Retrospective Identification of a Broad IgG Repertoire Differentiating Patients With S. aureus Skin and Soft Tissue Infections From Controls

Abstract

Background: Although the relevance of humoral immunity for protection against S. aureus skin and soft tissue infections (SSTIs) has been suggested by several animal and human studies, the question of which human antibodies may be protective has so far impeded the development of a safe and effective vaccine. Because most adults have developed certain anti-S. aureus antibodies due to S. aureus colonization or infection, we hypothesized that the titers of antibodies to S. aureus in uninfected controls would differ from those in infected patients and would also differ in infected patients from the time of acute infection to a 40-day convalescent serum. Methods: To test these hypotheses, we measured human antibody levels against a panel of 134 unique antigens comprising the S. aureus surfome and secretome in subjects with active culture-confirmed S. aureus SSTIs (cases) and in controls with no infection, using a novel S. aureus protein microarray. Results: Most S. aureus SSTI patients (n = 60) and controls (n = 142) had antibodies to many of the tested S. aureus antigens. Univariate analysis showed statistically weak differences in the IgG levels to some antigens in the SSTI patient (case) sera compared with controls. Antibody levels to most tested antigens did not increase comparing acute with 40-day serum. Multiple logistic regression identified a rich subset of antigens that, by their antibody levels, together correctly differentiated all cases from all controls. Conclusions: Antibodies directed against S. aureus antigens were present both in patients with S. aureus SSTIs and in uninfected control patients. We found that SSTI patients and controls could be distinguished only based on differences in antibody levels to many staphylococcal surface and secreted antigens. Our results demonstrate that in the studied population, the levels of anti-S. aureus antibodies appear largely fixed, suggesting that there may be some level of unresponsiveness to natural infection.

Keywords: Staphylococcus aureus; antibodies; antibody response; microarray; skin infection.

Figure 1

Showing IgG levels (anti-human IgG mg/ml) for each antigen (y-axis) derived through the calibration process for the SSTI cases at the two measured time points (Top) and for the control subjects (Bottom Right). Low values are shown in darker tones and higher values are shown in progressively lighter tones. Although few antigens are consistently associated with very low IgG levels, antibodies were detectable against most S. aureus antigens. Also, among case subjects, temporal changes in IgG were observed both across subjects and antigens (Bottom Left).

Figure 2

The mean and range of IgG levels for each antigen calculated across case and control subjects (Top); and the mean and range of the 134 IgG levels for each of the 30 culture-confirmed S. aureus SSTI case subject serum samples (Middle) and for each control subject serum sample (Bottom).

Figure 3

Venn diagram showing the antigens having statistically significant differences (p-value < 0.05) in analyses comparing the frequency distributions of the IgG levels measured in the 30 culture-confirmed S. aureus SSTI subject serum samples at D0 and D40 or between the IgG levels measured in the 30 SSTI serum samples and the samples from controls. The area of each circle is proportional to the number of antigens carrying significantly different IgG levels measured across subjects in each pair-wise comparison. A total of 11/134 (8%), 13/134 (10%), and 5/134 (4%) antigens carried different IgG levels for the D0-controls, D0-D40, and D40-controls pair-wise comparisons, respectively. For information about the antigens included here, (see the Supplementary Materials, Table S1).

Figure 4

Upper panels: regression sensitivity (red dots) at D0 (left) and D40 (Right) improved monotonically in the size of the antigen set selected by the penalized multiple logistic regression while specificity (blue dots) was roughly constant. The horizontal lines show the 95% probability intervals of the random classifier (40–61% for accuracy in black, 27–73% for sensitivity in red and 40–62% for specificity in blue lines). Bottom panels: the smallest fitted S. aureus SSTI probability among SSTI cases was 78% and the highest SSTI probability among the controls was 29%, showing a clear separation of the fitted infection status on the basis of the measured IgG levels.

Figure 5

Black dots: number of antigens (horizontal axis) selected by the logistic regression plotted vs. its sensitivity (vertical axis) generated by randomly mismatching the measured IgG profiles measured at D0 (Left) and D40 (Right) and in the controls from their corresponding SSTI infection status. The size of each dot is proportional to the frequency of its coordinate values over 10,000 randomizations. The red dot in each figure panel shows the number of antigens and corresponding sensitivity obtained by fitting the logistic regression to the IgG data and their corresponding SSTI infection status. The absence of black dots near the red dots shows that the probability that 100% sensitivity could be achieved by chance is numerically zero. These results lead us to reject the null hypothesis that the association between SSTI infection status and IgG profiles is not statistically significant.

Figure 6

Each dot depicts the S. aureus SSTI classification probabilities for the D0-controls (horizontal axis) and the D40-controls (vertical axis) when IgG is set to the value 1 μg/mL for the corresponding antigen and IgG = 0 μg/mL for all other antigens. Red circles [quadrant (B)] mark antigens (see list in Table S1) which higher IgG levels were associated with a decrease in the SSTI classification probabilities both when comparing the D0 data and the D40 data to the controls. Blue circles [quadrant (C)] represent antigens which higher IgG levels were associated with an increase in the SSTI classification probabilities. Empty circles [quadrants (A) and (D)] identify antigens associated with antibody responses having no association with SSTI classification probability in one or both case-control comparisons.