A KIR B centromeric region present in Africans but not Europeans protects pregnant women from pre-eclampsia

Significance

Pre-eclampsia is especially common in women of African ancestry and a major cause of maternal death. The killer-cell immunoglobulin-like receptor (KIR) genes that we analyzed are expressed by natural killer cells—immune cells that populate the uterus and are essential for successful pregnancy. KIR proteins bind HLA ligands on the implanting placental trophoblast cells. African and European women share similar risk associations for pre-eclampsia, but protection is associated with different KIR genes. African women are protected by a combination of KIR B haplotype genes that is present almost exclusively in Africans. This study emphasizes the importance of studying diseases in Africans, where the KIR/HLA genetic system is at its most diverse and maternal mortality rates are the highest in the world.

Abstract

In sub-Saharan Africans, maternal mortality is unacceptably high, with >400 deaths per 100,000 births compared with <10 deaths per 100,000 births in Europeans. One-third of the deaths are caused by pre-eclampsia, a syndrome arising from defective placentation. Controlling placentation are maternal natural killer (NK) cells that use killer-cell immunoglobulin-like receptor (KIR) to recognize the fetal HLA-C molecules on invading trophoblast. We analyzed genetic polymorphisms of maternal KIR and fetal HLA-C in 484 normal and 254 pre-eclamptic pregnancies at Mulago Hospital, Kampala, Uganda. The combination of maternal KIR AA genotypes and fetal HLA-C alleles encoding the C2 epitope associates with pre-eclampsia [P = 0.0318, odds ratio (OR) = 1.49]. The KIR genes associated with protection are located in centromeric KIR B regions that are unique to sub-Saharan African populations and contain the KIR2DS5 and KIR2DL1 genes (P = 0.0095, OR = 0.59). By contrast, telomeric KIR B genes protect Europeans against pre-eclampsia. Thus, different KIR B regions protect sub-Saharan Africans and Europeans from pre-eclampsia, whereas in both populations, the KIR AA genotype is a risk factor for the syndrome. These results emphasize the importance of undertaking genetic studies of pregnancy disorders in African populations with the potential to provide biological insights not available from studies restricted to European populations.

Although pre-eclampsia presents clinically with a diverse array of systemic symptoms, the underlying disease-causing mechanism starts with placentation when trophoblast cells invade the decidua. Here, they transform the uterine spiral arteries into large vessels that form the fetoplacental supply line (1, 2). In pre-eclampsia and other pregnancy disorders (fetal growth restriction, stillbirth, and recurrent miscarriage) known collectively as the great obstetric syndromes (GOSs), trophoblast fails to invade optimally (3). Pre-eclampsia and other GOSs occur in all populations, but women of African ancestry are significantly more at risk; thus, GOSs are responsible for much of the high maternal and fetal mortality rates seen in sub-Saharan Africa (SSA) (4). The genetic contribution to pre-eclampsia is supported by several studies and involves both maternal genes and paternal genes inherited by the fetus (5, 6).

The wall of the uterus is the territorial boundary between two genetically different individuals: the mother and the fetus. The uterine mucosal immune system appears to define this maternal/placental boundary. The decidua must control placentation, because in its absence, the trophoblast infiltrates to a dangerous extent, causing the condition of placenta percreta (7). The decidua contains an abundant population of specialized natural killer (NK) cells. These uterine NK cells (uNK) express killer-cell immunoglobulin-like receptors (KIRs) that recognize trophoblast HLA-C ligands (8, 9). Both KIR and HLA-C are genetically variable, resulting in many possible combinations of maternal KIR and fetal HLA-C ligands (10). The KIR region is defined by two groups of haplotype: A and B. The KIR A haplotype has seven KIR genes, all encoding inhibitory receptors apart from KIR2DS4. In contrast, the KIR B haplotype contains a variable number of additional KIR, most of which encode activating receptors (11, 12). All HLA-C allotypes are KIR ligands and can be divided into two groups (carrying either C1 or C2 epitopes) that are distinguished by a dimorphism at position 80 and recognized by different KIR (13). Within a human population, the combination of KIR and HLA diversity distinguishes individuals and this extremely high variation is particularly evident in SSA populations. They exhibit less linkage disequilibrium (LD) between the KIR genes than other populations (14–16), and the KIR genes have greater allelic diversity (15, 16). A variety of diseases and clinical conditions has been associated with combinations of HLA-C and KIR genes. In previous case–control studies of pre-eclampsia in pregnant European women, we showed that, when the fetus carries a C2 epitope, maternal KIR AA genotypes are risk factors for pre-eclampsia, whereas the KIR2DS1 gene of maternal KIR B haplotypes is protective (8, 17). In the case–control study reported here, we test the hypothesis that these factors confer similar risk and protection to pregnant SSA women.

Results

Clinical Characteristics of the Cohort.

This case–control study of pre-eclampsia involved 738 pregnant women at Mulago Hospital in Kampala, Uganda. More than 90% of cases and controls were Bantu, the largest ethnic group, with small numbers of Luo, Nilo-Hamites, and other ethnic groups. The ethnicity of the male partners and the sex ratios of the singleton babies in all of the groups were similar (Table S1). HIV+ women were not excluded from the analysis, because there were similar numbers in both pre-eclamptic and control pregnancies (∼5%) (Table S1), and similar results were found, even when HIV+ women were omitted (Table S2). As expected, gestational age at delivery and birth weight were significantly lower in the pre-eclamptic cases compared with controls (P < 0.001) (Fig. S1 and Table S1).

Unlike European Women, KIR B Centromeric Regions Containing KIR2DS5 Protect Ugandan Women from Pre-eclampsia.

Maternal KIR AA genotype is increased in pre-eclamptic pregnancies [P = 0.0256, odds ratio (OR) = 1.45] (Table 1), particularly when combined with the presence of fetal HLA-C alleles encoding the C2 epitope, similar to our findings in Europeans (P = 0.0318, OR = 1.49) (Fig. 1). We then analyzed which KIR B haplotype genes are protective. Three KIR B genes, KIR2DL2, KIR2DL5, and KIR2DS5, are more frequent in controls than in women with pre-eclampsia. Of these three genes, only KIR2DS5 is significantly protective for women with pre-eclampsia after Bonferroni correction (P = 0.0009, P_c = 0.0126, OR = 0.59) (Fig. 2A, Table 1, and Table S3). In comparable studies on European women, protection was seen with KIR2DS1 and not seen with KIR2DS5, which is shown here for African women (Table 1). Moreover, in Ugandans, the telomeric B (tB) genes KIR2DS1 and KIR3DS1 are at similar low frequency in cases and controls (Table 1).

Table 1.

Frequency of maternal KIR genotypes and KIR gene carriers

KIR genotypes or individual KIR genes present	Uganda pre-eclampsia cases (n = 251) n (%)	Uganda controls (n = 483) n (%)	P value*	OR (95% CI)	United Kingdom pre-eclampsia cases (n = 729) n (%)†	United Kingdom controls (n = 592) n (%)†	P value*	OR (95% CI)
KIR genotype
KIR AA	91 (36.3)	136 (28.2)	0.0256	1.45 (1.05–2.01)	266 (36.5)	163 (27.5)	0.0005	1.51 (1.20–1.91)
KIR AB	157 (62.5)	336 (69.6)	NS		456 (62.6)	424 (71.6)
KIR BB	3 (1.20)	11 (2.28)	NS		7 (0.96)	5 (0.84)
KIR genes
2DP1	247 (98.4)	474 (98.1)	NS		NA	NA
2DL1	247 (98.4)	476 (98.6)	NS		707 (97.0)	569 (96.1)	NS
2DL2	132 (52.6)	293 (60.7)	0.0365	0.72 (0.53–0.98)	348 (47.7)	313 (52.9)	NS
2DL3	222 (88.4)	414 (85.7)	NS		662 (90.8)	530 (89.5)	NS
2DL5	138 (55.0)	316 (65.4)	0.0061	0.65 (0.47–0.88)	330 (45.3)	330 (55.7)	0.0002	0.66 (0.53–0.82)
3DL1	248 (98.8)	473 (97.9)	NS		600 (95.5)‡	517 (94.3)‡	NS
3DS1	30 (12.0)	57 (11.8)	NS		211 (33.8)‡	242 (44.3)‡	0.0002	0.64 (0.51–0.81)
2DS1	52 (20.7)	114 (23.6)	NS		240 (32.9)	255 (43.1)	0.0002	0.65 (0.52–0.81)
2DS2	118 (47.0)	262 (54.2)	NS		349 (47.9)	317 (53.5)	NS
2DS3	56 (22.3)	118 (24.4)	NS		185 (25.4)	175 (29.6)	NS
2DS4	244 (97.2)	462 (95.7)	NS		703 (96.4)	560 (94.6)	NS
2DS4 del	73 (29.1)	144 (29.9)	NS		632 (89.9)	474 (84.8)	NS
2DS4 wt	171 (68.1)	318 (65.8)	NS		262 (37.3)	215 (38.5)	NS
2DS5	94 (37.5)	243 (50.3)	0.0009§	0.59 (0.43–0.81)	205 (28.1)	214 (36.1)	0.0023	0.69 (0.55–0.87)

CI, confidence interval; NA, not available; NS, not significant.

^*Fisher's exact test with mid-p adjustment.

^†Ref. 8.

^‡A number of individuals were not typed for this gene.

^§P = 0.0126 after Bonferroni correction.

Fig. 1.

Frequency of the KIR AA genotype alone and combined with the fetal HLA-C carrier group in Uganda and the United Kingdom. There was a significant difference in the KIR AA genotype frequencies between controls (light gray bars) and pre-eclampsia cases (dark gray bars) in both the Ugandan (*P = 0.0256, OR = 1.45) and the United Kingdom (***P = 0.0005, OR = 1.51) cohorts. The frequency of KIR AA genotypes is shown combined with either fetuses carrying a C2 epitope or those lacking C2 and carrying only C1-bearing HLA-C allotypes. There is a significant risk of pre-eclampsia when a KIR AA woman has a fetus carrying a C2 epitope for both cohorts. In Uganda, *P = 0.0318, and OR = 1.49, and in the United Kingdom, *P = 0.0267, and OR = 1.46.

Fig. 2.

Frequencies of the different genotypes carrying KIR2DS5 in controls and pre-eclampsia cases. (A) All controls (light gray bars) and pre-eclamptic cases (dark gray bars) were grouped according to whether they carried KIR2DS5. The presence of KIR2DS5 protects women from pre-eclampsia. ***P = 0.0009, OR = 0.59 (P_c = 0.0126 after Bonferroni correction). (B) Women were grouped according to the location of KIR2DS5 on the KIR B haplotype (cB, tB, contracted, or other unusual genotypes). KIR2DS5 on cB is significantly protective. **P = 0.0095, OR = 0.59. (C) The carrier frequencies of those KIR2DS5 alleles present on cB were compared between controls and pre-eclamptic cases. KIR2DS5*005C indicates those women in whom KIR2DS5 is located on cB. Only KIR2DS5*006 is significantly protective. **P = 0.0015, OR = 0.35.

Because KIR genes are in LD, KIR2DS5 could be itself protective or marking a nearby protective gene. KIR2DS5 can be found in both the KIR centromeric B (cB) and tB regions. To determine the location of KIR2DS5 in our cohort, we grouped individual genotypes according to their combination of centromeric and telomeric KIR regions based on previously described African KIR haplotypes (Materials and Methods and Fig. 3). Genotypes characteristic of expanded or contracted regions were also identified and shown to have similar frequencies in cases and controls.

Fig. 3.

Component genes of centromeric (c) and telomeric (t) KIR haplotype segments in African and European populations. The red segments together form the KIR A haplotype, and all other combinations of centromeric and telomeric motifs form KIR B haplotypes. The gene content motifs, shown for the (Left) centromeric and (Right) telomeric regions, are named according to (19), where KIR2DS3 and KIR2DS5 were considered as alleles of two genes, centomeric KIR2DS3/5 and telomeric KIR2DS3/5. The frequencies of the different KIR regions in representative African and European populations are also shown (15, 39). Afr, African; Eur, European.

Next, allele-level KIR2DS5 typing was performed, which identified 10 alleles that were assigned to cB or tB regions as described in Materials and Methods (Fig. 4). KIR2DS5*004, KIR2DS5*006, KIR2DS5*007, and KIR2DS5*010 are restricted to cB, whereas KIR2S5*002, KIR2S5*003, KIR2S5*008, KIR2S5*009, and KIR2S5*011 are restricted to tB. KIR2DS5*005 is the most frequent allele and the only one found in both cB and tB (Fig. 4), pointing to it being the progenitor of all other KIR2DS5 alleles. Our assignments of KIR2DS5 alleles to cB or tB agree with those defined by complete KIR haplotype sequences and analyses of African and African-American families (15, 18, 19). With all this information, we were able to determine the centromeric or telomeric location of KIR2DS5 for all KIR2DS5-carrying individuals.

Fig. 4.

Carrier frequencies of the different KIR2DS5 alleles found in the Ugandan population; cB alleles are in shades of red, tB alleles are in shades of blue, and KIR2DS5*005 (purple) is found in both cB and tB.

Comparison of the frequency of the centromeric and telomeric KIR2DS5 alleles in cases and controls shows that they differ in the protection that they provide against pre-eclampsia. KIR2DS5 is protective in Ugandan women when it is present in the cB region (cB01 or cB03; P = 0.0095, OR = 0.59) (Figs. 2B and 3 and Table S3). Furthermore, of all of the cB KIR2DS5 alleles, only KIR2DS5*006 is significantly more frequent in controls than in pre-eclamptic pregnancies (P = 0.0015, OR = 0.35) (Fig. 2C and Table S4). The dominant allele, KIR2DS5*005, has similar frequencies in both cases and controls, even when we can unequivocally assign its location to cB, and thus, seems neutral. Consistent with the low frequency of KIR2DS1 and KIR3DS1 in Africans, KIR2DS5 is less frequently present in tB than cB. When present in tB, it has no effect, being at similar frequencies in controls and cases (Fig. 2B and Table S3). Thus, the protective effect of KIR B is not just the absence of KIR A genes but also, the presence of genes belonging to a particular subgroup of cB regions: cB01 or cB03 (Fig. 3).

In Ugandan Women, Like European Women, Pre-eclampsia Associates with Maternal KIR AA Genotype Combined with Fetal Expression of Paternal HLA-C2.

We further examined the effect of different combinations of maternal KIR and fetal ligands: C1 and C2 epitopes of HLA-C allotypes. Considered alone, the C1 and C2 frequencies in mothers and babies do not significantly differ between cases and controls (Table S5). Using an extended Mantel–Haenszel test for linear trend, we find that KIR AB or BB genotype mothers carrying a C1C1 homozygous fetus have the least risk of pre-eclampsia, whereas a KIR AA mother carrying a C2 fetus has the greatest risk (P = 0.0122) (Fig. S2 and Table S6). Other genetic combinations have risks between these two extremes.

If the fetus has one more HLA-C allele encoding a C2 epitope than the mother, then the fetus must have inherited this C2 from the father. In this situation, the risk of pre-eclampsia in the absence of KIR2DS5 is increased (P = 0.0130, OR = 1.72) (Table 2). To explore this further, we defined the parental origin of the C2 for C1C2 heterozygous fetuses. When the single C2 is paternally inherited, the risk of pre-eclampsia associated with the absence of KIR2DS5 is greater (P = 0.0203, OR = 1.80) than when it is maternally inherited (not significant; OR = 1.16, 95% confidence interval = 0.72–1.84) (Table 2). Taken together, these findings show that there is an increased risk of pre-eclampsia in women with a KIR AA genotype lacking KIR2DS5 when the fetus has an HLA-C allele encoding a C2 epitope inherited from its father.

Table 2.

Risk of pre-eclampsia associated with the absence of KIR2DS5 for the different maternal/fetal HLA-C combinations

Parameter	P value*	OR (95% confidence interval)
Effect of relative dose of maternal and fetal HLA-C2
Fetus had fewer C2 than the mother	0.7085	1.09 (0.69–1.69)
Fetus had the same number of C2 as the mother	0.1612	1.28 (0.91–1.80)
Fetus had more C2 than the mother	0.0130†	1.72 (1.12–2.64)
Effect of origin of fetal HLA-C2
Paternal origin	0.0203†	1.80 (1.10–2.93)
Maternal origin	0.5222	1.16 (0.72–1.84)

^*Fisher's exact test with mid-p adjustment.

^†P < 0.05.

Recurrence of Pre-eclampsia in Ugandan Women Is Associated with Maternal KIR AA Genotype and Fetal Expression of Paternal HLA-C2.

The risk of recurrence of pre-eclampsia is known to be high (∼20%) (20, 21). In our cohort, there were 24 pre-eclamptic women who had recurrence of a hypertensive disorder of pregnancy, a condition on the same spectrum as pre-eclampsia. The 45.8% frequency of the KIR AA genotype in these women with recurrent pre-eclampsia was even higher than the frequencies of 36.3% in the full cohort and 28.2% in controls. Ten of eleven KIR AA pregnancies in this subcohort carried a C2 fetus.

Discussion

Our genetic study in an African population not only supports previous findings that certain combinations of maternal KIR and fetal HLA-C variants are associated with pre-eclampsia but also, reveals the benefits of studying multiple populations, including those most at risk for a disease. Pre-eclampsia occurs more commonly in African women, and the symptoms are of severe, early-onset disease associated with low birth weight and high mortality (4). Our findings have relevance to other disorders of pregnancy, because unexplained stillbirth, fetal growth restriction, and preterm labor are more common in women with African ancestry and share the same underlying problem of defective placentation with reduced maternal blood flow to the placenta (4).

There is considerably more genetic diversity of the KIR locus in Africans both at the level of KIR haplotypes and in the number of alleles at individual KIR genes (10, 15, 16). Despite this complexity, we find complete consistency with our studies of pre-eclampsia in Europeans: the risk is associated with a maternal KIR AA genotype combined with a paternally derived HLA-C allotype carrying a C2 epitope in the fetus (8, 17). Recurrent pre-eclampsia frequently occurs in African women (24.6% in a recent Tanzanian study), and the high frequency of KIR AA genotypes in the women in our study is striking (45.8% compared with 28.2% in controls) (21). The gene always present on the KIR A haplotype likely to confer this risk is KIR2DL1, which encodes for an inhibitory KIR with strict specificity for C2 epitopes (22). Thus, in women with a KIR AA genotype containing two copies of KIR2DL1, uNK will be strongly inhibited when confronted by trophoblast expressing C2-bearing HLA-C. There are at least 12 KIR2DL1 alleles located in the centromeric A (cA) region in Africans compared with 1–5 in other populations (15). Future analysis of larger cohorts including more women with recurrent pre-eclampsia should identify if there are particular KIR2DL1 alleles responsible.

One clear difference that might partially explain the increased risk of pre-eclampsia in Africans is the higher frequency of C2-bearing HLA-C allotypes across SSA compared with elsewhere in the world (14). The probability of African women having a C2-positive partner or fetus is 80% compared with 64% for European women. Similarly, the probability of African women having a fetus carrying a paternal C2 epitope is 55% compared with 40% for European women (Table S5). Given the selective pressure that pre-eclampsia imposes on a population, there must be other scenarios where C2 epitopes are beneficial. HLA-C and KIR are immune system genes with roles in outcome from viral infections, such as Hepatitis C and HIV (10, 23–25). In SSA, possession of C2 epitopes might be advantageous in responding to a range of pathogens, including malaria. Studies of how HLA-C and KIR variants affect responses to infection in SSA are still limited, especially in the crucial period from birth to adolescence.

We observed that tB regions containing KIR2DS1 provide a protective effect for pre-eclampsia in Europeans (8). In contrast, we now show that, in Ugandans, KIR cB regions characterized by KIR2DS5, KIR2DP1, and KIR2DL1 (cB01 and cB03) are protective. The low carrier frequency of KIR2DS1 in SSA (1.4–27.8%) compared with Europe (42.5%) also suggests that KIR2DS1 does not play an important role in pregnancy success in Africans (14). One explanation for the different protective effect is that KIR2DS5, an activating KIR that likely evolved from a KIR specific for C2, does function like KIR2DS1, although there is no evidence to date that the C2 epitope is a KIR2DS5 ligand (22). The single KIR2DS5 allele in Europeans, KIR2DS5*002, is in tight LD with KIR2DS1 and located in the tB region. Unlike Europeans, however, KIR2DS5 is polymorphic in Africans and African Americans. We found 10 alleles in Ugandans, consistent with previous reports from African Americans [the Immuno Polymorphism Database (IPD)], located in both cB and tB but most commonly found in those cB regions that also contain KIR2DP1 and KIR2DL1 (26, 27). The dominant allele, KIR2DS5*005, is the only allele found in both cB and tB, and it is probably ancestral; when in either location, it was similar in frequency between cases and controls. Of cB KIR2DS5 alleles, only KIR2DS5*006 is significantly associated with protection from pre-eclampsia. KIR2DS5 can be expressed by European peripheral blood NK cells, but we have been unable to show its expression on uNK using similar reagents (28–30). The functional effects of KIR2DS5 diversity await additional investigation, but certain KIR2DS5 allotypes do show different expression levels in transfected cells, similar to findings for other KIR variants (30). For example, allelic variation of KIR2DL1 affects protein expression levels at the cell surface, NK repertoire, and affinity of binding (22, 31, 32). Furthermore, although no binding has been shown of the European allele KIR2DS5*002 to any HLA ligand, KIR2DS5*006 might bind to C2-bearing HLA-C allotypes common in Africans (C*04, C*02, C*17, and C*18) (15).

Another possibility is that KIR2DS5 is in LD with other KIR on the protective cB01 and cB03 regions, notably KIR2DL1. The cB KIR2DL1 allele present in Europeans, KIR2DL1*004, gives a weak inhibitory signal compared with the common cA allele, KIR2DL1*003 (31). Thus, the protective effect of the cB01 and cB03 regions might be due to either KIR2DS5 activation or weaker KIR2DL1 inhibition, because both could counterbalance the strong inhibition conferred from cA KIR2DL1 alleles. For both KIR A and B haplotypes, the particular KIR2DS5 and KIR2DL1 alleles involved are, therefore, important, but to investigate this will require much larger, clinically well-characterized cohorts. Our method to infer KIR regions allows a fairly simple analysis of KIR data from clinical cohorts in SSA compared with the complex sequencing needed to define the exact haplotypes (15). Hence, although this analysis does not unravel the complete complexity of KIR variants found, it can point to the regions conferring risk or protection. In this clinical context, we have a clear pointer that the cB01 and cB03 regions containing KIR2DS5, KIR2DL1, and KIR2DP1 are providing protection from pre-eclampsia in Ugandan women.

In this African cohort, such as in Europeans, a paternal rather than a maternal origin of fetal C2 confers risk in women lacking KIR2DS5 (8). Whether this effect is caused by disparities between individual maternal and paternal HLA-C2 allotypes (allogeneic) and/or a dosage effect (more HLA-C alleles encoding C2 in the fetus than in the mother when C2 is paternally derived) is unresolved (8). To understand this effect will require high-resolution genotyping of C1C2 mothers who have C1C2 babies (where the dosage is identical) in a large cohort (2,000 cases and 4,000 controls would be required).

The great diversity of KIR and HLA-C variants in SSA is maintained by balancing selection (16). The two contrasting functions of these immune system gene families in reproduction and immune responses to infection mean that certain variants will be important at different stages of life in women, men, children, and adults and in geographical regions with a range of different pathogens. We have previously argued that the selective pressures from reproductive success and immune response to pathogens are competing and have driven evolution of the KIR A and B haplotypes in humans compared with other hominids (10). Our combined studies of KIR/HLA-C variants in diverse European and African populations now suggest that the unusual reproductive strategies characteristic of modern humans compared with other hominids could also be a cause of balancing selection. The evolution of the large neonatal brain relative to a pelvis adapted for bipedalism means that birth weight must be kept between two strictly defined limits. When babies are large (>95th centile), there is a risk of cephalopelvic disproportion and subsequent prolonged obstructed labor, birth asphyxia, and postpartum hemorrhage. Furthermore, these outcomes, like pre-eclampsia and other GOS, are also much more common in African women with associated features of pregnancy that favor smaller babies: earlier birth (the gestational age is reduced to 38 wk), the head engages late into the pelvis, and the baby matures earlier than in non-Africans (4). Thus, there is high mortality in mother and babies not only from pre-eclampsia (associated with low birth weight and still birth) but also, at the other end of the normal birth weight spectrum. Both mothers and their babies benefit if the latter have intermediate birth weights and the two extremes of very low and high birth weight are selected against. The balance between these two extremes is partially determined at placentation, when uNK allows trophoblast cells to access sufficient maternal oxygen and nutrients without starving the baby (defective trophoblast invasion) or risking uterine rupture (excessive trophoblast invasion) (3). In an African population, because of the greater risk of cephalopelvic disproportion (4), there is even greater selection for reduced fetal size with associated pre-eclampsia; this effect is consistent with the higher frequency of maternal KIR AA/paternal C2 combinations in SSA.

In Europeans, opposing KIR/HLA-C combinations are associated with the extremes of birth weight: a paternal C2 epitope is associated with both extremes, but in pre-eclampsia and low birth weight (less than fifth centile), the risk is with maternal KIR AA genotypes, whereas in high birth weight, the association is with maternal KIR2DS1 (33). Studies on how these genetic findings are translated in uNK functional differences are still limited, but we found that, when KIR2DS1+ uNKs (isolated from United Kingdom patients) are activated by target cells expressing HLA-C2, there is increased production of soluble factors [e.g., granulocyte-macrophage colony-stimulating factor (GM-CSF)] that enhance trophoblast invasion (34).

Thus, there is a balance between the KIR A and KIR B haplotypes in both populations, but they differ in the regions of the KIR B haplotype that correlate with protection from pre-eclampsia. tB regions and KIR2DS1 are infrequent in Africans compared with Europeans, but the opposite is true for cB regions containing KIR2DS5. During the out-of Africa migrations, it is possible that only individuals having tB with KIR2DS1 moved away from SSA. Introgression of KIR2DS1 from archaic humans is also a possibility (35). Our previous findings do indicate that KIR2DS1 and KIR3DS1 (both on tB) are selected against in SSA (14, 16). Studying disorders of pregnancy in an African setting is important and informative; the high rates of pre-eclampsia as well as other major disorders of pregnancy, including obstructed labor and stillbirth, and the greater genetic diversity of KIR in SSA mean that unraveling the role of the complex KIR and HLA systems will provide valuable genetic information to predict women who are at risk for a range of pregnancy disorders.

Materials and Methods

Ethics Statement.

Approval to conduct the study was given by the Higher Degrees Research and Ethics Committee of Makerere University College of Health Sciences and the Uganda National Council for Science and Technology. The participants gave written informed consent to participate in the study. Withdrawal from the study never jeopardized healthcare, which was provided free to all women.

Study Design.

This study was conducted at Mulago National Referral and Teaching Hospital located in Kampala, which functions as a tertiary referral center for Uganda. Mulago Hospital is the busiest maternity hospital in SSA, with over 30,000 deliveries a year. Genomic DNA was obtained from maternal blood from unrelated healthy women (n = 484) and women with pre-eclampsia or eclampsia (n = 254) between July of 2009 and June of 2011. Umbilical cord samples were obtained from the babies for genomic DNA isolation. Pre-eclampsia was defined as hypertension of 140/90 mmHg or more on more than one occasion at least 4 h apart plus proteinuria of +2 or more by dipstick both at 20 wk or more of gestation. Eclampsia was diagnosed when a patient with pre-eclampsia had generalized tonic–clonic convulsions. Controls were women with a normal first pregnancy delivering at term (≥38 wk) who were normotensive with no proteinuria. Excluded from controls were patients taking long-term medication and patients with other diseases, including chronic hypertension and renal disease but excluding HIV. Women who had received a blood transfusion within the last 3 months were also excluded. Cases and controls were consecutively recruited from the same catchment area during the study period. Data were collected at the time of clinical examination of the participants using an interviewer-administered questionnaire, and additional information was obtained from medical charts.

DNA Isolation and Genotyping.

Maternal genomic DNA was isolated from 5 mL blood using the QIAamp DNA Maxi Blood Kit (Qiagen). Fetal DNA was isolated from umbilical cord samples after overnight incubation with Proteinase K (Roche) and purification with a protein precipitation solution (Qiagen) followed by ethanol precipitation. Twelve maternal KIR genes were typed for presence or absence by PCR with sequence specific primers (PCR-SSP) using two pairs of primers per gene or allele as described previously (8, 14, 36). The KIR genes typed were 2DL1, 2DL2/3, 2DL5, 3DL1/S1, 2DP1, 2DS1, 2DS2, 2DS3, 2DS4 (including the deletion), and 2DS5. All of the samples were typed for KIR2DL1 and KIR2DP1 copy number, and 28 selected samples were further investigated for additional KIR (2DL4, 3DP1, 3DL2, and 3DL3) so that all 14 KIR genes were included (37). Individual genotypes were defined according to their combination of centromeric (cA and cB) and telomeric (tA and tB) KIR regions based on previously described African KIR haplotypes (14, 15, 18). We first discriminated KIR A from KIR B regions on the basis of the presence/absence of 2DS2, 2DL2/3, 2DP1, 2DL1, 3DL1/S1, 2DS1, and 2DS4. There are common cB regions in Africans (Fig. 3) that were identified in individuals with a cB region using information from the presence/absence of individual KIR genes and the copy number of KIR2DL1 and KIR2DP1 (18). Typically, cB01 and cB03 have 2DP1, 2DL1, 2DL5, and 2DS5 (or 2DS3), whereas cB02 lacks these genes. KIR2DS5 alleles were genotyped by pyrosequencing; targeting exons 5–7 (15). Then, by knowing which KIR2DS5 alleles are present in individuals homozygous for either cA or tA regions, we could assign each of 10 KIR2DS5 alleles to cB or tB (Fig. 4). C1 and C2 were defined in maternal and fetal samples based on the primers and methods described previously (8, 36). HLA-C low-resolution allelic typing was performed using a PCR-SSP method consisting of 21 reaction wells that was adapted from ref. 38. Each well contained a final reaction volume of 10 µL consisting of 5× Green GoTaq Flexi Buffer (Promega), 0.2 mM dNTPs (ThermoFisher), 1.25 mM MgCl2 (Promega), 0.4 U GoTaq DNA polymerase (Promega), 134 nM 63/64 control primer (Eurogentec), and ∼45 ng DNA. PCR products were run on a 1% agarose gel and visualized using a UV and ethidium bromide.

Statistical Analysis.

Unless otherwise indicated, categorical data were analyzed using the χ² and Fisher’s exact tests with two-tailed mid-p adjustment and Student’s t tests for continuous data. A P value of ≤0.05 was considered to be statistically significant. The magnitude of the effect was estimated by ORs and their 95% confidence intervals.