Single center experience suggests donor lymphocyte infusion may promote long-term survival in children with high-risk acute lymphoblastic leukemia

Abstract

Background:

Donor lymphocyte infusion (DLI) is often used to treat leukemic relapse after hematopoietic cell transplantation (HCT). However, the relationship between outcomes and distinct DLI cellular composition has not been previously reported. Additionally, there is limited published data on efficacy in pediatrics. We evaluated whether DLI cellular content and development of graft-versus-host disease (GVHD) impacted disease and influenced overall survival (OS) in children receiving DLI for recurrent leukemia.

Methods:

We performed an IRB-approved, retrospective study investigating all consecutive DLIs given to patients at the Children’s Hospital of Wisconsin between 1980-2018. Analyses were conducted using Mann-Whitney, Fisher’s exact, and chi-square testing.

Results:

Thirty patients ≤20 years old with hematologic malignancies [myeloid (AML/MDS/CML/JMML), n=23; lymphoid (ALL), n=7] received DLI to treat post-transplant relapse. We found no significant difference in OS or development of GVHD based on CD3, CD4, CD8, CD56, or CD19 DLI cellular composition. With a median follow-up of 0.69 (range, 0.04-16.61) years, OS at 5 years was 32 ± 9%. The lymphoid group had a 5-year survival rate at 71 ± 17% compared to the myeloid group at 22 ± 9%, although not statistically significant (p=0.11). The development of GVHD did not affect OS (p=0.62).

Conclusion:

Here we report a single-center, long-term experience of pediatric DLIs. Surprisingly, many children with ALL were able to achieve durable remissions. While cellular composition did not have a significant effect on GVHD or OS in our small study, engineering DLI products to maximize specific effector cell populations could be one strategy to improve efficacy.

Introduction

Allogeneic hematopoietic cell transplantation (HCT) can be curative for pediatric malignancies. However, if disease returns after HCT, there are limited curative options, and thus relapse remains the leading cause of death after HCT.¹ Therapies to induce post-transplant remission include novel chemotherapeutic drugs, a second transplant, or donor lymphocyte infusions (DLI).² Taking advantage of the prior transplant as an immunotherapeutic platform, DLI is frequently chosen for its ease of administration, improved toxicity profile, and decreased risk of graft ablation compared to other options.

The primary goal of DLI is to achieve a graft versus leukemia (GVL) effect. GVL is the ability of the immune cells within the DLI product to kill malignant cells via a host of mechanisms, including direct target recognition.³ This GVL effect is felt to be primarily mediated by donor T cells. For this reason, DLI is dosed based on its CD3+ content.⁴ However, traditional DLI products are not engineered and thus consist of a variety of different cell types. It is currently unknown which subsets of T cells are most critical for producing the GVL effect and whether other effector populations in the DLI product (such as natural killer (NK) cells) contribute to GVL.^5–7

The use of DLI is limited by the association between GVL and graft versus host disease (GVHD). Many previous publications have suggested that GVHD is a marker for GVL and correlates closely with disease response.^{3, 8–9} However, it is unclear whether eliciting GVHD should be considered an accurate goal and clinical readout of potent GVL, as different cell populations may contribute to each response.^3,10 For example, previous studies in murine models have established the ability of NK cells and NKT cells to induce an anti-malignant cell response without resulting in GVHD.⁶

Finally, DLIs are not consistently efficacious across all hematological diseases. Historically, they have been found to be more potent in myeloid compared to lymphoid diseases, with the best results seen in patients with chronic myelogenous leukemia (CML).^8–9,11 However, much of the literature reported the effect of DLI for adult malignancies with limited focus on the pediatric population. Given that pediatric ALL can express different cell surface ligands compared to adult ALL, it may be presumptuous to extrapolate adult DLI data when considering pediatric malignancies.¹² To better inform the ongoing use of DLI in children, this study evaluated our center’s pediatric DLI experience over a 27-year period in regard to cellular composition of the DLI product, GVHD status in relation to overall survival (OS), and underlying disease indication.

Methods

We conducted a retrospective study, approved by the Institutional Review Board at the Children’s Hospital of Wisconsin (CHW), analyzing DLIs given to patients at CHW between 1991-2018. At CHW, DLIs have been used to treat relapsed disease, promote donor chimerism/engraftment, and/or to improve immune reconstitution. Our study population included unmanipulated (defined as unmodified and uncultured) DLIs given to patients for the purpose of treating relapsed disease. Patients with non-hematopoietic solid tumors and non-malignant diseases were excluded from this analysis. Using these criteria, we identified 30 pediatric patients with hematologic malignancies, comprising both lymphoid (n=7) and myeloid (n=23) diseases, who were included in our final analyses.

Data was collected via chart review and entered into RedCap™ (http://www.project-redcap.org/), which is a secure, internet-based database. Each DLI was administered based on a CD3 per kg dose, but the entire infused product underwent cellular enumeration to also determine cell concentrations per kg of helper T cells (CD4), cytotoxic T cells (CD8), B cells (CD19), and NK cells (CD56). Many patients received more than one infusion of DLI, often part of a planned dose escalation. Disease response after first DLI was classified as having complete response (no morphological evidence of disease), partial or stable response (any improvement or stability of blast count), or progression (any increase of blast count). Acute and chronic GVHD were combined in this retrospective review for the endpoint of GVHD. Survival was defined as time between first DLI to last follow-up or death. The probability of OS (Figures 2 and 3) was estimated using Kaplan-Meier curves with log-rank test for comparison between groups. Mann-Whitney and Fisher’s Exact tests were used to determine statistical correlations between DLI cellular content and the development of GVHD, disease response, and underlying diagnosis. Results reported as median (range) and n (%). SPSS statistical software version 24 (Chicago, Illinois, USA) was used to analyze the data.

FIGURE 2:

Probability of survival based on disease type

FIGURE 3:

Probability of survival based on development of GVHD

Results

There were a total of 30 subjects who received 55 consecutive and unmanipulated DLIs after HCT. Nineteen of these patients received only one DLI, while 11 patients received ≥2 DLIs. Of these 11 patients, 10 received DLIs sequentially (separated anywhere from 1 – 3 months apart) as part of a planned dose escalation to maintain disease control. Human leukocyte antigen (HLA) match grade affected initial and any subsequent doses of DLI. When evaluating HLA-matched related (n=9) and unrelated (n=6) donors, the initial median CD3/kg doses were 5 x 10⁷/kg (range, 0.01 – 10 x 10⁷/kg) and 1 x 10⁷/kg (range, 1-5 x 10⁷/kg), respectively. For those patients receiving subsequent DLIs, the median CD3/kg dose was escalated to a dose of 8 x 10⁷/kg. However, for those patients receiving DLI from HLA-mismatched related (n=7) and unrelated (n=8) donors, the initial median CD3/kg dose was lower at 5 x 10⁵/kg (range, 1 x10⁴ – 5 x 10⁷/kg) and 1 x 10⁷/kg (range, 4 x10⁴ – 3 x 10⁸/kg), and any subsequent infusions were only escalated to a median of 5 x 10⁶/kg CD3/kg.

The study cohort was divided based on disease group into “myeloid” and “lymphoid” diseases. The myeloid group (n=23) included the diagnoses of acute myelogenous leukemia (AML) (n=14), CML (n=4), juvenile myelomonocytic leukemia (JMML) (n=3), and myelodysplastic syndrome (MDS) (n=2). The lymphoid group (n=7) was comprised of only acute lymphoblastic leukemia (ALL). Table 1 displays the demographics of the 30 subjects. The only demographic variables that were statistically different between the groups were the median age at transplant and median age at first DLI, with the lymphoid group being significantly older than the myeloid group in both cases, respectively [age at transplant: 14.7 (7-20) vs 7.5 (0-20) years, p = 0.022; age at 1^st DLI: 20 (7-22) vs 8 (0-21) years, p = 0.006]. There was no statistically significant difference between the two groups regarding sex, race, number of DLIs received, HLA-match grade, or interval from transplant to first DLI.

Table 1:

Patient demographics

Patient Characteristics	Myeloid (N=23)	Lymphoid (N=7)
Sex
Male	65%	71%	P=0.999
Female	35%	29%
Race
White non-Hispanic	91%	86%	P=0.451
Hispanic	0%	14%
Other	9%	0%
Age at Transplant (years), median (range	7.5 (0-20)	14.7 (7-20)	P=0.022
Age at 1st DLI (years), median (range)	8.0 (0-21)	20.0 (7-22)	P=0.006
Number of DLI – 55 total
1	61%	71%	P=0.878
2	13%	0%
3+ (max infusions 5)	26%	29%
HLA-Match Grade
10/10 or 8/8 related	30%	29%	P=0.999
10/10 or 8/8 unrelated	22%	13%
Mismatched related1	22%	29%
Mismatched unrelated	26%	29%
Interval from HCT to 1st DLI (years), median (range)	0.6 (0.1-4.1)	1.0 (0.3-5.8)	P=0.532

¹“Mismatched related” refers to HLA-haploidentical transplants from both parents & siblings.

Data was gathered on the cellular content of the DLI products given to each subject. Table 2 contains the cellular composition of the infusion products consisting of the CD3, CD4, CD8, CD19, and CD56 per kg concentrations. There were no statistically significant differences between the lymphoid and myeloid groups in regard to cellular composition of the DLI products for the first infusion. However, if the total number of cells infused over all DLIs received by a patient were combined, the myeloid group had significantly more cells infused for all cell types compared to the lymphoid group (p<0.05 for all cell types). Concerning GVHD, there was no statistical difference in cellular content of the DLI products given to patients who developed GVHD after the 1^st DLI as compared to those who did not (p>0.05 for all cell types) (Table 3). Survival status based on cellular composition of the first DLI also was not statistically different (Fig 1). When taking into consideration that some patients received multiple DLIs, there was no significant association between the total number of all CD3, CD4, CD8, CD19, or CD56 cells infused and survival status (p>0.05 for all cell types).

Table 2:

Cellular composition of DLI products by disease type

Cell Type		Myeloid	Lymphoid
CD3/kgx106	First DLI	15.8 (<0.01-316.0)	5.8 (<0.01-92.9)	P = 0.098
	Total DLI	49.9 (<0.01-650.5)	5.8 (<0.01-92.9)	P = 0.028
CD4/kgx106	First DLI	7.2 (<0.01-179.5)	4.4 (<0.01-48.3)	P = 0.080
	Total DLI	20.0 (<0.01-440.8)	4.4 (<0.01-48.3)	P = 0.031
CD8/kgx106	First DLI	4.5 (<0.01-140.0)	1.4 (<0.01-31.8)	P = 0.113
	Total DLI	9.4 (<0.01-217.2)	1.4 (<0.01-31.8)	P= 0.031
CD19/kgx106	First DLI	3.1 (<0.01-44.3)	0.4 (<0.01-15.0)	P = 0.063
	Total DLI	8.3 (<0.01-228.1)	0.8 (<0.01-15.0)	P = 0.022
CD56/kgx106	First DLI	1.7 (<0.01-46.7)	0.6 (<0.01-8.4)	P = 0.162
	Total DLI	8.2 (<0.01-210.0)	1.1 (<0.01-8.4)	P = 0.036

Table 3:

DLI cellular content by development of GVHD after 1st DLI

Cell Type	No GVHD	+ GVHD
CD3/kgx106	10.0 (<0.01-316.0)	30.0 (1.0-100.2)	P = 0.346
CD4/kgx106	5.0 (<0.01-179.5)	6.3 (0.7-69.1)	P = 0.464
CD8/kgx106	3.3 (<0.01-140.0)	4.1 (0.3-34.6)	P = 0.508
CD19/kgx106	1.0 (<0.01-44.3)	3.3 (0.3-37.7)	P = 0.603
CD56/kgx106	1.1 (<0.01-46.7)	1.6(0.2-22.5)	P = 0.656

FIGURE 1:

Survival status and DLI cellular content

Evaluation of disease response was available for 28 of the 30 patients; however, survival status is known for all 30 patients. Of the 28 patients in whom disease response is known, only 6 (21%) patients maintained their same level of disease (stable) or developed partial or complete remission. Of these six patients with positive responses, four had ALL and two had AML. Of the ALL patients, three developed complete remission and one had stable disease. Both AML patients developed complete remission. Those patients who achieved remission remained in remission at their last follow-up. Twenty-two of the 30 patients died after their last DLI (73%). With a median follow-up of 0.69 (0.04-16.61) years, the OS for all 30 patients was 32 ± 9% at 5 years. The lymphoid group had a 5-year OS rate at 71 ± 17% compared to the myeloid group at 22 ± 9%, although this was not statistically significant (p=0.11) (Fig 2). In our analysis, the development of GVHD after HCT had no effect on development of GVHD after 1^st DLI. Furthermore, GVHD after 1^st DLI had no effect on OS (p=0.62). Specifically, those patients who developed GVHD had a 5-year OS of 36 ± 15%, compared to those who did not develop GVHD who had an OS of 39 ± 15% (Fig 3).

Discussion

Given the relapse rate of up to 50% after HCT among some patients with hematological malignancies,¹ and the poor ability to salvage many of these post-HCT relapses, with one recent report citing a 3-year OS of 13% in pediatric patients with leukemia and MDS,¹³ identifying efficacious relapse therapies are vital to promoting long-term remission and durable survival. Although the use of DLI is considered a common treatment for disease relapse, its lack of consistent efficacy lends itself to question its routine use as a standard treatment. This question is more pronounced in the pediatric population, where its potential use is extrapolated from the adult literature, given few published studies focusing specifically on children and young adults. Moreover, most studies published do not analyze the exact heterogeneous cellular composition of each infused DLI nor do they explore how the cellular composition of the infusion product affects the patients’ disease course.

This retrospective study focused on evaluating long-term outcomes of DLI recipients at CHW, with a particular focus on the cellular composition of the DLI. We evaluated cellular composition of the DLI products to determine if improved outcomes could be correlated with cell content. Our results showed that the cellular composition of the infusion product had no significant effect on the development of GVHD, disease response, or survival. However, we likely need a much larger sample size to distinguish subtle differences in cellular composition between these groups.⁴ The wide range of cellular content of the DLI products could in part be due to selecting stratified and specific doses of CD3/kg based on a variety of factors, including HLA-match grade between patient and donor. In other words, to achieve a lower CD3/kg dose in such settings of HLA-mismatch, the volume of the DLI product was decreased, which also diminished the numbers of NK and B cells infused. Additionally, while the CD3/kg dose was often increased incrementally for a goal infusion amount, this same target consideration was not given to the other cellular types contained within the DLI product.

Similar to other studies examining leukemia and the use of DLI as a relapse treatment,^4,14 the OS of our study cohort was poor at 32% at 5 years. Interestingly, upon comparing survival in regards to myeloid versus lymphoid disease groups, DLI promoted longer survival in our pediatric lymphoid group, with 5-year survival of 71% compared to 22% in the myeloid group. This finding was surprising given prior literature in adult populations reporting poorer outcomes in patients with lymphoid diseases.⁸ One reason for this discrepant result could be due to disease burden difference at the time of DLI between the groups, suggesting a selection bias on the part of the health care team when choosing this treatment modality. Given that myeloid patients are felt to have a more encouraging response to DLI,⁹ providers may have been more willing to give DLI to riskier myeloid patients with a higher disease burden. The differences between pediatric and adult cancers should also be considered regarding our study’s observation. Pediatric cancers possess unique antigens and therefore, could respond to treatments differently than adult diseases.¹⁵ Thus, our findings suggest that adult data may not be always generalizable to the pediatric population. Lastly, it should be noted that the development of GVHD did not affect OS in our cohort. This is surprising given that GVHD is commonly used as a surrogate for GVL. Our data suggests that GVHD may not be as accurate a marker for GVL as has previously been suggested.^3,8–9 However, this finding must be taken with the caveat that all degrees of GVHD were combined, and it is possible that mild forms were not well-reported in the older patient charts.

Our study has several limitations relating to the retrospective collection of data comprising two decades of time. During this period, supportive care has changed, as has donor typing sensitivity. GVHD was a combined composite data point with both acute and chronic combined, and stage, severity, and chronicity of GVHD were not uniformly available for most early patients and thus not included in this analysis. There is no clear understanding of why some patients received more than one DLI infusion, and there was no consistent formula for dose augmentation of patients receiving more than one DLI. While collection of cellular data subtypes for all DLIs is a relative strength of our study, more sophisticated analyses to further distinguish subsets of T and NK cells were not performed. All of these factors could have influenced survival outcomes. Another limitation of this retrospective study is that there is inadequate documentation in the medical record regarding the use of additional post-transplant maintenance strategies, which may have contributed to prolonged survival. However, unlike our modern era in which several maintenance strategies are often incorporated post-transplant to maintain remission, this was not standard practice during the earlier time period reviewed.¹⁶ Finally, as mentioned earlier, DLIs were based on targeted CD3 per kg doses, and thus a much larger population with more prominent differences in the other cell populations would be needed to find true effects of these different cell doses. This would be more easily achievable in a larger prospective, dose-finding clinical trial setting.

Based on these results, it is clear from our single-center study that for many pediatric patients with leukemia, unmodified DLI is not consistently effective and is not vigorous enough to induce remission. However, for the ALL cohort, there is some suggestion that DLIs could promote durable survival after HCT, and thus this group should be further studied.¹⁷ Adoptive immunotherapy in the form of modified DLI is being developed as one potential strategy to overcome the limitations of traditional DLI products. Our group and others have suggested the infusion of a modified DLI consisting of purified donor NK cells after HCT is one way to augment GVL without causing GVHD.^7,18,19,20 NK cells can contribute to GVL effects by using a combination of activating and inhibitory receptors, ligand mismatches, and cytokine activation to target malignant cells without prior sensitization.²¹ Other exciting developments in field of immunotherapy include bispecific T-cell engagers (BiTE), which are molecules containing two antibodies that will result in binding of both the CD3 domain of T cells and a surface antigen on the malignant cell. When bound to both targets simultaneously, the BiTE will signal the T cell to degranulate resulting in damage to the malignant cell.^22,23,24 A pediatric phase I/II trial demonstrated a complete remission rate of 39% after two cycles of BiTE therapy.²⁵ Additionally, there is ongoing research on suicide gene therapy, which is a method of promoting T-cell mediated GVL while having an active mechanism to control severe GVHD. This is achieved through the genetic engineering of alloreactive T cells with the addition of a suicide gene that can be selectively targeted with a prodrug.²⁶ The safety and feasibility of suicide gene therapy, as well as the efficacy in controlling GVHD, has been demonstrated in multiple clinical trials.²⁷ Lastly, increasing focus is being placed on the development of chimeric antigen receptor T and NK cells as treatment for relapse after HCT to fulfill a more targeted cellular immunotherapy approach.^28,29 Multiple centers have demonstrated early remission rates after CAR T therapy for ALL in upwards of 70-90%, although longevity of remarkable response is variable.^30–32 As post-transplant relapse continues to plague survival outcomes, these new immunotherapy developments are exciting treatments that are extending the armamentarium of relapse options for high-risk pediatric patients.

Abbreviations

DLI

Donor Lymphocyte Infusion

HCT

Hematopoietic Cell Transplantation

HLA

Human Leukocyte Antigen

GVL

Graft Versus Leukemia

GVHD

Graft Versus Host Disease

CML

Chronic Myelogenous Leukemia

CHW

Children’s Hospital of Wisconsin

OS

Overall Survival

AML

Acute Myelogenous Leukemia

JMML

Juvenile Myelomonocytic Leukemia

MDS

Myelodysplastic Syndrome

ALL

Acute Lymphoblastic Leukemia