FROM BLOOD GROUPS AND DISEASE BY AE MOURANT.
COPYRIGHT 1977 OXFORD PRESS. (OUT OF PRINT)

This chapter is a discussion of a variety of diseases of either the pregnant woman or the fetus, in which the blood groups and other polymorphisms are involved. The connecting link between them is the immunization, actual or potential, of the mother by antigens in the fetus or abnormal conceptus, antigens which it has inherited from the father. The initial discovery from which all the others follow is that made by Dienst (1905) who showed that, in a woman carrying a fetus with an A or a B blood-group antigen which she herself does not possess, the titer of anti-A or anti-B (already, of course, present before pregnancy) becomes raised by immunization to the antigen in question. Dienst further tried to account, not very convincingly, for eclampsia, as being a result of such immunization. Had he concentrated on the fetuses rather than the mothers he might have come across a few cases of hemolytic: disease of the newborn among them, and so become the discoverer of the aetiology of this disease.

In 1939 Levine and Stetson, examining the serum of a woman who had given birth to a stillborn fetus, found in it a hitherto unknown hemagglutinin which they correctly attributed to immunization by a factor present in the blood of the fetus, and which it had inherited from the father. Following the discovery of the Rhesus blood-group antigen by Landsteiner and Wiener (1940), and of its immunizing effect by Wiener and Peters (1940), Levine, Katzin, and Buinham (1941) showed that hemolytic disease of the newborn was due to red-cell destruction by an anti-Rhesus antibody passively transferred from the mother. They also showed that the antibody found by Levine and Stetson (1939) was in fact anti-Rh (anti-Rhd or anti-D). With the subsequent discovery of the complexity of the Rh blood groups, and the finding of many new independent blood-group systems, it was shown that a great many other antigens could cause maternal immunization, and hence hemolytic disease of the newborn.

The outstanding example of association between blood groups and a disease is in fact that involving hemolytic disease of the newborn, but it has already been the subject of a vast literature of its own, including many thousands of papers and probably over 100 books. Only certain limited aspects of it, therefore, will be considered here.

HEMOLYTIC DISEASE OF THE NEWBORN AND NATURAL SELECTION

Soon after the discovery of the cause of hemolytic disease of the newborn, Wiener (1942) and Haldane (1942) independently showed that each death from hemolytic: disease of the newborn (which, untreated, is a highly fatal disease) causes the destruction of one Rh-positive (D) and one Rh-negative (d) gene. Hence, in the absence of some form of compensation, whichever gene was initially the commoner in a given population should theoretically alone survive after a number of generations. On this view all the populations of Europe, mostly containing about 40 per cent of d genes and 60 per cent of A should be unstable and evolving towards the total elimination of d

One possible explanation for the existing situation seemed to be that the present population of Europe was the result of fairly recent mixing between two main populations, one preponderantly Rh-positive and the other preponderantly Rh negative. There are many peoples in western Asia as well as in northern Africa which are very largely Rh-positive, but there was then no population known with more than 50 per cent of d genes.

Subsequently (Etcheverry, 1945; Mourant, 1947; Chalmers et al, 1948, 1949) it was shown that the Basques of northern Spain and south-western France constitute such a population. Historically, physically, and linguistically they are, indeed, probably the oldest distinct ethnic group at present living in western Europe, and it was suggested by Mourant that they are the relatively unmixed descendants of the late Palaeolithic populations of western Europe who subsequently hybridized with Neolithic newcomers. The population of parts of northern Scotland and the Hebrides, who also are descended from an ancient population group, have likewise a very high d frequency, though not as high as the Basques. The hypothesis that there were only two populations taking part in the mixing, and that the mixed population did not compensate in any way for the mortality from hemolytic disease of the newborn, is probably an over-simplified one, but Ammerman and Cavalli-Sforza (1971) support the hypothesis that the Basques do indeed represent the ancient peoples of Europe whom the Neolithic cultivators from the Near East met as they expanded through Europe.

In considering the effect of deaths from hemolytic disease of the newborn it should be pointed out (Li, 1953) that no simple genetic model, either without or with compensation, can account for a balanced polymorphic state such as may have existed before the present stage of effective prophylaxis and treatment of the disease.

ABO BLOOD GROUPS AND ABORTION

As implied in the previous pages, ABO incompatibility between husband and wife, in the sense that the husband possesses an A or a B antigen which is absent in the wife, is a well-recognized cause of hemolytic disease of the newborn, and, as such, has been discussed in numerous works on that disease. Such incompatibility is also, however, as is shown below, an important cause of early abortion, as well, probably, as of infertility. Some of the apparent infertility may, however, result from unrecognized very early abortion.

This chapter is based mainly on published statistical data of many kinds, but the only relevant data which fit into the main tables of this book are those of the ABO groups of women who have had abortions, and a very few records of the ABO groups of aborted fetuses. The three sets of maternal data differ considerably from one another, and when they are combined the only relative incidence to differ significantly from unity is AB/O, with a value of 1-83. However, even this is due entirely to one of the three sets, the other two both showing values below unity, as would be expected if an excess of aborted fetuses had carried an A or a B antigen not present in the mother. Nearly all the data of other kinds discussed below point to such an excess of A and B in aborted fetuses.

THE BLOOD GROUPS OF PARENTS AND LIVING CHILDREN

Hirszfeld and Zborowski (1925) were the first to observe that, in families where the father was of group A and the mother of group O, there was a deficiency of group A offspring whereas when the father was O and the mother A, the numbers of A and O children were as expected from genetical theory in the absence of selection. Waterhouse and Hogben (1947) observed a similar effect, analyzing a total of 453 families, and finding a 25 per cent deficiency of A children when the mother was group O.

ABO INCOMPATIBILITY BETWEEN MOTHER AND FETUS

We have already mentioned the very early work of Dienst (1905) on the effects of ABO incompatibility between a mother and her fetus. Other early investigations are those of Ottenberg (1923), Hirszfeld and Zborowski (1925), and Hirszfeld (1928). The discovery by Levine, Katzin, and Burnham (1941) of Rh isoimmunization as the principal cause of hemolytic disease of the newborn did not immediately lead to the realization that ABO isoimmunization caused what was fundamentally the same disease (in infants born at or near full term)-mainly because both the symptoms and the serology differ markedly from those characterizing the classic Rh disease. The first workers clearly to describe the ABO variety appear to have been Halbrecht (1944) and Polayes (1945).

However, before this, in 1943, Levine had identified ABO incompatibility as a cause of early abortions and stillbirths. From this time onwards numerous workers produced data suggesting, mainly on the grounds of a deficiency of A children, and an excess of abortions, in the families of O women married to A men, that the A fetuses produced by such matings were especially liable to be aborted. Some of the work was criticized, on grounds of statistical methodology, and in 1%1 Levene and Rosenfield undertook a critical survey of all that had previously been published on the subject, including studies both of full-term hemolytic disease and of early abortion. This paper is recommended to any reader requiring a fuller discussion of early work, especially that on the full-term disease, than can be given here. One aspect of the latter must, however, be mentioned since it (like other more recent work mentioned below) involves the interaction of two genetically independent polymorphisms. Levene and Rosenfield bring together the observations of Zuelzer and Kaplan (1954). Crawford et al. (1953), and Wiener et al. (1960) who all show that in those cases where the secretor status of infants suffering from ABO hemolytic disease has been ascertained, there is a marked deficiency of ABH non-secretors as compared with the general population. This suggests that the secretion of A or B substance by the fetus plays an important part in the immunization of the mother -presumably via the meconium and the amniotic fluid.

Levene and Rosenfield then give a highly critical and detailed analysis and recalculation of all available published data on the ABO groups of parents and offspring. Most of the information on possible loss of children from materno-fetal incompatibility can be derived from the frequencies of A and O children in A-O matings, comparing those matings where the mother is O with those where she is A. The combined data show a significant deficiency of 25 per cent of A children in the incompatible matings. Other matings, involving B, are less conclusive because of small numbers, but the overall conclusion is that 'there is a loss of between 14 per cent and 32 per cent of all A or B children from matings of an A, B (and presumably AB) father and an O mother, as compared with the reciprocal mating, and that the most likely value for this loss is 25 per cent.

Published data on total family size are also analyzed, and show smaller numbers in incompatible than in compatible matings. The mean deficiency (hard to ascertain because of varying criteria used in obtaining the primary data) is about 0.5 children per family. It will obviously be affected by deliberate compensation for fetal loss or a low rate of-conception in matings in which any form of pregnancy limitation is practiced.

Relatively few data were available on numbers of abortions in compatible and incompatible matings, but all series show an excess for the incompatible ones. Rather more abundant are the ABO data on matings which have produced two or more abortions. The data are somewhat heterogeneous but show an excess (42-6 per cent, compared with 35 per cent expected) of ABO incompatibility. It has been convenient to quote extensively (above) from the very thorough study of Levene and Rosenfield, incorporating the results of a very large body of previously published data.

Chung and Morton (1961), also, have surveyed the evidence available to them regarding natural selection at the ABO locus. They adversely criticize some previous investigations on both technical and statistical grounds but find, nevertheless, strong evidence for ante-natal and early post-natal selection related to the ABO groups. They, however, regard the reported associations between blood groups and adult disease, which form the main subject of this book, as having at most, second order selective effects.

Since the publication of these works many further papers on the subject have appeared. nearly all d them reaching similar conclusions. The principal papers consulted are those of Behrman et al. T1960). Wren and Vos (1961). Matsunaga (1962), Kircher (1965), Vos and Tovell (1967). Peritz (1967, 1971), Krieg and Kasper (1968). Cohen and Sayre (1968), Cohen (1970), Takano and Miller (1972), and Hiraizumi el al. (1973).

Cohen (1970) confirms the observation of most other workers, that there is increased early fetal loss in ABO incompatible matings as compared with ABO compatible ones, but in contrast with most others, finds no evidence that such loss affects the ABO frequencies among the full-term offspring. She does find. on the other hand, that in Rh incompatible matings, where there is little or no obvious early fetal loss, there is nevertheless a deficiency of Rh-positive (i.e. Rh-incompatible) offspring.

THE BLOOD GROUPS OF ABORTED FETUSES

The most conclusive evidence that fetuses incompatible with their mothers are more likely to be aborted than compatible ones would be the blood groups of the aborted fetuses themselves, compared with those of their mothers. Unfortunately very few workers have determined the blood groups of such fetuses in this context. Takano and Miller (1972) examined 62 aborted fetuses, but they do not report the actual ABO blood groups, stating only whether they were compatible or incompatible. They found that the number of ABO-incompatible abortuses. was significantly greater than expected, but in addition they noted that among the 43 compatible abortuses there were 8 abnormal fetuses (18-6 per cent) while among the 35 incompatible abortuses only 3 were abnormal. Krieg and Kasper (1968) report the blood groups of 61 aborted fetuses and of their mothers. Unlike Takano and Miller, they examined the blood groups only of fetuses for which no other causes of abortion could be found. Having already demonstrated, on a large sample of families, that the abortion rate was significantly higher in ABO matings in which the father was incompatible with respect to the mother than in those in which he was compatible, they found this not to be the case in this very small series in which the fetus was available, but the rate of incompatibility between mother and fetus was 9 per cent higher than expected by chance.

Allen (1964) has grouped twenty-seven products of abortion, together with the blood of the mother in all cases and of the father in fifteen. The abortus specimens were preserved in formalin, and were tested by suspending in a saline serum mixture followed by titration of the centrifuged supernatant. The author claims that this method gives correct results with secretor fetuses but false negative ones with non-secretors. Despite this bias against the detection of incompatibility, there was a marked excess above expectation, of B incompatible fetuses both from white and Negro mothers, A incompatibilities were as expected. The author draws attention to the conclusion, on indirect evidence, by McNeil el al. (1954), that B may be more important than A in abortions.

Perhaps the most complete study of spontaneously aborted fetuses is that of Lauritsen et al. (1975), who carried out a chromosome analysis of 288 consecutive products of abortion and determined the ABO blood groups of all those, seventy four in number, where this was at all possible, as well as those of the mothers, and in most cases the fathers. They found that 'there were too many group A and B abortuses among those with normal karyotypes, and too few among those with abnormal karyotypes. Furthermore, the frequencies of materno-fetal incompatible pregnancies and of incompatible matings were significantly higher in the group of abortions with karyotypically normal fetuses.

These investigations of rather small numbers of specimens have been described at some length since the results, on the whole consistent between the various investigations, provide the most direct evidence we possess regarding one of the main ways in which parental blood groups affect those of their surviving offspring.

ABO GROUPS AND INFERTILITY

Several workers have sought evidence of ABO incompatibility in couplesowhose infertility could not be otherwise explained. If. in the data of Grubb and Sjostedt (1954-5) one compares reciprocal mating types, e.g. A/O with O/A, then there is a slight. but not significant, excess of compatible matings among couples with involuntary sterility.

Bennett and Walker (1955-6) find childlessness more frequent (but not significantly so) among A and B than among O and AB women, which in no way suggests that incompatibility of the mother (either with the husband's spermatozoa or with very early embryos) is involved. Behrmann et al. (1960), on the other hand, find a significant excess of ABO incompatible couples in 102 persistently sterile matings (87.3 per cent) as compared with 171 fertile couples (38-6 per cent incompatible). They also find, in the case of 7 couples with markedly delayed fertility, that all the 9 children tested were of group O (and hence must have been compatible with the mother). These authors suggest that ABO-related infertility is due to the action of antibodies, in the secretions of the mother's genital tract, on incompatible spermatozoa. It is difficult to explain the marked discrepancies between the results of the different infertility studies, and there is a need for further data.

Kircher (1965) finds an increased proportion of O children in families with more than one child (as would be expected if mothers in ABO incompatible matings became progressively immunized against A or B). He finds that A2 behaves more like 0 than like A, in this respect. On the other hand, Wren and Vos (1961), studying abortions, find that A2 behaves like Al. However, in both cases the numbers of A, individuals studied is small.

The papers cited above show, on the whole, very strong evidence that in matings where the husband has an A antigen which the wife does not possess, there is a marked selection against the birth or survival of A (i.e. heterozygous) offspring. 'Me same is probably true of B. It has long been known that some such selection (mostly in fact in the neonatal period) takes place by means of the relatively rare deaths that result from ABO-related hemolytic disease of the newborn. It now appears that a much larger amount of selection takes place at an earlier stage of pregnancy, being manifested by selective abortion of incompatible fetuses, and by a deficiency of heterozygous children possibly due to very early unnoticed abortions. It is still not clear whether total sterility is often due to this cause. More data are needed on the blood groups of aborted fetuses (and on their sex-see 'The sex ratio' below) and of the partners in totally sterile matings. As pointed out by Grubb and Sjostedt, it is especially desirable to collect data, inevitably a very few at a time, of the groups of partners in situations where a mating has been sterile for at least 5 years and then both partners have had offspring by new partners.

THE SEX RATIO

A further observation, which may reflect selective abortion, is the variation in the sex ratio of the children classified by ABO combinations of mother and child. Allan (1959) has collected and given references for the observations of previous workers as well as those made by himself (Allan, 1958). There is a marked tendency for the sex-ratio (i.e. the ratio of males to females) to be higher in O babies of O mothers than in A babies of A mothers, both in European and in non-European families. Since these mother-child combinations are both ABO-compatible there can be no question of materno-fetal ABO incompatibility being involved. Allan (1972, 1973) has continued to make new observations and to keep his survey of the literature up to date. He has brought to light a highly significant unevenness in the sex-ratio according to the ABO groups of mothers and offspring, which must be taken into amount in any consideration of natural selection affecting these blood groups, but its meaning is at present far from clear. Allan (1973) sums up the matter by saying, 'babies of B mothers ... have a higher male/female ratio if of the same group as their mother than if of a different group (P < 0.05), and this is true also for the babies of O mothers (P < 0.05). In sharp contrast, on the other hand, babies of A mothers have a lower male/female ratio if of the same group as their mother than if of a different group (P < 0.0005), and this is true also of babies of AB mothers (P <0.01).

INTERACTION BETWEEN THE ABO AND Rh SYSTEMS

Levine, in 1943, in the same paper in which he showed that ABO incompatibility was a cause of abortions and stillbirths, pointed out that, in Rh-negative mothers of infants with hemolytic disease of the newborn, the proportion of ABO incompatible husbands is lower than in the general population. He and others soon interpreted this as a protective effect of ABO incompatibility against Rh immunization, since red cells of a fetus having inherited both the Rh(D) factor and A or B from the father would be destroyed rapidly by the mother's anti-A or anti-B, before they had time to cause Rh immunization. This in turn led to the devising, by Clarke and his colleagues in England and by Levine and his team in America, of the present effective means of prophylaxis against hemolytic disease of the newborn.

One of the most complete studies of the interactions of the Rh and ABO systems is that of Cohen (1970). In her first tables she attempts to avoid using data involving possible interactions between the ABO and Rh systems. Taking only Rh-compatible matings she compares fetal death rates in ABO-compatible and incompatible matings. Early fetal death rates are much higher in ABO-incompatible than in ABO-compatible matings, but late fetal losses are almost completely unaffected by ABO incompatibility as compared with ABO compatibility. Then, taking only ABO-compatible matings she shows the well-known effect of Rh incompatibility on late fetal losses but, rather surprisingly, a similar effect on early fetal losses. However, for early fetal losses the doubly incompatible matings show a lower rate than either of the singly incompatible classes, and close to that of the doubly compatible. It therefore appears that, as regards early fetal deaths, each type of incompatibility protects against the harmful effects of the other. For late fetal deaths the numbers are too small to be meaningful. It must also be stated that, despite the large numbers of matings included in the initial survey, the numbers of deaths are so small that only a few of the results mentioned reach statistical significance. In this clinically and scientifically important field of interaction between systems, there is therefore a need for further extensive surveys.

OTHER BLOOD GROUP SYSTEMS

In addition to the A and B antigens. and the D antigen of the Rh system, many other antigens of the Rh and other systems have, as already mentioned, sometimes caused maternal isoimmunization and hemolytic disease of the newborn. The antigens of the P system do not appear ever to be associated with such disease in the full-term newborn infant, but it has been claimed that isoimmunization to the P antigen (in the modern sense, formerly known as P1) is a cause of abortion. Most of the early examples of anti-P were found in pp women who had had an abortion, and it was concluded that the antibody was causing the abortion of heterozygous Pp fetuses. When numerous further examples of the antibody were found without a history of abortion, it was questioned whether this was so, but the latest summing-up by Race and Sanger (1975) concludes that anti-P (or possibly its constant companion anti-P2) is a potent cause of abortion. Vas et at (1964) have found a similar antibody in the serum of P1-positive women who had previously aborted and were threatened with a second abortion. As the phenomenon is not found in similar circumstances in other regions, Vos et at suggest that an environmental factor may be involved.

Though the finding bears no direct relation to abortion, it is convenient to mention here that the auto-antibody which causes paroxysmal cold hemoglobinuria usually has anti-P specificity (Levine et at,).

TOXEMIA OF PREGNANCY

The suggestion of Dienst (1905) that isoimmunization to the A or B antigen was a cause of eclampsia at first appeared to be supported by the finding by Pike and Dickins (1954) in women suffering from toxemias of pregnancy, 6f a significant excess of group O; subsequent surveys (including that of Hanington) have failed to confirm this. As in other cases where there is a discrepancy between the results of different surveys, it is desirable that further tests should be done, with particular attention to subtle points in the selection of patients and controls, differences in which may have given rise to the discrepancy.

Jenkins et at (1976) have tentatively suggested an association between pre-eclampsia and a failure of the patient to produce antibodies to her husband's HLA antigens, a process which is regarded as normal and presumably protective against fetal HLA antigens of paternal origin. This would suggest a possible role of HLA immunization as a protection against choriocarcinoma. Much work has been done on HLA immunization in this disease, work which has been summed up by Lawler (1976). It is clear that this is an important problem upon which more work is needed but as Lawler states, 'The choice of mate in relation to the HLA system appears not to influence a woman's chance of developing a choriocarcinoma".

When I started to write the above section on abortion I was firmly convinced that ABO incompatibility between mother and fetus was a major cause of abortion. Having read most of the relevant evidence and criticisms I still regard the hypothesis as valid, but there are certain unexplained apparent discrepancies between the findings of different workers. However, the most direct way of testing the hypothesis is by observing the blood groups of the spontaneously aborted fetuses of parents of known groups, and observations on these are unfortunately few and not fully conclusive. There are obvious difficulties in collecting such data, and it is therefore important that any workers who may have access to such fetuses, even a few, but preferably many, should carry out a cytological examination and the relevant ABO and Rh tests, as a minimum, and make sure that the results get published. We, however, do not know how far other blood groups (or indeed histocompatibility antigens) may be involved, and very full blood testing should be carried out if this is at all possible.