A cohort of suspected dengue cases admitted to three hospitals in the city of Recife in the state of Pernambuco, Brazil, has been established and described elsewhere ^[17]. Sequential blood samples collected during the first 30 days after the onset of symptoms were harvested and cryopreserved to supply serum and plasma samples at different stages of the infection. Plasma was collected in both heparin and citrate buffers and used according to the assay requirements. Dengue diagnosis was performed by combining virus isolation in C6/36 cells ^[18] and RT-PCR ^[19] with anti-dengue IgM/IgG ELISA. The classification of primary vs. secondary dengue infection was based on the kinetics of the IgM and IgG response, according to the following criteria: Primary infection was defined as the absence of specific anti-dengue IgG antibodies in the first serum samples during the acute phase, with IgM, virus, and/or viral RNA being detected and followed by the presence of anti-dengue IgG in convalescence serum samples. Secondary infection was defined as the detection of specific anti-dengue IgG in the first acute sample and an absence of anti-dengue IgM, associated with a positive RT-PCR and/or virus isolation ^[17], ^[20].

Dengue cases were clinically classified according to the World Health Organization (WHO) criteria into DF and DHF. DF cases were characterized by fever lasting up to 7 days and accompanied by at least two of the following symptoms: headache, retro-orbital pain, myalgia, arthralgia, and rash associated with a platelet level above 100,000/mm³. DHF cases were defined as having the same clinical manifestations as those in DF, but with evidence of hemorrhage, thrombocytopenia (platelet<100,000/mm³), and plasma leakage following deffervescence.

Since patient age, date of sample collection, and virus serotype and genotype are factors that can affect complement levels and potentially produce misleading results, we applied the following criteria in selecting patients: a) confirmed infection with the DENV-3 serotype (genotype III), b) age> = 13 years, and c) availability of a blood sample collected within 3 to 8 days of the onset of symptoms and defined as acute phase. A total of 74 dengue cases met the criteria: 47 patients with DF and 27 with DHF (grades I and II). In the case of the DF patients, 41% of the infections were primary and 59% secondary; in the case of the DHF patients, 58% were primary and 42% secondary (Table 1). A second set of samples from the same individuals collected 1 to 9 months after the onset of symptoms was used to represent physiologic levels after infection (termed “convalescent samples” in this report). Of the 74 cases identified, 48 (32 DF and 16 DHF patients) had convalescent samples available. The data presented in table 1 do not include all the information obtained from each patient and are not intended to define the clinical classification. The platelet data presented in the table give the count on the day of the sample used in this study and not the lowest count found in that particular patient. All DHF patients had thrombocytopenia at some time point. As a normal reference, we used samples from 30 healthy individuals, who were identified as eligible to receive yellow fever vaccine. The samples used here were collected from these individuals before immunization.

Written consent to participate in the study was obtained from each patient (or the patient's guardian) after a full explanation of the study was provided. All data were handled confidentially and anonymously. This study was reviewed and approved by the ethics committee of the Brazilian Ministry of Health (CONEP: 4909; Process n° 25000.119007/2002-03; CEP: 68/02). In addition, the Johns Hopkins University Institutional Review Board reviewed and approved the study as protocol JHM-IRB-3: 03-08-27-01.

Purified human complement protein C1q (Quidel), a chicken polyclonal antiserum against human C1q (Abcam), and HRP-conjugated sheep anti-human C1q (AbD Serotec) were used to measure C1q in plasma samples. A standard curve was established using a nonlinear regression curve (R²>0.95) with seven standard concentrations; the range of detection was 0.0001–10.0 µg/mL. Purified human factor H (Calbiochem), a goat polyclonal antiserum against human factor H (Calbiochem), a mouse monoclonal antiserum [HYB 268] against human factor H (Abcam), and HRP-conjugated goat anti-mouse IgG (Jackson Immunoresearch) were used to measure factor H. A standard curve was established using a linear regression curve (R²>0.95) with five standard concentrations; the range of detection was 3–25 ng/mL. Purified human complement protein C3 (Quidel), a chicken polyclonal antiserum against human complement C3 (Genway), rabbit polyclonal antiserum against human C3 biotinylated (Abcam), and HRP-conjugated streptavidin (Rockland) were used to measure C3. A standard curve was established using a nonlinear regression curve (R²>0.95) with seven standard concentrations; the range of detection was 0.0001–10.0 µg/mL. The plasma samples were diluted so that the values obtained fell within the standard curves.

Nunc microplates were coated with primary antibody and diluted in carbonate/bicarbonate buffer, pH 9.6. After washing with PBS (pH 7.2) containing 0.05% Tween 20 (PBS-T), the plates were blocked with 5% BSA (Sigma) for 1 h at 37°C. They were then washed again and incubated for 2 h at 37°C with either the plasma samples from dengue patients or with purified protein, in duplicate. The plates were again washed and incubated for 1 h at 37°C with secondary antibody. When a tertiary step was necessary, the plates were washed and incubated for 1 h at 37°C with HRP-conjugated component (antibody or streptavidin). After a final wash, the plates were developed with TMB (BD Biosciences) and read at 450 nm in a microplate reader (Safari², Tecan).

Plasma levels of complement factor D, anaphylotoxins, MBL oligomers, and total IgG immunocomplexes (CIC-C1q) were measured using the Duo Set ELISA development kit for human factor D (R&D systems), the Cytometric Bead Array kit for human anaphylotoxin (BD Biosciences), the MBL oligomer ELISA kit (Innate Immunity Diagnostics), and the CIC-C1q ELISA kit (Quidel), respectively, according to manufacturer directions. C-reactive protein was measured by immunoturbimetry using a Hitachi 912 (Roche Diagnostics).

Statistical analysis and plotting were carried out using Prism version 4 and R version 2.9.0 for Macintosh. In order to ensure that the data are Gaussian-distributed, concentration values of each complement protein were log-transformed. Gaussianity of the log-tranformed data was confirmed by means of cumulative frequency distribution plots and the Kolmogorov-Smirnov test (data not shown). Two-tailed unpaired t-tests were then used to compare the averages of concentration of different proteins, whereas two-tailed paired t-tests were used to compare the levels of different complement components in the same individual at different stages of infection (acute and convalescence). Pearson's correlation was used to assess the co-variation among the groups analyzed. Differences were considered statistically significant when p<0.05.

In order to quantify the imbalance in levels of factor H and factor D across different diagnostic categories, log-ratios between the concentration of factor H over the concentration of factor D were computed. Factor D acts by cleaving BB in the C3bBB complex to yield active C3bBb, whereas factor H regulates C3bBb by displacing the Bb portion of this C3 convertase, thus having the opposing effect of inactivating the AP convertase. If the log-ratios in concentration of factor H over factor D tend to be larger in a given diagnostic category than in the healthy category, then we may conclude that the balance between factors D and H is less favorable for the formation of the C3bBb C3 convertase in that diagnostic category than in healthy individuals, whereas smaller log-ratios mean that the balance is more favorable. Thus, the log-ratios in concentration of factor H over factor D allow the study of how these two proteins are regulating the alternative pathway C3bBb C3 convertase levels, however it does not directly indicate the amount of active convertase.

During complement activation, C3 is cleaved to form C3a and C3b; when C3 is cleaved faster than it can be replaced, the levels of intact C3 decay. This phenomenon often occurs when the complement system is highly active. In the present study, we measured the levels of C3 in the DF and DHF patients during the acute phase and compared their levels to those of control patients (Figure 1A, Table 2). The C3 levels in the DF patients in the acute phase were nearly identical to those in the healthy subjects (1.5+/−0.9 mg/mL vs. 1.5+/−0.8 mg/mL, respectively). However, the C3 levels in DHF patients (1.0+/−0.6 mg/mL) were significantly (p = 0.002) lower than those in the DF patients during acute dengue infection and also significantly (p = 0.0006) lower than the C3 levels in the healthy control group. In order to determine whether the lower levels of C3 were indeed the result of high levels of complement system activity during infection, the C3 levels of the DHF patients were also measured during the convalescent phase (Figure 1B). The data showed a significant (p = 0.005) increase in the C3 levels of the DHF patients after they recovered from the acute infection, and these levels were not significantly different from the levels found in control patients, confirming that the C3 was consumed at higher levels in DHF patients than in DF patients during the acute stage of the dengue infection.

In addition, we have performed analyses comparing the C3 levels on primary vs. secondary dengue infection and no significant differences were detected. Similarly we tested the effect of dengue infection history on the other complement factors we measured and we have not detected any influence within each clinical diagnosis assessed.

We also measured the plasma levels of C3a, a product of C3 cleavage, and C5a, a product of C5 cleavage, in samples from patients with confirmed dengue infection and different clinical diagnoses (Figure 2). Both C3a (Figure 2A) and C5a (Figure 2B) were present at approximately three-fold higher levels in the DHF patients than in the DF patients; these differences were statistically significant (p<0.001). The elevated levels of C3a and C5a that we observed in DHF patients are consistent with their decreased levels of C3 and further confirm the association of high complement activity with DHF.

The increased consumption of C3 and elevated levels of byproducts of complement activation (C3a and C5a) that we observed in DHF patients confirmed that higher complement activation is correlated with dengue severity. However, the pathway of complement activation was not defined. Therefore, we measured the levels of C4a, which correlates with the formation of the (C4b2b) C3 convertase, a hallmark for activation of the CP or LP (Figure 3). This result indicates that C4a was 3-fold elevated in DHF patients suggesting the involvement of either the CP or the LP. In order to identify the mechanism of the CP or LP activation we measured levels of C1q, CRP, CIC-C1q and oligomerized MBL in plasma samples from DF and DHF patients (Figure 4; Table 2). The levels of C1q and CRP in the DF and DHF patients were elevated during acute infection when compared to healthy control subjects, and they returned to normal levels during convalescence. However, levels of both proteins were similar between DF and DHF patients. There was also no difference in the total IgG immunocomplex levels (CIC-C1q) during the acute phase between the DF and DHF patients. This result suggests that the activation of the (C4b2b) C3 convertase was not mediated by differences on the levels of immune complexes in DF and DHF patients. However, other factors, such as significant differences in the IgG isotype distribution in the two groups could interfere with the interpretation of this result, since different immunoglobulins have different capacities to activate the CP.

The activity of the LP was evaluated by measuring the levels of oligomerized MBL, and a significant increase of MBL oligomerization was found in the DHF patients as compared to the DF patients during the acute phase of infection, which might have contributed to the elevated C4a level. These observations suggest that it is possible that MBL can be contributing to the increased activation of (C4b2b) C3 convertase in DHF through mechanisms that do not involve immune complexes.

Quantitation of two AP components, factors D and H, in DF, DHF, and control patients (Figure 5, Table 2) revealed that the levels of factor D were significantly (p = 0.01) lower in DF patients than in DHF patients during acute infection and also lower than healthy controls (p = 0.005, Figure 5A). During the convalescent phase, the mean factor D levels in the DF patients returned to that of DHF patients and control subjects (Figure 5B), suggesting that the levels of factor D in DF patients were reduced during acute dengue infection, limiting the rate of conversion of the C3bBB into the active C3 convertase, C3bBb, in these patients.

In contrast, the levels of factor H, which displaces Bb from C3bBb and inactivates the C3 convertase, were significantly lower (p = 0.03) in DHF patients than in DF patients during acute infection (Figure 5C). The levels of the factor H in DHF patients were also significantly lower (p = 0.01) than those in healthy individuals, whereas in DF patients, this was not observed. During the convalescent phase, the levels of factor H in the DHF patients had returned to the levels seen in healthy controls (Figure 5D). These data indicate a reduced ability to inhibit the amount of C3bBb C3 convertase in DHF patients. Taken together, the data regarding factors D and H suggest that DHF patients have a reduced ability to regulate the activity of the AP.

Based on the fact that factors D and H act together to influence the availability of the C3 convertase and consequently the activation of the complement system through this pathway, we analyzed the levels of both proteins in each patient through the log-ratio of their concentrations, as described in the Materials and Methods. The results of this analysis indicated that DF and DHF patients exhibited strikingly different patterns of AP activity during acute dengue infection (Figure 6). In this figure, a value of zero indicates that the levels of both factors were at proportions similar to that seen in healthy (ND) control subjects, negative values indicate a greater proportion of factor D in relation to factor H, and positive values indicate the opposite. The acute DF patients showed significantly (p = 0.046) larger log-ratios in relation to healthy controls, whereas the acute DHF patients showed significantly (p = 0.009) reduced log-ratios when compared to healthy controls. The difference in log-ratios between the acute DF with DHF patients was highly significant (p<0.0001). During the convalescent phase, the log-ratios of the DF and DHF patients approximated of those of the control group, and no significant differences were detected (Figure 6). Further supporting the association of the AP regulatory proteins with dengue disease severity, levels of C3 levels and factor H were higly correlated (Figure 7). These results suggest that during acute dengue infection, the regulatory proteins of the AP are modulated in order to maintain the homeostasis of complement activity, and when an imbalance between levels of factor D and factor H occurs, during the acute phase of the infection, the regulation of this pathway is decreased, leading to an abnormal activation of the complement system.

Platelet levels are reduced during acute dengue infection (both DF and DHF), and thrombocytopenia is associated with an increased severity of the dengue infection. The main mechanisms that have been proposed to account for the induction of thrombocytopenia in dengue infection include virus-mediated megakaryocyte suppression in the bone marrow ^[21] and increased platelet destruction ^[22]–^[24]. In the present analysis, the C3 levels were found to be directly correlated with platelet levels and inversely correlated with C3a and C5a levels (Table 3), indicating that the higher the complement system activity is, the lower the platelet level during acute dengue infection is. Factor H was also highly correlated with platelet levels in the dengue patients, whereas C1q, CIC-C1q, CRP, and MBL protein levels were not (Table 3). These results further support our conclusion that abnormal modulation of AP regulatory proteins is a significant factor contributing to dengue severity.