Evaluation of the Murex Hybrid Capture Cytomegalovirus DNA Assay versus Plasma PCR and Shell Vial Assay for Diagnosis of Human Cytomegalovirus Viremia in Immunocompromised Patients

Abstract

We evaluated a cytomegalovirus (CMV) 24-hour shell vial assay (SVA), the Murex Hybrid Capture CMV DNA assay (HCA), and a CMV plasma PCR for the detection of CMV viremia in renal and bone marrow transplant recipients and human immunodeficiency virus-infected patients. CMV viremia was detected by at least one method in 125 of 317 evaluable samples (39.4%) from 78 patients and was detected in 19.8% of samples by SVA, 26.8% by HCA, and 32.2% by plasma PCR. There was moderate to substantial agreement between the results of the different tests (kappa coefficient = 0.415 to 0.631). However, HCA and plasma PCR were significantly more sensitive than SVA (P = 0.001 and P < 0.0001, respectively; McNemar’s test), and plasma PCR was more sensitive than HCA (P = 0.031; McNemar’s test). HCA and plasma PCR were more consistently positive than SVA during viremic episodes (P = 0.0002 and P < 0.0001, respectively; McNemar’s test). The use of HCA or plasma PCR may therefore improve the diagnosis and management of CMV disease in susceptible patient groups.

Human cytomegalovirus (CMV) infection is responsible for significant morbidity and mortality in immunocompromised patients, including bone marrow and solid organ allograft recipients and AIDS patients. Early detection of CMV infection may guide the initiation of appropriate antiviral therapy. In addition, quantification of the CMV load may allow the prediction of disease relapse and the identification of drug-resistant virus.

Laboratory diagnosis of CMV infection is based primarily on the detection of CMV viremia. This may be achieved by accelerated culture methods such as the shell vial assay (SVA) (7). Recently, detection of CMV pp65 antigen in polymorphonuclear leukocytes has proved to be more sensitive (2, 5), but it is dependent on the rapid processing of blood samples (14) and is therefore unsuitable for use in laboratories such as ours which perform tests for geographically remote clinical units. There is also much interest in the detection and quantification of CMV DNA in leukocytes (6, 11–13, 16, 18) or plasma (16, 17, 19, 21) as a measure of viremia and viral load. Although potentially sensitive, PCR methods are poorly standardized at present (8), and the clinical significance of DNA detection by PCR has not been fully evaluated. In addition, PCR assays may be subject to false-positive results due to contamination of the reaction or false-negative results due to sample processing failure or the presence of enzymatic inhibitors in some samples.

The recently developed Hybrid Capture CMV DNA assay (HCA) is a quantitative DNA hybridization test. This may avoid problems of PCR contamination and inhibition. Use of a standardized assay may also aid multicenter studies and interlaboratory comparison of results. The aim of this study was to compare three methods for the detection of CMV infection: the CMV HCA, a 24-h SVA, and an in-house plasma PCR. These assays were evaluated with specimens from bone marrow transplant and renal transplant recipients and patients with AIDS.

MATERIALS AND METHODS

Patients and specimens.

Samples were obtained from 50 transplant recipients (32 renal transplant recipients and 18 allogeneic bone marrow transplant recipients). Paired heparinized and EDTA-anticoagulated blood samples were collected at weekly intervals during the first 3 months posttransplantation and thereafter when patients presented with symptoms consistent with CMV disease. One hundred ninety-eight specimens were received from the 50 transplant patients. Samples were also obtained from human immunodeficiency virus type 1-positive patients with a previous AIDS-defining illness and a CD4 count of less than 50 cells/mm³. Samples were collected at monthly intervals from outpatients and at weekly intervals from inpatients. One hundred sixty-seven specimens from 28 AIDS patients were received. All samples were tested by SVA, which is the standard diagnostic method for the detection of CMV viremia in our department, and the results were reported to clinicians. Samples for HCA and plasma PCR were tested retrospectively without knowledge of SVA results.

Twenty-four-hour SVA.

To 10 ml of heparinized blood was added a 1/10 volume of a 6% dextran solution. Blood tubes were incubated at a 45° angle for 10 to 15 min at 37°C. The supernatant (1.5 ml), which contained dextran-enriched leukocytes, was aspirated and the cells were washed with sterile phosphate-buffered saline. The cell pellet was resuspended in 1 ml of maintenance medium and inoculated onto shell vials containing semiconfluent HEL cell monolayers by centrifugation at 15,000 × g for 30 min at 4°C. After 18 to 24 h of culture at 37°C, the monolayers were fixed in acetone and were stained by direct immunofluorescence with fluorescein isothiocyanate-conjugated monoclonal antibody E13 (Tissue Culture Services, Bucks, United Kingdom) directed against a CMV immediate-early antigen; E13 was diluted 1:40 in phosphate-buffered saline–Evans Blue. A positive control consisting of cell culture-propagated CMV strain AD169 was included in each test batch.

HCA.

CMV genomic DNA was extracted from whole blood and was quantified by HCA (version 1; Murex Diagnostics Ltd., Dartford, United Kingdom) according to the manufacturer’s instructions. Briefly, leukocytes were recovered from 3.5 ml of whole EDTA-anticoagulated blood by two rounds of erythrocyte lysis and were then denatured and hybridized in solution to a CMV-specific complementary RNA probe. Hybrids were captured with a solid-phase bound monoclonal antibody specific for DNA-RNA hybrids and were detected with the same monoclonal antibody conjugated to alkaline phosphatase and a chemiluminescent substrate. Positive samples were defined as those giving twice the mean value for the negative control, and the genome copy number of positive samples was estimated by reference to three positive standards supplied with the assay.

Plasma PCR.

Total DNA was extracted from 200-μl volumes of EDTA-anticoagulated plasma with the QIAamp blood kit (Qiagen) and was reconstituted in 200-μl volumes according to the manufacturer’s instructions. For PCR, 5 μl of this DNA extract was amplified with primers 627:5′ and 459:5′ by the method of Wolf and Spector (21), with minor modifications. The sequence of primer 459:5′ was modified by omitting the two terminal 3′ nucleotides (5′-GGC AGC TAT CGT GAC TGG-3′) to reduce the formation of nonspecific amplification products. Primer 627:5′ was unchanged (5′-GAT CCG ACC CAT TGT CTA AG-3′). These primers amplify a target region of 152 bp from EcoRI fragment D, corresponding to nucleotides 135176 to 135326 of the CMV AD169 genome (1). All reactions were hot started by using a wax barrier to separate Taq polymerase and template from primers and deoxynucleoside triphosphates prior to thermocycling to further reduce nonspecific amplification. Amplification was performed in 50-μl reaction volumes of PCR Buffer II (Perkin-Elmer) containing 1.5 mM MgCl₂, 0.2 mM each deoxynucleoside triphosphate, 25 pmol of each primer, 1.25 U of AmpliTaq DNA polymerase (Perkin-Elmer), and 5 μl of DNA extract. Thermal cycling consisted of 35 cycles of denaturation (94°C, 1 min), primer annealing (55°C, 1 min), and primer extension (72°C, 1 min) and was performed with a Perkin-Elmer 480 thermal cycler. PCR products were visualized by agarose gel electrophoresis. Three positive controls, consisting of 55, 5.5, and 0.55 pg of CMV genomic DNA (Sigma Chemical Co.), corresponding to approximately 200,000, 20,000, and 2,000 genome equivalents, respectively, were included in each test batch. All had to be positive for the test run to be validated. Negative controls consisted of DNA extraction blanks and PCR reagent blanks in which sterile water or human placental DNA (Sigma Chemical Co.) was substituted for the test sample. Initially, the PCR product was confirmed as CMV by direct cycle sequencing with dye terminators and an ABI 377 DNA sequencer.

Statistical analysis.

The strength of agreement between the results of the different assays was measured by the κ test, with a κ coefficient of >0.4 indicative of moderate agreement and a κ coefficient of >0.6 indicative of substantial agreement. The significance of discordance was measured by McNemar’s test with Yate’s correction. Differences in the quantitative results of HCA among different sample groups were assessed by the Student t test.

RESULTS

Three hundred sixty-five samples from 78 patients were tested by all three methods. Twenty-five of the 78 patients (32.1%) remained negative by all tests during follow-up. Forty-eight samples were excluded from analysis because of test failures; for 44 samples (12.1%) SVA failed due to specimen toxicity, and for 4 samples (1.1%) the PCR results could not be interpreted due to the presence of nonspecific amplification products. Of the 317 remaining samples, 232 (73.2%) gave concordant results by all three tests, and 85 (26.8%) gave discordant results; i.e., positive results were obtained by only one or two of the tests (Fig. 1). Of these 85 samples giving discordant results, 76 were from 37 patients for whom at least one other previous or subsequent sample tested positive by at least one test. Only nine samples giving discordant results were from patients for whom no other sample tested positive by any method. Most of these samples tested positive by PCR only. The relationship between the results of the three assays was evaluated further by pairwise comparisons (Table 1).

FIG. 1.

Agreement between SVA, HCA, and PCR results.

TABLE 1.

Pairwise comparison of assays for CMV viremia

Assay comparisonb	% Con- cordance	% Discordance		P valuea	κ coeffi- cient
		Positive/negativec	Negative/positived
PCR vs SVA	77.0	17.7	5.4	<0.0001	0.415
PCR vs HCA	82.6	11.4	6.0	0.031	0.584
HCA vs SVA	86.8	10.1	3.2	0.001	0.631

^aMcNemar’s test. 

^bFirst assay versus second assay. 

^cPositive first assay result, negative second assay result. 

^dNegative first assay result, positive second assay result. 

During episodes of CMV viremia, defined as periods during which at least two consecutive samples tested positive by any method, SVA was less consistently positive than HCA or PCR. Of 102 samples taken during 26 viremic episodes from 25 patients, 53 (52.0%) samples tested positive by SVA, 70 (68.6%) tested positive by HCA, and 80 (78.4%) tested positive by PCR. Most discrepant results were observed at the beginning and end of viremic episodes. The sensitivity of SVA during viremic episodes was significantly lower than that of HCA (P = 0.0002; McNemar’s test) or PCR (P < 0.0001).

There was no significant difference between the viral loads of HCA-positive samples which were SVA or PCR positive and the viral loads of HCA-positive samples which were SVA or PCR negative (Table 2).

TABLE 2.

Relationship between viral load of HCA-positive samples and results of SVA and plasma PCR

Test and result	No. of HCA-positive samples	Mean viral load (no. of genomes/ ml [105])	95% confidence interval	Statistical significance (P)
SVA
Positive	53	2.91	±1.95 × 105
Negative	32	1.77	±1.29 × 105	0.36
PCR
Positive	66	2.18	±9.47 × 104
Negative	19	3.19	±5.15 × 105	0.51

DISCUSSION

Rapid methods for the detection of CMV viremia offer the prospects of an improved means of diagnosis and improved clinical management of CMV disease in immunocompromised patients. However, there is a lack of consensus regarding the most appropriate test and a lack of standardization of laboratory tests. Use of commercial assays may contribute to standardization. We have therefore evaluated the HCA for the detection of viremia and compared the results obtained by HCA with those obtained by SVA, a commonly used laboratory method, and plasma PCR, a potentially useful gene amplification method.

Although there was broad agreement between test results, this study revealed some disparity between the results of the three assays. Plasma PCR and HCA were significantly more sensitive than SVA. Plasma PCR was also significantly more sensitive than HCA. Similar findings were recently reported by Hebart et al. (9). During episodes of CMV viremia, SVA was less consistently positive than HCA or PCR with consecutive follow-up samples. Most samples giving discordant results were obtained from patients for whom samples collected earlier or later also tested positive by at least one test. It is therefore probable that the majority of positive results are true positives, indicative of CMV viremia, and that discordant results reflect differences in the relative sensitivities of the assays rather than low specificities. However, the sensitivities of HCA and SVA, which detect leukocyte-associated virus, may have been affected by variations in blood leukocyte counts. Additionally, SVA detects only actively replicating virus, and its sensitivity may have been adversely affected by delays in specimen transport and processing. The use of an additional shell vial culture stained after 48 h of incubation may have increased the sensitivity of SVA.

Quantitation of CMV viremia may also play a role in predicting CMV disease and monitoring therapy. Although there was moderate agreement between the results of HCA and the results of SVA or PCR in this study, HCA-positive samples which were SVA or PCR negative did not have significantly lower viral DNA loads than samples which were SVA or PCR positive. This may indicate that the viral DNA load in whole blood as measured by HCA is not directly related to viral infectivity as measured by SVA or the presence of viral DNA in plasma as measured by PCR. CMV DNA in leukocytes may appear earlier and persist longer than CMV DNA in plasma. It is also possible that the presence of substances inhibitory to the PCR gave rise to false-negative PCR results for blood samples which were HCA positive. In contrast, others have found that the level of CMV DNA determined by HCA correlates with the level of pp65 antigenemia (15, 20).

HCA is a potentially useful means of detecting and quantifying CMV viremia. The test format is standardized, and the test is not subject to contamination or inhibition. The results are objective and quantitative. Blood samples can be stored for 6 to 8 h and then processed, stored, and batch tested. Plasma PCR may also be useful for the detection of CMV viremia. The in-house assay used here requires small sample volumes, is technically straightforward, and has a low cost. We elected to evaluate CMV plasma PCR rather than leukocyte PCR because others have found that the results of plasma PCR correlate with the results of antigenemia testing and the presence of CMV disease (9, 17, 19). In contrast, several groups have found that leukocyte PCR more frequently gives positive results for patients without evidence of CMV disease (3, 4), although quantitative leukocyte PCR may be more useful (6). A commercial CMV plasma PCR test has recently become available (10), and this should allow standardization between laboratories.

It will be important to correlate the results of these assays with clinical information in order to determine the clinical value of these assays for the early detection and monitoring of CMV infection and disease. This work is now in progress in a prospective study.