Giving the Genes a Shuffle: Using Natural Variation to Understand Host Genetic Contributions to Viral Infections.

The laboratory mouse has proved an invaluable model to identify host factors that regulate the progression and outcome of virus-induced disease. The paradigm is to use single-gene knockouts in inbred mouse strains or genetic mapping studies using biparental mouse populations. However, genetic variation among these mouse strains is limited compared with the diversity seen in human populations. To address this disconnect, a multiparental mouse population has been developed to specifically dissect the multigenetic regulation of complex disease traits. The Collaborative Cross (CC) population of recombinant inbred mouse strains is a well-suited systems-genetics tool to identify susceptibility alleles that control viral and microbial infection outcomes and immune responses and to test the promise of personalized medicine.

Viral Disease Is a Complex Trait

Viral infections pose a major threat to human and animal health, causing significant morbidity and mortality every year. During the past decades, the emergence of several highly pathogenic zoonotic viruses has demonstrated the fragility of the species barrier in protecting human and farm animal populations from pathogens that exist in the animal kingdom. While basic virological research offers the possibility to detect viral strains that are ‘poised’ for emergence and identify mutations that might promote pathogen emergence, accurate prognoses are confounded by our inability to predict disease severity and virulence. Importantly, viruses with identical genome sequences do not always cause the same set of clinical manifestations in humans. Moreover, the complex interplay between environmental, viral, and host genetic factors drives differences in interindividual disease progression, severity, and outcome. These factors change over the course of a lifetime and some, like individual health status, comorbidities, and environmental factors 1, 2, are difficult if not impossible to control. However, perhaps one of the most important key players in the fragile balance of microbial pathogenesis centers around host genetic susceptibility alleles that dramatically influence the course of disease in different individuals. In humans, a growing number of genetic factors like entry receptors, receptor-modifying enzymes, and innate and adaptive immune-related proteins that regulate influenza virus, norovirus, rotavirus, respiratory syncytial virus (RSV), HIV, hepatitis B and C viruses, herpes virus, and other acute and chronic virus disease outcomes have been identified (Table 1 ; more detailed list in [3]).

Table 1

Genes with Significant Associations with Viral Disease in Humans

Pathogen	Phenotype	Causal gene	Refs
Dengue virus (DENV)	DENV shock syndrome	MICB, PLCE1	[73]
Epstein–Barr virus (EBV)	EBNA-1 IgG titer	HLA-DRB1, HLA-DQB1	[74]
Hepatitis B virus (HBV)	Chronic infection	HLA-DPA1, HLA-DPB1	[75]
	Persistence	INST10	[76]
Hepatitis C virus (HCV)	Spontaneous clearance	IL28B	[77]
	Development of hepatocellular carcinoma	TLL1	[78]
	Progression to hepatocellular carcinoma	DEPDC5	[79]
HIV-1	Viral load	HLA-B, HLA-C	[80]
	Viral load control	HCP5	[81]
Influenza A virus (IAV)	Reduced restriction of viral replication	IFITM3	[82]
	Increased incidence and increased risk of viral pneumonia	TNF	[83]
Norwalk virus (NoV)	Resistance	FUT2	[84]
Respiratory syncytial virus (RSV)	Bronchiolitis	SFPA/D	85, 86
West Nile virus (WNV)	Resistance	CCR5	[87]

Accordingly, research on host genetics is a promising tool for understanding susceptibility and virulence patterns in human populations and refining pandemic-preparedness efforts. Furthermore, the discovery of innovative prophylactic or diagnostic and therapeutic treatment options for viral disease that can be leveraged across different host genetic susceptibility patterns can lead to improved personalized medicine.

In this review we discuss historic and new platform strategies designed to unravel the interplay between the complex host and viral genetic determinants that regulate disease severity. Moreover, we discuss recent developments in the field of complex genetics designed to resolve quantitative trait loci (QTLs) (see Glossary) and rapidly identify single candidate genes and alleles that regulate microbial pathogenesis.

The Laboratory Mouse in Viral Disease Research

Animal models offer a strategy to reduce system-wide complexity through standardizing environmental influences without losing the integrity of a functional biological system. By far, most in vivo viral pathogenesis studies are conducted in inbred mouse models. Not only are the husbandry and breeding of mice cost-efficient, but genome sequences, as well as many species-specific immunologic, molecular, and biochemical reagents, are available to the research community. However, disease spectra in inbred mouse models are narrow compared with the diverse spectra noted in outbred populations like humans. For respiratory viral infections, additional phenotypic variations must also be considered, as mice do not sneeze, cough, or develop fever following infection. Rather, they exhibit loss of body weight, reduced respiratory function, and decreased locomotive activity. As human pathogens often replicate less efficiently in mice, it is frequently necessary to use mouse-adapted viral strains selected for increased replication and disease in inbred mouse strains, which may or may not replicate disease symptoms seen in humans [4]. The most commonly used parameters for viral infection intensity in mice are changes in body weight and survival rate. Detailed analysis of disease progression can be undertaken utilizing time-course experiments during which samples of interest are collected and disease kinetics revealed. Most recently, mouse genetic reference populations with diverse genetic backgrounds have been developed that replicate the genetic and disease outcome variability found in outbred populations like humans. These new platforms are further supported by technologies that allow targeted genetic modifications, enabling researchers to study how complex traits regulate microbial pathogenesis.

Genetic Manipulation of the Mouse Genome

Using the mouse as a model organism offers two distinct types of genetic approaches: reverse and forward genetic studies. The most commonly used application in reverse genetic approaches involves the specific ablation of a single gene, either by deleting parts of or the entire gene or by replacing coding exons. Gene trapping by comparison offers the possibility to insert reporter genes into the gene of interest, disrupting its function. Both techniques require the DNA construct of choice to be transfected into mouse embryonic stem cells (mESCs), injection of screened mESC clones into a blastocyst, and transfer of this blastocyst into the uterus of the host animal. Most recently, innovative techniques like transcription activator-like effector nucleases (TALENs) [5] and clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) [6] are accelerating the process of targeted genome editing, including the introduction of susceptibility alleles that improve virus replication and pathogenesis in the mouse [7].

Genetically engineered mouse strains have represented the gold standard to investigate the role of specific host genes in regulating disease and immune outcomes following virus infections (e.g. 8, 9). Furthermore, genetically modified strains can be used [10] to understand human monogenic diseases like cystic fibrosis, polycystic kidney disease, and sickle cell disease. Although complete (biallelic) knockout of genes in humans is a relatively rare event, a recent study showed that healthy people have about 100 loss-of-function variants and 20 completely inactivated genes [11]. Consequently, the majority of human diseases are complex traits whose disease outcomes are influenced by multiple genetic factors.

Human genomes are characterized by extensive natural variation, which has accumulated over time through, for example, mutation events (external damage to DNA or internal errors during replication), gene flow, and sexual reproduction, which drives phenotypic interindividual differences across populations. Differences in DNA sequence can modify transcript and protein levels by altering their functional properties, timing, level, and site of expression [12]. Natural variation can be used in forward genetic studies to identify novel genes involved in a variety of quantitative traits and diseases [13].

The occurrence of spontaneous mutations in laboratory mouse strains was the first platform employed for forward genetic approaches [14]. There are various methods to increase the likelihood of mutations by treating male mice with mutagens such as N-ethyl-N-nitrosourea (ENU) [15] or chlorambucil 16, 17, by irradiation 18, 19, and by utilizing transposons such as the sleeping beauty [20] or piggyback [21] system, named according to their transposases, to insert specific DNA sequences. Breeding of those mutated mice allows selection of those with an altered phenotype in the trait of interest. This approach mimics natural genetic variation in humans in the controlled setting of the mouse model and has the power to reveal genomic variation and networks of genes influencing a phenotype rather than analyzing the effect of the absence of a single-gene product, making the translation from mouse to human more likely.

Genetic mapping studies are undertaken to identify the genomic region that causes the altered phenotype of interest. Commonly, QTL analysis is used to reveal genotype–phenotype associations [22]. F2 crosses between an inbred strain carrying the aberrant phenotype of interest and another inbred mouse strain lacking this particular phenotype have been widely performed. The identified chromosomal regions can be large, containing several hundreds of genes. The size of the chromosomal region exclusively depends on the number of recombination breakpoints in the cross and the genetic complexity of the region. Chromosomal locations can be narrowed using consomic, conplastic, congenic, recombinant inbred (RI), or recombinant congenic mouse strains [23] (Figure 1 ).

Figure 1

Various Genetic Mouse Model Approaches to Narrow Quantitative Trait Locus (QTL) Regions. (A) Congenic mouse strains are produced by transferring a genomic region from an inbred donor strain to an inbred recipient strain through repeated backcrossing. (B) Consomic mouse strains contain an entire chromosome from a donor strain and are generated via backcrossing to the recipient strain. (C) In conplastic mouse strains, the entire mitochondrial DNA is derived from a donor strain and is generated through backcrossing of females from the donor strain to males from the recipient strain. (D) Recombinant inbred mouse strains are generated by crossing two inbred strains to obtain an F1 generation. These F1 mice are crossed to create an F2 generation, which is brother–sister mated for at least 20 generations to achieve mice with a fixed genetic background and equal contributions of the two parental strains. (E) Recombinant congenic strains are produced by crossing two inbred mouse strains. The resulting F1 generation is backcrossed twice (BC1 and BC2) before they are brother–sister mated for an additional 14 generations. The genome composition of the final strains is skewed towards one parental strain in a 7:8 ratio.

The alternative to dealing with large chromosomal regions and the narrowing process is to use mouse genetic reference populations (GRPs). RI stains of mice are popular due to their long-term genetic stability, which helps in integrating data collected in different settings and reproducibility over a long time. The most extensively used mouse GRP is the BXD family of recombinant inbred strains. They are derived from a cross between C57BL/6J and DBA/2J mice [24]. To obtain an F1 generation, C57BL/6J and DBA/2J mice were crossed. The resulting F1 generation is mated to achieve the F2 generation, which is subsequently brother–sister mated for at least 20 generations to generate inbred mouse strains with a fixed genetic architecture. Completely inbred strains are then called BXD strains, which are genotyped once (one animal per line, not every individual) and can be phenotyped indefinitely for every trait of interest. Currently, there are 156 BXD strains available [25]. For decades, recombinant inbred strains such as the BXD family have been used extensively as tools for genetic mapping of Mendelian and quantitative traits. To identify single genes that are responsible for the observed phenotype fine mapping, sequence analysis, expression profiling, and functional studies are typically performed [26]. However, the identification of causal gene variants remains challenging due to the large size and the number of genes under the identified QTL region, coupled with the fact that the parental strains were identical by descent resulting in so-called ‘blind spots’ for genetic mapping. Wild-derived strains other than Mus musculus domesticus need to be employed to cover those spots and increase genetic variation [23]. Various resources have been established to address this issue, among them the heterogeneous stock (HS), which is derived from eight founder strains (A/J, AKR/J, BALBc/J, CBA/J, C3H/HeJ, C57BL/6J, DBA/2J, and LP/J) and maintained through random mating. No inbred mouse lines are created and therefore each mouse exhibits a unique combination of alleles with the goal of containing random variation similar to the human population [27]. The caveat of this mouse population is that every individual mouse needs to be genotyped, which might be too expensive for some researchers. However, genotyping technologies are evolving constantly and costs are decreasing rapidly.

The Collaborative Cross

Another mouse GRP that includes other M. musculus subspecies is the CC Mouse Resource (Figure 2 ), which has already been used to successfully identify highly promising candidate genes that are influencing susceptibility or resistance to viral infections (Table 2 ).

Figure 2

Generation of the Collaborative Cross Resource. (A) The Collaborative Cross panel of recombinant inbred mouse strains is a multiparental population that is derived from eight founder strains. Among these founder strains are classical laboratory mouse strains (A/J, C57BL/B6, 129S1/SvImJ), mouse models for human diseases (NOD/ShiLtJ – type 1 diabetes, NZO/HlLtJ – obesity), and wild-derived mouse strains (CAST/EiJ, PWK/PhJ, WSB/EiJ). Every mouse strain was assigned a letter (A–H) and a particular color that are used by the entire research community. (B) Breeding-funnel design of the Collaborative Cross that guarantees equal distribution of founder alleles to the resulting CC strain. Depicted is the specific breeding funnel for chromosome 19 of the CC strain CC001. (C) Genome architecture of CC001 with founder contributions displayed in their respective colors. Photograph: Klaus Schughart.

Table 2

QTLs of CC Studies Using Different Viruses

Pathogen	Phenotype	QTL region	% Variation	Number of genes under QTL	Refs
SARS-CoV	Vascular cuffing	HrS1: Chr. 3: 18286790–26668414	26%	26 [narrowed to one (Trim55)]	[40]
	Viral titer	HrS2: Chr. 16: 31583769–36719997	22%	92 (narrowed to 48)
	Eosinophil infiltration	HrS3: Chr. 15: 72103120–75803414	26%	63 (narrowed to 25 – functional change only in Bai1)
	Vascular cuffing	HrS4: Chr. 13: 52822984–54946286	21%	30 (narrowed to nine – Cdhr2)
	D3% weight	HrS5: Chr. 18: 27108062 – 58694005	6.6%	158 [narrowed to one (Ticam2)]	[61]
	D4% weight	HrS5: Chr. 18: 27108062–58694005	8.5%
	Log titer	HrS5: Chr. 18: 27108062–58694005	12.9%
	Hemorrhage	HrS5: Chr. 18: 24762824–7829634	6%
	D3% weight	HrS6: Chr. 9: 116476207–telomere	7%	–
	Log titer	HrS7: Chr. 7: 55169841–11722358	12.3%	–
	Log titer	HrS8: Chr. 12: 81649471–108529109	5.4%	–
	Hemorrhage	HrS9: Chr. 15: centromere–64.430001	9.1%	–
WNV	Frequency of CD73+ Tregs	HI1: Chr. X: 166 Mb–telomere	–	43 (narrowed to 22)	[88]
	Decreased frequency of CXCR3+ Tregs, CXCR3+ CD4+, and CD8+ T cells	HI2: Chr. X: 100–106 Mb	–	42 (narrowed to 26)
	Increased frequency of ICOS+ Tregs in spleen	HI3: Chr. X: 140–145 Mb	–	18 (narrowed to 11)
IAV	D4 weight, log titer, IHC score, D3 clinical, airway inflammation, airway damage	HrI1: Chr. 16: 97.5 Mb–98.2 Mb	41.67%	Ten (including Mx1)	[39]
	D4 weight	HrI2: Chr. 7: 89.1 Mb–96.7 Mb	9.7%	69
	Pulmonary edema	HrI3: Chr. 1: 21.7 Mb–29 Mb	29.73%	24
	Airway neutrophils	HrI4: Chr. 15: 77.4 Mb–86.6 Mb	22.7%	206

To expand genetic variation in GRPs, an innovative strategy for a multiparent population (MPP) of mice was conceptualized in the early 21st century and developed over the next decade [28]. Use of an octoparental crossing scheme between genetically distinct mouse strains was proposed and modifications through the research community were integrated. Breeding of this novel GRP specifically designed for complex genetics started in 2002 [28]. The eight founder strains of the CC include three classical laboratory strains (A/J, C57BL/6J, and 129S1/SvImJ), which have been used extensively in biological research and build the genetic backbone on which most of the knockout mouse strains are generated. Moreover, two mouse models for common human diseases were included (NOD/ShiLtJ for type 1 diabetes and NZO/HlLtJ for obesity) to address research questions of comorbidities. The addition of three wild-derived mouse strains (CAST/EiJ, PWK/PhJ, and WSB/EiJ) not only increased the genetic diversity by adding new alleles that are not present among the classical inbred and disease model strains, but also covered different phylogenetic origins of the mouse species (CAST/EiJ – Mus musculus castaneous, PWK/PhJ – Mus musculus musculus) to encompass 90% of genetic variation present in the M. musculus species [23]. Accordingly, the CC population reaches a level of genetic diversity comparable with the diversity found in the human population.

To guarantee equal contributions of all eight founder strains to each of the resulting CC strains, a specific breeding funnel was elaborated with around 135 unique recombination events and segregating polymorphisms every 100–200 bp [28]. In this way susceptibility alleles are scrambled in new ways, allowing novel allelic combinations leading to an extension of phenotypic range beyond the scope observed in the parental strains. Breeding was performed at three different locations: Oak Ridge National Laboratory in Oak Ridge, TN, which moved to the University of North Carolina at Chapel Hill [29]; the International Livestock Research Institute in Nairobi, Kenya, which moved to the Tel Aviv University in Tel Aviv, Israel [30]; and the Western Australian Institute for Medical Research/Geniad Ltd in Perth, Australia [31]. Although the theoretical plan was perfectly elaborated, hundreds of CC strains became extinct, almost half from problems in male infertility [32].

Prior to the development of the final CC resource, incipient CC lines that were not fully inbred yet (pre-CC lines) were used in genetic mapping studies to provide proof of concept and to show the potential of this newly designed GRP (Box 1 ). Candidate genes for various phenotypes, such as susceptibility to Aspergillus fumigatus infections [33], energy balance traits [34], differences in hematological parameters [35], susceptibility to Klebsiella pneumoniae [36], and neutrophilic inflammation due to house dust mite-induced asthma [37], were successfully identified. Moreover, mapping efforts revealed expression QTLs for extreme host responses to influenza A virus (IAV) infections [38], host response QTLs to IAV [39], and severe acute respiratory syndrome coronavirus (SARS-CoV) pathogenesis [40]. A common observation in all of the studies was that pre-CC lines exhibit an enhanced phenotypic range compared with the variation observed in the eight founder strains or other classical inbred strains and that it was possible to dissect traits that were thought to be inseparably entwined.

As an expansion to their first paper identifying QTLs for SARS pathogenesis in pre-CC mice, the authors utilized an alternative genetic approach to understand host genetic contributions to the course and outcome of SARS-CoV infection [61]. Two pre-CC lines with divergent outcome after SARS-CoV infection were identified. CC003/Unc is resistant to SARS-CoV whereas CC053/Unc is highly susceptible. To dissect host genetic factors that lead to the different outcomes, these two strains were bred and 264 F2 animals generated. Loss of body weight, viral titer in lungs, pulmonary hemorrhage, and histopathological changes were analyzed at multiple time points after infection. Overall, F2 mice exhibited a broader phenotypic range than the parental strains. Five significant QTLs across all analyzed phenotypes were identified, one of which affected multiple SARS-CoV response phenotypes. This QTL (HrS5) on chromosome 18 (27.1–58.6 Mb) was selected for follow-up studies. Integration of different bioinformatics and database approaches led to the identification of Ticam2 as a highly promising candidate gene. Ticam2 is a toll-like receptor (TLR) adapter protein that had not been shown to play an important role in SARS-CoV pathogenesis. Utilizing another tool from the geneticist’s toolbox, Ticam2 knockout mice (Ticam2 −/−) were employed to investigate its effect on disease progression and outcome. Ticam2 −/− mice exhibited significantly more weight loss, similar viral load in lungs on day 4 after infection (however, significantly higher titers on day 2 after infection were reported previously [8]), and similar histopathological findings but significantly increased pulmonary hemorrhage on day 4 after infection. Thus, the authors successfully showed that a screening approach in CC mice in combination with an F2 follow-up study can lead to single candidate genes that can be confirmed using reverse genetic tools. It remains to be determined whether allele swaps can be used to identify the allele driving this disease phenotype.

Alt-text: Box 1

Highlighting another powerful characteristic of the CC, completely new mouse models for spontaneous colitis [41], Ebola-associated hemorrhagic fever [42], novel neurological responses to Theiler’s murine encephalomyelitis virus (TMEV) [43], and persistent West Nile virus (WNV) infection in the brain [44] were discovered, a harbinger of new model systems that may emerge over time.

The Diversity Outbred (DO) population of mice is a GRP complementary to the CC derived from the same set of founder strains. Instead of inbreeding, an outbred population is maintained through random mating, which enhances the mapping resolution even further through the acquisition of even more recombination sites [45]. Early studies utilizing DO mice led to the identification of a specific isoform of Apobec1 contribution to atherosclerosis [46] and sulfotransferases as candidate genes for benzene-induced genotoxicity [47].

Designed specifically for the analysis of complex traits, the CC and DO populations of genetically highly diverse mice provide the first true systems-genetic platform for cumulative and integrated data collection [48]. Systems genetics, as an innovative strategy to investigate the role of host genetics that encompasses diverse molecular omics data [49], also catalyzed the development of a variety of analytic and informatics tools and methods 50, 51, 52, 53 and provides state-of-the-art tools for genetic mapping and candidate gene identification and an opportunity to test the promise of predictive genomic medicine.

From Complex Screens to Candidate Genes: A Recipe for Complex Genetic Studies

There is increasing evidence for host genetic regulation of viral and microbial pathogenesis in humans (Table 1). Although classical approaches like cell culture experiments or knockout mouse studies have been successfully used to understand infectious disease pathogenesis, the contribution of variation in host genetics can no longer be omitted because it more accurately phenocopies the human condition and is critical to pave the way to personalized medicine. The hurdles for most researchers who are not using complex genetics until now are the lack of expertise in designing complex genetic studies and the inevitable bioinformatics barriers until now. However, complex genetic approaches become exponentially more valuable as data is accumulated and compared across experimental setups and pathogens and combined to gain deeper knowledge. In this review, we provide an experimental framework for the rapid implementation of complex-trait genetic strategies in the laboratory setting, including design of the overall study, best utilization of the CC resource, and ways to narrow down QTLs to pinpoint single candidate genes influencing the trait of interest (reviewed in [54]) (Figure 3 ).

Figure 3

Schematic of Experimental Approach for Complex Genetic Studies of Viral Infections. eQTL, expression QTL; QTL, quantitative trait loci.

The classic CC hypothesis is that ‘my disease phenotype’ is a complex oligogenic trait regulated by multiple genetic polymorphisms in different genes, which interact to regulate disease outcomes in natural populations. In a second step, the breadth and duration of disease-associated phenotypes is selected and measured in a subset of mouse strains, which typically is either a random selection of CC strains or the eight CC founder strains (RSV – Buntzman 2016, IAV [55]). Analyses of phenotypic baselines are crucial to be able to distinguish between phenotypes that are caused by differences in genome architecture or those that are related to viral pathogenesis. Additionally, multiple sample collection is advised to maximize phenotypic comparisons across the final population. This is important as it is impossible to predict phenotypic outcome as different gene variants segregate in the final population and traits that are closely linked might be broken apart. Another important factor that might complicate experimental design is the impact of known genes of major effect, like Mx1 for IAV [39] or Oas1b for flaviviruses [56]. Selection of CC strains that carry a variation at that locus or even a null allele might help to uncover rather small effects by other modulating genes. Utilizing up to eight different alleles at any given locus in the genome allows the discovery of gene of major effect-dependent and -independent processes (IAV [55], WNV [57]). There is a risk that genetic mapping might exclusively identify the gene of major effect if no attention is paid to these genes while selecting CC strains for the respective study. Genetic mapping offers the possibility to add covariates like batch effects to the calculation and randomization should be achieved whenever possible. Sample collection should anticipate that the CC model is likely to identify CC strains that progress to different stages in disease severity, replicating phenotypes seen in human populations and allowing the identification of susceptibility alleles that regulate disease progression from mild to severe to chronic infections in vivo. This is important as it has been shown that novel disease phenotypes might be discovered in the CC (severe neuroinvasive disease and chronic WNV infection [44]). New models for diseases that are completely unrelated (spontaneous colitis [41], human leukocyte adhesion and recruitment deficiencies [55]) or related to the study design (SARS-CoV [40], Ebola [42]) will generate a panel of variable disease-state mouse models that capture the different phases of disease seen in human populations.

After all phenotype measurements are obtained, genetic mapping is utilized to calculate the likelihood of every position in the genome being associated with the analyzed phenotype, and regions with the highest logarithm of the odds (LOD) score are identified as QTLs. Zooming into the QTL region, bioinformatics tools can be employed to generate allele-effect plots, which show the contribution of each of the eight possible alleles to the QTL. Analyses of allele effects are a good way to determine which parental alleles are driving high and low responses. Afterwards, variations that are unique to the identified strain can be analyzed utilizing the Sanger SNP browseri. Additionally, other resources, like BioGPSii or the gene browser included in Pubmediii, are of great use to gain information on known expression patterns and biological functions. In the line of a systems-genetics approach, transcriptional analyses can be used to identify differentially expressed genes or respective pathways influenced by a particular QTL (WNV [58]). However, these downstream analyses should only be performed on a subset of extreme phenotype samples, as this approach is highly informative and inexpensive. Using different pathogens for the same type of analysis can lead to the discovery of common or unique features across different viral species (IAV and SARS-CoV [59]). Additional tools of the complex genetic toolkit such as diallels of the CC founders and their reciprocal F1 hybrids allow the identification of different types of heritable effects (IAV [60]). Alternatively, once a highly promising candidate gene is identified the use of knockout mouse strains to validate the contribution of this particular gene to the observed phenotype can be used (SARS-CoV 40, 61). The International Knockout Mouse Consortium (IKMC) will soon achieve its goal of having a mouse mutant or a targeted mESC for every gene (http://www.mousephenotype.org), providing a crucial resource for functional annotation and validation of candidate genes. Importantly, it has been shown that the genetic background on which the knockout is created plays an important role 62, 63 and with recent advances of CRISPR/Cas9 technology the development of the identical knockout on different mouse backgrounds may be achievable in a time- and cost-efficient manner. Additionally, CRISPR/Cas9-generated allele swaps between mouse strains offer enhanced specificity, applicability, and translatability compared with complete knockout of genes. Translational aspects can be addressed by comparing identified SNPs with human SNP and gene databases or conserved structural elements and functional motifsiv,v,vi.

Concluding Remarks and Future Directions

The contribution of host genetic factors to the progression and outcome of viral infections has not only been proved but represents a powerful new tool to reveal the complex interplay between novel genes and their polymorphisms and disease severity. Although critical voices were raised in the scientific community [64] as well as in the media, it has been shown that data collected during viral infections in mice and in humans are highly correlated. For example, comparing signature genes derived from transcriptome analyses of IAV-infected CC founder strains, but not classic inbred mouse strains, with infected human volunteers revealed that gene expression in the blood of infected CC mice reproduces much more representative human signature profiles [65].

Human genetic studies are conducted using either candidate-gene approaches or genome-wide association studies (GWASs) [66]. Depending on the research question, both study designs have their own advantages, disadvantages, and limitations. Nevertheless, GWASs revealed many loci that are associated with human diseasevii. However, small numbers of participants, limited sample size, and technical differences in sample collection, and the inability to validate candidate gene–disease associations, make cross-study comparison challenging and reproducibility difficult. Additionally, the low number of accessible human samples often results in lack of statistical significance and/or reproducibility 67, 68. Recommendations on how to enhance the transparency of human GWASs [Strengthening the Reporting of Genetic Association Studies (STREGA)] have been published [69].

However, although human genetic factors influencing infectious disease susceptibility and outcome have been identified, significant difficulties remain in the verification of these factors in humans, mostly due to small effect sizes. Moreover, there are phenotypic traits, especially in the infectious disease context, that are difficult or even impossible to investigate in humans. The only way to unravel the effect of alleles associated with diseases that influence a specific molecular process is to investigate those alleles in controlled experimental settings either individually or in combination [12]. The mouse as the model organism of choice is currently the only mammalian systematic platform that offers not only the resources but also the technology to identify susceptibility and resistance loci for viral infections.

The CC population recombinant inbred mouse panel represents a paradigm shift for microbial pathogenesis studies, immunology, and studies of the role of complex genetic traits in disease. It offers increased mapping resolution compared with classical mouse GRPs and has already been successfully used in the field of infectious diseases, suggesting that an even greater impact exists for the field of immunology [70]. Systems-biology approaches in which a plethora of genetic and genomic (omics) data for the identical experimental condition is collected allows integration of data and offers the possibility of complex modeling and, ultimately, the identification of key factors driving differences in disease progression and outcome [71]. Being a relatively young and novel platform for systems genetics, the CC and its complementary resources have accelerated not only the identification of QTLs from a wide range of phenotypes but also pushed the development of genotyping and databases as well as bioinformatics tools [72]. We would like to highlight that the more that incipient lines of the CC are used in different fields to explore different phenotypes and answer different question, the more valuable the resource itself becomes (see Outstanding Questions). Data generated by laboratories from different fields might be of use for others, shortening their way to obtain results and saving money along the way. In summary, GWASs in mice can help to derive working hypotheses and direct human studies and, accordingly, studies in the two species are highly complementary.