Infected and uninfected Vero E6 cells were harvested 3–5 days after inoculation, inactivated by fixation and gamma irradiation (2 × 10⁶ rad), and processed for standard, immunolabeling electron microscopy (IEM) or ISH EM as previously described (6,14). For standard EM, glutaraldehyde- and osmium tetroxide–fixed specimens were embedded in Epon-substitute and Araldite (Ted Pella, Inc., Redding, CA) and sections were stained with uranyl acetate and lead citrate. Some infected and uninfected cultures were treated with 5% tannic acid solution before being embedded for standard EM (15). Specimens prepared for IEM and ISH assays were fixed in paraformaldehyde and glutaraldehyde and embedded in LR White resin (Ted Pella, Inc.), and sections were collected on nickel mesh grids.

A BAL specimen was obtained from a 47-year-old man within the first week of the onset of symptoms. A portion of the specimen was centrifuged at 2,000 rpm for 10 min, and the pellet was processed for standard EM.

IEM and ISH assays were performed essentially as described for Nipah virus (14). In brief, for IEM assays, sections were reacted with hyperimmune mouse ascitic fluid raised against SARS-CoV and then with a goat anti-mouse antibody conjugated to 12-nm colloidal gold particles (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Negative-sense riboprobes for the ultrastructural ISH assays were prepared as previously described (16,17). Riboprobes were directed against the nucleocapsid or polymerase protein portions of the SARS-CoV genome (Table) and incorporated digoxigenin-11-dUTP. Because of the nested set structure of the coronavirus genomic RNA (genRNA) and messenger RNAs (mRNAs), the nucleocapsid riboprobe would detect all viral RNAs (18). Sections were reacted with a pool of nucleocapsid and polymerase probes and then with a sheep anti-digoxigenin antibody conjugated to 6-nm colloidal gold particles (Electron Microscopy Sciences, Hatfield, PA). To obtain negative controls, we performed both assays with uninfected Vero E6 cells, and infected cells were reacted with an unrelated antibody and probe for IEM and ISH procedures, respectively.

During the global SARS outbreak of 2002 to 2003, a virus was isolated from human patients and identified by EM as belonging to the family Coronaviridae (6,7). Detailed studies described here on the morphogenesis of the SARS-CoV by thin-section EM found many characteristics previously described for coronaviruses (19,21,22). Virus particles formed upon membranes of the “budding compartment,” a term used to describe the continuous membrane system from the rough endoplasmic reticulum to the Golgi complex (23,24). Virions accumulated in dilated vesicles that appeared to migrate to the cell surface where the virus particles were released or remained adherent to the plasma membrane. Additional cytoplasmic structures associated with coronavirus infections included nucleocapsid inclusions and double-membrane vesicles, which have been proposed as the replication complex for coronaviruses (20), and arteriviruses (25), a closely related virus family that, in addition to coronaviruses, is a member of the order Nidovirales. IEM and ultrastructural ISH assays detected viral proteins and mRNA or genRNA associated with virions, double-membrane vesicles, and nucleocapsid inclusions. Coronaviruses are known to synthesize a nested set of subgenomic mRNAs (26), such that the nucleocapsid riboprobe used here allowed detection of all viral mRNAs in addition to genRNA. Indeed, considerable amounts of RNAs were detected in the ultrastructural ISH assays performed on SARS-CoV–infected cells.

As has been reported for other coronaviruses, additional cytoplasmic features were associated with SARS-CoV–infected cells. Tubular structures were occasionally seen within virus-containing vesicles (27,28); and cytoplasmic tubuloreticular structures, known to occur with numerous other infections (see 29), were also found. Large granular areas of cytoplasm, relatively devoid of organelles and containing viral proteins and RNA, were noted in SARS-CoV–infected cells; such features have not been described previously for coronaviruses. While the role of these cytoplasmic areas is unclear, the close proximity of cellular ribosomes with viral proteins and RNA suggests that they may be viral translation centers. Future ultrastructural ISH and IEM studies to characterize these areas, using riboprobes and monoclonal antibodies to specific SARS-CoV genes and gene products, should help clarify this issue.

Many of these ultrastructural findings were also observed in a BAL specimen from a SARS patient (Figure 5B) (6). Characteristic virions in vesicles and lining the cell surface and the presence of double-membrane vesicles provided clear evidence of a coronavirus infection and suggested that viral replication was occurring in the lower airways early in the course of infection. EM examination of BAL specimens may prove to be a useful tool in the diagnosis of SARS-CoV, analogous to the use of BAL specimens to diagnose influenza infections. Recent studies have reported finding coronavirus particles in lung and gastroenteric tissues of SARS patients and experimentally infected macaques (7,30–33), although the viral nature of these structures has not been confirmed by IEM or ultrastructural ISH assays. Coronavirus particles may be confused morphologically with other nonviral structures routinely found in cells, including coated vesicles, multivesicular bodies, perichromatin granules, glycocalyceal bodies, and cellular projections (29). Therefore, a cautious approach is advisable when examining clinical specimens.

The SARS outbreak is a prime example of an emerging infectious disease that can rapidly and easily spread, reaching global proportions. With SARS, as with previous investigations of outbreaks involving such viruses as Ebola (34–36), Hendra (37), Nipah (38), and more recently, monkeypox (39), EM played an essential role in determining the specific virus family of the pathogen involved. In all of these cases, tissue culture amplification of a virus isolate facilitated the ultrastructural examination. Thus, traditional microbiologic and EM approaches proved pivotal in determining the etiologic agents, thereby guiding subsequent laboratory and epidemiologic investigations.