Cytomegalovirus Cultured From Different MajorLeukocyte Subpopulations: Association With ClinicalFeatures in CMV Immunoglobulin G-Positive RenalAllograft Recipients

Peter Schafer,1* Werner Tenschert,2 Liana Cremaschi,2 Matthias Schroter,1 Kai Gutensohn,3 andRainer Laufs1

1Institut fur Medizinische Mikrobiologie und Immunologie, Universitats-Krankenhaus Eppendorf,Hamburg, Germany

2Urologische Klinik, Universitats-Krankenhaus Eppendorf, Hamburg, Germany3Abteilung fur Transfusionsmedizin, Universitats-Krankenhaus Eppendorf, Hamburg, Germany

Cytomegalovirus (CMV) cultured from periph-eral blood mononuclear cells (PBMCs) wasshown to be associated more closely with clini-cal manifestations than infectious CMV in poly-morphonuclear leukocytes (PMNLs) of renal al-lograft recipients with secondary CMV infection.Shell vial culture was carried out with ficoll-purified PBMCs and PMNLs of 71 CMV IgG-positive patients after kidney transplantation.Thirty-six patients experienced active CMV in-fections. Of these, 17 developed clinical symp-toms. The diagnostic value of PMNLs andPBMCs viremia was determined in comparisonto pp65 antigenemia, leukoDNAemia, plasmaDNAemia, and detection of cytomegalic endo-thelial cells. In both PMNLs and PBMCs (with orwithout detectable endothelial cells), frequen-cies and levels of viremia were significantlyhigher among symptomatic patients. Regardingthe occurrence of clinical CMV manifestations,the sensitivity of culture from PMNLs and fromPBMCs fractions was 100%. Viremia in PBMCs,however, was far more specific (94%) than inPMNLs (74%). Cutoff values established previ-ously for pp65 antigenemia and leukoDNAemia,standard markers in the laboratory, had similarspecificity (96% each) to PBMCs viremia, butwere less sensitive (88% each). Plasma DNA-emia was both less sensitive (82%) and less spe-cific (69%) than PBMCs viremia. Detection of en-dothelemia showed maximal specificity (100%),but inferior sensitivity (47%). All patients hadPBMCs viremia before the onset of symptoms.In conclusion, infectious CMV present in PBMCsmay prove to be a determinant of clinical CMVmanifestations in seropositive immunocompro-mised individuals. Factors involved in PBMCstropism may help to understand the pathoge-

netic mechanisms of CMV dissemination in thisgroup of patients. J. Med. Virol. 61:488–496,2000. © 2000 Wiley-Liss, Inc.

KEY WORDS: CMV disease; viremia; endothe-lial cells; antigenemia; DNA-emia; quantitative PCR

INTRODUCTION

Viral components (infectious particles, pp65 anti-gens, and DNA) circulating in the peripheral blood areaccepted widely as markers for the diagnosis of cyto-megalovirus (CMV) infections in immunocompromisedindividuals. Although virtually all immunosuppressedpatients with primary CMV infections are at risk forclinical complications, detection of the virus in patientswith preexisting CMV immunoglobulin G (IgG) anti-bodies does not necessarily indicate clinical CMV fea-tures [for reviews, see Schafer and Laufs, 1996; Boeckhand Boivin, 1998]. It is therefore important to find re-liable determinants of a symptomatic course of CMVinfection in these patients.

The level of circulating viral components is recog-nized increasingly as an important criterion for assess-ing the patients’ risk of clinical CMV manifestations.Thus, protocols for quantitative determination of pp65antigenemia, DNAemia, and viremia have been de-vised. These techniques are useful in terms of predict-ing symptomatic CMV infections [Schafer and Laufs,1996; Boeckh and Boivin, 1998]. From the diagnostic

*Correspondence to: Peter Schafer, Institut fur MedizinischeMikrobiologie und Immunologie, Universitats-Krankenhaus Ep-pendorf, Martinistr. 52, D-20246 Hamburg, Germany.E-mail: pschaefe@uke.uni-hamburg.de.

Accepted 10 January 2000

Journal of Medical Virology 61:488–496 (2000)

© 2000 WILEY-LISS, INC.

standpoint, however, the importance of virus culturehas diminished compared with pp65 antigenemia andPCR, because results are available by shell vial cultureonly after 1–2 days, and the sensitivity has been re-ported to be limited [Erice et al., 1992].

Despite the progress made in recent years, misjudg-ment of the outcome of patients still occurs. Most im-portant, there are patients who have only low levelantigenemia and DNAemia but nevertheless developCMV disease [Gerna et al., 1994]. Thus, it is possiblethat other factors, apart from the absolute quantity ofcirculating viral components, should be associated withclinical CMV features.

In this context, increasing attention is being paid tothe nature of viral interactions with blood compart-ments. It is known that pp65 antigens and CMV DNAare predominantly harbored in polymorphonuclearcells (PMNLs) during active CMV infection [Saltzmanet al., 1988; Grefte et al., 1992; Gerna et al., 1992b;Hackstein et al., 1996; Schafer et al., 1998a,b]. On theother hand, plasma DNAemia, although its exact char-acter is defined poorly, is said to indicate active CMVinfection [Boeckh and Boivin, 1998] and seems to re-flect approximately high-level leukoDNAemia [Zipetoet al., 1995].

Regardless of the decreasing usage of CMV culture inthe diagnostic laboratory, infectious virus present inthe peripheral blood plays a crucial role in virus dis-semination. CMV has been isolated with highly vari-able frequency from both PMNLs and peripheral bloodmononuclear cell (PBMCs) subsets of leukocytes pre-pared by ficoll density gradient centrifugation [Rinaldoet al., 1977; Howell et al., 1979; Saltzman et al., 1988;Erice et al., 1992; Gerna et al., 1992b]. In polymorpho-nuclear cells, the presence of infectious CMV is attrib-uted to virus uptake from surrounding sites of replica-tion [Gerna et al., 1992b; Grefte et al., 1994; Hacksteinet al., 1996]. In PBMCs fractions, however, there areadditional possible origins of infectious virus. First,productive infection of mononuclear leukocytes hasbeen reported [Larsson et al., 1998]. Second, circulat-ing cytomegalic endothelial cells, that are fully permis-sive of CMV infection, have been identified recently inPBMCs fractions of a proportion of immunosuppressedindividuals. These cells are strongly associated withend-organ disease. From the diagnostic point of view,however, the significance of cytomegalic endothelialcells is limited because their detection is highly labori-ous and is also of low sensitivity [Percivalle et al., 1993;Grefte et al., 1993a,b; Salzberger et al., 1997; Gerna etal., 1998].

To date, no information except for endothelial cells isavailable on the association of infectious CMV presentin different leukocyte populations with the clinical out-come. Therefore, the aim of our study was to correlateshell vial culture results from the major PBL subpopu-lations of CMV IgG-positive renal-allograft recipientswith clinical manifestations. In addition, the signifi-cance was evaluated in comparison to other establisheddiagnostic markers (endothelemia, antigenemia, and

DNAemia). Unlike PMNLs, mononuclear leukocytesharboring infectious CMV indicated clinical features.

MATERIALS AND METHODSPatients

Seventy-one hospitalized CMV IgG-positive renal-allograft recipients were monitored up to 6 monthsposttransplantation. Only patients without current an-tiviral treatment at the time of first investigation wereenrolled in the study. Patients without evidence of ac-tive CMV infection who were treated prophylacticallywith ganciclovir during steroid-resistant graft rejectiontherapy (anti-T-lymphocyte globulin or OKT3) were ex-cluded as well. Samples were collected twice a weekfrom all patients during the primary admission to hos-pital (between 21 and 50 days after transplantation).From 7 patients samples were drawn every 2 days for2 weeks upon readmission due to active CMV infectionsbetween days 75 and 150 posttransplantation. Allsamples were analyzed prospectively.

Diagnostic and Therapeutic Criteria

Active CMV infection was confirmed by positive pp65antigenemia, CMV culture from PBL, detection of cy-tomegalic endothelial cells, or culture from organ biop-sies (see below). Qualitative CMV DNAemia alone wasnot rated as active infection. CMV syndrome was diag-nosed according to criteria described recently [Ljung-man and Plotkin, 1995] by at least two of the following:unexplained fever >38°C for >3 days, arthralgia, hy-perhidrosis, graft dysfunction without histological evi-dence of rejection, leukopenia, thrombocytopenia, orliver enzyme elevation. CMV disease was diagnosedwhen organ involvement was confirmed (pneumonitisor gastrointestinal ulceration with virus isolation frombiopsies [Ljungman and Plotkin, 1995]).

Patients with symptomatic CMV infection at admis-sion or experiencing laboratory-confirmed active CMVinfection combined with steroid-resistant graft rejec-tion were treated with ganciclovir for a minimum of 14days at 10 mg/kg of body-weight daily adjusted for re-nal function. In addition, cutoff levels of quantitativepp65 antigenemia ($50 positive cells/2 × 105 leuko-cytes) and leukoDNAemia ($1,000 genome equiva-lents/106 copies b-globin) in mixed PBL (see below)were used during CMV monitoring as criteria for ini-tiation of preemptive antiviral treatment before onsetof symptoms. In previous investigations, these cutofflevels had been shown to be highly associated with asymptomatic course of CMV infection [Kuhn et al.,1994; Schafer and Laufs, 1996].

Leukocyte Isolation

Four hundred ninety-six blood samples from the 71patients were investigated. The laboratory methodsused in this study were carried out with mixed PBLsand with PMNLs and PBMCs subsets. To avoid labo-rious and cost-intensive cell sorting techniques, ficoll-metrizoate density centrifugation was used for separa-tion of PMNLs and PBMCs. Thus, when interpreting

Infectious CMV in Leukocyte Subsets 489

the culture results it must be considered that PMNLsfractions were contaminated with about 5% PBMCs(monocytes and lymphocytes) on average, and thatPBMCs contained monocytes and lymphocytes, about5% contaminating PMNLs, and cytomegalic endotheli-al cells (when present).

Ten milliliters of whole blood were collected in EDTAtubes and processed within 1 hr post-drawing. Differ-ential leukocyte counts were obtained with a hemato-logical cell counter. Total leukocytes were isolated bycentrifugation of 2 ml of whole blood at 700 × g for 3min and subsequent lysis of erythrocytes with 0.8%ammonium chloride. Pelleted leukocytes were washedwith phosphate-buffered saline (PBS), pH 7.4, andquantitated. PMNLs and PBMCs were semi-purifiedby ficoll-metrizoate density centrifugation using Poly-morphprep and Lymphoprep (Nycomed, Oslo, Norway)according to the manufacturer’s instructions, andcounted as well. Fractions were adjusted with PBS tothe concentration equivalent to 107 total leukocytes perml, respectively.

Virus Culture

Quantitative shell vial culture was carried out essen-tially as described [Storch et al., 1994]. MRC-5 cell(Boehringer Ingelheim BioWhittaker, Verviers, Bel-gium) monolayers grown in shell vials were inoculatedwith 100-ml aliquots (equivalent to 106 total cells) ofleukocyte subsets by centrifugation at 700 × g for 45min. After incubation at 37°C for 48 hr the cells werefixed and permeabilized with 5% paraformaldehyde–0.5% Nonidet-P40 (Sigma, Munchen, Germany). In-fected fibroblasts were quantitated by indirect immu-nofluorescence staining as described previously [Gernaet al., 1992a; Storch et al., 1994] using a monoclonalantibody (9221; DuPont de Nemours, Bad Homburg,Germany) directed against the CMV immediate earlyantigen p72, and counting the positive nuclei.

Antigenemia in Leukocytes

Forty microliter aliquots (each equivalent to 2 × 105

total leukocytes) of the adjusted mixed PBLs, PMNLs,and PBMCs fractions were centrifuged (Cytospin 2;Shandon Southern Products Ltd., Runcorn, UK) ontoglass slides at 700 × g for 6 min. One slide of eachfraction was stained by Giemsa for morphologicalanalysis. The average purity of leukocyte fractions wasabout 95%. Two mixed PBLs, PMNLs, and PBMCsslides each were screened for CMV pp65 antigens (us-ing Clonab C10/C11; Biotest, Dreieich, Germany) andfor p72 antigens (using antibody 9221), respectively, byindirect immunofluorescence as described [Gerna etal., 1992a].

Detection of CEC

Five PBMCs slides of each patient sample (the totalamount derived from 106 mixed PBL) were screened forgiant (35–45 mm) cells. The endothelial origin of thesecells was confirmed by direct immunofluorescence withtetramethyl-rhodamine-conjugated Ulex europaeus ag-

glutinin I (UEA I) (Sigma, Munchen, Germany) as de-scribed [Grefte et al., 1993b]. Five additional slideswere screened for CMV p72 antigens present in giantcells by indirect immunofluorescence as describedabove. CEC counts were expressed as the total numberof CMV p72-positive giant cells on 5 slides.

PCR

DNA was extracted from 100-ml aliquots (equivalentto 5 × 105 total cells) of leukocyte subset preparationsby digestion with 100 mg/ml Proteinase K [Schafer etal., 1993] in a 50-ml reaction volume. After being boiledfor 10 min, a 10-ml supernatant sample was subjectedto PCR.

Plasma was obtained by centrifugation of 1 ml ofwhole blood at 700 × g for 10 min to remove cells. Su-pernatants (approximately 500 ml) were centrifuged fora second time to remove cell debris and then extractedonce with one volume of phenol-chloroform and subse-quently with one volume of chloroform-isoamyl alcohol(24:1). DNA from 100 ml of the aqueous phase was pre-cipitated with 0.1 volumes of sodium acetate, pH 5.2,and 2.5 volumes of absolute ethanol. After centrifuga-tion at 15,000 × g for 30 min pellets were washed with70% ethanol and resuspended in 100 ml of H2O. A 10-mlsample was subjected to PCR.

CMV and b-globin target sequences were quanti-tated by competitive PCR using cloned standard (ST)sequences as described previously [Schafer et al., 1993;Kuhn et al., 1994] with minor modifications.

Ten-microliter samples of DNA extracts from leuko-cyte or plasma preparations were added to PCR reac-tion mixtures [Schafer et al., 1993] to a total volume of100 ml containing 2 × 105 copies of the b-globin ST andeither 103 copies (high ST) or 50 copies (low ST) of theCMV ST [Kuhn et al., 1994]. CMV target sequenceswere amplified in 20 cycles with external CMV-specificprimers E1 (58-TCCAACACCCACTAGACCGGT-38)and E2 (58-CGGAAACGATGGTGTAGTTGG-38). Tenmicroliters each of the external reaction was reampli-fied in a second round of PCR for 30 cycles with theinternal CMV-specific primers TGGE1B (58-CCGGAT-CCCGCCGCCCGCCCCGCGCCCGCCGCGGCAGC-ACCTGGCT-38) and TGGE2E (58-GCGAATTCG-TAAACCACATCACCGTGGA-38), and for leukocyteextracts with the b-globin-specific primers 1aB (58-CCGGATCCCGCCGCCCGCCCCGCGCCCCTGCCG-TTACTGCCCTGT-38) and 1bE (58-GCGAATCCTAT-TGGTCTCCTTAAACCTG-38), respectively [Kuhn etal., 1994]. Standard and wild-type CMV and b-globinPCR amplimers were quantitated by hybridization to asingle strand-labeled standard sequence, separation bytemperature gradient gel electrophoresis, and densito-metric analysis of autoradiographs [Schafer et al.,1993]. For CMV wild-type copies $500 in 10 ml thefigures from the high-standard reaction were used, andfor CMV wild-type copies <500, those from the low-standard reaction were used.

In mixed leukocyte samples, CMV/cellular DNA ra-tios were expressed as CMV copies/106 copies b-globin

490 Schafer et al.

(the theoretical maximum DNA yield from 5 × 105

cells). To allow precise comparison of CMV DNA levelsin the leukocyte subsets, CMV DNA copies were re-lated to the expected b-globin DNA copy numbers pres-ent in the mixed PBL population. For example, if 3 ×105 PMNLs were counted in 5 × 105 mixed PBL, thetheoretical maximum cellular DNA yield from PMNLswas 6 × 105 copies b-globin. Thus, in this case the CMV/b-globin ratio in the PMNLs fraction was expressed asCMV copies/6 × 105 copies b-globin.

Plasma CMV copy numbers were expressed as theCMV genome equivalents present in 10 ml.

Statistical Analysis

Frequencies of positive results in the laboratory as-says between patients with symptomatic and withasymptomatic CMV infections were compared by the x2

test. Differences in the frequencies of positive resultsbetween PMNLs and PBMCs fractions were tested forsignificance by McNemar’s test. Differences in quanti-tative results of viremia, antigenemia, and DNAemiawere compared by the Kruskal–Wallis test combinedwith Dunn’s test. Correlation of endothelial cell countswith quantitative results of other laboratory assayswas determined by Spearman regression analysis. Inall test procedures P values <0.05 were considered to bestatistically significant.

RESULTSActive CMV Infections

Three hundred ten blood samples were drawn fromthe 36 patients with active CMV infection. Two hun-dred seventy specimens met the laboratory criteria foractive CMV infection. Of these, 131 belonged to 17 pa-tients with clinical features (Table I). In 4 patients thesymptoms were interpreted as CMV disease (2 indi-viduals presented with CMV pneumonitis, 2 with coli-tis). CMV syndrome was diagnosed in 13 patients.

In 19 patients CMV infection went asymptomatic.From these patients 139 samples indicating activeCMV infection were available.

All of the symptomatic and 5 of the asymptomatic

patients were treated with ganciclovir. In 2 symptom-atic patients steroid-resistant graft rejection was addi-tionally confirmed. These patients were also treatedwith anti-T-lymphocyte globulin and OKT3.

Among the 5 ganciclovir-treated patients withasymptomatic CMV infection, pp65 antigenemia andleukoDNAemia levels lay above the cutoff values gov-erning the initiation of preemptive therapy in 2 cases.Another patient was treated due to a febrile episodebut altogether did not meet the criteria for diagnosingsymptomatic CMV infection. In the remaining twocases ganciclovir therapy was administered because ofsteroid-resistant graft rejection (data not shown).

One hundred eighty-six specimens were investigatedfrom 35 patients that revealed no evidence of activeCMV infection.

Symptomatic Vs. Asymptomatic CMV Infection:Frequencies of Positive Laboratory Results

All results of pp65 and p72 antigenemia and of leu-koDNAemia were determined separately in the differ-ent leukocyte fractions to allow correct interpretationof the viremia data obtained from these subsets.

Three important observations were made on theoverall qualitative laboratory results (presented inTable I). First, although viremia was significantly morefrequent in both PMNLs and PBMCs fractions of symp-tomatic compared with asymptomatic patients, forpp65 and p72 antigenemia such differences were seenonly in the PBMCs fractions. Second, viremia and an-tigenemia (pp65 and p72) was significantly more fre-quent in PMNLs than in PBMCs of all actively CMV-infected patients. Third, endothelemia was found onlyin PBMCs fractions of symptomatic, not of asymptom-atic patients. No significant differences between symp-tomatic and asymptomatic patients were observed forthe rates of leukoDNAemia (100% in all fractions) orplasma DNAemia.

Specimens in which cytomegalic endothelial cellswere detected bore a second origin of infectious CMV inaddition to the leukocytes. To allow correct interpreta-tion of their influence, in the subsequent qualitative

TABLE I. Positivity Rates of Laboratory Assays: Differences Between Symptomatic and Asymptomatic Courses ofCMV Infection

Clinical course ofCMV infection

No. ofpatients

No. ofsamplesa

Leukocytesubset

No. of samples positive (%)

Culture CECAntigenemia DNAemia

pp65 p72 Leukocytes PlasmaSymptomatic 17 131 105 (80.1)

Mixed PBL 114 (87.0) 129 (98.5) 94 (71.8) 131 (100)PMNL 114 (87.0) 129 (98.5) 93 (71.0) 131 (100)PBMC 57 (43.5) 43 (32.8) 35 (26.7) 34 (26.0) 131 (100)

Asymptomatic 19 139 101 (72.7)Mixed PBL 55 (39.6) 135 (97.1) 97 (69.8) 139 (100)PMNL 54 (38.8) 135 (97.1) 97 (69.8) 139 (100)PBMC 10 (7.2) 0 16 (11.5) 12 (8.6) 139 (100)

aNumber of samples that fulfilled the laboratory criteria for active CMV infection. Symptomatic vs. asymptomatic CMV infections: viremia inmixed PBL, P < 0.001 (x2 test); in PMNL, P < 0.001; in PBMC, P < 0.001. CEC in PBMC, P < 0.001. pp65 antigenemia in PBMC, P 4 0.002;p72 antigenemia in PBMC, P < 0.001. PMNL vs. PBMC, symptomatic CMV infections: viremia, P < 0.001 (McNemar’s test); pp65 antigenemia,P < 0.001; p72 antigenemia, P < 0.001. PMNL vs. PBMC, asymptomatic CMV infections: viremia, P < 0.001; pp65 antigenemia, P < 0.001; p72antigenemia, P < 0.001.

Infectious CMV in Leukocyte Subsets 491

and quantitative data analyses comparisons weremade between i) endothelemia-positive samples (allfrom symptomatic patients), ii) endothelemia-negativespecimens from symptomatic patients, and iii) samplesfrom asymptomatic patients (all endothelemia nega-tive).

Regarding PBMCs, the fraction that harbored cyto-megalic endothelial cells, virus isolation was more fre-quent from endothelemia-positive samples than fromendothelemia-negative samples (Fig. 1). From endothe-lemia-negative samples of symptomatic patients, how-ever, culture rates were significantly higher as well,compared with asymptomatic individuals. Antigen-emia rates among the symptomatic patients did notdiffer between endothelemia-positive and endothe-lemia-negative specimens.

No significant differences between endothelemia-positive and endothelemia-negative samples in PMNLsfractions from symptomatic patients were observed forviremia or antigenemia frequencies. No influence of en-dothelial cells on the rates of DNAemia (leukocytes orplasma) was observed (data not shown).

No differences were evident in the frequency of anypositive laboratory measurement between patientswith CMV disease or CMV syndrome (data not shown).

Symptomatic Vs. Asymptomatic CMV Infection:Quantitative Laboratory Results

The levels of virus culture, pp65 and p72 antigen-emia, and DNAemia in the separate leukocyte fractionswere compared for endothelemia-positive samples, en-dothelemia-negative specimens from symptomatic in-dividuals, and samples from asymptomatic patients(Fig. 2).

Counts of infectious foci in virus culture determinedin total PBLs fractions were significantly higher in en-dothelemia-positive samples than in endothelemia-negative specimens (Fig. 2A). Furthermore, culturelevels from endothelemia-negative samples of symp-

tomatic patients were significantly higher than fromsamples of asymptomatic patients. These observationscould be attributed to the PBMCs. In contrast, withPMNLs fractions differences in infectious foci wereonly observed between symptomatic and asymptomaticpatients, with no evident influence of endothelial cells.

pp65 and p72 antigen-positive cell counts in mixedPBL from symptomatic patients compared with asymp-tomatic patients were mainly attributable to thePMNLs (Fig. 2B, 2C). In PBMCs fractions, only lowamounts of antigen-positive cells were found. Never-theless, they were significantly higher in symptomaticthan in asymptomatic individuals. No significant dif-ferences in the antigenemia levels were noted betweenendothelemia-positive and endothelemia-negativesamples of symptomatic patients.

CMV DNA levels were significantly higher in mixedleukocyte samples from patients experiencing symp-tomatic CMV infections compared with asymptomaticpatients (Fig. 2D). These observations were clearly at-tributable to the PMNLs. In contrast to viremia andantigenemia, no differences between symptomatic andasymptomatic patients were noted for DNAemia inPBMCs. Among the symptomatic patients, no signifi-cant differences of DNAemia levels were seen betweenendothelemia-positive and endothelemia-negativesamples regarding leukocytes of any subset.

Overall plasma DNAemia levels were significantlyhigher (P < 0.05) in symptomatic (range, 35–170 copies/10 ml [median, 95]) than in asymptomatic patients(range, 5–125 copies/10 ml [median, 45]), with no dif-ference between endothelemia-positive and endothe-lemia-negative samples. A reliable cutoff level as es-tablished for leukoDNAemia, however, that wouldenable discrimination of symptomatic from asymptom-atic courses of active CMV infection, could not be de-fined (data not shown).

The numbers of cytomegalic endothelial cells deter-mined in PBMCs samples ranged from 3–46 (median,

Fig. 1. Association of CEC with viremia andantigenemia in PBMC fractions from symptom-atic and asymptomatic patients, respectively.Horizontal bars indicate statistically significantdifferences between corresponding groups (x2

test). pp65 antigenemia: CEC-positive vs. CEC-negative/symptomatic, not significant. p72 anti-genemia: CEC-positive vs. CEC-negative/symptomatic, not significant.

492 Schafer et al.

14). No significant correlation of endothelial cell countswas observed with the levels of any other parameterdetermined in corresponding endothelemia-positivesamples (data not shown).

No obvious differences in the levels of any laboratoryparameter were observed between patients who suf-fered from CMV disease compared with those who ex-perienced CMV syndrome (data not shown).

Association of CMV Culture From LeukocyteSubsets With the Clinical Outcome

The clinical outcome of the 36 actively CMV-infectedpatients was correlated with the three possible originsof virus isolated from leukocyte preparations (endothe-lial cells, polymorphonuclear leukocytes, and mono-nuclear cells). Parameters were rated as positive in apatient if they were found in at least one blood sampleduring the course of the CMV infection. Criteria fortrue leukocytic viremia were i) that regarding PBMCs,virus had to be isolated from a sample in which noendothelial cells were detected, and ii) that p72 anti-gen-positive PMNLs or PBMCs were detected in thecorresponding fraction (indicating that these cells in-deed harbored CMV DNA).

Following these criteria, no significant differenceswere evident between the rates of PBMCs and PMNLsviremia among patients who experienced symptomaticCMV infections (Table II). This applied both to the 4patients who experienced CMV disease (2 had endothe-lial cells, 3 had PBMCs viremia, 4 had PMNLs viremia)and to the 13 patients with CMV syndrome (6 had en-dothelial cells, 9 had PBMCs viremia, 13 had PMNLsviremia), respectively (data not shown).

In asymptomatic patients, however, PMNLs viremiaoccurred with significantly higher frequency thanPBMCs viremia (P 4 0.003). This was the case both forthe 14 patients who received no ganciclovir (10 hadPMNLs viremia, only 2 had PBMCs viremia) and forthe 5 patients (4 with PMNLs viremia, only one withPBMCs viremia) who received preemptive therapy(data not shown).

Significance of Virus Culture FromLeukocyte Subsets

Sensitivity, specificity, positive predictive value, andnegative predictive value of symptomatic CMV infec-tion were determined separately for virus culture from

Fig. 2. Comparison of quantitative results of shell vial culture (A),pp65 antigenemia (B), p72 antigenemia (C), and leukoDNAemia (D)in samples from patients with symptomatic and asymptomatic CMVinfections, respectively. Vertical bars indicate median values. Errorbars span minimum and maximum values. Horizontal bars indicatestatistically significant differences between corresponding groups(Dunn’s test). (A) Mixed PBL: CEC-positive vs. asymptomatic, P <0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01; CEC-positive vs. CEC-negative/symptomatic, P < 0.05. PMNL: CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomatic vs.asymptomatic, P < 0.01. PBMC: CEC-positive vs. asymptomatic, P <0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01; CEC-positive vs. CEC-negative/symptomatic, P < 0.01. (B) Mixed PBL:

CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomaticvs. asymptomatic, P < 0.01. PMNL: CEC-positive vs. asymptomatic, P< 0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01.PBMC: CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01. (C) Mixed PBL: CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomatic vs.asymptomatic, P < 0.01. PMNL: CEC-positive vs. asymptomatic, P <0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01. PBMC:CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomaticvs. asymptomatic, P < 0.01. (D) Mixed PBL: CEC-positive vs. asymp-tomatic, P < 0.01; CEC-negative/symptomatic vs. asymptomatic, P <0.01. PMNL: CEC-positive vs. asymptomatic, P < 0.01; CEC-negative/symptomatic vs. asymptomatic, P < 0.01.

Infectious CMV in Leukocyte Subsets 493

the different leukocyte subsets, respectively (Table III).The diagnostic utility was compared with i) CEC de-tection, ii) the routine cutoff approach for pp65 anti-genemia, iii) the routine cutoff approach for leuko-DNAemia, and iv) plasma DNAemia.

Unlike mixed PBLs and PMNLs preparations, posi-tive results for viremia in PBMCs fractions were highlyindicative (in terms of specificity and positive predic-tive value) of a symptomatic course of CMV infection.PBMCs viremia was equivalent with pp65 antigenemiaand leukoDNAemia in that respect (Table III). The sen-sitivity of viremia in PBMCs fractions was maximal,however, in contrast to pp65 antigenemia, DNAemia(leukocytes and plasma), and detection of endothelialcells. All predictive parameters of PBMCs viremia werehigher than those of plasma DNAemia.

Patients who experienced clinical CMV manifesta-tions had viremia in mononuclear leukocytes before theonset of symptoms (data not shown).

DISCUSSION

The data indicate that culturable CMV present inmononuclear leukocytes is of greater significance com-pared with PMNLs regarding the clinical outcome ofCMV-seropositive immunocompromised individuals.This observation is relevant from both the diagnosticand the pathogenetic point of view and should drawfurther attention to factors involved in PBMCs tropismand their influence on CMV dissemination.

In peripheral blood, leukocytes (PMNLs and PBMCs)and cytomegalic endothelial cells can harbor infectiousCMV and thereby contribute to viral disseminationduring active infection. In the PMNLs fractions ana-lyzed in our setting, infectious virus was sure to origi-nate from leukocytes, because endothelial cells are notfound in these fractions. The potential of carrying in-fectious virus is reflected by p72 positivity, that re-quires that these leukocytes contain viral DNA andhave not merely acquired pp65 tegument protein. Al-though pp65 and p72 positivity were almost as fre-quent in asymptomatic as in symptomatic patients andCMV DNA was present in all samples from these pa-tients, virus was recovered significantly more often andwith higher levels from symptomatic than from asymp-tomatic individuals. This can be explained by the sig-nificantly higher amounts of antigen-positive cells andof CMV DNA copies in the samples from symptomatic

individuals. It has been reported that a distinct per-centage of pp65-positive PMNLs is also p72 positive[Gerna et al., 1990; Gerna et al., 1992b; Hackstein etal., 1996] and DNA positive [Hackstein et al., 1996].These observations probably reflect initial transcrip-tion events after uptake from surrounding sites of virusproduction. Most investigators have found no evidenceof full CMV replication in PMNLs [Grefte et al., 1994;Larsson et al., 1998]. It is unlikely that contaminatingPBMCs had relevant influence on PMNLs viremia, be-cause only about 5% PBMCs were present in PMNLsfractions, and the numbers of p72-positive PBMCswere also much lower compared with p72-positivePMNLs counts.

In PBMCs fractions, leukocytes and endothelial cellscontributed to viremia. To discriminate between endo-thelemia and leukocytic viremia, endothelemia-positive and endothelemia-negative samples fromsymptomatic patients and samples from asymptomaticpatients (that were all endothelemia negative) werecompared. Positivity for cytomegalic endothelial cellswas defined as positivity for CMV p72 antigen-positivegiant cells (35–45 mm), because this meant that theyharbored CMV DNA and that transcription events oc-curred. Although no double immunofluorescence wasperformed to confirm their endothelial origin, the onlycells between 35 and 45 mm in size were found to beUEA I positive on separate slides.

No correlation was observed between the numbers ofendothelial cells and of infectious foci from PBMCsfractions. The significantly higher frequency and levelsof viremia in endothelemia-positive compared with en-dothelemia-negative PBMCs samples, however, indi-cate that cytomegalic endothelial cells contributed toinfectious virus cultured from PBMCs fractions. On theother hand, the significantly higher frequency and lev-els of CMV culture and of p72 antigenemia in mono-nuclear leukocytes of symptomatic patients comparedwith asymptomatic individuals suggest that PBMCsalso play a substantial role. Contaminating PMNLshad minor influence on culture results from PBMCsfractions, because no significant differences of p72-positive PMNLs counts were observed between PBMCsfractions from symptomatic and asymptomatic pa-tients, respectively (data not shown).

TABLE II. Association of Clinical CMV ManifestationsWith Sources of Infectious Virus

Clinical courseof CMV infectiona

Total no. ofpatients

No. of patients positiveb

CECLeukocyte culturePBMCc PMNL

Symptomatic 17 8 12 17Asymptomatic 19 3 14aFor definitions see Materials and Methods.bIn at least one sample during the course of CMV infection.cCulture from samples negative for CEC (to assure the leukocyticorigin of infectious CMV).PBMC vs. PMNL: asymptomatic infection, P 4 0.003 (McNemar’stest).

TABLE III. Prediction of Symptomatic CMV Infection (n 417): Comparison of Laboratory Methods

ParameterSe(%)

Sp(%)

PPV(%)

NPV(%)

ViremiaMixed PBL 100 74 55 100PMNL 100 74 55 100PBMC 100 94 85 100

CEC 47 100 100 86pp65 ($50 cells)a 88 96 88 96DNAemia

Mixed PBL ($1,000 copies)b 88 96 88 96Plasma 82 69 45 93

Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV,negative predictive value.aPer 2 × 105 mixed PBL.bPer 106 copies b-globin.

494 Schafer et al.

Interestingly, in PBMCs fractions no differences ofCMV DNA copy numbers were noted between samplesfrom symptomatic (endothelemia-positive or endothe-lemia-negative) and from asymptomatic individuals,although pp65- and p72-positive leukocyte counts diddiffer significantly. This is a sharp contrast to the ob-servations made with PMNLs. Furthermore, althoughthe range of infectious foci from PBMCs was similar tothat observed for PMNLs preparations, CMV DNAcopy numbers in PMNLs reached almost 15,000 copies/106 copies b-globin, compared with 500 copies/106 cop-ies b-globin in PBMCs. The reasons for these discrep-ancies are unclear, but might be due partly to the lowabundance of endothelial cells in the background of to-tal DNA isolated from PBMCs fractions. Regarding theendothelemia-negative samples, a possible explanationis that productively infected PBMCs [Larsson et al.,1998] could yield infectious foci at amounts similarlyhigh to those detected in PMNLs (transmitting viruswithout de novo synthesis) despite much lower baselineDNA load.

PBMCs viremia was detected in much less asymp-tomatic patients compared with PMNLs. In contrast, inall patients who experienced CMV-specific clinical fea-tures PBMCs viremia or cytomegalic endothelial cellswere detectable. Shell vial culture from ficoll-purifiedPBMCs fractions detected infectious virus both fromPBMCs and from endothelial cells. As endothelemiawas detected only in 8 of the 17 symptomatic patients,CMV culture from mononuclear leukocytes can also beregarded as highly predictive of symptoms.

The underlying mechanisms of CMV leukocyte tro-pism are largely unclear. In this context, differencesbetween the glycoprotein B genotypes have been de-scribed [Meyer-Konig et al., 1998]. No clear correlationwith the clinical outcome, however, was observed inthis study. Moreover, cell tropism was only investi-gated to the DNA level, so no discrimination betweenproductively and abortively CMV-infected cells waspossible. Thus, it remains to be seen whether the gBgenotype indeed determines the clinical outcome of sys-temic CMV infections in immunosuppressed individu-als.

From the diagnostic view, reasonable utility of cutoffapproaches with antigenemia and leukoDNAemia forpatient monitoring has been shown [for reviews, seeSchafer and Laufs, 1996; Boeckh and Boivin, 1998].However, some patients with CMV disease are stillmissed using these methods, as also reported by othergroups [Gerna et al., 1994]. Thus, it remains necessaryto proceed in the search for markers of clinical CMVmanifestations. In this context, culture from PBMCsfractions was maximally sensitive regarding symptom-atic CMV infections, thus being superior to the cutoffapproaches of pp65 antigenemia and leukoDNAemia inthat respect. Importantly, the specificity of PBMCs cul-ture was not impaired. Although results were availableonly after 2 days, in all cases of clinical CMV manifes-tations PBMCs viremia was detected before the pa-tients first developed symptoms.

Plasma PCR was inferior to PBMCs viremia, both interms of sensitivity and specificity. Others have re-ported higher specificity of plasma PCR for diagnosis ofCMV disease [Boeckh and Boivin, 1998], mainly inAIDS patients [Shinkai et al., 1997; Boivin et al., 1998;Spector et al., 1998]. The reasons for these discrepan-cies are not clear, but might have to do with the factthat i) we did not investigate AIDS patients, and ii)that no subjects who experienced primary CMV infec-tion were enrolled. Furthermore, the studies are onlypartly comparable because pp65 antigenemia and shellvial culture were not performed with PBLs in all stud-ies.

Cytomegalic endothelial cells were detected in thisstudy with nearly the same frequency in patients withCMV syndrome as in those with CMV disease, whereasothers [Percivalle et al., 1993; Gerna et al., 1998;] de-scribed endothelemia to be mainly associated with late-stage organ disease. The overall detection rate in ourstudy on renal-allograft recipients, however, was lowcompared with reports on AIDS patients [Percivalle etal., 1993; Grefte et al., 1993b] or bone marrow trans-plant recipients [Salzberger et al., 1997]. The sensitiv-ity of endothelial cell detection might be improved by atechnically demanding cell sorting procedure describedrecently [Kas-Deelen et al., 1998]. Nevertheless, as thepatients who experienced symptomatic CMV infectionswere identified by virus culture from PBMCs fractionswith high specificity, we see no need to routinely screenrenal-allograft recipients for endothelemia.

It was not possible by any laboratory method to dis-criminate between those patients who developed CMVdisease and those who had CMV syndrome. As CMVsyndrome is a much less specific diagnosis, one mightask what the significance of such a monitoring systemis. In our opinion it is very important to assess anyform of CMV-related syndrome reliably so that diseaseprogression in patients who otherwise might not betreated in time can be prevented.

In this context some attention should be paid to the5 patients who did not meet the definition of symptom-atic CMV infection but were nevertheless treated withganciclovir. It is reasonable that in these cases CMV-related symptoms were prevented by the therapy. Re-garding the viremia data before administration oftherapy, however, 4 of the 5 preemptively treated pa-tients had PMNLs viremia, but only one of them hadPBMCs viremia. These proportions are highly compa-rable to those in the other group of asymptomatic pa-tients who never received therapy (10 of 14 had PMNLsviremia, only 2 of 14 had PBMCs viremia), but aredifferent compared with the patients who experiencedsymptomatic CMV infections (all of whom had PBMCsviremia). One might argue that the specificity ofPMNLs viremia may have increased if these 5 patientshad remained untreated until symptoms occurred. Onthe other hand, it is just as likely that PBMCs viremiaalso would have emerged under such conditions. Takentogether, our data still suggest that the absence ofPBMCs viremia is linked to clinical courses that do not

Infectious CMV in Leukocyte Subsets 495

meet the criteria for diagnosis of symptomatic CMVinfection.

In summary, CMV culture from mononuclear leuko-cytes seems to be strongly associated with clinicalmanifestations. In contrast, PMNLs culture is more of-ten positive in asymptomatic individuals. By virus cul-ture from ficoll-purified PBMCs fractions infectious vi-rus both in mononuclear leukocytes and in endothelialcells can be assessed. Thus, CMV culture from ficoll-purified PBMCs may help to define of immunocompro-mised patients for clinical CMV manifestations. More-over, closer investigation of factors involved in PBMCsviremia may be a key to better understanding of thepathogenetic mechanisms responsible for virus dis-semination in the CMV-seropositive immunocompro-mised subject.

REFERENCES

Boeckh M, Boivin G. 1998. Quantitation of cytomegalovirus: method-ological aspects and clinical applications. Clin Microbiol Rev 11:533–554.

Boivin G, Handfield J, Toma E, Murray G, Lalonde R, Bergeron MG.1998. Comparative evaluation of the cytomegalovirus load in poly-morphonuclear leukocytes and plasma of human immunodeficien-cy virus-infected subjects. J Infect Dis 177:355–360.

Erice A, Holm MA, Gill PC, Henry S, Dirksen CL, Dunn DL, HillamRP, Balfour HHJ. 1992. Cytomegalovirus (CMV) antigenemia as-say is more sensitive than shell vial cultures for rapid detection ofCMV in polymorphonuclear blood leukocytes. J Clin Microbiol 30:1763–1767.

Gerna G, Furione M, Baldanti F, Sarasini A. 1994. Comparativequantitation of human cytomegalovirus DNA in blood leukocytesand plasma of transplant and AIDS patients. J Clin Microbiol32:2709–2717.

Gerna G, Revello MG, Percivalle E, Morini F. 1992a. Comparison ofdifferent immunostaining techniques and monoclonal antibodiesto the lower matrix phosphoprotein (pp65) for optimal quantita-tion of human cytomegalovirus antigenemia. J Clin Microbiol 30:1232–1237.

Gerna G, Revello MG, Percivalle E, Zavattoni M, Parea M, BattagliaM. 1990. Quantification of human cytomegalovirus viremia by us-ing monoclonal antibodies to different viral proteins. J Clin Mi-crobiol 28:2681–2688.

Gerna G, Zavattoni M, Baldanti F, Furione M, Chezzi L, Revello MG,Percivalle E. 1998. Circulating cytomegalic endothelial cells areassociated with high human cytomegalovirus (HCMV) load inAIDS patients with late-stage disseminated HCMV disease. J MedVirol 55:64–74.

Gerna G, Zipeto D, Percivalle E, Parca M, Revello MG, Maccario R,Peri G, Milanesi G. 1992b. Human cytomegalovirus infection ofthe major leukocyte subpopulations and evidence for initial viralreplication in polymorphonuclear leukocytes from viremic pa-tients. J Infect Dis 166:1236–1244.

Grefte A, Blom N, van der Giessen M, van Son W, The TH. 1993a.Ultrastructural analysis of circulating cytomegalic cells in pa-tients with active cytomegalovirus infection: evidence for virusinfection and endothelial origin. J Infect Dis 168:1110–1118.

Grefte JMM, van der Giessen M, van Son WJ, The TH. 1993b. Circu-lating cytomegalovirus-infected endothelial cells in patients withan active cytomegalovirus infection. J Infect Dis 167:270–277.

Grefte JMM, Harmsen MC, van der Giessen M, Knollema S, van SonWJ, The TH. 1994. Presence of human cytomegalovirus (HCMV)immediate early mRNA but not ppUL83 (lower matrix proteinpp65) mRNA in polymorphonuclear and mononuclear leukocytesduring active HCMV infection. J Gen Virol 75:1989–1998.

Grefte JMM, van der Gun BTF, Schmolke S, van der Giessen M, vanSon WJ, Plachter B, Jahn G, The TH. 1992. The lower matrixprotein pp65 is the principal viral antigen present in peripheralblood leukocytes during an active HCMV infection. J Gen Virol73:2923–2932.

Hackstein H, Kirchner H, Jahn G, Bein G. 1996. The intracellularlocalization of human cytomegalovirus DNA in peripheral bloodleukocytes during active infections by high-resolution fluorescencein situ hybridization. Arch Virol 141:1293–1305.

Howell CL, Miller MJ, Martin WJ. 1979. Comparison of rates of iso-lation from leukocyte populations from blood by conventional andFicoll-Paque/Macrodex methods. J Clin Microbiol. 10:533–537.

Kas-Deelen AM, Harmsen MC, de Maar EF, van Son WJ, The TH.1998. A sensitive method for quantifying cytomegalic endothelialcells in peripheral blood from cytomegalovirus-infected patients.Clin Diagn Lab Immunol 5:622–626.

Kuhn JE, Wendland T, Schafer P, Mohring K, Wieland U, Elgas M,Eggers HJ. 1994. Monitoring of renal allograft recipients by quan-titation of human cytomegalovirus genomes in peripheral bloodleukocytes. J Med Virol 44:398–405.

Larsson S, Soderberg-Naucler C, Moller E. 1998. Productive cytomeg-alovirus (CMV) infection exclusively in CD13-positive peripheralblood mononuclear cells from CMV-infected individuals. Trans-plantation 65:411–415.

Ljungman P, Plotkin SA. 1995. Workshop on CMV disease; defini-tions, clinical severity scores, and new syndromes. Scand J InfectDis Suppl 99:87–89.

Meyer-Konig U, Vogelberg C, Bongarts A, Kampa D, Delbruck R,Wolff-Vorbeck G, Kirste G, Haberland M, Hufert FT, von Laer D.1998. Glycoprotein B genotype correlates with cell tropism in vivoof human cytomegalovirus infection. J Med Virol 55:75–81.

Percivalle E, Revello MG, Vago L, Morini F, Gerna G. 1993. Circulat-ing endothelial giant cells permissive for human cytomegalovirus(HCMV) are detected in disseminated HCMV infections with or-gan involvement. J Clin Invest 92:663–670.

Rinaldo CR, Black PH, Hirsch MS. 1977. Interaction of cytomegalo-virus with leukocytes from patients with mononucleosis due tocytomegalovirus. J Infect Dis 136:667–676.

Saltzman RL, Quirk MR, Jordan MC. 1988. Disseminated cytomega-lovirus infection. Molecular analysis of virus and leukocyte inter-actions in viremia. J Clin Invest 81:75–81.

Salzberger B, Myerson D, Boeckh M. 1997. Circulating cytomegalovi-rus (CMV)-infected endothelial cells in marrow transplant pa-tients with CMV disease and CMV infection. J Infect Dis 176:778–781.

Schafer P, Braun RW, Mohring K, Henco K, Kang J, Wendland T,Kuhn JE. 1993. Quantitative determination of human cytomega-lovirus target sequences in peripheral blood leukocytes by nestedpolymerase chain reaction and temperature gradient gel electro-phoresis. J Gen Virol 74:2699–2707.

Schafer P, Kuhn JE, Tenschert W, Eing B, Schroter M, Laufs R.1998a. Polymerase chain reaction (PCR) from ficoll-purified poly-morphonuclear leukocytes for monitoring cytomegalovirus infec-tions in renal allograft recipients: superior sensitivity and similarspecificity compared with plasma PCR. J Infect Dis 178:1544–1545.

Schafer P, Laufs R. 1996. Experience with quantitative PCR for themanagement of HCMV disease. Intervirology 39:204–212.

Schafer P, Tenschert W, Cremaschi L, Gutensohn K, Laufs R. 1998b.Utility of major leukocyte subpopulations for monitoring second-ary cytomegalovirus infections in renal-allograft recipients byPCR. J Clin Microbiol 36:1008–1014.

Shinkai M, Bozzette SA, Powderly W, Frame P, Spector SA. 1997.Utility of urine and leukocyte cultures and plasma DNA polymer-ase chain reaction for identification of AIDS patients at risk fordeveloping human cytomegalovirus disease. J Infect Dis 175:302–308.

Spector, SA, Wong R, Hsia K, Pilcher M, Stempien MJ. 1998. Plasmacytomegalovirus (CMV) DNA load predicts CMV disease and sur-vival in AIDS patients. J Clin Invest 101:497–502.

Storch GA, Buller RS, Bailey TC, Ettinger NA, Langlois T, Gaud-reault-Keener M, Welby PL. 1994. Comparison of PCR and pp65antigenemia assay with quantitative shell vial culture for detec-tion of cytomegalovirus in blood leukocytes from solid-organ trans-plant recipients. J Clin Microbiol 32:997–1003.

Zipeto D, Morris S, Hong C, Dowling A, Wolitz R, Merigan TC, Ras-mussen L. 1995. Human cytomegalovirus (CMV) DNA in plasmareflects quantity of CMV DNA present in leukocytes. J Clin Mi-crobiol 33:2607–2611.

496 Schafer et al.