High Absolute T Regulatory Cell Counts and Resistance of CD8 Central Memory Cells to Killing by Fludarabine Predicts for Poor Responses to DLI in the Context of a Pre-DLI Lymphoreduction Strategy

Abstract 1912

The response to donor leukocyte infusions (DLI) can be improved by the use of pre-DLI lymphoreduction. The mechanism underlying this is poorly explained. In order to investigate this further we examined the T cell subsets in patients entered into a clinical trial testing the utility of pre-DLI lymphoreduction using fludarabine.

Inclusion criteria were: 1. mixed chimerism (MC) in whole blood (< 95% donor), 2. previous reduced intensity transplant for a haematological malignancy or failure to respond to a previous dose of DLI and 3. a lymphocyte count of >1.0 × 10⁹/l. Patients consenting to the trial received three doses of oral fludarabine as an outpatient at 25mg/m² on day −7, −6 and −5. DLI was given on day 0. The starting dose of DLI for patients with MC alone was 5 × 10⁵ CD3/kg (unrelated donor) or 1 × 10⁶ CD3/kg (sibling donor). Higher doses were used in those with evidence of both disease and MC. Samples were collected at D-7, D0, D+7, D+28, D+60 and D90 for T cell subset analysis.

Samples were analysed for a combination of markers for CD4, CD8 and T regs using multicolour flow cytometry. CD4 and CD8 were analysed for subset: central memory (CM 62L+RA-), effector memory (EM 62L-RA-), naïve (N 62L+RA+) and effectors (EFF 62L-RA+). Tregs were characterised as CD4+25hi127loFOXP3+.

The study was a phase II pilot study with 18 patients required. At the end of the trial, 17 patients were eligible for analysis (1 early death due to leukaemia relapse). The disease categories were: AML (7), ALL (1), MDS (4), T-PLL (1), HD (2), DLBCL (1) and MCL (1). Only 2 patients had co-existing evidence of disease (morphological relapse of AML, 2% positive by immunophenotyping in T-PLL). The median age was 55 years (range: 22–66). There were 10 males and 7 females. 12 patients had HLA-matched sibling donors and 5 had 10/10 matched unrelated donors (UD). The majority of patients (13) had conditioning with fludarabine, melphalan and alemtuzumab, with cyclosporine post transplant.

Each received a single dose of DLI on the trial. The median time to DLI post-transplant was 7.4 months (range: 4–11.7). The median percentage of donor chimerism pre-DLI was 85% (range: 18–93). The median lymphocyte count pre-DLI was 1.3 (range: 1.0–3.8).

Of the 17 eligible patients, 11 (65%) responded to a single dose of DLI (CR=9 (>95%), PR=2 (chimerism 94% donor at study end)). Five patients developed GvHD – grade 1 (n=2), grade 2 (n=1), grade III/IV (n=2). GVHD resolved completely in all cases. 3 patients reactivated a virus (CMV, 2 × EBV), although not all required treatment.

No pre-DLI clinical factor had a significant impact on DLI responsiveness including: disease type, age, gender, donor type, product source (sibling, UD), pre-DLI chimerism, pre-DLI lymphocyte count, time to DLI or source of DLI (mobilised or non-mobilised cells). Post DLI GvHD did not have a significant association with response.

There was a significant decrease in the lymphocyte count, as well as the absolute numbers of CD8 (unpaired t test, p<0.0001) and CD4 (unpaired t test, p<0.005) cells between day D-7 and D0 in all patients. The absolute number of Tregs was not significantly different between D-7 and D0, however, DLI responders had a significantly higher absolute Treg count at all time points compared to non-responders (unpaired t test, p<0.005). In addition, DLI responders had a significantly higher Treg/CD8 ratio of means compared to non-responders at all time points (unpaired t test p<0.05). There were no significant differences in CD4 or CD8 counts between responders and non-responders. CD8 subset analysis showed a significant reduction in the CD8 CM cells in the responders (unpaired t test, p<0.001) between D-7 and D0, while this was not seen in the non-responders (p=0.3).

In conclusion, lymphoreduction with fludarabine pre-DLI is associated with high response rates to a single dose of DLI and a low incidence of GvHD. Interestingly, Treg counts do not appear to be reduced by fludarabine and higher Tregs counts at all time points were associated with responsiveness to DLI. Conversely, CD8 subsets were reduced following fludarabine, but relative resistance in CD8 memory cells was associated with poor responses to DLI. These findings suggest future strategies to test novel delivery of DLI as well as potential mechanisms of action.