Characterization of Three Variants of SARS-CoV-2 In Vivo Shows Host-Dependent Pathogenicity in Hamsters, While Not in K18-hACE2 Mice

Animal models are used in preclinical trials to test vaccines, antivirals, monoclonal antibodies, and immunomodulatory drug therapies against SARS-CoV-2. However, these drugs often do not produce equivalent results in human clinical trials. Here, we show how different animal models infected with some of the most clinically relevant SARS-CoV-2 variants, WA1/2020, B.1.617.2/Delta, B.1.1.529/Omicron, and BA5.2/Omicron, have independent outcomes. We show that in K18-hACE2 mice, B.1.617.2 is more pathogenic, followed by WA1, while B.1.1.529 showed an absence of clinical signs. Only B.1.1.529 was able to infect C57BL/6J mice, which lack the human ACE2 receptor. B.1.1.529-infected C57BL/6J mice had different T cell profiles compared to infected K18-hACE2 mice, while viral shedding profiles and viral titers in lungs were similar between the K18-hACE2 and the C57BL/6J mice. These data suggest B.1.1.529 virus adaptation to a new host and shows that asymptomatic carriers can accumulate and shed virus. Next, we show how B.1.617.2, WA1 and BA5.2/Omicron have similar viral replication kinetics, pathogenicity, and viral shedding profiles in hamsters, demonstrating that the increased pathogenicity of B.1.617.2 observed in mice is host-dependent. Overall, these findings suggest that small animal models are useful to parallel human clinical data, but the experimental design places an important role in interpreting the data. Importance: There is a need to investigate SARS-CoV-2 variant phenotypes in different animal models due to the lack of reproducible outcomes when translating experiments to the human population. Our findings highlight the correlation of clinically relevant SARS-CoV-2 variants in animal models with human infections. Experimental design and understanding of correct animal models are essential to interpreting data to develop antivirals, vaccines, and other therapeutic compounds against COVID-19.


Introduction
Worldwide transmission of the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in the human population has contributed to the persistence of the COVID-19 pandemic [1]. The emergence of new variants refers to viral genome adaptation in which mutations in the spike gene RBD (receptor binding domain) drastically increase binding affinity in the RBD-human angiotensin-converting enzyme 2 (hACE2) complex, while also transmitting rapidly in human populations [2,3]. The emergence of new variants in dominantly immunologically naïve populations suggests that adaptive mutations in the viral genome continue to improve viral fitness in unvaccinated and vaccinated individuals.
Further, CD4+ T cells are critical to aid in the development of antigen-specific memory B cells [1,29]. Thus, understanding how different variants stimulate immunological memory is critical. Of interest, the Omicron variant is known to be able to infect hosts regardless of the ACE2 status [19]. In this report, we investigate the immunological response to Omicron infection in two mouse strains: one which includes the human ACE-2 receptor (K18-hACE2), and the strain of origin (C57BL/6J) which lacks this receptor.
There is now growing evidence that Omicron causes a less severe pathology than the ancestral strain and other VOCs [13,30]. It has been shown that alpha, beta, gamma, and Delta VOCs replicate efficiently in the lungs of Syrian hamsters and to a similar level as the ancestral strain [20,22,23]. Here, we show the infectivity of the Omicron and Delta variants versus the WA1/2020 ancestral strain D614 in mouse and hamster models. We sought to investigate the infective dose (ID 50 ) and/or lethal dose (LD 50 ) of the three most clinically relevant variants of SARS-CoV-2: B.1.617.2/Delta, the first and most prevalent in the human population in 2021 (known for being highly pathogenic in humans); B.1.1.529/Omicron (BA.1), the variant that displaced Delta and became the most prevalent in 2022 (less pathogenic but more transmissible [19]); and Ancestral WA1/2020, no longer in the population but is the strain that currently approved vaccines are based on, as well as used to test therapeutics against COVID-19. We conducted mouse infection experiments using the ancestral WA1/2020 isolate (containing ancestral D614), B.1.617.2 (also known as 21A Delta or 21A/S:478K with the D614G substitution), and B.1.1.529 (also known as 21K Omicron with D614G and 35 more mutations). The hACE2 transgenic mice (K18-hACE2) ( Figure 1A) were used in three independent experiments with several serial dilutions of our viral stocks. Here, we summarize the LD 50 for WA1/2020 compared to the lower dose of B.1.617.2/Delta and the higher dose of B.1.1.529/Omicron. Mortality started at days 8-9 (8-9 days postinfection (d.p.i.)), where 50% of the animals infected with 1 × 10 3 TCID 50 WA1/animal succumbed to infection. Animals infected with a dose of 1 × 10 3 TCID 50 Delta/animal presented 100% mortality by day 8 (Supplemental Figure S1). Interestingly, the lower dose of 1 × 10 1 TCID 50 Delta/animal presented 100% mortality by day 11. Conversely, the higher dose (our straight viral stock) of 9 × 10 6 TCID 50 Omicron/animal presented no SARS-CoV-2 disease-associated mortality, as shown by the Kaplan-Meier survival analysis ( Figure 1B). The body weight loss of the 1 × 10 1 TCID 50 /Delta-infected mice was significantly greater (p = 0.0371) than that of the 9 × 10 6 TCID 50 /Omicron-infected mice and similar to the 1 × 10 3 TCID 50 /WA1 across 14 dpi, on average ( Figure 1C). Viral burden in lungs was measured by a TCID 50 infectivity assay at 3 dpi with 1 × 10 3 TCID 50 /WA1 (n = 7), 1 × 10 1 TCID 50 /B.1.617.2/Delta (n = 7), and 9 × 10 6 TCID 50 / Omicron (n = 4) on randomly selected animals after interim sacrifice. The variant with the highest viral burden was WA1, with an average of 7 × 10 6 TCID 50 /mg, followed by Delta with 2.5 × 10 5 TCID 50 /mg, and Omicron with 7.9 × 10 3 TCID 50 /mg. However, due to high variability and small sample size, no statistically significant differences were found regarding viral burden after the ANOVA analysis ( Figure 1D). Therefore, we were unable correlate higher titers and pathogenicity. Furthermore, the animals infected with 1 × 10 1 TCID 50 of B.1.617.2/Delta that met the predetermined criteria for moribundity on day 6 were euthanized to avoid unnecessary pain and distress. Lungs were collected from these moribund animals for histopathological evaluation. Moribund lungs were observed to have multifocal necrosis, neutrophilic infiltrates, hemorrhage, edema, loss of pneumocytes, epithelial hyperplasia, infiltration with neutrophils, pulmonary intraalveolar macrophages (PIMs), and red blood cells when compared to the lungs of the mock-infected animals ( Figure 1E).

Infection in Hamsters
In contrast to other rodents, the Syrian hamsters are naturally susceptible to infection with SARS-CoV-2, allowing for them to be rapidly developed into a common model. The hamster SARS-CoV-2 model recapitulates many characteristic features seen in human patients with a moderate, self-limiting course of the disease, such as specific patterns of respiratory tract inflammation, while not producing the outcome of death [12,20,22]. We aimed to elucidate whether the increased pathogenicity of B.1.617.2 compared to WA1 observed in mouse experiments translates to the nonlethal hamster model. To address this, we intranasally infected hamsters with Delta and WA1 at the mouse-lethal dose of 5 × 10 3 TCID 50 and the higher dose of our viral stocks for Omicrons BA.1 and BA.5 of 6.5 × 10 5 TCID 50 /animal. We found that the same dose of 5 × 10 3 TCID 50 for Delta and WA1 ( Figure 2A) produced similar outcomes. The body weight loss was similar for both variants in the hamster model. Animals had a weight loss of 4% on average at 4 dpi ( Figure 2B). Additionally, viral replication kinetics in lungs (estimated by RT-qPCR on lung tissues from 3 and 6 dpi for WA1 and 4 dpi for Delta) reflect similar copy numbers in both strains ( Figure 2C). In hamsters, we were also able to measure the viral shedding profiles (through estimation of viral copy number by RT-qPCR) in oropharyngeal swabs at different time points after infection. Shedding profiles from hamsters infected with 5 × 10 3 TCID 50 of Delta, WA1, or 6.5 × 10 5 TCID 50 /animal of BA.1 or BA.5 show a peak viral load 2 dpi and a consistent decrease in viral load until 4-5 dpi; no significant differences regarding variants were observed. These data suggest that the observed increase in pathogenicity in mice by Delta is relative to the host and not directly related to the viral variant ( Figure 2D). In contrast to mice, hamsters appeared to suffer an acute peak of infection before they recover. Virus shedding was consistent between strains on day 2, the highest viral burden in lungs was measured on day 3, and the highest body weight loss was on day 4, regardless of viral strain ( Figure 2).

Reduced Pathogenicity of B.1.1.529/Omicron BA1-Infected Mice
We next assessed the infectious dose (ID 50 ) for B.1.1.529/Omicron in C57BL/6J mice to investigate the strain's reported adaptation (due to multiple genetic changes) to no longer require human ACE2 for infectivity [13]. We conducted mouse infection experiments in K18-hACE2 transgenic mice and the nontransgenic strain of origin, C57BL/6J ( Figure 3A). Serial dilutions of B.1.1529/Omicron viral stock, from 9 × 10 6 , 1 × 10 5 , 1 × 10 4 , to 1 × 10 3 TCID 50 /animal in K18-hACE2 mouse, and 9 × 10 6 , 1 × 10 5 , to 1 × 10 4 TCID 50 /animal in C57BL/6J, were utilized. The K18-hACE2 transgenic mice inoculated with the higher dose of 9 × 10 6 TCID 50 was the only group that presented a slight reduction in body weight, with a peak of −1.4% body weight loss at day 7, but on day 8 the animals recovered, and further weight was gained by day 12 ( Figure 3B). However, none of the other dosing groups (either K18-hACE2 or C57BL/6J) presented a consistent body weight loss ( Figure 3C) or had any other clinical signs. We next compared the viral burden in the lungs by RT-qPCR ( Figure 3D) and TCID 50 . The K18-hACE2 mice sustained moderate levels of infection, as shown for the average of 1 × 10 4 TCID 50 /mg in the lungs ( Figure 3E). The levels of viral RNA were comparable to the TCID 50 assay ( Figure 3E,D). Similar results were observed for the C57BL/6J mice infected with 9 × 10 6 TCID 50 . This group also presented similar levels of viral burden in lungs to K18-hACE2 mice. In this model, we aimed to have an additional readout due to a lack of clinical signs, body weight change, or mortality. For ethical reasons, we were not able to perform nasal washes in mice to obtain viral shedding profiles. However, we performed oropharyngeal swabs in alert animals from days 2 to 4 postinfection. The shedding profiles obtained from the oral swabs show that viral RNA had similar levels (on average) in the K18-hACE2 and C57BL/6J mice infected with 9 × 10 6 TCID 50 across time points (p > 0.005, multiple regression) ( Figure 3F).

Spike-Specific T Cell Phenotypes in Omicron-Infected K18-hACE2 and C57BL/6J Mice
Finally, we assessed the inflammatory responses in the spleens of K18-hACE2 mice at 21 dpi. SARS-CoV-2 spike-specific T cell responses were first measured using flow cytometry activation-induced marker (AIM) assay ( Figure 4A and Figure S2) by stimulating cells with an overlapping peptide pool of Omicron's spike domain [31][32][33]. When comparing AIM+ phenotypes, there was no difference in the percentage of CD8+ AIM+ cells, regardless of the mouse model or Omicron challenge status. In contrast, in K18-hACE2 mice, the baseline of uninfected CD4+ AIM+ T cells was significantly lower than in infected K18-hACE2 mice (inner left, Figure 4A). This was not observed in C57BL/6J CD4+ T cells. However, when unstimulated controls were subtracted from the Omicron-spike-peptidetreated samples, this effect was abolished ( Figure 4A). This phenomenon displays that there was not a marked observed difference between antigen-stimulated and unstimulated wells, regardless of challenge. It may also demonstrate that baseline antigen specificity was altered by the presence of the challenge in K18-hACE2 mice that was not present in the C57BL/6J mice.

Discussion
In COVID-19 patients with acute respiratory illness, the main clinical manifestation is pneumonia [4]. Consistent with previous studies, our data suggest that WA1/2020, B.1.617.2/Delta, and B.1.1.529/Omicron SARS-CoV-2 variants replicate efficiently in the lungs of the K18-hACE2 heterozygous mouse and Syrian hamster models [4]. Our results indicate that, in mice, Delta is most pathogenic, followed by WA1, and Omicron was absent of clinical signs. This is similar to what has been described for human infections population-wise [4]. Interestingly, Omicron was able to infect C57BL/6J mice that do not have the human ACE2 receptor. Viral shedding profiles and viral titers in the lungs were similar between the K18-hACE2 and the C57BL/6J. These data indicate viral adaptation to a new host and a confirmation of asymptomatic SARS-CoV-2 carriers that accumulate and shed the virus even in the absence of clinical signs.
Next, we show how the same dose of Delta and WA1 have similar viral replication kinetics, pathogenicity, and viral shedding profiles in hamsters. This demonstrates that the increased pathogenicity observed in mice is host-dependent and not associated with viral replication.
In contrast to similar viral shedding profiles and titers between K18-hACE2 and C57BL/6J mice, our data suggest that K18-hACE2 mice, but not C57BL/6J mice, have a change in the CD4 memory T cell repertoire towards effector memory cells and an increase in CD8 T cell cytokine production after Omicron infection. Since K18-hACE2 mice demonstrated changes in the T cell immune response, while C57BL/6J mice did not, this suggests that the hACE2 receptor is necessary for generating immunological memory and T cell cytokine production against SARS-CoV-2.
While we did not observe significant changes in the CD8 T cell memory markers, only cytokine production, this may be due to the timing of splenocyte collection. Typically, CD8 memory T cells spike 8 dpi after the initial encounter with a pathogen and reduce to 5-10% of that initial population over the following weeks [34]. Due to this natural reduction over time, 21 dpi may have been enough time for these memory T cells to return near to baseline levels. Of note, we observed a similar trend to the CD4 memory cells (an increase in Tem) and a slight decrease in Temra, but these changes were not significant. We do observe antigen-specific cytokine responses, which does suggest that memory CD8 T cells were generated from the initial challenge and restimulated with the Omicron spike peptides. Further studies should be performed using peptides from other regions of Omicron (such as the nucleocapsid) and at multiple time points to determine the full T cell response against the virus in K18-hACE2 mice, as it has been reported that memory CD4 T cells can bind the M, spike, N, nsp3, nsp4, ORF3a, and ORF8 regions (among others), while CD8 can also bind the M region and multiple ORFs [35]. Overall, these findings suggest that small animal models are useful to parallel human clinical data, but the experimental design places an important role in interpreting the data.

Materials and Methods
No statistical methods were used to predetermine the sample size. Investigators were not blinded to allocation during experiments and outcome assessment. Randomized animals were assigned to groups based on body weights that will produce similar group mean values by using Ascentos ® version 2.0, (PDS Pathology Data Systems, Inc., Basel, Switzerland). No animal body weight varied from the mean of the group body weight by more than 20%.

Ethics Statement
Animal studies were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals (National Research Council, 2011) and the Public Health Service Policy on Humane Care and Use of Laboratory Animals (NIH, 2002). The protocols were approved by the Institutional Animal Care and Use Committee at IITRI (Assurance Number D16-00299 A3475-01). Virus inoculations were performed under induced anesthesia and maintained with isoflurane or ketamine hydrochloride and xylazine, and all efforts were made to minimize animal suffering. SARS-CoV-2 in vivo and in vitro studies were performed in the animal biosafety level 3 (ABSL3) facility at IITRI. All experiments complied with all relevant ethical regulations.

Hamster Experiments
Eight-to twelve-week-old male Syrian golden hamsters of about 90-105 g were obtained from Envigo. Animals were single-housed and fed standard #2014C Teklad Global 14% Protein Rodent Diet. Under ketamine/xylazine (IP) anesthesia, hamsters were intranasally challenged with a volume of 50 µL containing 5 × 10 3 TCID 50 /animal of WA1/2020 or 5 × 10 3 TCID 50 /animal of B.1.617.2, or 6.5 × 10 5 TCID 50 /animal of B.1.1.529 (BA.1) or 6.5 × 10 5 TCID 50 /animal of BA5.2 SARS-CoV-2 variants. Prior to challenge, hamsters were monitored for 3 days to measure body weight (average of 2 closest days to challenge were used for establish baseline). Body weights were monitored every day for 4 dpi. Oral swabs were collected on 1 to 4 dpi for analyzing viral shedding profiles by RT-qPCR. On 4 dpi, all animals were euthanized. Immediately, the left lung was inflated for histopathological analysis, and the right lung was processed and flash-frozen for RT-qPCR and TCID 50 assay.

Histopathological Analysis
Fixed tissues were processed through paraffin blocks. Each section from the right cranial, right accessory, right caudal, and right middle lobes were trimmed, embedded in paraffin, sectioned at approximately 5 µm, and stained with hematoxylin and eosin (H&E). All paraffin H&E slides were evaluated microscopically and graded for presence and severity of pathology by a board-certified (ACVP) veterinary pathologist by PAI-Charles River Laboratories, Inc.

Oral Swabs
On alert animals, an oropharyngeal (OP) sample was taken using Puritan™ PurFlock™ Ultra Flocked Swabs (Thermo Fisher Scientific cat. No. 22-025-192). Swabs were vortexed for 10 s in 300 µL of 1× DNA/RNA Shield TM (Zymo Research, cat. No. R1100) and centrifuged for 5 min at 5000× g. Samples were extracted using Quick-RNA Viral Kit (Zymo Research cat. No. R2141) and the viral genome copy numbers were evaluated by RT-qPCR analysis.

Nasal Washes
Hamsters were anesthetized with ketamine (100 mg/kg) and xylazine (5 mg/kg) mixture (KX) via intraperitoneal (IP) injection. Antisedan ® , an α-2-antagonist, was used to reverse of KX after the procedure. Once the hamster was anesthetized, 0.16 mL of sterile 1X PBS containing penicillin (100 U/mL), streptomycin (100 µg/mL), and gentamicin (50 µg/mL) was injected into the nostrils 20 µL at a time and pipetted out into a collection tube. The recovered nasal wash volume was collected and recorded. Samples were stored at ≤−65 • C until analysis.

TCID 50 Assay
Lung tissue samples had viral titers determined using the TCID 50 assay. Briefly, clarified tissue supernatants were diluted 1:10, 1:100, or 1:1000 followed by 2-fold serial dilutions and added to a 96-well plate preseeded with VeroE6 cells. Each sample was plated in triplicate. The plates were read after 72 h for cytopathic effect (CPE) and immunostained.

One-Step RT-qPCR Assay
Following left lung tissue homogenization and brief centrifugation, a sample of 100 µL of the supernatant of the homogenate was mixed with 100 µL of the 2× DNA/RNA Shield TM (1:1 v/v). The concentration of viral genome copy number in harvested tissue supernatants was determined by RT-qPCR assay. Briefly, RNA was extracted from samples stored in RNA/DNA Shield using the Quick-RNA Viral Kit (Zymo Research) according to the manufacturer's protocol. The following One-step RT-PCR cycling conditions were used: 50 • C for 15 min (RT), then 95 • C for 2 min (denature), then 40 cycles of 10 s at 95 • C, 45 s at 62 • C.
Primers/probe used for SARS-CoV-2 detection are given below: 2019-nCoV_N1-F: 5 -GACCCCAAAATCAGCGAAAT-3 for two populations were assessed via Student's t-test and statistics for greater than two populations were assessed via one-way ANOVA on GraphPad Prism 9.