The volunteer (YN, an author) was a 35-year old Japanese man. He had previously experienced Plasmodium infection only once, when he was infected with P. vivax, one year before the present study 21. Control subjects were composed of 21 healthy adult Bangkok residents (9 male and 12 female, median age: 33 years) who had never been infected with any Plasmodium species.

In January 2001, 2 ml of whole blood was obtained from an adult patient who was infected with P. vivax near Thailand-Myanmar border. PCR confirmed that this blood was infected with P. vivax but not with other Plasmodium species 22. Fifty female anopheline mosquitoes were fed on this blood in a membrane feeder. One month later, sporozoites were detected in the salivary glands of two randomly selected mosquitoes.

The volunteer was challenged by another infected mosquito on day 0. The mosquito was tested by PCR, confirming the presence of P. vivax only. The volunteer was continuously monitored in Mahidol University Hospital for Tropical Diseases in Bangkok, a non-malarious region. The volunteer experienced microscopically positive parasitaemias of P. vivax from day 14. From day 20, a regimen of artemisinin (200 mg/day for 3 days) and doxycycline (100 mg/day for 6 days) was administered to radically terminate the blood-stage parasites. The blood collection was resumed on day 29, when the anti-malarial drugs should have been completely eliminated from the volunteer's blood 2324. Subsequently YN relapsed, and parasites were detected in the peripheral blood microscopically on day 36. After experiencing an acute phase, he was treated with chloroquine (1,500 mg on day 52; 500 mg on day 53 and 54), and with primaquine (15 mg per day for 14 days) to radically kill both the blood-stage parasites and the hypnozoites. The illness, especially headache, was subjectively much milder than in the previous infection one year before; nevertheless, the peak parasitaemia was three-times higher than the previous one.

Venous blood was collected every day as follows: (i) the pre-challenge period (days -6 to 0); (ii) the initial infection period (days 1 to 20); (iii) the relapse period (days 29 to 53). Blood was collected 7, 29 and 29 times during the periods (i), (ii) and (iii), respectively. The sera collected during the period (i) provided the baseline. The volunteer was phlebotomized once per day when the body temperature did not show a noticeable elevation; otherwise twice per day (one at a peak of body temperature and one at about 12 hours after the peak). Blood was allowed to coagulate at 25°C for 20 minutes, and centrifuged. Serum was preserved at -70°C until use. No drugs were administered from day -20 and throughout the entire period of the infection, except for the above medications.

Plasmodium falciparum was cultured in vitro, in each of the volunteer's sera, as follows. K1 strain of P. falciparum was synchronized into the ring form stage 25, and subsequently mixed with the incomplete medium and the purified type-O red blood cells, so that the P. falciparum parasitaemia was set to 0.5% and the haematocrit was adjusted to 50%. In the each well on a 96-well flat-bottomed plate, 10 μl of this parasite preparation, 45 μl of the incomplete medium, and 35 μl of each of the volunteer's sera were mixed (final volume 90 μl; hematcrit 5.6%; serum concentration 39% (v/v)). The plate was incubated under 5% CO₂pressure at 37°C. After 24 hours, 60 μl of the supernatant was replaced by 35 μl of the the same serum and 45 μl of incomplete medium. After a further 28 hours, two thin films were prepared from each well. The percent reduction from the baseline was defined as the growth inhibition of the each serum. As a result, the volunteer's sera collected during the relapse period exhibited growth inhibitory effects, significantly elevated above the baseline level (Figure 1c). This was particularly marked between days 46 and 51 (mean: 42%, maximum: 68%), and to a lesser degree, between days 33 and 35.

To investigate the degree of cross-reactive immunity in the sera, we measured IgM and IgG levels against P. falciparum schizont extract and Merozoite Surface Protein (MSP)-1₁₉ of P. falciparum. ELISA was carried out as previously described 26. P. falciparum schizont extract and MSP-1₁₉(Wellcome sequence) were obtained as described in 262728. Each IgG plate contained a standard dilution series of a laboratory reference hyper-immune adult Gambian plasma pool. For IgM, a standard dilution series of non-specific IgM purified from human serum (Sigma-Aldrich) was coated in one set of wells. Of note, IgM values against P. falciparum schizont extract and MSP-1₁₉ were elevated after day 47 (Figure 2a,b). The temporal correlation between pairs of time-series variables, such as growth inhibitory effect and antibody level, were quantified as follows. First, simple linear regression was applied. Next, since the blood was sampled at non-random time points and temporally non-independent, we also employed Prais-Winsten's method for time-series regression 29. In doing so, the time-axis was divided into discrete time-steps of a fixed length (half day, one day and two days). When a time-step included more than one sample, the averaged value was used. As a result, the in vitro growth inhibitory effects on P. falciparum and the levels of IgM against P. falciparum (especially, schizont extract) showed strong temporal correlation, in any time-step lengths examined (Table 1).

Interferon (INF) -γ, Interleukin (IL) -1β, IL-2, IL-4, IL-6, IL-12p40, TNF-α were measured, at least in duplicate, using commercial ELISA kits (Diaclone, Besancon). TNF-α showed highly variable values among multiple measurements (data not shown). Among the cytokines examined, IL-12 showed the highest correlation with P. vivax parasitaemia (compare Figure 3a vs Figure 1a). When IL-12 was regressed against the parasitaemia, Prais-Winsten's R² at half-day time-steps was 0.62 (P < 0.0001). This suggested that IL-12 is closely associated with schizont rupture. IFN-γ was extremely spiky and appeared only in the first few paroxysms (defined as a body temperature above 37.5°C), both in initial and relapse periods (Figure 3b). Other cytokines showed weaker association with paroxysm or parasitaemia (Figure 3c,d,e).