartments, indicating an increased rate of proliferation, but there was little significant change in labeling of the early

Proc. Natl. Acad. Sci. USA 86

(1989)

A

12001 E z I C-I

Ca

a: Cl a)

.0

MB+PMC

al

B 4-J

CD

a) a) a)

:I

MB+PMC

MC

MMC

GC

E+L

TOTAL

FIG. 1. Bone marrow differentials and [3H]thymidine labeling data for two patients. (A) Patient 1 (B.M.). (B) Patient 2 (R.L.). Distribution of the marrow cell populations (i) before treatment (first column of each triplet) and (ii) 48 hr (second column) and (iii) 71 hr (third column) after the start of rhG-CSF. The crosshatched portion of each column and the numbers surmounting them indicate the relative number and percentage, respectively, of each cell type labeled with [3H]thymidine. GC, proliferating granulocytic cells; E + L, erythroid + lymphoid cells; total, all bone marrow cells. [3H]Thymidine labeling was in vitro for the first two columns of each triplet and in vivo for the third column. Relative cell numbers were calculated from the percentage differential cell counts assuming the previously observed increase of about 20o in the net marrow

cellularity (4).

cells. Grain counts over MB/PMC and MC increased from pretreatment values of 144 ± 91 (mean ± SD) and 55 ± 28, respectively, to 189 ± 37 and 114 ± 105 after 48 hr of rhG-CSF treatment. The in vivo labeling indices obtained at 3 days compared well with those obtained in vitro at 2 days. Thus, although the grain counts were some 50% lower, the in vivo labeling efficiency was satisfactory, and valid comparisons between the two can be drawn. MB/PMC numbers doubled under the influence of rhGCSF, but the size of the MC compartment was unchanged. MMC numbers remained unchanged over 2 days but increased over the third day. In spite of little change in the labeling index, the increased grain count over the early myeloid cells suggests more proliferative activity leading to population amplification in these early compartments, but a greater input of cells from the progenitor into the MB compartment cannot be ruled out. The increased proliferation rate of the MC and MMC is balanced by a faster maturation rate, which maintains a stable MC population.

Medical Sciences: Lord et al.

Proc. Natl. Acad. Sci. USA 86 (1989)

The numbers of erythroid and lymphoid cells fell about 25%, and a moderate compensatory increase in the proliferation rate was seen (labeling index, Fig. 1). The granulocytic/ erythroid cell ratio increased 2.7-fold, due almost entirely to increased granulopoiesis. Blood. Within 24 hr of starting rhG-CSF infusion, the leukocyte count increased from 4 x 106/ml to 19 x 106/ml (Fig. 2A), indicating an early release of the mature forms. This increase continued for 1 day after stopping the infusion and increased to a peak of 41 x 106/ml before falling rapidly toward normal levels. Greater than 80% of this wave was neutrophilic. Early release was maintained under the influence of G-CSF; labeled neutrophils appeared in the blood by 24 hr and reached a maximum of 30o labeled by 48 hr (Fig. 2B). This wave of labeling then declined, probably only coincidentally with chemotherapy since the cellularity was already waning rapidly (Fig. 2A) but rose again on restimulating with rhG-CSF 1 day after starting chemotherapy (Fig. 2B). This second wave of labeling presumably arose from cells labeled at a very early stage of granulopoiesis, possibly the GM-CFC stage, since otherwise a single continuous wave would have been expected. Patient 2. Bone marrow. A similar picture was obtained from this second patient (Fig. 1). Pretreatment granulocytic cell differentials and labeling were comparable to patient 1, but there was a relatively larger erythroid/lymphoid component. In this case, the increase in MB/PMC numbers was less dramatic, but the labeling indices were considerably increased. MC were increased (in contrast with patient 1) by over 50%, and their labeling index increased from 15 to 43%. By 3 days, the erythroid/lymphoid components were severely depressed, but this was again compensated by some A

9501

increase in proliferative activity; 15% were labeled. The granulocytic/erythroid cell ratio was increased 8-fold. MB/ PMC grain counts increased from 32 + 9 (mean ± SD) to 47 ± 18, and those of MC increased from 24 ± 11 to 47 ± 18. These data suggest an increased amplification by way of increased proliferation throughout the maturation sequence, possibly related partly to the higher dose of rhG-CSF used in this patient. Blood. As with patient 1, the neutrophil count rose rapidly within 24 hr, peaking after 4 days of rhG-CSF treatment at 32 x 106/ml (Fig. 2C). Again, labeled cells appeared in the blood by 24 hr, but the wave of labeling was prolonged due to the extended rhG-CSF treatment and returned to zero at day 4 (Fig. 2D). ""'Tc Labeling Studies. From the curves of lung activity upon injection of leukocytes labeled with `mTc, the 50% washout time was calculated to be 550 sec for the first study and 590 sec for the second study. No significant holdup in the lung was observed. Following injection, therefore, labeled cells passed rapidly through the lungs and equilibrated between the circulating and marginating neutrophil pools of the liver and the spleen within 5-30 min. In vitro elution rates were fairly constant in different experiments and became marked at 9 hr (up to 45%). The viability of the labeled cells at this time, as assessed by trypan blue exclusion, was >95%. Infusion of G-CSF was followed by a rapid and transient decrease in circulating neutrophils to less than 500 mlbetween 5 and 10 min, with a return to pretreatment levels within 1 hr. This phenomenon has already been described for both G-CSF (4) and GM-CSF (10) and, unlike the early leukopenia often observed after endotoxin administration (11), it affects only neutrophils. In both studies, the curves B 3 HTdR

ID x 0

rE U)

Cu

Days Post 3HTdR Labelling C

D

CD

x

30

-

'-.

20

url U1

-2

-1

0

1

2

3

4

Days Post 3HTdR Labelling

Days Post 3HTdR Labelling

FIG. 2. (A and B) Patient 1 (B.M.). (C and D) Patient 2 (R.L.). (A and C) Labeled neutrophils (LNC), total neutrophils (NC), and leukocyte (WBC) counts in the peripheral blood with respect to labeling with [3H]thymidine (3HTdR) at time 0. (B and D) Percentage of labeled neutrophils entering the blood. The arrows indicate the application and duration of rhG-CSF, [3H]thymidine, and chemotherapy (CT). =

9502

Medical Sciences: Lord et al.

Proc. Natl. Acad. Sci. USA 86

from the delayed (second) dynamic study showed a decrease in radioactivity in the spleen 8-15 min after administration of rhG-CSF (Fig. 3). This was statistically significant (>3 SD) and suggested "egression" of marginated neutrophils from the spleen. It was also observed, but was less marked, when G-CSF was administered 9 hr after the infusion of labeled cells, in spite of considerable elution of the radiolabel (data not shown). No equivalent change in liver or lung activity was seen and neither was the preferential uptake by the lungs as described for GM-CSF (10) seen.

DISCUSSION The principles of analyzing cell kinetics using [3H]thymidine were established in the late 1950s, and the basic kinetic parameters of blood cell production were established in normal individuals (12-16). We have now used a similar approach to examine hemopoietic cell kinetics following infusion of a myeloid cell growth factor, which is known to stimulate abnormally high levels of neutrophil production both in experimental animals (2, 17, 18) and in humans (3-7). Since, for ethical reasons, only two patients were investigated, no statistical analysis of the results is warranted. On the other hand, some rather crude calculations (see below) based on the changes in marrow differential cell counts and the movement of [3H]thymidine-labeled cells through the marrow and into the circulation were sufficient to illustrate a basic mechanism governing the responses to rhG-CSF. In both patients, hemopoiesis before administration of rhG-CSF, as judged by marrow differentials and blood counts, was normal and there was no evidence of marrow involvement by carcinoma. A 2-day treatment with rhG-CSF in patient 1 (B.M.) increased the numbers of MB and PMC,

100-

-

s

v

iv

Spleen

75-I

0 0

0x

E E100 co

~~~~~~~Liver

Q

c

c

c

co

CZ

c)

0)

0 8

E 0

75-

C

II.

0

100-

i

Lung

75-

4 0

30

15

45

Min. FIG. 3. Organ-specific distribution of 99mTc-HMPAO activity before and after the rhG-CSF priming bolus. Activity values have been corrected for 99mTc decay and normalized to 100%1. Data shown are for the second dynamic of the first study, taken at 2 hr after the injection of cells. Data for the second dynamic of the second study (at 9 hr) are not shown, but they present the same picture.

(1989)

suggesting an increased input from the progenitor compartment into the recognizable granulocytic maturation phase and/or an early increased rate and maintenance of proliferation in these compartments. This latter mechanism may be the more important since the autoradiographic grain counts were also increased and a previous study (4) indicated only a small increase in committed progenitor cell proliferation. The lack of change in MC numbers, coupled with the increased labeling index and grain count-comparable to those in studies on dogs (19)-however, suggested that both their proliferation and maturation accelerated, a feature reinforced by the emergence of labeled granulocytes in the peripheral blood within 1 day. This is to be compared with a normal emergence time of 4 days, range 2-5 days (12, 15, 16). The increase in peripheral neutrophil levels within 24 hr of administering G-CSF (Fig. 2 and ref. 4) itself indicates an early release of stored, mature granulocytes from the marrow, but the equally rapid release of cells labeled subsequent to this release indicates a continued reduced marrow transit time of the maturing population. The second patient (R.L.) showed a similar pattern. Labeled cells again appeared in the circulation within 24 hr. Increased proliferation in the bone marrow was more dominant, suggesting that, in this case, increased proliferation through the maturation stages was the major response to G-CSF. The earlier conclusion that most of the expansion in cell numbers probably occurs after the GM-CFC stage (4) was therefore probably valid. Nevertheless, since the GMCFC/burst-forming unit-erythroid ratio was increased, some contribution from increased feed from the progenitor compartment could not be excluded. Although there was no question of infection in these patients and the G-CSF itself was free of pyrogen (3), some aspects of the increased granulopoiesis are analogous to those following an endotoxin injection. For example, serum from endotoxin-injected animals contains a substance (not the endotoxin itself) that induces neutrophil release (20). In humans, the MC-to-blood transit time may be as short as 48 hr during infection (15) and, in a number of recent cases, increased levels of G-CSF were detected during periods of bacterial infection, levels that fell corresponding to the elimination of infection (21). In mice, the generative cycle of early precursors is shortened in the face of increased granulocytopoietic stimulation (22). The induction of sterile abscesses in dogs (19) led to an initial granulopenia, the recruitment of a larger fraction of MC in DNA synthesis, and an increase in the rate of DNA synthesis in MC; the increase in grain count was comparable with that seen here. It is interesting to speculate, therefore, that the early release of mature cells may be the result of decreasing the endothelial coverage of the sinus wall by adventitious cells, such as is observed to occur in the marrow of animals given endotoxin (23). By using the principles established by Killmann et al. (14), the maximum mean blood transit time (bttmax) for the newly formed neutrophils is obtained from the 24-hr rate of increase of labeling index (ANL) as 100/ANL. This gives bttmax values of 141 and 80 hr for B.M. and R.L., respectively. These calculations assume that only labeled cells enter the blood, however, and a minimum value can be obtained from the peripheral half-life of the neutrophils as tl/2/0.693. From the rate of loss of labeled cells in the blood, the calculated half-lives were 10.2 and 5 hr for B.M. and R.L., respectively, thus giving minimum blood transit times of 14.7 and 7.2 hr. These are, however, underestimates since the calculation is strictly valid only for the most heavily labeled cells. Half-life measurements were comparable with other normal measurements (7-8 hr; refs. 24 and 25)-indicating that the granulocytosis is not simply the result of early release-and function of the stimulated neutrophils appeared normal (4). The rapid fall in absolute neutrophil count when rhG-CSF is stopped

Medical Sciences: Lord

et

al.

suggests that G-CSF interferes with the normal processes of margination, to maintain a relatively normal lifespan. However, the 9'9Tc labeling study indicated that there was no abnormal distribution of the neutrophils, and it appears reasonable to conclude that rhG-CSF stimulated neutrophils function normally in all respects. It is not possible to calculate the mean cell cycle time (tQ) of the proliferating granulocytic cells from the [3H]thymidine labeling index alone. However, assuming 13 hr for the DNA synthesis time (ts) (26, 27), cell cycle estimates of 40 and 33 hr, respectively, were calculated (tc = 100 x ts/% labeled cells). If ts is also reduced, as suggested by the increased grain counts over the myelocytic cells, the cell cycle times will be correspondingly less. These are to be compared to 48 hr for their normal counterparts (28) and, therefore, represent a considerable decrease in the cycle time and an increase, correspondingly, in the proliferation rate. The net enhancement in neutrophil kinetics can be obtained from the rate of increase in neutrophils and their half-clearance time. Thus, for patient 1 (B.M.), over 24 hr from the time of labeling (to), the increase in neutrophils was 7 x 106 cells per ml (Fig. 2A). With t1/2 = 10.2 hr, the neutrophil value of 30 x 106 cells per ml would have fallen to 6 x 106 cells per ml, so that the real output over 24 hr equals 31 x 106 cells per ml. By a remarkable coincidence, due to the shorter t1/2 of 5 hr, the 10 x 106 cells per ml increase in neutrophils in R.L. should be corrected also to 31 x 106 cells per ml (Fig. 2C). Assuming a normal neutrophil count of 3.75 x 106/ml and a t1/2 of 8 hr (25), then over 24 hr, 87% of the neutrophils (3.3 x 106/ml) would be lost. In normal steadystate conditions, this would equal the generation rate. The rhG-CSF-stimulated neutrophilia is therefore about 9.4 times the normal level of neutrophils, which is equivalent to 3.2 extra amplification divisions in its development sequence, some of which may occur prior to the myeloblast stage. At the same time, increased labeling in the marrow MMC pool indicates that proliferation has been extended further along the maturation sequence than normal, and this may also account for some of the extra divisions. The fact that there is not a consequential buildup of numbers in the compartment is compatible with that very early release into the circulation also suggested by previous studies (4). The increased production and release of neutrophils by the bone marrow together with a normal circulating half-life suggest that more neutrophils are migrating into tissues from the vascular compartment per unit time than under normal steady-state conditions. Since no preferential uptake of 9'9Tc-HMPAO-labeled neutrophils by important reticuloendothelial organs (spleen, liver, and lung) was found, their fate after leaving the circulation is not known with certainty. However, it is mainly in the tissues that these cells fulfill their functional role. Finally, it may be noted that this increase of 3.2 extra divisions is a relatively modest requirement, and the fears of some workers that the dramatic responses to G-CSF might place undue strain on the GM-CFC and the pluripotent stem

cell compartments are probably unfounded. We thank Mr. Gary Owen of the Department of Epithelial Biology for kindly preparing the autoradiographs. The work was supported by the Cancer Research Campaign and the Leukaemia Research Fund.

Proc. Natl. Acad. Sci. USA 86 (1989)

9503

1. Souza, L. M., Boone, T. C., Gabrilove, J., Lai, P. H., Zsebo, K. M., Murdoch, D. C., Chazin, V. R., Bruszewsski, J., Lu, H., Chen, K. K., Platzer, E. & Moore, M. A. S. (1986) Science 232, 61-65. 2. Moore, M. A. S. & Warren, D. J. (1987) Proc. Natl. Acad. Sci. USA 84, 7134-7138. 3. Bronchud, M. H., Scarffe, J. H., Thatcher, N., Crowther, D., Souza, L. M., Alton, N. K., Testa, N. G. & Dexter, T. M. (1987) Br. J. Cancer 56, 809-813. 4. Bronchud, M. H., Potter, M. R., Morgenstern, G., Blasco, M. J., Scarffe, J. H., Thatcher, N., Crowther, D., Souza, L. M., Alton, N. K., Testa, N. G. & Dexter, T. M. (1988) Br. J. Cancer 58, 64-69. 5. Bronchud, M. H., Howell, A., Crowther, D., Hopwood, P., Souza, L. & Dexter, T. M. (1989) Br. J. Cancer 60, 121-125. 6. Gabrilove, J. L., Jakubowski, A., Scher, H., Sternberg, C., Wong, G., Grous, J., Yagoda, A., Fain, K., Moore, M. A. S., Clarkson, B., Oettgen, H. F., Alton, K., Welte, K. & Souza, L. (1988) N. Engl. J. Med. 318, 1414-1422. 7. Morstyn, G., Campbell, L., Souza, L. M., Alton, N. K., Keech, J., Green, M., Sheridan, W., Metcalf, D. & Fox, R. (1988) Lancet i, 667-672. 8. Peters, A. M., Danpure, H. J., Osman, S., Hawker, R. J., Henderson, B. L., Hodgson, H. J., Kelly, J. D., Neirinckx, R. D. & Lavender, J. P. (1986) Lancet ii, 946-949. 9. Goldsmith, P. L. & Woodward, R. R. (1964) Cumulative Sum Techniques, ICI Monograph 3 (Oliver & Boyd, Edinburgh,

U.K.). 10. Devereux, S., Linch, D. C., Campos-Costa, D., Spittle, M. F. & Jelliffe, A. M. (1987) Lancet H, 1523-1524. 11. Craddock, C. G., Perry, S., Ventzki, L. G. & Lawrence, J. S. (1960) Blood 15, 840-855. 12. Cronkite, E. P., Fliedner, T. M., Bond, V. P. & Rubini, J. R. (1959) Ann. N.Y. Acad. Sci. 77, 803-820. 13. Cronkite, E. P., Bond, V. P., Fliedner, T. M., Rubini, J. R. & Killmann, S. A. (1960) Proc. Ninth Int. Cong. Radiol. 1, 894-905. 14. Killmann, S. A., Cronkite, E. P., Robertson, J. S., Fliedner, T. M. & Bond, V. P. (1963) Lab. Invest. 12, 671-684. 15. Fliedner, T. M., Cronkite, E. P. & Robertson, J. S. (1964) Blood 24, 402-414. 16. Fliedner, T. M., Cronkite, E. P., Killmann, S. A. & Bond, V. P. (1964) Blood 24, 683-700. 17. Welte, K., Bonilla, M. A., Gillio, A. P., Boone, T. C., Potter, G. K., Gabrilove, J. L., Moore, M. A. S., O'Reilly, R. J. & Souza, L. M. (1987) J. Exp. Med. 165, 941-948. 18. Cohen, A. M., Zsebo, K. M., Inoue, H., Hines, D., Boone, T. C., Chazin, V. R., Tsai, L., Ritch, T. & Souza, L. M. (1987) Proc. Natl. Acad. Sci. USA 84, 2484-2488. 19. Cronkite, E. P., Burlington, H., Chanana, A. D., Joel, D. D., Reinke, U. & Stevens, J. (1977) Exp. Hematol. Today 5, 41-49. 20. Boggs, D. R., Marsh, J. C., Chervenick, P. A., Cartwright, G. G. & Wintrobe, M. M. (1968) Proc. Soc. Exp. Biol. Med. 127, 689-693. 21. Watari, K., Asano, S., Shirafuji, N., Kodo, H., Ozawa, K., Takoku, F. & Kamachi, S. (1989) Blood 73, 117-122. 22. Niskanen, E., Tyler, W. S., Symann, M., Stohlman, F., Jr., & Howard, D. (1974) Blood 43, 23-31. 23. Weiss, L. (1970) Blood 36, 189-208. 24. Athens, J. W., Haab, 0. P., Raab, S. O., Mauer, A. M., Ashenbrucker, H., Cartwright, G. E. & Wintrobe, M. M. (1961) J. Clin. Invest. 40, 989-996. 25. Vincent, P. C. (1977) Clin. Haematol. 6, 695-716. 26. Stryckmans, P., Cronkite, E. P., Fache, F., Fliedner, T. M. & Ramos, J. (1966) Nature (London) 211, 717-720. 27. Vincent, P. C., Cronkite, E. P., Greenberg, M. L., Kirsten, C., Schiffer, L. M. & Stryckmans, P. A. (1969) Blood 33, 843-850. 28. Killman, S. A., Cronkite, E. P., Fliedner, T. M. & Bond, V. P. (1964) Blood 24, 267-280.