Predicting Humoral Alloimmunity from Differences in Donor and Recipient HLA Surface Electrostatic Potential

Abstract

In transplantation, development of humoral alloimmunity against donor HLA is a major cause of organ transplant failure, but our ability to assess the immunological risk associated with a potential donor-recipient HLA combination is limited. We hypothesized that the capacity of donor HLA to induce a specific alloantibody response depends on their structural and physicochemical dissimilarity compared with recipient HLA. To test this hypothesis, we first developed a novel computational scoring system that enables quantitative assessment of surface electrostatic potential differences between donor and recipient HLA molecules at the tertiary structure level [three-dimensional electrostatic mismatch score (EMS-3D)]. We then examined humoral alloimmune responses in healthy females subjected to a standardized injection of donor lymphocytes from their male partner. This analysis showed a strong association between the EMS-3D of donor HLA and donor-specific alloantibody development; this relationship was strongest for HLA-DQ alloantigens. In the clinical transplantation setting, the immunogenic potential of HLA-DRB1 and -DQ mismatches expressed on donor kidneys, as assessed by their EMS-3D, was an independent predictor of development of donor-specific alloantibody after graft failure. Collectively, these findings demonstrate the translational potential of our approach to improve immunological risk assessment and to decrease the burden of humoral alloimmunity in organ transplantation.

FIGURE 1.

Schematic overview of the computational approach for quantification of surface electrostatic potential differences between HLA molecules. Bioinformatics approach to enable HLA structure prediction, surface electrostatic potential calculation, and quantification of electrostatic potential differences between two HLA molecules. (A) The atomic resolution structure of a given HLA class I or class II molecule is calculated using homology modeling (MODELLER) based on information derived from high-quality HLA structures resolved by x-ray crystallography. (B) The electrostatic potential in 3D space surrounding an HLA structure is calculated numerically by solving the linearized Poisson–Boltzmann equation for each point on a cubic grid (spacing of 0.33 Å, solvent ionic strength of 0.15 M [pH: 7.4]). (C) Electrostatic potential comparisons consider cubic grid points within a defined region or layer of space (of thickness δ = 3Å) at a distance σ (4 Å) above the van der Waals surface of the HLA molecule. Quantitative comparison of the electrostatic potential between two HLA molecules of interest are performed using the Hodgkin similarity index for grid points within the intersection of their layers (depicted in gray) after the two structures are superimposed, and values are converted into a distance (ESD). (D) Derivation of EMS-3D. A mismatched donor HLA class I molecule is compared electrostatically to each of the recipient HLA class I molecules to derive the respective ESD, and the minimum ESD value (denoted by the dashed red frame) is taken to represent the EMS-3D (interlocus comparison). Similarly, for HLA class II alloantigens, the mismatched donor HLA is compared electrostatically to each of the recipient HLA within the same locus to derive the ESDs, and the minimum value is taken to represent the EMS-3D (intralocus comparison).

FIGURE 2.

Distribution of HLA mismatches in the LIT patient cohort. This figure shows the percentage of patients (lymphocyte donors) with 0, 1, or 2 mismatches within HLA-A, -B, -DRB1, -DQ, and -DP loci.

FIGURE 3.

Donor-specific alloantibody responses after LIT against mismatched HLA expressed on donor lymphocytes. The figure depicts alloantibody binding, as detected using Luminex single HLA beads, against mismatched HLA expressed on donor lymphocytes for the entire cohort. Donor–recipient HLA mismatches (n = 1381) are grouped according to HLA locus (242 HLA-A, 266 HLA-B, 213 HLA-C, 156 HLA-DP, 247 HLA-DQ, and 257 HLA-DRB1), and the MFI of DSA binding detected in recipient sera is shown on the y-axis. The median (SD) MFI of Ab responses against mismatches within individual HLA loci was 5270.3 (4625.6) for HLA-A, 2303.8 (4091.9) for HLA-B, 40.0 (1115.0) for HLA-C, 102.4 (2868.2) for HLA-DP, 2372.4 (5435.0) for HLA-DQ, and 803.1 (4056.2) for HLA-DRB1.

FIGURE 4.

Frequency of HLA class I and class II mismatches in the LIT patient cohort according to their EMS-3D. The figure depicts the frequency of donor–recipient HLA mismatches according to their EMS-3D, grouped by HLA locus. The median (IQR) EMS-3D for individual loci was HLA-A: 0.32 (0.27–0.36); HLA-B: 0.28 (0.22–0.33); HLA-C: 0.32 (0.22–0.40); HLA-DRB1: 0.20 (0.17–0.24); HLA-DQ: 0.35 (0.20–0.42); and HLA-DP: 0.19 (0.18–0.24).

FIGURE 5.

Probability of donor-specific alloantibody response after LIT according to the EMS-3D of mismatched HLA on donor lymphocytes. The relationship between the immunogenic potential of donor HLA, as determined by EMS-3D, and the probability of a donor-specific alloantibody response after LIT was examined using logistic regression modeling. Each panel shows a logistic regression model with 95% CI (dotted lines) for individual HLA loci (A–E) and for HLA class I and class II loci combined (F and G). DSA responses against HLA-C mismatches were infrequent and were not examined. Donor-specific alloantibody responses were defined using MFI cut-off thresholds ≥2000 (A–F) and ≥8000 (G). Wide CIs for alloantibody responses against HLA-DR and -DP alloantigens reflect the relatively low number of observations for HLA-DR and -DP mismatches with high EMS-3D scores in the LIT patient cohort. Relatively few alloantibody responses with MFI ≥8000 were noted against HLA-DP alloantigens (n = 7), and, therefore, HLA-DP mismatches were not included in the (G) model.

FIGURE 6.

Relationship between donor HLA EMS-3D and the donor-specific MFI binding level of alloantibodies developed after LIT. The relationship between EMS-3D of mismatched HLA on donor lymphocytes and the magnitude of DSA binding, as assessed based on MFI, detected in the Luminex single Ag bead assay, is depicted. Donor HLA mismatches are grouped according to EMS-3D and the box plots depict the median MFI (horizontal blue line) and IQR (box) of MFI values (the lines show maximum MFI values) for DSA binding within each group. Median regression analysis showed that donor HLA with increasing EMS-3D were associated with progressively stronger (higher MFI) alloantibody responses following LIT (p < 0.001). Alloantibody responses against donor HLA-C and -DP mismatches were infrequent and of low MFI value and are, therefore, not included in this analysis.