The central nervous system (CNS) is invaded early in the evolution of HIV-1 disease  and is one of the potential sanctuaries of the virus . HIV-1 RNA can be detected in cerebrospinal fluid (CSF)  and its presence may reflect a local production, leakage from plasma or both [4-7]. Furthermore, CSF viral load has been correlated with the degree of AIDS dementia [8-10]. If leakage from plasma is the explanation for the presence of HIV-1 RNA in CSF, the fall in HIV-1 RNA in CSF in response to antiretroviral therapy [11,12] might not indicate control of HIV-1 infection in the CNS. Conversely, if the hypothesis of local production is true, lumbar puncture might be necessary to measure the level of HIV-1 RNA in this compartment before starting antiretroviral therapy and thereafter to monitor the response.
The analysis of the correlation of the levels of HIV-1 RNA in plasma and CSF in patients with integrity of the blood-brain barrier (BBB) could be a useful tool for investigating the origin of HIV-1 RNA in the CSF. Examination of CSF in asymptomatic patients with early stages of chronic infection would permit correlation of the CSF HIV-1 RNA level with the markers of CNS HIV-1 infection, such as CSF total proteins and white cell counts. If the hypothesis of leakage is true, when the integrity of BBB is preserved two groups of patients with similar levels of plasma viral load should also have similar levels of CSF viral load. Furthermore, the level of HIV-1 RNA in CSF should not necessarily correlate with markers of HIV-1 CNS infection such as CSF protein levels and white cell counts.
In spite of the relevance of these issues, almost all the studies correlating plasma and CSF viral load have been performed in small groups  of patients or in those with a neurological disease or in advanced stages of their HIV-1 disease [5-10]. Many of these patients might have an increase in the permeability of the BBB ; this makes it difficult to assess if the presence of HIV-1 RNA in CSF is related to leakage from plasma or to local production.
The present study analyses several parameters in CSF and plasma in 70 consecutive asymptomatic patients with stable chronic HIV-1 infection and a CD4 T cell count >500×106cells/l in order to assess CSF HIV-1 RNA levels and correlate them with plasma viral load and CNS HIV-1 infection markers.
All consecutive patients screened for two clinical trials (antiretroviral-naive adults with chronic asymptomatic HIV-1 infection with CD4 lymphocyte count >500×106cells/l and viral load as HIV-1 RNA >5000copies/ml) were eligible for a lumbar puncture if written informed consent was granted. These additional studies were approved by the hospital ethical committee. A complete evaluation included a physical examination and a history of risk behaviour, time from first diagnosis of HIV-1 infection, previous diseases and immunizations.
At day 0, a simultaneous sample of CSF and plasma were obtained. Levels of albumin, total IgG (measured by kinetic nephelometry) and titres of HIV-1 specific antibodies (IgG) against p24 antigen were measured in parallel. Plasma and CSF antibody titres were defined as the reciprocal value of the highest dilution that showed positive staining by enzyme-linked immunosorbent assay (ELISA, Wellcozyme Antip24, Abbott Diagnostics, Wierbaden-Delkenheim, Germany). The integrity of the BBB was estimated using the albumin index (CSF albumin/serum albumin×1000; normal value <7). To estimate the intrathecal production of antibodies against HIV-1, an antibody index defined by the formula [titre of CSF antibody against HIV-1 p24 antigen (reciprocal)× total plasma IgG]/[total CSF IgG× titre of plasma antibody against HIV-1 p24 antigen (reciprocal)] . A value greater than 1.0 suggested specific intrathecal synthesis. HIV-1 RNA was measured in samples of plasma and CSF that had been stored frozen (-70ºC) using the polymerase chain reaction (PCR) (Amplicor HIV Monitor, Roche Diagnostics Systems, Branchburg, NJ, USA). Those samples with less than 200copies/ml (lower limit of detection) were retested using the Amplicor HIV-1 Monitor Ultra Sensitive Specimen Preparation Protocol Ultra Direct Assay (Roche Molecular Systems, Inc., Somerville, NJ, USA) with a lower limit of quantification of 20copies/ml. To avoid between-assay variation, all samples from each individual were run in parallel on the same bath. CSF glucose and protein (normal value <0.45g/l) determinations and a white cell count (normal value <5x106cells/l) were also performed. Hepatitis C virus (HCV) RNA was measured simultaneously in the plasma and CSF samples using the PCR technique (Amplicor HCV Monitor, Roche Diagnostics Systems). CD4 T cell counts were measured at the same time point by flow cytometry.
Data were entered into a data base (DBase IV; Borland, Scotts Valley, CA, USA) and analysed using the SPSS statistical package (SPSS version 4.0; SPSS, Chicago, IL, USA). For the purpose of analysis, RNA values reported as undetectable (<20copies/ml) were considered equivalent to 20copies/ml. Pearson‚s correlation coefficient was calculated to assess the correlation between levels of HIV-1 RNA in plasma and CSF. The patients were classified into two groups according to the difference between HIV-1 RNA levels in plasma and CSF. Group A patients had a difference in the levels of <1.5log10 copies/ml and group B had a difference of ≥1.5. The baseline characteristics between these groups were compared in order to detect the data that could predict a difference in the levels of HIV-1 RNA between plasma and CSF of ≥1.5log10 copies/ml. A second sub-analysis was performed classifying the patients into two groups according to the median level of CSF HIV-1 RNA: group I <1698copies/ml and group II >1698 copies/ml. The analysis of continuous variables with normal distribution was performed with the t test; log transformation of the data was performed with data with skewed distribution. Non-parametric tests (Mann-whitney U test) were used for continuous variables with unequal variances or non-normal distribution. Categorical data were analysed with the chi-square test. Those statistically significant variables in the univariate analysis were introduced in a multivariate analysis (logistic regression model).
Of the 70 patients included in the study, 52 (74%) were males and 13 (19%) intravenous drug users. The mean age was 34 years (SD 8). HIV-1 infection was diagnosed from 6 to 170 months before enrolment (median 20 months). Ten patients (14%) had a past medical history of a symptomatic HIV-1 acute infection. In 40 patients, a determination of antibodies against HCV was available. Ten of these 40 patients (25%) were co-infected with HCV. The CD4 lymphocyte count (mean± SD) was 664(±179)×106cells/l and the plasma viral load (mean± SD) was 4.53±0.53log10 copies/ml. Table 1 shows the baseline characteristics of the patients.
Cerebrospinal fluid characteristics
In 59 out of 70 patients (84%), the albumin index was <7, suggesting integrity of the BBB. The antibody index was >1 in 55 out of 70 individuals (78%), suggesting intrathecal synthesis of IgG antibodies against HIV-1 p24 antigen. The viral load in CSF (mean± SD) was 3.13±0.95log10 copies/ml (median 3.23; range 1.3-5.33). CSF viral load was <200 copies/ml in 25% of the patients but <20copies/ml in only three patients (4%). The CSF glucose, proteins and white cells values (mean± SD) were 630±10g/l, 0.44±0.17g/l and 9(±10)×106cells/l, respectively. Thirty-two patients (48%) had a CSF protein level above the upper limit of normality and 38 patients (54%) had a white cell count above the upper limit of normality. Overall, 48 out of 70 patients (69%) had a protein level or white cell count above the normal level. None of the 10 patients co-infected by HCV had HCV RNA detectable in CSF, despite having very high plasma HCV RNA levels (mean±SD 6.36±0.29copies/ml; range 6.09-6.73).
Correlation between HIV-1 RNA in plasma and cerebrospinal fluid
Plasma viral load was significantly higher than CSF values: 4.53±0.53 versus 3.13±0.95log10 copies/ml HIV-1 RNA (P=0.0001) (Fig. 1). CSF viral load was greater than plasma viral load only in three patients and the difference was always lesser than 0.5log10 copies/ml (Fig. 2). There was a significant correlation (r=0.43, P=0.0001) between plasma and CSF viral load (Fig. 3). Despite this, it was possible to define two groups (A and B) with almost identical plasma levels [the difference of the means of plasma HIV-1 RNA between group A and B was -0.01 (95% confidence interval, -2.8 to 0.25; P=0.9)] but with a difference between the HIV-1 RNA levels in plasma and CSF <1.5 (group A) or ≥1.5 (group B) log10 copies/ml (Fig. 2; Table 1). Specifically, in 29 out of 70 patients (41%), HIV-1 RNA levels in plasma were ≥1.5log10 copies/ml above CSF levels. There were no differences between the groups A and B with regard to age, gender, risk factor for HIV-1 infection, time from first diagnosis of HIV-1 infection, history of symptomatic acute HIV-1 infection, co-infection by HCV or CD4 T cell counts (Table 1).
Correlation between viral load, white cell count and protein levels in cerebrospinal fluid
If patients were divided into two groups according to median value of CSF viral load (group I <1698copies/ml and group II >1000 copies/ml), those in group I had significantly lower levels (mean ± SD) of CSF proteins, 0.37±0.14g/l, and white cells, 4(±6)×106cells/l, than patients of group II [proteins 0.5±0.2g/l and white cells 13(±12) x106cells/l] (P=0.01 and 0.001, respectively). When the 11 patients with a disrupted BBB were excluded from the analysis, the results were very similar (data not shown). In a logistic regression analysis, the level of proteins and white cells in CSF, but not the plasma viral load, were the only factors independently associated with the level of HIV-1 RNA in CSF (P=0.0001).
HIV-1 RNA can be detected almost always in CSF from HIV-1 infected patients. Its presence could indicate up to three different clinical conditions, as described by Enting et al. . The first is patients with low plasma CD4 T cell counts and a CNS opportunistic infection. This could be cryptococcal meningitis and the presence of high levels of HIV-1 RNA could reflect CNS immune activation, as happens with plasma viral load during acute opportunistic infections or following an immune stimulation [16,17]. The second clinical situation is patients with HIV-1 encephalitis. In this case, the level of CSF HIV-1 RNA has been correlated with the degree of cognitive impairment [8-10]. The third situation, and by far the most frequent, is the group of neurologically asymptomatic patients. In these, almost all the studies of CSF viral load have been performed in small groups  or in advanced stages of HIV-1 disease [5-7]. Many of these patients might have an increase in the permeability of the BBB , which is difficult to assess because the presence of HIV-1 RNA viremia in CSF could be related to leakage from plasma or occur as a result of local production.
Our cohort of 70 consecutive stable asymptomatic patients with a CD4 T cell count >500×106cells/l offers the opportunity to analyse CSF HIV-1 RNA levels and its correlation with plasma viral load and with CNS HIV-1 infection markers without the drawback of disruption of the BBB or the presence of opportunistic infections. Our data strongly argue against the conclusions reached by Annunziata et al. that early nervous system invasion by HIV-1 in a study with an animal model may be explained by an early alteration of the permeability of BBB . The integrity of the BBB was preserved (according to the albumin index) in 84% of our patients. In spite of this, almost all of them had indirect evidence of CNS involvement since 69% had CSF protein levels or white cell counts above the normal level. In addition, most of our patients (78%) showed an intrathecal production of antibodies against HIV-1 p24 antigen (measured by the antibody index), as reported by other authors [19,20].
In 25% of the patients CSF HIV RNA was undetectable as measured with the widely used Amplicor HIV Monitor (with a lower limit of detection of 200copies/ml). However, HIV RNA was detectable in 96% of the patients using an ultrasensitive method (lower limit of detection 20copies/ml). The values of CSF viral load were similar to previous reports in patients in early stages of disease . The median of HIV-1 RNA in CSF was 1698log10copies/ml. CSF levels were never 0.5log10 copies/ml higher than plasma levels, and only three patients had absolute levels of viral load in CSF higher than in plasma (Fig. 2).
Different study populations and the influence of antiretroviral therapy could explain some of the conflicting results in the correlation between plasma and CSF viral load. In many studies, a positive correlation has been found [21-23], while in others this correlation was not observed [10,11,24]. The positive correlation between CSF and plasma viral load would suggest a leakage from plasma as the most likely explanation for the presence of HIV-1 RNA in CSF. In this case, the patients maintaining the integrity of the BBB and with similar levels of HIV-1 RNA viremia should also have similar levels of HIV-1 RNA in CSF. We have found a positive correlation between plasma and CSF viral load (r=0.43; P= 0.0001; Fig. 3). However, in spite of this significant correlation, it was possible to define two groups with almost identical plasma viral load and differences between HIV-1 RNA plasma and CSF levels, below (41%) or above 1.5 log10 copies/ml. Apart from this substantial difference between plasma and CSF viral load, these two groups were homogeneous with regard to demographic and clinical characteristics. Furthermore, CSF viral load was independently correlated with CSF markers of CNS HIV-1 infection (CSF protein level and white cell count). Conversely, plasma viral load was not an independent predictor of CSF viral load. Therefore, the correlation between CSF and plasma viral load suggests a high level of virus production in both compartments rather than a leakage of virus from plasma to CSF.
In order to strengthen the evidence of a local production as the source of CSF HIV-1 RNA, we have also studied the presence in CSF of HCV. This is a RNA virus smaller than HIV-1 (35nm and 80nm, respectively). If the hypothesis of leakage is correct, it would be expected that HCV would also be found in CSF. However, in none of the 10 patients co-infected by HCV could HCV RNA be detected in CSF, while the mean of plasma HCV RNA was 6.36log10copies/ml.
In summary, HIV-1 RNA can be detected almost always in CSF even in asymptomatic patients in early stages of chronic HIV-1 infection with preserved integrity of the BBB. This preserved integrity of BBB found in most of our patients argues against the hypothesis that the early nervous system invasion by HIV-1 may be a consequence of an early alteration of the permeability of the BBB. The important discrepancies between plasma and CSF viral load, and the independent association between CSF abnormalities and CSF viral load, support the hypothesis of a local production of HIV-1.
We are indebted to Dr Jordi Yagüe for the measurement of the albumin index and the intrathecal production of antibodies, to Dr Josep Costa for the hepatitis C virus RNA measurements and to Dr Francesc Graus for comments on the manuscript.
1. Chiodi F, Keys B, Albert J, et al. Human immunodeficiency virus type 1 is present in the cerebrospinal fluid of a majority of infected individuals. J Clin Microbiol
2. Pialoux G, Fournier S, Moulignier A, et al. Central nervous system as sanctuary of HIV-1 in a patient treated with AZT + 3TC + indinavir. Fourth Conference on Retroviruses and Opportunistic Infections
. Washington, DC, January 1997
3. Cinque P, Scarpellini P, Vago L, Linde A, Lazzarin A. Diagnosis of central nervous system complications in HIV-infected patients: cerebrospinal fluid analysis by the polymerase chain reaction
4. Krivine A, Esmilaire L, Lebon P, Force G. HIV-RNA in CSF and permeability of the blood-brain barrier. Fourth Conference on Retroviruses and Opportunistic Infections
. Washington, DC, January 1997
5. Ghorpade A, Nukuna A, Che M, et al. Human immunodeficiency virus neurotropism: an analysis of viral replication and cytopathicity for divergent strains in monocytes and microglia
. J Virol
6. Martin C, Albert J, Hansson P, et al. Cerebrospinal fluid mononuclear cell counts influence CSF HIV-1 RNA levels. J Acquir Immune Defic Syndr Hum Retrovirol
7. Cinque P, Vago L, Mengozzi M, et al. Elevated cerebrospinal fluid levels of monocyte chemotactic protein-1 correlate with HIV-1 encephalitis and local viral replication
8. Brew BJ, Pemberton L, Cunningham P, Law MG. Levels of human immunodeficiency virus type 1 RNA in cerebrospinal fluid correlate with AIDS dementia stage. J Infect.Dis
9. di Stefano M, Monno L, Fiore J, et al. Neurological disorders during HIV-1 infection correlate with viral load in cerebrospinal fluid but not with virus phenotype. AIDS
10. Cinque P, Vago L, Ceresa D, et al. Cerebrospinal fluid HIV-1 RNA levels: correlation with HIV encephalitis. AIDS
11. Foudraine N, Hoetelmans R, Lange JMA, et al. Cerebrospinal fluid HIV-1 RNA and drug concentrations after treatment with lamivudine plus zidovudine or stavudine. Lancet
12. Garcia F, Gatell JM. A randomized study comparing triple versus double antiretroviral therapy or no treatment in HIV-1 infected patients in very early stages. The Spanish Earth-1 Study. Fourth International Congress on Drug Therapy in HIV infection
. Glasgow, November 1998 [abstract OP7.4].
13. Gisslen M, Hagberg L, Fuchs D, Norkrans G, Svennerholm B. Cerebrospinal fluid viral load in HIV-1-infected patients without antiretroviral treatment: a longitudinal study. J Acquir Immune Defic Syndr Hum Retrovirol
14. Potasman I, Resnick L, Luft BJ, Remington JS. Intrathecal production of antibodies against Toxoplasma gondii in patients with toxoplasmic encephalitis and the acquired immunodeficiency syndrome (AIDS). Ann Intern Med
15. Enting RH, Hoetelmans R, Lange JMA, et al. Antiretroviral drugs and the central nervous system. AIDS
16. Bush CE, Donovan RM, Markowitz N, Kvale PA, Saravolatz LD. A study of HIV RNA viral load in AIDS patients with bacterial pneumonia. J Acquir Immune Defic Syndr Hum Retrovirol
17. Garcia F, Vidal C, Gatell JM, et al. Changes in HIV-1 RNA viral load following tuberculin skin test. J Acquir Immune Defic Syndr Hum Retrovirol
18. Annunziata P, Cioni C, Toneatto S, Paccagnini E. HIV-1 gp120 increases the permeability of rat brain endothelium cultures by a mechanism involving substance P. AIDS
19. Goudsmit J, Wolters EC, Bakker M. Intrathecal synthesis of antibodies to HTLV-III in patients without AIDS or AIDS related complex. Br Med J
20. Andersson M, Begstrom T, Blomstrand C, et al. Increasing intrathecal lymphocytosis and immunoglobulin G production in neurologically asymptomatic HIV-1 infection. J Neuroimmunol
21. Calvez V, Suarez S, Dubard T, et al. Cerebro-spinal fluid HIV-1 RNA load and AIDS dementia. Fourth Conference on Retroviruses and Opportunistic Infections
. Washington, DC, January 1997 [Abstract].
22. McArthur JC, McClermon D, Cronin M, et al. Relationship between human immunodeficiency virus-associated dementia and viral load in cerebrospinal fluid and brain. Ann Neurol
23. Bossi P, Dupin N, Coutellier A, et al. The level of human immunodeficiency virus (HIV) type 1 RNA in cerebrospinal fluid as marker of HIV encephalitis. Clin Infect Dis
24. Sei S, Stewart SK, Farley M, et al. Evaluation of human immunodeficiency virus (HIV) type 1 RNA levels in cerebrospinal fluid and viral resistance to zidovudine in children with HIV encephalopathy
. J Infect Dis