METHODS

Study area and cohort

The study was conducted in Prampram (population 8000), approximately 50 km east of Accra on the south coast of Ghana, in an area of coastal savanna. Malaria transmission in the area is perennial (although with some seasonal variation), stable, and of low to moderate intensity (5–10 infectious bites per person per year).¹⁸P falciparum accounts for 98% of all detected infections. Informed consent was obtained from all volunteers and ethical approval for all procedures was obtained from the ethical review committee of the London School of Hygiene and Tropical Medicine, the institutional review board of the Noguchi Memorial Institute of Medical Research, University of Ghana, and the Ghanaian Ministry of Health. After giving informed consent, a consecutive sample of 110 mothers and their newborn infants was recruited into the study. A heparinised heel prick blood sample was obtained from each child on the day of delivery, at 2, 4, and 6 weeks of age, and then every four weeks. The child’s axillary temperature was measured and a health questionnaire was completed by the mother every two weeks. If the child was febrile (temperature ⩾37.5°C), a blood film was made and examined immediately so that anti-malarial chemotherapy could be instituted if necessary. All children with fever and a parasite positive blood film were treated with a full course of chloroquine. As blood films from clinically well children were not read until some weeks later, asymptomatic infections were not treated. As free treatment was provided by study staff and was continuously available, we believe that very few infections were treated without our knowledge. Antimalarial chemoprophylaxis was not available to pregnant women in Prampram during the study period and thus transfer of antimalarials across the placenta, or via breast milk, was estimated to be minimal. Infants did not receive malaria prophylaxis.

Parasite detection

Giemsa stained thick blood films were examined by oil immersion microscopy. The number of parasitised erythrocytes per 300 leucocytes was counted and the number of parasites per μl of blood calculated was based on an average leucocyte count of 13 000/μl in children under 1 year of age.¹⁹ Slides were classified as negative only after a minimum of 1000 leucocytes had been counted; the minimum parasite density that could reliably be detected by microscopy was 40 parasites/μl.

To improve the sensitivity of parasite detection, DNA was extracted from the red cell pellet²⁰ and amplified using primers to the multicopy, subtelomeric 7H8/6 gene of P falciparum, as described previously.¹ This method had a resolution of approximately one parasite per μl,¹ offering greater than 10-fold sensitivity over microscopy alone.

Examination of APOE genotype

DNA prepared from the red cell pellet contained also human DNA extracted from white blood cells trapped in the pellet, and so it was usable for determination of APOE genotypes. This was done using the method of Wenham et al.²¹

Statistical analysis

The ages (in weeks) at which infection was detected were evaluated using the Kaplan-Meier survival analysis method in conjunction with the log rank test. This is a statistical procedure for estimating time to event models in the presence of differing follow up times and “censored” cases (that is, in this context, of infants who remained infection free at the termination of the study). Mean times to infection were compared between APOE genotypes.

The maximum parasite counts detected in each infected infant were compared using one way analyses of variance. As these counts were very positively skewed, a natural logarithmic transformation was applied to normalise the data distribution. Geometric mean maximum parasite densities were also compared between APOE genotypes and allele frequencies.

In the presentation of results, all summary statistics are accompanied by their 95% confidence limits.

RESULTS AND DISCUSSION

Infection occurred in 96 (87.3%) of the 110 infants during the study period. The mean time to infection was 48 weeks (95% confidence interval 41 to 55). Among the 110 children for whom an APOE genotype was obtained, the allele frequencies were 14.5%, 61.4%, and 28.5% for ε2, ε3, and ε4, respectively. Overall, mean time to infection differed (statistically) significantly between genotypes (log rank test: χ²(5)=15.69, p=0.008), indicating that APOE genotype affects risk of malaria infection. A more detailed evaluation showed that the latter result was the result of the mean time to infection being significantly shorter for infants with a ε2/ε2 genotype compared to all other genotypes, but the differences between the remaining five genotypes were all non-significant (log rank test: χ²(4)=2.94, p=0.568). However, this result must be interpreted with great caution as only four infants had an ε2/ε2 genotype. Calculation of the expected frequency of APOE ε2 homozygotes in this population, using the Hardy-Weinberg distribution, gives a value of only 2.6%, so clearly to evaluate our findings a much larger study would be needed to yield an adequate number of these homozygotes.

A secondary analysis was carried out based on the presence of a particular type of allele (table 1). No significant relationship was found between mean time to infection and the presence of a ε2 allele (log rank test: χ²(1)=0.48, p=0.491), the presence of a ε3 allele (log rank test: χ²(1)=0.10, p=0.755), or the presence of a ε4 allele (log rank test: χ²(1)<0.01, p=0.985). Thus, if APOE ε2 does confer a risk, possibly by affecting entry of sporozoites into the liver, it does so only in homozygotes.

To find if the number of maturing liver schizonts (and thus parasite density in the blood) depended on APOE genotype, we examined parasite density. In most of the infants, the density was very low, being detectable by PCR but not reliably so by microscopy. Values were mainly in the range of tens or hundreds of infected red blood cells per microlitre, with relatively few in the low thousands and only one very high value, 60 000/μl. Table 2 shows the (geometric) mean maximum parasite counts for each of the genotypes. The average parasite density was slightly lower for those infants who were APOE ε2/ε2 homozygotes, but overall the differences between the subgroups were not statistically significant (F(5,90) = 0.580, p=0.715). Similarly, maximum parasite counts were not significantly affected by the presence of an ε2 allele (F(1,94) = 0.055, p=0.815), the presence of a ε3 allele (F(1,90) = 0.036, p=0.849) or the presence of a ε4 allele (F(1,90) = 0.883, p=0.350).

Thus, our preliminary data suggest that APOE may affect susceptibility to malaria infection, in contrast with the previous findings on HSV1, HCV, and HIV,^5–9 which show that APOE does not influence susceptibility but instead influences the extent of damage caused by these agents. Whether the severity of disease caused by malaria infection is determined by APOE genotype remains to be discovered, by comparing genotypes of subjects with advanced disease with those found here for mild disease.

The high overall APOE ε2 allele frequency of our group of infants is noteworthy (table 3). All allele frequencies were calculated using values for all the 110 infants investigated; since the survey was a prospective cohort study and the infants comprised a random sample of the population, allele frequencies can be taken as population values. The value of 14.5% for APOE ε2 is substantially higher than almost all the other known values for sub-Saharan African populations (Pygmies, Nigerians (apart from group 3), Sudanese, Ethiopian, and South African Khoi San), but almost identical to the Tswana speaking South Africans (table 3). The Prampram APOE ε4 allele frequency of 24.1 is substantially lower than the values for Pygmies, group 3 Nigerians, Ethiopians, and South African Khoi San, but very similar to the Tswana value, while both the APOE ε2 and ε4 allele frequencies of the Prampram infants are much higher than those for UK subjects (table 3).

In summary, this study points to a possible correlation between susceptibility to malaria infection and homozygous carriage of the APOE ε2 allele, in that APOE ε2 homozygous infants became infected with malaria earlier than other children; this was accompanied by a relatively high frequency of the APOE ε2 allele in the study population. One interpretation of these data might be that early malaria infection is beneficial in the long term, as has been suggested by epidemiological studies,^24,25 perhaps allowing children to acquire protective immune responses while being relatively protected from severe malaria by passively transferred maternal antibodies or by physiological factors.²⁶ Clearly, further studies of the role of APOE in susceptibility to malaria are warranted using a larger numbers of infants, and in particular identifying more infants with an ε2/ε2 combination in order to improve the precision of the estimates of both time to infection and maximum parasite counts. Also, it would be of interest to examine the APOE genotypes of older children or adults with much more severe disease, including cerebral malaria, to find out if APOE does in fact determine susceptibility to severity of damage caused by the protozoon.