Three-dimensional structure discrepancy between HLA alleles for effective prediction of aGVHD severity and optimal selection of recipient-donor pairs: a proof-of-concept study.

The optimal selection of recipient-donor pair and accurate prediction of acute graft-versus-host disease (aGVHD) severity are always the two most crucial works in allogeneic hematopoietic stem cell transplantation (allo-HSCT), which currently rests mostly with HLA compatibility, the most polymorphic loci in the human genome, in clinic. Thus, there is an urgent need for a rapid and reliable quantitative system for optimal recipient-donor pairs selection and accurate prediction of aGVHD severity prior to allo-HSCT. For these reasons, we have developed a new selection/prediction system for optimal recipient-donor selection and effective prediction of aGVHD severity based on HLA three-dimensional (3D) structure modeling (HLA-TDSM) discrepancy, and applied this system in a pilot randomized clinical allo-HSCT study. The 37 patient-donor pairs in the study were typed at low- and high-resolution levels for HLA-A/-B/-DRB1/-DQB1 loci. HLA-TDSM system covering the 10000 alleles in HLA class I and II consists of the revised local and coordinate root-mean-square deviation (RMSD) values for each locus. Its accuracy and reliability were confirmed using stably transfected Hmy2.CIR-HLA-B cells, TCR Vβ gene scan, and antigen-specific alloreactive cytotoxic lymphocytes. Based on the preliminary results, we theoretically defined all HLA acceptable versus unacceptable mismatched alleles. More importantly, HLA-TDSM enabled a successful retrospective verification and prospective prediction for aGVHD severity in a pilot randomized clinical allo-HSCT study of 32 recipient-donor transplant pairs. There was a strong direct correlation between single/total revised RMSD and aGVHD severity (92% in retrospective group vs 95% in prospective group). These results seem to be closely related to the 3D structure discrepancy of mismatched HLA-alleles, but not the number or loci of mismatched HLA-alleles. Our data first provide the proof-of-concept that HLA-TDSM is essential for optimal selection of recipient-donor pairs and effective prediction of aGVHD severity before allo-HSCT.

The RMSD parameters were calculated by Visual Molecular Dynamics (VMD, http://www.ks.uiuc.edu/ research/vmd), which was aided by National Institutes of Health(NIH), developed by Beckman institute of University of Illinois at Urbana-Champaign, and used for biophysics computing of protein, nucleic acid, lipid et. al. VMD software can be programmed, scripts were used to compute the RMSD of HLA-alleles high-throughputly and accurately. RasMol is developed by R. Sayle in the University of Edinburgh for the visualization of biomacromolecule, which can deal with the atomic coordinate data. In our study, RasMol was used to visualize the structure differences of HLA -alleles between patient-donor pairs in PC [3][4][5][6][7].
Due to the facts that amino acid residues (AAR) differences at key site between mismatched HLA-alleles at each HLA locus can changed the polarity and hydrogen bonding of the site, thus, some important 3D structures made up of either the altered amino acid itself or the nearby residues, for peptide binding and/or TCR interaction, were significantly changed through alterations to van der Waals forces, short-range interaction and/or long-range interaction [16]. AAR diversity at HLA class I and II molecules are mainly focused on the region of antigens binding groove which are consisted of 2 alpha helix and 8 beta sheet, especially in the regions of the first alpha helix and beta sheet which are extremely important for TCR binding and antigen peptide presenting (Tables 1-4 and Figure 1A-1B in the Supplementary Appendix). As we all know that not every AAR participate in antigen peptide presenting or TCR binding especially AAR in region of random coil, but all these AAR were still involved in the coordinate RMSD calculation. For example,when calculating the coordinate RMSD between HLA-B*07:13 and HLA-B*08:09 allleles, AAR in random coil region such as residues 39-44 ( Figure  2 in the Supplementary Appendix) will be greatly affectting the last calculated results. Thus, The parameters of RMSD calculation were revised according to the structure features of HLA alleles, and the amino acids residues located in α1/ α2 domain at HLA I or α1 domain at HLA II interacting with neither peptide nor TCR were excluded in the calculation of the revised RMSD [17][18][19][20].

Construction, transfection, expression and characterization of pcDNA3.1/HLA-B eukaryotic expression vector
The full-length cDNA coding for HLA-B alpha chain was synthesized from the donor's blood cells (RNA LA PCR kit, TaKaRa Bio, Otsu, Shiga, Japan), ligated with PGEM-T-easy vector and then cloned

Measurement of serum Anti-EBNA antibody and affinity analysis between EBNA3 nonapeptide with HLA molecules
The levels of serum antibodies to Epstein-Barr virus nuclear antigen (EBNA3) from healthy volunteer blood were measured by ELISA (EUROIMMUN, AG, GEM) according to the manufacture's instruction. There are lots of methods for calculation of the affinity between molecules, and the prediction of the Interaction of peptide-MHC based on molecule docking algorithm to get the conformation in the lowest free energy state was popular. To avoid the system deviation, three different algorithm: NetMHC Server, MMBPred, and SYFPEITHI were employed to evaluate the affinity between EBNA3 nonapeptide with HLA molecules as previously described. The three methods were all based on net server and convenient to use peptides and HLA allele were submitted properly, and the feedbacks would be given with internet.

Induction of antigen-specific alloreactive CTLs in vitro
The synthesized EBNA3 nonapeptide (FLRGRAYGL) was used as stimulating antigens and its purity was greater than 95% determined by mass spectrometry and high-performance liquid chromatography. Stock solutions of EBNA3 nonapeptide solubilized in dimethyl sulfoxide (DMSO) were kept at −70°C. The transfected Hmy2.CIR cells with either pcDNA3.1 HLA-B*1502 or -B*1518 or -B*3503 or -B*4403 allele pulsed with EBNA3 nonapeptide, were used as APC cells to induce antigen-specific alloreactive CTLs in vitro respectively. Peptide pulsing was performed for 1 hour at room temperature at a final concentration of 10 μmol/L and washed twice. Mitomycin C (Kyowa Hakko Kogyo Inc., Japan) treating was performed for 30 min at room temperature at a final concentration of 30 μg/ ml and washed twice. The antigen-specific alloreactive CTLs was induced in vitro as previously described. Briefly, human purified T cells were isolated from heparinized peripheral blood of healthy EBV-seropositive individuals as previously described. For the generation of EBV-specific CTLs, the following stimulation procedures were used. Firstly, purified T cells in RPMI 1640 supplemented with 10% heat-inactivated human AB serum were incubated in 12-well plates (Costar) at a final concentration of 2 × 106 cells/well with 2 × 104 stimulator cells prepared already. And then, hrIL-2 (50U/ ml, Quangang Pharmaceutical Co., Ltd, Shandong,China), mouse anti-human CD3 mAb (1 μg/mL, Abcam Inc., UK), hrIL-4 (30 ng/ml, Sigma Inc., USA) and PHA-P (5 μg/ mL, Sigma Inc., USA) were supplemented after 24 hrs. On days 3 and 7 of cultures, each well received 1 × 105 stimulator cells, and cultures were fed every 3-4 d with RPMI 1640 completed medium containing hrIL-2, anti-CD3 mAb ,rhIL-4 and PHA-P. So like this cultured T cells were harvested after 14 days and used for EBV-specific CTLs in vitro experiments.

TCR Vβ gene scan analysis
Total RNA was prepared from the transfected Hmy2.CIR cells and cDNA was synthesized by reverse transcriptase PCR (RT-PCR) with an oligo dT primer (RNA LA PCR kit, TaKaRa Bio, Otsu, Shiga, Japan). Human Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) cDNA (350 bp) was amplified as the control. For TCR Vβ family-specific amplification, PCR was performed using 24 Vβ family-specific primers and two anti-sense primers binding in the TCR Vβ C region primer (5′ fluorescently labeled with Fam and Hex respectively). PCR products of about a 255 bp fragment (range 240-290 bp) were generated on a 2.5% agarose gel. For TCR Vβ gene sequence analysis, 1 μl of PCR product was mixed with 25 μl formamide and 0.3 μl of the internal size standard Genescan-2500 Rox (Applied Biosystems, Warrington, UK). PCR products were denatured for 5 min at 95°C and size separated on a high resolution polyacrylamide gel on the ABI automated gene sequencer (ABIPRISM 310 Genetic Analyzer) and analyzed using the GENESCAN software.
Fluorescent PCR products are represented as electrophoregrams where relative fluorescence intensities are plotted as a function of PCR fragment size. Products from clonal cell populations were visualized as one or two sharp peaks of fluorescence corresponding to the PCR amplified clonal rearranged alleles, while cDNA extracted from normal polyclonal peripheral blood cells was visualized as a fluorescence spectrum composed of polyclonal PCR fragments of different sizes in a normal Gaussian distribution with peaks spaced by 3 bp corresponding to selected in-frame rearrangements. Oligoclonal profiles were visualized as a modification of the Gaussian distribution where three or more distinct peaks appeared above the normal polyclonal background.

Flow cytometric analysis of CD4 + and CD8 + T cells
The induced CTLs were harvested and different subsets were determined by flow cytometric assay after induced 14 days by the transfected Hmy2.CIR cells as previously described [21,22]. Briefly, the induced CTL and uninduced cells were incubated with various monoclonal antibodies including FITC-anti-human CD3, PerCP-anti-human CD8 and PE-anti-human CD4 (Becton Dickinson, USA) for 1 h at RT in the dark. After three washes with PBS, the cells was resuspended in 500 μl PBS/2% paraformaldehyde solution and analyzed by FACSCalibur instrument and CellQuest software.

Proliferation assays of the induced CTLs
To investigate the proliferation inhibitory activity of the induced CTLs, the transfected Hmy2.CIR cells pulsed with EBNA3 nonapeptide on 96-well plates (2 × 10 4 /ml) were co-cultured with the induced CTLs cells (1 × 10 5 /ml) for 3 hrs at 37°C in 5% CO2 air as previously described [21,22]. CellTiter 96 ® Aqueous One Solution Cell Proliferation Assay kit (Promega, Madison, WI, USA) was used to detect the viable cells in the co-culture. After 3 hrs incubation, the O.D. at 570 nm was measured with an ELISA plate reader. The proliferation was determined as follows: (viable cells in the killer and target cell coculture/numbers of viable cells in the control and target cell co-culture) × 100%.

Cytotoxicity of the induced CTLs
To investigate cytotoxicity of the induced CTLs, the transfected Hmy2.CIR cells with pulsed with EBNA3 nonapeptide on 96-well plates (2 × 10 4 /ml) were cocultured with the induced CTLs cells (1 × 10 5 /ml) for 3 hrs at 37°C in 5% CO 2 air as previously described [21,22]. The transfected Hmy2.CIR cells, the induced CTLs, the transfected Hmy2.CIR cells treated with Lysis Solution, 1640 complete medium, and 1640 complete medium treated with Lysis Solution were used as target cell spontaneous LDH release control, effector cell spontaneous LDH release control, target cell maximum LDH release control, culture medium background control, and volume correction control respectively. CytoTox 96 ® Non-Radioactive Cytotoxicity Assay kit (Promega, Madison, WI, USA) was used to detect Lactate Dehydrogenase (LDH) quantitatively that released upon cell lysis. After 3 hrs incubation, 50 μl culture supernatants were collected and added with 50 μl Cytotoxicity Assay Buffer in enzymatic assay plate for 30 mins at RT in the dark. The O.D. at 570 nm was measured with an ELISA plate reader. The proliferation was determined as follows: Cytotoxicity(%) = (Experimental-Effect Spontaneous-Target Spontaneous)/(Target Maximum-Target Spontaneous) × 100%

Study patient-donor pairs
We studied 37 patient-donor pairs from several treatment centers were evaluated, comprising 33 sibling and 4 unrelated recipient-donor transplant pairs. Of these, 25 pairs were prospective (5 patients did not complete the transplant owing to either disease deterioration or economic constraints) and 12 pairs were retrospective. Of the 32 pairs who completed the transplant, 12 had 1 HLA-allele mismatch, 13 had 2 HLA-allele mismatches, 5 had 3 HLA-allele mismatches, and 2 were haploidentical at the HLA allele. Of the 32 transplant recipients, 5 had 2-6 alternative potential donors with comparable HLAallele mismatch to be selected respectively. For 3 of these recipients, the alternates were unrelated donors from a cord blood donor bank, while for the other 2 recipients, the alternates were sibling donors. Both recipient and donor details are summarized in Table 2.
This study of prospective prediction and retrospective verification of aGVHD severity by HLA-TDSM system was based in the Beijing 307 Hospital and the protocol was approved by the Academy of Medical Sciences Review Board. Informed consent was obtained from all patients and donors or their guardians in accordance with the Declaration of Helsinki.

HLA typing at low-resolution and sequencebased typing
The patients and donors were typed at low-and high-resolution levels by PCR-SSP and sequence-based typing methods, respectively, for HLA-A/-B/-DRB1/-DQB1 loci as previously described.

GVHD prophylaxis and evaluation
Patients were prepared for transplantation with the use of standard myeloablative conditioning regimens and all patients received cyclosporine(CsA)+methotrexate (MTX) as aGVHD prophylaxis as previously described [23]. aGVHD was diagnosed and graded according to established clinical criteria [24,25].

Statistical analysis
Data were analyzed using the SPSS11.5 software program (SPSS, USA). Group comparisons were performed using the chi-square test or the Mann-Whitney test for independent samples (non-parametric data). Probability (P) values < 0.05 were considered significant.