Plasma Interferon-Gamma–Inducible Protein 10 Can Be Used to Predict Viral Load in HIV-1–Infected Individuals

To the Editors: Interferon-gamma–inducible protein 10 (IP-10), also known as CXCL10, is a chemokine involved in both innate and acquired immune responses that direct T cells to sites of inflammation.1,2 Plasma IP-10 has been shown to correlate closely with inflammation, liver fibrosis, and hepatitis C virus infection,3 reflect HIV load in cerebral spinal fluid,4 and be a useful marker for early HIV disease progression.2,5 In addition, genital tract levels of IP-10 may reflect vaginal HIV load.6 Herein, we report on plasma IP-10 concentrations correlating with plasma viral load that can be used as a predictive marker of viral replication in HIV-infected adults. Plasma concentrations of IP-10 were evaluated by Luminex assay using a Bio-Plex Cytokine reagent kit (BIO RAD Laboratories, Hercules, CA) in 51 HIV-uninfected and 55 HIV-infected individuals, including 7 participants on antiretroviral (ARV) treatment. The median plasma IP-10 concentration in HIV-uninfected controls was 340 ng/mL [interquartile (IQR): 249–468], and in all HIV-infected individuals (including those on ARV), there was a significantly higher concentration (P < 0.0001 using Mann–Whitney) of 1160 ng/mL (IQR: 779–2088). There was also a significantly higher concentration of IP-10 when comparing HIV-uninfected and HIV-infected individuals receiving ARV (median: 778 ng/mL, IQR: 534–924, P = 0.0009), although all those on ARV were below the threshold of 400 RNA copies per milliliter. Relative to ARV-naive HIV-infected individuals, the lower IP-10 concentration in patients receiving ARV is compatible with reduced inflammation upon viral suppression.7 However, the finding that IP-10 is significantly higher in ARV-treated individuals, relative to uninfected controls, also suggests that inflammatory signals are not completely dampened and may reflect viral activity below the 400 copies per milliliter threshold. When grouping all HIV-infected individuals, there was a significant positive correlation (P < 0.0001) between IP-10 and viral load [r = 0.71, 95% confidence interval (CI): 0.52 to 0.84; Fig. 1A], compatible with the close association between this chemokine and viral burden. We also measured IP-10 longitudinally in 25 of the 55 HIV-infected individuals over 9 months, where a significant correlation was observed between changes in both IP-10 and viral loads over time (r = 0.65; P = 0.049), accentuating the parallel course of viral load and IP-10. We used receiver-operating characteristic (ROC) curves to identify the possible predictive nature of IP-10. A significant area under the curve (AUC) was noted when selecting viral loads above or below 5000 RNA copies per milliliter (Fig. 1B, AUC = 0.88, 95% CI: 0.77 to 0.98). This was also reflected when a threshold of 2000 RNA copies per milliliter was chosen (AUC of 0.888, 95% CI: 0.78 to 0.996, P = 0.0009, not shown). Together, these data show the robust ability of plasma IP-10 concentrations to predict different levels of viremia with good sensitivity and specificity. Using a crossover plot to precisely identify the point where specificity and sensitivity intersects, an IP-10 concentration of 900 ng/mL could predict viral loads above and below 5000 RNA copies per milliliter (Fig. 1C). These data highlight that IP-10 is a suitable host marker for predicting viral load and may potentially be used instead of the latter assay. IP-10 has also been proposed as a useful biomarker for sputum clearance in tuberculosis patients,8 and there are reports showing that IP-10 is useful to track disease progression in HIV-infected individuals.5 Our data would concur with this, and we extend the analysis to show a direct association with in vivo viral replication, as has been similarly shown in hepatitis C infection.9 When we extended the analysis to CD4 counts, we identified that there was no association between IP-10 and CD4 count changes over time, and ROC curve analysis showed an AUC of 0.58, with no significant predictive capability. Thus, although plasma IP-10 is an excellent proxy and predictor for viremia, it is unrelated to CD4 numbers.FIGURE 1: Plasma IP-10 concentrations correlate and predict viral load in HIV-1–infected individuals. A, Spearman rank correlation between IP-10 concentrations and log10 RNA copies per milliliter; B, ROC curve showing significant AUC for IP-10 above and below 5000 copies per milliliter; C, specificity/sensitivity crossover plots showing the intercept at which the concentration of IP-10 is predictive (denoted by hashed arrow line).In summary, using ROC curve analysis, we have shown that IP-10 is a good predictor of viral load, is useful for tracking changes in viremia, including a response to ARV, and may be an alternative to the more costly viral load measurements. Whether there is a casual biological association between HIV and IP-10 is unclear, but we speculate that IP-10 is secreted as a direct effect of viral replication and activation of tumor necrosis factor alpha and interferon-gamma in host leukocytes.1 As a result of this direct association, plasma IP-10 concentration can be used as an accurate proxy for viral load and could potentially be adapted to use as a point-of-care test. This is of great importance in a country such as South Africa, which has one of the largest ARV rollouts in the public health sector. Having a relatively quick colorimetric assay to measure this host-derived chemokine would obviate the need to send samples to regional or tertiary centers for viral load measurements. ACKNOWLEDGMENTS The authors would like to thank Drs Landon Myer and Catherine Riou for useful discussions.