Re American Journal of Freie Metin and Patons December 1984 Ana Maris Di Lonardo Cris ian Orrego* Pierre Darlu Mary-Claire King* Max Baur Human genetics an d human rights Identifying the families of kidnapped children ABSTRACT Between 1975 and 1983 in Argentina, at least 145 children were kidnapped with their parents or born in capt vity to imprisoned women and then separated from their mothers. The parents of these children gener- ally remain among the missing persons. However, the grandparents of the kidnapping victims and the Argen- tinian government are concerned with identifying these children. Genetic analysis can ad in determining whether a child who may be among the kidnap victims is related biologically to a particular family. Laboratory analysis of genetic markers in human blood enables te calculation of an “index of grandpaternity.” This index reflects the probability that a child shares genes with a specified set of grandparents because he is their grandchild eompared to the probability he and they share similar genes only by chance. The most useful system for this analysis is HLA, although otter genetic markers including blood groups, red cell enzymes, plasma proteins, and DNA polymor- pphisms ean be tested to provide adequate information in families with only one or two living grandparents. This approach has been applied successfully in Argentina, with an index of prandpaternity for one family of 99.9%, based on HLA typing only, From Laboratorio Inmunolgis. Hospital Carlos G. Durand Bucnos Airs, Argentine (AMDL) Groupe de Recherches de Genetigue Epidemiciogque, INSERM, Chatese de Longehamp, Pars France (PD Bometrsche Abtellung és Diabetes orschungsintuts, Dussiéork West Germany (MB): Laboratory of Molecular Blogs. National Institutes St Health, Bethsds, Marsland (CO); and Schoo of Pubic Heath. Unversty of Calfornia Berkely, California (Mt ra) INTRODUCTION In Argentina, since the early 1970s and particularly following the March 1976 military coup, “disap- pearance” virtually replaced formal arrest and im- prisonment of real or imagined opponents of the government.()) The involvement of the military government in the disappearances subsequent to the coup was documented in 1980 by the Inter-Ameri- can Commission on Human Rights (ICHR) of the Organization of American States. The ICHR af- firmed itself to be“... morally convinced that, in general, these (governmental) authorities could not have been ignorant of the events (disappearances) as they were occurring and did not adopt the nec- essary measures to terminate them.") Following democratic elections in Argentina in 1983, the National Commission on the Disappearance of Persons was established in December of that year. It has received testimony concerning 8,800 persons who disappeared in the preceding 10 years.) The ‘Commission, as well as human rights organizations in Argentina, believe the actual number of “miss- ing” persons to be considerably higher. These abductions, carried out by special intelli- * Member ofa scientific delegation sponsored by the American {Association for the Advancement of Science that traveled to ‘Argentina in June 1984, atthe request of the National Com mission on the Disapperance of Persons of the Republic of Argentina 339Human genetics and human rights gence units of the military and police, also included children and pregnant women.) One hundred forty-five cases of missing children have been doc- umented by the Grandmothers of the Plaza de Mayo.) a human rights group formed in 1977, Sources of information about these children have been primarily witnesses later released from de- tention and torture centers. In addition, the Grandmothers have taken note of potentially forged birth certificates with information that matched disappeared infants. Of the missing children, 120 are known to have been born during the detention of their mothers. Nearly all of these mothers remain among the missing. The rest of the children were abducted with their parents. To date, the Grand- mothers have located 25 children and have evidence that some of them are living with families of the same military personnel responsible for the disap- pearance of their parents. Circumstantial evidence is generally not adequate to establish the biological relationship of such children and their possible rel- atives. Therefore, the Argentinian National Com- mission on the Disappearance of Persons and the Grandmothers of the Plaza de Mayo asked us to investigate how genetic analysis could be used to test whether a child is biologically related toa known set of grandparents. METHODS 1e approach The use of genetic analysis to determine paternity in cases of disputed parentage is well established in civil courts in Europe and North America. !2) An analogous method can be used to establish grand- paternity. The biological basis of the approach is that whereas many features of a child can be al- tered—his clothes, his name, his hair color, even his morphological features—his genes never change, nor do the genes of his true family. Thus the evi- dence for relationship of a child to his true, biological family remains with that child and his biological relatives forever. Similarly, the evidence that can prove that a child is nor the biological child of the couple with whom he is living remains with these individuals forever. If the family with whom the kidnapped child now lives falsely maintains that they are the biological parents, itis possible to disprove such a claim, Furthermore, it is possible to deter- mine the probability that a family looking for their grandchild has in fact located their true grand- child. ‘A hypothetical family of interest is illustrated in Figure 1: other family structures are of course pos- 340 i ir ‘A hypothetical family ofa kidnapped child, The child indicated by the square at the bottom of the pedigree may be the grandson of the four grandparents at the top of the figure, The child's Parents are assumed to be dead. Squares represent males: circles females; slashes deceased persons. Designations I, 5, ete, indicate genotypes at a hypothetical genetic marker with eight alleles. Using genetic analysis, itis possible to determine the probability thatthe childs 1,8 because hes related to this family rather than sharing alleles 1 and 5 with these individuals only by chance, sible as well. In this ease, a child was kidnapped or born in captivity to two parents who have disap- peared and are presumed dead. Suppose all four grandparents are living, but the child has no siblings, uncles, or aunts. A child has been located who is potentially the grandchild in this family. Blood samples are drawn from each of these five people by venipuncture, Twenty to thirty millimeters are drawn into liquid heparin or acid citrate dextrose and delivered to an appropriate tissue typing labo- ratory within 24-48 hours. The genes examined may include those coding for human leukocyte antigens, of HLA," blood groups,'4) red cell enzymes,(15) and plasma proteins. In addition, genetic markers ‘can be examined at the DNA level using restriction enzyme analysis of DNA. polymorphisms.(!6.!7) Table | indicates the most informative genetic markers. Each of these genes is present in various forms in nearly every ethnic group.('*) Each genetic variant is benign, and their physiological functions are irrelevant to testing genetic relationships. Every person has two copies of each gene, one copy in= herited from his mother and one from his father. The parents, in turn, have inherited their genes from the grandparents. Thus for each gene, a grandchild in- herits one copy from his maternal grandparents and one copy from his paternal grandparents. Consider only one gene in the family described Suppose this gene exists in eight forms, or alleles, and the tested individuals have the combinations of alleles, or genotypes, indicated below: Paternal grandfather 12 Paternal grandmother 34 Maternal grandfather 56 Maternal grandmother 78 Hypothetical grandchild 1s The American Journal of Forensic Medicine and PathologyDi Lonardo et al. TABLE 1 Gonetic Markers Particularly Useful for Estimati of Far ies and Kidnapped Children Genetic Marker Abbreviation ‘No. of Alleles or Haplotypes 1. Human leukocyte antigens HLA-A, B, DR ~8,000 2. Blood groups Rhesus BH 8 Duty FY 3 Kidd uk 2 MNSs MNS 4 Kell K 2 3. Red cell enzymes and plasma proteins Phosphoglucomutase Pomt 4 Acid phosphatase ACPI 3 Transferrin TF 5 Vitamin D binding globulin Go 3 Galactose-1-phosphate uridy! transferase GALT 3 Adenylate kinase AKA 3 ‘Transcobalamin I T2 4 Esterase 0 ESD 3 Proteaso inhibitor PL 5 Glutamate-pyruvate transaminase ert 4 Haptoglobin HPA 4 Peptidace A EPA 2 Third component of complement os 3 ‘Adenosine deaminase ‘ADA 3 Immuno globin heavy chain om 8 4. DNA polyniorphisms {AS of 1984, approximately 150 different DNA polymorphisms are known. The number of alleles ps marker varies trom two to dozens. New DNA polymorphisms are detected monthly. At ths ime, each laboratory screening ONA polymorphisms Uses a fraction of the markers available, so the DNA polymorphisms to be teste fora family will depend on the collaborating laboratory. Frequencies for DNA polymorphisms are now being estimated by several groups. At present, DNA polymor- phisms need to be used with caution, because litle information on gene frequencies is available. In the future, ONA polymorphisms will be extremely informative, DNA trom family members or kidnapped children can be preserved and Aditional DNA markers screened in the future as necessary. * Genes with th most alleles are generally the most informative. Markers with codominant alleles are moro informative than markers with dominance. For many families, HLA testing wil be sufficient. For sil more families, screening HLA ard blood groups willbe adequate. Tho ‘other genotic markers can be used for lames with few surviving relatives. Suppose also that the eight alleles of this gene are equally frequent in the population from which this family is drawn. Clearly this child could be the grandchild of these grandparents, because his alleles (1 and 5) are among those of the grandparents. However. in principle, he could have these forms of this gene sir ply by chance, especially if alleles | and 5 are common in the population. We have chosen to work throuzhout our studies with genetic markers for which all alleles can be distinguished from one another: i.e., codominant markers for which gen- types are equivalent to phenotypes. For such ge- netic markers, this information is used to calculate the probabi ity P(gp|c) that these ancestors are the true biological grandparents (gp), given the phe- notype of the child (c). From Bayes theorem we know that Pax Plo Peaple) = 22% Pele) , Poi X (clan) December 1984 where Po; is the prior probability of relationship of this child to the ith set of grandparents in the pop- ulation, and P(c|gp)) is the probability of the child’s phenotype given the ith set of grandparents. If all members of the population are hypothetically equally probable ancestors, then: PoX Plclep) "0X PCe|gp) + (IPP) where P, is the prior probability of relationship of this child to these grandparents, P(c|gp) is the probability of this child’s phenotype if these are his true grandparents, and P(c) is the probability of this child's phenotype if he were drawn at random from the population, rather than this particular family, We call P(gp|c) the “index of grandpaternity.” Of the three probabilities contributing to the index of grandpaternity P(gp|c), P(c|gp) is derived directly from the pedigree data as described below. P(c) is based on the frequencies of the child's alleles in the population, which are published for the larger Plgple) = 341Human genetics and human rights, populations of the world,("?18-' and Po is the prior probability of relationship of this child to these grandparents, based on nongenetic evidence. For kidnapped children, such evidence might include suspicious birth certificates and testimony of wit- nesses. Because Po is not based on genetic analysis, it is inevitably somewhat arbitrary. However, the choice of Po has little influence on the final estimate, be- cause the estimate of P(gp|c) from the first genetic marker is used as the prior probability P, for the estimate of the probability of grandpaternity using the second genetic marker, and so on in a stepwise fashion. The final probability of grandpaternity does not depend on the order in which genetic markers introduced, and is little influenced by the original estimate of Py when many markers are investi- gated.®) By analogy with the index of paternity,(® we set P, = 0.50 for the first genetic marker, so that: —Piclep)__ P(lgp) + Plc) For the family in Figure 1, itis straightforward tocalculate P(gp|c). First, we calcualte P(c|gp), the probability of this child given these grandparents, ‘The probability that the paternal grandparents transmit allele | to the child, if he is in fact their grandchild, is 0.25, since there are four alleles present in this couple and one of the four is inherited by the grandchild. Similarly, the probability that the maternal grandparents transmit allele 5 to the child is 0.25, Thus the probability that this child has phenotype 1,5, if he is a grandchild in this family, is P(c|gp) = (0.25) (0.25) = 0.625. Next, we calculate P(c), the probability that this child has genotype 1,5 without being related to this family at all. We have assumed that alleles 1 through 8 at this genetic marker are equally fre- quent in this population. Therefore fi = fs = 0.125. Furthermore, P(c) = 2fyfs because a child of ge- chosen at random from the population allele 1 from his mother and 5 from his father, or vice versa. Thus, P(c) = 2(0.125) (0.125) = 0.03125. Therefore, for this marker: 0.625 0.625 + 0.03125 ‘That is, based on this one marker only, the proba- bility that this child is the grandchild of these grandparents is 66.7%. ‘Suppose now we test another, independent genetic marker in the same family, with the following re- sults: Peeplc! Pgple 667 Paternal grandfather Wl 342 Paternal grandmother 1 Maternal grandfather 1 Maternal grandmother 1 Hypothetical grandchild 1 ‘Suppose also that for this marker allele 1 is very common in this population, with a frequency of 0.99, and that allele 2 is rare, with a frequency of 0.01. Then P(c|gp) = (1.0) (0.25) = 0.25, P(c) = 2(0.99) (0.01) = 0.0198, and P, = 0.667, so: (0.667) (0.02: (0.667) (0.25) + 1(1-0.667) (0.0198) = 0.962 Pigple) Thus, based on two independent genetic markers, the index of grandpaternity in this hypothetical family is 96.2%. The most informative situations are clearly those in which a child shares a rare allele with a grand- parent. It may be necessary to test many genetic markers to find such an allele, in which case all markers are incorporated into the probability esti- mate. Of course, if a child carries an allele at any marker not present in one of the hypothetical grandparents, that is adequate evidence to exclude the possibility of relationship, assuming tests are properly conducted and no new mutations have oc- curred. Common alleles shared by the child and his hypothetical grandparents will have little influence on P(gp|c). Incorporating linked markers The most informative genetic markers, such as HLA, consist of two or more genes closely linked on the same chromosome, so they are inherited as a unit. Under the assumption that the child is related to the hypothetical grandparents, it is generally possible to construct all potential HLA haplotypes for each relative given the antigens present in each individual. Therefore, P(c|gp) can be calculated as before, considering HLA to be just one marker. Next, to estimate the probability P(c), that this child exists in this population independently of these grandparents, we must consider all possible com- binations of haplotypes. That is, if an individual carries two antigen specificities for HLA-A, two for HLA-B, and two for HLA-DR, eight pairs of hap- lotypes are possible. To estimate P(c) for each possible pair of haplotypes i and j, we calculate 2fifj, and P(c) is the sum of up to eight terms of this form, For each possible haplotype i, f; will be very small, so P(c) will be small despite the several terms that contribute to it. Haplotype frequencies for HLA and for MNSs are published for most of the world’s populations, so these calculations are feasi- ‘The American Journal of Forensic Medicine and Pathology