A systematic review of Cynomolgus macaques, Rhesus macaques, and ferrets as large animal models for COVID-19

Author: Alexandra Witt, Queens University


Layperson Summary

COVID-19 is a disease that has quickly caused a global pandemic, and millions of scientists are currently working to create vaccines to fight it. In fact, they already have a few on the market, as you probably know! The process to create a vaccine is a long one. Before they test anything in human patients, scientists have to test their work on animals; you’ve probably heard about mice and rats being used a lot. Those small animals are pretty different from humans though, and so researchers often test their vaccines in larger animals as an intermediate step before humans. This can include pigs, dogs, monkeys, and ferrets. But can those animals actually be infected with the virus that causes COVID-19? That’s an important question.

To see if there is an ideal large animal model for COVID-19, one that presents with the same symptoms as humans, we looked into all of the studies performed prior to May of 2020, working with Cynomolgus macaques (a breed of monkey, also known as a non-human primate), Rhesus macaques (another non-human primate), and ferrets. It turns out that none of the three animals are similar to humans when they are infected with COVID-19. What that means is that scientists have to be really careful when they start testing anything on humans, because the animals and the humans might react really differently.


Abstract 

With animal models at the forefront of biomedical research for transmission studies, vaccine testing, and pathogenesis discovery, there remains a need to identify a large animal model for COVID-19. Without clear diagnostic criteria for the disease, six human studies were first used to identify the symptomatic criteria required in the animal model. Following this, sixteen studies were then used to examine the symptom prevalence in Cynomolgus macaques, Rhesus macaques, and ferrets. While some of the animals modeled independent symptoms sufficiently, the differences in overall presentation of disease were too different (p>0.001) from that of humans to successfully identify one large animal as a leading model for COVID-19. 

 Introduction 

Human research of disease presents a number of ethical dilemmas. This prompted scientists to use animal models in their research, the primary goal of which is to enhance the understanding of a human or animal disease or phenomenon. Animal models have been instrumental to our understanding of pathologies, the assessment of novel vaccines, and the testing of acute therapies. Of the past 216 Nobel prizes awarded in the physiology and medicine categories since 1901, all but 36 have been a direct result of animal-based research. 

Insects, nematodes, fish, amphibians, and numerous mammals have enabled some of the most important advances in physiology and medicine since their introduction in disease research. Through genetic modification, surgical adaptation, xenograft, chemical induction, and infection models, these animals are able to model human phenomena. While particular animal species are often chosen based on their ability to meet specific criteria in line with the research question, their size remains a key determining factor. 

Small animals are often preferred in laboratory settings for their ease of use, including their small size, shorter life cycle, easier handling and care, and quick gestation. Rodent models remain the most commonly used animal for the study of human diseases for these very reasons, although they frequently fail to fully mimic the clinical signs and significant pathologic hallmarks of human diseases (Egan, Vesterinen, Beglopoulos, Sena, & Macleod, 2016). For this reason, many researchers have turned to large animal models. Non-human primates in particular have been extremely useful in reproducing the clinical signs of human diseases, as a result of their close phylogenetic relationship to humans and resulting genetic, behavioural, and biochemical similarities (Gerdts et al., 2015). 

On March 11, 2020, the World Health Organization declared SARS-CoV-2 a pandemic. SARS-CoV-2 is a novel coronavirus causing symptoms similar to, but distinct from, those found in those infected with SARS, the 2003 pandemic. To date, this coronavirus has infected millions of individuals with the COVID-19 disease. As scientists race to identify the pathogenesis of the disease and develop a vaccine to fight it, finding the ideal animal that replicates the human disease is essential. 

Much of the research on COVID-19 thus far has been aided by previous SARS research. In both SARS and SARS-CoV-2 studies, mice (Roberts et al., 2007; Sun et al., 2020) and hamsters (Imai et al., 2020; Roberts et al., 2005) were used as the small animal models of choice. As large animals, numerous ferrets, cats, pigs, chickens, dogs, and non-human primates have been tested for their reproducibility of the coronaviruses, with varying degrees of commendation (McAuliffe et al., 2004; Shi et al., 2020; Weingart et al., 2004). While there is likely no perfect animal model of this viral infection, there remains a need to identify at least one of these large animals as a frontrunner in reproducibility of the clinical signs and significant pathologies of SARS-CoV-2 infection. 

For this reason, the present article focuses on summarizing the results of multiple studies on ferret, rhesus macaque, and cynomolgus macaque infection with SARS-CoV-2. With the aim of highlighting the animals which best replicate the human clinical and laboratory symptoms of COVID-19, the results of six animal studies have been synthesized to provide, to our knowledge, the most comprehensive analysis of the models to date. 

Materials and Methods 

Search strategy for human symptoms 

Without a standard for the diagnosis of COVID-19 at the time of this paper’s writing, there was a requirement to identify the major symptoms of the disease in order to identify an animal model that accurately replicates COVID-19. A literature search was conducted in May 2020 using Google Scholar with the aim of identifying observational studies commenting on the diagnosis of COVID-19 since 2019. The database was searched using the following MeSH terms: SARS-CoV-2, COVID-19, coronavirus, symptoms, diagnosis, pathology. To identify key indicators of human disease, six of the most relevant and frequently cited studies for COVID-19 were chosen to generate a baseline of the clinical symptoms of the human disease. 

Search strategy for animal models 

A second literature search was conducted in May 2020 using Google Scholar with the aim of identifying animal experiments modelling COVID-19 since 2019. The databases were searched using the following MeSH terms: SARS-CoV-2, COVD-19, coronavirus, animal model, monkey, rhesus, cynomolgus, macaque, ferret, pathogenesis, symptom, and symptoms. During the literature search, studies were excluded if the article was a duplicate of another article and used the same cohorts, if the animals in question were geriatric, or if the article could not be accessed. 

The literature search yielded 9 of papers, with 23 individual studies between them. In determining a final sample for analysis, studies were filtered on the basis of three criteria: (1) the age and bodyweight of the animals used, (2) the inoculation route used, and finally (3) the intervention method. Data from studies were excluded if the animals used were juvenile, or had greater than a 5% variation in normal bodyweight for their age. Growth charts (Ribeiro Andrade et al., 2004; Triple F Research, 2015) were used to extrapolate the age of an animal when only weight was provided. Where gender was listed, the appropriate growth curve was used. For those articles wherein gender was not provided, an averaged value was used. Articles that provided neither age nor weight were excluded. 

The remaining studies (n = 21) were then filtered for inclusion based on inoculation route. Experiments were only included if animals had been inoculated via at least one mucosal route of inoculation (Lawler et al., 2006), including intratracheal, intranasal, conjunctival, and oral inoculation routes (any viral strain was accepted). Studies where animals were inoculated intravenously were excluded from the data, although if intravenous inoculation was used in combination with one of the aforementioned mucosal routes, the data was included. 

Finally, studies were filtered based on SARS-CoV-2 intervention. In the examined papers, animals were exposed to the virus in one of three ways: inoculation with SARS-CoV-2, direct contact with inoculated cage-mates, or indirect contact with inoculated animals in adjacent cages. Within a paper, individual studies were excluded for animals whose interventions resulted from direct or indirect contact, as the exact titre of the virus to which they were exposed could not be calculated. The final dataset included 16 individual studies contained in 6 papers. 

Data extraction from human studies 

From each paper, symptoms cited as being prevalent in at least 25% of patients from any given study were included. The investigated symptoms were fever, non-productive cough, myalgia, malaise/lethargy, chills, headache, diarrhea, dyspnoea, radiographic findings of pneumonia or opacification, lymphopenia, thrombocytopenia, increased D-dimer, CRP, and LDH levels, and decreased albumin. 

Data were extracted for the number of patients with each clinical symptom (including radiographic findings). In addition, the average of all applicable serum levels was documented. 

When progression of symptoms was reported (Day 1-7 or “on admission” versus “later”), the initial symptoms were used to be consistent with publications that only referenced symptoms on admission. Where Day 1 was not the first recorded day of infection, data from the earliest reported day (Day 2 or 3) were used. 

For publications that listed laboratory findings and symptoms of ICU and non-ICU patients (or similarly juxtaposed conditions), total values were used to extract the data from the inclusive population. 

Data extraction from animal studies 

Based on the results from human studies, animal model experiments were evaluated for the clinically relevant signs of either disease (those present in more than 25% of humans). The investigated symptoms were fever, non-productive cough, malaise/lethargy, radiographic findings of pneumonia or opacification, lymphopenia, increased CRP and LDH levels, and decreased albumin. 

The data from the final 16 studies were stratified by animal model, and data within each animal stratification was inclusive of the different/combination mucosal routes and viral titres, which were all in the 105-108 TCID50/PFU range for macaques and 103-105.5 TCID50/PFU range for ferrets. Although the original intent was to conduct a meta-analysis, based on communication with experts in the field, data manipulation was not recommended. For this reason, there was no attempt to convert and combine the data of the studies for a meta-analysis. 

Data from studies of COVID-19 in humans described initial symptoms present on admission, which has been shown to average around Day 7 with the virus (WHO, 2020). To account for the faster viral lifecycle seen in the animals, data was extracted from animal studies on the third day after infection. Where Day 3 data was not available, Day 2 data was used (Yu et al., 2020). Radiographic data was taken from any scan between Days 2 and 14. 

Data that was presented only in graphical form was not used for interpolation (Yu et al., 2020). 

Calculations 

Where the median and interquartile range of a serum analysis were provided in the place of a mean, the mean was estimated using the following formula10, where 𝑥̅ = estimated mean, q1 = lower quartile, m = median, and q3 = upper quartile: 

To determine the overall average of a symptom, serum level averages of a study were multiplied by patient/animal number to adjust power, before all COVID-19 studies were averaged to determine the clinical relevance of each symptom. The same was done with the percent prevalence of the clinical symptoms. Where 𝑋̅ = overall average, 𝑥̅ = study average, and n = the number of papers that data was extracted from, the calculation of each indicator was as follows: 

Assumptions 

A number of values found in these studies were clinically unreliable, as they were thousands of times larger or smaller than normal values, or far out of the reference ranges. In order to maintain as large a data set as possible, every attempt was made to contact authors to determine whether these values were valid, or reported in error. When it was clear that the wrong unit had been reported, the data was adjusted. When D-dimer values were reported in DDU, they were multiplied by 2 to obtain values in FEU that could be compared to the values from other papers. When validity of the data could not be ascertained, data points were excluded from the analysis. 

Statistical analysis 

With the objective of accepting a null hypothesis of similarity between human and animal symptom representation, smaller mean differences and larger significance values (p = 1.000) were sought. 

For each animal, symptoms were recorded as “Present,” “Not Present,” or “Not Tested.” Present and Not Present responses were then converted to the numbers 0 and 1, respectively; answers recorded as N/A were left blank. 

Using SPSS, a univariate ANOVA was performed on each symptom individually to compare each animal to the human baseline, with a Bonferroni correction made to produce a more conservative threshold for statistical significance. 

Another ANOVA and subsequent Bonferroni post-hoc test was run with all symptoms combined to generate an overall picture of likeness in disease presentation. 

Results 

An initial analysis of 6 human trials provided the basis for which symptoms would be investigated in animal models. More than 25% of all patients (n = 1,527) were found to present with the following observable symptoms: pneumonia, fever, cough, and lethargy (Table 1). In addition to the observable symptoms, patients’ blood showed signs of lymphocytopenia, increased CRP and LDH levels, and decreased albumin levels (Table 2). 

A similar analysis of the animal trials was also conducted. These values were then compared to those found in the human trials (see Table 1). Many of the symptoms seen in human patients were not visible or tested in the animals, and, with one exception (Munster et al., 2020), studies did not report blood work for the animals. In Munster and colleagues’ (2020) study, the lymphocyte and albumin levels of rhesus macaques were measured, and the authors found a decrease in the levels of both from the lower limit, similar to human presentation of the disease (Table 2). 

Where clinical symptoms were concerned, human percentages were compared to those of the animals using a Bonferroni correction (Table 3). With the objective of accepting a null hypothesis of similarity between human and animal symptom representation, smaller mean differences and larger significance values were sought. Cynomolgus macaques were significantly similar (p = 1.000) to humans in their prevalence of pneumonia and dyspnoea, but differed significantly (p < 0.05) where fever and cough were concerned (Table 3). Rhesus macaques showed no similarities to humans, with p values less than 1.000 for all tested symptoms (Table 3). Ferrets showed a similar prevalence of lethargy (p = 1.000) but did not mimic the other clinical symptoms (Table 3). 

As a final step, the clinical symptoms were combined, for each animal separately, and compared to the clinical symptoms of humans. Results indicated that each animal presented differently than humans (Table 4). Specifically, cynomolgus macaques (p = 0.000, 95% CI [-0.73, -0.17]), rhesus macaques (p = 0.000, 95% CI [-0.41, -0.13]), and ferrets (p = 0.000, 95% CI [-0.45, -0.17]) showed significant differences in their presentation of COVID-19 when compared to human trials. 

Conclusion 

In this systematic review, six studies of SARS-CoV-2 animal models were compared to the human presentation of the disease. In both clinical (Table 1) and laboratory (Table 2) symptomatic comparisons, cynomolgus macaques, rhesus macaques, and ferrets failed to fully mimic the severity of the presentation of COVID-19 in humans. 

Where clinical symptoms were concerned, cynomolgus macaques seem a frontrunner with significantly similar (p = 1.000) presentations of pneumonia and dyspnea (Table 3). Unfortunately, it was rhesus and not cynomolgus macaques that were tested for laboratory symptoms. While they had similarly decreased lymphocyte and albumin levels when compared to humans (Table 1), they were untested for CRP or LDH, two other indicators of the human presentation. Ferrets had only a significantly comparable prevalence of lethargy, but did not mimic any of the other human clinical symptoms (Table 3). 

The three large animal models were also compared in an multivariate ANOVA with Bonferroni-corrected (Table 4). In all instances, animals presented disease in significantly different manners (p = 0.000 for all three comparisons). At this time, no animal presents as an ideal model of COVID-19. 

These results highlight one of the greatest challenges in animal-based research—the differences between species. In these studies, and others before it concerning SARS (Nagata, Iwata-Yoshikawa, & Taguchi, 2010), it has been demonstrated that the virus has a faster viral life cycle in animals. Their uniquely subdued presentation of the disease points to a different genetic susceptibility and reaction to the virus. 

One limitation of the present study is that the sample size of the current review is quite small. It is possible that with a greater number of studies, we would find an animal that is an ideal model of COVID-19. However, numerous COVID-19 studies were excluded from this review because viral titres were too small, the age of the animal was unknown, the inoculation route was intravenous or otherwise non-mucosal, or data was available only in graphical form. If had there been an international guide on experimental methods and a requirement for consistency in reporting, these studies would have been included in the present review and might have altered the results. As is evident in Tables 1 and 2, there are multiple instances in which a symptom was tested by only one study (or in none). The power of this analysis is weakened by these gaps in knowledge, and alone would prevent us from highlighting any one animal as the ideal model for COVID-19. 

Standardization in reporting has been an issue for some time now, without great change. The ARRIVE (Animals in Research: Reporting In Vivo Experiments) guidelines were published in 

2010 with multiple goals, one of which was to maximize the utility of the information gained from every animal experiment (Kilkenny, Browne, Cuthill, Emerson, & Altman, 2010). Two years later, it was determined that endorsement of the guidelines was insufficient to promote their adherence (Baker, Lidster, Sottomayor, & Amor, 2014). This same predicament can be seen in the papers excluded from the present review, which omit key parameters of methodology and prevent reproducibility of experiments as well as the use of their data in any matter of systematic review or meta-analysis. 

A meta-analysis of animal model experimentation was the original intent of this study, and would, in fact, be the best use of resources through which to determine an ideal animal model for COVID-19. Unfortunately, lack of standardization in reporting standards has proven a meta-analysis of viral experiments to be impossible, with variations in assays, units, and vague statistical comparisons highlighting some of the inherent challenges faced. Even between clinical symptoms, there is a potential for misleading data interpretation where the standards are unclear. D-dimer, for example, is a prominent prognostic indicator in many diseases, but could not be singled out regarding COVID-19 due to a lack of standardization in its reporting (Thachil et al., 2020). Where the magnitudes of units are easily converted, even they should be standardized in such instances where data needs to be shared. 

The aim of this study was to identify an ideal large animal model for COVID-19. Based on the available and included studies, it has been determined that neither cynomolgus macaques, rhesus macaques, or ferrets can be singled out as an ideal model of the disease. The greatest obstacle to this discovery is the lack of experimental and reporting standards. Based on the information already published and the current data, the only thing that can be concluded is that data needs to be shared more easily, and that every study must be published and registered to maximize animal use and data comparison. The path to find an ideal animal model for any infectious disease begins with standardization. 

References:

Baker, D., Lidster, K., Sottomayor, A., & Amor, S. (2014). Two Years Later: Journals Are Not Yet Enforcing the ARRIVE Guidelines on Reporting Standards for Pre-Clinical Animal Studies. PLoS Biology, 12(1). https://doi.org/10.1371/journal.pbio.1001756&nbsp;

Egan, K. J., Vesterinen, H. M., Beglopoulos, V., Sena, E. S., & Macleod, M. R. (2016). From a mouse: systematic analysis reveals limitations of experiments testing interventions in Alzheimer’s disease mouse models. Evidence-Based Preclinical Medicine, 3(1), e00015. https://doi.org/10.1002/ebm2.15&nbsp;

Gerdts, V., Wilson, H. L., Meurens, F., Van den Hurk, S. van D. L., Wilson, D., Walker, S., … Potter, A. A. (2015). Large animal models for vaccine development and testing. ILAR Journal, 56(1), 53–62. https://doi.org/10.1093/ilar/ilv009&nbsp;

Imai, M., Iwatsuki-Horimoto, K., Hatta, M., Loeber, S., Halfmann, P. J., Nakajima, N., … Kawaoka, Y. (2020). Syrian hamsters as a small animal model for SARS-CoV-2 infection and countermeasure development. Proceedings of the National Academy of Sciences117(28), 202009799. https://doi.org/10.1073/pnas.2009799117&nbsp;

Kilkenny, C., Browne, W. J., Cuthill, I. C., Emerson, M., & Altman, D. G. (2010). Improving bioscience research reporting: The arrive guidelines for reporting animal research. PLoS Biology, 8(6). https://doi.org/10.1371/journal.pbio.1000412&nbsp;

Lawler, J. V, Endy, T. P., Hensley, L. E., Garrison, A., Fritz, E. A., Lesar, M., … Paragas, J. (2006). Cynomolgus Macaque as an Animal Model for Severe Acute Respiratory Syndrome. PLoS Medicine, 3(5), e149. https://doi.org/10.1371/journal.pmed.0030149&nbsp;

McAuliffe, J., Vogel, L., Roberts, A., Fahle, G., Fischer, S., Shieh, W. J., … Subbarao, K. (2004). Replication of SARS coronavirus administered into the respiratory tract of African Green, rhesus and cynomolgus monkeys. Virology, 330(1), 8–15. https://doi.org/10.1016/j.virol.2004.09.030&nbsp;

Munster, V. J., Feldmann, F., Williamson, B. N., van Doremalen, N., Pérez-Pérez, L., Schulz, J., … de Wit, E. (2020). Respiratory disease in rhesus macaques inoculated with SARS-CoV-2. Nature, 585(7824), 268–272. https://doi.org/10.1038/s41586-020-2324-7&nbsp;

Nagata, N., Iwata-Yoshikawa, N., & Taguchi, F. (2010). Studies of severe acute respiratory syndrome coronavirus pathology in human cases and animal models. Veterinary Pathology, 47(5), 881–892. https://doi.org/10.1177/0300985810378760&nbsp;

Roberts, A., Deming, D., Paddock, C. D., Cheng, A., Yount, B., Vogel, L., … Subbarao, K. (2007). A Mouse-Adapted SARS-Coronavirus Causes Disease and Mortality in BALB/c Mice. PLoS Pathogens, 3(1), e5. https://doi.org/10.1371/journal.ppat.0030005&nbsp;

Roberts, A., Vogel, L., Guarner, J., Hayes, N., Murphy, B., Zaki, S., & Subbarao, K. (2005). Severe Acute Respiratory Syndrome Coronavirus Infection of Golden Syrian Hamsters. Journal of Virology, 79(1), 503–511. https://doi.org/10.1128/jvi.79.1.503-511.2005&nbsp;

Shi, J., Wen, Z., Zhong, G., Yang, H., Wang, C., Huang, B., … Bu, Z. (2020). Susceptibility of ferrets, cats, dogs, and other domesticated animals to SARS-coronavirus 2. Science, 368(6494), 1016–1020. https://doi.org/10.1126/science.abb7015&nbsp;

Sun, S. H., Chen, Q., Gu, H. J., Yang, G., Wang, Y. X., Huang, X. Y., … Wang, Y. C. (2020). A Mouse Model of SARS-CoV-2 Infection and Pathogenesis. Cell Host and Microbe, 28(1), 124-133.e4. https://doi.org/10.1016/j.chom.2020.05.020&nbsp;

Thachil, J., Longstaff, C., Favaloro, E. J., Lippi, G., Urano, T., & Kim, P. Y. (2020). The need for accurate D‐dimer reporting in COVID‐19: Communication from the ISTH SSC on fibrinolysis. Journal of Thrombosis and Haemostasis, 18(9), 2408–2411. https://doi.org/10.1111/jth.14956&nbsp;

Triple F Research. (2015). Triple F Farms Ferret Growth Curve

Weingart, H. M., Copps, J., Drebot, M. A., Marszal, P., Smith, G., Gren, J., … Czub, M. (2004). Susceptibility of Pigs and Chickens to SARS Coronavirus. Emerging Infectious Diseases, 10(2), 179–184. https://doi.org/10.3201/eid1002.030677&nbsp;

Yu, P., Qi, F., Xu, Y., Li, F., Liu, P., Liu, J., … Qin, C. (2020). Age‐related rhesus macaque models of COVID‐19. Animal Models and Experimental Medicine, 3(1), 93–97. https://doi.org/10.1002/ame2.12108&nbsp;

Appendix 

Leave a Reply