Approach to cerebrospinal fluid (CSF) biomarker discovery and evaluation in HIV infection

Abstract

Central nervous system (CNS) infection is a nearly universal facet of systemic HIV infection that varies in character and neurological consequences. While clinical staging and neuropsychological test performance have been helpful in evaluating patients, cerebrospinal fluid (CSF) biomarkers present a valuable and objective approach to more accurate diagnosis, assessment of treatment effects and understanding of evolving pathobiology. We review some lessons from our recent experience with CSF biomarker studies. We have used two approaches to biomarker analysis: targeted, hypothesis-driven and non-targeted exploratory discovery methods. We illustrate the first with data from a cross-sectional study of defined subject groups across the spectrum of systemic and CNS disease progression and the second with a longitudinal study of the CSF proteome in subjects initiating antiretroviral treatment. Both approaches can be useful and, indeed, complementary. The first is helpful in assessing known or hypothesized biomarkers while the second can identify novel biomarkers and point to broad interactions in pathogenesis. Common to both is the need for well-defined samples and subjects that span a spectrum of biological activity and biomarker concentrations. Previously-defined guide biomarkers of CNS infection, inflammation and neural injury are useful in categorizing samples for analysis and providing critical biological context for biomarker discovery studies. CSF biomarkers represent an underutilized but valuable approach to understanding the interactions of HIV and the CNS and to more objective diagnosis and assessment of disease activity. Both hypothesis-based and discovery methods can be useful in advancing the definition and use of these biomarkers.

Fig. 1

Schematic diagram of salient systemic and CNS HIV disease components. CNS HIV infection and immune responses initially are extensions across the blood–brain and blood-CSF barriers (horizontal dashed line) of their systemic counterparts, though with variable selection and local evolution depending on the stages of disease. Within the nervous system, both the virus and immune responses may impact the CNS and its function in the depicted triangle. In the small font are examples of biomarkers for each of the main pathogenic components. Plasma HIV RNA and blood CD4 cells in systemic disease have proved to be essential biomarkers in systemic management, while CSF HIV RNA, neopterin and NFL have served as principal guide biomarkers in our studies of CNS infection and injury

Fig. 2

Changes in the three guide markers with disease progression and suppressive treatment studied cross-sectionally. The panels show the changes in CSF HIV RNA, neopterin and NFL with infection as systemic disease progresses, showing HIV uninfected controls (HIV-), neuroasymptomatic (NA) subjects with progressively lower blood CD4+ T cells, patients with HAD, and a group on suppressive therapy (N = 20 in each group except HAD with 12 and suppressed with 19). CSF HIV RNA was elevated in all untreated groups, and highest in HAD though in this small study only the treated-suppressed differed from the other infected groups (P < 0.05–<0.001). CSF neopterin was elevated in all the untreated HIV subjects, though highest in the HAD group (all HIV-infected groups differed from HIV- except CD4 > 350; HAD differed from suppressed and CD4 > 350; suppressed differed from all untreated except CD4 >350) (p values). NFL was highest in the HIV group (above normal in all) but NA subjects with lower CD4+ T cells showed a substantial prevalence of elevated levels indicative of subclinical CNS injury (HAD, CD4 <50 and 50–199 differed from HIV- while HAD differed from all groups except CD4 <50) (stat p values). Statistical comparison used Kruskal-Wallis test and Dunn’s post hoc test of multiple comparisons; graph and statistics prepared on Prism 6 (Graphpad Software Inc, San Diego, CA). Box plots show median and intraquartile range, ‘+’ shows mean and error bars the 10–90 percentiles

Fig. 3

Protein correlations with CSF neopterin and pathway diagram. The heat map on the left-side diagrams the significant protein correlations (proteins with R values either >0.3 or < −0.3 by Spearman analysis) with decreasing concentrations of CSF neopterin. The upper nine rows show normalized (Z scores) protein concentration groups with positive correlations (green to red) while those below show negative correlations (red to green). The right diagram shows results of a pathway analysis that included the highest number of previously defined relationships of these neopterin correlating proteins and identified APP (amyloid precursor protein) as a ‘node’ in these relationships using Ingenuity Pathway Analysis (http://www.ingenuity.com/). Thus, using neopterin as an external biomarker we were able to identify an unrelated CSF neuronal protein that had previously shown to correlate with HAD, thus validating this overall approach with respect. The pathway analysis also suggests possible links in the pathogenesis related to these protein changes, based on previously reported interactions of the correlating proteins. Abbreviations use standard nomenclature. Undefined abbreviations on right panel include: APP amyloid precursor protein; CD14 cluster of differentiation 14, also monocyte differentiation antigen that serves as lipopolysaccharide (LPS) co-receptor. Data are from Angel et al. (Angel et al. 2012) where the findings are described in greater detail

Fig. 4

Comparison of biomarker sample distributions of two studies. In each panel the sample results have been sorted by value from highest to lowest for the listed biomarkers. The top three panels show data from the cross-sectional study outlined earlier and presented by group in Fig. 2 using the same color scheme that is also defined in the middle panel. The lower three panels show sorted samples from the longitudinal study, with those from each individual subject now identified by color as shown in the middle panel. The X axes differ because of the larger number of samples in the cross-sectional study. The CSF HIV RNA (left panels) shows a similar distribution in the two studies from the level of quantitation (40 copies per mL) to >100,000 copies per mL. Similarly, the CSF neopterin concentrations of both sample sets span a similar distribution. The CSF NFL values in the two studies show differences that relate to two factors. First, the assays differed in the two studies with the cross-sectional study (actually performed more recently) using the newer, more sensitive assay (Peluso et al. 2013) accounting for both the higher values and the continued value spread at the lower end of concentrations (including within the normal value distribution below 890 ng per L), whereas in the second study the lower half of the sample was below the detection limit of 125 ng/L) for the older assay (Gisslen et al. 2007a). Second, the longitudinal study had a larger number of samples from neurologically abnormal subjects as a result of selection of HAD patients and those with low blood CD4+ cells for study. Overall, both studies show a continuous distribution of values for each of the CSF biomarkers