HLA are cell surface proteins that present peptides to T lymphocytes as part of the immune recognition process, which is fundamental to the adaptive immune response. HLA are encoded by genes within the class I and class II regions of the major histocompatibility complex (MHC) located on the short arm of chromosome 6.1 These genes are characterized by extreme levels of variation, with almost 5000 alleles described for HLA-B and thousands of alleles described for the other HLA genes.
Sensitization, the presence of anti-HLA antibodies, is present in approximately 25% of candidates on the kidney and heart transplant waiting lists.2,3 These antibodies develop after exposure to nonself HLA through pregnancy, blood transfusion, or a previous organ transplant.4 After transplant, anti-HLA antibodies directed against the allograft can cause rejection within minutes to hours of reperfusion, termed hyperacute rejection,5 as well as early antibody-mediated rejection that can occur within days to weeks after transplant. These forms of rejection can lead to allograft dysfunction and failure,6-9 and consequently methods for predicting anti-HLA antibody cytotoxicity have been evolving10 since Patel and Terasaki11 first described the crossmatch almost 50 years ago.
Application of the principles of HLA population genetics to transplantation has led to the development of a numeric metric that predicts the likelihood of a positive crossmatch, the calculated panel reactive antibody (CPRA). CPRA measures the fraction of incompatible donors present in the population, given a set of antigens that the candidate has antibodies against.
In this review, we first discuss the fundamental principles of human population genetics. We then apply these principles to HLA and demonstrate how the CPRA is derived mathematically using these principles. Next, the data supporting the use of CPRA in solid organ transplantation are presented. Finally, we describe emerging refinements to the way that HLA and antibody specificities may be defined and used in the future.
FUNDAMENTAL PRINCIPLES OF POPULATION GENETICS
The DNA contained within each human cell is separated into 23 different chromosomes. The genome of humans is diploid; there are 2 copies of each nonsex (or autosomal) chromosome. However, gametes (oocyte or spermatazoon) are haploid and contain only 1 copy of each autosomal chromosome, plus the sex chromosome (X for the oocyte, X or Y for the spermatazoon). At fertilization, each embryo receives half of its genes via the oocyte and half via the spermatazoon.
Fundamentally, variant gene sequences, better known as alleles, exist within the population due to evolutionary processes such as mutation and recombination, which result in the generation of new DNA sequences. These new sequences then diffuse through the population via the offspring’s progeny, with varying success depending on the relative fitness conferred by the variant. Polymorphic genes are those that have more than 1 allele throughout the population as a whole.
One can derive mathematical relationships between haploid allele frequencies and diploid genotype frequencies within populations. For the simplest case of a polymorphic gene with only 2 alleles with allele frequencies p and q, the sum of the allele frequencies in the population will equal 100% (Eq 1). Given a diploid genome, 2 alleles will give rise to 3 possible genotypes: pp, pq, and qq. The frequencies of these genotypes given random mating were described by Hardy and Weinberg in 1908 (Eq 2).12
Under circumstances where there is random mating, an infinitely large population, no mutation, no migration and no selection, these allele and genotype frequencies will remain constant between successive generations.13 This principle is known as Hardy-Weinberg equilibrium (HWE). Because the above assumptions are not always met for human population samples, the allele frequencies for some genes deviate from the Hardy-Weinberg proportions. Deviation from the Hardy-Weinberg proportions may indicate population substructure or the action of evolutionary forces like selection or mutation.
If the alleles for 2 different genetic loci are inherited together more frequently than would be expected by random recombination, their genes are said to be in linkage disequilibrium.14 The physical explanation for linkage disequilibrium is that the 2 polymorphic loci are more likely to be inherited as a block if they are near to one another on the same chromosome, because the 2 alleles are less likely to be broken up by recombination. Sets of alleles that co-occur in different loci on the same chromosome are known as haplotypes, a term originally coined for the HLA genes.15
Because HLA genes are located near one another in the MHC, some HLA loci are in high linkage disequilibrium. As such, the pattern of HLA alleles that each person possesses, the HLA genotype, are best described in terms of their 2 HLA haplotypes, with 1 haplotype inherited from each parent. Because the MHC encompasses 3.1 million bases,1 the distance between HLA genes is too large to allow for experimental determination of haplotypes (also known as phasing) in individuals. Pedigree analysis of family genotypes can help in haplotype determination. In samples of unrelated individuals, haplotype frequencies can be estimated using mathematical models based on HWE, such as the expectation-maximization algorithm.16
STRUCTURE AND GENETICS OF HLA
The MHC was discovered in humans by Jean Dausset in 1958.17,18 The MHC spans several gene clusters,1 including most notably the class I and class II regions which contain the genes for the HLA–A, -B, -C and –DR, -DQ, and -DP proteins, respectively (Figure 1). The HLA class I antigens (HLA–A, -B, -C) are composed of a polymorphic α chain encoded by their respective gene (HLA-A, -B, -C), and an invariant β chain (encoded by the gene B2M). The HLA class II antigens are composed of a polymorphic α chain (except for HLA-DR which has minimal polymorphism of its α chain) and a β chain, both encoded by separate genes. The HLA-DRA, -DQA1, and -DPA1 genes encode the α chain and the HLA-DRB1, -DQB1, and -DPB1 genes encode the β chain.
In addition to the class II antigens, HLA-DR has 3 additional serological specificities: HLA-DR51, -DR52, and -DR53. These antigens are composed of the common α chain encoded by the HLA-DRA gene and a β chain encoded by 1 of 3 different genes: HLA-DRB5, HLA-DRB3, and HLA–DRB4, respectively. The HLA-DRB3/4/5 genes are in nearly 100% linkage disequilibrium with certain HLA-DRB1 allele families and each haplotype has at most 1 of the 3 genes, and some haplotypes (specifically DR1, DR8, and DR10) have none.
HLA POLYMORPHISM AND POPULATION GENETICS
HLA are the most polymorphic genes in the human genome.19 The degree of polymorphism is not uniform among HLA genes, with the class I genes being substantially more polymorphic than the class II genes20 (Table 1). About 40% of alleles are characterized by synonymous nucleotide substitutions that do not affect the amino acid sequence of the protein, resulting in more distinct DNA nucleotide sequences (alleles) than amino acid sequences (proteins) for each gene. The other 60% of alleles encode nonsynonymous substitutions that lead to distinct HLA proteins.
In HLA-typed population samples, significant deviations from HWE are often observed.21,22 One of the explanations for this deviation is the action of balancing selection,22,23 in which persons who are heterozygotes for 1 or more HLA genes have increased fitness. This increased fitness leads to propagation of these alleles throughout the population over many generations. Heterozygotes are thought to be favored because of an enhanced ability, as compared with a homozygote, to present a diverse set of peptides to the immune system. Specifically, this model is known as pathogen-driven balancing selection—a significant positive correlation between the diversity of the HLA-A and -B alleles in a subpopulation and pathogen richness within that subpopulation has been shown.24 Thus, the extreme polymorphism of HLA genes within the population is likely due to the advantage carried by immune systems that can recognize a greater diversity of pathogens. Human mate choice is not random and is constrained by many factors including geography. As a result, persons of similar ethnogeographic origin share HLA haplotypes, and the frequency of HLA haplotypes in the population varies according to ethnic group.25
BACKGROUND AND MATHEMATICAL BASIS OF THE CPRA
Multiple technologies have been developed to identify which of the anti-HLA antibodies that a sensitized recipient possesses will be cytotoxic and cause allograft injury. The first of these technologies was the crossmatch, in which a cytotoxic reaction between recipient serum and donor lymphocytes was found to strongly correlate with graft failure.11 This technique was extended to transplant candidates on the waiting list to identify those sensitized patients with anti-HLA antibodies.26 The level of sensitization was quantified by determining the panel-reactive antibody (PRA): the number of cytotoxic reactions using recipient serum and a panel of third-party lymphocytes that was representative of different HLA phenotypes in the population divided by the total number of reactions.27 Patients with high levels of sensitization, as indicated by a higher PRA, were more likely to have a positive crossmatch at the time of transplantation.
The development of solid-phase methods of antibody identification led to the ability to determine the individual specificities of potentially cytotoxic anti-HLA antibodies present within the blood of a sensitized candidate.10 HLA antigens that are bound by these antibodies are then excluded from the potential donor pool. At the time of waiting list addition for transplantation in the United States, these antigens must be specified (in the case of renal transplant) or can be specified (in the case of cardiac transplant) as unacceptable, and all donors bearing these unacceptable HLA antigens (UA-HLA) will be automatically excluded. This process is known as the “virtual” crossmatch.28,29
As compared with measuring sensitization with panels of live cells (vis-a-vis the PRA), a different approach is needed to determine the level of sensitization using solid-phase methods. In 1985 Zachary and Braun30 described a mathematical method for calculating “a predictive value for transplantation,” which subsequently has become known as the CPRA. Here we will describe this method in detail.
We start with a transplant candidate with between 1 and n unacceptable antigens, encoded by alleles An and with allele frequencies pn. This candidate can receive an allograft with compatible antigens, encoded by alleles Ac and with allele frequencies pc. The frequency for these compatible alleles, pc, is equal to the frequency of alleles remaining from the total pool when the frequency of unacceptable alleles are removed (Eq 3). Applying the Hardy-Weinberg equation (Eq 2), the probability of finding genotypically compatible donors in the population and thus a negative crossmatch, PNegative XM, is equal to the frequency of these compatible alleles squared (Eq 4).
This equation can then be extended to the case where the unacceptable antigens are controlled by 2 genes i and j, with allele frequencies pi and pj. As above, the frequency of the compatible alleles, pc, is equal to the frequency of alleles remaining from the total pool when the frequency of unacceptable alleles for both genes are removed (Eq 5), and P is equal to this frequency squared, given the diploid genome (Eq 6). When we expand the binomial in Equation 6, we find that again P is equal to the square of the frequency of alleles remaining from the total pool when the frequency of unacceptable alleles for i and j are removed, but we have to add back the frequency of persons who inherit i and j together (Eq 7). Because of the high linkage disequilibrium between the HLA genes, this later quantity, pipj, is not negligible, and thus the impact of additional antibodies across loci is not linearly related to the allele frequencies, because certain alleles co-occur across loci as part of the same haplotype. The allele frequencies for pi, pj and pipj are taken from the haplotype frequencies for the HLA loci.
Equation (7) becomes more complex because there are 5 major HLA genes, HLA–A, -B, -C, -DRB1, and -DQB1, so the two-locus equation must be extended to the five-locus case (Eq 8). The expansion of this polynomial equation, which is detailed in the SDC, https://links.lww.com/TP/B454, includes the haplotype frequencies for all one (eg, A, B, C), two (eg, A-B, A-C, A-C), three (eg, A-B-C), four (eg, A-B-C-DR), and five (eg, A-B-C-DR-DQ) gene combinations (SDC Eq 9, https://links.lww.com/TP/B454). Given that the allele frequencies for each of these gene combinations have the same sign, they can be grouped and summarized (S1 for the 1 gene haplotype frequencies, S2 for the 2 gene haplotype frequencies, etc). This equation gives the probability of finding genotypically compatible donors in the population (ie, a negative crossmatch). However, by convention, we are interested in the opposite, the probability of a positive crossmatch, so the final equation for CPRA is given as Equation (10).
Because HLA allele frequencies vary strongly by ethnic group, PPositive XM must be calculated separately for each major ethnic group using their respective HLA haplotype frequencies. The ethnic groups used by the United Network for Organ Sharing (UNOS) are European American, African American, Hispanic/Latino, and Asian/Pacific Islander. The PPositive XM for each ethnic group is then multiplied by the percentage of the US donor population of that ethnic group, f, and these values are summed (Eq 11). The CPRA is equal to this final sum. We have included an example CPRA calculation in the SDC, https://links.lww.com/TP/B454.
UTILITY OF CPRA IN TRANSPLANTATION
The CPRA is a highly useful metric of the degree of a candidate’s sensitization because it summarizes the percentage of the potential donor population with the candidate’s unacceptable HLA antigens for both the class I and class II specificities as a single numeric value. The CPRA, which is a function of the UA-HLA entered by the transplant center for a candidate, replaced the PRA as the official metric of sensitization for the US Kidney Allocation System in October 200931 and Eurotransplant in 2016.32 The US Kidney Allocation System was updated in 2014 so that sensitized candidates are awarded allocation points on a sliding scale, beginning at CPRA of 20%, and with a high number of points awarded to candidates with CPRA of 98% or greater.33 The effect of this policy change has been a markedly increased rate of transplantation for highly sensitized candidates.34
We have recently shown that the CPRA also predicts outcomes on the heart transplant waiting list.35 Using a data set of 3855 candidates entered on the heart transplant waiting list with UA-HLA between 2006 and 2013, we found that as the CPRA increased, the percentage of candidates that received a transplant decreased, whereas the percentage that were still waiting for a transplant increased, as did the percentage that were removed from the waiting list, or died (Figure 2). The group of candidates with CPRA over 80% displayed a markedly decreased incidence of transplantation (hazard ratio, 0.37) and an increased risk of removal from the waiting list or death (hazard ratio, 2.18) as compared with CPRA of 10% or less. Currently, the US Adult Heart Allocation System does not account for candidate sensitization, although the Canadian Allocation System does have a specialized urgency status for sensitized patients.36
There are several limitations of the CPRA that require further discussion. First, the CPRA, as currently implemented by UNOS does not include antigens for the loci HLA–DQA1, -DPA1, or -DPB1. Tinckam et al37 found that using the Canadian CPRA calculator, which includes these antigens, CPRA increased by 5% or more in 38% of candidates with these UA-HLA. Revision of the UNOS CPRA to include these antigens may improve prediction of adverse immunologic events after transplantation. However, extended haplotype data containing these HLA loci are not currently available for the US population.
For candidates that are very highly sensitized (CPRA ≥ 98%), the CPRA is not sufficiently granular to describe the level of sensitization. Recently, we have proposed the use of the CPRA with decimals (CPRAd) to improve recognition of highly sensitized candidates.38 Although CPRA is an integer, CPRAd is as a probability value expressed from 0 to 1, with as many decimal places as needed. CPRAd also facilitates calculation of the opposite of CPRA, the probability of a negative crossmatch, which we have termed the likelihood of a compatible donor (LCD). LCD is a useful metric to describe access to transplantation for transplant candidates with CPRA of 100%. For candidates on the UNOS kidney transplant waiting list with CPRA of 100%, the LCD spans a wide breadth, from candidates with LCD of 1 in 200 to LCD of 1 in 50 000.38
FUTURE APPLICATIONS OF POPULATION GENETICS IN SOLID ORGAN TRANSPLANTATION
The development and widespread utilization of single-antigen solid-phase antibody testing methods has the potential to change the approach to the listing of UA-HLA and consequently calculation of CPRA in solid organ transplantation. Single-antigen solid-phase assays are coated with specific HLA proteins, for example allele HLA-B*35:01 (in HLA nomenclature the colon separates the first field, which designates the serological antigen or allele family from the second field, which designates the distinct protein). For a transplant candidate with an antibody that recognizes this allele, the approach using the virtual crossmatch has been to exclude all donors with the corresponding serological antigen (eg, HLA-B35). An alternative approach is to only exclude donors with the HLA-B*35:01 allele and all other HLA alleles containing equivalent amino acid epitopes.39 Because serological antigens encompass many unique proteins (eg, HLA-B35 encompasses hundreds of alleles with differing amino acid sequences), excluding specific alleles rather than antigens would have the effect of increasing the size of the potential donor pool (Table 2).
For transplant candidates with antibodies against HLA–DQ, the ability to exclude individual alleles within a serologic antigen group may be particularly important. Tambur et al found that one-third of these candidates can have an 11 to 40% reduction in CPRA by excluding specific HLA alleles instead of the serologic HLA-DQ specificity.40
The strategy of excluding HLA alleles has been implemented in the Eurotransplant Acceptable Mismatch Program for kidney transplantation41 as well as in pediatric kidney transplantation in Australia. Using such a strategy, Kausman et al42 reported no episodes of rejection and a rate of donor-specific antibody development similar to controls, however, longer term follow up is needed.
The next step, widespread implementation of epitope-based assignment of UA-HLA will require the development of new data management tools to quickly handle the complexities of HLA epitope analysis. It will also require refinement of the CPRA to manage HLA specificities at the unique protein level rather than the antigen level. Although an allele-specific CPRA is not yet available, given that high-resolution HLA haplotypes have been determined for multiple ethnic groups,25 such a metric could be developed in the near future. Epitope-based assignment of UA-HLA has the potential to increase the donor pool for sensitized candidates.
The HLA genes are the most polymorphic genes in the human genome, with thousands of different alleles described for the HLA loci. For patients that develop anti-HLA antibodies before transplantation, the CPRA is the preferred metric to determine the level of sensitization, describing the probability of a positive crossmatch as a function of UA-HLA, which are determined from solid-phase antibody testing for each transplant candidate. Patients with high CPRA values are given priority in the US Kidney Allocation System and Eurotransplant, but this is not accounted for in the current Adult Heart Allocation System. Improvements to the CPRA, such as inclusion of the loci HLA-DQA1, -DPA1, and -DPB1 and refinements, such as an allele-specific CPRA, will increase the utility of this metric for measuring access to a compatible solid organ transplant.
1. Horton R, Wilming L, Rand V, et al. Gene map of the extended human MHC. Nat Rev Genet
2. Sapir-Pichhadze R, Tinckam KJ, Laupacis A, et al. Immune sensitization and mortality in wait-listed kidney transplant candidates. J Am Soc Nephrol
3. O'Connor MJ, Keeshan BC, Lin KY, et al. Changes in the methodology of pre-heart transplant human leukocyte antibody assessment: an analysis of the United Network for Organ Sharing database. Clin Transplant
4. Redfield RR, Scalea JR, Zens TJ, et al. The mode of sensitization and its influence on allograft outcomes in highly sensitized kidney transplant recipients. Nephrol Dial Transplant
5. Kissmeyer-Nielsen F, Olsen S, Petersen VP, et al. Hyperacute rejection of kidney allografts, associated with pre-existing humoral antibodies against donor cells. Lancet
6. Williams GM, Hume DM, Hudson RP, et al. “Hyperacute” renal-homograft rejection in man. N Engl J Med
7. Weil R 3rd, Clarke DR, Iwaki Y, et al. Hyperacute rejection of a transplanted human heart. Transplantation
8. Orandi BJ, Chow EH, Hsu A, et al. Quantifying renal allograft loss following early antibody-mediated rejection. Am J Transplant
9. Clerkin KJ, Farr MA, Restaino SW, et al. Donor-specific anti-HLA antibodies with antibody-mediated rejection and long-term outcomes following heart transplantation. J Heart Lung Transplant
10. Tambur AR, Lavee J. Incorporating human leukocyte antibody results into clinical practice. J Heart Lung Transplant
11. Patel R, Terasaki PI. Significance of the positive crossmatch test in kidney transplantation. N Engl J Med
12. Mayo O. A century of Hardy-Weinberg equilibrium. Twin Res Hum Genet
13. Gourraud PA, Barnetche T, Vidan-Jeras B, et al. Introduction to statistical analysis of population data in immunogenetics. Transpl Immunol
14. Kauppi L, Jeffreys AJ, Keeney S. Where the crossovers are: recombination distributions in mammals. Nat Rev Genet
15. Ceppelini R, Curtoni ES, Mattiuz PL, et al. Genetics of leukocyte antigens: a family study of segregation and linkage. In: Curtoni ES, Mattiuz PL, Tosi RM, eds. Histocompatibility testing. 1967. 1967; 149.
16. Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol
17. Dausset J. Iso-leuko-antibodies. Acta Haematol
18. Davies DA, Manstone AJ, Viza DC, et al. Human transplantation antigens: the HL-A (Hu-1) system and its homology with the mouse H-2 system. Transplantation
19. 1000 Genomes Project Consortium, Abecasis GR, Altshuler D. A map of human genome variation from population-scale sequencing. Nature
20. Robinson J, Soormally AR, Hayhurst JD, et al. The IPD-IMGT/HLA database—new developments in reporting HLA variation. Hum Immunol
21. Chen JJ, Hollenbach JA, Trachtenberg EA, et al. Hardy-Weinberg testing for HLA class II (DRB1, DQA1, DQB1, and DPB1) loci in 26 human ethnic groups. Tissue Antigens
22. Solberg OD, Mack SJ, Lancaster AK, et al. Balancing selection and heterogeneity across the classical human leukocyte antigen loci: a meta-analytic review of 497 population studies. Hum Immunol
23. Bronson PG, Mack SJ, Erlich HA, et al. A sequence-based approach demonstrates that balancing selection in classical human leukocyte antigen (HLA) loci is asymmetric. Hum Mol Genet
24. Sanchez-Mazas A, Lemaître JF, Currat M. Distinct evolutionary strategies of human leucocyte antigen loci in pathogen-rich environments. Philos Trans R Soc Lond B Biol Sci
25. Gragert L, Madbouly A, Freeman J, et al. Six-locus high resolution HLA haplotype frequencies derived from mixed-resolution DNA typing for the entire US donor registry. Hum Immunol
26. Oldfather JW, Anderson CB, Phelan DL, et al. Prediction of crossmatch outcome in highly sensitized dialysis patients based on the identification of serum HLA antibodies. Transplantation
27. Ting A. Problems of the strongly sensitized patient. Transplant Proc
28. Zangwill SD, Ellis TM, Zlotocha J, et al. The virtual crossmatch—a screening tool for sensitized pediatric heart transplant recipients. Pediatr Transplant
29. Zangwill S, Ellis T, Stendahl G, et al. Practical application of the virtual crossmatch. Pediatr Transplant
30. Zachary AA, Braun WE. Calculation of a predictive value for transplantation. Transplantation
31. Cecka JM, Kucheryavaya AY, Reinsmoen NL, et al. Calculated PRA: initial results show benefits for sensitized patients and a reduction in positive crossmatches. Am J Transplant
32. Eurotransplant Foundation. Chapter 10: Histocompatibility Testing. https://www.eurotransplant.org/cms/mediaobject.php?file=H10+Histoocompatibility+November+2016+v4.11.pdf
. Published August 4 2016. Accessed May 16 2017.
33. United Network for Organ Sharing. OPTN Policies - Chapter 8. https://optn.transplant.hrsa.gov/media/1200/optn_policies.pdf#nameddest=Policy_08
. Published April 6 2017. Accessed May 16 2017.
34. Stewart DE, Kucheryavaya AY, Klassen DK, et al. Changes in deceased donor kidney transplantation one year after KAS implementation. Am J Transplant
35. Kransdorf EP, Kittleson MM, Patel JK, et al. Calculated panel-reactive antibody predicts outcomes on the heart transplant waiting list. J Heart Lung Transplant
36. Clarke B, Ducharme A, Giannetti N, et al. Multicenter evaluation of a national organ sharing policy for highly sensitized patients listed for heart transplantation in Canada. J Heart Lung Transplant
37. Tinckam KJ, Liwski R, Pochinco D, et al. cPRA increases with DQA, DPA, and DPB unacceptable antigens in the Canadian cPRA calculator. Am J Transplant
38. Kransdorf EP, Pando MJ. Calculated panel reactive antibody with decimals: A refined metric of access to transplantation for highly sensitized candidates. Hum Immunol
39. Duquesnoy RJ, Kamoun M, Baxter-Lowe LA, et al. Should HLA mismatch acceptability for sensitized transplant candidates be determined at the high-resolution rather than the antigen level? Am J Transplant
40. Tambur AR, Leventhal JR, Walsh RC, et al. HLA-DQ barrier: effects on cPRA calculations. Transplantation
41. Claas FH, Rahmel A, Doxiadis II. Enhanced kidney allocation to highly sensitized patients by the acceptable mismatch program. Transplantation
42. Kausman JY, Walker AM, Cantwell LS, et al. Application of an epitope-based allocation system in pediatric kidney transplantation. Pediatr Transplant