Serial Section Array Scanning Electron Microscopy Analysis of Cells from Lung Autopsy Specimens following Fatal A/H1N1 2009 Pandemic Influenza Virus Infection

Generally, it is difficult to observe IAV particles in postmortem samples from patients with seasonal influenza. In fact, only a few viral antigens are detected in bronchial epithelial cells from autopsied lung sections. Previously, we detected many viral antigens in AEC-IIs from the lung. This was because the majority of A/H1N1/pdm09 in the lung tissue harbored an aspartic acid-to-glycine substitution at position 222 (D222G) of the hemagglutinin protein. A/H1N1/pdm09 harboring the D222G substitution has a receptor-binding preference for α-2,3-linked sialic acids expressed on human AECs and infects them in the same way as H5N1 and H7N9 avian IAVs. Here, we report the first successful observation of virus particles, not only in AEC-IIs, but also in Ms/Mϕs and Neus, using electron microscopy. The finding of a M/Mϕ harboring numerous virus particles within vesicles and at the cell surface suggests that Ms/Mϕs are involved in the pathogenesis of IAV primary pneumonia.

to ␣-2,6-linked sialic acid, which is expressed abundantly on epithelial cells lining the human respiratory tract (except the alveoli) (3). Therefore, seasonal IAVs rarely infect alveolar epithelial cells (AECs). A/H1N1 2009 pandemic influenza virus (A/H1N1/pdm09) was the causative IAV of a novel influenza pandemic in 2009 (4). Since then, A/H1N1/ pdm09 has continued to circulate as a seasonal virus, along with A/H3N2 influenza virus. IAVs accumulate point mutations during replication, thereby giving rise to quasispecies. One of these mutations results in changes to the HA amino acid sequence. Since the first appearance of A/H1N1/pdm09, an amino acid substitution from aspartic acid to glycine at position 222 (D222G) in the HA protein has been identified sporadically (5)(6)(7). The D222G substitution alters the receptor-binding preference of HA so that it binds to ␣-2,3-linked sialic acids, which are expressed abundantly on human AECs; therefore, this virus has the potential to cause viral pneumonia (5)(6)(7).
Previously, we reported a fatal case of A/H1N1/pdm09 infection (8). The results of next-generation sequencing revealed that a majority of the viruses detected in the lung carried the D222G substitution in HA (9). The patient presented with primary viral pneumonia, which developed into severe acute respiratory distress syndrome (ARDS), resulting in death from respiratory failure within 7 days of disease onset (8). Histopathological examination of the lung revealed diffuse alveolar damage, along with the presence of many viral antigens, genomic RNA, and mRNA in AECs (8,10,11). In this case, a large number of virus particles remained in the tissue at the time of death; thus, virus particles were observed in lung specimens by transmission electron microscopy (TEM) (8).
Here, we analyzed the same lung specimens using a novel scanning electron microscopy (SEM) method in addition to TEM (12). Whereas conventional SEM does not require embedding and sectioning steps before imaging and observes only the surfaces of specimens, the novel SEM technique used in this study involves mounting single or serial ultrathin sections on a glass slide and staining them with heavy metals, thereby enabling imaging at a resolution comparable with that of TEM by recording backscattered electrons. By digitally "stitching" together contiguous SEM images, a large-scale two-dimensional (2D) image can be obtained at high resolution (13). Serial section array (SSA)-SEM analyzes each cross-sectional image throughout a single whole cell to reveal the distribution of specific structures (12,14,15). After manual tracing of specific structures, such as virus particles, cytoplasm, and nucleus, in each serial section, the structures are visualized in a three-dimensional (3D) model (12,14,15). We focused on type II alveolar epithelial cells (AEC-IIs) as virus antigen-positive cells and on monocytes/macrophages (Ms/Ms) and neutrophils (Neus) as immune cells that respond to influenza virus infection. We used TEM and SEM to examine the distribution of influenza virus particles and matrix 1 (M1) protein-associated intranuclear dense tubules (16)(17)(18)(19)(20) within these cell types in specimens of autopsied lung in which A/H1N1/pdm09 harboring the D222G substitution (A/H1N1/pdm09-D222G) had proliferated. Here, even though we were able to examine only a single human case, we present the first successful observation of influenza virus particles and M1-associated intranuclear dense tubules in large-scale 2D images and 3D images.

RESULTS
TEM imaging of type II alveolar epithelial cells. Most of the AEC-IIs observed in the lung tissue specimen prepared for electron microscopy (EM) observation were desquamated from the basement membrane; 50 TEM images of AEC-IIs were analyzed. Virus particles were found in the alveolar space and around AEC-IIs ( Fig. 1a and c). The virus particles were about 95 nm in diameter and surrounded by glycoprotein spikes and envelopes; in addition, they contained virus ribonucleoprotein (RNP) complexes (Fig. 1b). Although no budding virus particles were identified, 6/50 TEM images of AEC-IIs revealed virus particles around the cell surface (Table 1). In these images, virus particles could be differentiated from microvilli (Fig. 1d). The virus particles were characterized by high-electron-density RNPs surrounded by glycoprotein spikes (Fig.  1d, inset). Notably, dense helical tubules were observed in the nucleus (Fig. 1d and e); these intranuclear tubules have been reported in IAV-infected cells and identified as an M1 protein-associated structure (16)(17)(18)(19)(20). We also observed these structures in the degenerated nuclei of AEC-II cells ( Fig. 1f and g). Intranuclear dense tubules were identified in 23/50 TEM images ( Table 1). The tubules were 108 to 698 nm in length, with a diameter of 22 to 39 nm.
TEM imaging of monocytes/macrophages. Ms/Ms were less common than AEC-IIs; therefore, 30 TEM images of Ms/Ms were analyzed. In four of them, several filamentous and spherical virus particles were observed within cytoplasmic vesicles  Table 1). The mean diameters of the filamentous and spherical virus particles were 75 nm and 95 nm, respectively. Also, the longest filamentous virus was 676 nm. The observation that the diameter of a filamentous virus was less than that of a spherical virus is consistent with the results of a previous report (21). No budding virus particles or intranuclear dense tubules were observed in any of the 30 images.
TEM imaging of neutrophils. Neus were present in numbers similar to those of Ms/Ms; therefore, 30 TEM images of Neus were analyzed. In three of them, Neus were  closely associated with desquamated and degenerated AEC-IIs, suggesting the occurrence of phagocytosis (Fig. 2c). Six images showed virus particles around the cell surface ( Table 1). Some of these virus particles contained nucleocapsids (mean diameter, 94 nm), whereas others were hollow (mean diameter, 68 nm) (Fig. 2d). No budding virus particles, virus particles in vesicles, or intranuclear dense tubules were observed in any of the 30 TEM images. Stitching together SEM images of an alveolus. An entire alveolus is too large (in this example, 209 by 206 m) to be presented in cross-section in a single TEM or SEM image. Therefore, we built a set of contiguous SEM images and digitally stitched them together to generate a cross-section of an entire alveolus at high resolution. Briefly, a single 200-nm-thick section was mounted on a glass slide and stained with toluidine blue to facilitate selection of an appropriate region by light microscopy. The section was then stained with heavy metals, thereby enabling a 10-by-14 array of contiguous SEM images to be taken and digitally stitched together (Fig. 3). In the alveolus shown in Fig. 4, type I and type II AECs and Ms/Ms are observed in the alveolar lumen. Neus and Ms/Ms are also present in a capillary blood vessel. The patient suffered chronic heart failure, and hemosiderin-laden macrophages were present in the alveolar lumen. Desquamated AEC-IIs had little visible chromatin in the expanded nucleus. At low magnification, the stitched SEM image was similar to the optical microscopic image stained with hematoxylin-eosin ( Fig. 4a and e); however, the stitched SEM image could be highly magnified using virtual-slide software (NDP.view2) so that virus particles (Fig.  4b) and intranuclear dense tubules (Fig. 4c) could be visualized. Virus particles were present at three extracellular sites, and intranuclear tubular structures were found in FIG 3 Comparative illustration of TEM and SEM workflows. Small pieces of fixed lung tissue were embedded in epoxy resin. Ultrathin sections 70 nm thick were mounted on a grid and stained with heavy metals for TEM. A novel SEM method, stitching together of contiguous SEM and SSA-SEM images, is shown. Ultrathin single or serial sections (each 200 nm thick) were mounted on a conductive glass slide and stained with toluidine blue for light microscopy. After staining with heavy metal, a 10-by-14 array of contiguous SEM images was digitally stitched together to generate an entire alveolus at high resolution, sufficient to confirm the presence of virus particles. For SSA-SEM images, sections that included the objective cells were colored yellow and stained with heavy metal prior to SEM. The SSA-SEM images had to be aligned for 3D reconstruction. Subsequently, the cytoplasm, nuclei, and viruses were segmented by tracing their boundaries. The images were then imported into the 3D software and visualized in a 3D model. eight AEC-IIs in the alveolus (Fig. 4a). In another alveolus with a similar level of inflammation, immunohistochemistry revealed that the distribution of AEC-IIs positive for the nucleoprotein of A/H1N1/pdm09 (Fig. 4f) was similar to that of AEC-IIs harboring intranuclear dense tubules (Fig. 4a).
SSA-SEM imaging of type II alveolar epithelial cells. SSA-SEM analysis revealed the distribution of specific structure throughout a single cell. Briefly, 100 serial sections (each 200 nm thick) were mounted on a platinum-palladium-coated glass slide and stained with toluidine blue. Then, sections containing the object cells were marked in yellow. Next, each marked serial section was captured by SEM after heavy metal staining (Fig. 3). Six AEC-IIs, six Ms/Ms, and three Neus were analyzed. The numbers of SSA-SEM images containing specific features are shown in Table 2, and the numbers of virus particles and intranuclear dense tubules in each SSA-SEM image are shown in Table 3.   Six randomly selected AEC-IIs were analyzed by SSA-SEM. All were desquamated from the basement membrane. Virus particles were observed around the surfaces of three of six AEC-IIs (Table 2). Intravesicular virus particles were observed only in AEC-II-6, and all were spherical/oval (Tables 2 and 3 and Fig. 5). Intranuclear dense tubules were observed in four of six AEC-IIs. Although it is possible that virus particles around 100 nm in diameter might be missed by analyzing serial sections that were 200 nm thick, neither virus particles nor dense tubules were observed in AEC-II-3 and AEC-II-4 ( Table 2).  Table 2). (c) Representative sections (S11, S13 to S15, and S20) from 37 consecutive sections (each 200 nm thick) of M/M-2. The boxed areas (left) are shown at higher magnification on the right. S13 to S15 show a vesicle that contains several virus particles. Spherical and filamentous virus particles are seen around the cell surface and in the vesicle (S11, S13 to S15, and S20). Multiple virus particles are observed in the extracellular region in which the cell membrane enters the cytoplasm (S20). (Left) Scale bars ϭ 1 m. (Right) Scale bars ϭ 100 nm.

SSA-SEM imaging of neutrophils.
Among the 100 serial sections used for SSA-SEM imaging, Neus were less common than Ms/Ms. Thus, three randomly selected Neus were analyzed by SSA-SEM (Tables 2 and 3). Virus particles were observed around the surfaces of Neu-1 and Neu-2, and intravesicular virus particles were found in Neu-1 (Tables 2 and 3). Neu-1 harbored several spherical virus particles within large vesicles (Fig. 7). No intranuclear dense tubules were seen in any of the 100 SSA-SEM images from the three whole cells examined (Table 2).
3D Visualization of SSA-SEM images. Specific structures, including cytoplasm (light blue), nucleus (purple), virus particles within the cell (green), and virus particles outside the cell (red), in each SSA-SEM image from serial sections were traced manually (Fig. 3). All SSA-SEM images were imported into 3D Slicer software for 3D visualization. The distribution of virus particles within desquamated AEC-II-6 ( Fig. 5b and 8a; see Movie S1 in the supplemental material), M/M-2 (Fig. 6b, and 8b; see Movie S2 in the supplemental material), and Neu-1 ( Fig. 7b and 8c; see Movie S3 in the supplemental material) was observed in the 3D images.

DISCUSSION
EM enables visualization and identification of the morphology of virus particles, although it has a limited observable range and can provide only partial observation of large-scale structures. Most EM images of influenza virus particles have been captured using infected cultured cells (22)(23)(24)(25). Several reports show TEM images of virus particles in autopsied human lung specimens, but the resolution of the images is limited and dependent on the condition of the specimens (18,(26)(27)(28)(29). Here, we used TEM and SEM to observe virus particles and virus-related structures in autopsied lung specimens from a patient who had been infected with A/H1N1/pdm09 and subsequently died from viral pneumonia complicated by ARDS (8). A/H1N1/pdm09-D222G proliferated in the lung, and many viral antigens and much viral RNA were detected in AEC-IIs (Fig. 4f) (8,10,11). It should be noted that this study is based on limited analyses of a single autopsied human lung sample. Because sufficient virus particles remained in the sample, the novel SEM method was able to show the detailed distribution of virus particles and intranuclear dense tubules in AEC-IIs, Ms/Ms, and Neus. Figure 3 shows how the same specimens were analyzed using TEM and the novel SEM method. The resolution of TEM images is higher than that of SEM images ( Fig. 1  and 2), although the observation range is obstructed by the framework of the grid (Fig.  3). In contrast, specimens used for SEM are mounted on a glass slide, which enables stitching together of contiguous SEM images (Fig. 4a). By digitally stitching together SEM images, we observed an entire cross-section of a single alveolus and were able to identify virus particles and intranuclear dense tubules at high resolution. The stitched SEM image enables visualization from nanoscale to microscale within a single data set, making it a very useful tool for analyzing the distribution of virus particles and specific structures associated with viral infection.
In this study, we focused on AEC-IIs, Ms/Ms, and Neus and identified influenza virus particles and intranuclear dense tubules (M1-associated structures within the nucleus, which are a morphological characteristic of influenza virus-infected cells) in each cell type (16)(17)(18)(19)(20). The numbers of TEM and SSA-SEM images containing specific features and the numbers of virus particles and dense tubules in each cell are summarized in Tables 1 and 2 and in Table 3, respectively. It should be noted that all AEC-IIs in the tables were desquamated from the basement membrane due to the host immune response, suggesting arrest of viral replication. We think that this is why we saw no virus budding in any of the images (Table 1). Also, intravesicular virus particles were observed only in AEC-II-6 (Tables 2 and 3). This might be because influenza viruses fused rapidly in the cytoplasm after adsorption to the cells, making it difficult to capture  Tables 2 and 3). Cytoplasm, light blue; nucleus, purple; virus particles within the cell, green; virus particles outside the cell membrane, red. images of intravesicular virus particles (30,31). In contrast, a lot of dense tubules (M1 proteins) were observed in the nuclei of four of six AEC-IIs (Tables 2 and 3). This is consistent with a report that M1 proteins remain in the nucleus for a long time (30). Taken together, the results show that IAV infects AEC-IIs. In contrast, neither budding of IAV particles nor intranuclear dense tubules were observed in Ms/Ms and Neus, suggesting no IAV infection of these cells. However, four of six Ms/Ms and one of three Neus harbored many virus particles within vesicles (Tables 2 and 3).
We found no evidence that IAV replicated in Ms/Ms within the autopsied lung specimen. In vitro IAV infection of human Ms/Ms results in productive infection and release of infectious particles (32,33) but can also result in abortive infection (34,35). The results of infection may depend upon the particular subtype or strain of virus involved, as well as on the subset of Ms/Ms (36)(37)(38).
There are least two mechanisms by which IAVs can enter cells. A spherical IAV may enter a host cell through clathrin-dependent receptor-mediated endocytosis (39,40). In addition, IAVs can induce macropinocytosis as an alternative entry pathway, independent of viral morphology (41,42). Here, SSA-SEM revealed the presence of many virus particles in vesicles and around the surfaces of Ms/Ms. In the 20th SSA-SEM image of M/M-2, multiple virus particles, including filamentous viruses, were observed in the extracellular region in which the cell membrane entered the cytoplasm; this suggests induction of micropinocytosis (Fig. 6c). Furthermore, 3D reconstruction revealed that many virus particles were located on one side of the M/M-2 surface (Fig. 8b). Further investigation is required to determine whether this arrangement was coincidental or related to the functions of the M/M. This study has several limitations. First the sample examined was obtained from a single human case. Second, virus particles and intranuclear dense tubules were identified morphologically but not confirmed by immunoelectron microscopy (immuno-EM). Unfortunately, we have no remaining samples available for immuno-EM studies. Since the specimens used for EM had been fixed with glutaraldehyde and osmium, their antigenicity has been destroyed. However, we point out that this is a common limitation of previous studies reporting EM images of influenza virus particles in human tissue (18,(26)(27)(28)(29). There are no studies reporting immuno-EM of influenza virus in human lung tissue. Third, because autopsied specimens were obtained on day 7 post-disease onset, most of the AECs were desquamated from the basement membrane due to the host immune response. Therefore, evaluation of negative findings in AEC-IIs was difficult. Finally, we did not capture images of virus budding. Capturing images of virus budding is extremely difficult; indeed, no such images in human lung tissue have been reported to date. Acquisition of such images in postmortem lung tissue is dependent on the time after death and on the site of the sample used for EM analysis.
Despite being based on a single human case, these viruses were observed in postmortem influenza virus-infected human tissues. Conducting studies using postmortem biopsied lung tissue that is appropriate for image analysis of viral particles is extremely difficult. Thus, previous EM analyses of influenza viruses were conducted mainly in animal models or cultured cells. Despite the difficulties, we succeeded in obtaining clear images of the virus in human lung tissue.
Additionally, the study used novel experimental methods that are unprecedented in the field of influenza research. First, we successfully obtained EM images of an entire alveolus, which can be viewed at low and high magnification, by stitching together SEM images. This technique enabled visualization of both virus particles and intranuclear dense tubules (M1 proteins) in a single EM image. Second, 3D reconstruction of SSA-SEM images of an AEC-II, a M/M, and a Neu revealed virus distribution within a single whole cell. Thus, the analysis provides insight that could not be obtained by conventional 2D imaging. This is the first report of the use of novel SEM methods to examine the distribution of IAV particles throughout a single whole cell in sections from an autopsy specimen. The differences between AEC-IIs, Ms/Ms, and Neus in terms of distribution of virus particles and virus-related structures suggest that each cell plays a different role in A/H1N1/pdm09 pneumonia.

MATERIALS AND METHODS
The patient. The subject of the study was a 33-year-old man who died of respiratory failure resulting from A/H1N1/pdm09 pneumonia complicated by development of ARDS on the 7th day after onset of disease. An autopsy was carried out 3 h after death. The detailed pathological and molecular-biological findings have been reported (8)(9)(10)(11). This study was approved by the institutional medical ethical committee of the National Institute of Infectious Diseases, Japan (approval no. 247).
Fixation and tissue processing. To prepare the specimens for EM, small parts of the lung tissue were excised and prefixed with 2% glutaraldehyde and 2.5% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 2 h at 4°C and then postfixed in 1% osmium tetroxide, dehydrated through a graded series of alcohols and propylene oxide, and embedded in epoxy resin. Other lung tissue specimens were routinely fixed in 10% buffered formalin and embedded in paraffin to prepare them for optical microscopy. Hematoxylin-eosin staining was performed for histological analysis, and immunohistochemistry was performed to evaluate the IAV nucleoprotein antigen, as previously described (8,10,11).
Sectioning for TEM and SEM. After trimming the epoxy resin-embedded tissue, ultrathin sections 70 nm thick were cut using an ultramicrotome (EM UC 6; Leica, Wetzlar, Germany), mounted on grids, and stained with 4% uranyl acetate and lead citrate for TEM using an H-7650 or an HT7700 (Hitachi, Tokyo, Japan). For SEM, ultrathin sections (200 nm) were cut using an ultramicrotome fitted with a histo-Jumbo diamond knife (DiAtome, Biel, Switzerland). Serial sections were placed on platinum-palladium-coated (MC1000; Hitachi, Tokyo, Japan) glass slides and allowed to adhere at 55°C for 30 min. The sections were stained with 1% toluidine blue for light microscopy and scanned with a NanoZoomer 2.0-RS digital slide scanner (Hamamatsu Photonics, Shizuoka, Japan). The observation regions were selected using NDP.view2 viewing software (Hamamatsu Photonics, Shizuoka, Japan). The sections were stained with 1% uranyl acetate and lead citrate and observed by ultra-high-resolution field emission SEM (SU8010; Hitachi, Tokyo, Japan). Regions of the ultrathin sections were imaged using backscattered electrons with an accelerating voltage of 1.5 kV. Images were taken at magnifications of ϫ7,000 to ϫ9,000, with a working distance of 3 mm. The workflows are shown in Fig. 3.
Stitching together the SEM images. To observe a whole cross-section of one alveolus, a 10-by-14 array of images was taken under the following conditions: accelerating voltage, 1.5 kV; working distance, 3 mm; magnification, ϫ5,000. After acquisition, the images were automatically aligned using the image stitching plugin in Fiji/Image J software (https://imagej.net/Image_Stitching) and viewed with the NDP.view2 viewing software. The workflows are shown in Fig. 3.
3D reconstruction of SSA-SEM images. Serial section images were aligned automatically using the StackReg plugin in Fiji/Image J (https://imagej.net/StackReg). The aligned images were carefully checked visually and corrected manually if necessary. Subsequently, they were imported into Photoshop software (Photoshop CS5.1; Adobe Systems, San Jose, CA, USA), and the cytoplasm, nucleus, and viruses were manually selected by tracing their boundaries. After identification of the target structures in the serial images, 3D surface-rendering reconstruction was performed in 3D Slicer (https://www.slicer.org/).

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.