Fig 1.
Study group background characteristics.
Except for the first panel (Fig 1A), the graphs (Fig 1B–1O) use the same format that is applied to subsequent figures showing the Olink-generated protein measurement profiles across the study groups. This includes the symbols and the median and IQR bars. The dashed horizontal lines in Fig 1K–1O indicate the mean +2 SD of the uninfected, seronegative (HIV-) control group in the measured variable as a rough estimate of their upper normal limit and as a visual guide to compare groups; this convention is also used in the Olink CSF protein measurement figures showing the group profiles in subsequent figures. A. Study sites. Most of the samples were obtained in the context of scheduled outpatient visits during long-term cohort studies in Gothenburg, Sweden (GOT) and San Francisco, California, USA (SF) that included lumbar punctures (LPs) as part of the study protocols (Fig 1A). This included CSF specimens from the HIV-, elite controllers (elites), five neuroasymptomatic CD4-defined groups, two ART-treated and virally suppressed groups (one with normal CSF NfL and one with elevated CSF NfL, designated as RxNFL+ and RxNFL-) and the asymptomatic escape (AsE) group. The HIV-associated dementia (HAD), neurosymptomatic escape (NSE) and secondary escape (2ryE) samples were obtained during clinical evaluations after informed consent. Additionally, the NSE and HAD groups were augmented by CSF specimens from Milan and Pune to attain a size comparable to the other main groups. When there was a larger number of CSF specimens available for a given group than required for the study, samples were chosen at random from archived collections (e.g., CD4-defined groups). An exception to this random selection was the choice of the HIV- group from GOT: all of these were from individuals taking pre-exposure prophylaxis (PrEP) because of their self-identified risk for HIV-1 infection. HIV- controls from SF were seronegative individuals from the same clinical site and demographic population as the people living with HIV (PLWH) included in the local cohort, but not on PrEP. The Rx NFL+ and the AsE groups were almost all from GOT related to the local interest in these conditions. The small number of secondary escape (2ryE) were also all from GOT. In general, the imbalances in the subject group sizes related to scarcity of available samples (e.g., elites and 2ryE, resulting in smaller sample sizes), or to inclusion of larger numbers of specimens to augment particularly important groups, including the HIV- group, related to its comparative utility, and HAD and NSE groups because of our interest in more severe HIV-1-related CNS injury. These imbalances, along with those of calendar years and subject ages (B and C below), emphasize that this exploratory study used a convenience sample rather than one that was balanced with respect to individual demographic variables. B. Calendar year. The dates of study varied among the groups. In part this reflected the history of our cohort studies in concert with the changes in acceptance and efficacy of treatment regimens over the collection period. These time variations impacted the length of untreated chronic infection, susceptibility to disease progression and presence of viral suppression among our study cohorts. For example, there were nearly 10 years separating the median years of the HAD group from both treatment-suppressed groups (RxNFL- and RxNFL+). The NSE group sampling also reflected later recognition and attention to this clinical entity. C. Age. The ages of the subjects were not specifically selected and consequently also varied among groups, with group medians ranging from 39 to 50 years with largely similar ranges. Overall, the median ages of the treated PLWHs were similar to those of the HIV-negative (HIV-) controls, while the untreated groups were generally younger. The balance of sex was not taken into account with specimen choices, so distribution also varied among the groups. D. Blood CD4+ T-lymphocytes. The median blood CD4+ T-cell count in the elites was above that of the uninfected controls and the other HIV-1-infected groups, including the two treatment suppressed groups. The blood CD4+ T-cell concentrations in the 5 groups defined by these counts show the stepwise decrease from >500 to <50 CD4+ T-cells/μL dictated by the study design. The HAD group had a median CD4+ T-cell count of 108 cells/μL that was similar to the 50–199 CD4+ group median of 106 cells/μL, in keeping with the role of advanced immunosuppression in the development of this subacute disorder and its underlying HIVE pathology. The CD4+ T-cell counts of the two treatment-suppressed groups were nearly equal, with both having blood CD4+ T-cell counts above those of the viremic CD4-defined groups except for the group with the highest blood CD4+ T-cell counts (CD4>500 group), indicating the partial CD4+ cell preservation or recovery with suppressive treatment. By contrast, the blood CD4+ T-cell counts in the three CSF escape groups were lower than the two treatment-suppressed groups, consistent with less robust recovery or ongoing loss during the escape episodes. However, these levels were not as low as in the untreated HAD group. The CD4+ blood counts of the 2ryE subjects were also low, indeed below the other two escape groups, likely contributing to their vulnerability to intercurrent infections (herpes zoster in three and HSV2 meningitis in the fourth) and factoring into their clinically-directed LPs. E. Blood CD8+ T lymphocytes. Blood CD8+ T-cell counts were on the high side of normal in the elites, rose as blood CD4+ cells decreased in the untreated CD4-defined groups except for those in the CD4 <50 group (median of 380 CD8+ T-cells per μL). They were relatively increased in the HAD group (median 750 CD8+ T-cells per μL) and in all the treated groups, including the two virally suppressed and the NSE and ASE groups (medians in 800s per μL). F. Blood CD4+/CD8+ T-cell ratio. The differences in the trajectories of the two T-cell subpopulations resulted in lowered ratios in all infected groups compared to the HIV- controls, reflecting variable reductions in CD4+ and elevations of CD8+ T-lymphocytes. Reductions were most severe in the CD4 -defined groups, most notably in those with CD4+ T-cell counts below 200 cells per μL and in the HAD group. These ratios were relatively increased in the four treated groups (0.47 to 0.77), though all remained below the level of the uninfected controls (median ratio of 1.75). Reduced ratios persisted in the two ART-suppressed groups (RxNFL- and RxNFL+) and the AsE and NSE groups. G. Nadir blood CD4+ T-lymphocytes. Nadir CD4+ T-cell counts are shown only for the ART-treated groups, since CD4+ T-cell values in the untreated groups at their study visits were generally at or near their nadirs. The treated groups showed varying degrees of presumed blood CD4+ T-cell recovery, higher in the two virally suppressed (Rx NFL- and NFL+) groups (medians of 590 and 600 CD4+ T-cells per μL) than in the CSF escape NSE and ASE groups (medians of 425 and 412 CD4+ T-cells per μL, respectively), the latter perhaps either contributing to development of CSF escape or, alternatively, reflecting an impact of escape on CD4+ T-cell dynamics. H. CSF HIV-1 RNA. Highest CSF HIV-1 RNA concentrations were present in the HAD group, while the CD4-defined groups showed the ‘inverted U’, or lymphoid, pattern of change with lower concentrations at the extremes (in the CD4+ >500 and CD4+ <50 groups) than in the middle ranges (CD4+ 350–499, 200–345 and 50–199 groups). We previously reported the association of this pattern with that of the CSF WBC counts as also noted in this study (see Fig 1K below), suggesting that these two findings were causally related [45]. In this study the correlation of CSF HIV-1 RNA to CSF WBC count across the CD4-defined groups was significant (Pearson correlation P = 0.006). This lymphoid pattern of infection and cell response likely underlies this same pattern in many of the CSF proteins included in the Olink Explore 1536 panel as shown in later figures. The CSF HIV-1 RNA concentrations of three escape groups were generally below those of the untreated groups except for the CD4 <50 group. CSF HIV-1 RNA concentrations were all below detection in the elites. I. Blood HIV-1 RNA. In the untreated individuals (not including the elites) there was a steady increase in blood HIV-1 RNA through the full range of CD4+ T-cell loss that reached its highest mean levels in the HAD group. Unlike with CSF, there was no decrement in the CD4 <50 group, showing a distinct difference between the CSF and the systemic blood viral dynamics. The presence of low-level blood HIV-1 RNA in the NSE group may have reflected systemic partial drug resistance in some or spillover of the ‘escaped’ CNS/CSF infection into the blood. J. CSF:Blood HIV-1 RNA differences. This panel shows the differences in viral loads between the two fluids (calculated as the log10 CSF HIV-1 RNA copies per mL–log10 plasma HIV-1 RNA copies per mL) and emphasizes the largely consistent relationship between CSF and blood HIV-1 RNA levels in untreated infection at blood CD4+ T-cell levels between 50 and >500 cells/μL in which the CSF concentrations were approximately 10-fold lower than those in blood, and thus maintaining a nearly 1:10 ratio of CSF to blood HIV-1 RNA. This may relate to the kinetics in the influx of infected CD4+ T-cells and of viral release into the CSF over this blood CD4+ T-cell range. This ratio was disrupted when blood CD4+ T-cells fell to <50 cells/μL and the CSF WBC count decreased to negligible levels; in this group the CSF:blood HIV-1 RNA ratio decreased nearly 10-fold to an overall ratio of <1:100. This underscores the importance of lymphocyte traffic and likely direct virus release by trafficking infected CD4+ T-cells in the determining the CSF HIV-1 levels in the neurologically asymptomatic individuals, particularly those with blood CD4+ T-cell counts above 50 cells per μL. This relationship changed markedly with the development of HAD and the underlying direct neuropathic parenchymal CNS infection, i.e., HIVE, in which the CSF HIV-1 RNA concentrations reached high levels, and the differences between it and the blood viral load decreased. Thus, in the HAD group local brain infection rather than hematogenous sources likely was responsible for the high levels of CSF HIV-1 RNA. The reversed CSF:blood HIV-1 RNA ratios in the three CSF escape groups were consonant with their definitions predicated on CSF > blood HIV-1 RNA concentrations that indicated the direct CNS sources of the measured CSF HIV-1 RNA in the face of systemic viral suppression. K. CSF WBC counts. The median CSF cell counts rose and then fell over the course of untreated infection in the CD4-defined groups. CSF WBC counts were highest in the 200–349 group and declined to lowest levels when blood CD4+ T-cells fell below 50 cells/μL, defining the lymphoid pattern. As discussed above, this likely was causally linked to the changes in CSF HIV-1 RNA in these groups. Both neurologically defined groups were associated with augmented local inflammation: CSF WBCs were elevated in HAD, and even higher in NSE (median counts of 10 and 22 cells per μL, respectively). In these two settings the CSF WBC counts likely involved a response to local CNS infection and injury. L. CSF neopterin. This pteridine is predominantly, though perhaps not exclusively, a macrophage-related activation marker [143]. CSF levels showed a steady median increase as CD4 cells declined, including the highest levels in the CD4+ <50 group among the CD4-defined sequence of groups, thus providing a prototypical example of the myeloid pattern of CSF biomarker change. CSF concentrations were even higher in the HAD group, and, notably, were highest in the NSE group, while suppressive treatment brought CSF neopterin concentrations back to near normal levels in the absence of symptomatic CSF escape. M. Blood neopterin. In blood, neopterin concentrations showed a similar, though more restricted elevation with CD4 decline but were notably increased in both the CD4 < 50 and HAD groups, presumably indicating systemic macrophage activation in these settings that either links these two sites of infection or marks parallel changes in myeloid cell populations in the CNS and systemically. In contrast to CSF, the blood neopterin was only mildly elevated in the NSE and 2ryE groups in keeping with the underlying compartmentalized CNS HIV-1 and intercurrent infections in these two groups in the face of systemic HIV-1 suppression. N. CSF: blood albumin ratio. This ratio provides an index of blood-brain barrier integrity, though with considerable individual and age-related variability [144]. In this sample set, only direct HIVE in the HAD group and, to a lesser degree, in the NSE group was associated with elevated median albumin ratios. Minor, though not significant, increases were present in the RxNFL+ group. One-way ANOVA with Tukey’s multiple comparison test found significant differences (P<0.05) only for HAD (CD4 >500: P = 0.0004; Elites: P = 0.0028; AsE: P = 0.0067; CD4 200–349: P = 0.0069; RxNFL-: P = 0.0116; CD4 50–199: P = 0.0202; CD4 350–499 P = 0.0228) and NSE (CD4 >500: P = 0.0023; Elites: P = 0.0075; AsE: P = 0.0173; CD4 200–349 P = 0.0211; HIV-1- P = 0.024; RxNFL- P = 0.0337; CD4 50–99 P = 0.0479). Other intergroup comparisons were not significant. Overall, the albumin ratio elevations were small, and likely to have had only limited, if any, impact on the CSF proteins measured in this panel. O. CSF NfL. Prior to this study, CSF NfL was measured using the UMAN ELISA method (UmanDiagnostics, Umeå, Sweden) in 209 of the 307 (71.3 percent) of the specimens. In this figure, the NfL values have been adjusted to age 50 years using the normative data and methods of Yilmaz and colleagues [68] to allow more direct subject comparisons and estimations of abnormal values. While these data were incomplete and superseded in this study by the Olink measurements performed on all samples, they guided our initial definition of the RxNFL+ group. The highest levels of NfL were in the HAD group, and there were elevated levels in a substantial proportion of the CD4 <50 group presumably related to subclinical neurological injury. Notably, the NfL elevation of the HAD group had nearly a 10-fold higher median value than that of the NSE group. These observations were confirmed and extended by the measurement of NfL within the Olink Explore panel as shown later in S2 Fig.
Fig 2.
Overview of CSF proteins across subject groups.
The assembled panels depict some of the main features of the CSF protein changes across the set of subject groups. A. This hierarchical cluster analysis displays the full set of the Olink CSF protein measurements in a heatmap. The study groups (and the individual CSF samples within each group) are arranged on the vertical axis, while the proteins are grouped across the figure after segregation into three clusters designated by colors at the top within BLUE, RED and GREEN columns of the measurements of individuals proteins in each sample. The heatmap relative scale from -1 (blue) to 0 (yellow) to 1 (red) is shown to the right of the main figure. The individual proteins are color-scaled according to their relative concentrations. The most conspicuous gradations in scale are within the GREEN cluster (right column) in which many samples from the HAD group and the two symptomatic escape groups (NSE and 2ryE) stand out with higher (red) concentrations compared to the generally lower concentrations (blue and yellow) in the non-viremic individuals (HIV-, elites, and RxNFL+ and RxNFL-) with some mixed colors over this spectrum in the CD4-defined groups. By contrast, the RED cluster (middle column) showed a more even and mottled spectrum except for some of the proteins shown on the left side of this cluster in which there was red predominance in the HAD and two symptomatic escape groups similar to, but less pronounced than, in the GREEN cluster. The changes among groups in the BLUE cluster (left column) are less distinct but with some low (blue) concentrations in the CD4 <50, HAD and two symptomatic escape groups (NSE and 2ryE), and higher concentrations in the HIV- group, i.e., showing some patterns that are opposite to (or inverted in comparison to) those in the GREEN cluster. Overall, the HAD and two symptomatic escape groups accounted for the greatest concentration differences from the controls among the proteins, most prominently in the GREEN cluster. B. This PCA analysis using results of all the measured CSF proteins to assess effects of the overall CSF measurements on individual subjects and their groups provides a different perspective on the protein changes across the groups. 1. This panel includes the results from all the groups and shows a ‘gradient’ downward and to the right of the panel associated with HIV-1 systemic disease progression (as indicated by falling blood CD4+ T-cell counts) and, more particularly, with the symptomatic neurological disease groups, so that individuals with HAD and NSE reach farther in this direction. Because of the superimposition of many of the individual in this panel, three smaller panels (Fig 2B2-4) separate these groups for better visual definition of these changes. 2. the aviremic groups including the HIV-, the largely superimposed elites, treated-suppressed (RxNFL- and RxNFL+) and the ASE group are all grouped together with overlap in the upper left of the plot; 3. the five CD4-defined groups are also largely clustered in the same upper left area but with some individuals ‘moving’ down and to the right as immunosuppression advanced with lower blood CD4+ T-cell counts; 4. the HAD and two symptomatic escape groups (NSE and 2ryE) showed some overlap with the other groups but with further movement down and to the right (higher PC1 and lower PC2 values), again consistent with these neurological groups importantly ‘driving’ the main group differences in the aggregate CSF protein changes across the large cohort sample. Thus, these groups, defined clinically, are also, not surprisingly, separated in this analysis by the changes in their CSF proteomes. C. Proteins and pathways influential in distinguishing subject groups. The two bar plots show the proteins and pathways that most influenced the overall differences among the subject groups. The left bar graph lists the most influential proteins, all 10 of which are markers of inflammatory processes. The right bar graph lists the most influential defined pathways extracted from the protein changes, and again their designations emphasize that they identify inflammatory processes, including several involving T-cells. While, in part this may reflect the large number of inflammatory proteins that were included in the Explore 1536 panel, many with intersecting or overlapping functions, it also clearly shows that changes in inflammatory profiles largely determine the contours of the overall measured CSF proteome as systemic disease progressed and, particularly, as overt CNS disease with HIVE developed. D. Influential protein patterns across study subjects. These plots show the subject group concentrations of the 10 most influential proteins (from the left side of Fig 2C above) across the subject groups using the color schema introduced in Fig 1 and used throughout this report. This includes the color symbols, medians and IQR bars, and the dashed horizontal line that indicates the mean + 2 SD of the HIV- control group in this study as an estimate of the upper limit of these variables and aids as visual reference across the groups. Protein concentrations are expressed in log2 NPX units with varying scales. The plots show the prominence of the increases in the HAD and NSE groups and their dominating influence; these two groups had the highest concentrations of all 10 of these CSF proteins, with lower and more variable patterns of elevation in the CD4-defined and other groups. As annotated in the panels, of the 10 listed proteins, the measurements in six were all (100%) above the LODs for the assay. The exceptions included LILRA5, SH2D1A, PCDC1, and CLEC6A in which 99.1%, 47.3%, 88.9% and 61.0%, respectively, of the proteins were above the LODs (shown in the individual panels here as in subsequent figures of protein patterns across the subject groups). These influential proteins are all within the GREEN cluster in Fig 2A and are listed in the Glossary of Featured Proteins in Fig 2 included in S1 Appendix with brief functional descriptions extracted from UniProt database (https://www.uniprot.org/uniprotkb?query=*) [140], along with comments on selected features of the individual panels.
Fig 3.
CSF proteins in relation to correlations of CSF NEFL and HIV-1.
This figure explores the relationships of the CSF proteins to two major variables (or pathogenic vectors) of HIV-1-related CNS disease: CSF HIV-1 RNA, an ambiguous index of CNS infection, and CSF NEFL (NfL measured by Olink in the Explore 1536 panel), an objective measurement of active CNS injury. A. Correlation of NEFL with CSF HIV-1 RNA. This plot shows a significant (P<0.0001), but relatively weak correlation (R2 = 0. 0.07797) between NEFL and CSF HIV-1 RNA by linear regression. Among the features underlying this limited correlation are the high CSF RNA values without elevation of NEFL in the untreated CD4-defined groups, the elevations of CSF NEFL in both the treated NSE group despite their relatively lower CSF HIV-1 RNA levels (related to the partial suppression from ART) and the untreated HAD patients with high levels of HIV-1 RNA. Separate analysis of the untreated and treated groups showed little improvement of the correlations (both also with P< 0.0001, but with R2 values of 0.1699 and 0.1373, respectively). Analysis of HAD patients alone showed p = 0.1709 (R2 = 0.07087). B. High correlating individual proteins. The three bar graphs show the 10 CSF proteins with: the highest correlations to 1. CSF HIV-1 and 2. CSF NEFL; and 3. the sum of the CSF HIV-1 and CSF NFL correlations. For 27 of the listed proteins, the measurements are above the LODs. Exceptions include SH2D1A in which only 47.3% of the proteins are above the LODs, and consequently this protein is omitted from further descriptions and the plotting of the proteins across subject groups below. The other two exceptions are PCDC1 (88.89% above LODs) and LILRA5 (99.1 above the LODs) which are included in the further discussion and illustration. Notably some of these proteins were also previously identified as strongly impacting the overall segregation across groups as defined earlier in Fig 2. C. Correlations of the full set of CSF proteins with CSF NEFL and HIV-1 RNA. This figure includes a total of 1662 proteins, i.e., all of the Olink-measured proteins except NEFL, plotted according to their correlations with CSF NEFL and CSF HIV-1 RNA. It provides a broad overview of the measured proteins in the context of these two important pathogenetic vectors. Each protein symbol is colored according to their grouping in the three clusters (GREEN, BLUE and RED) defined earlier in the hierarchical cluster analysis (Fig 2A) with the solid symbols showing the proteins with ≥60% of the proteins above their LODs and the open symbols showing the proteins with <60% of proteins above their LODs. Exceptions to this symbol color coding are the top five proteins identified in the three groups in Fig 3B for which the symbols have been individually enlarged, color-coded using the bar colors in Fig 3B (all were originally identified in the GREEN cluster), and labelled with their gene names. These highly correlating proteins are located at the extremes on the top, right and right upper sections of the plot (CD48 appears in both the HIV-1 and HIV-1 + NEFL groups and is colored according to the dual correlation). The dotted grid lines mark the 0 and 0.5 correlation levels for both the Y and X axes. Overall, the GREEN cluster proteins populate the highest NEFL and HIV-1 RNA correlation areas with a far larger number in the ≥0.5 of CSF NEFL correlation (Y axis) (375 proteins) than in the ≥0.5 area of CSF HIV-1 RNA correlation (X axis) (55 proteins). Twenty-two of these proteins are in the shared area in the upper right with correlations ≥0.5 for both CSF NEFL and CSF HIV-1 RNA. A minority of the GREEN proteins had <60% of their measurements above the LODs as indicated by the open green circles that exhibit NEFL correlations <-0.5 and CSF HIV-1 RNA correlations between 0.25 and 0.5. None of the 294 BLUE-clustered proteins are in the ≥0.5 HIV-1 RNA correlation area, and indeed most exhibit negative correlations with HIV-1 RNA; 52 of the BLUE-clustered proteins are in the ≥0.5 NEFL correlation area. Only a small number of the BLUE-clustered proteins show <60% of measurements above the LODs. Of the 566 RED-clustered proteins 10 are in the ≥0.5 correlations with NEFL and none correlate with HIV-1 RNA ≥0.05. These RED-clustered proteins thus are concentrated along the vertical line correlation designating a 0.0 correlation with CSF HIV-1 RNA. Additionally, many the RED-clustering proteins have <60% of proteins above the LOD (open symbols), likely contributing to their lack of correlations with either CSF NEFL or CSF HIV-1 RNA (with most grouped along the X axis at 0 correlation with CSF HIV-1 RNA). This figure with the ‘cloud’ of protein placement in relation to the two pathogenetic vectors emphasizes that the correlations of most of the Olink-measured proteins with CSF NEFL and HIV-1 RNA are largely ‘dissociated’. Only a limited number lay along the diagonal dotted line delineating equal correlations with the two major disease variables. Additionally, the much larger number of proteins correlating with NEFL at levels above 0.5 than with HIV-1 RNA, emphasizes the major influence of CNS injury on the CSF proteome. This is consistent with the strong effect of HAD and NSE on CSF proteins as can be seen by the examples below and with many of the proteins identified in other sections of the study. By contrast, high HIV-1 RNA levels in some of the CD4-defined groups with normal CSF NEFL values also contributes to this dissociation. D. Patterns of high-correlating proteins across subject groups. This Panel shows the patterns of CSF protein concentrations of the top five proteins in the three categories identified in Fig 3B. Again, the dashed horizontal lines are at the mean + 2SD of the HIV-1- group as a visual reference for comparison of group value distributions. The top row shows the five proteins with the highest correlations with HIV-1 RNA. Here the highest protein concentrations vary between the HAD and NSE groups while the CD4-defined groups exhibit the lymphoid pattern with a reduction in the protein levels in the CD4 <50 group compared to the CD4 50–199 and 200–350 groups. This is not surprising because this is the pattern of the CSF HIV-1 RNA concentrations and CSF WBC counts in these groups (see Fig 1I and 1K). The patterns in the middle row of proteins with highest correlations with NEFL exhibit two major features: highest concentrations in the HAD group and a gradual increase across the CD4-defined groups with highest values within these in the CD4 <50 group, i.e., with the myeloid pattern. This circumstantially supports the association of CNS injury with myeloid cell inflammation. The final row with highest combined HIV-1 and NEFL correlations shows a mixture of the two patterns just outlined: variability in whether HAD or NSE had highest concentrations and in the presence of either the lymphoid or myeloid patterns in the CD4-defined groups. The identified proteins in Fig 3D are listed in S1 Appendix with brief functional descriptions extracted from UniProt database (https://www.uniprot.org/uniprotkb?query=*)[140].
Fig 4.
CSF proteins favoring NSE vs HAD.
This figure examines CSF protein differences between NSE and HAD. A. Comparison of CSF proteins in NSE and HAD groups. Both of these conditions are forms of HIVE with elevations of multiple CSF proteins. For the most part, the protein elevations do not differ significantly, and this volcano plot shows a high degree of protein similarity (or lack of significant differences) within the context of the full complement of measured proteins (identified by simple blue symbols) on each side of the central vertical axis. However, there are a few identified protein differences, indicated by the enlarged labelled symbols formatted using the NSE and HAD symbols. The remaining panels of the figure focus on these distinguishing proteins. B. CSF proteins distinguishing NSE and HAD subject groups. Proteins identified by t stat of ≥ +/- 5 are the same those singled out in Fig 4A volcano plot, with eight proteins favoring NSE and four favoring HAD. C. NSE and HAD distinguishing proteins in the context of their correlations with CSF NEFL and CSF HIV. This panel shows all the measured Olink proteins (except NEFL) in the context of their correlations with CSF NEFL and CSF HIV RNA using the same format as Fig 3C, but, for clarity, with a neutral yellow color for the full array of proteins rather than the GREEN, BLUE and RED designations from the hierarchical cluster analysis. The significant proteins identified in the volcano plot and t stat above (Fig 4A and 4B) are indicated by enlarged symbols. All are from the GREEN group of the cluster analysis. The proteins favoring both neurological conditions are similarly variable across a range of correlations with CSF HIV-1 RNA but located in different vertical strata with respect to NEFL correlation. Those favoring HAD are in a higher range of correlation with NEFL than those favoring NSE. This may relate, at least in part, to the generally lower levels of NEFL in NSE than HAD (Fig 2A). D. Cellular associations of proteins favoring NSE and HAD. This simple presentation draws on literature sources describing the cell associations of 10 of the proteins identified in Fig 4A and 4B, https://www.proteinatlas.org/ and [156] with the darkness of the square indicating strength of the cell relations. The proteins listed as favoring NSE are largely associated with lymphocytes (B, T and NK lineages), while those favoring HAD associate with myeloid cells, consistent with a different balance of inflammatory cells in these two forms of HIVE. E. Different patterns of high-correlating proteins in NSE and HAD across subject groups. The format of the graphs in this panel is the same as in previous figures showing Olink protein concentrations across the subject groups. The upper two rows plot the protein concentrations of the 8 proteins favoring NSE while the bottom row plots the proteins favoring HAD as identified in Fig 4A and 4B above. In the upper two rows, the median concentrations in the NSE group are greater than those of the comparable HAD group which are either normal or near the highest values among the CD4-defined groups. The CD4-defined groups exhibit the lymphoid pattern with varying levels of increase in the middle brackets. The bottom row shows the patterns of the four proteins that favor HAD. Most notably, the values of HAD group are higher than those of NSE group, and the CD4-defined group patterns are either flat or show an increase in the CD4<50 group above the groups with higher CD4+ T-cell counts, closest to the myeloid or perhaps neuronal patterns, but with clear variation in the magnitude of differences of the CD4<50 from the other CD4-defined groups. The dashed horizontal line in each panel again designates the mean +/- 2 SD of the HIV- control group for visual reference. The identified proteins in Fig 4E are listed in S1 Appendix with brief functional descriptions extracted from UniProt database (https://www.uniprot.org/uniprotkb?query=*) [140].