Gene expression before HAART initiation predicts HIV-infected individuals at risk of poor CD4+ T-cell recovery

Abstract

Objective: To identify a pre-HAART gene expression signature in peripheral blood mononuclear cells (PBMCs) predictive of CD4 T-cell recovery during HAART in HIV-infected individuals.

Design: This retrospective study evaluated PBMC gene expression in 24 recently HIV-infected individuals before the initiation of HAART to identify genes whose expression is predictive of CD4 T-cell recovery after 48 weeks of HAART.

Methods: The change in CD4 T-cell count (DeltaCD4) over the 48-week study period was calculated for each of the 24 participants. Twelve participants were assigned to the 'good' (DeltaCD4 > or = 200 cells/microl) and 12 to the 'poor' (DeltaCD4 < 200 cells/microl) CD4 T-cell recovery group. Gene expression profiling of the entire transcriptome using Illumina BeadChips was performed with PBMC samples obtained before HAART. Gene expression classifiers capable of predicting CD4 T-cell recovery group (good vs. poor), as well as the specific DeltaCD4 value, at week 48 were constructed using methods of Class Prediction.

Results: The expression of 40 genes in PBMC samples taken before HAART predicted CD4 T-cell recovery group (good vs. poor) at week 48 with 100% accuracy. The expression of 22 genes predicted a specific DeltaCD4 value for each HIV-infected individual that correlated well with actual values (R = 0.82). Predicted DeltaCD4 values were also used to assign individuals to good vs. poor CD4 T-cell recovery groups with 79% accuracy.

Conclusion: Gene expression in PBMCs can be used as biomarkers to successfully predict disease outcomes among HIV-infected individuals treated with HAART.

Figures

Fig. 1. Gene expression in peripheral blood mononuclear cell samples taken from HIV-infected participants before HAART was used to predict after 48 weeks of HAART: (a) the good (ΔCD4 ≥ 200 cells/μl) vs. poor (ΔCD4 + T-cell recovery group or (b) the specific ΔCD4 value for each patient

(a) When predicting recovery group, classification accuracies achieved using different multivariate classification methods were determined for gene expression classifiers containing different numbers of genes. A full description of classification methods can be found in the BRB-Array Tools manual [26]. Support vector machines (SVMs) were the most robust multivariate classification method in which 40 genes attained 100% classification accuracy. (b) When predicting the specific ΔCD4 value for each HIV-infected patient, the LASSO algorithm was used in a cross-validated approach to identify 22 genes whose expression could predict ΔCD4 values that correlated well with actual ΔCD4 values. Predicted ΔCD4 values were then used to assign participants to the good and poor CD4+ T-cell recovery groups based on the ΔCD4 threshold of 200 cells/μl (dotted lines). This allowed the number of true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN) to be determined. Classification accuracy was calculated as [(TP + TN)/(TP + FP + TN + FN)] × 100 and was equivalent to 79%. R refers to the Pearson's correlation coefficient.