Early reports indicate that non-human primates (NHPs) are suitable models for the pathology of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection/coronavirus disease 2019 (COVID-19), and for evaluating vaccine candidates1,2,3,4,5,6. These studies delineated the associated histopathology and morbidity at terminal end points for rhesus macaques, cynomolgus macaques and African green monkeys. To further expand the existing understanding of the NHP models of SARS-CoV-2 infection1,2,3,4,5,6, we also evaluated baboon and old marmoset models. Furthermore, we expanded on the rhesus macaque model of SARS-CoV-2 infection using computed tomography (CT) imaging and flow-cytometry-based cellular phenotyping, as well as by analysing alveolar and peripheral cytokine responses during the course of the disease development and resolution. We also complemented the data with matched immunological and histopathological findings in organs from the terminal end points. We sought to characterize early cellular immune events in the lungs after SARS-CoV-2 infection in three NHP genera: Indian rhesus macaques (old and young), baboons (old and young) and common marmosets (old). In the early stages, macaques developed clinical signatures of viral infection and systemic inflammation, early evidence of viral replication, mild to moderate interstitial and alveolar pneumonitis, and moderate progressive pneumonia. In longitudinal studies over two weeks, young and old macaques showed early signs of COVID-19 with recovery in both age groups. Analysis of blood and bronchoalveolar lavage (BAL) revealed a complex early inflammatory milieu with influx in the lungs of innate and adaptive immune cells, particularly myeloid cells, and a prominent type-I interferon (IFN) response. Whereas rhesus macaques exhibited moderate disease, baboons were susceptible to SARS-CoV-2 infection with extensive pathology after infection, and marmosets demonstrated mild infection. Thus, different NHP species exhibit heterogeneous responses to SARS-CoV-2 infection. Rhesus macaques and baboons develop different, quantifiable disease attributes making them essential models to test vaccines and therapeutics against COVID-19.


Heterogeneity in SARS-CoV-2 viral loads in young and old NHPs

Macaques, baboons and old marmosets were infected by multiple routes (ocular, intratracheal and intranasal) with sixth-passage, fully sequenced and authenticated virus at a target dose of 1.05 × 106 plaque-forming units (p.f.u.) per animal. SARS-CoV-2 viral RNA (vRNA) was detected early in all species at 3 d post-infection (d.p.i.; Fig. 1a–o and Supplementary Table 1) and decreased thereafter at variable rates. Comparable BAL vRNA levels were detected in young and old macaques at 3 d.p.i. (5/6 each; Fig. 1a). Almost no BAL vRNA was detected at 9 d.p.i. (1/12) and none was detected at 12 d.p.i. (Fig. 1a). vRNA in nasopharyngeal swabs (NS) could be detected in 50% of animals at 3 d.p.i., 10/12 (6 young, 4 old) at 9 d.p.i. and 6/12 at the study end (Fig. 1b). vRNA was detected in 4/12 and 2/12 animals, respectively, from rectal swabs (RS; Fig. 1c) and in buccopharyngeal swabs (BS; Extended Data Fig. 1a) at 3 and 6 d.p.i., but infrequently at later time points. vRNA was detected in the lungs of 8/12 (3 young, 5 old) macaques at necropsy (14–17 d.p.i.; Fig. 1d). No vRNA was detected in any plasma samples (Extended Data Fig. 1b) or in randomly selected urine samples (Extended Data Fig. 1c). We also detected no SARS-CoV-2 subgenomic RNA (a correlate for infectious/replicating virus) in either rhesus (Extended Data Fig. 1d) or baboon (Extended Data Fig. 1e) lungs in the longitudinal study (Fig. 1m). Thus, despite vRNA persistence in the lungs of immunocompetent macaques, the absence of replicative virus indicates that macaques control SARS-CoV-2 infection.

Fig. 1: SARS-CoV-2 RNA and histopathology in rhesus macaques, baboons and marmosets.
figure 1

ad, vRNA in BAL fluid (a), NS (b) and RS (c) collected longitudinally, and lung tissue homogenates (d) collected at the end point (14–17 d.p.i.) from rhesus macaques infected with SARS-CoV-2. eh, vRNA in BAL fluid (e), NS (f) and RS (g) collected longitudinally, and lung tissue homogenates (h) at the end point (14–17 d.p.i.) from baboons infected with SARS-CoV-2. n = 12. il, Comparison of vRNA in BAL fluid (i), NS (j), RS (k) and lungs (l) of rhesus macaques and baboons infected with SARS-CoV-2. m,n, To estimate the persistence of replicative virus, we performed subgenomic RNA estimation on end-point lung samples of rhesus macaques (m) and baboons (n). n = 12. o,p, vRNA in nasal wash (o) and oral (p) longitudinal swabs. n = 6 (0–3 d.p.i.) and n = 4 (6–14 d.p.i.). q, vRNA was also measured in lung homogenates of marmosets at the end point. n = 2 (3 d.p.i.) and n = 4 (14 d.p.i.). Statistical analysis was performed using one-way (ac and eg) and two-way (ik) repeated-measures analysis of variance (ANOVA) with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8). r, Histopathological analysis in infected rhesus macaques revealed regionally extensive interstitial lymphocytes, plasma cells, lesser macrophages and eosinophils expanding the alveolar septa (bracket) and alveolar spaces filled with macrophages (asterisks). The normal alveolar wall is indicated (arrow) for comparison (top). Bottom, alveolar spaces with extensive interstitial alveolar wall thickening by deposits of collagen (asterisks) and scattered alveolar macrophages (arrow). s, Histopathological analysis in infected baboons also revealed regionally extensive interstitial lymphocytes, plasma cells, lesser macrophages and eosinophils expanding the alveolar septa (bracket) and alveolar spaces filled with macrophages (asterisk) (top). Bottom, alveolar wall thickening by interstitial deposits of collagen (asterisk), alveoli lined by occasional type-II pneumocytes (arrowheads) and alveolar spaces containing syncytial cells (arrows) and alveolar macrophages. t, Histopathological analysis in marmosets revealed a milder form of interstitial lymphocytes, and macrophages recruited to the alveolar space. Scale bars, 100 μm (r, s (top), t (top)) and 50 μm (s (bottom) and t (bottom)). u,v, Comparison of the lung inflammation score (u) and end-point viral titre (v) of infected rhesus macaques and baboons. n = 12. Statistical analysis was performed using one-tailed Mann–Whitney U-tests (m, n, u and v). For aq, u and v, the shapes indicate old (triangles) and young (diamonds) rhesus macaques; old (inverted triangles) and young (squares) baboons; and old marmosets (hexagons). The different colours represent individual animals (Supplementary Table 1). For a, b, eg, ik, mo, u and v, P values are indicated above the plots. Data are mean (i, k and q) or mean ± s.e.m. (ah, mo, u and v).

The pattern of vRNA in baboons mimicked that of macaques (Fig. 1e–h). At 3 d.p.i., 8/12 and 10/12 baboons had detectable vRNA in BAL (Fig. 1e) and NS (Fig. 1f), respectively. A comparatively higher number of old baboons (4/6) harboured BAL vRNA at 9 d.p.i. relative to young baboons (0/6) (Fig. 1e). Despite these differences, peak vRNA in BAL (~4 log) was detected in both age groups at 3 d.p.i. (Fig. 1e). vRNA levels detected in NS (Fig. 1f) and RS (Fig. 1g) of old baboons were higher compared with swabs from young baboons. Peak vRNA levels were detected at 3 d.p.i. (6/6 old and 4/6 young baboons) in NS. At 6 and 9 d.p.i., 4/6 old (mean > 3 log) and only 1/6 young (mean < 1 log) baboons, respectively, were positive for vRNA in NS (Fig. 1f). Peak vRNA was detected at 9 d.p.i. for both age groups of baboons in RS (Fig. 1g), with an average of ~5 log in old baboons, which was higher than in young baboons (~3 log; Fig. 1g). vRNA was also detected in the lungs of five old baboons at necropsy (14–17 d.p.i.; Fig. 1h) but no SARS-CoV-2 subgenomic RNA was detected (Fig. 1n and Extended Data Fig. 1e). Although no statistically significant differences were observed for BAL vRNA levels (Fig. 1i) and NS vRNA levels (Fig. 1j) between baboons and macaques, RS showed substantial differences; baboons harboured greater levels (by several log) throughout the infection protocol (Fig. 1k). Differences in lung viral titres were not significantly different between rhesus and baboons (Fig. 1l).

Less than 4 log of vRNA was detected in NS from infected old marmosets, peaking at 3 d.p.i., and 1/6 animals was also positive at 6 d.p.i. (Fig. 1o). No vRNA was detected in BS throughout the study (Fig. 1p). In comparison to macaques and baboons, marmosets had accelerated clearance of vRNA. Only low lung vRNA levels were detected at study end (Fig. 1q).

Gross examination at necropsy (14–17 d.p.i.) identified red discoloration of the lung lobes in 50% of macaques (Extended Data Fig. 2b) and 100% of baboons (Extended Data Fig. 3a,b). Supplementary Tables 2 (macaques) and 3 (baboons) summarize the histopathological findings. The lungs were the most affected organ in each case (Extended Data Figs. 2a–d and 3, and Supplementary Tables 2 and 3). Multifocal minimal to mild interstitial mononuclear inflammation was seen in 11/12 macaques (Fig. 1r, Supplementary Table 2 and Extended Data Fig. 2c) and 12/12 Baboons (Fig. 1s and Extended Data Fig. 3d–g), generally composed of macrophages and lymphocytes that expanded the alveolar septa (Fig. 1r,s and Extended Data Fig. 2d–g), with variable neutrophil infiltrates (Extended Data Figs. 2e and 3d,e,h,i) or fibrosis (Fig. 1r,s and Extended Data Figs. 2f,g and 3h,i). A subset of six marmosets was euthanized at 3 d.p.i. (n = 2), while others were necropsied at 14 d.p.i. Interstitial and alveolar pneumonitis was observed in the marmosets (Fig. 1t), although not as prevalent as in macaques or baboons. Thus, our results show that three NHP genera develop different degrees of COVID-19 when evaluated side by side, with baboons exhibiting moderate to severe pathology, macaques exhibiting moderate pathology and old marmosets exhibiting mild pathology. The pulmonary inflammation score was significantly different between macaques and baboons, with greater inflammation in baboon lungs (Fig. 1u); however the vRNA levels did not differ statistically (Fig. 1v).

Longitudinal SARS-CoV-2 infection in rhesus macaques, baboons and marmosets demonstrates heterogeneity in radiological and clinical outcomes across age and species

All SARS-CoV-2-infected macaques (Fig. 2a and Supplementary Table 4) and baboons (Fig. 2b and Supplementary Table 5) exhibited low baseline chest X-ray (CXR) scores (Fig. 2a,b and Supplementary Tables 4 and 5) with no age differences (Fig. 2a,b), but there were significantly higher CXR scores in baboons compared with in macaques (Fig. 2c) at day 9 and the study end. Several infected macaques showed changes that were consistent with pneumonia (Supplementary Table 4) with peak severity at 3–6 d.p.i. and a decline by the study end (Fig. 2a and Supplementary Table 4). Examples of extensive pneumonia were seen in CXRs in macaques at 6 d.p.i. with subsequent resolution at the end point (Supplementary Fig. 1a–c). Several animals exhibited multilobe alveolar infiltrates and/or interstitial opacities at 6 d.p.i., while others exhibited progressive, moderate to severe interstitial and alveolar infiltrates at 6 d.p.i. that resolved by day 14. By contrast, CXRs of all procedure control animals (which underwent repeated BAL procedures) revealed minimal to no findings.

Fig. 2: Radiology and correlates of SARS-CoV-2 infection in rhesus macaques and baboons.
figure 2

ac, CXR scores in macaques (a; n = 12) and baboons (b; n = 12), and comparative CXR scores of infected rhesus macaques and baboons (c). d, CT scores in macaques. n = 6. e,i, Three-dimensional (3D) reconstruction of the region of interest (ROI) volume representing the location of the lesion at day 6 (e) and day 12 (i). fh,jl, Representative longitudinal (f,j), vertical (g,k) and transverse (h,l) images for the quantification of lung lesions at day 6 (fh) and day 12 (jl); the teal area represents normal-intensity voxels and the yellow areas represent hyperdense voxels. m, The percentage change in lung hyperdensity in animals infected with SARS-CoV-2 over 6 d.p.i. compared with over 12 d.p.i. n = 6. ns, Clinical correlates of early SARS-CoV-2 infection in rhesus macaques (circles) over 0–3 d.p.i. showing changes in serum CRP (n), albumin (ALB) (o), haemoglobin (HGB) content (p) in peripheral blood; vRNA in BAL fluid (q) and NS (r) and BS (s) collected longitudinally. t,u, vRNA (t) and subgenomic RNA (u) was measured in lung tissue homogenates at the end point (3 d.p.i.). v, Side-by-side comparison of vRNA and subgenomic RNA. For nv, the different colours represent individual animals (Supplementary Table 1). For nv, n = 4. w,x, Comparison of the viral titre (w) and lung subgenomic RNA (x) of infected rhesus macaques between a short (3 d) and a long (14–17 d) study. n = 4 (3 d) and n =12 (14–17 d). Statistical analysis was performed using Mann–Whitney U-tests. y, Haematoxylin and eosin staining was performed on formalin-fixed paraffin-embedded lung sections from infected animals for pathological analysis. Top, histopathological analysis revealed bronchitis characterized by infiltrates of macrophages, lymphocytes, neutrophils and eosinophils that expanded the wall (bracket) and, along with syncytial cells (arrows), filled the bronchiole lumen and the adjacent alveolar spaces. Bottom, suppurative interstitial pneumonia with type-II pneumocyte hyperplasia (arrowheads); the alveolar space is filled with neutrophils, macrophages and fibrin (asterisk). The bracket denotes the alveolar space. Scale bars, 100 μm (top) and 50 μm (bottom). z, Multilabel confocal immunofluorescence microscopy of the lungs (top) and nasal epithelium (bottom) at ×63 magnification with nucleocapsid (N)-specific antibodies (turquoise) 4,6-diamidino-2-phenylindole (DAPI) (blue) and ACE2 (magenta). Scale bars, 20 μm. Undetectable viral titres were represented as one copy. Statistical analysis was performed using one-way (a, b and ns) and two-way (c) repeated-measures ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8); and with one-tailed Mann–Whitney U-tests (d, m and vx). For ad, ms and vx, P values are indicated above the plots. Data are mean (tv) or mean ± s.e.m. (a, b, d, ms, w and x).

High-resolution lung CT imaging was performed before and after SARS-CoV-2 infection in six young and six old macaques (Fig. 2d–m). CT scans for baboons and both CT scans and CXR imaging for marmosets were not feasible. Pneumonia was present in all animals after infection, but to a significantly higher degree in old macaques compared with in young macaques (Extended Data Fig. 4 and Supplementary Table 6). At 6 d.p.i., severe patchy alveolar patterns were observed in some lobes, while others had milder interstitial patterns, with moderate to severe ground-glass opacities primarily in old macaques (Extended Data Fig. 4a–f). Resolution of many ground-glass opacities and nodular as well as multifocal lesions was observed in all animals at 12 d.p.i. (Fig. 2d and Extended Data Fig. 4a,b,d–f). At 12 d.p.i., all but one of the old macaques exhibited a normal or nearly normal CT scan. Findings in one old macaque were considerably improved but retained patchy ground-glass opacities in all lobes and alveolar patterns in some lobes at 12 d.p.i. (Extended Data Fig. 4c). This animal had the highest overall score on the basis of CT scans (Fig. 2d) and CXRs (Fig. 2a,b). These results suggest that multilobe pneumonia may persist for longer in some old macaques. Hyperdensity analysis revealed a significant progressive increase in involved lung volume at 6 d.p.i., which normalized by 12 d.p.i. (Fig. 2e–m).

Fig. 3: The alveolar compartment over 0–3 d.p.i. in infected rhesus macaques.
figure 3

a,e, CXR (a) and CT (e) scores of rhesus macaques (circles) over 0–3 d.p.i. The different colours represent individual animals (Supplementary Table 1). bd, Representative CT scan images performed on day 0–2 d.p.i. show transverse (b), vertical (c) and longitudinal (d) views of the left caudal lobe ground-glass opacity at 1 d.p.i. (middle), 2 d.p.i. (bottom) and the baseline at 0 d.p.i. (top). The CT scans in bd revealed evidence of pneumonia and lung abnormalities in the infected animals relative to controls, which resolved between 1 and 2 d.p.i. (red arrows). f, 3D reconstruction of the ROI volume representing the location of the lesion. gi, Representative longitudinal (g), vertical (h) and transverse (i) images for the quantification of lung lesions; the teal area represents normal-intensity lung voxels, while the yellow areas represent hyperdense voxels. j, The percentage change in lung hyperdensity in SARS-CoV-2-infected animals over days 1–3 d.p.i. compared with the baseline. kv, Simultaneous analysis of multiple cytokines using Luminex technology in BAL fluid of rhesus macaques over 0–3 d.p.i. revealed that SARS-CoV-2 induced alveolar inflammation, showing increased levels of IL-6 (k), IFNα (l), IFNγ (m), IL-8 (n), perforin (o), IP-10 (p), MIP1a (q), MIP1b (r), IL-12p40 (s), IL-18 (t), TNFα (u) and IL-1Ra (v), expressed as log10-transformed concentration in pg ml−1 of BAL fluid. n = 4. For a and kv, statistical analysis was performed using one-way repeated-measures ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8). For e and j, statistical analysis was performed using ordinary one-way ANOVA with Dunnett’s post hoc test. n = 4 (0–2 d.p.i.) and n = 2 (3 d.p.i.). For a, e and jv, P values are indicated above the plots; data are mean ± s.e.m.

Early events in SARS-CoV-2 infection of rhesus macaques

We performed an independent acute 3 d infection study of rhesus macaques that included extensive imaging and immunological readouts. Four rhesus macaques (Supplementary Table 1) were infected as described above and all of the animals developed clinical signs of viral infection with a doubling of serum C-reactive protein (CRP) levels relative to the baseline (Fig. 2n); significantly decreased serum albumin (Fig. 2o) and haemoglobin (Fig. 2p) levels, indicating viral-induced anaemia; and progressively increasing total serum CO2 levels (Extended Data Fig. 5a), indicative of pulmonary dysfunction.

As before, vRNA was detected in BAL, NS and BS at 1–3 d.p.i. (Fig. 2q–s) and in RS at 1 d.p.i. (Extended Data Fig. 5i). At necropsy (3 d.p.i.), SARS-CoV-2 vRNA was detected in 23/24 random lung sections analysed. We detected 6–8 log-transformed copies per 100 mg of lung tissue from every lobe (Fig. 2t). The ~4 log increase in BAL vRNA from 1 to 2 d.p.i. (Fig. 2q) was consistent with early active replication of SARS-CoV-2, a finding that was verified by significant subgenomic vRNA levels (6 log) at 3 d.p.i. (Fig. 2u), which was ~2 log lower than detected vRNA (Fig. 2v). Thus, SARS-CoV-2 induces a rapid replicative response in the lungs that is evident within 3 d and that subsides by two weeks, with low vRNA (Fig. 2w) and no subgenomic vRNA detected in lung samples at 14–17 d.p.i. (Fig. 2x). The vRNA titres in BAL and NS of rhesus macaques demonstrated a strong positive correlation with the vRNA titres in the lungs at both end points (Extended Data Fig. 1f–h).

Examination at day 3 necropsy revealed findings of interstitial and alveolar pneumonia (Fig. 2y). A summary of the histopathological findings is provided in Supplementary Table 7. The lungs were again the most affected organ (Fig. 2y, Supplementary Table 7 and Extended Data Fig. 6). Multifocal, mild to moderate interstitial pneumonia characterized by infiltrates of neutrophils, macrophages, lymphocytes and eosinophils was present in all four animals (Fig. 2y (bottom) and Extended Data Fig. 6d,e,g,h), and was accompanied by variable fibrosis (4/4; Extended Data Fig. 6e), fibrin deposition (3/4; Extended Data Fig. 6c) and vasculitis (3/4; Extended Data Fig. 6f). All four macaques exhibited the following: (1) syncytial cells in the epithelial lining and/or alveolar lumen (Extended Data Fig. 6e,g,k); (2) bronchitis characterized by infiltrates of eosinophils within the bronchial wall and epithelium (Fig. 2y (top) and Extended Data Fig. 6i–k); (3) bronchus-associated lymphoid tissue hyperplasia (Extended Data Fig. 6i); and (4) minimal to moderate lymphoplasmacytic and eosinophilic tracheitis and rhinitis.

Fluorescence immunohistochemical analysis revealed SARS-CoV-2 proteins in the lungs (Fig. 2z (top) and Supplementary Figs. 2a,g and 3a,d,g,j), nasal epithelium (Fig. 2z (bottom) and Supplementary Figs. 2b,h and 3b,e,h,k) and tonsils (Supplementary Figs. 2c,i and 3c,f,i,l). In all tissues, including the lungs (Fig. 2y (top) and Supplementary Fig. 2a,g), nasal epithelium (Fig. 2z (top) and Supplementary Fig. 2b,h) and tonsils (Supplementary Fig. 2c,i), N antigen was detected in cells expressing ACE2 and in adjacent cells, consistent with SARS-CoV-2 using ACE2 to bind to, internalize and proliferate inside cells7,8. ACE2 protein levels were much lower in lung tissues derived from naive animals compared with SARS-CoV-2-infected animals (Supplementary Fig. 2m,n), a finding that is consistent with a recent RNA-sequencing analysis showing that ACE2 is upregulated after infection, especially in young macaques9. The majority of S signal was detected in the epithelial layer with discrete distribution throughout the lung tissue (Supplementary Fig. 3a,d). In the nasal cavity, virus was observed in epithelial lining cells (Supplementary Fig. 3b,h), but was distributed throughout tonsillar tissue (Supplementary Fig. 2c,i). Thus, viral replication is supported in the upper and lower lung compartments during the first 3 d of infection and viral antigens are detected at high levels in the lungs.

CXRs of all four infected macaques showed progressive increase in CXR abnormality scores (Fig. 3a and Extended Data Fig. 7a). The CXR scores at 2 and 3 d.p.i. were significantly increased relative to the baseline (Fig. 3a), with partial resolution of specific lesions at 2 or 3 d.p.i. (Extended Data Fig. 7a). There were mild to severe multifocal interstitial-to-alveolar patterns with soft tissue opacities in various lobes or diffusely in some animals, with more-severe abnormalities in the lower lung lobes, and with the most severe findings at 3 d.p.i. (Extended Data Fig. 7a). Lung CT scans showed increased multifocal pulmonary infiltrates with ground-glass opacities in various lobes, linear opacities in the lung parenchyma, nodular opacities in some lung lobes and increased soft tissue attenuation extending primarily adjacent to the vasculature within 1 d.p.i. (Fig. 3b–d and Extended Data Fig. 7b–e). In some of the animals, multifocal alveolar pulmonary patterns and interstitial opacities were observed in lobe subsections, with soft tissue attenuation and focal border effacement with the pulmonary vasculature. Features intensified at 2–3 d.p.i., primarily in the lung periphery, but also adjacent to the primary bronchus and the vasculature (Fig. 3b–d and Extended Data Fig. 7b). In other animals, progressive alveolar or interstitial pulmonary patterns were observed at 2 d.p.i. (Fig. 3b–d). Although ground-glass opacities intensified at 2 d.p.i. in some lobes, others resolved (Extended Data Fig. 7c,d). In one animal, the individual nodular pattern at 1 d.p.i. evolved to a multifocal soft tissue nodular pattern in multiple lobes with associated diffuse ground-glass opacities (Extended Data Fig. 7d). At 3 d.p.i., persistent, patchy, fairly diffuse ground-glass pulmonary opacities existed in many lung lobes with multifocal nodular tendency (Extended Data Fig. 7e). Overall, CT abnormality scores continuously increased over 3 d.p.i. (Fig. 3e). The percentage change in the hyperdensity volume was calculated using CT scans to quantify pathological changes over time10. We observed a substantial increase in lung hyperdense areas from 1 to 3 d.p.i. (Fig. 3f–j). Together, CXR and CT scans revealed moderate multilobe pneumonia in all of the infected animals (Fig. 3j), confirming the histopathology results (Fig. 2y and Extended Data Fig. 6) in the very early phase of SARS-CoV-2 infection in macaques.

IL-6, IFNα, IFNγ, IL-8, perforin, IP-10, MIP1α and MIP1β (Fig. 3k–v) were all significantly elevated in BAL fluid of acutely infected macaques. Of particular interest was the elevated IFNα (Fig. 3l), which has critical anti-viral activity, including against SARS-CoV-2 (ref. 11). Expression of a downstream type-I-IFN-regulated gene CXCL10 (which encodes IP-10), which promotes the recruitment of CXCR3+ T helper 1 (TH1) cells, was also increased (Fig. 3p). Macaques therefore mount an early anti-viral response to SARS-CoV-2 infection. Type-I IFNs and IL-6 are key components of a ‘cytokine storm’ that promote acute respiratory-distress syndrome, which is associated with both SARS-CoV-1 and SARS-CoV-2 when induced uncontrollably12. Of the plasma cytokine levels tested (Extended Data Fig. 8), IFNα and IP-10 were also significantly elevated at 2 and 3 d.p.i. (Extended Data Fig. 8b,f). Thus, clinical, imaging, pathology and cytokine analyses provided evidence for an acute SARS-CoV-2 infection in macaques that leads to moderate pneumonia with early activation of anti-viral responses.

Myeloid cell response in the lungs of infected rhesus macaques

BAL predominantly comprises alveolar macrophages (AMs) in healthy lungs13. BAL and peripheral blood cellular composition14,15 of SARS-CoV-2-infected macaques in our longitudinal study showed substantially altered immune cell composition and responses at necropsy. Infection moderately increased the proportions of BAL myeloid cells, interstitial macrophages (IMs; Fig. 4a,e), neutrophils (Fig. 4c,g) and plasmacytoid dendritic cells (pDCs; Fig. 4d,h)) at 3 d.p.i. with no age effect. By contrast, resident AM levels decreased significantly at 3 d.p.i. (Fig. 4b,f). The increase in myeloid subpopulations at 3 d.p.i. was highly correlated with vRNA levels (Fig. 4i,j). Multilabel confocal imaging of Ki67 staining showed that few of the virally infected cells in lung tissue actively proliferated (Fig. 4k–m). Neutrophils (Fig. 4k–m and Extended Data Fig. 9a,i), macrophages (Fig. 4n–p and Extended Data Fig. 9b,j) and pDCs (Fig. 4q–s and Extended Data Fig. 8c,k) recruited to the lung harboured high levels of viral proteins (Fig. 4k–s and Extended Data Fig. 9). Furthermore, resident epithelial cells, type-1 and type-2 pneumocytes, all harboured viral antigens (Fig. 4t–v and Extended Data Fig. 9d,l). These are also reported to be the primary alveolar cells with high ACE2 expression and SARS-CoV-2 signal in NHPs16 and humans17. These data suggest that early infection of resident alveolar cells might trigger the cytokine storm that induces rapid lung influx of specialized myeloid cell subsets that are known to express type-I IFNs and other proinflammatory cytokines, and is a key event in the control of SARS-CoV-2 infection. Relative to AMs, IMs have a shorter half-life, exhibit continuous turnover, help to maintain homeostasis and protect against continuous pathogen exposure from the environment18. Increased lung recruitment of pDCs suggests that they are a potentially important feature of protection against advanced COVID-19 because they are a major source of type-I IFNs, such as IFNα, which is elevated in BAL within 1–3 d.p.i. (Fig. 3l and Extended Data Fig. 8b).

Fig. 4: Accumulation of myeloid cells in BAL of infected rhesus macaques.
figure 4

ah, Flow cytometry analysis of BAL IMs (a,e), AMs (b,f), neutrophils (c,g) and pDCs (d,h). For ad, data are combined for age (n = 12). For eh, data are split by age (n = 6). Statistical analysis was performed using one-way (ad) and two-way (eh) repeated-measures ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8). The shapes indicate old (triangles) and young (diamonds) animals. The different colours represent individual animals (Supplementary Table 1). n = 12. i,j, Spearman’s rank correlation between the cellular fraction and log10-transformed vRNA copy number in BAL was calculated (i) and the corresponding values for the Spearman’s rank correlation coefficient (j, left) and P values (j, right) are shown. Multilabel confocal immunofluorescence microscopy of formalin-fixed paraffin-embedded lung sections from rhesus macaques infected with SARS-CoV-2 with a high viral titre at 3 d.p.i., stained with DAPI (blue) (kv); SARS-CoV-2 spike-specific antibodies (ks) (turquoise); Ki67 (magenta) and neutrophil marker CD66abce (yellow) (km); pan-macrophage marker CD68 (magenta) (np); HLA-DR (magenta) and pDC marker CD123 (yellow) (qs); SARS-CoV-2 nucleocapsid-protein-specific antibodies (turquoise), pan-cytokeratin (magenta) and thyroid transcription factor-1 (yellow) (tv). For kv, scale bars, 20 μm (×63 magnification; k, n, q and t), 50 μm (×20 magnification; l, o, r and u) and 100 μm (×10 magnification; m, p, s and v). For ah, P values are indicated above the plots; data are mean ± s.e.m.

T-cell, antibody and cytokine responses in the lungs of infected rhesus macaques

Macaque infection resulted in a significant T-cell influx into the alveolar space by 3 d.p.i., which normalized by 9 d.p.i. (Fig. 5a–c) and correlated significantly with BAL viral titre (Fig. 5d–f). After infection, CD4+ T cells in BAL expressed significantly lower levels of antigen experience/tissue residence (CD69), TH1 (CXCR3), memory (CCR7) and activation (HLA-DR; Fig. 5g–k) markers. By contrast, the level of CD4+ T cells expressing PD-1 and LAG-3 (Fig. 5l,m) was significantly increased. CXCR3 expression on CD4+ T cells showed significant negative correlation with viral titre, whereas PD1 expression was positively, albeit not significantly, correlated (Fig. 5n–p). A similar effect was observed in BAL CD8+ T-cell subsets, in which CD69, CXCR3 and CCR7 expression (Fig. 5q–u) was significantly reduced after infection, whereas PD-1 and LAG-3 expression (Fig. 5v,w) was significantly increased. PD1 expression on CD8 subsets was positively correlated with BAL viral titre, whereas CXCR3 expression was inversely correlated (Fig. 5n–p). There were no age-related differences in T-cell responses. Rapid influx of myeloid cells expressing type-I IFNs probably results in immune control of SARS-CoV-2 infection, but enables viral antigens to persist, which recruits T cells with a profile associated with immune activation in the effector phase.

Fig. 5: T cells in BAL of infected rhesus macaques.
figure 5

ac, Frequencies of CD3+ T cells (a), CD4+ T cells (b) and CD8+ T cells (c) in BAL. df, Spearman’s rank correlation of the cell types in ac between the cellular fraction and the log10-transformed vRNA copy number in BAL was calculated (d) and the corresponding Spearman’s rank correlation coefficient values (e) and P values (f) are shown. gm, CD4+ T-cell subsets expressing the early activation marker CD69 (g), CXCR3 (h), the memory marker CCR7 (i), CCR5 (j), HLA-DR (k), PD-1 (l) and LAG-3 (m). np, Spearman’s rank correlation between PD1 and CXCR3 expression on CD4+ and CD8+ T cells and log10-transformed vRNA copy number in BAL was calculated (n) and the corresponding Spearman’s rank correlation coefficient values (o) and P values (p) are shown. qw, CD8+ T-cell subsets expressing the early activation marker CD69 (q), CXCR3 (r), the memory marker CCR7 (s), CCR5 (t), HLA-DR (u), PD-1 (v) and LAG-3 (w). The shapes indicate old (triangles) and young (diamonds) animals. The different colours represent individual animals (Supplementary Table 1). n = 12. For ac, gm and qw, data are mean ± s.e.m. n = 12. Statistical analysis was performed using two-way repeated-measures ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8). For ac, gm and qw, P values are indicated above the plots.

To extrapolate from phenotype to function, we explored proliferation, immune mediator production and memory phenotypes. BAL CD4+ and CD8+ T cells exhibiting proliferative (Fig. 6a,g) and memory markers (Fig. 6b,h) were significantly increased after infection, whereas those expressing naive (Fig. 6c,i) and effector (Fig. 6d,j) phenotypes were significantly reduced. The percentage of CD4+ (Fig. 6e) and CD8+ (Fig. 6k) T cells expressing IL-2 and granzyme B (GZMB; Fig. 6f,l) were significantly elevated. No significant effect of age was observed, although IL-2 expression on T cells was higher for young macaques compared with old macaques. These results suggest that the induction of robust T-cell immune responses (both CD4+ and CD8+ T cells) is generated in the lungs (BAL) as early as day 3 and maintained at 9 d.p.i. in many cases. After ex vivo restimulation of T cells from BAL at 9 d.p.i. with CoV-specific peptide pools, CD4+ and CD8+ T cells expressing IL-2, GZMB, IFNγ, IL-17 and TNFα were not significantly elevated beyond baseline values until day 12 (Fig. 6).

Fig. 6: Memory T cells in BAL of infected rhesus macaques.
figure 6

af, Frequencies of the following CD4+ T-cell subsets: expressing Ki67 (a), memory (b), naive (c), effector (d), expressing IL-2 (e) and expressing GZMB (f) in BAL. gl, Frequencies of the following CD8+ T-cell subsets: expressing Ki67 (g), memory (h), naive (i), effector (j), expressing IL-2 (k) and expressing GZMB (l). mp, Cells from BAL were stimulated overnight (12–14 h) with mock control (U), phorbol 12-myristate 13-acetate (PMA)-ionomycin (P/I) or SARS-CoV-2-specific peptide pools of the nucleocapsid (N), membrane (M) and spike (S) proteins. Antigen-specific cytokine secretion in T cells was estimated using flow cytometry. The fraction of CD4+ T cells secreting IL-2 (m) and GZMB (n); and CD8+ T cells secreting IL-2 (o) and GZMB (p). The shapes indicate old (triangles) and young (diamonds) animals. The different colours represent individual animals (Supplementary Table 1). n = 12. Statistical analysis was performed using two-way repeated-measures ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8). q, The SARS-CoV-2 spike-protein-specific antibody titre in the plasma of rhesus macaques at the end point compared with the naive control. The shapes indicate old (triangles) and young (diamonds) animals. The different colours represent individual animals (Supplementary Table 1). n = 4 (naive), n = 6 (old) and n = 6 (young). Statistical analysis was performed using ordinary one-way ANOVA with Dunnett’s post hoc test (GraphPad Prism 8). For al and q, P values are indicated above the plots. Data are mean ± s.e.m. (aq).

The immunophenotyping results were confirmed by studying cytokine production in BAL and plasma longitudinally19 (Extended Data Fig 10) and using spike-protein-specific antibodies at necropsy (Fig. 6q). IFNα, IL-1Ra and IL-6 (Extended Data Fig. 10a–c) were elevated in BAL after infection, but levels rapidly normalized after the 3 d.p.i. peak. IFNα levels were also induced in plasma, but the levels of the other cytokines studied were not (Extended Data Fig. 8g). Overall, the longitudinal study immunological results were consistent with the acute infection study in the expression of type-I proinflammatory cytokines (IFNα) and IL-6. The plasma of these macaques contained high levels of viral spike-protein-specific IgG at necropsy (Fig. 6q). The levels were significantly higher in young versus old animals. Altered antibody production, coupled with T-cell exhaustion, in the setting of chronic type-I IFN signalling has been reported earlier for other chronic viral infections20,21,22. In the face of chronic T-cell exhaustion, IL-10 responses may result in poor downstream antibody formation, as observed in our results.

As pulmonary pathology was significantly higher in baboons compared with in rhesus macaques (Fig. 1u), we studied the expression of various proinflammatory and protective cytokines and chemokines in BAL fluid derived from both young and old baboons. MIF, IL-6, CRP and IP-10 (Supplementary Fig. 4a–d) expression was significantly induced; the magnitude was greater in old animals (Supplementary Fig. 4a–d). BAL and serum levels of several other cytokines, chemokines and inflammatory markers were also increased (Supplementary Figs. 4e–m and 5), including IL1Ra, perforin and IL-8. Plasma IP-10, RANTES and IFNα levels were elevated (Supplementary Fig. 5a–d). The higher persistence of vRNA in some compartments (Fig. 1i–l) coupled with a greater severity of inflammation in the lungs (Fig. 1u) suggests that baboons, especially older ones, develop more-severe and longer-lasting disease than macaques, which is supported by the higher levels of proinflammatory cytokines in BAL of old baboons relative to young baboons (Supplementary Fig. 4a–d).


Rhesus macaques, baboons and marmosets can all be infected with SARS-CoV-2 but show differential progression to COVID-19. Whereas older marmosets have a mild infection, macaques developed moderate progressive pneumonia that resolves, accompanied by a marked reduction in lung and nasal viral loads. Baboons have the most severe lung pathology, and the greatest viral load in RS. SARS-CoV-2 infection is associated with dynamic lung influxes of specific myeloid cell subsets, particularly IMs, neutrophils and pDCs, and viral proteins can be detected in these cells. This helps to explain the development of COVID-19 pneumonia and also the subsequent control through expression of a strong type-I IFN response. Control of infection is accompanied by resolution of viraemia and radiological lesions but viral antigens and remnant histopathological lesions persist over the course of two weeks. Integration of state-of-the-art CT scanning and innovative algorithms to assess the extent of lung involvement with viral loads in NS and BAL have enabled reproducible and quantifiable metrics of infection. Our experimental models have been useful for preclinical testing of candidate vaccines23 and therapies24 for COVID-19.

We were unable to detect replicative virus in the lungs using either plaque assays or subgenomic PCR after two weeks of infection, although live virus was readily recovered after 3 d.p.i. using both methods. However, various SARS-CoV-2 protein antigens were detected after two weeks in the tissues of NHPs, suggesting antigenic but not viral persistence. This is supported by the finding that the correlation between PD-1 expression on CD8+ T cells with vRNA and a lack of induction of antigen-specific immune effector cytokine production by these cells might also be due to the comparatively shorter duration of this study, as longer studies have clearly established a protective T-cell immunity in SARS-CoV-2 infection25,26. Although there was viral clearance and improvement in radiological scores over the course of the study, the inflammation parameters studied and histology were more protracted resulting in a limited correlation between end-point histological and radiological scores. Furthermore, pathways related to angiogenesis and thrombosis are enriched in the lungs of infected macaques compared with healthy macaques9. Histopathology analysis was performed at 3 d.p.i. and 12/14 d.p.i.; we expect that radiologic lesions at 6 d.p.i. are a combination of the features described above.

As COVID-19 disproportionately affects older humans, we included age as an independent variable in our studies. Age-related effects were more pronounced in baboons than macaques. Baboons developed more-severe inflammatory lesions compared with macaques. Baboons are also a preferred model for cardiovascular and metabolic diseases including diabetes27,28,29; further development of the baboon model may therefore prove especially useful for the study of such comorbidities with COVID-19.

Age-related differences during infection in macaques was striking. Older animals generated substantially reduced amounts of SARS-CoV-2-specific antibodies compared with young macaques. Despite comparable viral replication and immune responses, the effect of age on neutralizing immunity may be a factor in the more pronounced COVID-19 disease in the elderly population. We propose that old macaques could be a useful model for testing therapies and vaccines for elderly humans.


Study approval

All of the infected animals were housed under Animal Biosafety Level 3 or 4 (ABSL3, ABSL4) facilities at the Southwest National Primate Research Center, where they were treated according to the standards recommended by AAALAC International and the NIH Guide for the Care and Use of Laboratory Animals. Sham controls were housed under ABSL2. The animal studies in each of the species were approved by the Animal Care and Use Committee of the Texas Biomedical Research Institute and as an omnibus Biosafety Committee protocol.

Animal studies and clinical evaluations

Sixteen (eight young and eight old; Supplementary Table 1) Indian-origin rhesus macaques (Macaca mulatta) and twelve (six young and six old) African-origin baboons (Papio hamadryas), all from the Southwest National Primate Research Center (SNPRC) breeding colonies, were exposed through multiple routes of inoculation (ocular, 100 μl; intranasal, 200 μl, using a MADgic Intranasal Mucosal Atomization Device (Teleflex); intratracheal, 200 μl, using a paediatric-size laryngo-tracheal Mucosal Atomization Device (Teleflex)) to 500 μl of an undiluted stock of SARS-CoV-2, which had a titre of 2.1 × 106 p.f.u. per ml, resulting in the administration of 1.05 × 106 p.f.u. SARS-CoV-2. SARS-CoV-2 generated from isolate USA-WA1/2020 was used for animal exposures. A fourth cell culture passage of SARS-CoV-2 was obtained from Biodefense and Emerging Infections Research Resources Repository (BEI Resources, NR-52281; GenBank: MN985325.1) and propagated at Texas Biomed. The stock virus was passaged for a fifth time in Vero E6 cells at a multiplicity of infection (m.o.i.) of approximately 0.001. This master stock was used to generate a sixth cell culture passage exposure stock by infecting Vero E6 cells at a m.o.i. of 0.02. The resulting stock had a titre of 2.10 × 106 p.f.u. per ml and was attributed the lot number 20200320. The exposure stock was confirmed to be SARS-CoV-2 using deep sequencing and was identical to the published sequence (GenBank: MN985325) strain USA-WA1/2020 (BEI Resources, NR-52281). Six Brazilian-origin common marmosets (Callithrix jacchus) were also infected through the combined routes (80 µl intranasal; 40 µl ocular (20 µl per eye); 40 µl oral, performed twice for a total of 160 µl intranasal, 80 µl ocular, 80 µl oral; and 100 µl intratracheal once) of the same stock. The total target dose presented to marmosets was 8.82 × 105 p.f.u. per ml. We included sham-infected animals, which underwent all of the procedures (with the exception of necropsy) to control for the impact of multiple procedures over the course of this study (Supplementary Table 1). Four macaques, baboons and marmosets were each sham-infected with DMEM-10 medium (the storage vehicle of the virus) to be used as procedural controls. Infected animals were euthanized for tissue collection at necropsy, and control animals were returned to the colony. Macaques were enrolled from a specific-pathogen-free colony maintained at the SNPRC and were tested to be free from SPF-4 (simian retrovirus D, SIV, STLV-1 and herpes B virus). All of the animals, including the baboons and the marmosets, were also free of Mycobacterium tuberculosis. Animals were monitored regularly by a board-certified veterinary clinician for rectal body temperature, weight and physical examination. Collection of blood, BAL, nasal swab and urine, under tiletamine–zolazepam (Telazol) anaesthesia, was performed as described below, except that BAL was not performed in marmosets. Four macaques were sampled daily until euthanasia at 3 d.p.i. All of the other macaques and all of the baboons were sampled at 0, 3, 6, 9 and 12 d.p.i. and at euthanasia (BAL was performed weekly). Blood was collected for complete blood cell analysis and specialized serum chemistry analysis. Animals were observed daily to record alert clinical measurements. Nasal (longitudinal) or nasopharyngeal (acute) swabs and BALs were obtained to measure viral loads in a longitudinal manner, as described previously14. In brief, in a sitting position, the larynx was visualized and a sterile feeding tube was inserted into the trachea and advanced until it was met with resistance. Up to 80 ml of warm sterile saline was instilled, divided into multiple aliquots. Fluid was aspirated and collected for analysis.


Clinical radiographic evaluation was performed as follows: the lungs of all of the animals were imaged by conventional CXR as previously described30. Three-view thoracic radiographs (ventrodorsal, right and left lateral) were performed at all of the sampling time points. High-resolution computed tomography (CT) was performed daily over 3 d.p.i. in 4 infected macaques and on 6 and 12 d.p.i. in 3 young and 3 old macaques as described in the next section. Images were evaluated by a board-certified veterinary radiologist and scored as normal, mild moderate or severe disease. The changes were characterized as to location (lung lobe) and distribution (perivascular/peribronchial, hilar, peripheral, diffuse, multifocal/patchy).

CT imaging and quantitative analysis of lung pathology

The animals were anaesthetized using Telazol (2–6 mg kg−1) and maintained by inhaled isoflurane delivered through the Hallowell 2002 ventilator anaesthesia system (Hallowell). Animals were intubated to perform end-inspiratory breath-hold using a remote breath-hold switch. Lung field CT images were acquired using the Multiscan LFER150 PET/CT (MEDISO) scanner. Image analysis was performed using 3D ROI tools available in Vivoquant (Invicro). The percentage change in lung hyperdensity was calculated to quantify lung pathology1,2. The lung volume involved in pneumonia was quantified as follows. In brief, lung segmentation was performed using a connected thresholding feature to identify the lung ROI by classifying all of the input voxels of the scan in the range of −850 HU to −500 HU. Smoothing filters were used to reassign every ROI voxel value to the mode of the surrounding region with defined voxel radius and iterations to reconstruct the lung ROI. Thereafter, global thresholding was applied to classify the voxels within the lung ROI in the range of −490 HU to +500 HU to obtain the lung hyperdensity ROI. The resultant ROIs were then rendered in the maximum intensity projection view using the VTK feature.

vRNA determination

vRNA from plasma/sera, BAL, urine, saliva, swabs (nasal/nasopharyngeal, oropharyngeal, rectal) and lung homogenates was determined using quantitative PCR with reverse transcription (RT–qPCR) and vRNA isolation was performed as previously described for Middle East respiratory syndrome (MERS)-CoV and SARS-CoV12,27,28. RNA extraction from fluids was performed using the epMotion M5073c Liquid Handler (Eppendorf) and the NucleoMag Pathogen Kit (Macherey-Nagel). Test samples (100 µl) were mixed with 150 µl of 1× DPBS (Gibco) and 750 µl TRIzol LS. Inactivation controls were prepared with each batch of samples to ensure that no cross contamination occurred during inactivation. Samples were thawed at room temperature and then—for serum, swabs and urine samples—10 µg yeast tRNA was added, along with 1 × 103 p.f.u. of MS2 phage (Escherichia coli bacteriophage MS2, ATCC). DNA LoBind Tubes (Eppendorf) were prepared with 20 µl of NucleoMag B-Beads (NucleoMag Pathogen Kit, Macherey-Nagel) and 975 µl of buffer NPB2 (NucleoMag Pathogen Kit, Macherey-Nagel). After centrifugation, the upper aqueous phase of each sample was transferred to the corresponding new tube containing NucleoMag B-Beads and buffer NPB2. The samples were mixed using a HulaMixer (Thermo Fisher Scientific), rotating for 10 min at room temperature. The samples were then transferred to the sample rack on a epMotion M5073c Liquid Handler (Eppendorf) for further processing according to the NucleoMag Pathogen kit instructions. For vRNA determination from tissues, 100 mg of tissue was homogenized in 1 ml TRIzol Reagent (Invitrogen, Grand Island) with a Qiagen steel bead and Qiagen Stratagene TissueLyser. To detect infectious virus, in brief, tissues were homogenized 10% (w/v) in viral transport medium using Polytron PT2100 tissue grinders (Kinematica). After low-speed centrifugation, the homogenates were frozen at −70 °C until they were inoculated on Vero E6 cell cultures in tenfold serial dilutions. The SARS-CoV-2 RT–qPCR was performed using a CDC-developed 2019-nCoV_N1 assay with the TaqPath 1-Step RT–qPCR Master Mix, CG (Thermo Fisher Scientific). The assays were performed on a QuantStudio 3 instrument (Applied Biosystems) with the following cycling parameters: hold stage, 2 min at 25 °C, 15 min at 50 °C, 2 min at 95 °C; PCR stage, 45 cycles of 3 s at 95 °C and 30 s at 60 °C. The primers and probe used were as follows: 2019-nCoV_N1-F: GACCCCAAAATCAGCGAAAT (500 nM); 2019-nCoV_N1-R: TCTGGTTACTGCCAGTTGAATCTG (500 nM); 2019-nCoV_N1-P FAM/MGB probe: ACCCCGCATTACGTTTGGTGGACC (125 nM).

RNA extraction for subgenomic vRNA determination using RT–qPCR

Samples were inactivated using TRIzol LS Isolation Reagent (Invitrogen) as follows: 250 µl of test sample was mixed with 750 µl TRIzol LS. Inactivation controls were prepared with each batch of samples. Before extraction, 1 × 103 p.f.u. of MS2 phage (E. coli bacteriophage MS2, ATCC) was added to each sample to assess extraction efficiency. RNA extraction was performed using the epMotion M5073c Liquid Handler (Eppendorf) and the NucleoMag Pathogen kit (Macherey-Nagel). Extraction controls were prepared with each batch of samples. After processing, the presence of the eluate was confirmed and the extracted RNA was stored at −80 ± 10 °C.

Determination of viral load using RT–qPCR

RNA samples (5 µl) were reacted using duplex RT–qPCR to detect both SARS-CoV-2 and MS2 phage. Two assays were used to assess the presence of SARS-CoV-2 in the samples. The CDC-developed 2019-nCoV_N1 assay was used to target a region of the N gene. SARS-CoV-2_N1 probe (ACCCCGCATTACGTTTGGTGGACC) was labelled with 6-FAM fluorescent dye. The forward primer sequence was GACCCCAAAATCAGCGAAAT and the reverse primer sequence was TCTGGTTACTGCCAGTTGAATCTG. A secondary qPCR assay to measure subgenomic RNA was also performed to target a region of E (envelope)31,32. The probe was also labelled with 6-FAM fluorescent dye (ACACTAGCCATCCTTACTGCGCTTCG). The forward primer sequence was CGATCTCTTGTAGATCTGTTCTC and the reverse primer sequence was ATATTGCAGCAGTACGCACACA. The MS2 probe was labelled with VIC fluorescent dye. Both assays used the TaqPath 1-Step RT–qPCR Master Mix, CG (Thermo Fisher Scientific) and were performed using a QuantStudio 3 instrument (Applied Biosystems). The QuantStudio design and analysis software (Applied Biosystems) was used to run and analyse the results. Cycling parameters were set as follows: hold stage, 2 min at 25 °C, 15 min at 50 °C, 2 min at 95 °C; PCR stage: 45 cycles (N1 assay) or 40 cycles (E assay) of 3 s at 95 °C and 30 s at 60 °C. The average Ct value for MS2 phage was calculated for all of the processed samples and SARS-CoV-2 quantification was performed only in samples in which the MS2 Ct value was lower than average MS2 + 5%.


Animals were euthanized and complete necropsy was performed. Gross images (lungs, spleen, liver) and organ weights (lymph nodes, tonsils, spleen, lungs, liver, adrenal glands) were obtained at necropsy. Representative samples of lung lymph nodes (inguinal, axillary, mandibular and mediastinal), tonsils, thyroid gland, trachea, heart, spleen, liver, kidneys, adrenal glands, digestive system (stomach, duodenum, jejunum, ileum, colon and rectum), testes or ovary, brain, eyes, nasal tissue and skin were collected for all animals. Tissues were fixed in 10% neutral-buffered formalin, processed to paraffin, sectioned at a thickness of 5 μm, stained with haematoxylin and eosin using standard methods, and evaluated by a board-certified veterinary pathologist.

Tissue processing, flow cytometry, multiplex cytokine analyses, immunohistochemistry, multicolour confocal microscopy and SARS-CoV-2 specific-antibody response detection for immune evaluations

Flow cytometry was performed as previously described14,33 on blood and BAL samples collected on days 3, 6, 9 and 12, and at the end point, which occurred at 14–17 d.p.i. for various animals. A comprehensive list of antibodies used in these experiments is provided in Supplementary Table 8. For evaluations of peripheral blood, peripheral blood mononuclear cells were prepared as previously described. In brief, Cellular phenotypes were studied using the following antibodies: anti-CD3 (SP34–2), anti-CD4 (L200), anti-CD69 (FN50), anti-CD20 (2H7), anti-CD95 (DX2), anti-Ki67 (B56), anti-CCR5 (3A9), anti-CCR7 (3D12), anti-CD28 (CD28.2), anti-CD45 (D058-1283), anti-CXCR3 (1C6/CXCR3), anti-HLA-DR (L243), anti-CCR6 (11A9), anti-LAG-3 (polyclonal, R&D Systems), anti-CD123 (7G3), anti-CD14 (M5E2), anti-CD206 (206), anti-CD16 (3G8), anti-CD163 (GHI/61), anti-CD66abce (TET2, Miltenyi Biotech), anti-CD40 (5C3), anti-IL-2 (MQ1-17H12), anti-GZMB (GB11), which were all purchased from BD Biosciences unless specified. Anti-CD8 (RPA-T8), anti-CD11c (3.9), anti-TNFα (MAb11), anti-IFNγ (B27), anti-IL-17 (BL168) and anti-PD-1 (EH12.2H7) antibodies were purchased from BioLegend. For antigenic stimulation, cells were cultured overnight with SARS-CoV-2-specific peptide pools of the nucleocapsid, membrane and spike proteins (PepTivator SARS-CoV-2 peptide pool, Miltenyi Biotech). A description of the detailed gating strategy for the detection and enumeration of various cellular phenotypes is provided in Supplementary Fig. 6.

Immunohistochemistry was performed on sections (thickness, 4 μm) of lung, nasal cavity and tonsils. The sections were baked at 65 °C for 30 min followed by deparaffinization using xylene and subsequent hydration with decreasing gradations of ethanol as described previously14,34. Heat-induced antigen retrieval was performed using sodium citrate buffer (10 mM, pH 6.0) followed by blocking (3% BSA in TBST for 1 h at 37 °C). For SARS-CoV-2 detection, specimens were incubated with rabbit anti-SARS-CoV-2 spike antibodies (ProSci, 1:200, 37 °C for 2 h) or anti-SARS-CoV-2 nucleocapsid antibodies (Sino Biologicals, 1:100, 2 h at 37 °C). Anti-human-ACE2 antibodies (R&D Systems, 1:50, 2 h at 37 °C) were used to identify ACE2. Mouse anti-human-CD66abce PE-conjugated antibodies (Miltenyi Biotech, 1:20, 2 h at 37 °C) were used to identify neutrophils; mouse anti-CD68 antibodies (Thermo Fisher Scientific, 1:100, 2 h at 37 °C) for macrophages; and pDCs were identified by costaining with PE-conjugated mouse anti-human-CD123 (BD Biosciences, 1:20, 37 °C for 2 h) and mouse anti-human-HLA-DR (Thermo Fisher Scientific, 1:100, 2 h at 37 °C) antibodies. Furthermore, mouse anti-Ki67 antibodies (BD Biosciences, 1:50, 2 h at 37 °C) were used to detect actively proliferating cells. Pan-cytokeratin mouse monoclonal antibodies (AE1/AE3) Alexa Fluor 488 (Thermo Fisher Scientific, 1:50, 2 h at 37 °C) were used to detect type-1 pneumocytes and epithelial cells, and anti-TTF-1 mouse monoclonal antibodies (Thermo Fisher Scientific, 1:200, 2 h at 37 °C) were used to detect type-2 pneumocytes. Chicken anti-rabbit IgG (H+L) Alexa Fluor 488 conjugate; goat anti-mouse IgG (H+L) Alexa Fluor 647 conjugate; donkey anti-mouse IgG (H+L) Alexa-Fluor 555 conjugate; donkey anti-goat IgG (H+L) Alexa Fluor 555; goat anti-mouse IgG1 secondary antibody Alexa Fluor 555 (Thermo Fisher Scientific, 1:400, 1 h at 37 °C) were used to label spike and nucleocapsid, Ki67 and HLA-DR, CD68, ACE2 and TTF-1 primary antibodies, respectively. Tissue sections were then stained with DAPI (Thermo Fisher Scientific, 1:5,000, 5 min at 37 °C) with subsequent mounting using Prolong Diamond Antifade mountant (Thermo Fisher Scientific). A Ziess LSM 800 confocal microscope was used to visualize the stained sections (×10, ×20 and ×63 magnification). RNA was isolated for qPCR and data were analysed as described previously19.

The antibody response against SARS-CoV-2 was measured using an enzyme-linked immunosorbent assay (ELISA) using the SARS-CoV-2 spike S1 subunit protein (Sino Biological) as the capture antigen. Each well of a 96-well microtitre plate (Corning, 2592) was coated with 100 ng of protein in 100 μl carbonate–bicarbonate buffer (pH 9.6) and incubated overnight at 4 °C. The next day, the plate was washed twice with 1× PBS containing 0.05% Tween-20 (PBST) followed by blocking with 5% non-fat dry milk in PBST at room temperature for 2 h. Twofold serially diluted heat-inactivated plasma from infected and control macaques was added to duplicate wells and incubated at 37 °C for 1 h. After five washes with 1× PBST, 100 μl of rabbit anti-monkey IgG peroxidase-conjugated antibodies (1:5,000 dilution in blocking buffer) was added to each well and incubated at 37 °C for 1 h. The plate was washed five times with 1× PBST. SureBlue TMB peroxidase substrate (100 μl) was then added to each well and incubated at room temperature for 10 min. The reaction was stopped by adding 100 μl TMB stop solution and the plate was immediately read at 450 nm. The end-point ELISA titre of binding antibodies was defined as the reciprocal of the serum dilution that resulted in a positive optical density reading, which is at least two times the mean optical density reading with no plasma control wells. The detection limit of the ELISA was considered to be the starting dilution (1:100) of the test sera.

Statistical analyses

Graphs were prepared and statistical comparisons were applied using GraphPad Prism v.8. Various statistical comparisons were performed as follows. Two-tailed Student’s t-tests, ordinary ANOVA or one-way or two-way repeated measure ANOVA with Geisser–Greenhouse correction for sphericity and Tukey post hoc correction for multiple testing (GraphPad Prism 8) was applied where applicable and as described in the figure legends. For correlation analysis, Spearman’s rank tests were applied. Statistical differences between groups were reported to be significant when the P value was less than or equal to 0.05. Data are presented as mean ± s.e.m.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.