Comparison of Three Nucleic Acid Amplification Assays of Cerebrospinal Fluid for Diagnosis of Cytomegalovirus Encephalitis

Abstract

The diagnostic reliabilities of three cytomegalovirus (CMV) nucleic acid amplification assays of cerebrospinal fluid (CSF) were compared by using CSF samples from human immunodeficiency virus-infected patients with a postmortem histopathological diagnosis of CMV encephalitis (n = 15) or other central nervous system conditions (n = 16). By using a nested PCR assay, the quantitative COBAS AMPLICOR CMV MONITOR PCR, and the NucliSens CMV pp67 nucleic acid sequence-based amplification assay, sensitivities were 93.3, 86.6, and 93.3%, respectively, and specificities were 93.7, 93.7, and 87.5%, respectively. The COBAS AMPLICOR assay revealed significantly higher CMV DNA levels in patients with diffuse ventriculoencephalitis than in patients with focal periventricular lesions.

Human cytomegalovirus (CMV)-induced complications of the central nervous system (CNS) in patients with human immunodeficiency virus (HIV) infection mainly consist of encephalitis (CMV-E) and polyradiculomyelitis. Histologically, patients with CMV-E can present with ventriculoencephalitis, focal parenchymal lesions, or microglial nodules; in addition, isolated cells containing typical CMV inclusions can also be recognized (1). Although diffuse encephalopathy is the clinical correlate of CMV-E, its clinical presentation may not be specific and a definitive diagnosis requires the identification of the virus in the brain or cerebrospinal fluid (CSF). In this regard, PCR of CSF for the detection of CMV DNA is the most reliable method for the diagnosis of CMV-related CNS lesions in vivo in HIV-infected patients, with most of the studies indicating sensitivity and specificity values of 80 and 100%, respectively (4–6, 8–10, 16). The aim of the study described here was to compare the diagnostic reliabilities of two new commercially available assays for the detection of CMV genomes in peripheral blood, the COBAS AMPLICOR HIV Monitor PCR (the COBAS AMPLICOR PCR) and the NucliSens CMV pp67 nucleic acid sequence-based amplification assay (NASBA), with that of an in-house nested PCR test for CSF analysis (4, 5).

(The content of this article was presented at the Neuroscience of HIV Infection 2000, Edinburgh, United Kingdom, 22 to 24 June 2000.)

Patients belonged to a cohort of HIV-infected patients with CNS symptoms admitted to the San Raffaele Hospital, Milan, Italy, between 1992 and 1998. Thirty-one patients with or without CMV encephalitis were selected because of the availability of both postmortem tissues and a CSF sample that had been taken during the 180 days preceding death and that had been stored at −80°C and never previously thawed. Six patients (four with CMV-E and two with other CNS diseases) were included in a previous study (5). Upon admission, the patients underwent a complete neurological and neuroradiological examination and CSF analysis. Additional CSF specimens were drawn as necessary during the subsequent follow-up. All of the patients who died in hospital underwent postmortem examination (5). At autopsy, CMV-E was defined as the presence of inflammatory and/or necrotizing brain lesions, with CMV identified by histology on the basis of typical viral inclusions and confirmed by immunochemistry with a monoclonal mouse anti-CMV antibody (clone CCH2, diluted 1:25; Dakopatts, Glostrup, Denmark).

The nested PCR assay was performed as described previously (4). This test was used to examine the CSF at the time of the diagnostic workup but was repeated for the present study at the same time of the other two assays and using the same stored CSF samples. CSF samples were tested by the COBAS AMPLICOR test with a 200-μl volume and by the manufacturer's recommendation for analysis of plasma (13, 14). For the Organon Teknika NucliSens CMV pp67 NASBA, an initial CSF sample volume of 100 μl was used and the analysis was performed as recommended for analysis of blood (2, 3).

At neuropathological examination, 15 patients had CMV-E and 16 patients had other CNS conditions. All of the patients with CMV-E had ventriculoencephalitis, and 11 also had focal parenchymal lesions. Further details of the postmortem findings are shown in Tables 1 and 2. The analytical sensitivity of the nested PCR and the COBAS AMPLICOR test, evaluated by using 10-fold dilutions of purified full-length CMV Towne DNA, showed a lower detection limit of 100 copies/ml for the nested PCR and 1,000 copies/ml for the COBAS AMPLICOR PCR. By the nested PCR, the COBAS AMPLICOR PCR, and NASBA, the diagnostic sensitivities were 93.3, 86.6, and 93.3%, respectively, and the specificities were 93.7, 93.7, and 87.5%, respectively. In five patients discordant CSF and histopathology results were found by one or more of the assays (Table 1, patients 14 and 15; Table 2, patients 16 to 18). By the nested PCR, the results of the retrospective and prospective analyses were identical, showing interassay reproducibilities of 100%.

TABLE 1.

Histopathological and CSF findings for 15 patients with CMV-E

Patient no.	CMV ventriculo-encephalitisa	CMV parenchymal lesions (localization)	CMV infected cell scoreb	Other CNS findingsc	Other CMV localizationsc	Anti-CMV treatment ongoing at the time of CSF samplingd	Interval from CSF sampling to death (days)	Result by nested CMV PCR of CSF	No. of CMV copies/ml of CSF by COBAS AMPLICOR PCR	Result by NucliSens CMV pp67 NASBA of CSF
1	Diffuse	Multiple (brain, cerebellum, brain stem)	+++	None	Lung, adrenal glands	None	58	Positive	>100,000	Positive
2	Diffuse	None	+++	HSV-1e encephalitis	Retina	None	20	Positive	>100,000	Positive
3	Diffuse	Single (brain)	++	HSV-2f encephalitis	None	None	47	Positive	>100,000	Positive
4	Diffuse	Multiple (brain, cerebellum, brain stem)	++	Non-Hodgkin's lymphoma	None	Ganciclovir	93	Positive	>100,000	Positive
5	Diffuse	Multiple (brain, cerebellum, brain stem)	+++	None	Adrenal glands, retina	None	37	Positive	99,800	Positive
6	Diffuse	None	++	None	Adrenal glands	None	150	Positive	70,300	Positive
7	Diffuse	Multiple (brain stem)	+++	None	Retina	Ganciclovir	4	Positive	60,900	Positive
8	Diffuse	Multiple (brain, cerebellum, brain stem)	+++	None	Lung, adrenal glands, retina	Foscamet	3	Positive	47,800	Positive
9	Diffuse	Multiple (brain)	++	HIV encephalitis	Retina	Ganciclovir	42	Positive	44,900	Positive
10	Diffuse	Multiple (brain, cerebellum)	+	HSV-1 encephalitis	Adrenal glands, retina	Ganciclovir	29	Positive	15,100	Positive
11	Diffuse	None	+++	None	None	Ganciclovir	12	Positive	11,900	Positive
12	Diffuse	Multiple (brain, cerebellum, brain stem)	++	None	Lung, adrenal glands	Foscamet	178	Positive	6,400	Positive
13	Focal	None	+	Toxoplasmosis	None	None	8	Positive	1,900	Positive
14	Focal	Single (brain)	+++	None	Adrenal glands, retina	Ganciclovir	17	Positive	<400	Positive
15	Focal	Multiple (brain)	++	Mycobacterium tuberculosis meningoencephalitis	Lungs, brain, adrenal glands, esophagus	None	21	Negative	<400	Negative

^aDiffuse ventriculoencephalitis, present throughout the periventricular lining of a tissue section; focal ventriculoencephalitis, only small foci with aggregates of CMV-infected cells. 

^b+, 1 to 5 infected cells per 10 high-power fields; ++, 6 to 20 infected cells per 10 HPF high-power fields; +++, >20 infected cells per 10 high-power fields. 

^cAt postmortem examination. The retina was not examined at autopsy, and retinitis was diagnosed in vivo by ophthalmological examination. 

^dAnti-CMV drugs were administered between the time of CSF sampling and death for all but one patient (patient 15) (median time, 42 days). 

^eHSV-1, herpes simplex virus type 1. 

^fHSV-2, herpes simplex virus type 2. 

TABLE 2.

Histopathological and CSF findings for 16 patients without CMV infection of CNS^a

Patient no.	Other CNS findingsa	Other CMV localizationsa	Anti-CMV treatment ongoing at the time of CSF samplingb	Interval from CSF sampling to death (days)	Results by nested CMV PCR of CSF	No. of CMV DNA copies of CSF by COBAS AMPLICOR PCR	Result by NucliSens CMV pp67 NASBA of CSF
16	Nonspecific subependymal gllosis	Adrenal glands, colon, lungs	None	42	Positive	3,500	Negative
17	HIV encephalitis	Adrenal glands	None	5	Negative	<400	Positive
18	Cryptococcus neoformans meningoencephalitis	Adrenal glands	None	52	Negative	<400	Positive
19	Ischemic lesions	None	Ganciclovir	59	Negative	<400	Negative
20	Aspergillosis	Adrenal glands	Foscarnet, ganciclovir	36	Negative	<400	Negative
21	No lesions	Lung, adrenal glands, retina	Ganciclovir	28	Negative	<400	Negative
22	Unspecific myelitis	None	None	8	Negative	<400	Negative
23	HIV encephalitis	Retina	Ganciclovir	28	Negative	<400	Negative
24	No lesions	None	None	25	Negative	<400	Negative
25	Toxoplasmosis	None	None	46	Negative	<400	Negative
26	Crytococcus neoformans meningoencephalitis	Adrenal glands	None	19	Negative	<400	Negative
27	Edema	None	None	12	Negative	<400	Negative
28	No lesions	None	None	77	Negative	<400	Negative
29	Toxoplasmosis	Lung	None	38	Negative	<400	Negative
30	Non-Hodgkin's lymphoma	Lung, stomach, heart	None	43	Negative	<400	Negative
31	Non-Hodgkin's lymphoma	None	None	11	Negative	<400	Negative

^aAt postmortem examination. The retina was not examined at autopsy, and retinitis was diagnosed in vivo by ophthalmological examination. 

^bAnti-CMV drugs were administered between the time of CSF sampling and death for five patients (patients 16, 19 to 21, and 23) (median, 32 days). 

By the COBAS AMPLICOR PCR, the median CMV DNA load was 65,600 copies/ml (range, 6,400 to 100,000 copies/ml) for 12 patients with diffuse ventriculoencephalitis and <400 copies/ml (range, <400 to 1,900 copies/ml) for three patients with a focal form (P < 0.01 by the Mann-Whitney test). No significantly different distributions of the CMV DNA loads were observed according to the presence of associated parenchymal lesions, CMV-infected cell score, extracerebral CMV localization, other CNS diseases, or ongoing anti-CMV therapy, although untreated patients tended to have higher viral loads (Table 1).

The results obtained in this study by all three techniques closely correlated with the postmortem diagnoses and showed similar diagnostic sensitivity and specificity values. These values were similar to those obtained for two other patient series in which CSF findings were compared with the histopathological diagnoses (4, 5). Overall, the analytical sensitivities of the assays therefore appeared to be adequate to specifically distinguish patients with CMV-E. Furthermore, the results of the present retrospective evaluation by nested PCR were identical to those prospectively obtained by the same assay at the time of patient evaluation 2 to 8 years earlier, which had allowed a premortem diagnosis for the majority of patients. This indicates the high interassay reproducibility of this technique and the good rate of recovery of amplifiable genomes after longterm storage at −80°C.

The COBAS AMPLICOR PCR is a quantitative assay developed to measure CMV levels in plasma or peripheral blood leukocytes (13, 14). To our knowledge, only Roche's other commercially available PCR assay, the qualitative plasma AMPLICOR CMV test, but not the COBAS AMPLICOR PCR assay, has been described for analysis of CSF (11, 15). In the present study, in which we compared the findings obtained with CSF with the neuropathological findings, we found that the analytical and diagnostic sensitivities of the COBAS AMPLICOR PCR were only slightly lower than those of the nested PCR and NASBA, and the COBAS AMPLICOR test failed to recognize only one patient with focal ventriculoencephalitis. The threshold of 400 copies/ml, calculated as representing the best compromise between sensitivity and specificity in identifying HIV-infected patients at risk for the development of systemic CMV disease (12), therefore also seems to be useful for the identification of patients with CMV-E. Furthermore, by the COBAS AMPLICOR assay we found a significant association between high CSF CMV DNA levels and the presence of diffuse rather than focal periventricular lesions and found that the presence of focal parenchymal lesions did not seem to affect the viral load in CSF significantly. These findings are consistent with the hypothesis that the high CSF CMV DNA burden in patients with ventriculoencephalitis is probably due to the direct release of virus from the adjacent infected ependymal and subependymal tissues. From a practical point of view, the ability to discriminate extended from milder lesions may be important for prognostic considerations.

Whereas the nested PCR and the COBAS AMPLICOR PCR detect viral DNA, the NASBA detects late CMV transcripts. A recent evaluation of the NucliSens CMV pp67 NASBA with CSF from HIV-infected patients with a clinical diagnosis of CMV-associated or other opportunistic CNS diseases showed that the test was highly specific but less sensitive than PCR of DNA for the diagnosis of CMV-related neurological complications (17). The results of the present study, specifically designed to compare the results obtained with CSF with those obtained by histopathological examination, not only confirm the high degree of specificity of the NASBA but also show that its sensitivity is not less than that of the PCR assays with DNA.

Discordance between the findings obtained with CSF and postmortem findings were obtained for five patients. Positive PCR results in the absence of brain lesions caused by CMV can be explained by the possible effects of anti-CMV drugs administered between the time of sampling and death, by the inevitable limitations of neuropathological analysis to sample all the microscopic lesions, or by the possible presence of active CMV replication in the brain or CSF in the absence of established lesions. The detection in CSF of viral DNA originating from the blood might be another explanation for these discordant findings. However, previous reports indicated that CMV DNA is generally recovered in the CNS irrespective of CMV DNA detection in blood or of an extracerebral CMV involvement (4, 5, 10). We constantly used strict laboratory procedures to avoid contamination, and so the presence of exogenous nucleic acids in these CSF-positive samples with discordant results was unlikely. The time that elapsed between CSF sampling and postmortem diagnosis may have resulted in a negative CSF result in the presence of CMV-E. On the other hand, the presence of amplification inhibitors in the clinical specimens was unlikely because both of the commercial assays include amplification controls and no inhibition was ever observed.

In conclusion, the two commercial assays examined in the present study are useful additional tools for the diagnosis of CMV-E in HIV-infected patients. Furthermore, the COBAS AMPLICOR assay enables quantification of DNA in CSF, which not only is important for diagnostic and prognostic factors but might also provide precious information for monitoring of antiviral treatment (7). Although the prevalence of CMV-E in HIV-infected patients has decreased since the introduction of highly active antiretroviral therapy, we believe that monitoring of CMV infection of the CNS is still important because a number of patients cannot tolerate or fail to respond to these regimens. Furthermore, although not evaluated in the present study, these new tests might prove to be useful for the diagnosis of other CMV-associated neurological diseases, such as those observed in transplant recipients, neonates with congenital CMV infection, or immunocompetent patients (1).