Diagnosis of cytomegalovirus encephalitis in patients with AIDS by quantitation of cytomegalovirus genomes in cells of cerebrospinal fluid

Cytomegalovirus (CMV) frequently involves the CNS in patients with AIDS. CMV-related encephalitis (CMVE) is recognized in up to 40% of cases in autopsy series.^1-3 Antemortem diagnosis of CMVE is difficult because of variability of clinical features, nonspecific CSF findings, and insensitivity of viral culture from CSF.⁴ The polymerase chain reaction (PCR) has recently been successfully used to diagnose CMV-induced neurologic manifestations in patients infected with the human immunodeficiency virus type 1 (HIV-1). Detection of CMV DNA in CSF had a high diagnostic sensitivity and specificity for active CMV infection of the CNS in most^5-10 but not all studies.^11,12 Most investigators used total CSF or the supernatant of fractionated CSF to detect the presence of viral DNA. Because active CMV disease in AIDS patients may result from reactivation of virus in latently infected cells, quantitative assessment of CMV DNA in cells of CSF could be a useful means to differentiate latent from symptomatic CMV infection of the CNS and to evaluate the severity of disease. In peripheral blood, this approach allows early diagnosis of systemic CMV-related organ manifestation in patients with AIDS.^13,14

In this study, we used a nested PCR assay to determine the number of CMV genomes in serial dilutions of CSF cell lysates obtained from both AIDS patients with and without active CMVE and from HIV-1-seronegative individuals. Alternatively, using a nonquantitative PCR protocol with reduced sensitivity, we tested if detection of a "cutoff" value of cellular CMV quantities in CSF was valid to reliably identify patients with CMVE.

Methods. Patients. We retrospectively tested CSF samples from 19 patients with AIDS who underwent LP for assessment of neurologic dysfunction. CMV IgG titers were positive in 12 patients and negative in 3 patients. In four cases CMV serology was not available. Five patients had active CMV infection of the brain as confirmed by autopsy. Postmortem histopathologic analysis occurred within 3 to 7 weeks of CSF collection. Diagnosis of CMVE was based on the presence of cytomegalic inclusion-bearing cells, focal necrosis, microglial nodules, and necrotizing ventriculoencephalitis. None of the subjects with CMVE had concomitant CNS opportunistic processes. Two patients with autopsy-confirmed CMVE (Patients 3 and 4) received ganciclovir or foscarnet treatment at the time of CSF sampling, and the remaining three individuals were without specific therapy. Fourteen AIDS patients had no evidence of clinically relevant CMV disease of the CNS, among them two subjects with CMV retinitis (Patients 14 and 19) who were on maintenance ganciclovir treatment. In this group of patients, brain biopsy or autopsy results were not available. Etiologies considered to be reponsible for neurologic symptoms included cerebral toxoplasmosis (n = 3), HIV-1-related dementia (n = 2), HIV-1-related neuropathy (n = 2), aseptic meningitis (n = 2), and cryptococcal meningitis (n = 1). One individual was investigated after a seizure. Two subjects were neurologically asymptomatic and underwent LP to exclude neurosyphilis (n = 2).

As control subjects, CSF samples were obtained from 12 HIV-1-seronegative patients with inflammatory polyradiculitis (n = 2), bacterial meningitis (n = 1), headache (n = 2), Herpes simplex virus encephalitis (n = 1), aseptic meningitis (n = 1), myelitis (n = 1), MS (n = 1), neuropathy (n = 1), hydrocephalus (n = 1), and papilledema (n = 1). Among control subjects, two individuals had positive CMV IgG titers, six patients were seronegative for CMV, and in four cases CMV serology was not available. Clinical diagnoses, CSF cell count, and CMV serology are indicated intable 1.

Sample preparation. DNA preparation from CSF cells was performed as previously described.¹⁵ The number of CSF cells was determined by light microscopy using a Fuchs-Rosenthal chamber. Samples containing RBCs were not tested to exclude leakage of CMV-positive peripheral blood leukocytes to the CSF. Cells from 0.5 to 3 mL of CSF were resuspended in 15 to 30 µL lysis buffer, containing 50 mM KCl, 10 mM Tris hydrochloride (pH 8.3), 2 mM MgCl₂ 0.5% Tween 20, and 20µg/mL of proteinase K. Cell lysates were stored at -20 C until processed by PCR.

PCR method. Gene amplification was carried out as nested PCR using two sets of oligodeoxynucleotide primers (P1-P4) specific for the glycoprotein B gene of CMV.¹⁶ Nested PCR yielded a final product of 183 bp. Alternatively, external primers only were used for amplification of a 268-bp fragment to reduce sensitivity of the protocol.

PCR was performed in an OmniGene thermal cycler (MWG-Biotech, Ebersberg, Germany). Primary PCR was carried out in a volume of 30 µL, and nested PCR was performed in a volume of 50 µL using 2.5 µL from the first amplification round as target DNA. The reaction mixtures contained 16 mM (NH₄)₂SO₄, 20 mM Tris hydrochloride (pH 8.5), 1.7 mM MgCl₂, 80 µM (100 µM) of each deoxynucleotide, 75 nM (150 nM) of each deoxynucleotide primer and 0.5 units (1.0 units) of Taq polymerase (AGS, Heidelberg, Germany). (Modifications in the second PCR are indicated in parentheses.) Primary PCR proceeded with 94 C for 3 minutes, followed by 40 cycles of 94 C for 50 seconds, 56 C for 45 seconds, 72 C for 50 seconds, and a final extension step at 72 C for 3 minutes. Nested PCR was performed with 25 cycles of 95 C for 45 seconds, 54 C for 45 seconds, and 72 C for 45 seconds, followed by 72 C for 3 minutes.

Ten microliters of amplified products was analyzed by agarose gel electrophoresis (3% agarose, containing 0.3 µg/mL ethidium bromide) and visualized under ultraviolet light. To avoid contamination, all recommended precautions were taken, and negative controls were included in each reaction.

Quantitation of CMV genomes. Quantitation of viral copies was performed as previously described.^15,17 Aliquots and serial dilutions of CSF cell lysates containing DNA from a known number of cell equivalents were tested in subsequent sets of 8 to 10 identical PCR reactions. Our aim was to obtain a maximum of two thirds of positive amplifications to make sure that one positive PCR is equivalent to a single viral copy. Viral load was calculated using the number of cell equivalents present in sets of reactions with up to 6 of 10 or up to 5 of 8 positive PCRs, respectively. Reaction sets with a higher number of positive results were not taken into consideration to exclude samples that contained more than one single target molecule. As can be expected from a Poisson distribution, even in the presence of 50% positive PCRs, two viral genomes are present before amplification in approximately one third of total PCR reactions.¹⁸ Thus, the calculated cellular infection rate should be higher than indicated below. The viral load was expressed as approximate number of CMV copy per cell equivalents and adjusted to the number of CMV genomes per 10⁵ cells.

Sequence determination. CMV PCR products from several reactions were assessed by nucleotide sequence analysis. Fragments were purified using Qiaquick spin columns (Qiagen, Düsseldorf, Germany). Sequencing was carried out by the fluorescent dideoxy chain termination method using an automated DNA sequencer (ABIPrism 310, Applied Biosystems, Weiterstadt, Germany). Primers P1 or P3 were used as sequencing primer.

Results. Detection of CMV DNA. CMA DNA amplification was positive in CSF specimens from all patients with AIDS and from 5 of 12 HIV-1-seronegative individuals. Of 24 patients with detectable CMV, DNA CMV serology was positive in 13 patients and negative in 6 patients. From five subjects, IgG titers were not available. CMV DNA was consistently undetectable in all internal contamination controls.

The specificity of the PCR protocol was verified by direct nucleotide sequence analysis of PCR products obtained from several patients. Sequencing confirmed amplification of authentic CMV DNA.

The sensitivity of the nested PCR assay was assessed using serial dilutions of the 6,7 kilobase plasmid PQ 4 (kindly provided by J.E. Kühn, Department of Virology, University of Cologne, Germany). We were able to detect single copies of target DNA, even in the presence of 0.5µg genomic DNA (data not shown). Using external primers only for gene amplification, the sensitivity of the PCR protocol was reduced to a detection limit of approximately 500 copies of plasmid DNA in the presence of 0.5µg genomic DNA (data not shown).

Quantitation of CMV DNA. Titration analysis of CSF samples is shown in table 1. In cells of CSF obtained from five AIDS patients with proven CMVE, viral genomes were detected at a median value of 3,333/10⁵ (range, 1,667 to 5,333/10⁵ cells; mean, 3,558/10⁵ cells) compared with a median value of 125/10⁵ cells (range, 9 to 1,000/10⁵ cells; mean, 281/10⁵ cells;p < 0.01) for the 14 AIDS patients without clinical evidence of CMVE and a median value of 19/10⁵ cells (range, 0 to 562/10⁵ cells; mean, 52/10⁵ cells; p < 0.01) for the 12 HIV-1-seronegative control subjects. Significance was determined by means of Wilcoxon's two-sample rank test. All five AIDS patients with high amounts of CMV genomes in CSF leukocytes and autopsy-confirmed CMVE had histopathologic signs of extensive necrotizing CMV infection of the CNS. The number of viral copies did not clearly differ in patients who received CMV-specific treatment at the time of CSF sampling, and there was no correlation between CSF pleocytosis and the amount of cellular CMV DNA.

PCR with reduced sensitivity. CSF cell lysates from 14 patients were additionally tested by the simplified single round amplification PCR protocol with a detection limit of approximately 500 CMV genomes(table 2). Assessment included samples from four patients with AIDS and autopsy-proven CMVE, five patients with AIDS without evidence of CMV-related CNS disease, and five HIV-1-seronegative control subjects. CMV DNA detection in the CSF was positive in two independent reactions in six cases, among them all subjects with CMVE and two AIDS patients with non-CMV-associated neurologic symptoms. Negative results were obtained in the CSF of three patients with AIDS and in all five seronegative control subjects.

Discussion. As shown in this study, quantitative assessment of CMV genomes in CSF cell lysates is a sensitive method to reliably allow diagnosis of CMVE in patients with AIDS. Quantities of CMV copies in CSF leukocytes ranged significantly higher in AIDS patients with autopsy-proven CMVE (median, 3,333/10⁵ cells) as compared with AIDS patients without evidence of CMV-related neurologic symptoms (median, 125/10⁵ cells) and a group of HIV-1-seronegative control subjects with various neurologic disorders (median, 19/10⁵ cells). Thus, individuals with active CMV disease of the CNS had viral copies greater than 1,500/10⁵ CSF cells, whereas in subjects without CMV-associated CNS disease, cellular CMV genomes were below or equal to 1,000/10⁵ CSF cells. In control subjects, CMV copies were below 100/10⁵ in all except one case. All patients with CMVE had low CSF cell counts, indicating that the higher CMV copy number in this group reflected the presence of increased levels of virus in the small number of cells rather than a larger percentage of infected cells within the CSF. In keeping with our findings, several studies performed in peripheral blood of AIDS patients revealed a correlation between the cellular CMV DNA burden and symptomatic CMV disease. High quantities of CMV DNA in peripheral blood leukocytes of AIDS patients were associated with visceral organ involvement,¹³ and higher levels of virus were found in WBCs of HIV-1-positive individuals with CMV-related retinitis as compared with patients without symptomatic CMV infection.¹⁹ Furthermore, immunocompromised patients with CMV viremia have significantly higher levels of viral DNA in leukocytes.¹⁴

Previous studies mostly used total or cell free CSF to correlate the presence of virus with symptomatic CMV infection of the nervous system in patients with AIDS.^5-12 Quantitative assessment of CMV copies within CSF cells has the potential to differentiate latent from productive viral replication and therefore may increase diagnostic sensitivity and specificity of CMV-related neurologic manifestations. As revealed by our observations, patients with ongoing intracerebral CMV infection are more likely to have higher levels of intracellular viral genomes and may be identified early in the disease course where conventional CMV PCR may fail to detect CMV DNA in CSF.^5,9,20 Alternatively, the absence of high quantities of viral copies in a patient with positive CMV DNA detection in CSF may reflect nonspecific gene amplification and may argue against active CMVE. Furthermore, in peripheral blood, quantitative monitoring of CMV DNA in leukocytes has proved more useful to monitor efficacy of antiviral treatment than plasma CMV DNA levels.²¹

In our study, the high levels of cellular viral genomes in patients with autopsy-proven CMVE coincided with the presence of marked histopathologic changes. Similar findings were reported in few studies that addressed quantitation of CMV genomes in total or cell free CSF and brain of AIDS patients with neurologic CMV-related symptoms.^16,20,22,23 High levels of virus were unequivocally indicative of active intracerebral CMV replication and correlated with the presence of extensive necrotizing ventriculoencephalitis rather than minor CMV-induced lesions.^20,22

Five of 12 HIV-1-seronegative control subjects had detectable CMV copies in CSF cells. CSF lysates harbored small numbers of viral genomes in four individuals (Patients 22, 27, 28, and 31) and a higher burden of 562/10⁵ cells in one subject (Patient 29). CMV serology was negative in four cases and positive in one case. Similarly, detection of CMV DNA was reported in peripheral blood monocytes/macrophages of HIV-1-seronegative persons with negative CMV serology.²⁴ Our observations contradict those by Studahl et al.,²⁵ who detected CMV DNA in either CSF cells or CSF supernatant in a small number of immunocompetent patients with likely CMV infections of the CNS. CMV DNA was not found in CSF of 59 asymptomatic CMV-seropositive patients and in cells and supernatant of only one of three patients with bacterial meningitis and very high CSF cell counts. These differences are most likely explained by the higher sensitivity of our PCR protocol, which allowed identification of single CMV copies in latently infected cells.

To test whether patients with definite CMVE could be reliably identified using a cutoff value of cellular CMV, DNA CSF specimens from a subset of patients were additionally assessed by nonquantitative single amplification round PCR with a detection limit of approximately 500 CMV genomes. CMV DNA detection was repeatedly positive in CSF cell lysates of all patients with proven CMVE, in two of five AIDS patients without evidence of intracerebral CMV involvement, and in none of five HIV-1-seronegative control subjects. Viral loads of AIDS patients with CMV-unrelated symptoms and with positive CMV DNA detection were 1,000/10⁵ cells and 541/10⁵ cells and ranged higher than the average CSF cellular CMV burden of the remaining subjects of this group who were included in that part of the study. Both individuals had a diagnosis of cerebral toxoplasmosis and responded to specific therapy. We cannot exclude, however, intracerebral coinfection with CMV. The number of CSF samples tested by modified single round amplification PCR was too small to draw definite conclusions as to its diagnostic validitiy for CMVE. Because the sensitivity is influenced by the number of CSF cells in a given sample and the quantity of total CSF available for PCR, a reliable differentiation of latent from productive CMV infection may not always be possible. Results indicate that this approach may rapidly identify patients at risk for CMV-related neurologic manifestations who should be further tested by quantitative assessment of CSF cellular viral burden.

Acknowledgment

We are grateful to Dr. F. Huber, Service d'Infectiologie, Hôpital Claude Bernard Bichat, Paris, France, for providing CSF samples of Patients 3, 4, and 5.