The Number of Nucleated Cells Reflects the Hematopoietic Content of Umbilical Cord Blood for Transplantation

Home , Hematopoietic stem cell

Bone Marrow Transplantation, (1999) 24, 965–970  1999 Stockton Press All rights reserved 0268–3369/99 $15.00 http://www.stockton-press.co.uk/bmt The number of nucleated cells reﬂects the hematopoietic content of umbilical cord blood for transplantation

FTH Lim1,2, JM v Beckhoven3, A Brand3, JC Kluin-Nelemans1, JMH Hermans4, R Willemze1, HHH Kanhai2 and JHF Falkenburg1

Departments of 1Hematology, 2Obstetrics and 4Medical Statistics from the Leiden University Medical Center; and 3Red Cross Blood Bank, Leidsenhage, The Netherlands

Summary: irradiated recipients. Since the colony-forming unit granulocyte–monocyte (CFU-GM) defined in a semi-solid A single umbilical cord blood (UCB) collection may con- culture assay, has been associated with hematopoietic tain sufficient hematopoietic stem cells to achieve engraftment of human bone marrow cells, this assay has engraftment and repopulation of the hematopoietic sys- been used as a quality control method for hematopoietic tem of children and adults after myeloablative therapy. stem cell grafts.12,13 The hematopoietic potential of a UCB unit is often The human hematopoietic stem cells reside in the frac- defined by the number of CD34+ cells or the number of tion of cells expressing the CD34 antigen.14,15 CD34+ cells colony-forming units as measured in semisolid hemato- can be identified by flow cytometry, but the various tech- poietic progenitor cell (HPC) cultures. However, these niques for CD34 enumeration that have been advocated assays are relatively difficult to standardize between may result in different estimated concentrations.16–18 Both UCB banks. The number of nucleated cells infused per the CD34+ cell content, and the numbers of hematopoietic kilogram body weight of the recipient is also reported progenitor cells (HPC) as determined in semi-solid medium to be a significant factor in the speed of recovery of culture assays are frequently used to predict the hematopoi- neutrophils and platelets after transplantation. To ana- etic potential of a stem cell transplant, and to estimate the lyze which parameters could be used to evaluate the viability of the stem cells after processing or cryopreserv- hematopoietic potential of a UCB graft, we evaluated ation. However, both the functional enumeration of HPC almost 300 UCB units that were collected for banking and the CD34+cells enumeration are difficult to standardize. for unrelated transplantation. A strong correlation was + The number of nucleated cells infused per kilogram of found between the frequencies of CD34 cells and the the recipient has been reported to be a significant factor HPC as measured in semi-solid medium cultures. From in the speed of recovery of neutrophils and platelets after the various leukocyte subpopulations, the concentration transplantation. In reports by Wagner et al19 on UCB trans- and total numbers of nucleated cells correlated best 20 + plants in 44 siblings, and Kurtzberg et al on 25 unrelated with both the HPC content and the number of CD34 UCB transplantations, a trend was found for the time to cells. Differentiation of these nucleated cells into subsets myeloid or platelet engraftment to decrease with the dose of leukocytes offered no advantage for better prediction + of clonogenic precursors or CD34 cells infused, but the of HPC or CD34+ cells. These results indicate that the correlations were not statistically significant. Gluckman et nucleated cell count probably reflects the hematopoietic al21 reported a significant correlation between the number potential of a UCB graft, and may for that reason corre- of nucleated cells per kg body weight infused and neutro- late with the speed of engraftment after transplantation. phil recovery. Recently Wagner et al22 showed a correlation Keywords: umbilical cord blood; transplantation; hema- + topoietic cells between the number of CD34 cells and the speed of recovery in 143 patients transplanted with unrelated UCB transplants. In this study, we evaluated the correlation between the Human umbilical cord blood (UCB) can be used as a source two methods of enumeration of HPC, and analyzed whether of hematopoietic stem cells for transplantation purposes.1 the nucleated cell counts or numbers of leukocyte subpopu- A single cord blood collection may contain enough hemato- lations correlated with the CD34+ cells or HPC content of poietic stem cells to achieve engraftment and repopulation UCB units. From the Cord Blood Bank of the Red Cross of the hematopoietic system of both children and adults.2–11 Blood Bank, Leidsenhage, which is part of the national Functionally, hematopoietic stem cells can be defined as umbilical cord blood bank EuroCord Nederland, we evalu- cells that are capable of lymphomyeloid reconstitution of ated data from the first 300 UCB units. We found that the number of nucleated cells correlated well with HPC and CD34 content of a UCB unit. Therefore we conclude that Correspondence: Dr JHF Falkenburg, Department of Hematology, Build- as long as the enumeration of CD34+ cells and cultured ing 1, C2-R, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands HPC from an UCB transplant are not standardized, the Received 24 November 1998; accepted 16 June 1999 number of nucleated cells can be used as an alternative to Hematopoietic content of umbilical cord blood FTH Lim et al 966 indicate the hematopoietic potential of UCB units for the gated CD14−/CD66e− population, the CD34+ cells could transplantation. be identified in a CD34 vs SSC dot plot (Figure 1). The sample that was stained with the mixture containing mouse- IgG1, was used as a negative control. In the calculation of + Materials and methods CD34 cells, the percentage of isotype control binding cells was subtracted from the percentage of CD34+ cells. Con- + Collection of umbilical cord blood centrations of CD34 cells and leukocyte subpopulations were calculated by multiplying the percentage with the After informed consent was obtained from the mother number of nucleated cells per ml UCB counted by the Sys- before delivery, UCB was collected from the umbilical cord mex. White blood cells (WBC) were defined as lympho- after the expulsion of the placenta. After puncture of the cytes, monocytes and granulocytes; mononuclear cells umbilical vein, the UCB was collected into a 250 ml collec- (MNC) were defined as lymphocytes and monocytes. tion bag containing 15 ml of Hanks’ balanced salt solution (BioWhittaker, Verviers, Belgium) with 75 IU/ml of pre- servative-free heparin as anticoagulant. Two samples of the Hematopoietic progenitor cell cultures UCB were sent to two different laboratories. The cell counts and phenotype were independently analyzed in one For the culture of HPC, UCB samples were diluted 10 times laboratory, while the HPC culture frequencies were defined in Iscove’s modified Dulbecco’s medium (IMDM) by the other laboratory. (BioWhittaker). Cells were seeded in three 35-mm tissue culture dishes with different concentrations of cord blood ␮ ␮ ␮ Cell counts and FACS analysis cells. Quantities of 10 l, 20 lor40 l of the diluted UCB were added to 1.5 ml of a semisolid culture-mixture Nucleated cells were counted using a Sysmex K1000 (Toa of IMDM, containing 30% fresh-frozen AB-heparin Medical Electronics, Kobe, Japan), which was validated in plasma, 0.5% deionized bovine serum albumin (BSA), quality rounds. human transferrin (0.47 g/l) (Behringwerke, Marburg, × −5 Phenotype analysis of umbilical cord blood was perfor- Germany) saturated with FeCl3-H2O, 5 10 m beta-mer- med within 36 h after harvest, using a Becton Dickinson captoethanol (Sigma Chemicals, St Louis, MO, USA), FACScan flow cytometer. For the determination of CD45 1.1% methylcellulose (Methocel 4000 cps; Fluka, Freiburg, and CD34 expressing cells, a method for whole blood coun- Germany), and the following human hematopoietic growth ting was used which was based on a protocol for CD34+ factors: recombinant granulocyte colony-stimulating factor cell enumeration with flow cytometry, described by the (G-CSF) 10 ng/ml, kindly provided by Amgen (Thousand national workshop SIHON.16 Oaks, CA, USA); granulocyte–macrophage colony-stimul- Three samples of 100 ␮l blood were labeled with combi- ating factor (GM-CSF) 10 ng/ml, Sandoz (Basel, nations of directly fluorescein isothiocyanate (FITC)- or Switzerland); stem cell factor (SCF) 50 ng/ml, kindly pro- phycoerythin (PE)-labeled monoclonal antibodies: CD45 vided by Amgen; interleukin-3 (IL-3) 25 ng/ml, a gift from FITC + CD14 PE for leukocyte subsets, CD14 Sandoz; and human recombinant erythropoietin (Epo) 2 FITC/CD66e FITC + CD34 PE for CD34+ cell assessment, U/ml, a kind gift from Organon Technica (Turnhout, and CD14 FITC/CD66e FITC + mouse-IgG1 PE for isotype Belgium). After 14 days of culture in 35 mm tissue culture ° control. The CD45 FITC/CD34 PE combination, CD14 dishes (37 C, fully humidified atmosphere, 5% CO2), col- FITC, mouse-IgG1 PE and CD34 PE were from Becton onies were counted. Colonies were defined as aggregates Dickinson (San Jose, CA, USA); CD66e FITC, was of more than 50 cells. Erythroid burst-forming unit (BFU- obtained from the Central Laboratory of the Blood Trans- E) colonies were defined as bursts of colonies consisting fusion Services (Amsterdam, The Netherlands). Cellquest of hemoglobinized cells. Granulocyte–macrophage colony- software was used to determine percentages lymphocytes, forming unit (CFU-GM) colonies contained granulocytes or monocytes, granulocytes and CD34+ cells. Briefly, LDS macrophages or both. Granulocyte–erythroid–macrophage– 751 (Exciton, Daton, OH, USA)23 was used to set a live megakaryocyte CFU (CFU-GEMM) colonies were defined gate, discriminating between damaged and intact nucleated as aggregates containing at least both erythroid and myeloid cells (LDS 751+ cells). Within this live gate, 50000 events cells. HPC was defined as the sum of BFU-E, CFU-GM and were analyzed, and a distinction between subpopulations of CFU-GEMM. From the three samples plated in different CD45+ leukocytes was made; CD45− nucleated cells were concentrations, the concentrations of HPC/ml UCB were considered nucleated red blood cells. First, monocytes were calculated. The mean of the three results was used to calcu- defined as CD45+ and CD14+, and granulocytes and lym- late the concentration of HPC for each UCB unit. phocytes as CD45+ and CD14−. These CD45+/CD14− cells were gated and then plotted in a forward sideward light scatter pattern, which showed two subpopulations, granulo- Statistics cytes with a high sideward light scatter (SSC) pattern, and lymphocytes with a low SSC. In the samples that were Correlations were evaluated with Spearman correlation stained with CD14 FITC/CD66e FITC + CD34 PE , an FL coefficients. Comparison of two independent correlation histogram gate was set on the population that was negative coefficients was done with Fisher’s Z-test. Comparison of for CD14 and CD66e, to exclude mature myeloid (CD66e+) test results of two groups was done with the Mann– and monocytic cells (CD14+) from further analysis. From Whitney test. Hematopoietic content of umbilical cord blood FTH Lim et al 967

Figure 1 Flow cytometric enumeration of CD34+ cells in UCB. (a) Forward-sideward scatter pattern. (b) Labeling with CD45 FITC and CD14 PE for leukocyte subsets. (c) Labeling with CD14 FITC/CD66e FITC and CD34 PE, FL histogram gate on the population that is negative for CD14 (monocytic cells) and CD66e (mature myeloid cells). (d) Identiﬁcation of CD34+ cells in the CD14−/CD66e− population.

Results HPC culture by different technicians. Therefore, we compared test results from the two technicians who performed Only UCB collections that exceeded 50 ml were accepted most of the HPC culture assays, 104 and 105 UCB units, as UCB units for transplantation and entered in this evalu- respectively. The HPC cultures from the remaining 75 UCB ation. Table 1 shows the volumes, cell counts, CD34+ cell units were scored by eight other technicians. The results frequencies and HPC content of the 295 UCB units that indicate that there were significant interpersonal differences were evaluated, expressed as concentration as well as total in scoring HPC expressed in relation to CD34 counts. The numbers per unit harvested. correlation coefficient for the first technician (r = 0.84) dif- Figure 2 illustrates the significant correlation between the fered significantly from the correlation coefficient for the number of CD34+ cells and the number of HPC in the UCB second technician (r = 0.63) (P Ͻ 0.001). The CD34 con- units. The correlation of r = 0.68 appeared to be influenced tent of the units analyzed by both technicians were similar, by subjective interpretation of scoring of colonies in the 3.30 × 106 and 3.36 × 106 CD34+ cells, respectively. How-

Table 1 Volumes and cell counts of 295 UCB units

Median Range Mean s.d.

Volume (ml) 87 50–227 92.5 27.8 Nucleated cells concentration (× 106/ml) 12.1 5.1–33.1 12.6 4.2 total numbersa (× 109) 1.60 0.2–5.1 1.20 0.62 White blood cells concentration (× 106/ml) 11.1 4.8–32.0 11.6 3.8 total numbers (× 109) 0.95 0.2–4.9 1.09 0.55 Mononuclear cells concentration (× 106/ml) 5.2 0.9–20.8 5.5 2.0 total numbers (× 109) 0.45 0.07–3.2 0.51 0.29 CD34+ cells concentration (× 103/ml) 37.0 0–264.5 45.4 36.0 total numbers (× 106) 3.32 0–36.8 4.40 4.12 HPC concentration (× 103/ml) 15.4 0.7–87.0 20.3 16.3 total numbers (× 106) 1.37 0.07–14.8 1.94 1.91 BFU-E concentration (× 103/ml) 5.7 0.1–50.5 7.7 7.6 total numbers (× 106) 0.49 0.01–7.0 0.75 0.86 CFU-GM concentration (× 103/ml) 6.4 0.1–43.0 8.8 7.2 total numbers (× 106) 0.56 0.01–6.4 0.84 0.83 CFU-GEMM concentration (× 103/ml) 2.5 0–23.6 3.8 3.8 total numbers (× 106) 0.24 0–3.0 0.36 0.42 aTotal numbers per unit harvested. Hematopoietic content of umbilical cord blood FTH Lim et al 968

+ Figure 2 Correlation between the total numbers of CD34 cells and Figure 3 Correlation between the total numbers of CD34+ cells and HPC. b = technician 1, ć = technician 2, ¼ = other technicians. Corre- nucleated cells (NC) in individual UCB units. n = 295, r = 0.71, P Ͻ 0.01. lation for all technicians: n = 295, r = 0.68, P Ͻ 0.01; for technician 1: n = 104, r = 0.84, P Ͻ 0.01; for technician 2: n = 105, r = 0.63, P Ͻ 0.01. Discussion ever technician 1 scored significantly lower numbers of HPC (median number of 1.33 × 106 HPC) than technician The use of UCB as a source of hematopoietic stem cells 2 (median of 1.66 × 106 HPC) (P = 0.014). This illustrates for transplantation, is limited by the volume of cord blood that the HPC culture assay is partly dependent on a that can be harvested. Therefore the estimation of the hema- subjective component. topoietic potential of a UCB unit to be used for transplan- In Table 2, the correlations of the frequencies and total tation is critical for deciding which graft should be used numbers of nucleated cells, leukocyte subpopulations with for transplantation if more than one suitable unit is CD34+ cells and cultured HPC in UCB were evaluated. The available.24,25 best correlation was found between the total numbers of In this study, we evaluated whether a correlation could CD34+ cells and the total numbers of nucleated cells, as be found between leukocyte subset cell counts and the two illustrated in Figure 3. Discrimination of the leukocytes into most commonly used methods of enumeration of progenitor subpopulations including mononuclear cells, monocytes, cells, namely flow cytometric counts of CD34+ cells and lymphocytes or granulocytes did not improve the culture of HPC. These parameters are considered to indicate correlation. the hematopoietic cell content, although none of these parameters indicate long-term repopulation. The cobblestone area forming cell assay (CAFC)26 and the long-term culture Table 2 Correlations between hematopoietic progenitor cells and leu- initiating cell assay (LTC-IC)27 may be methods to predict kocyte subpopulations more accurately the frequency of stem cells but they are n = 295 As concentration As total numbers time and skilled man-power consuming, and therefore less Subsets suitable for standard tests for routine use in cord blood + + banking. As shown in Table 2, the correlation between the CD34 HPC CD34 HPC + cells cells frequencies of CD34 cells and HPC in semi-solid medium cultures was 0.68. In scoring of HPC in semi-solid medium CD34+ 0.62 0.68 cultures, however, subjective factors in the enumeration of Nucleated cells 0.61 0.50 0.71 0.60 progenitor cells were most obvious, as illustrated by the White blood cells 0.55 0.44 0.68 0.57 significant interpersonal differences found. The correlation Mononuclear cells 0.46 0.31 0.64 0.48 coefficient of 0.84 between the HPC scores of one tech- Lymphocytes 0.39 0.25 0.60 0.44 + Monocytes 0.45 0.31 0.60 0.47 nician and the enumeration of CD34 cells by flow cytome- Granulocytes 0.40 0.35 0.55 0.49 try may indicate that it is technically possible to achieve standardization of the HPC enumeration, but that on a rou- Hematopoietic content of umbilical cord blood FTH Lim et al 969 tine basis the assay may be too dependent on long lasting References experience of the technicians. Correlations have been found between the numbers of 1 Broxmeyer HE, Douglas GW, Hangoc G et al. Human umbili- nucleated cells as well as CD34 positive cells with clinical cal cord blood as potential source of transplantable hematopoi- outcome of allogeneic stem cell transplantation.14,19–22 etic stem/progenitor cells. Proc Natl Acad Sci USA 1989; 86: Therefore, we analyzed whether leukocyte subsets also 3828–3832. 2 Gluckman E, Broxmeyer HA, Auerbach AD et al. Hematopoi- reflected the hematopoietic content of UCB grafts. Since etic reconstitution in a patient with Fanconi’s anemia by UCB grafts can contain a significant number of nucleated means of umbilical cord blood from an HLA-identical sibling. red cells, both the total numbers of nucleated cells, the total New Engl J Med 1989; 321: 1147–1178. + numbers of CD45 white blood cells, lymphocytes, mono- 3 Rubinsten P. Placental blood-derived hematopoietic stem cells cytes and granulocytes were compared with frequencies of for unrelated bone marrow reconstitution. J Hematother 1993; CD34+ cells and HPC in each unit. The best correlation 2: 207–210. was found between the total numbers of nucleated cells and 4 Harris DT. Experience in autologous and allogeneic cord the number of CD34+ cells or HPC. Although it is known blood banking. J Hematother 1996; 5: 123–128. that hematopoietic precursors reside in the mononuclear 5 Kogler G, Callejas J, Hakenberg P et al. Hematopoietic trans- fraction of leukocytes, any differentiation of leukocytes plant potential of unrelated cord blood: critical issues. J Hema- + tother 1996; 5: 105–116. decreased the correlation with CD34 cells or HPC. Vari- 6 Laporte JP, Gorin NC, Rubinstein P et al. Cord-blood trans- ation of hematopoietic precursors in UCB does not directly plantation from an unrelated donor in an adult with chronic influence the number of nucleated cells in UCB by their myelogenous leukemia. New Engl J Med 1996; 335: 167–170. small absolute numbers, but certain factors like the course 7 Kernan NA, Schroeder ML, Ciavarella D et al. Umbilical cord of parturition, may influence both the absolute numbers of blood infusion in a patient for correction of Wiskott–Aldrich progenitor cells as well as the absolute numbers of syndrome. Blood Cells 1994; 20: 245–248. nucleated cells.28 8 Kurtzberg J, Graham M, Casey J et al. The use of umbilical In conclusion, flow cytometric analysis of CD34+ cells cord blood in mismatched related and unrelated hemopoietic in UCB transplants appeared to be more suitable to enumer- stem cell transplantation. Blood Cells 1994; 20: 275–283. ate the frequency of hematopoietic precursors in UCB units 9 Lazzari L, Corsini C, Curioni C et al. The Milan Cord Blood Bank and the Italian Cord Blood Network. J Hematother than the HPC culture assay, due to interpersonal differences 1996; 5: 117–122. in scoring HPC. Moreover, flow cytometry is less time- 10 Hows J, Nicol A, Denning-Kendall P et al. Cord blood as an consuming, and therefore cheaper than HPC culture. The alternative source of haematopoietic stem cells. Ann Oncol total number of nucleated cells correlated with the number 1996; 7 (Suppl. 2): 47–51. of CD34+ cells, although there was still considerable vari- 11 Rubinsten P, Taylor PE, Scaradavou A et al. Unrelated pla- ation. Because it is still difficult to achieve a standard of cental blood for bone marrow reconstitution: organization of CD34 enumeration performed by all providers of UCB the placental blood program. Blood Cells 1994; 20: 596–600. transplants, the total number of nucleated cells of UCB 12 Gordon MY. Human hematopoietic stem cell assays. Blood units at the time of harvest may be used to compare the Rev 1993; 7: 190–197. hematopoietic potential of the units from different UCB 13 Amos TAS, Gordon MY. Source of human hematopoietic + stem cells for transplantation – a review. Cell Transplant banks. The number of CD34 cells may help as an 1995; 4: 547–569. additional parameter to choose when there are more suit- 14 Berenson RJ, Bensinger WI, Hill RS et al. Engraftment after able UCB units available. Although HPC cultures appeared infusion of CD34+ marrow cells in patients with breast cancer not to be very suitable for selecting UCB units, comparison or neuroblastoma. Blood 1991; 77: 1717–1722. of HPC cultures before and after cryopreservation plays an 15 Brugger W, Henschler R, Heimfeld S et al. Positively selected important role in quality control of cryopreservation and autologous blood CD34+ cells and unseparated peripheral storage of UCB units in umbilical cord blood banks. Deter- blood progenitor cells mediate identical hematopoietic mination of the HPC frequencies in a small aliquot from engraftment after high-dose VP16, ifosfamide, carboplatin and the unit left aside during freezing and storage may indicate epirubicin. Blood 1994; 84: 1421–1426. 16 Gratama JW, Kraan J, Levering W et al. Analysis of variation the viability of the unit selected for use at the time of the + in results of CD34 hematopoietic progenitor cell enumeration transplantation. in a multicenter study. Cytometry 1997; 30: 109–117. 17 Farley TJ, Rooney W, Kuhns E et al. An intralaboratory quality control program for quantitation of CD34+ cells by flow cytometry. J Hematother 1997; 6: 303–308. + Acknowledgements 18 Johnson HE. Toward a worldwide standard for CD34 enumeration? J Hematother 1997; 6: 83–84 (letter). 19 Wagner JE, Kernan NA, Steinbuch M et al. Allogeneic sibling This study was supported by grants from the JA Cohen Institute umbilical cord blood transplantation in children with malig- for Radiopathology and Radiation Protection, and the Dutch Can- nant and non-malignant disease. Lancet 1995; 346: 214–219. cer Society. We are grateful to M Aalders and Y Breuger for 20 Kurtzberg J, Laughlin M, Graham ML et al. Placental blood obtaining informed consent from patients at the Department of as a source of hematopoietic stem cells for transplantation into Obstetrics, to A vd Veen and M Blijleven for harvesting, pro- unrelated recipients. New Engl J Med 1996; 335: 157–166. cessing and preservation of the UCB units, to A vd Marel, J Over- 21 Gluckman E, Rocha V, Boyer-Chammard A et al. Outcome devest, G d Groot-Swings and G Brouwer-Mandema for flow of cord-blood transplantation from related and unrelated cytometric analysis and to R Bongaerts and E Neerincx for the donors. New Engl J Med 1997; 337: 373–381. HPC culture assays. 22 Wagner JE, Kurtzberg J. Unrelated donor umbilical cord blood Hematopoietic content of umbilical cord blood FTH Lim et al 970 transplantations at Duke University and Minnesota: results in 26 Breems DA, Blokland EA, Neben S, Ploemacher RE. Fre- 143 patients. Proc 3rd Eurocord Transplant Concerted Action, quency analysis of human primitive haematopoietic stem cell Annecy, France, May, 1998. subsets using a cobblestone area forming cell assay. Leukemia 23 Terstappen LWMM, Shah VO, Conrad MP et al. Discriminat- 1994; 8: 1095–1104. ing between damaged and intact cells in fixed flow cytometric 27 Sutherland HJ, Lansdorp PM, Henkelman DH et al. Func- samples. Cytometry 1988; 9: 477–484. tional characterization of individual human hematopoietic 24 Rubinsten P, Dobrila L, Rosenfield RE et al. Processing and stem cells cultured at limiting dilution on supportive marrow cryopreservation of placental/umbilical cord blood for unre- stromal layers. Proc Natl Acad Sci USA 1990; 87: 3584–3588. lated bone marrow reconstitution. Proc Natl Acad Sci USA 28 Lim FTH, van Winsen L, Willemze R et al. Influence of deliv- 1995; 92: 10119–10120. ery on numbers of leukocytes, leukocyte subpopulations, and 25 Harris DT, Schumacher MJ, Rychlik S et al. Collection, separ- hematopoietic progenitor cells in human umbilical cord blood. ation and cryopreservation of umbilical cord blood for use in Blood Cells 1994; 20: 547–559. transplantation. Bone Marrow Transplant 1994; 13: 135–143.