Rapid immune reconstitution after a reduced-intensity conditioning regimen and a CD3-depleted haploidentical stem cell graft for paediatric refractory haematological malignancies


Gregory A Hale MD, Division of Bone Marrow Transplantation, St Jude Children's Research Hospital, Mail Stop 260, 332 N Lauderdale St, Memphis, TN 38105-2794, USA. E-mail: gregory.hale@stjude.org


The main obstacles to successful haploidentical haematopoietic stem cell transplantation from a mismatched family member donor are delayed immune reconstitution, vulnerability to infections and severe graft-versus-host disease (GvHD). We designed a reduced-intensity conditioning regimen that excluded total body irradiation and anti-thymocyte globulin in order to expedite immune reconstitution after a CD3-depleted haploidentical stem cell transplant. This protocol was used to treat 22 paediatric patients with refractory haematological malignancies. After transplantation, 91% of the patients achieved full donor chimaerism. They also showed rapid recovery of CD3+ T-cells, T-cell receptor (TCR) excision circle counts, TCRβ repertoire diversity and natural killer (NK)-cells during the first 4 months post-transplantation, compared with those results from a group of patients treated with a myeloablative conditioning regimen. The incidence and extent of viremia were limited and no lethal infection was seen. Only 9% of patients had grade 3 acute GvHD, while 27% patients had grade 1 and another 27% had grade 2 acute GvHD. This well-tolerated regimen appears to accelerate immune recovery and shorten the duration of early post-transplant immunodeficiency, thereby reducing susceptibility to viral infections. Rapid T-cell reconstitution, retention of NK-cells in the graft and induction of low grade GvHD may also enhance the potential anti-cancer immune effect.

Haploidentical hematopoietic stem cell transplantation (HaploHSCT) from a mismatched family member (MMFM) donor offers an alternative option for patients who lack an human lecucocyte antigen (HLA)-matched donor. The main obstacles are graft rejection, delayed immune reconstitution, graft-versus-host disease (GvHD) and vulnerability to infections (Spitzer, 2005). Intensive chemo-radiotherapy has customarily been given before HaploHSCT to reduce the burden of disease and induce immunosuppression in the host (Satwani et al, 2005). Such conditioning regimens reduce graft rejection but they can cause considerable mortality due to severe toxicity, delayed immune restoration and severe infection (Spitzer, 2005), especially in elderly or heavily pre-treated patients. Moreover, disease-free survival estimates are not appreciably improved by aggressive chemo-radiotherapy (Resnick et al, 2005), as recurrent or refractory malignancies have usually become resistant to chemotherapy. These observations have encouraged the reassessment of conditioning strategies for HaploHSCT. Newer strategies aim to minimise toxicity while allowing rapid engraftment and expediting immune reconstitution during the early post-transplant period, thereby protecting the host from infection and perhaps generating a graft-versus-tumour (GvT) effect against disease relapse.

Several groups have overcome the HLA barrier in HaploHSCT, even in the non-myeloablated host, by administering large doses of CD34+ selected stem cells (Rao et al, 1997; Reisner & Martelli, 2000; Reisner et al, 2005). However, the emergence of new markers (Bonde et al, 2004; Shmelkov et al, 2005) targeting stem cells that are negative for CD34 (Guo et al, 2003) has brought into question the use of the CD34 antigen as the sole target for stem cell selection. Further, the loss of other immune components, especially natural killer (NK)-cells, during the selection process makes this approach less than optimal for generating anti-tumour and anti-viral effects. In recent years, several groups have used reduced-intensity or non-myeloablative conditioning regimens for HaploHSCT. They reported reduced mortality and an acceptable rate of engraftment. However, delayed immune reconstitution, severe GvHD and infection continue to be impediments (Spitzer, 2005).

We designed a reduced-intensity conditioning regimen (RICR) for HaploHSCT from MMFM donor. The regimen excluded total body irradiation (TBI) and antithymocyte globulin (ATG). Our rationale was that elimination of TBI may reduce damage to organs that generate immune cells (Friedman et al, 2001), while avoidance of ATG may prevent complications that include delayed immune reconstitution and Epstein-Barr virus (EBV)-associated lymphoproliferative disease (Bacigalupo, 2005; Beiras-Fernandez et al, 2005). The graft was immunomagnetically depleted of mature T-cells that can cause overwhelming GvHD and contained a large number of stem cells (≥2 × 106/kg) to reduce the risk of graft failure, while retaining CD34-negative stem cells and most other immune cells. To examine whether the non-TBI/non-ATG-based RICR would expedite immune reconstitution during the early post-transplant period, this protocol was used to treat 22 paediatric patients with refractory haematological malignancies. Results were compared with those obtained from a group of patients that received a myeloablative conditioning regimen. We also looked at the influence of this protocol on incidence of viral infection and GvHD.

Patients and methods

Patients and transplant regimens

Twenty-two patients who had refractory haematological malignancies (persistent disease to standard chemotherapy and conditioning regimen) or who had failed prior transplantation and required a second or third transplant, or who had therapy-related myelodysplasia or acute myeloid leukaemia (AML), received a RICR. The regimen comprised fludarabine, thiotepa, melphalan and OKT3 and did not include TBI and ATG. Mycophenolate mofetil was used for GvHD prophylaxis.

A group of twelve concurrently treated patients who received a myeloablative conditioning regimen (MCR) were used as a comparison group because their ages, pretransplant levels of immune competent cells and graft composition were similar to those in the RICR group (Tables I and II). The patients in the MCR group were in remission of haematological malignancies (blasts <5%; platelets ≥100 × 109/l and absolute neutrophil count >1 × 109/l) and were receiving a first transplant. Their conditioning regimen comprised TBI (12 Gy), cyclophosphamide, thiotepa, ATG and OKT3. Cyclosporine was utilised for GvHD prophylaxis.

Table I.   Characteristics of the patients and transplant details.
Patient no.DiagnosisAge at HaploHSCTSexHLA matches (n/6)CD34+cell dose (×106/kg)CD3+ cell dose (×106/kg)
  1. AML, acute myeloid leukaemia; ALL, acute lymphoblastic leukaemia; CML, chronic myeloid leukaemia; MDS, myelodysplastic syndrome; P-value: level of ≤0·05 was considered to be significant.

RICR group
19AML astrocytoma20M3/68·60·25
Mean (±SE) 10 ± 1  17·4 ± 10·14 ± 0·02
MCR group
Mean (±SE) 11·1 ± 1  8·3 ± 10·07 ± 0·02
P-value 0·54  0·00250·0336
Table II.   Mean immune indicator values before HaploHSCT
GroupsjTRECVBSPSCSCD3+ cellsCD4+ cellsCD8+ cellsCD19+ cellsCD56+/CD16+/CD3 cells
  1. RICR, reduced intensity conditioning regimen; MCR, myeloablative conditioning regimen; VBSPSCS, mean Vβ spectratype complexity score; sjTREC count: mean copies ± SE, per ml blood; Cell counts: mean absolute number ± SE, × 109/l blood; P-value: ≤0·0125 was considered to be significant.

RICR8092 (±4117)145 (±13)0·604 (±0·199)0·337 (±0·119)0·246 (±0·076)0·102 (±0·05)0·063 (±0·013)
MCR6711 (±1947)197(±4)0·835 (±0·107)0·378 (±0·066)0·422 (±0·06)0·117 (±0·076)0·067 (±0·013)

Both groups received grafts of mobilized peripheral blood stem cells obtained from MMFM donors. The grafts were depleted of CD3+ cells by using the ClinicMacs® system as previously described (Handgretinger et al, 2001a; Gordon et al, 2002). The characteristics of the patients and the grafts they received are listed in Table I. This study was approved by the St Jude Children's Research Hospital Institutional Review Board and informed consent was obtained from donors, patients, parents, or guardians, as appropriate. The patients in both groups underwent HaploSCT over a 2-year time period at St Jude Children's Research Hospital.

Acute and chronic GvHD were diagnosed, staged and graded according to the standard criteria (Glucksberg et al, 1974; Shulman et al, 1980; Przepiorka et al, 1995).

The assays for detecting immune reconstitution and engraftment

Signal joint T-cell receptor excision circles (sjTREC) were measured monthly by real-time polymerase chain reaction (PCR) as described previously (Hazenberg et al, 2000; Chen et al, 2005). The size distribution of the T cell receptor β (TCRβ) complementarity-determining region 3 (CDR3) was determined by using TCRβ CDR3 size spectratyping as described previously (Chen et al, 2005) and was performed quarterly in parallel with the sjTREC assay. The overall complexity of TCRβ subfamilies was calculated by spectratype complexity score (SCS) as described by Wu et al (2000). The phenotypes of immune cells were analysed by flow cytometry, as previously described (Chen et al, 2005). Donor and recipient alleles were discriminated by using quantitative chimaerism assay reported previously (Kreyenberg et al, 2003).

Detection of cytomegalovirus (CMV), Epstein-Barr virus (EBV) and adenovirus (ADV)

All study participants were monitored at least weekly for all three viruses for the first 100 d after transplantation, and as clinically indicated thereafter. ADV and EBV were measured in peripheral blood mononuclear cells by real-time PCR, as described previously (Kimura et al, 1999; Gu et al, 2003). CMV was quantified in whole blood by using the COBAS AMPLICOR CMV MONITORTM TEST (Roche Diagnostic Corp, Indianapolis, IN, USA). Any positive PCR results were considered to indicate viremia for the purpose of this study.

Statistical analyses

Age, CD34+ and CD3+ cell dose in the two groups were compared with using Student's t-test. The exact Wilcoxon rank sum test was used to independently assess whether there were any differences on the measurements of immune reconstitution in the two groups in the first 4 months after the transplantation. The P-values were considered to be significant at α = 0·0125 (0·05/4) using Bonferroni adjustment for multiple comparisons. Measurements of log10 transformed immune reconstitution indicator value during the first 4 months post-HaploHSCT were compared between two groups by using the mixed effects model with repeated measurements.

The cumulative incidence of GvHD in the two groups was estimated by Gray's method. The length of time at risk of grade 3–4 acute GvHD (aGvHD) or chronic GvHD (cGvHD) was measured from the date of the transplantation to the date of occurrence of the GvHD, death, or most recent contact, whichever occurred first. Death from any cause was considered a competing event.

The criterion for significance in all analyses other than exact Wilcoxon rank sum test was a P-value significant at level of α = 0·05. All statistical analyses were performed with the statistical software package SAS, Release 9.1 (Cary, NC, USA).



Twenty of 22 patients (91%) in the RICR group and 11 of 12 patients (92%) in the MCR group had full-donor chimaerism after transplantation. Two patients (9%) in the RICR group and one patient (8%) in the MCR group experienced graft failure. Average chimaerism was not found to differ significantly between the two groups in any month tested.

Recovery of sjTREC counts

We analysed sjTRECs before the start of transplant-related treatment and at monthly intervals during the first 12 months post-transplantation. Before the conditioning, the mean sjTREC counts of the two groups before HaploHSCT showed no significant difference (Table II). During the first month after transplantation, five patients in the RICR group had detectable sjTRECs while their chimaerism was 100% donor. At this time the mean sjTREC number was 35 copies/ml blood in the RICR group and 0 copies/ml blood in the MCR group (Fig 1). During the third and fourth months post-transplantation, the mean sjTREC counts of the RICR group continued to increase and were six to nine times the mean counts in the MCR group. There was a trend toward a significant difference (P = 0·06) between the two groups in log10-transformed sjTREC numbers over the first 4 months post-HaploHSCT. The sjTREC numbers in the RICR group increased steadily over the next 8 months and reached a normal level at 5 months. There was no significant difference in sjTREC counts between two groups after 5 months.

Figure 1.

 Recovery of signal joint T-cell receptor excision circles (sjTREC) count in patients who received reduced-intensity conditioning regimen (RICR) or myeloablative conditioning regimen (MCR). The y-axis shows mean sjTREC counts (copies/ml blood) in the patients treated with RICR (•) or with MCR (open triangle).

Restoration of TCRβ repertoire diversity.

Before conditioning, most patients in both groups had a Gaussian-like TCRβ repertoire pattern. However, the average SCS in the RICR group (145) was less than that in the MCR group (197) (P = 0·0064) (Table II; Fig 2). During the first month after HaploHSCT, the SCS descended sharply in both groups, but the SCS in the RICR group (40) was 3·7 times that in the MCR group (12) (P = 0·0001). Three months after HaploHSCT, the mean SCS had increased to 117 in the RICR group, but remained significantly lower in the MCR group (SCS: 58) (P = 0·0075). After 6 months, the SCS continued to increase in the RICR group and reached 188, near the normal range, at 9 months. No significant difference in SCS was observed between the two groups after 6 months post-HaploHSCT.

Figure 2.

 T-cell receptor(TCR)β repertoire diversity. The y-axis shows mean TCRβ spectratype complexity score (SCS) in patients treated with reduced-intensity conditioning regimen (•) or with myeloablative conditioning regimen (open triangle).

Measurement of immune cell reconstitution by flow cytometry

We measured T-, B- and NK-cell reconstitution by flow cytometry. Before transplantation, the mean CD3+, CD4+ and CD56+ cell counts were below the normal ranges (CD3+: 0·9–3·7 × 109/l blood; CD4+: 0·55–2·15 × 109/l blood; CD56+: 0·1–0·75 × 109/l blood) in both groups (P = 0·024, 0·12, and 0·95 respectively). The mean CD8+ and CD19+ cell counts in both groups were within normal ranges (CD8+: 0·200–1·4 × 109/l blood; CD19+: 0·1–1·62 × 109/l blood) (Table II).

During the first month after HaploHSCT, the T-cell populations (CD3+, CD4+, CD8+) were dramatically decreased in both groups, but their numbers were much higher in the RICR group than in the MCR group. The cell numbers began to increase at the second month in the RICR group, while remaining significantly lower in the MCR group (P ≤ 0·0046) (Fig 3A–C). The log10-transformed counts over the first 4 months post-HaploHSCT differed significantly between the two groups (P ≤ 0·0037). During the subsequent 8 months, T-cell numbers increased steadily and no significant difference in T-cell numbers was observed between the two groups.

Figure 3.

 Reconstitution of immune competent cells tested by flow cytometry. The y-axis shows mean cell counts ( × 109/l) in the patients treated with reduced-intensity conditioning regimen (•) or with myeloablative conditioning regimen (open triangle). (A) CD3+ T-cells; (B) CD4+ T-cells; (C) CD8+ T-cells; (D) CD56+/CD16+/CD3 cells; (E) CD19+cells.

Natural killer-cell numbers in both groups returned to the normal range during the first month. However, the mean number in the RICR group was four times that in the MCR group (P = 0·0074). During the subsequent 3 months, the NK-cell numbers in the RICR group were one to three times those in the MCR group (Fig 3D).

The mean B-cell counts of both groups returned to the normal range within 3 months (Fig 3E) and were not found to differ significantly. The B-cell numbers continued to increase over the subsequent 9 months and remained in the normal range in the RICR group.

Incidence of GvHD

Six patients (27%) in the RICR group and two patients (17%) in the MCR group had grade 1 aGvHD. Another six patients (27%) in the RICR group had grade 2 aGvHD. Two patients (9%) in the RICR group developed grade 3 aGvHD, but no grade 4 aGvHD was seen (Table III). Most patients with GvHD in the RICR group had only skin involvement. Three patients had gut involvement, one had liver involvement and two had both gut and liver involvement. Five patients in the RICR group had cGvHD. No significant difference was found between the two groups in the incidence of aGvHD (grade 3–4) or cGvHD when the cumulative incidence of GvHD was estimated. At the end of this study, none of the patients had died because of GvHD.

Table III.   Incidence of aGvHD in patients.
GroupaGvHD grade IaGvHD grade IIaGvHD grade IIIaGvHD grade IV
  1. RICR, reduced intensity conditioning regimen; MCR, myeloablative conditioning regimen; aGvHD, acute graft-versus-host disease.

RICR (n = 22)6 (27%)6 (27%)2 (9%)0
MCR (n = 12)2 (17%)000


We monitored patients for EBV, CMV, and ADV viremia. Virus was detected in 16 of 22 patients (73%) in the RICR group and 11 of 12 patients (92%) in the MCR group. In the RICR group, ten patients were positive for one type of virus and six for two types of virus. In the MCR group, all three viruses were detected in four patients, two viruses in six patients and one virus (CMV) in a single patient. Figure 4 shows mean monthly viral loads post-transplant, with each data point representing the log10-transformed quantitative mean of all samples collected during that month. The mean copy number of all viruses was greater in the MCR group than in the RICR group. The log10-transformed viral loads over the first 4 months post-HaploHSCT differed significantly in the two groups (CMV: P = 0·0486; EBV: P < 0·0001; ADV: P = 0·0018). During the remaining 8 months, ADV was undetectable in the RICR group and EBV loads in this group were far below the threshold at which anti-EBV treatment is required. CMV was detected in only one patient in the RICR group and undetectable in all of the patients tested after 6 months post-HaploHSCT. At the end of this study, no patient had died of viremia.

Figure 4.

 Viral load during viremia. The y-axis shows log10 mean viral copy numbers in the patients treated with reduced-intensity conditioning regimen (•) or with myeloablative conditioning regimen (open triangle). (A) Cytomegalovirus (CMV) infection; (B) Epstein-Barr virus (EBV) infection; (C) Adenovirus (ADV) infection.


This study demonstrated, for the first time, rapid immune reconstitution in children with refractory haematological malignancy who receive a non-TBI-ATG based RICR before receiving CD3-depleted HaploHSCT from a MMFM donor.

With our conditioning regimen, most patients achieved full donor chimaerism and retained it for the duration of the study. These results are comparable with those described in the literature (Spitzer, 2005) or achieved with the myeloablative conditioning in our comparison group. This success may reflect the adequacy of fludarabine, thiotepa, melphalan and OKT3 in inhibiting host immune responses or may reflect the large number of stem cells in the graft, which can overcome HLA engraftment barriers.

T-cell reconstitution is usually delayed after intensive conditioning. The patients in our study showed a more rapid increase in sjTREC counts, TCR Vβ diversity scores and T-cell counts during the first 4 months after transplantation than did the MCR group or patients described in reports of other HaploHSCT studies, in which T-cell recovery was not detected until 5–6 months post-transplant (Ball et al, 2005). The increases began during the second or third month after transplantation, which is crucial for protecting patients from infections and disease relapses. The simultaneous rise in sjTREC numbers and the appearance of full donor chimaerism suggest that the T-cells were derived from donor precursors regenerated mainly through the thymus-dependent pathway, rather than through peripheral expansion of mature donor T-cells. We reported previously (Chen et al, 2005) that rapid recovery of thymus-derived T-cells was dependent on normal thymic function before transplantation. Therefore, it is possible that the rapid T-cell recovery reflected high thymic function before transplantation. However, these patients suffered from refractory haematological malignancies and 59% had poor thymic function with sjTREC counts below the normal range prior to transplantation. Moreover, sjTREC counts increased rapidly in patients with both normal and low pretransplant thymic function. Therefore, rapid thymus-dependent T-cell recovery may be facilitated by the RICR. Exclusion of TBI and ATG may reduce toxicity to immune organs and to pre-activated T-lymphocyte blasts. Immune cells may therefore develop and expand more quickly after transplantation.

Natural killer-cells usually recover rapidly within 1 month after transplantation. However, NK-cells regenerated from CD34+ selected stem cells may be immature during the first few months post-transplantation, with impaired cytotoxicity to recipient mis-matched AML blasts (Nguyen et al, 2005). In our study, NK-cell numbers rebounded strongly to the upper normal range during the first month and remained in the normal range for the next 11 months. Because we did not deplete donor NK-cells from the graft, the NK-cell repletion observed during the first few months probably included both graft-derived mature NK-cells and immature NK-cells regenerated from donor stem cells. Mature donor NK-cells may retain the capacity to efficiently kill tumour cells and viruses. The robust NK-cell restoration in the RICR group may also reflect the exclusion of ATG, which can exert cytotoxicity against NK-cells (Bacigalupo, 2005; Beiras-Fernandez et al, 2005).

B-cell recovery in the RICR group was also rapid, requiring only a 3-month period to reach the normal range. This recovery time was comparable with that in the MCR group and with that reported in the literature (Handgretinger et al, 2001b).

Severe viremia is often a problem after myeloablative conditioning for HaploHSCT. In our study, the overall incidence of CMV, EBV and ADV viremia in the RICR group was less than that in the MCR group. Both the number of patients with viremia and the number of patients positive for multiple viruses was less in the test group. The viral loads were also significantly lower in the RICR group and no lethal viremia was seen. EBV counts in the RICR group were consistently lower than the threshold (2000 copies/μg mononuclear cells) at which anti-EBV treatment is initiated. It is possible that the earlier recovery of thymus-regenerated T-cells and the strong rebound of mature NK-cells may have protected these patients from viral infection and helped to reduce the severity or duration of viremia when it occurred.

Although severe GvHD causes substantial mortality after HLA-mismatched transplantation, it is also associated with a beneficial GvT effect (Lundqvist & Childs, 2005). Therefore, investigators have tried to elicit GvT without inducing GvHD. Transfusion of donor NK-cells is helpful to eliminate tumour cells that lack major histocompatibility complex (MHC)-class I matching killer immunoglobulin-like receptors (KIR) (Nguyen et al, 2005). Production of tumour-specific T-cell clones is laborious and problematic to produce (Li et al, 2005), and such cells recognise only tumour-specific antigens, which may not be stably expressed. Moreover, NK-cells and tumour-specific T-cells may lose their effectiveness as most tumour antigens are non-mutant self-proteins carrying MHC class-I and are poorly immunogenic (Li et al, 2005). GvHD is the immune response against recipient antigens, including those tumour antigens carrying non-mutant self-proteins and MHC class-I antigen. Mild aGvHD (grade 1 or 2) is an acceptable compensation of anti-tumour treatment for those patients for whom intensive chemo-radiotherapy, NK-cell infusion or even anti-tumour specific T-cell therapy has failed.

In our RICR group, aGvHD was mild: only 9% patients had grade 3 aGvHD and no grade 4 aGvHD was seen. This incidence is much lower than that described in the literature (Spitzer, 2005). Further, no patient had died of GvHD at the end of this study (G. Hale, unpublished observations). Grade 1–2 aGvHD, on the other hand, developed in 54% of patients. As discussed above, the mild aGvHD may help to reduce residual malignancy as a GvT effect may arise simultaneously (Lundqvist & Childs, 2005). Five patients in the RICR group have experienced progression to cGvHD, but the cumulative incidence of cGvHD does not differ significantly between the two groups. As reported previously, the occurrence of cGvHD can reduce malignancy relapse (Perez-Simon et al, 2003; Kröger et al, 2004). Patients in the RICR group had refractory malignancies resistant to chemo-radiotherapy. Acute and chronic GvHD together may efficiently kill tumour cells and enable patients to survive refractory tumours. Indeed, 31·8 ± 10·7% of patients in the RICR group survived and 25·5 ± 9·8% of patients survived disease-free 2 years after transplantation. The relapse rate in the RICR group was 50% (unpublished observations). The overall survival, disease-free and relapse rates for RICR group are very promising in this set of high-risk patients.

In conclusion, the non-TBI/non-ATG-based RICR used in our study accelerated the immune reconstitution of T-, B- and NK-cells after HaploSCT. We also observed limited viremia and limited severe aGvHD in those patients. Furthermore, the development of grade 1 or 2 aGvHD may provide an anti-cancer immune effect. Success in this vulnerable and highly pre-treated group of patients may provide an example for other populations who require transplantation and have not been highly pre-treated.


X Chen collected, collated and analysed all of the data in the article and wrote the article; R Handgretinger designed the RICR protocol and was responsible for the conception of the article; G Hale designed the MCR protocol and was responsible for conception of the article; R Handgretinger and G Hale were fully responsible for both protocols; R Barfield collected clinical data and was responsible for the conception and writing of part of the article; E Benaim participated in designing the RICR protocol; X Chen was responsible for sjTREC and Vβ spectratype reporting; J Knowles performed the Vβ spectratyping; R Handgretinger, G Hale, R Barfield, E Benaim, W Leung, EM Horwitz, P Woodard, K Kasow, U Yusuf treated patients on both protocols and contributed clinical data reports; FG Behm was responsible for lymphocyte subset analysis by flow cytometry; RT Hayden was responsible for aspects of the study related to detection and quantification of viral infection, and wrote related sections of the article; SA Shurtleff was responsible for the chimaerism reporting; V Turner was responsible for the HLA typing reporting; K Srivastava performed the statistical analysis.


We thank all scientists, physicians and nurses in the Division of Bone Marrow Transplantation for their contribution to this study. We thank Dr PW Eldridge and the staff of the Human Applications Laboratory for stem cell selection. We thank X Tong in the Department of Biostatistics for statistical analysis. We also thank the scientists in the Department of Pathology and in the Hartwell Center at St Jude Children's Research Hospital for performing the chimaerism, HLA typing, flow cytometry and viral detection assays, and for making the primers used in this study. We appreciate the editorial assistance of S Naron.

This work was supported in part by grant CA21765 from National Institutes of Health, by the Assisi Foundation of Memphis and by the American Lebanese Syrian Associated Charities (ALSAC).