Although rosetting is the best-documented cytoadherence phenotype associated with severe malaria in African children, little is known on the response acquired to rosetting parasites in endemic areas. The present study in Benin confirms the elevated seroprevalence of VarO in semi-immune children and immune adults observed in a Senegalese population . Children with severe or uncomplicated malaria had a much lower anti-varO response than semi-immune asymptomatic children, consistent with the conclusion that these antibodies are associated with protection against clinical malaria in the age group that progressively mounts a protective response.
A very high prevalence of the antibodies reacting with the VarO-IE surface was observed in AP children. The mean age of the AP children was 6.5 ± 1.3 y, an age at which children living under such transmission conditions are semi-immune but are still at risk in developing clinical malaria. As such, and with the caveats of comparisons between different studies, the seroprevalence to VarO-IE in AP children from Ouidah seems higher than the response to other parasite lines, including the rosette-forming FCR3S1.2 parasites  or a single variant A4var line observed in semi-immune Kenyan children living under similar transmission conditions , or to the response to local isolates reported in Tanzanian children living in low and moderate transmission conditions . It is also higher than the response observed against a panel of local isolates in Ghanaian children living in more intense transmission conditions who had supposedly acquired earlier in life an expanded antibody repertoire . While these data require further confirmation, they are consistent with VarO being a so-called "frequent" or "prevalent" serotype [42, 43] usually associated with severe malaria, which seems to be a feature of group A var genes [23, 44–47] to which the varO gene belongs. It is not known at present whether the reaction observed with the VarO-IE surface and/or the various -varO-derived recombinant domains is strictly varO-specific or reflects a broad cross-reactivity to other "rosetting variants" some of which also belong to group A var genes [46, 48].
Prevalence rates and levels in all VarO-related assays (IE surface, total IgG, IgG1 and IgG3 to the three recombinant domains) were much lower in the children with clinical malaria than in AP children. This difference remained significant in the multivariate analysis, i.e. after correcting for age and parasite density at enrolment. It may be possible that the different exposure of children partly contribute to these findings. However, AP children were recruited in an area where transmission intensity was estimated to be lower than in the Cotonou area where symptomatic children recruited at hospital lived, although transmission in Cotonou is quite heterogeneous. Be that as it may, this difference would translate into a delayed acquisition of antibodies in Ouidah compared to Cotonou, as higher transmission intensity is clearly associated with a more rapid acquisition of an expanded antibody repertoire [23–25]. To further document the association of VarO-reacting antibodies with protection against clinical malaria, a longitudinal follow-up of children is needed to show that the presence of such antibodies prevents disease caused by P. falciparum parasites expressing this serotype or cross-reacting serotypes.
Of the three recombinant domains studied here, NTS-DBL1α1 had the highest seroprevalence and the highest antibody levels. In view of the known mosaic structure of the var genes, this probably indicates that parasites expressing the varO-NTS-DBL1α1 domain or a related cross-reacting domain may not co-express varO-CIDRγ-like or varO-DBL2βC2-like domains. Each domain elicits antibody in the context of where it is presented, i.e., not necessarily associated with the same domains as those in the varO gene. This might account for the limited correlation of the response against the individual domains in the children with clinical malaria. Although it is tempting to speculate that the higher seroreactivity to varO-NTS-DBL1α1 may reflect the generally high conservation of DBL1α1 sequences across the var repertoire and between isolates compared to CIDRγ, for example, which are quite diverse, further study is needed to test this hypothesis. Furthermore, comparison of reactivity between different antigens must be interpreted with caution because detection depends on the sensitivity of the assay, which may vary from one to the other, and because it is difficult to compare arbitrary units. Notwithstanding these reservations, the observation of a higher response to NTS-DBL1α1 is consistent with recent reports on the related R29-DBL1α1  and the responses to individual domains of the A4var gene , although this was not observed in serological surveys using an array of domains from a number of var genes [23, 24].
VarO-IE surface reactivity correlated best with the anti-NTS-DBL1α1 IgG, and less strongly with the other domains. The recombinant NTS-DBL1α1 domain mediates rosetting, but a very small percentage of adults from Benin disrupted more than 50% of the VarO rosettes, differing in this regard from immune Senegalese adults who consistently displayed high VarO rosettes disrupting capacity . None of the children sera disrupted VarO rosettes, including sera from AP children. This confirms previous observations with Senegalese asymptomatic children of the same age range, although most had VarO-IE surface-reactive antibodies . The disconnection of rosette-disrupting antibodies with surface-reacting antibodies observed here in Benin has been reported in studies conducted in Kenya and Gabon with other rosette-forming parasites . This suggests that if recognition of the IE-surface participates in protection against clinical malaria like anti-PfEMP1var2csa antibodies against placental malaria [8, 16], other mechanisms than prevention/reversion of cytoadherence are brought about against rosette-forming parasites, including possibly complement-mediated lysis or phagocytosis of the IE. In the Saimiri sciureus monkey model, cytophilic antibodies targeting the IE surface and promoting IE phagocytosis of Palo Alto varO parasites were associated with protection against experimental blood stage challenge and protect animals when passively administered [33, 34]. In this context, it is interesting to note that indeed a significant IgG1 and IgG3 response to each of the three recombinant varO domains could be documented in the Beninese children. It is thus possible that cytophilic antibodies to the IE surface contribute to parasite clearance in individuals that have not (yet) acquired rosette-disrupting antibodies. Both likely contribute to protection, but their acquisition may be sequential and/or depend on endemicity and transmission intensity.
SM and UM children presented similar responses in all serological assays used here. No changes in the conclusions of the analysis were brought when considering the sub-group of CM children. Based on previous studies in Gambian  or Gabonese children  such a difference might have been anticipated. The intensity of surface reaction with VarO-IE tended to be lower in the SM than in UM Beninese children studied here, but this difference did not reach significance. The high parasite density in SM children may be associated with capture of antibodies onto the parasite antigens and could account for the observed lack of differences between SM and UM children. Because of the small volume of plasma available, the isotype of surface reacting antibodies and the IgG2 and IgG4 antibodies to the recombinant domains were not measured. Therefore, the possibility exist that SM and UM children differ with regard of these isotypes. IgG3 antibodies to surface variant antigens have been reported , although other isotypes are produced as well . In one study, IgG4 responses were reported as contributing to protection . Further studies are needed to clarify this issue and evaluate the respective role of anti-surface, anti-rosetting and antibody isotype in the anti-varO response and their potential contribution to protection against severe or uncomplicated malaria.
In asymptomatic carriers, the frequency of some antibody responses decreased between 2006 and 2008, while anti-NTS-DBL1α1 IgG1 and anti-CIDRγ IgG did not. The level of antibodies to the recombinant antigens did not drop. This might reflect the short-lived response and/or the need of sustained asymptomatic infection documented in other settings [52, 55]. The children identified as P. falciparum asymptomatic carriers in 2006, were parasite-free when recruited in January 2008, except for six children. It is unknown whether these parasite-free children had been and for how long, asymptomatic carriers during the 13 months elapsed between the blood samplings. The 2006-2008 period matched precisely with a large-scale distribution of long lasting insecticidal nets  to vulnerable populations (pregnant women and children under 5 years old) in most towns and villages in the south of Benin. This scaling-up of control measures has very likely reduced transmission - a sign of this impact might be the very low rate of asymptomatic infections in the children recruited in 2008. A reduced circulation of P. falciparum parasites in the field, including VarO or VarO-related parasites, resulting in reducing asymptomatic carriage and its consequences on maintenance of immune responses during this period is therefore plausible.