Centre for Clinical Vaccinology and Tropical Medicine and Wellcome Centre for Human Genetics, Churchill Hospital, University of Oxford, Oxford, United Kingdom
We calculated the number and growth rate of Plasmodium falciparum parasites emerging in recipients of candidate preerythrocytic malaria vaccines and unvaccinated control subjects undergoing mosquito-bite challenge. This was done to measure vaccine efficacy and to distinguish the effects on blood-stage multiplication from those on liver-stage parasites. Real-time polymerase chain reaction measurements of parasite densities were analyzed by nonlinear regression and mixed-effects models. Substantial reductions in numbers of liver parasites resulted from the use of 2 immunization regimens: FP9 boosted by modified virus Ankara (MVA) encoding the malaria epitopethrombospondinrelated adhesion protein insert (92% reduction) and RTS,S/AS02 used in heterologous prime-boost immunization regimens, with MVA encoding the circumsporozoite protein (97% reduction). Forty-eighthour growth rates in blood from control subjects were not different from those in blood from any vaccination group (mean, 14.4-fold [95% confidence interval, 1119-fold]).
An effective malaria vaccine is urgently needed, and several candidate vaccines are in clinical development [1]. Sporozoite challenge of human volunteers can assess the efficacy of preerythrocytic vaccines before large field trials [2]. Volunteers were challenged with bites by Plasmodium falciparuminfected mosquitoes, monitored by use of regular blood-film examinations, and treated after the first positive blood film. Parasitemia-free intervals after sporozoite inoculation were analyzed as a survival function. A longer parasitemia-free interval for a vaccine group has been considered to represent a reduced inoculum of parasites seeding the blood from the liver [3, 4]. Polymerase chain reaction (PCR) detects lower numbers of parasites than does blood-film examination; blood-film examination detects 20,00050,000 parasites/mL, and PCR detects 20 parasites/mL (L.A., R.F.A., D.W., S.D., M.W., P.B., A. Hunt-Cooke, G. Bergson, F. Sanderson, A.V.S.H., and S.C.G., unpublished data). Analysis of these data allowed us to test whether vaccine-induced immunity reduced either numbers of parasites emerging from the liver or subsequent growth rates.
Of the immunization regimens described here, 5 included the preerythrocytic antigen malaria epitopethrombospondinrelated adhesion protein (ME-TRAP), encoded by 3 vectors: FP9 (an attenuated strain of fowlpox), modified vaccinia virus Ankara (MVA), or plasmid DNA. A second preerythrocytic antigen, the circumsporozoite protein, was delivered by recombinant MVA and RTS,S/AS02 (a particulate vaccine fusing most of the circumsporozoite protein to hepatitis B surface antigen) with a proprietary adjuvant [5]. The particulate Apovia vaccine, ICC-1132, comprises hepatitis B core antigen containing T and B cell epitopes from the circumsporozoite protein with Montanide ISA720 adjuvant [6, 7].
Heterologous prime-boost regimens including DNA and MVA encoding ME-TRAP were associated with significant delays in time to parasitemia, as determined by blood-film examination, but predate quantitative PCR [4]. A similar regimen (2 sequential DNA immunizations followed by MVA and then FP9; DDMF), among others, has been assessed here. Two FP9 immunizations followed by MVA (FFM) and 2 prime-boost regimens including RTS,S/AS02A and MVA encoding the circumsporozoite protein protected some volunteers and significantly delayed parasitemia in others (D.W., S.D., J. Vuola, S. McConkey, I. Poulton, L. Andrews, L.A., R. Ebensen, T. Berthoud, S. Keating, P.B., G. Butcher, R. Sinden, S.G.C., and A.V.S.H., unpublished data, and S.D., M.W., J. Vuola, D.W., S. Keating, T. Berthoud, L.A., P.B., I. Poulton, G. Butcher, K. Watkins, R. Sinden, A. Leach, P. Moris, N. Tornieporth, J. Schneider, F. Dubovsky, E. Tierney, J. Williams, D. Heppner, S.C.G., J. Cohen, and A.V.S.H., unpublished data). Other ME-TRAPbased regimens2 DNA immunizations followed by FP9 and then MVA (DDFM), a single FP9 immunization before MVA (FM), and MVA before FP9 (MF)did not delay the onset of parasitemia, as determined by blood-film examination (D.W., S.D., J. Vuola, S. McConkey, I. Poulton, L. Andrews, L.A., R. Ebensen, T. Berthoud, S. Keating, P.B., G. Butcher, R. Sinden, S.G.C., and A.V.S.H., unpublished data). The protection data available from studies of blood film data are summarized in table 1.
Parasites first emerge from the liver 6.5 days after inoculation of sporozoites by mosquitoes [8, 9]. Numbers of parasites in peripheral blood then depend on multiplication rates and sequestration of parasites onto blood-vessel walls. Parasites sequester during the second half of their 48-h life cycle; the progeny of sequestered parasites then appear in peripheral blood until they, too, sequester. If parasites were in perfect synchrony, they would be absent from peripheral blood on alternate days. However, partial synchrony is usual, with marked periodic fluctuations in parasitemia. Fluctuations in blood-film parasitemia, measured in P. falciparumtreated patients with neurosyphilis, have been modeled by use of the sine wave function, for sequestration, and a logarithmic function, for growth [10]. Another approach splits parasites into broods, depending on the time of sampling, and has been applied to the study of volunteers receiving identical blood-stage inocula, to calculate growth rates [11]. We adapted both of these models to determine growth rates and liver-to-blood inocula and, thereby, the protective efficacy resulting from various immunization strategies.
SUBJECTS AND METHODS
Challenge
Vaccinated and unvaccinated volunteers underwent challenge with bites by 5 P. falciparum (strain 3D7)infected Anopheles stephensii mosquitoes. Twice-daily blood-film examination and PCR measurements were performed from day 6 until day 14 and then once daily until day 21. These vaccination and challenge trials are reported in full elsewhere [4, 6] (D.W., S.D., J. Vuola, S. McConkey, I. Poulton, L. Andrews, L.A., R. Ebensen, T. Berthoud, S. Keating, P.B., G. Butcher, R. Sinden, S.G.C., and A.V.S.H., unpublished data, and S.D., M.W., J. Vuola, D.W., S. Keating, T. Berthoud, L.A., P.B., I. Poulton, G. Butcher, K. Watkins, R. Sinden, A. Leach, P. Moris, N. Tornieporth, J. Schneider, F. Dubovsky, E. Tierney, J. Williams, D. Heppner, S.C.G., J. Cohen, and A.V.S.H., unpublished data). Visualization of a single parasite, by blood-film examination, prompted immediate treatment with standard doses of chloroquine over the course of 3 days and defined the study end point. Volunteers were considered to be fully protected if they were still aparasitemic at day 21. During the trial, neither the managing clinicians nor the microscopists were aware of PCR data. The PCR method has been described elsewhere (L.A., R.F.A., D.W., S.D., M.W., P.B., A. Hunt-Cooke, G. Bergson, F. Sanderson, A.V.S.H., and S.C.G., unpublished data). Briefly, EDTA anticoagulated blood samples were filtered to remove leukocytes, and DNA was purified from 0.5 mL of filtered blood and was eluted into 50 L. A portion of the multicopy 18S (small-subunit) ribosomal RNA genes of P. falciparum was amplified by PCR, and the increase in PCR product was detected by binding the fluorescent dye SYBR green, by use of the Rotor-Gene Real-Time PCR machine (Corbett Research), with 5 L of extracted DNA in duplicate. All volunteers gave informed consent for immunizations, challenge studies, and blood sampling. Procedures were reviewed by Oxford Research Ethics Committee, the local ethics committee.
Modeling
Individual modeling using the sine wave function.
Volunteers with <6 PCR readings and those completely protected were excluded from the analysis. Of 80 volunteers, 6 were completely protected and 6 had data from <6 time points. Data for the logarithm of each PCR result, described by time of sampling as a continuous variable and by volunteer, were analyzed by use of Stata 8 (Timberlake). Nonlinear least-squares regression, for determination of best fit, was used: log(P) = tm + a + [c × sin(t + k)], where P is the number of parasites per milliliter, t is time minus 6.5 days after sporozoite inoculation, m (gradient) is the logarithm of the daily multiplication rate, a (intercept) is the logarithm of the starting number of parasites per milliliter of peripheral blood, c (sine wave amplitude) is the logarithm of the number of sequestered parasites, and k (phase shift) is the time between day 7 and the first peak. Hence, 10(c+a) approximated the first peak of parasites per milliliter as the number of parasites emerging from the liver after day 6.5. A 5-L volume of blood was assumed to calculate total parasite numbers; 95% confidence intervals (CIs) were calculated by use of the root mean square of the SE of each parameter, which were derived by asymptotic approximation. Day-6.5 readings were excluded, since the change between day 6.5 and day 7 was generated by parasites emerging from the liver rather than by growth or sequestration. Samples obtained before the first detection of parasites were excluded from analysis. In cases in which parasites were not detected but had been detected at previous time points, the lower limit of parasite detection was used (20 parasites/mL).
We assessed methods for estimating the liver burden by use of these parameters, using a step-function simulation of parasites emerging from the liver, with a 10-fold multiplication rate in each 48-h life cycle and 24-h patent, nonsequestered period. We simulated 12-hourly sampling over the course of 4 days and found 10(c+a) to be an accurate approximation of total liver burden, with an average error of 10%.
Group modeling using the sine function.
A mixed-effects model applied separately to each group allowed volunteers with <4 PCR readings to be included in the analysis. Mixed-effects modeling was used for groups with 8 volunteers. A group mean and each volunteer's deviation from the mean were calculated, under the assumption of normal distributions for the group. Parameters m, c, a, and k are analogous to those in individual fits, where m0 represents the mean value for the group, and mi represents the separate factor for each volunteer: log(P) = t(m0 + mi) + a0 + ai + (c0 + ci) × sin(t + k0 + ki). The mixed-effects modeling methodology has been described elsewhere [12]. Extensions of this model allowed us to examine periodicity for the overall group, log(P) = t(m0 + mi) + a0 + ai + (c0 + ci) × sin[t(p + p0) + k0 + ki], and change in growth over time, log(P) = t(m0 + mi)(v0 + vi) + a0 + ai + (c0 + ci) × sin[t(p + p0) + k0 + ki], where p is the time between peaks, and v is the change in growth rate over time.
A Bayesian Markov Chain Monte Carlo method (WinBUGS; version 1.4; MRC Biostatistics Unit and Imperial College, London) was used. Fully protected volunteers were excluded, and day-6.5 values and undetectable values were treated as above.
Group modeling using divided parasite populations.
An alternative model links the peaks, troughs, and midpoints of the oscillations (under the assumption of 48-h periodicity) by use of a logarithmic growth function. The volunteers in these trials underwent 4 blood tests during each 48-h period and therefore were considered to be infected by 4 partially overlapping populations of parasites. For example, parasites observed in peripheral blood on day 9 are the progeny of those observed on day 7. Data for only a few volunteers were from a sufficient number of time points to allow modeling of 4 different parasite broods, but mixed-effects modeling allowed analysis of a group of volunteers, each with 4 broods. Normal distributions were assumed for parameters, and WinBUGS (version 1.4) was used for analysis.
Statistics
Liver-to-blood inocula and growth rates for volunteers were compared with those from unvaccinated volunteers, by use of a nonparametric test (Stata 8). Kruskal-Wallis nonparametric testing was used to compare vaccination regimens within the individual modeling.
RESULTS
Individual modeling using the sine wave function.
The numbers of parasites emerging from the liver are displayed by volunteer and are grouped into different vaccination groups (figure 3). There were no significant differences between the control group and the groups that received ICC-1132, DDFM, FM, or MF. DDMF was associated with an 86% decrease in liver burden (P = .074). The groups that received FFM or RRM/MRR had reductions in liver-to-blood inocula of 92% and 97%, respectively, compared with that in the control group (P = .004 and P = .001, Kruskal-Wallis). Growth rates were similar between groups (figure 4), with an average increase of 14.4-fold (95% CI, 1119-fold) per 48 h. For sequestration of parasites or time to parasites emerging from the liver, there were no differences between groups. Estimates of the reduction in numbers of liver-stage parasites corresponded with estimates of vaccine efficacy by use of blood film data. Correlating estimates of liver burden with time to blood-film patency resulted in r = -0.552 (P < .0005). No correlation was seen with growth rate.
Group modeling using the sine function.
A group model allowed inclusion of volunteers whose data were from too few time points for individual modeling. The reduction in parasite load was 89% for the group that received RRM/MRR and 82% for the group that received FFM (figure 5). By weighted means, including completely protected volunteers, the reductions were 93% for the group that received RRM/MRR and 85% for the group that received FFM. The mean 48-h growth rate was 11.5-fold (95% CI, 6.620.9-fold). Forty-eighthour growth rates were not significantly different by vaccination group: control, 12.3-fold (95% CI, 8.316.6-fold); ICC-1132, 23-fold (95% CI, 1160-fold); FFM, 9.6-fold (95% CI, 823-fold); and RRM/MRR, 15.8-fold (95% CI, 8.730-fold). In models examining periodicity, a cycle of 2.2 days was found (95% CI, 1.52.9 days). There was no change in growth rate over time (0.97-fold/day; 95% CI, 0.71.3-fold/day).
Group modeling using divided parasite populations.
By assuming a 48-h life cycle, it was possible to fit a logarithmic growth pattern to observations taken 48 h apart for each volunteer. A mixed-effects model for 4 different broods/volunteer showed reductions in parasite load of 95% for the group that received RRM/MRR and 86% for the group that received FFM (figure 5). The overall growth rate was 11.5-fold (95% CI, 5.822.9-fold). By-group growth rates were as follows: control, 10.5-fold (95% CI, 4.723-fold); ICC-1132, 23-fold (95% CI, 860-fold); FFM, 10.5-fold (95% CI, 5.220-fold); and RRM/MRR, 15.9-fold (7.931-fold).
DISCUSSION
Estimation of 4 parameters by use of an average of only 10 (range, 616) time points/volunteer produced large SEs (figures 3 and 4). Parameters were fitted as log values, amplifying SEs when described in linear terms. This contributed to the large range of values returned. Nevertheless, the efficacy of immunization with FFM and RRM/MRR is clear and attributable to a reduction in parasites emerging from the liver rather than to reduced blood-stage growth rates. The mixed-effects models (models 2 and 3) produced equally large SEs. We used group models to avoid bias resulting from excluding from individual modeling those volunteers whose data were from few points and did not select volunteers whose data best fit the model, to reduce SEs, as is done occasionally [10]. Results were similar to those of the individual model using the sine wave function (figure 5); we therefore emphasize the more parsimonious, individual model.
Morphological studies suggest that 20,000 merozoites are released by a single schizont [13, 14] and that 11 sporozoites are injected by each mosquito [15]; our challenge model used 5 mosquitoes. Therefore, if every sporozoite reaches a hepatocyte, 55 sporozoites would, theoretically, produce >106 merozoites, which is 3 times the PCR-based estimate. However, the average unvaccinated subject was found to be positive by blood-film examination on day 11. Under the assumption that parasites emerged from the liver on day 6.5, grew in a 5-L volume of blood at a rate of 10-fold per 48 h, and were detected by blood-film microscopy at 10 parasites/L, the initial inoculum from the liver to blood would be 280,000 merozoites, which is very similar to the PCR-based estimate (figure 6). This suggests that only a fraction of sporozoites successfully invade liver cells.
The average liver-to-blood inoculum in the RRM/MRR group was in the range of 10,00030,000 merozoites, suggesting that these volunteers had only 1 infected hepatocyte. Volunteers clearing 13 of 14 hepatocytes might have had immunity similar to that in volunteers clearing all 14, perhaps explaining difficulties in correlating individual immunity with protection [16]. However, there are several caveats to this calculation: (1) estimates of nascent merozoites within a single schizont are not precise and do not allow for interschizont variation in vivo, (2) only a fraction of the 20,000 merozoites will invade red blood cells, and (3) it is possible that schizonts may be damaged but not killed by vaccine effects, resulting in release of reduced numbers of parasites. Delay in schizonts releasing merozoites rather than reduction in numbers of released merozoites would also reduce the intercept. However, there was no difference in the timing of the first peak (as determined by modeling) between the group that received FFM and the control group (P = .61) or between the group that received RTS,S and the control group (P = .75), and, similarly, there was no difference in the timing of the first PCR-positive result between the groups (P = .45 and P = .52, respectively). It is therefore unlikely that, in vaccinees, these results reflect prolonged hepatic schizogony rather than killing of parasites.
We considered that the random effects of sampling of low numbers of parasites present in the volume of blood drawn and of gene copies in the DNA eluted might be a source of variability in our data. Estimates from the Poisson distribution, the volume of blood sampled, and the fraction of DNA analyzed suggest that parasites were reliably sampled (i.e., with a probability of detection >95%) at >10 parasites/mL and that quantitation was accurate to within 20% at 250 parasites/mL.
Growth rates here were slightly higher than the 8-fold that was previously calculated by use of blood-film data [10] but were similar to the higher growth rates estimated by use of PCR (range, 1225-fold) after inoculation of low numbers of blood-stage parasites [11, 17]. PCR examines lower densities of parasites, suggesting that, at lower densities, growth may be more rapid. However, in a mixed-effects model, variation in growth rates over time did not occur during the monitoring period. It is possible that growth begins to slow at higher densities.
The protective vaccine regimens assessed here have been shown to act by reducing liver parasite load rather than by altering growth patterns or the time of emergence of parasites from the liver. This is an important end point of phase 2a malaria vaccine trials and complements data obtained by the traditional blood-film microscopy end point. The substantial effect quantified here on parasite liver burden of each of the 2 partially protective vaccination regimens (FFM and RRM/MRR) suggests that relatively minor improvements of either of these preerythrocytic vaccination approaches could lead to a major increase in sterile protection rates.
References
1. Moorthy V, Hill AV. Malaria vaccines. Br Med Bull 2002; 62:5972. First citation in article
2. Herrington DA, Clyde DF, Murphy JR, et al. A model for Plasmodium falciparum sporozoite challenge and very early therapy of parasitaemia for efficacy studies of sporozoite vaccines. Trop Geogr Med 1988; 40:1247. First citation in article
3. Witney AA, Doolan DL, Anthony RM, Weiss WR, Hoffman SL, Carucci DJ. Determining liver stage parasite burden by real time quantitative PCR as a method for evaluating pre-erythrocytic malaria vaccine efficacy. Mol Biochem Parasitol 2001; 118:23345. First citation in article
4. McConkey SJ, Moorthy V, Webster D, et al. Enhanced T-cell immunogenicity of plasmid DNA vaccines boosted by recombinant modified vaccinia virus Ankara in humans. Nat Med 2003; 9:72935. First citation in article
5. Stoute JA, Slaoui M, Heppner DG, et al. A preliminary evaluation of a recombinant circumsporozoite protein vaccine against Plasmodium falciparum malaria. RTS,S Malaria Vaccine Evaluation Group. N Engl J Med 1997; 336:8691. First citation in article
6. Walther MD, Keating S, Vuola S, et al. Safety, immunogenicity and efficacy of a pre-erythrocytic malaria candidate vaccine, ICC-1132, formulated in Seppic ISA 720. Vaccine (in press). First citation in article
7. Birkett A, Lyons K, Schmidt A, et al. A modified hepatitis B virus core particle containing multiple epitopes of the Plasmodium falciparum circumsporozoite protein provides a highly immunogenic malaria vaccine in preclinical analyses in rodent and primate hosts. Infect Immun 2002; 70:686070. First citation in article
8. Daubersies P, Thomas AW, Millet P, et al. Protection against Plasmodium falciparum malaria in chimpanzees by immunization with the conserved pre-erythrocytic liver-stage antigen 3. Nat Med 2000; 6:125863. First citation in article
9. Hermsen CC, Telgt DS, Linders EH, et al. Detection of Plasmodium falciparum malaria parasites in vivo by real-time quantitative PCR. Mol Biochem Parasitol 2001; 118:24751. First citation in article
10. Simpson JA, Aarons L, Collins WE, Jeffery GM, White NJ. Population dynamics of untreated Plasmodium falciparum malaria within the adult human host during the expansion phase of the infection. Parasitology 2002; 124:24763. First citation in article
11. Cheng Q, Lawrence G, Reed C, et al. Measurement of Plasmodium falciparum growth rates in vivo: a test of malaria vaccines. Am J Trop Med Hyg 1997; 57:495500. First citation in article
12. Burton P, Gurrin L, Sly P. Extending the simple linear regression model to account for correlated responses: an introduction to generalized estimating equations and multi-level mixed modelling. Stat Med 1998; 17:126191. First citation in article
13. Meis JF, Rijntjes PJ, Verhave JP, et al. Fine structure of the malaria parasite Plasmodium falicparum in human hepatocytes in vitro. Cell Tissue Res 1986; 244:34550. First citation in article
14. Meis JF, Verhave JP, Meuwissen JH, Jap PH, Princen HM, Yap SH. Fine structure of Plasmodium berghei exoerythrocytic forms in cultured primary rat hepatocytes. Cell Biol Int Rep 1984; 8:75565. First citation in article
15. Beier JC, Beier MS, Vaughan JA, Pumpuni CB, Davis JR, Noden BH. Sporozoite transmission by Anopheles freeborni and Anopheles gambiae experimentally infected with Plasmodium falciparum. J Am Mosq Control Assoc 1992; 8:4048. First citation in article
16. Stoute JA, Kester KE, Krzych U, et al. Long-term efficacy and immune responses following immunization with the RTS,S malaria vaccine. J Infect Dis 1998; 178:113944. First citation in article
17. Pombo DJ, Lawrence G, Hirunpetcharat C, et al. Immunity to malaria after administration of ultra-low doses of red cells infected with Plasmodium falciparum. Lancet 2002; 360:6107. First citation in article
Identification of Multiple Chondroitin Sulfate A (CSA)Binding Domains in the var2CSA Gene Transcribed in CSA-Binding Parasites
A Combination of Antigenic Regions of Toxoplasma gondii Microneme Proteins Induces Protective Immunity against Oral Infection with Parasite Cysts
Are Iron-Scavenging Parasites Protective against Malaria
Malaria-specific transgenic CD4+ T cells protect immunodeficient mice from lethal infection and demonstrate requirement for a protective threshold of antibody production for parasite clearance
Receptor-binding residues lie in central regions of Duffy-binding–like domains involved in red cell invasion and cytoadherence by malaria parasites
Disruption of Plasmodium berghei merozoite surface protein 7 gene modulates parasite growth in vivo
Distribution of Varicella-Zoster Virus DNA and Gene Products in Tissues of a First-Trimester Varicella-Infected Fetus
Coronaviruses in the Limelight
Female-to-Male Infectivity of HIV-1 among Circumcised and Uncircumcised Kenyan Men
Antimicrobial-Resistant Nontyphoidal Salmonella Is Associated with Excess Bloodstream Infections and Hospitalizations