Immune Reconstitution following Myeloablative Allogeneic Hematopoietic Stem Cell Transplantation: The Impact of Expanding CD28negative CD8+ T Cells on Relapse Academic research paper on "Clinical medicine"

ASBMI

American Society for Blood and Marrow Transplantation

Immune Reconstitution following Myeloablative Allogeneic Hematopoietic Stem Cell Transplantation: The Impact of Expanding CD28negative CD81 T Cells on Relapse

12 3 2 3 1 2

Ibrahim Yakoub-Agha, ' ' Pasquine Saule, ' Leonardo Magro, Pascale Cracco, Alain Duhamel,4 Valerie Coiteux,1 Benedicte Bruno,1 Françoise Dufosse,2

1 2 3 2 3 2 3

Jean-Pierre Jouet, ' ' Jean-Paul Dessaint, , Myriam Labalette ,

Allogeneic stem cell transplantation has become standard therapy for hematologic malignancies through the positive immunologic graft-versus-leukemia effect. Initial immune recovery relies on peripheral expansion of infused T cells, which switch to a memory-like phenotype. This study prospectively investigated whether changes in subset composition precedes complications after myeloablative HLA-matched transplantation for hematologic malignancies. Of 80 allograft recipients, 18 were still free of clinical complication throughout 395 to 1564 days of follow-up. Compared with this complication-free subgroup, patients who developed chronic graft-versus-host disease (cGVHD) without relapsing recovered similar numbers of circulating T cells with predominance of CD81 T cells lacking CC-chemokine receptor-7 and CD28 expression throughout the first year after transplantation. Conversely, poor CD81 T cell recovery with diminished numbers of CD28neg CD8 T cells 1/4th of that of relapse-free patients) preceded occurrence of malignant relapse. In multivariate analysis, lower CD28neg CD8 T cell counts by day 60 postallograft were associated with a greater risk of subsequent relapse (hazard ratio [HR] 0.33; 95% confidence interval [CI]: 0.14-0.76; P = .01). Enumeration of CD28neg CD81 T cells in patients could assist in predicting risk of relapse and help build an algorithm for accelerating the immune recovery by reducing the immunosuppressive treatment and considering the introduction of preemptive donor lymphocyte infusions.

Biol Blood Marrow Transplant 15: 496-504 (2009) © 2009 American Society for Blood and Marrow Transplantation

KEY WORDS: Allogeneic stem cell transplantation, Homeostatic peripheral expansion, CD8, CD28, Relapse, chronic GVHD

INTRODUCTION

Improved prophylactic measures, supportive care, and anti-infectious therapies have contributed to reduce treatment-related morbidity and mortality

From the 'Service des Maladies du Sang, CHU, Lille, France;

Service d'Immunologie, CHU, Lille, France; EA2686, IFR114, Universite Lille 2, Lille, France; and Centre d'Etudes et de Recherche en Informatique Medicale, Universite Lille 2, Lille, France. The first 2 authors contributed equally to this work and should be

considered as first authors. Financial disclosure: See Acknowledgments on page 503. Correspondence and reprint requests: Ibrahim Yakoub-Agha, MD, PhD, Service des Maladies du Sang, UAM allogreffes de CSH, CHRUde Lille, F-59037 Lille, France (e-mail: i-yakoub-agha@ chru-lille.fr).

Received July 31, 2008; accepted November 27, 2008

©2009 American Society for Blood and Marrow Transplantation

1083-8791/09/154-0001$36.00/0

doi:10.1016/j.bbmt.2008.11.038

(TRM) in the early period following allogeneic stem cell transplantation (allo-SCT). However, relapse of the underlying hematologic malignancy and chronic graft-versus-host disease (cGVHD) remain major causes of treatment failure. Other cell lineages notwithstanding, mature T cells contained in or derived from the allograft exert powerful antitumor effects through the graft-versus-leukemia (GVL) reaction [1], but T cell-mediated responses can also induce GVHD [2]. Initial T cell recovery is primarily dependent on peripheral expansion of donor mature T cells (termed homeostatic peripheral expansion), because de novo generations of naive T cells from donor hemato-poietic stem cells is compromised following cytore-ductive regimens and cannot have a full impact on immune reconstitution for 1 or several years, depending on recipient age [3-7]. Homeostatically expanding T cells show memory-like phenotypes [3-8] and the high homeostatic proliferation following lymphopenia can facilitate expansion of antitumor immunity [7]. Therefore, malignant relapse might result from

insufficient homeostatic peripheral expansion, and conversely, peripheral expansion with a too strong immune reaction against host antigens might induce GVHD.

The purpose of this prospective study was to address whether differences in the dynamics of CD4 and CD8 T cell recovery within the first year after allo-SCT can be related to the absence or occurrence of malignant relapse and GVHD. The expression pattern of surface markers in peripheral blood was analyzed longitudinally in 80 patients after myeloabla-tive HLA-matched allo-SCT for hematologic malignancies, to categorize T cells with a naïve phenotype and the phenotypically distinct subsets of memorytype T cells (ie, central memory and effector memory cells) [9]. Expression of the costimulation receptor, CD28, was also analyzed, because CD28 expression is lost after several rounds of cell division, and segregating CD81 T cells by CD28 staining might be indicative of their functional capacities [10,11]. In a first step, we concentrated on 3 subgroups of recipients according to whether or not they had developed only 1 posttransplant complication: 1 without evidence of infection, GVHD, or relapse throughout the whole period of follow-up; 1 with only cGVHD; and 1 with only evidence of a malignant relapse after transplantation. These subgroups allowed comparing kinetics of T cell recovery in the absence of possible interferences by other clinical complications. Poor recovery of the CD81 T cell compartment by 2 months postgraft, resulting in reduced numbers of CD81 T cells that lacked CD28 expression, was significantly correlated with later occurrence of malignant relapse, as verified in a second step by univariate and multivariate analyses of the entire cohort.

MATERIALS AND METHODS

Patients

From October 2003 on, absolute counts and relative frequencies of peripheral blood T cell subsets were determined prospectively in 80 allo-SCT recipients. Their demographic and clinical characteristics are given in Table 1. All patients (and parents of children) provided informed consent, with approval from our institutional ethics committee. All patients were treated for a hematologic malignancy; all were in complete hematologic remission at time of transplantation, and none were at high risk of relapse after allo-SCT. A slight majority of patients received myeloablative conditioning including radiation and chemotherapy instead of myeloablative chemotherapy alone. All patients underwent T cell-replete allo-SCT followed by standard GVHD prophylaxis with cyclosporine (CsA) and short-term methotrexate (MTX). For all patients, it was their first transplantation. Stem cell source was bone marrow in the majority of patients

(n = 67) and otherwise granulocyte colony stimulating factor (G-CSF)-mobilized peripheral blood stem cells (PBSCS). Patients were monitored prospectively in terms of clinical complications, including graft rejection, acute GVHD (aGVHD) and cGVHD, infection, and relapse. This includes weekly cytomegalovirus (CMV) pp65 antigenemia assays to monitor for CMV reactivation during the first 3 months postgraft. No prophylaxis or preemptive treatment for antigenemia in the absence of other clinical or biologic CMV-related symptom was given.

Enumeration of CD41 and CD81 T Cell Subsets

One milliliter of the final graft product was obtained immediately before infusion for evaluation of graft composition. Peripheral blood samples were collected from recipients 30 6 2, 60 6 3, 90 6 5, 180 6 10, and 365 6 15 days after allo-SCT. All living patients had at least 3 samples analyzed. Whole-blood and graft samples were stained without further separation shortly after collection, acquired by multiparameter flow cytometry and categorized by their profiles of CD45RA, CC-chemokine receptor-7 (CCR7), and CD28 expression, as described previously [12,13]. All reagents and instrumentation were from Beckman Coulter (Fullerton, CA), with the exception of the monoclonal antibody (mAb) to CCR7, from R&D systems (Minneapolis, MN). Absolute counts and percentages of naïve (CD45RA1 CCR71 CD281) and differentiated T cells with a central memory (TCM, CD45RAneg CCR71), effector memory (Tem, CD45RAneg CCR7neg), and CD45RA1 effector memory (TemRA, CD45RA1 CCR7neg) phenotype [9] were determined within the CD41 and the CD81 T cell populations. The proportion of cells coexpress-ing CD28 was determined within each of these subsets. The degree of donor chimerism among CD31 T cells was assessed by real-time polymerase chain reaction (PCR) with a sensitivity <1% [14].

Statistical Analysis

Patient characteristics are presented as median (range) unless otherwise stated. Normalization of the absolute counts of a given lymphocyte population in the peripheral blood of a patient was defined as reaching the fifth percentile ofage-matched reference values of this laboratory [15]. The results were first summarized with descriptive statistics, including medians and interquartile ranges (25th-75th percentiles; interquartile range [IQR]). An initial analysis of the data was performed univariably using rank-order nonpara-metric tests and log-likelihood ratio G-test or Fisher's exact test, as appropriate. Statistical correlations were performed using Spearman's rank test. For the patient outcome study, probabilities were estimated by the Kaplan-Meier product-limit method, and the log-rank

Table 1. Characteristics of donor/recipient pairs.

All Patients Complication Free* Only Relapse Only Chronic GVHDf n

No. of patients (male/ female) 80 (43 / 37) 18(7/ 11) 15 (12/3) 13 (9 / 4) —

Age of recipients: median (range), years 32 (4-56) 24 (4-56) 33 (5-55) 25 (14- 55) NS

No. <16 years of age 15 6 4 2 NS

Underlying disease, no. of patients:§

Myelogenous proliferation (AML / CML / MDS) 53 (28/7/18) 14(9/2/3) 9 (8 / — / 1) 6 (3 / 1 / 2) NS

Lymphoid proliferation (ALL / CLL / NHL) 27 (24/1/2) 4 (4 / — / —) 6 (6 / — / —) 7 (6 / 1 / —)

Age of donors: median (range), years 35 (2-63) 31 (2-58) 37 (9-55) 30 (9- 57) NS

Donor-Recipient sex mismatch, no.^ 23 3 3 7 NS

Donor-Recipient CMV-serostatus (either or both positive), no. 49 12 7 9 NS

Donor/recipient pairs, no.:

HLA-identical sibling 54 14 9 11

HLA-identical (10/10), unrelated 20 4 4 2 NS

HLA-Cw mismatched (9/10), unrelated 6 — 2 —

Source of graft, no.:

Bone marrow 67 17 12 10

G-CSF-mobilized PBSC 13 1 3 3 NS

Infused CD34+ stem cells x 107/ kg recipient's weight: median (range) 2.98 (1.05-13.9) 3.38 (1.23-8.5) 3.44 (1.22-7.4) 2.69 (1.23- 8.5) NS

Infused T cells x 106 / kg recipient's weight: median (interquartile range):

all CD3+ T cells 24.3 (18.0-35.1) 22.5 (24.3-36.1) 27.9 (20.0-42.1) 27.1 (20.7-35.8) NS

CD4+ T cells 13.0 (8.8-18.4) 10.6 (6.0-14.8) 17.7 (9.8-22.7) 14.8(11.2-19.5) NS

CD8+ T cells 11.0 (8.4-17.2) 10.0 (7.6-15.2) 13.0 (10.2-19.8) 11.0 (9.2-17.5) NS

Myeloablative conditioning regimens:t

Busulfan- Cyclophosphamide 32 12 4 4 .03

Cyclophosphamide— 12-Gray TBI 48 6 11 9 **

Cyclosporine—short-term Methotrexate prophylaxis for acute GVHD, no. of patients: 80 18 15 13 —

Day 60 chimerism: median (interquartile range), percentage of donor T cells 98.5 (93-100) 96 (88- 99) 94 (98-100) 97.5 (96-100) NS

CMV reactivation, no. of patients: median (range) days postgraft: 15 — — —

45 (5-141)

Acute GVHD, no. of patients: median (range) days postgraft: 23 — — —

31 (13-61)

Chronic GVHD, no. of patients: median (range) days postgraft: 30 — — 13

184 (57- 971) 202 (105-971)

Relapse, no. of patients: median (range), days postgraft: 26 — 15 —

217 (85- 561) 215 (95- 561)

Whole follow-up period: days postallograft 502 (64-1474) 897 (395-1564) 487 (122-841) 1012 (542-1389)

AML indicates acute myelogenous leukemia; CML, chronic myelogenous leukemia; GVHD, graft-versus-host dieases; C-GSF, granulocyte-colony stimulating factor; NHL, non-Hodgkin lymphoma; CMV, cytomegalovirus; MDS, myelodysplastic syndrome; ALL, acute lymphoblastic leukemia; CLL, chronic lymphocytic leukeima; PBSC, peripheral blood stem cells; TBI, total body irradiation.

*Complication-free patients: those who did not develop any transplantation-related complication (severe infection, viral infection, acute or chronic GVHD, graft rejection) and who did not relapse over the whole period of follow-up.

fChronic GVHD without prior acute GVHD, scored according to standard criteria (10 limited, 3 extensive). ^Comparisons between the 3 subgroups of patients, using the Kruskal-Wallis test or log-likelihood ratio G-test as appropriate.

§AML, CML in chronic phase and MDS without blast crisis; ALL, NHL (low grade). All patients with acute leukemia were in complete remission at the time of transplantation. Log-likelihood ratio G-test compared AML/CML/MDS versus ALL/NHL.

^Defined as a male patient undergoing transplantation with cells of a female donor.

^Busulfan plus cyclophosphamide (200 mg/kg) or cyclophosphamide (60 mg/kg) plus fractioned TBI (12 Gray).

**Post hoc test revealed a significant difference in the type of myeloablative conditioning regimen between the complication-free patients and those who only relapsed, whereas myeloablative preparative regimen did not differ between patients who relapsed and those who had only cGVHD.

«о os

с —

О О.

о -о

test was used to evaluate differences between stratified patient subgroups. Potential prognostic factors for relapse were brought into a multivariate analysis using Cox proportional-hazards model. Candidate risk factors included in the model were the logarithmic transformation of lymphocyte subset counts, age of the recipient, myeloablative conditioning regimen, source of stem cells, donor-patient histocompatibility, disease type, CMV serostatus, and all other clinical and biologic factors as listed in Table 1. Results were confirmed using a competitive risk regression model [16]. All results were considered statistically significant at the 5% level (2-sided).

RESULTS

Overall, 18 of the 80 myeloablative allo-SCT recipients were free of any clinical complication by the time of last follow-up (395-1564 days postgraft). These patients less commonly received total body irradiation (TBI) containing preparative regimen, and all received grafts from 10 of 10 HLA-matched donors. Two other subgroups were distinguished, 1 with only cGVHD and 1 with only relapse. These 2 subgroups were comparable as regards the proportion of patients younger or older than 16 years of age, of patients receiving preparative TBI, of marrow versus PBSC allografts, and of sibling versus unrelated donors. There were, however, 2 patients who relapsed after receiving a graft with 1 HLA-Cw mismatch, whereas all patients in the subgroup with only cGVHD received a 10 of 10 HLA-matched allograft (Table 1).

Recovery of T Cell Subsets in Long-Term Complication-Free Patients

Inasmuch as immune recovery can be affected by the occurrence of postgraft complications and their treatment, the long-term complication-free patients allowed assessing T cell reconstitution in the absence of interfering factors. Absolute counts of CD41 T cells remained below the fifth or the 10th percentile of age-matched reference values throughout the first year postgraft, with subsequent reversal of the normal CD4:CD8 ratio (not depicted). Absolute counts of CD81 T cells increased rapidly, reaching the fifth percentile of age-matched reference values within 3 months in 8 of the 18 complication-free patients and otherwise by 6 months (Figure 1a). Virtually all circulating T cells were of donor origin (Table 1). There were no significant differences in the rate of recovery comparing complication-free patients younger or older than 16 years of age, marrow, and PBSC grafts, sibling and unrelated donor status, and whether mye-loablative conditioning regimen included TBI or not.

Figure!. Median absolute counts of peripheral blood CD8 Tcells(a) and their CD28-negative subset (b) by days 30,60,90, 180, and 365 after myeloablative HLA-matched allo-SCT, comparing patients free of any clinical complication throughout 395 to 1564 days of follow-up (diamonds and thick solid line) with patients who experienced only cGVHD (triangles and thin solid line) and patients with relapse (squares and dashed line) exclusive of other complication. Error bars indicate the 25th to 75th percentiles. The number of patients studied who were complication-free, who had only cGVHD or only malignant relapse were 17, 13, and 14, respectively, on days 30; 18, 13, and 15 on day 60 and on day 90; 16, 13, and 10 on day 180; and 8, 10, and 8 on day 365. Positions are depicted next to each other around the time point to prevent overlaying 1 subgroup value with another. The horizontal dotted line in (a) denotes the fifth percentile of reference values for the median age of the recipients. *P # .05; **P # .01; ***P # .001 by Kruskal Wallis test with Dunn's posttest evaluation to compare complication-free patients with those who relapsed; differences between complication-free patients and patients with only cGVHD were not significant whatever the time point.

At all time points the vast majority of circulating T cells were lacking CC-chemokine receptor-7 (CCR7) expression, indicative of an effector memory-like phenotype. Within the CD41 population, the CD45RAneg CCR7neg Tem phenotype was disproportionately abundant, resulting in an inversion (median of 0.42 by day 60; IQR 0.30-0.70) of the normal TCM:TEM ratio evidenced in the corresponding graft sample (1.42; IQR: 0.67-2.49) and in age-matched healthy controls. Within CD81 T cells, TEM represented a median of 25.9% of all CD81 T cells (IQR 16.6%-39.6%) and CD81 TemRA 49.6% (IQR 45.4%-66.1%) by day 60 postgraft.

Circulating CD81 lacking CD28 expression accumulated steadily over the first year (Figure 1b). Whether in adult or in younger recipients, the percentages of CD28neg CD81 T cells oversized the percentages determined in paired graft samples, whatever the memory-type subset and whatever the elapsed

time since transplantation (P # .003 by Friedman test). There was no significant correlation between the total number of CD81 T cells infused with the graft and absolute counts of peripheral blood CD28neg CD81 T cells in the recipient (p = .09; P = .75 by day 60). Overall, CD28neg cells predominated within the CD81TEM and TEMRA subsets and altogether peaked at a median of 69.4% (IQR 54.5%-78.5%) of all CD81 T cells by day 60. The proportion of CD28neg CD81 T cells slightly diminished thereafter to a median of 55.2% (IQR 41.3-56.8) by day 365, still remaining within the upper quartile of reference values in adult recipients, whereas pediatric patients kept throughout a higher percentage than normal for their age. Absolute numbers of CD28neg CD8 T cells in patients who were younger than 16 years of age at time of transplantation and those 16 years and older were compared, but no statistically significant difference was evidenced (P = .09 by a Mann-Whitney test). All these 18 recipients were devoid of clinical and biologic signs of CMV reactivation, and combinations of donor and recipient CMV serostatus were not significantly associated to circulating CD28neg CD81 T cell counts.

Recovery of T Cell Subsets in Patients with cGVHD

Peripheral blood CD41 T cell counts increased as slowly and CD81 T cell counts as rapidly in patients who developed de novo cGVHD as in the complication-free subgroup (Figure 1a). Frequencies of CD41 and CD81 T cell subsets, as categorized by their profiles of CD45RA and CCR7 expression, were also substantially similar (data not shown). Absolute numbers and percentages of CD81 T cells lacking CD28 expression tended to be higher by day 60, but there was substantial overlap with values from the complication-free subgroup, and formal statistical significance was not achieved (Figure 1b). The difference remained nonsignificant when the comparison included only patients who underwent bone marrow transplantation. Individual CD41 and CD81 T cell subsets in peripheral blood sampled after cGVHD was diagnosed were compared with values from the complication-free subgroup sampled at similar times. No statistically significant difference was evidenced, even if the patients with ongoing cGVHD tended to have greater numbers of CD28neg CD81 T cells by day 365 (P = .09).

There were 7 patients who had aGVHD preceding cGVHD without other clinical complications. Their median count of CD28neg CD81 T cells was 180 per mm3 (range: 140-246) at day 60 (P = .44 compared with patients experiencing de novo cGVHD). Including these patients together with those with de novo cGVHD did not evidence a significant difference with the complication-free subgroup.

Recovery of the Main T Cell Subsets in Patients Who Relapsed

Full donor T cell chimerism was achieved among all patients who relapsed exclusive of any other clinical event (Table 1). Relative and absolute counts of CD41 T cells and their subsets were low and evolved similar to those of complication-free patients (data not shown). Absolute counts of CD81 T cells increased more slowly (Figure 1a), at around half of the values from complication-free patients, with subsequent higher CD4:CD8 ratios. There was a trend toward a smaller percentage of cells with a CD81 TemRA phenotype that achieved formal statistical significance (P = .01) only on day 90 postgraft. Absolute numbers and percentages of all CD81 T cells lacking CD28 expression were significantly diminished from day 30 onward (Figure 1b). The median percentage of CD28neg CD81 T cells represented around 1/4th that of complication-free patients at all time points. The 3 patients who relapsed after receiving PBSC grafts had as low CD28neg CD81 T cell numbers as those who relapsed after marrow grafts (range at day 60: 46-165 and 6-227/mm3, respectively; P = .78).

Five additional patients relapsed 71 to 316 days after being diagnosed first with de novo cGVHD. Their CD81 T cell counts and absolute numbers and percentages of CD28neg CD81 T cells (median of 94/mm3 at day 60) ranged intermediate between corresponding values from the patients who relapsed without previous cGVHD (median of 47.5/mm3) and the patients who were devoid of any clinical complication (median of 179/mm3).

Univariate and Multivariate Analysis of Factors Affecting Relapse

All the patient and donor factors listed in Table 1 as potential predictors of relapse fell short of significance at the .05 level in univariate analysis comparing the subgroups of patients without complication, with only cGVHD and with only relapse. Although the proportion of patients receiving preparative TBI was lower in the complication-free subgroup, it was comparable between patients who experienced cGVHD and those who relapsed. Likewise, the proportion of patients receiving PBSC instead of bone marrow grafts was similar between patients with cGVHD and those who relapsed (Table 1).

Risk of developing relapse was further examined univariably at day 60 postgraft by categorizing all the 80 patients by the median of CD28neg CD81 T cell counts from the complication-free subgroup (179 cells/mm3). Using day 60 values was elected because a complete set of phenotypic analyses was available for all the recipients at this time point, whereas some values were missing on day 30, because of insufficient lymphocyte counts for detailed subset analysis. All

but 5 of the 26 patients who subsequently relapsed had absolute numbers of CD28neg CD81 T cells <179 cells/mm3, a significant difference by log-rank test (Figure 2).

Cox regression analysis for relapse as the endpoint was performed to compare the effect of recovering low CD28neg CD81 T cell counts with the recipient age and disease type, the presence or absence of TBI included in the myeloablative conditioning regimen, the donor type (sibling versus unrelated), the source of stem cells (marrow versus PBSC), and quantities of CD81 T cells infused into patients. Prior to all analyses, proportional hazard assumptions were tested and met in the subgroups with and without relapse. Considering the entire patient cohort, likelihood of relapse correlated inversely with absolute CD28neg CD81 T cell counts by day 60 (n = 80, adjusted hazard ratio of 0.33; 95% confidence interval [CI]: 0.14-0.76; P = .01). CD28neg CD81 T cell counts being predictive of relapse was confirmed by competitive risks regression analysis after adjustment for the other factors presented in Table 1 (P = .01). The inverse correlation was also verified by day 90 (0.29; 95% CI 0.130.65; P = .003).

None of the donor and patients characteristics was significantly associated with relapse in multivariate modeling. Marrow graft transferred fewer numbers of mature T cells to the recipient than did G-CSF mobilized PBSC. However, when patients who received PBSC grafts were excluded and only marrow recipients were tested in a multivariate model, the inverse correlation between likelihood of relapse and low absolute CD28neg CD81 T cell counts by day 60 remained strong (hazard ratio: 0.41; P = .02). Likewise, not including in the model the patients who relapsed after receiving a graft mismatched at 1 HLA-Cw locus still evidenced only CD28neg CD81 T cell counts as associated with relapse (hazard ratio

Figure 2. Kaplan-Meier estimate of malignant relapse after myeloablative HLA-matched allo-SCT. The 80 patients were categorized as having absolute CD28neg CD81 T cell numbers by day 60 postgraft falling lower (solid line) versus equal to or higher than (dashed line) the median (179 CD28neg CD281 T cells/mm3 of peripheral blood) determined in the subgroup of 18 patients free of any clinical complication by the time of last follow-up. The P value is for the log-rank test.

0.33; P = .03). Finally, when pediatric patients were excluded to analyze patients aged 16 years or older at the time of transplantation, the inverse correlation between likelihood of relapse and CD28neg CD81 T cell counts was still significant (hazard ratio 0.29; P = .03).

DISCUSSION

Successful T cell recovery substantially reduces morbidity and mortality following allo-SCT for hematologic malignancies, but disease relapse and cGVHD remain major causes of late morbidity and mortality. This prospective analysis found that poor CD81 T cell recovery with diminished absolute CD28neg CD8 T cell counts in the first 2 months after myeloablative allo-SCT was significantly associated with later development of malignant relapse, whereas patients who developed progressive or de novo cGVHD without relapsing recovered T cell counts similar to patients free of any clinical complication.

The subgroup of patients who did not develop any transplantation-related complication and who did not relapse over a period of follow-up longer than 1 year allowed analyzing the kinetics of T cell reconstitution in the absence of possible interferences by infections and GVHD that can affect the number and memory phenotypes of T cells [10,17]. As expected, absolute counts of CD41 T cells remained low throughout the first year posttransplantation, whereas absolute counts of CD81 T cells increased rapidly over 3 to 6 months, however, with predominance of cells lacking CCR7 and CD28 expression and persisting low levels of cells with a naive (CD45RA1 CCR71 CD281) phenotype. There were presumably too few pediatric recipients in this series to detect any effect of age on recovery over the first year posttransplantation, but it is known that renewal of thymopoiesis cannot be evidenced before 6 months in young children and is slower in adolescents [3,5]. Accordingly, for the first several months after myeloablative transplantation, the number of circulating T cells depends primarily on homeostatic peripheral expansion rather than on the thymic function. There were also too few PBSC recipients to detect any faster recovery effect of this source of graft. As regards CD28neg CD81 T cells, however, their absolute counts were reported to be similar in the first year in marrow and in PBSC recipients [18]. Bone marrow grafts transfer around 10 times less mature T cells to the patient, but 3 to 4 more cell divisions would suffice to yield the same count of CD28neg CD81 T cells as in patients given PBSC grafts. CMV-specific CD81 T cells predominantly reside in the CCR7neg and CD28neg subsets [10], and we cannot formally exclude that CMV-reactive T cells disproportionately expanded whenever the donor and/or the recipient was

CMV-seropositive. However, this subgroup ofpatients was devoid of clinical and biologic signs of CMV reactivation, and there was no significant association between levels of CD28neg CD81 T cells and combinations of donor and recipient CMV serostatus. Homeostatic cytokines can also induce a stable loss of CD28 expression on expanding CD81 T cells [19,20], which in the absence of CMV disease suggests that homeostatic peripheral expansion directly contributed to the generation and accumulation of CD28neg CD81 T cells throughout the first year after allo-SCT. Accordingly, sustained accumulation of CD28neg CD81 T cells has been reported after highdose chemotherapy or depleting mAb [4,21,22].

Kinetics of T cell reconstitution in patients who developed cGVHD was fairly similar to that in complication-free recipients. It is generally considered that the continuing risks of developing cGVHD may persist until 12 or 15 months postgraft. Our follow-up period was longer, so that incidence of symptomatic cGVHD was not underestimated. Our observation confirms earlier reports showing that CD81 T cell counts tended to be higher [23] or not significantly associated [24] with cGVHD, and that patients with poor early lymphocyte recovery were not at higher risk of aGVHD or cGVHD [25]. A few cross-sectional studies have variously reported on decreased CD41 TCM [26,27], elevated CD41 TEM [26], increased [28] or decreased [29] CD81 TCM, and increased CD81 TemRA [29], in patients who had cGVHD at time of sampling. There also exists uncertainty on whether the proportion of CD8 T cells lacking CD28 (or the overlapping CD571 subset) is increased [30] or not [24] during ongoing cGVHD. One possible explanation is that all these reports analyzed mononu-clear cells (instead of whole blood as in this study), given that Ficoll purification can induce a bias in subset distribution, particularly among CD81 T cells [31]. Another difference is that some reports included patients after nonmyeloablative conditioning, who could be generally less lymphopenic, and therefore undergoing lesser homeostatic peripheral expansion.

cGVHD may involve both alloreactivity and au-toreactivity against the host [3]. In fully HLA-matched donor/recipient pairs, donor T cells that have gone through homeostatic peripheral expansion while being exposed to ubiquitous minor histocompatibility antigens or self-peptides on host cells could be driven to differentiate into alloreactive effectors or to display pathogenic autoreactivity [32], respectively. Patients who will go on to develop cGVHD might therefore be expected to accumulate memory-like T cells at least as quickly as patients who undergo immune reconstitution without massively expanding pathogenic clones, notwithstanding expected differences in the respective clonal composition of their peripheral T cells [33,34].

Contrary to cGVHD, malignant relapse was consistently preceded by diminished absolute CD81 T cell counts, at around half of the values from patients who did not relapse. This is in agreement with other reports showing that marrow recipients with a high repopulation of CD81 T cells have a lower incidence of leukemia relapse [25,35]. Absolute CD28neg CD81 T cell counts were diminished at 1/4th of that of patients who did not relapse. This difference was significant from 30 days onward, long before the patients went on to relapse (at a median of 215 days in the patients who relapsed without prior cGVHD and after 152 days or more in the patients with cGVHD preceding relapse). Multivariate proportional hazard analyses revealed low absolute CD28neg CD81 T cell counts by day 60 postgraft as the only significant risk factor of malignant relapse, and their predictive value was confirmed by using competitive risk multivariate modeling. All patients included in this study were transplanted in complete hematologic remission at transplantation, the distribution of underlying diseases was comparable whether or not patients went on to relapse, and no patient was at high risk of relapse. The proportion of PBSC recipients in subgroups that did or did not relapse was also comparable. Furthermore, multivariate proportional hazard analysis including only bone marrow recipients confirmed the inverse correlation between likelihood of relapse and absolute CD28neg CD81 T cell counts by day 60.

The GVL effect to prevent relapse can be mediated variously by donor-derived T cells and natural killer cells. Missing killer immunoglobulin-like receptor recognition is not expected to predominate after 10 of 10 HLA-matched grafts, and excluding from multivariate analysis the few patients receiving a graft with 1 HLA-Cw mismatch did not ablate the significance of the association between CD28neg CD81 T cell counts and risk of relapse. Homeostatic proliferation following lymphopenia may facilitate expansion of antitumor immunity. Then, insufficient expansion might result from insufficient interaction of donor T cells with antigens present in the host, including hematopoietic tissue-restricted minor histocompatibility antigens and leukemia-specific antigens that would expand tumor-specific T cell clones [1,6,7]. CD28neg CD81 T cells are endowed with high cytotoxic capacities and diversified cytokine production [8-11]. Lower absolute CD28neg CD81 T cell counts might therefore reflect lesser generation of T cells with eradicating potential against the hematologic malignancy, whether directly or through a reduced bystander (cytokine-dependent) effect.

This prospective study is the first to associate a diminished absolute number of circulating CD28neg CD81 T cells with a greater risk of malignant relapse after a myeloablative conditioning. Enumeration of circulating CD81 T cells lacking CD28 expression

by day 60 postgraft could assist in predicting risk of relapse. This can help build an algorithm aimed at accelerating the immune recovery by reducing the intensity of posttransplant immunosuppressive regimen [36] and considering the introduction of posttransplant immunotherapy (preemptive donor lymphocyte infusions [DLI]) to overcome the risk of disease relapse after the acute risks of allo-SCT have passed.

ACKNOWLEDGMENTS

Financial disclosure: This work received a grant from a PHRC régonal - Nord Pas-De-Calais.

The authors thank Julia Salleron, Stephane Darre, the hematology fellows, and other senior physicians. They are especially grateful to Virginie Dutriez, Véronique Lekeux, Naoual Sabaouni, Jacques Trauet, and staffs involved in the care of the BMT unit. They also wish to thank all the donors and patients who agreed to be enrolled in this study.

REFERENCES

1. Bleakley M, Riddell SR. Molecules and mechanisms of the graft-versus-leukaemia effect. Nat Rev Cancer. 2004;4:371-380.

2. Shlomchik WD. Graft-versus-host disease. Nat Rev Immunol. 2007;7:340-352.

3. Mackall CL, Fleisher TA, Brown MR, et al. Distinctions between CD81 and CD41 T-cell regenerative pathways result in prolonged T-cell subset imbalance after intensive chemotherapy. Blood. 1997;89:3700-3707.

4. Storek J, Joseph A, Espino G, et al. Immunity of patients surviving 20 to 30 years after allogeneic or syngeneic bone marrow transplantation. Blood. 2001;98:3505-3512.

5. de Vries E, van Tol MJ, van den Bergh RL, et al. Reconstitution of lymphocyte subpopulations after paediatric bone marrow transplantation. Bone Marrow Transplant. 2000;25:267-275.

6. Peggs KS, Mackinnon S. Immune reconstitution following haematopoietic stem cell transplantation. Br J Haematol. 2004;124: 407-420.

7. Williams KM, Hakim FT, Gress RE. T cell immune reconstitution following lymphodepletion. Semin Immunol. 2007;19: 318-330.

8. Goldrath AW, Luckey CJ, Park R, Benoist C, Mathis D. The molecular program induced in T cells undergoing homeostatic proliferation. Proc Natl Acad Sci USA. 2004;101:16885-16890.

9. Sallusto F, Geginat J, Lanzavecchia A. Central memory and effector memory T cell subsets: function, generation, and maintenance. Annu Rev Immunol. 2004;22:745-763.

10. van Lier RA, ten Berge IJ, Gamadia LE. Human CD8 T-cell differentiation in response to viruses. Nat Rev Immunol. 2003; 3:931-939.

11. Romero P, Zippelius A, Kurth I, et al. Four functionally distinct populations of human effector-memory CD8 T lymphocytes. J Immunol. 2007;178:4112-4119.

12. Yakoub-Agha I, Saule P, Depil S, et al. A high proportion of donor CD4 T cells expressing the lymph node-homing chemo-kine receptor CCR7 increases incidence and severity of acute graft-versus-host disease in patients undergoing allogeneic stem cell transplantation for hematologic malignancy. Leukemia. 2006;20:1557-1565.

13. Yakoub-Agha I, Saule P, Depil S, et al. Comparative analysis of naive and memory CD41 and CD81 T-cell subsets in bone marrow and G-CSF-mobilized peripheral blood stem cell allografts: impact of donor characteristics. Exp Hematol. 2007;35:861-871.

14. Alizadeh M, Bernard M, Danic B, et al. Quantitative assessment of hematopoietic chimerism after bone marrow transplantation by real-time quantitative polymerase chain reaction. Blood. 2002;99:4618-4625.

15. Saule P, Trauet J, Dutriez V, et al. Accumulation of memory T cells from childhood to old age: central and effector memory cells in CD4 versus effector memory and terminally differentiated memory cells in CD81 compartment. Mech Ageing Dev. 2006;127:274-281.

16. Kim HT. Cumulative incidence in competing risks data and competing risks regression analysis. Clin Cancer Res. 2007;13: 559-565.

17. Hakki M, Riddell SR, Storek J, et al. Immune reconstitution to cytomegalovirus after allogeneic hematopoietic stem cell transplantation: impact of host factors, drug therapy, and subclinical reactivation. Blood. 2003;102:3060-3067.

18. Storek J, Dawson MA, Storer B, et al. Immune reconstitution after allogeneic marrow transplantation compared with blood stem cell transplantation. Blood. 2001;97:3380-3389.

19. Borthwick NJ, Lowdell M, Salmon M, Akbar AN. Loss of CD28 expression on CD81 T cells is induced by IL-2 receptor gamma chain signalling cytokines and type IIFN, and increases susceptibility to activation-induced apoptosis. Int Immunol. 2000;12: 1005-1013.

20. Chiu WK, Fann M, Weng NP. Generation and growth of CD28nuIlCD8+ memory T cells mediated by IL-15 and its induced cytokines. J Immunol. 2006;177:7802-7810.

21. Fagnoni FF, Lozza L, Zibera C, et al. T-cell dynamics after high-dose chemotherapy in adults: elucidation of the elusive CD8 subset reveals multiple homeostatic T-cell compartments with distinct implications for immune competence. Immunology. 2002;106:27-37.

22. Trzonkowski P, Zilvetti M, Chapman S, et al. Homeostatic repopulation by CD28~CD8+ T cells in alemtuzumab-depleted kidney transplant recipients treated with reduced immunosup-pression. Am J Transplant. 2008;8:338-347.

23. Abrahamsen IW, Semme S, Heldal D, Egeland T, Kvale D, Tjennfjord GE. Immune reconstitution after allogeneic stem cell transplantation: the impact of stem cell source and graft-versus-host disease. Haematologica. 2005;90:86-93.

24. Baron F, Storer B, Maris MB, et al. Unrelated donor status and high donor age independently affect immunologic recovery after nonmyeloablative conditioning. Biol Blood Marrow Transplant. 2006;12:1176-1187.

25. Ishaqi MK, Afzal S, Dupuis A, Doyle J, Gassas A. Early lymphocyte recovery post-allogeneic hematopoietic stem cell transplantation is associated with significant graft-versus-leukemia effect without increase in graft-versus-host disease in pediatric acute lymphoblastic leukemia. Bone Marrow Transplant. 2008;41: 245-252.

26. Yamashita K, Choi U, Woltz PC, et al. Severe chronic graft-versus-host disease is characterized by a preponderance of CD4 effector memory cells relative to central memory cells. Blood. 2004;103:3986-3988.

27. Fukunaga A, Ishikawa T, Kishihata M, Shindo T, Hori T, Uchiyama T. Altered homeostasis of CD41 memory T cells in allogeneic hematopoietic stem cell transplant recipients: chronic graft-versus-host disease enhances T cell differentiation and exhausts central memory T cell pool. Biol Blood Marrow Transplant. 2007;13:1176-1184.

28. Yamashita K, Horwitz ME, Kwatemaa A, et al. Unique abnormalities of CD41 and CD81 central memory cells associated with chronic graft-versus-host disease improve after extracorpo-real photopheresis. Biol Blood Marrow Transplant. 2006;12:22-30.

29. D'Asaro M, Dieli F, Caccamo N, Musso M, Porretto F, Salerno A. Increase of CCR72 CD45RA1 CD8 T cells (TEMRA) in chronic graft-versus-host disease. Leukemia. 2006;20:545-547.

30. Fukuda H, Nakamura H, Tominaga N, et al. Marked increase of CD8+S6F1+ and CD8+CD57+ cells in patients with graft-versus-host disease after allogeneic bone marrow transplantation. Bone Marrow Transplant. 1994;13:181-185.

31. Appay V, Reynard S, Voelter V, Romero P, Speiser DE, Leyvraz S. Immuno-monitoring of CD8 T cells in whole blood versus PBMC samples. J Immunol Methods. 2006;309: 192-199.

32. Baccala R, Theofilopoulos AN. The new paradigm ofT-cell ho-meostatic proliferation-induced autoimmunity. Trends Immunol. 2005;26:5-8.

33. O'Keefe CL, Gondek L, Davis R, et al. Molecular analysis of al-loreactive CTL post-hemopoietic stem cell transplantation. JImmunol. 2007;179:2013-2022.

34. Beck RC, Wlodarski M, Gondek L, et al. Efficient identification of T-cell clones associated with graft-versus-host disease in

target tissue allows for subsequent detection in peripheral blood. Br J Haematol. 2005;129:411-419.

35. Dolstra H, Preijers F, Van de Wiel-van Kemenade E, SchattenbergA, GalamaJ, de WitteT. Expansion ofCD8+CD57+ T cells after allogeneic BMT is related with a low incidence of relapse and with cytomegalovirus infection. Br J Haematol. 1995; 90:300-730.

36. Bacigalupo A, Vitale V, Corvo R, et al. The combined effect of total body irradiation (TBI) and cyclosporin A (CyA) on the risk of relapse in patients with acute myeloid leukaemia undergoing allogeneic bone marrow transplantation. Br J Haematol. 2000; 108:99-104.