Patients and Methods
Patients and transplantation
Adult patients with acute myeloid leukemia (AML) or non-Hodgkin’s lymphoma (NHL) undergoing first allogeneic marrow or filgrastim-mobilized blood cell transplantation in Calgary between December 2005 and September 2008 and who consented to participate and give blood were included. Their day 56 (range = 52-62) blood was collected and mononuclear cells (MNCs) cryopreserved. We chose to analyze MNCs collected on day 56 as this would enable us to evaluate all patients prior to developing relapse (the earliest day of relapse in our cohort was day 75). We did not choose an earlier time point as our previous experience with enumerating EBV specific T cells on day 28 showed that these were detectable in less than 50 % of patients (2). Patients undergoing HCT after September 2008 were not included as we strived for a minimum follow up of 2 years, as most relapses occur within 2 years posttransplant. The above criteria were met by 61 AML patients and 30 NHL patients. Of the 61 AML patients, 13 were excluded due to GVHD development by day 56 and 6 were excluded due to development of second malignancy (5 posttransplant lymphoproliferative disorders (median day of diagnosis = 62, range = 48-85) and 1 signet ring carcinoma diagnosed on day 235), leaving 42 patients for analyses. Patients who developed GVHD by day 56 were excluded as these patients would not be considered for preemptive therapy such as withdrawal of immunosuppression or donor lymphocyte infusion. We also excluded patients who developed second malignancy as these patients received treatments altering their immunity (eg, withdrawal of immunosuppression, donor lymphocyte infusion, rituximab or chemotherapy) confounding our ability to find an association between anti-herpesviral immunity and anti-leukemic immunity. Graft failure was also considered an exclusion criterion, but no patient developed graft failure. Of the 30 NHL patients, 6 were excluded due to GVHD development by day 56 and 1 was lost to follow up on day 155, leaving 23 patients for analysis. As the number of NHL patients was small (n=23), we could not answer the question of potential association between herpesvirus specific T cells and NHL relapse conclusively. Therefore, this report focuses on the 42 AML patients. For preliminary data on the potential association between herpesvirus specific T cells and NHL relapse, see Master’s Thesis entitled “High Herpesvirus Specific T cell Counts are Associated with Near-Zero Likelihood of Leukemia Relapse After Hematopoietic Cell Transplantation” published on the National Library of Canada’s website; www.nlc-bnc.ca. To maximize the power of finding associations between herpersvirus specific T cells and infections, both the AML and the NHL patients were included into the analysis of infections presented in this letter to the editor. The transplantation protocol and definitions of outcomes have been previously described (1). The study was approved by the University of Calgary Research Ethics Board.
EBV, VZV and HHV6 serology
EBV and VZV serostatus in both donor and recipient was determined within 2 months pretransplant by a routine clinical laboratory from fresh sera. HHV6 serostatus was determined retrospectively from frozen sera of donors and recipients for which frozen serum was available. Presence of IgG for each virus was determined by enzymoimmunoassay - Captia (Trinity Biotech, Bray, Ireland) for EBV, Enzygnost (Siemens, Marburg, Germany) for VZV, and HHV6 IgG ELISA kit (Advanced Biotechnologies, Columbia, MD, USA) for HHV6.
Enumeration of herpesvirus specific T cells
Cytokine production was assessed after 18 h incubation of the MNCs with viral antigen (“18h assay”). The ability of virus specific T cells to proliferate was measured after 7 day incubation with viral antigen (“7 day assay”). EBV lysate (Virusys, Taneytown, MD) was used as stimulus at 5 µg/ml in the 18h assay and at 10 ug/ml in the 7 day assay, HHV6 lysate (Bioworld Consulting Laboratories, Mt. Airy, MD) at 12.5 µg/ml in the 18h assay and 5 ug/ml in the 7 day assay, and VZV lysate (Virusys, Taneytown, MD) at 5ug/ml in the 18h assay and 7.5 ug/ml in the 7 day assay. Overlapping 15-mer peptide pools spanning the following EBV proteins were used: BZLF1, EBNA3A+B+C (EBNA3) or LMP1+LMP2 (LMP1+2) or the HHV6 protein U56 protein. All peptides were purchased from JPT (Berlin, Germany) and were used at a final concentration of 2µg/ml in both the 18h and 7 day assays. Other stimuli were staphylococcal enterotoxin B (Sigma, Toronto, ON) at final concentration of 5 mg/ml as positive control, HIVgag 15-mer overlapping peptide pool (JPT, Berlin, Germany) at a final concentration of 2 mg/ml as negative control for overlapping peptides (all patients and donors were pretransplant HIV seronegative) and complete media as negative control for viral lysate. In preliminary experiments we showed that the background was similar when using complete medium alone or complete medium with lysate from corresponding uninfected cell lysate, so medium alone was used as negative control for virus lysates. Due to limited number of MNCs from 5 AML patients, these were not stimulated with LMP1+2 in the 18h assay (2 relapse, and 3 no relapse patients) and the 7 day assay was not performed with MNCs from 6 AML patients (1 relapse, and 5 no relapse patients).
The 18h assay has been described (2). The following fluorochrome-labeled antibodies were used for staining: FITC-TNFα, APC-IL2, PC7-IFNγ, eFluor605-CD8, eFluor450-CD4, PC7-CD3 (all from eBioscience, San Diego, CA), Alexa700-CD3 (BD Bioscience, Mississauga, ON, Canada) and PE-IFNγ (Miltenyi Biotec, Auburn, CA). Enumeration of T cell subsets was performed using FacsDIVA software (BD Biosciences, San Jose, CA).
In the 7 day assay MNCs were labeled with 3 µM CFSE (Sigma-Aldrich, St. Louis, MO) for 7 min at room temperature. Cells were washed in PBS with 10 % FBS and one million MNCs were incubated with each viral antigen (except EBV lysate) in a final volume of 0.6 ml of complete media. Three million MNCs were incubated with EBV lysate in a final volume of 2 ml of complete media. After 7 day incubation, cells were washed and stained as previously described (2), using the following antibodies: PC7-CD3, PE-CD4 and eFlour450-CD8 (all from eBioscience, San Diego, CA). The number of precursor T cells (number of T cells in the starting population that were able to proliferate in response to stimulus) can be calculated from the estimated number of cells in each division. This was determined using the proliferation tool in FlowJo version 7.6.1 (Tree Star Inc., Ashland, OR). The absolute count of precursor herpesvirus specific CD4 and CD8 T cells was determined from the absolute lymphocyte count on the day of blood draw, the acquired cell proportion on day 7 of the assay (determined as the acquired proportion of fluorospheres, eg, 0.8 if 40,000 of the 50,000 fluorospheres were acquired), and the number of herpesvirus specific precursor cells.
All flow cytometric data acquisition was performed on a FacsAria IIu (BD Biosciences, San Jose, CA). A total of 50,000 CD3+CD4+ cells were acquired whenever possible. In the 18h assay we enumerated CD4 and CD8 T cells specific for 7 different stimuli and producing 7 different combinations of cytokines, resulting in 98 herpesvirus specific T cell subsets. In the 7 day assay we enumerated CD4 and CD8 T cells capable of proliferating in response to 7 viral antigens, resulting in 14 herpesvirus specific T cell subsets. Thus, a total of 112 herpesvirus specific T cell subsets were enumerated.
Statistics
Fisher’s exact test was used to test significance of differences between patient groups for binary variables, and the Mann-Whitney-Wilcoxon (MWW) rank sum test for continuous variables. To determine whether there was an association between a specific T cell subset count (e.g., the count of EBV lysate specific CD4 T cells producing IFNγ) and subsequent relapse, for each T cell subset we determined whether there was a significant difference in the T cell subset count between patients who did vs did not relapse using MWW test. To determine whether there was a significant correlation between a T cell subset count and infection rate (number of infections per 365 days at risk), for each T cell subset and for each infection type (any, viral, bacterial, fungal) we performed Spearman’s rank correlation test.
The problem of multiple comparisons was approached in the following two ways: 1. When accepting p<0.05 in the above 112 MWW or Spearman comparisons, 5% (6/112) of the comparisons may show association of herpesvirus specific T cell subset count with relapse or infections by chance alone. If >5% (>6/112) comparisons showed the association, we calculated the 95% exact binomial confidence interval (CI) for the proportion of comparisons yielding the significant association. If the CI did not include 5%, we concluded that at least some of the associations were not by chance (that at least some of the subset counts were truly associated with relapse or infections). 2. Westfall-Young permutation analysis was performed to determine which subsets were likely associated with relapse/infections. This analysis, adjusting p values for multiple comparisons, takes into account the interdependency among observations (as it is possible that if herpesvirus specific T cell subset A is associated with relapse/infections, herpesvirus specific T cells subset B is associated as well).
Subsets identified with a p value <.05 in the Westfall-Young analysis were considered definitely associated with relapse, and subsets identified with a p value between .05 and .15 were considered as possibly associated with relapse. Subsets that were definitely (p<.05) and possibly (.05 ≤ p ≤ .15) were combined into a scoring system. For each subset, a score of 1 was assigned to patients who had T cell subset counts above the median, versus a score of 0 to patients who had T cell subset counts equal to or below the median. Multivariate analyses were performed using Poisson regression, to assess whether T cell subsets were significantly associated with infection rates in addition to known factors associated with risk of infections. MedCalc software (Mariakerke, Belgium) was used for chi square, Fisher’s and MWW tests. Stata software (StataCorp, College Station, Texas) was used for calculating Spearman’s rank sum test and Poisson regression. SAS software (SAS, Cary, NC) was used to perform the Westfall-Young permutation analysis.
Acknowledgements
We thank the patients for participating in research that could not benefit them but only future patients. This study could not happen without the dedication of the staff of the Alberta Blood and Marrow Transplant Program, especially Polly Louie, Lynne Fisk, Judy Wu, Diana Quinlan, Jan McLaughlin, Maggie Young, as well as all inpatient nurses (headed by Lorraine Harrison, Joanne Leavitt and Jody Seerup), outpatient nurses (headed by Marie-Josee Paquin and Naree Ager) and physicians including Drs. Ahsan Chaudhry, Nancy Zacarias, Ping Yue, Nizar Bahlis, Chris Brown, Andrew Daly, Peter Duggan, Michelle Geddes, Lynn Savoie, Douglas Stewart, Mona Shafey, Melaku Game, Loree Larratt and Robert Turner. We also thank Provincial Lab, especially Kevin Fonseca for continued support. Finally we thank Drs Don Morris, Chris Brown and Pere Santamaria for invaluable feedback during this study.
References
1. Podgorny PJ, Ugarte-Torres A, Liu Y, Williamson TS, Russell JA, Storek J. High rabbit-antihuman thymocyte globulin levels are associated with low likelihood of graft-vs-host disease and high likelihood of posttransplant lymphoproliferative disorder. Biol Blood Marrow Transplant 2010 Jul; 16(7): 915-926.
2. Hoegh-Petersen M, Goodyear D, Geddes MN, Liu S, Ugarte-Torres A, Liu Y, et al. High incidence of post transplant lymphoproliferative disorder after antithymocyte globulin-based conditioning and ineffective prediction by day 28 EBV-specific T lymphocyte counts. Bone Marrow Transplant, in press, doi:10.1038/bmt.2010.272.