Table 1.
Cohort Characteristics.
Fig 1.
The frequency of HIV specific CD4 T-cells expressing IFNγ in the CSF was higher than in peripheral blood.
PBMC (1a) and CSF (1b) cells from the same HIV+ donor were cultured overnight with HIV p24 antigen and stained for intracellular IFNγ expression.
Table 2.
T-cell cytokine expression is often higher in CSF than in peripheral blood.
Fig 2.
CD8+ T-cells from NC impaired subjects expressed increased levels of constitutive and p24 induced IFNγ, IL-2, and TNFα at higher CSF HIV RNA levels (>400 copies/ml).
Constitutive and HIV p24 antigen induced CSF CD8+ IFNγ, IL-2, and TNFα expression by normal (triangles) and neurocognitive impaired (circles) subjects were compared at 3 CSF HIV RNA levels. Expression levels were square root transformed. Differences between impaired and unimpaired subjects were significant at higher HIV RNA levels (>400 copies/ml) in CSF (WRS = Wilcoxon Rank Sum p<0.001)
Table 3.
Neurocognitive impairment is associated with a low CD4/CD8 ratio in plasma and higher CD8+T-cell cytokine expression in CSF.
Table 4.
Division of data according to CSF HIV RNA detectability reveals distinct CD8 T-cell and monocyte inflammatory responses correlations with NC impairment in the HIV RNA detectable and HIV RNA undetectable conditions.
Fig 3.
When HIV RNA was detectable in CSF, constitutive CD8+IFNγ and IL-2 expression correlated positively with NCI (Fig. 3a) and CD4+ and CD8+CD107a expression correlated negatively with NCI (3b) Percent cytokine (3a) or lytic marker (3b) expression were plotted as a function of neurocognitive global deficit score (GDS).
Expression levels and GDS were square root transformed. a: When HIV RNA was detectable in CSF, there was a positive correlation between constitutive CSF CD8+IFNγ (solid line, diamonds r = 0.57, p = 0.004) and IL-2 expression (dotted line, circles r = .49, p = 0.01) and NCI. b: In contrast, CSF CD4+CD107a+ (solid line, triangles r = 0.49, p = 0.001) and CD8+CD107a (dotted line, circles r = .44, p = 0.004) expression were negatively associated with NCI.
Fig 4.
Constitutive and HIV p24 induced CD8+CD107a expression (lytic activity) in CSF and peripheral blood were only detected when HIV RNA levels were <1000 copies/ml and CD4 levels were >400/μl.
a: The percentage of HIV p24 induced CD4+ (solid diamond) and CD8 (open circle) and constitutive (cCD107a) CD4+ (open triangle) and CD8+ (x) expression in CSF were plotted as a function of log HIV RNA in CSF. CD107a expression >1% was only detected when HIV RNA was <1000 copies/ml. b: The percentage of constitutive CD8+CD107a+ in CSF was plotted as a function of plasma CD4 levels. CD8+CD107a expression >1% was detected when CD4 levels were >400/μl. Cytokine expression and CD4 levels were square root transformed.
Fig 5.
CD8+IFNγ + expression and soluble CD163 in CSF represent two significant and independent correlates of NCI (5a). The composite association is stronger than either alone (5b).
a: The percentage of constitutive CSF CD8+IFNγ expression (solid circles, dotted line, r = 0.59 p = 0.004) and soluble CSF CD163 (open diamonds, solid line, r = 0.42, p = 0.004) were plotted as a function of neurocognitive global deficit score (GDS). b: The percentage of constitutive CSF CD8+IFNγ expression and pg of soluble CSF CD163 were summed for visits at which both values were available and the composite score plotted as a function of neurocognitive GDS (solid triangles, r = 0.83, p<0.0001). Expression levels and GDS were square root transformed.
Fig 6.
When HIV RNA is suppressed, the percent of CD8+ T-cells in CSF (6a) and the percent of HIV specific CD8+ IFNγ+ T-cells in CSF (6b) both correlate with severity of NCI (GDS).
a, b: When HIV RNA was suppressed in CSF, CD8+ T-cells (6a: r = 0.66, p = 0.002) and HIV p24 induced CD8+ T-cell IFNg expression (6b: r = .53, p = 0.02) correlated positively with NCI (GDS). Cell numbers, expression levels and GDS were square root transformed.
Table 5.
When HIV RNA is suppressed, NC impairment is associated with higher CD8+ T-cell levels, higher CD14 levels, lower T-cell lytic activity and lower beta chemokine levels in CSF.
Fig 7.
Neurocognitive outcome (GDS) can be viewed as a function of levels of chemoattractant (CXCL10) in CSF combined with the phenotype of CD8+ T-cells (CD107a or IFNγ+) available for recruitment into the CNS.
Using all patient visits, best fit lines/curves for constitutive CSF CD8+CD107a (r = 0.28 p = 0.05), CD8+IFNγ expression (r = 0.47 p = 0.001) (both left axis), and CSF CXCL10 (r = 0.51 p = 0.004) (right axis) were plotted as a function of neurocognitive GDS. The dotted line indicates the conventional cut-off for clinically significant NCI, crossing near the inflection point for CXCL10 and the intersection of falling CD107a levels and rising CD8+IFNγ expression. CD8+T-cell expression levels, CXCL10, and GDS were square root transformed.
Table 6.
Longitudinal analysis confirms correlation between NC impairment and CD8+IFNγ+ (positive) and lytic activity (negative) at baseline and reveals positive association between NC impairment and residual HIV RNA in CSF at follow-up.
Table 7.
The most significant predictor of neurocognitive decline was higher CD8+IFNγ response to HIV p24 antigen at baseline.