Immediate myeloid depot for SARS-CoV-2 in the human lung

In the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic1, considerable focus has been placed on a model of viral entry into host epithelial populations, with a separate focus upon the responding immune system dysfunction that exacerbates or causes disease. We developed a precision-cut lung slice model2,3 to investigate very early host-viral pathogenesis and found that SARS-CoV-2 had a rapid and specific tropism for myeloid populations in the human lung. Infection of alveolar macrophages was partially dependent upon their expression of ACE2, and the infections were productive for amplifying virus, both findings which were in contrast with their neutralization of another pandemic virus, Influenza A virus (IAV). Compared to IAV, SARS-CoV-2 was extremely poor at inducing interferon-stimulated genes in infected myeloid cells, providing a window of opportunity for modest titers to amplify within these cells. Endotracheal aspirate samples from humans with the acute respiratory distress syndrome (ARDS) from COVID-19 confirmed the lung slice findings, revealing a persistent myeloid depot. In the early phase of SARS-CoV-2 infection, myeloid cells may provide a safe harbor for the virus with minimal immune stimulatory cues being generated, resulting in effective viral colonization and quenching of the immune system.


1
In the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic 1 , 2 considerable focus has been placed on a model of viral entry into host epithelial populations, 3 with a separate focus upon the responding immune system dysfunction that exacerbates or 4 causes disease. We developed a precision-cut lung slice model 2,3 to investigate very early 5 host-viral pathogenesis and found that SARS-CoV-2 had a rapid and specific tropism for

24
Here, we used precision-cut lung slices (PCLS) obtained from human lungs to study early host-25 pathogen responses in a system replete with the full repertoire of lung stromal and immune 26 cells.

28
We have previously developed a model of PCLS in the mouse lung 2,3 that we have now applied 29 to human lungs donated for research (Supplementary Table 1). A lung lobe was inflated using 30 2% low melting point agarose and 300 µm PCLSs were produced for tissue culture and direct 31 infection with SARS-CoV-2 (USA-WA1/2020; MOI 0.1 -1) for up to 72 hours (Fig. 1a). Imaging

41
Using both imaging and flow cytometry, we also observed spike and dsRNA signal in lung 42 immune cells (Fig. 1c and e), the former prevalent from the earliest 48-hour timepoint. Spike 43 was colocalized to CD45 + ACE2 + cells ( Fig. 1c; non-infected in Extended Data Fig. 1b) and similarly dsRNA was found in these cells (Extended Data Fig. 2b), supporting that concept that 45 immune cells may either be infected by SARS-CoV-2 or phagocytose the virus. Flow cytometry 46 allowed us to further characterize spike + and dsRNA + immune cells (Fig. 1e). We observed 47 significant dsRNA and spike signal in lung myeloid cells (CD45 + CD3 -CD19 -HLA-DR + CD14 + 48 cells, including interstitial macrophages, monocytes and monocyte-derived dendritic cells 7 ) at

68
Aligning the scRNA-seq data on the two viral genomes revealed that the main targets for IAV 69 infection were epithelial cells and fibroblasts (Fig. 2c), consistent with the observed loss of 70 these populations in PCLSs (Fig. 2b). In addition, IAV reads were also sporadically distributed

88
To investigate how SARS-CoV-2 affects human lung myeloid cells, we focused on the 89 dominant AMs, which are also anatomically within the airspaces and so directly exposed to

97
10% of the AMs (Fig. 3d, Extended Data Fig. 4a). Somewhat surprisingly, an MOI of 1 did not 98 significantly increase spike + AM percentage compared to MOI 0.1 (Fig. 3d), suggesting that 99 either cells were somehow protected at higher titers-perhaps due to increase antiviral sensing and subsequent ISGs-or that a plateau was reached in the number of cells that were capable 101 of being infected. At 48h, viability of AMs was high in both MOI groups, an effect that suggests 102 the virus was not inducing AM cell death (Extended Data Fig. 4f) contrary to observations in 103 blood monocytes from COVID-19 subjects 12 . However, of the spike + AMs, the majority were 104 ACE2 + (Extended Data Fig. 4g) pointing to a specific but not obligate role for ACE2 in licensing 105 AM viral entry. In support of this, when we used an ACE2 blocking antibody 13 incubated with 106 AMs 2h prior to SARS-CoV-2, we always observed a significant decrease in spike + AMs ( Fig.   107 3e) but this was rarely complete, even despite using saturating concentrations of the ACE2 108 antibody. Taken together, these data indicate that SARS-CoV-2 entry into AMs was 109 significantly mediated by ACE2 expression and did not require epithelial cells as an 110 intermediate host, since our BAL preparation lacked these cells.

112
Macrophage infection by viruses such as IAV has long been described as abortive, but several 113 studies have shown that the IAV H5N1 strain is capable of virus production in macrophages 14 .

114
To study the ability of AMs that were exposed to virus to produce and release new viruses, SARS-CoV-2 has been continuously evolving, resulting in the emergence of several variants 129 of concern (VOC), the most pathogenic of which has been the delta variant, which we directly 130 compared to ancestral USA-WA1/2020 in our system. As previously done, BAL cells were 131 infected with SARS-CoV-2 viruses for 48h. To determine SARS-CoV-2 viral production by BAL 132 cells, we used a plaque assay on BAL supernatant (Fig. 3h). BAL cells increased viral titer for 133 both ancestral and delta (Fig. 3h, Table 2) and analyzed 149 by scRNA-seq (Fig. 4a). Clustering showed that the cellular population was predominantly 150 composed of myeloid cells (Fig. 4a, far right), which contained macrophages, neutrophils, and 151 some DCs (Fig. 4c). Analysis of SARS-CoV-2 normalized expression revealed that SARS-

152
CoV-2 reads localized mainly to macrophages, but in some cases were also found in 153 neutrophils and T cells (Fig. 4b), similar to prior results 15 . The ETA samples were obtained at 154 different times after intubation, but the timing did not correlate with the quantity or SARS-CoV-155 2 reads (Extended Data Fig. 6). In fact, one subject was sampled at 40 days after intubation but still had detectable viral reads in macrophages, which may point to a long-lasting depot 157 effect. As in the PCLS model, several macrophage subpopulations (Extended Data Fig. 7) 158 were found to have SARS-CoV-2 reads with almost 25% of AMs being virus positive (Fig. 4d).

160
Finally, we analyzed differential gene expression in infected versus uninfected AMs obtained 161 from COVID-19 ETA samples. Multiple interferon-stimulated genes (ISGs) were increased in 162 infected compared to non-infected AMs in ETA samples (Fig. 4e), consistent with their being 163 exposed to viruses more profoundly than neighboring cells. To ask whether ISG expression 164 was a prominent feature of early infection, we returned to the PCLS system and compared 165 uninfected, IAV, and SARS-CoV-2 exposure. We indeed found that AMs exposed to SARS-

177
We also found multiple macrophage subpopulations with SARS-CoV-2 viral reads, and it is 178 certainly possible that these non-AM cells are also capable of productive infections. As ACE2 179 is an interferon-stimulated gene 17 , initial infection may be self-propagating, since our results

267
After agarose consolidation, 1 cm 3 lung tissue was placed on the precision compresstome VF-