Combined effects of Gm or Km immunoglobulin

Microbes and Infection 13 (2011) 771e775 www.elsevier.com/locate/micinf

Short communication

Combined effects of Gm or Km immunoglobulin allotypes and age on antibody responses to Plasmodium falciparum VarO rosetting variant in Benin Florence Migot-Nabias a,b,c,d,*, Adjimon G. Lokossou a,b,c,d, Ine`s Vigan-Womas e, Evelyne Guitard f, Micheline Guillotte e, Julien M. Noukpo a,c,d, Odile Mercereau-Puijalon e, Jean-Michel Dugoujon f, Andre´ Garcia a,b,c,d a

Institut de Recherche pour le De´veloppement, UMR 216 Me`re et enfant face aux infections tropicales, 4 avenue de l’Observatoire, 75006 Paris, France b Faculte´ de Pharmacie, Universite´ Paris Descartes, 4 avenue de l’Observatoire, 75006 Paris, France c Institut des Sciences Biome´dicales Appliqueés, 01 BP 918 Cotonou, Benin d Laboratoire de Parasitologie, Faculte´ des Sciences de la Sante´, 01 BP 188 Cotonou, Benin e Institut Pasteur, Unite´ d’Immunologie Molećulaire des Parasites, CNRS URA 2581, 28 rue du Dr Roux, Paris, France f Laboratoire d’Anthropologie Molećulaire et Imagerie de Synthe`se, UMR 5288, Universite´ Paul Sabatier Toulouse III, 37 alleés Jules-Guesde, 31073 Toulouse, France Received 21 January 2011; accepted 4 April 2011 Available online 13 April 2011

Abstract Clinical protection of Beninese children against Plasmodium falciparum malaria was shown to be influenced by immunoglobulin (IG) Gm and Km allotypes, and related to seroreactivity with the rosette-forming VarO-antigenic variant. IgG to the VarO-infected erythrocyte surface, IgG1 and IgG3 to PfEMP1-NTS-DBL1a1-VarO were higher in the under 4-year-old children carrying the Gm 5,6,13,14;1,17 phenotype. In contrast, surface-reactive IgG, total IgG, IgG1 and IgG3 to NTS-DBL1a1- and DBL2bC2-VarO domains were lower in the above 4-year-old children harbouring the Km1 allotype. These data outline an age-related association of antibodies against malaria antigens and IG allotype distribution. Ó 2011 Institut Pasteur. Published by Elsevier Masson SAS. All rights reserved. Keywords: Malaria; Plasmodium falciparum; Gm and Km allotypes; Palo Alto VarO strain; Antibody responses

1. Introduction Individual quantitative and qualitative differences in specific immunoglobulin (IG) responses of the IgG class include the genetic polymorphism of Gm and Km allotypes, located in the constant regions of the gamma and the kappa chains, respectively. Overall, 18 Gm allotypes are described on * Corresponding author. Florence MIGOT-NABIAS, IRD UMR216, Faculte´ de Pharmacie, 4 avenue de l’Observatoire, 75006 Paris, France. Tel.: þ33 1 70 64 94 34; fax: þ33 1 53 73 96 17. E-mail address: [email protected] (F. Migot-Nabias).

the constant region of the heavy gamma1 (4 G1m allotypes), gamma2 (1 G2m allotype) and gamma3 (13 G3m allotypes) chains of the IgG1, IgG2 and IgG3 subclasses, respectively. Three allotypes, encoded by the IGKC gene on chromosome 2 (2p 12), constitute the Km polymorphism of the kappa light chain. Polymorphic Gm allotypes are encoded by alleles of closely linked genes, located on chromosome 14 (14q32.3), which are inherited in fixed combinations called haplotypes [1]. A limited number of Gm haplotypes are observed worldwide, with highly variable frequency between human populations from different continents [2]. Gm allotype-related levels of IgG subclasses have been observed in Caucasians [3]

1286-4579/$ - see front matter Ó 2011 Institut Pasteur. Published by Elsevier Masson SAS. All rights reserved. doi:10.1016/j.micinf.2011.04.001

772

F. Migot-Nabias et al. / Microbes and Infection 13 (2011) 771e775

and African Americans [4], suggesting differences in the subclass specificity of the antibody responses to infectious agents. Individual variations of Gm allotype frequencies are also described, and may be related to the modulation of the IG functionality by particular allotypes [5]. An involvement of the Gm/Km allotypes in the susceptibility to parasitic diseases was indicated by the negative and positive association of the carriage of the Km3 and Km1,3 phenotypes, respectively, with susceptibility to infection by the filarial parasite Onchocerca volvulus in Ecuador [6]. Differences in the prevalence rates recorded among populations from two nearby villages from Philippines where lymphatic filariasis is endemic were putatively attributed to a distinct distribution of Gm phenotypes in these populations [7]. A recent genomewide linkage study performed among West Africans showed that a region on chromosome 2p influenced the intensity of O. volvulus infection [8], although this region was located at too great a genetic distance from the Km allotypes to account for the linkage [9]. In the case of Plasmodium falciparum malaria, a preferential expression of G3m10, G3m11 and G3m14 allotypes and not G3m5 was reported for the IgG3 sensitising the erythrocytes of Gambian children [10]. Recently, higher prevalence rates of the Gm 5,6,13,14;1,17 phenotype were observed among the Masaleit tribe of Sudan, compared with Fulani people, who presented lower values of malariometric indices [11]. Investigation in another Fulani population from Sudan outlined a greater susceptibility to clinical malaria of carriers of the Gm 5,6,13,14;1,17 phenotype [12]. In a previous study [13], we observed an association of IgG allotype distribution with clinical presentation of falciparum malaria in Beninese children living in a malaria-endemic area. An inverse relationship was found between the Gm 5,6,13,14;1,17 phenotype and uncomplicated malaria, while Km1 was mainly found in children with severe malaria. In the same group of children, we recently showed that the specific antibody response to the VarO-antigenic variant that forms rosettes and autoagglutinates was associated with protection against clinical malaria [14]. In the study reported here, we have investigated how the VarO-reactive antibodies related with the particular Gm and Km allotypes associated with either resistance (Gm 5,6,13,14;1,17) or susceptibility (Km1) to clinical malaria in these children.

2. Methods 2.1. Population study One hundred fifty-four Beninese children (mean age [range] ¼ 4.5 [0.2e13.0] years) were considered in the study, 88% of them belonging to the Gbe ethno-linguistic group and 53% to the more restrictive Fon ethnic group [13]. Informed consent was obtained from parents or guardians of the children. The study was authorised by the ethic committee of the Faculte´ des Sciences de la Sante´ (FSS) from the Universite´ d’AbomeyCalavi (UAC) in Benin.

2.2. Determination of Gm and Km immunoglobulin allotypes Gm and Km allotypes for G1m (1, 2, 3, 17), G3m (5, 6, 10, 11, 13, 14, 15, 16, 21, 24, 28) and Km (1) determinants in plasma samples were analysed by a standard haemagglutinationinhibition method [15]. Only G3m (5, 6, 13, 14, 21) were considered for the definition of Gm phenotypes because of their sufficient value for the discrimination of the major worldwide distributed haplotypes and their previously described association with the function of the IG [3]. According to IMGT-ONTOLOGY [16], Gm phenotypes are written by grouping together the markers that belong to each IgG subclass, by the numerical order of the marker and for the subclass on the gene order; markers belonging to different subclasses are separated by a semicolon, while allotypes within a subclass are separated by commas. 2.3. Antibody measurements Surface immunofluorescence was used to assay the IgG reactivity of each plasma sample (diluted 1/20) to the surface of a monovariant culture of infected erythrocytes expressing the PfEMP1-VarO adhesion. Data were analysed by flow cytometry that allowed expressing the IgG surface reactivity in arbitrary units (AU), calculated from the percentage of IgG-bound infected erythrocytes. The positivity threshold was set at 20 AU, which is the mean surface reactivity plus 3 standard deviations observed with plasma samples from 20 non immune individuals living in France [17]. Total IgG, IgG1 and IgG3 reactivity to three soluble recombinant PfEMP1-VarO domains including the NTS-DBL1a1 domain and the DBL2bC2 domain (produced in baculovirus/ insect cells system) and the CIDR-g domain (produced in Pichia pastoris) was determined by ELISA in children plasma diluted 1/100. Results were expressed in AU calculated from the optical density value at 450 nm of tested plasma relative to negative and positive controls included in each assay. A positivity threshold was defined for each class/subclass and each antigen [13]. 2.4. Statistical analysis Differences in proportions were tested using the Chi square test. Statview 5.0 (SAS Institute Inc., Cary, NC) was used for these calculations. The associations between antibody levels and Gm and Km immunoglobulin allotypes taking into account age and ethnic group were investigated by a multiple linear regression analysis using STATA (Stata-Corp. Release 8.0). The existence of an interaction between age and Gm and Km immunoglobulin allotypes was systematically tested. For all tests, P values of less than 0.05 were considered significant. 3. Results 3.1. Univariate analysis of Gm and Km phenotypes and anti-VarO antibody levels Gm 5,6,13,14;1,17 was the main phenotype, found in 73 (47%) children, and Km1 was present in 105 (68%) children


[13]. Analysis of the distribution of the antibody responses by Gm phenotypes identified a trend towards higher prevalence rates among carriers vs. non-carriers of Gm 5,6,13,14;1,17 for surface-reactive IgG to VarO-infected erythrocytes (P ¼ 0.07), IgG3 to NTS-DBL1a1 (P ¼ 0.08), IgG and IgG1 to CIDRg (P ¼ 0.09 and 0.10, respectively) (Fig. 1A). An inverse trend was noticed when dichotomising antibody data according to presence or absence of the Km1 allotype: antibody responders were more numerous among non-carriers of Km1 compared to carriers of this allotype, especially in the case of IgG responders to CIDRg (P ¼ 0.07) and DBL2bC2 (P ¼ 0.005) (Fig. 1B). 3.2. Multiple linear regression on Gm and Km phenotypes and anti-VarO antibody levels As age had a very strong effect on antibody levels (P < 103 in all cases: data not shown), all analyses were adjusted for age effect. Moreover, when we found a significant effect of Gm 5,6,13,14;1,17 or Km1 phenotypes on antibody responses (see below) there was a significant interaction between age and the mentioned phenotype, meaning that the classical effect of age on antibody levels differed between children harbouring or not, the phenotype. To exemplify this effect, the results have been presented by stratifying the analyses considering two age groups: less than or equal to 4 years (n ¼ 75) and greater than 4 years

773

(n ¼ 79). According to these age categories, Gm 5,6,13,14;1,17 was present in 27 and 46 children of 4 years and >4 years, respectively and Km1 was present in 52 and 53 children of 4 years and >4 years, respectively. Within each age group, the analyses were adjusted for ethnic group using a multiple linear regression model. As we showed previously that Gm and Km phenotypes (i.e. our exposure factor) were associated with clinical presentation and parasite density [13], we did not include these factors as covariates in multivariate analyses. Indeed, it has been demonstrated that the inclusion of such factors (that could change the estimated effect of exposure but may themselves be affected by the exposure) as explicit terms in a model, can severely bias measures of the exposure-associated risk [18,19]. The Gm 5,6,13,14;1,17 phenotype was associated with higher levels of surface-reactive IgG to VarO-infected erythrocytes in children aged less than 4 years (P ¼ 0.031) (Table 1). Presence of Gm 5,6,13,14;1,17 was associated with higher NTSDBL1a1-specific IgG1 and IgG3 levels in the under 4 year old children (P ¼ 0.029 and P ¼ 0.018 respectively). No effect of the carriage of Gm 5,6,13,14;1,17 phenotypes on antibody levels to neither CIDRg nor to DBL2bC2 domain was found. The carriage of Km1 was associated with lower levels of VarO-surface reactive IgG in children older than 4 years (P ¼ 0.003). It was associated with total IgG (P ¼ 0.034), IgG1 and IgG3 to NTS-DBL1a1 (P ¼ 0.021 and P ¼ 0.009,

Fig. 1. Prevalence rates of surface-reactive IgG to VarO-infected erythrocytes and of domain-reactive IgG, IgG1 and IgG3 in Beninese children presenting or not the Gm 5,6,13,14;1,17 phenotype (A) and the Km1 allotype (B) Bars denote the 95% confidence interval; asterisks indicate P values estimated using the Chi-square test as follows: *, P < 0.10; **, P < 0.05.

774


for IgG1 and IgG3 respectively) with a lower antibody level for children older than 4 years. Furthermore, the carriage of Km1 was associated with lower antibody levels in children older than 4 years for DBL2bC2-total IgG (P ¼ 0.037), -IgG1 (P ¼ 0.012) and -IgG3 (P ¼ 0.038). No effect of the carriage of Km1 phenotypes on antibody levels to CIDRg was found. The combined effect of Gm and Km phenotypes could not be evaluated because of too small effectives in subgroups to be compared. Indeed, 50 children carried both Gm 5,6,13,14;1,17 and Km1 phenotypes, and distributed into 19 and 31 children of 4 years and >4 years, respectively. For all analyses belonging to the Fon ethnic group had no significant effect. 4. Discussion This is the first demonstration of an age-related effect of the IG allotypes on acquisition of antibodies directed to the variant antigen displayed into the surface of the P. falciparum-infected erythrocyte. A positive relationship between Gm 5,6,13,14;1,17 carriage and antibody responses directed to some P. falciparum asexual blood-stage antigens was recently observed in Sudan, where there was a negative relationship with Km1,3 allotypes [12]. Paradoxically, a greater susceptibility to malaria attacks prevailed in this study among carriers of the Gm 5,6,13,14;1,17 phenotype. The influence of age could not be evaluated in the Sudanese study, because of the large age range (from 8 to 45 years) of the investigated population.

An age-related impact of host genetic factors on either the rate of acquisition or the overall level of antimalarial antibodies was previously demonstrated for the sickle-cell trait carriage [20] or for HLA class II alleles [21]. Due to a complex interaction between age and the genetic factor involved, the effect of age on the evolution of the antibody response could differ according to the presence or the absence of the genetic factor, the influence of which being more easily detectable in young children, having not yet acquired their natural immunity against malaria. Some mechanistic explanations have been proposed for a better comprehension of the interaction between the constant regions of the gamma chains (containing Gm allotypes) and the variable region of the IG, to which the antigen-binding affinity and specificity has been for a long time attributed in exclusivity [22]. In the same way, in spite of their localisation on the constant region of the kappa chain, Km markers may be involved in antibody specificity by means of linkage disequilibrium between alleles coding for definite allotypes and determinants of the variable kappa region [23]. This study reinforces the notion of the impact of the carriage of particular Gm and Km allotypes in antibody response against pathogens. In malaria, protective genetic factors play a major role in young children and their influence is subsequently diminished as protective immune responses are acquired. Inversely, genetic factors that negatively impact on immune response have an influence at an older age, as the child accumulates exposure and immune responses to multiple infections. Interestingly both types

Table 1 Effects of Gm 5,6,13,14;1,17 and Km1 immunoglobulin allotypes on surface-reactive IgG to VarO-infected erythrocytes and on domain-reactive IgG, IgG1 and IgG3, in Beninese children. Antigens

Antibody levels

Age groups

Gm 5,6,13,14;1,17

Km1 a

Parameters of the best fitting model

Parameters of the best fitting model

VarO-IE surface

IgG

NTS-DBL1a1

IgG IgG1 IgG3

DBL2bC2

IgG IgG1 IgG3

CIDRg

IgG IgG1 IgG3

4 >4 4 >4 4 >4 4 >4 4 >4 4 >4 4 >4 4 >4 4 >4 4 >4

years years years years years years years years years years years years years years years years years years years years

Cons

b

CI95

2.3 3.5 38.3 72.3 27.5 70.8 30.8 54.1 17.2 46.3 24.2 51.8 18.6 39.6 5.2 38.5 16.3 45.2 23.4 38.2

þ0.6 þ0.2 þ7.3 2.7 þ13.8 5.6 þ15.2 5.6 þ1.8 2.1 þ3.9 þ4.1 þ1.5 2.9 þ4.7 5.6 þ11.3 0.1 1.2 6.2

[þ0.1 e þ1.1] [0.3 e þ0.6] [8.3 e þ22.9] [20.7 to þ15.3] [þ1.5 e þ26.2] [21.7 e þ10.5] [þ2.7eþ27.8] [20.0 e þ8.9] [9.2 e þ12.8] [19.9 e þ15.7] [8.2 e þ16.1] [11.8 e þ20.0] [10.9 e þ13.9] [18.7 e þ12.4] [4.8 e þ14.2] [25.0 e þ13.9] [0.3 e þ22.9] [18.0 e þ17.8] [14.8 e þ12.4] [20.6 e þ8.2]

b

Pc

Cons

b

CI95

P

0.031 0.46 0.35 0.77 0.029 0.49 0.018 0.44 0.74 0.81 0.52 0.61 0.81 0.70 0.33 0.57 0.06 0.90 0.86 0.39

2.4 4.0 29.9 83.1 28.3 79.7 40.6 63.0 20.8 57.2 22.0 67.2 18.5 48.3 9.7 44.7 22.4 51.1 27.0 39.5

þ0.1 0.6 þ6.5 19.4 þ6.2 18.9 þ5.6 19.0 4.1 18.9 þ5.1 20.2 þ0.9 16.2 3.9 14.7 2.6 9.4 5.6 7.7

[0.4 e þ0.7] [1.0e0.2] [19.5 e þ32.5] [37.3 to 1.5] [7.5 e þ19.8] [34.9 to 3.0] [19.5 e þ8.4] [33.2 to 4.9] [15.9 e þ7.8] [36.6 to 1.2] [8.0 e þ18.2] [35.8 to 4.6] [12.5 e þ14.3] [31.5 to 0.9] [14.1 e þ6.4] [34.4 e þ5.0] [15.4 e þ10.2] [27.6 e þ8.8] [20.2 e þ9.0] [22.4 e þ7.0]

0.64 0.003 0.61 0.034 0.37 0.021 0.43 0.009 0.50 0.037 0.44 0.012 0.89 0.038 0.46 0.14 0.68 0.31 0.45 0.30

a For each antibody response the b coefficient indicates the increase (positive sign) or the decrease (negative sign) of the mean antibody level for individuals harbouring the Gm 5,6,13,14;1,17 phenotype or the Km1 allotype in the indicated age group. The P value indicates whether the increase/decrease is significant. Example: mean level of IgG3 against NTS-DBL1a1 for children 4 years is (30.8 þ 15.2) arbitrary units (AU) and 30.8 AU respectively for children harbouring or not the Gm 5,6,13,14;1,17 phenotype. b CI95, 95 percent confidence interval of the b parameter. c In bold for significant results.


of effects have been observed in the Beninese children studied here. In the Beninese population sample under study, we previously showed (i) an imbalanced distribution of the Km1 phenotype which was mainly found among children hospitalised for severe malaria compared to children with uncomplicated malaria or with P. falciparum asymptomatic carriage and (ii) a reduced probability of developing uncomplicated malaria with increasing age for carriers of the Gm 5,6,13,14;1,17 phenotype carriers compared to non-carriers [13]. Higher prevalence and level of VarO-reactive antibodies were observed in asymptomatic than in symptomatic children, and no anti-VarO-rosetting activity was found in any case, heightening the contribution of cytophilic antibodies in the decrease of the parasite load, via the opsonization of infected erythrocytes [14]. Our present observation that Gm 5,6,13,14;1,17 and Km1 phenotypes were differently associated with acquired antibodies reacting with the VarO-infected erythrocyte surface or with VarO recombinant domains reinforces the association of this particular antibody specificity with protection against clinical malaria. Indeed, the VarO variant belongs to the VarA group of the PfEMP1 antigen, encoded by the var multigene family, and isolates expressing group A var genes are frequently associated with severe malaria in African children [17]. Our demonstration of subtle differences in the age-related evolution of anti-VarO antibodies (namely cytophilic antibodies which are impacted by the G1m and G3m allotypic polymorphism) according to Gm and Km allotypes, exemplifies the importance of host genetic factors in individual differences in both severity of clinical malaria and implemented antibody responses, at a crucial young age when specific immunity is not yet acquired. Acknowledgements We thank the participating children and their families; medical staff at the hospitals as well as school directors and teachers at the primary schools; the UMR 216 team of Cotonou, for performing field activities; and A. Massougbodji, for interceding with local authorities. Financial support was provided by the Institut de Me´decine et d’Epide´miologie Appliqueé, the Faculte´ de Me´decine Xavier-Bichat (grant 5950MIG90), the French National Research Agency (Agence Nationale de la Recherche grant MIME 021 01-02), the Institut Pasteur, the Institut de Recherche pour le De´veloppement and the Centre National de la Recherche Scientifique. References [1] M.-P. Lefranc, G. Lefranc, Molecular genetics of immunoglobulin allotype expression. in: F. Shakib (Ed.), The Human IgG Subclasses: Molecular Analysis of Structure, Function and Regulation. Pergamanon Press, Oxford, 1990, pp. 43e78. [2] J.-M. Dugoujon, S. Hazout, F. Loirat, B. Mourrieras, B. Crouau-Roy, A. Sanchez-Mazas, GM haplotype diversity of 82 populations over the world suggests a centrifugal model of human migrations, Am. J. Phys. Anthropol. 125 (2004) 175e192. [3] J.P. Pandey, M.A. French, GM phenotypes influence the concentrations of the four subclasses of immunoglobulin G in normal human serum, Hum. Immunol. 51 (1996) 99e102.

775

[4] J.P. Pandey, J.C. Gunsolley, J.G. Tew, H.A. Schenkein, GM allotypes and IgG subclass concentrations in blacks, Exp. Clin. Immunogenet. 13 (1996) 126e130. [5] S. Whittingham, D.N. Propert, Gm and Km allotypes, immune response and disease susceptibility, Monogr. Allergy 19 (1986) 52e70. [6] J.P. Pandey, L.H. Elson, S.E. Sutherland, R.H. Guderian, E. Araujo, T.B. Nutman, Immunoglobulin kappa chain allotypes (KM) in onchocerciasis, J. Clin. Invest. 96 (1995) 2732e2734. [7] M.A. Kron, B. Ramirez, V. Belizario Jr., J.P. Pandey, Immunoglobulin allotypes among the Bicolanos of Sorsogon province, Luzon, Philippines: implications of phenotypes for filariasis, Exp. Clin. Immunogenet. 16 (1999) 65e71. [8] C. Timmann, E. van der Kamp, A. Kleensang, I.R. Ko¨nig, T. Thye, D.W. Bu¨ttner, C. Hamelmann, Y. Marfo, M. Vens, N. Brattig, A. Ziegler, R.D. Horstmann, Human genetic resistance to Onchocerca volvulus: evidence for linkage to chromosome 2p from an autosome-wide scan, J. Infect. Dis. 198 (2008) 427e433. [9] J.P. Pandey, Immunoglobulin kappa chain locus on chromosome 2p12 and onchocerciasis, J. Infect. Dis. 199 (2009) 286e287. [10] C.A. Facer, Direct antiglobulin reactions in Gambian children with P. falciparum malaria. III. Expression of IgG subclass determinants and genetic markers and association with anaemia, Clin. Exp. Immunol. 41 (1980) 81e90. [11] J.P. Pandey, A. Nasr, K.M. Rocca, M. Troye-Blomberg, G. Elghazali, Significant differences in GM allotype frequencies between two sympatric tribes with markedly differential susceptibility to malaria, Parasite Immunol. 29 (2007) 267e269. [12] H.A. Giha, A. Nasr, N.C. Iriemenam, D. Arnot, M. Troye-Blomberg, T.G. Theander, K. Berzins, G. ElGhazali, J.P. Pandey, Antigen-specific influence of GM/KM allotypes on IgG isotypes and association of GM allotypes with susceptibility to Plasmodium falciparum malaria, Malar. J. 8 (2009) 306. [13] F. Migot-Nabias, J.M. Noukpo, E. Guitard, J. Doritchamou, A. Garcia, J.M. Dugoujon, Imbalacanced distribution of GM immunoglobulin allotypes according to the clinical presentation of P. falciparum malaria in Beninese children, J. Infect. Dis. 198 (2008) 1892e1895. [14] I. Vigan-Womas, A. Lokossou, M. Guillotte, A. Juillerat, G. Bentley, A. Garcia, O. Mercereau-Puijalon, F. Migot-Nabias, The humoral response to Plasmodium falciparum VarO rosetting variant and its association with protection against clinical malaria in Beninese children, Malar. J. 9 (2010) 267. [15] L.L. Field, J.-M. Dugoujon, Immunoglobulin allotyping (Gm and Km) of GAW5 families, Genet. Epidemiol. 6 (1989) 31e34. [16] M.-P. Lefranc, WHO-IUIS nomenclature subcommittee for immunoglobulins and T cell receptors report, Immunogenetics 59 (2007) 899e902. [17] I. Vigan-Womas, M. Guillotte, C. Le Scanf, S. Igonet, S. Petres, A. Juillerat, C. Badaut, F. Nato, A. Schneider, A. Lavergne, H. Contamin, A. Tall, L. Baril, G.A. Bentley, O. Mercereau-Puijalon, An in vivo and in vitro model of Plasmodium falciparum rosetting and autoagglutination mediated by varO, a group A var gene encoding a frequent serotype, Infect. Immun. 76 (2008) 5565e5580. [18] C.R. Weinberg, Toward a clearer definition of confounding, Am. J. Epidemiol. 137 (1993) 1e8. [19] M.A. Hernan, S. Hernandiez-Diaz, M.M. Werler, A.A. Mitchell, Causal knowledge as a prerequisite for confounding evaluation: an application to birth defect epidemiology, Am. J. Epidemiol. 155 (2002) 176e184. [20] J.-Y. Le Hesran, I. Personne, P. Personne, N. Fievet, B. Dubois, M. Beyeme´, C. Boudin, M. Cot, P. Deloron, Longitudinal study of Plasmodium falciparum infection and immune responses in infants with or without the sickle cell trait, Int. J. Epidemiol. 28 (1999) 793e798. [21] A. Johnson, R. Leke, L. Harun, C. Ginsberg, J. Ngogang, A. Stowers, A. Saul, I.A. Quakyi, Interaction of HLA and age on levels of antibody to Plasmodium falciparum rhoptry-associated proteins 1 and 2, Infect. Immun. 68 (2000) 2231e2236. [22] M. Torres, A. Casadevall, The immunoglobulin constant region contributes to affinity and specificity, Trends Immunol. 29 (2008) 91e97. [23] G. Moxley, R.S. Gibbs, Polymerase chain reaction-based genotyping for allotypic markers of immunoglobulin kappa shows allelic association of Km with kappa variable segment, Genomics 13 (1992) 104e108.