1. Supplementary Figure 2. Kinetics of HIV-specific IgG in the plasma over time in the tenofovir and placebo arms from 3 months to 39 months post-infection. Responses to gp70 V1V2 (A), Consensus Group M gp120 (B), Clade C V3 (C), Consensus Clade C Env gp140 (D), gp41(E) and p24 (F) Plasma response magnitude is presented as MFI*dilution. Boxes denote the 25th and 75th percentile and whiskers denote the minimum and maximum values. The horizontal line represents the median response. Responses with MFI<100 are below the limit of detection for this assay. P<0.05 were statistically significant.

2. Supplementary Figure 3. Kinetics of HIV-specific IgA in the plasma over time in the tenofovir and placebo arms from 3 months to 39 months post-infection. Responses to gp70 V1V2 (A), Consensus Group M gp120 (B), Clade C V3 (C), Consensus Clade C Env gp140 (D), gp41(E) and p24 (F). Plasma response magnitude is presented as MFI*dilution. Boxes denote the 25th and 75th percentile and whiskers denote the minimum and maximum values. The horizontal line represents the median response. Responses with MFI<100 are below the limit of detection for this assay. P<0.05 were statistically significant.

3. Supplementary Figure 4. Kinetics of HIV-specific IgG in the genital tract over time in the tenofovir and placebo arms from 3 months to 39 months post-infection. Responses to gp70 V1V2 (A), Consensus Group M gp120 (B), Clade C V3 (C), Consensus Clade C Env gp140 (D), gp41(E) and p24 (F). CVL response magnitude is presented as specific activity [Log10 (MFI* dilution factor/(ng ml-1)]. Responses with MFI<100 are below the limit of detection for this assay. P<0.05 were statistically significant.

4. Supplementary Figure 5. Kinetics of HIV-specific IgA in the genital tract over time in the tenofovir and placebo arms from 3 months to 39 months post-infection. Responses to Consensus Group M gp120 (A), p24 (B) and p66 (C). CVL response magnitude is presented as specific activity [Log10 (MFI* dilution factor/(ng ml-1)]. Responses with MFI<100 are below the limit of detection for this assay. P<0.05 were statistically significant.

5. Supplementary Figure 6. Comparison of identified LASSO signature to 1000 other combinations of 8 antibody specificities. The LASSO-identified signature was significantly better for distinguishing tenofovir and placebo arms, as cross-validation (A) and calibration (B) accuracy were both 99th percentile rank compared to other models generated from 8 non-LASSO antibody specificities. Shown are histograms of cross-validation and calibration accuracy for 1000 different models generated from 8 non-LASSO antibody specificities from this dataset.