Of the 48 NPAs studied, we observed no cytopathic effect on HEp-2, MDCK, human embryonic lung fibroblast, and BGM monolayers. Eleven (22.9%) specimens showed cytopathic effect of diffuse refractile rounding of cells on Vero cell monolayers 2–4 days after incubation, progressed rapidly, and involved the whole monolayer within 12–24 hours. The same cytopathic effect was reproducible on passage to Vero cells, and appeared 1 to 2 days after incubation. These Vero cell cultures were all positive by the coronavirus RT-PCR. The Vero cell culture supernatants showing cytopathic effect were randomly selected for electron microscopy examination, and coronavirus particles were seen.

Five specimens showed cytopathic effect of focal refractile rounding of cells in LLC-MK2 monolayers. All these LLC-MK2 cell cultures had been incubated at 37°C. Unlike the cytopathic effect observed in Vero cells attributable to coronavirus, this cytopathic effect developed after 10 to 12 days of incubation and progressed slowly to detachment from the cell monolayer.

The HMPV RT-PCR examination of cell cultures was negative for human embryonic lung fibroblast, BGM cells, and Vero cells (including those positive for coronavirus). In contrast, HMPV RT-PCR showed a PCR product of the expected size (89 bp) from 25 (52.1%) isolation materials injected with specimens. The nucleotide sequences of the PCR products were identical to the F-gene fragment of HMPV (GenBank accession no. NC 004148) (1). We retrospectively examined the first round of PCR products of all positive samples. Those positive samples derived from direct NPAs did not show positive band, indicating a nested RT-PCR was necessary. However, most (27 [90%] of 30) of those derived from cell cultures showed a positive band of the expected size from the first round of PCR. The distribution of HMPV RT-PCR results on direct detection of NPAs and from different cell culture types is shown in the Table. Overall, the sensitivity of direct detection of NPAs using HMPV RT-PCR was 2 (8.0%) of 25 samples. In one of these two samples, we isolated the virus from three cell lines. In the other sample, we isolated virus from HEp-2 and LLC-MK2. Overall, HEp-2 was the most sensitive cell lines (22 [88.0%] of 25 HMPV positive samples); LLC-MK2 cells detected 6 (24.0%) of 25 samples, and MDCK cells detected 2 (8.0%) of 25 samples. Most (with the exception of three LLC-MK2–positive samples) showed positive results in HEp-2 cells. All six LLC-MK2 cell cultures positive for HMPV were incubated at 37°C; three of these positive cultures that had had the corresponding LLC-MK2 cell cultures incubated at 33°C showed positive results.

To ascertain that cell cultures with HMPV RT-PCR–positive results represented the isolation of HMPV, all LLC-MK2 (incubated at 37°C), HEp-2, and MDCK cell cultures, regardless of the HMPV RT-PCR findings, were passaged to LLC-MK2 cells for a prolonged incubation of 28 days. HEp-2 cells were not used for this purpose because HEp-2 cell monolayers are often difficult to maintain for >2 weeks. The results showed that all passages from HMPV RT-PCR–positive cell cultures showed cytopathic effect of focal refractile rounding of cells that occurred after 10 to 22 days of incubation (Figure 1). The cytopathic effect progressed slowly to detachment from the cell monolayer (Figure 2). The supernatants of these passages were also positive by the HMPV RT-PCR and had visible HMPV viral particles on electron microscopy examination (Figure 3). The passages from HMPV RT-PCR–negative supernatants did not show positive results by the above tests. We also passaged five Vero cell cultures that were positive for coronavirus to LLC-MK2 cells in a similar way. All of these passages did not show cytopathic effect and were negative by the HMPV RT-PCR.

To reconfirm the fact that HMPV infections detected by this combination approach represented genuine infections, we coated HMPV-infected LLC-MK2 cells onto slides for antibody detection using the immunofluorescence technique. All HMPV culture–positive patients who had serologic evidence of infection had a more than fourfold rise in antibody titers, and 15 patients seroconverted.

Overall, our results indicated that the combination approach of using conventional virus isolation and molecular detection could be successfully applied to the isolation of HMPV (Figure 4). With this approach, we found that among the 48 study participants, 6 (12.5%) had both HMPV and coronavirus isolated from NPAs, 19 (39.6%) had HMPV, and 5 (10.4%) had coronavirus. Eighteen (37.5%) had no virus isolated from the cell lines that we used.

On the basis of a combination of conventional virus isolation system and molecular techniques, we found that 52.1% (25/48) of patients with SARS admitted to our hospital had HMPV infection. Isolation of HMPV is known to be difficult, which is why the virus could not be detected until recently. The first report on HMPV by van den Hoogen et al. (1) showed that the virus produced syncytia formation in tertiary monkey kidney cells, followed by rapid internal disruption of the cells and subsequent detachment from cell monolayer. The virus replicated poorly in Vero cells and human lung adenocarcinoma (A-549) cells and could not be propagated in MDCK cells or chicken embryo fibroblasts (1). In the study from Boivin et al. (7), multiple cell lines including LLC-MK2, HEp-2, MDCK, human foreskin fibroblast, Vero, Mink lung, A-549, human rhabdomyosarcoma (RD), transformed human kidney (293), and human colon adenocarcinoma (HT-29), were used for isolation of HMPV. The results showed that HMPV only grew on LLC-MK2 cells with cytopathic effect of round and refringent cells but without syncytia formation in most cases, an observation in agreement with our results. In that study, HEp-2 cell monolayers did not show cytopathic effect. Since the HEp-2 cells were not tested for HMPV RNA, we do not know whether our findings on HEp-2 cells were also observed by Boivin et al. In another study reported by Peret et al. (6), LLC-MK2, MDCK, and NCI-H292 cells were used; those researchers found that only LLC-MK2 cells produced cytopathic effect of focal rounding and without syncytia formation, which is also similar to our observation. The major difference in our approach for HMPV isolation compared to previous studies is the use of RT-PCR to enhance the detection of HMPV isolated from cell cultures. With this approach, we found that HEp-2 cells, a widely available and commonly used cell line, support the growth of HMPV. When RT-PCR was used to follow up all cell cultures, the sensitivity of HEp-2 cells was higher than LLC-MK2 cells, the cell line most commonly used in previous studies for HMPV. However, even using our approach, LLC-MK2 cells cannot be discarded, as in 12% of cases HMPV was only isolated from LLC-MK2 cells. In contrast, in the presence of HEp-2 cells, MDCK cells gave little additional value, as both specimens positive by MDCK cells had the viruses isolated from HEp-2 cells. In addition, our initial incubation of 14 days might not be optimal for isolating HMPV because Boivin et al. reported that the cytopathic effect took a mean incubation time of 17.3 days to develop (7). By prolonging the initial incubation of LLC-MK2 cells to 21 or 28 days, more HMPV infections might have been detected from our “negative” group.

Because all our study samples were collected from patients related to the outbreak of SARS that occurred in our hospital, one cannot simply infer that this in vitro growth property can be applied to all HMPV strains in general. Nevertheless, our approach of including HEp-2 cells, a widely available cell line, to search for HMPV, in particular for those cases related to SARS, needs to be considered. In our study, six patients were co-infected with HMPV and coronavirus. Although the number was limited, our findings suggest that HMPV and coronavirus have different in vitro tropisms, and the isolation of one virus does not affect the recovery of the other from different cell lines.

Overall, we confirmed that 25 (52.1%) of 48 patients admitted to our hospital with SARS had HMPV infections, with 6 also co-infected with coronavirus. However, the data on such high prevalence of HMPV should be interpreted cautiously. Our study population was based on persons and their family members who had been exposed in the index ward in our hospital. Thus, a co-circulation of two pathogens within our study group was possible. While the clinical presentations of all our study participants fulfilled the World Health Organization definition for a probable case of SARS (13), one should not infer, at this stage, that the prevalence of HMPV is similarly high in SARS outbreaks occurring elsewhere. On the other hand, the possibility of an important role of HMPV in the current worldwide outbreak of SARS should not be neglected. HMPV has also been detected in five of six SARS patients living in Canada (15). In that study series, coronavirus was also detected in five of six patients and four patients were co-infected with HMPV and coronavirus. A few recent studies implicate a strong association of a novel coronavirus with the worldwide outbreak of SARS (16–18). While both HMPV and coronavirus infections may result in severe respiratory tract diseases, their transmission efficiency may not be the same. This urgent question must be answered because the answer affects the priority for immediate development of control strategies.

During this outbreak of SARS, we have applied this combination approach of conventional virus isolation and molecular detection to establish the viral infection status of other patients hospitalized for SARS. We are in the process of analyzing a larger cohort to elucidate their clinical conditions, treatment responses, and epidemiologic links with respect to the infection status for both HMPV and coronavirus. Similar work in other parts of the world is needed.