August 31, 2016

Amir Ahmed Toor, MD

In Partnership With:

Investigators at Virginia Commonwealth University Massey Cancer Center are developing dynamical models that may allow clinicians to move beyond simple human leukocyte antigen matching to personalize immunosuppressive regimens and thus improve donor selection.

Amir Ahmed Toor, MD

Bone Marrow Transplant Program

Massey Cancer Center

Current clinical practice in allogeneic stem cell transplantation (SCT) is to identify donors based on the degree of human leukocyte antigen (HLA) matching. With this practice, immunosuppressive regimens are chosen using stochastic models based on the probability of disease relapse or development of graft-versus-host disease (GVHD).

Now, investigators at Virginia Commonwealth University Massey Cancer Center are developing dynamical models that may allow clinicians to move beyond simple HLA matching to personalize immunosuppressive regimens and thus improve donor selection.

This approach will hopefully lead to better clinical outcomes and the ability to bring more patients to transplant by reducing the likelihood of GVHD and other complications. Already, a clinical trial has been initiated based on the dynamical systems model of transplant outcomes (NCT02593123), and its results will further help guide an understanding of GVHD pathogenesis.

The emerging model grew out of studies conducted at VCU Massey Cancer Center that uncovered a large body of antigenic variation between transplant donors and recipients for which HLA matching does not account. This research was done using next-generation DNA sequencing to examine the exomes of stem cell transplant donor-recipient pairs (DRPs).

Figure 1. Evolution of State in Dynamical Versus Stochastic Systems

In dynamical model (top), deterministic growth of T cells over time leads to the successive states of the patient (eg, GVHD present or not). In the stochastic system, outcome probability changes with changing state of the system, and cannot be predetermined beyond a certain level of precision due to random influences.

Using matrix mathematics, a computer program has been developed that allows simulation of SCT donor T-cell response to recipient antigens.

Understanding GVHD Biology

The computational output from this program appears to correlate with clinical outcomes in stem cell transplant recipients, thus enabling clinicians to develop immunosuppressive regimens based on the antigenic variation in each DRP.Allogeneic SCT recipients are susceptible to the competing risks of GVHD, opportunistic infections, and relapsed primary malignancy. The common variable in all of these clinical outcomes is donor-derived immune reconstitution whereby recipient antigens, specifically minor histocompatibility antigens (mHAs), are recognized by nontolerant, donor-derived T cells.

Clinical manifestations of GVHD are mixed, with either an acute, rapidly progressive phenotype, or a more indolent, chronic phenotype. In the former, the skin, gastrointestinal tract, and the liver are frequent targets; in the latter, skin, mucosal surfaces, muscle, fascia, lungs, and liver are commonly involved.

GVHD occurrence, while potentially life threatening, is often associated with maintenance of cancer remission, presumably through donor T-cell recognition of tumor-associated and normal recipient antigens on the malignant cells. The treatment of GVHD, on the other hand, can predispose patients to opportunistic infections, which threaten survival.

Examining the Antigenic Landscape

In short, GVHD biology represents a therapeutic conundrum that has not been resolved to satisfaction despite advances in tissue histocompatibility matching techniques and continues to compromise clinical outcomes. What is more, the ability to perform haploidentical and HLA-mismatched umbilical cord blood transplantation indicates that there is more to the challenge of optimizing posttransplant outcomes than donor-recipient identity at the major histocompatibility (MHC) locus.Recognizing the role of donor T cells in transplant biology, the Bone Marrow Transplant Program at VCU Massey Cancer Center initiated a clinical trial in 2008 combining a polyclonal anti-T cell antibody preparation, rabbit antithymocyte globulin (ATG), with 4.5 cGy total body irradiation, followed by HLA-matched related or unrelated donor SCT, with donor lymphocyte infusions given for mixed T-cell chimerism.

In these patients, clinical outcomes were associated with the magnitude and rate of recovery of donor-derived T cells.^1,2 Further, when lymphocyte reconstitution was evaluated, it demonstrated logistic kinetics, with a sigmoid growth curve largely attributable to donor-derived T-cell proliferation. The T-cell repertoire emerging in patients following transplant had a complex, fractal organization^3,4 and raised the question of whether the underlying antigenic landscape might not be equally as varied.

To investigate this question, a donor and recipient whole-exome sequencing study was proposed in 2011, with the premise that nonsynonymous single nucleotide polymorphisms (nsSNP) present in the recipient but absent in the donor would lead to immunogenic peptides that might trigger donor T-cell activation. In aggregate, these polymorphisms would constitute an alloreactivity potential between the specific donors and recipients.

An initial cohort of 9 HLA-matched DRPs demonstrated an average of nearly 6000 nsSNPs across the entire exome.⁵ To determine whether HLA presentation “filters out” irrelevant nsSNPs, the flanking amino acid (AA) sequence for each of the nsSNP-coded-AA was established in order to obtain a polymorphic-oligopeptide library.

Figure 2. Targeted T-cell Expansion After Transplant

Illustration depicts HLA-dependent presentation of polymorphic recipient peptides and deterministic expansion of donor T cells in response to alloantigen. When this occurs for several hundred different antigens, it leads to eventual clinical outcome (state of the system).

APC indicates antigen-presenting cell; HLA, human leukocyte antigen; mHA, minor histocompatibility antigen.

This, in turn, enabled the determination of peptide-HLA binding affinity for each oligopeptide to the specific class I HLA molecules in each DRP in silico, using the publicly available NetMHCpan and IEDB software. A large array of putative mHA was identified using this technique with a range of HLA binding affinities in each individual.⁶

New Model for Transplants

These studies have now been extended in a larger cohort of patients (78 DRP) with verification of the earlier findings. This observation raised the question: if there is a large magnitude of potential antigens in each DRP, why do all patients not develop some measure of GVHD?One possible answer lies in modeling STCs as dynamical systems, which are physical systems where the components of the system evolve according to defined rules and the state of the system at any given time will predict the state to follow.⁷

Dynamical system modeling is unlike purely stochastic modeling where the state of the system is a probability function of the state at a previous time (Figure 1). In transplantation, the donor-recipient chimera constitutes the system, with the recipient tissues, donor antigen presenting cells, and T cells being the components.

In transplant recipients, mHA released following tissue injury are taken up by antigenpresenting cells and presented on HLA molecules to donor T cells which, in the event of antigen recognition, proliferate in a deterministic fashion and home back to the target tissue, initiating a graft-versus-host response (Figure 2).

Their collective state would then predict clinical outcomes. For example, the more that polymorphic antigens are presented on the HLA with greater avidity, the more alloreactive T cells are likely to proliferate, the more likely GVHD becomes, and vice versa. Since T-cell growth appears to clinically follow logistic dynamics— assuming that the T cells will react to specific mHA-HLA complexes—one may mathematically simulate the T-cell response to the previously described mHA arrays identified in each DRP in order to more precisely determine the alloreactivity potential that a unique DRP might have. Indeed, this hypothesis appears to be the case. Simulation of the CD8-positive T-cell response to the in silico—derived mHA–HLA complexes in 34 DRPs has led to the identification of very distinct patterns of putative cytotoxic T-cell reactivity.⁸

Further, these patterns were consistent with lymphocyte reconstitution patterns observed in the clinical setting, and they correlated with clinical outcomes such as survival and relapse in these patients. This model represents a first step in developing the ability to simulate posttransplant alloreactivity in a computer to more precisely predict the immunosuppression needs of an individual patient before the actual transplant.

In contrast to current clinical practice, dynamical system modeling of immune reconstitution based on whole-exome sequencing data would enable the development of patient-specific immunosuppressive regimens posttransplant by more precisely calibrating the GVHD risk that unique donors might pose to that patient. This modeling would allow us to identify a donor with optimal alloreactivity potential and develop a GVHD prophylaxis regimen of optimal intensity to achieve maximal likelihood of good clinical outcome for each patient.

In so doing, dynamical systems understanding could attenuate the unpredictability to which the current, largely probability-based models of outcomes prediction expose patients. Randomness within the dynamical system will remain a problem for the foreseeable future, but a reduction in its influence would tremendously impact clinical outcomes.

John Wallace, MS, contributed to the editing of this article.

Acknowledgements: The author would like to gratefully acknowledge colleagues and students without whose expertise, collaboration, and dedicated hard work the model development described here would not have been possible. These include Catherine H. Roberts, PhD, Michael C. Neale, PhD, Masoud Manjili, DVM, PhD, Roy T. Sabo, PhD, Gregory A. Buck, PhD, Vishal N. Koparde, PhD, Nihar Sheth, MS, Max Jameson-Lee, PhD, Jared Kobulnicky, MD, Allison Scalora, MS, Juliana Sampson, MS, Jeremy Meier, BS, and Badar Abdul Razzaq, BS.

References

1. Toor AA, Sabo RT, Chung HM, et al. favorable outcomes in patients with high donor-derived T cell count after in vivo T cell-depleted reduced-intensity allogeneic stem cell transplantation. Biol Blood Marrow Transplant. 2012;18(5):794-804.

2. Toor AA, Sabo RT, Roberts CH, et al. Dynamical system modeling of immune reconstitution following allogeneic stem cell transplantation identifies patients at risk for adverse outcomes. Biol Blood Marrow Transplant. 2015;21(7):1237-1245.

3. Berrie JL, Kmieciak M, Sabo RT, et al. Distinct oligoclonal T cells are associated with graft versus host disease after stem-cell transplantation. Transplantation. 2012;93(9):949-957.

4. Meier J, Roberts C, Avent K, et al. Fractal organization of the human T cell repertoire in health and following stem cell transplantation. Biol Blood Marrow Transplant. 2013;19(3):366-377.

5. Sampson JK, Sheth NU, Koparde VN, et al. Whole exome sequencing to estimate alloreactivity potential between donors and recipients in stem cell transplantation. Br J Haematol. 2014;166(4):566-570.

6. Jameson-Lee M, Koparde VN, Griffith P, et al. In silico derivation of HLA-specific alloreactivity potential from whole exome sequencing of stem cell transplant donors and recipients: understanding the quantitative immunobiology of allogeneic transplantation. Front Immunol. 2014;5:529. doi:10.3389/fimmu.2014.00529.

7. Toor AA, Kobulnicky JD, Salman S, et al. Stem cell transplantation as a dynamical system: are clinical outcomes deterministic? Front Immunol. 2014;5:613. doi:10.3389/fimmu.2014.00613.

8. Abdul Razzaq B, Scalora A, Koparde V, et al. Dynamical system modeling to simulate donor T cell response to whole exome sequencing- derived recipient peptides demonstrates different alloreactivity potential in HLA-matched and mismatched donor-recipient pairs. Biol Blood Marrow Transplant. 2016;22(5):850-861