New Perspective on the Geographic Distribution and Evolution of Lymphocytic Choriomeningitis Virus, Central Europe

Lymphocytic choriomeningitis virus (LCMV) is an Old World mammarenavirus found worldwide because of its association with the house mouse. When LCMV spills over to immunocompetent humans, the virus can cause aseptic meningitis; in immunocompromised persons, systemic infection and death can occur. Central Europe is a strategic location for the study of LCMV evolutionary history and host specificity because of the presence of a hybrid zone (genetic barrier) between 2 house mouse subspecies, Mus musculus musculus and M. musculus domesticus. We report LCMV prevalence in natural mouse populations from a Czech Republic–Germany transect and genomic characterization of 2 new LCMV variants from the Czech Republic. We demonstrate that the main division in the LCMV phylogenetic tree corresponds to mouse host subspecies and, when the virus is found in human hosts, the mouse subspecies found at the spillover location. Therefore, LCMV strains infecting humans can be predicted by the genetic structure of house mice.


Lymphocytic choriomeningitis virus (LCMV) is an Old
World mammarenavirus found worldwide because of its association with the house mouse. When LCMV spills over to immunocompetent humans, the virus can cause aseptic meningitis; in immunocompromised persons, systemic infection and death can occur. Central Europe is a strategic location for the study of LCMV evolutionary history and host specifi city because of the presence of a hybrid zone (genetic barrier) between 2 house mouse subspecies, Mus musculus musculus and M. musculus domesticus. We report LCMV prevalence in natural mouse populations from a Czech Republic-Germany transect and genomic characterization of 2 new LCMV variants from the Czech Republic. We demonstrate that the main division in the LCMV phylogenetic tree corresponds to mouse host subspecies and, when the virus is found in human hosts, the mouse subspecies found at the spillover location. Therefore, LCMV strains infecting humans can be predicted by the genetic structure of house mice.
In 2010, Albariño et al. (16) investigated the genetic diversity and distribution of LCMV variants by analyzing 29 genomes. They demonstrated that LCMV is highly diverse and forms 4 distinct lineages (I-IV) but found little correlation of those lineages with time or place of isolation. From their dataset, only 3 strains (Marseille12-2004, Yale-1977, and Michigan-2005) originated from wild mice, but those strains were not assigned to subspecies. Furthermore, the place of isolation is a poor proxy for the origin of spillover to human hosts. For example, focusing on lineage II of Albariño et al., strains M1 and M2 were isolated in Japan in 2005, but came from a wild-derived strain originating from M. musculus musculus mice caught in Illmitz, Austria, in 1985 (17). Likewise, the Dandenong-Yugoslavia LCMV strain (18) was isolated in Australia from a human spillover, but that person returned from the former Yugoslavia before becoming ill and dying. The Bulgaria 1956 strain (19) was isolated from a human spillover, but geographic origin was not mentioned in the original study; a contact of that patient was treated for the same symptoms in a hospital in Vidin, Bulgaria, suggesting spillover origin in northwestern Bulgaria. Finally, the last LCMV strain in lineage II, LE-FRANCE (20), was isolated from a pregnant woman in France (i.e., within M. musculus domesticus mouse territory), but the person worked in a pet store, making strain origin uncertain because other rodent species, especially hamsters, are known to be LCMV carriers (3,4,20). In summary, for 3 of 4 LCMV strains in lineage II, the potential spillover origin is consistent with M. musculus musculus mouse territory despite diverse viral isolation locations. Similarly, in LCMV lineage I, strains were found in laboratory mice, essentially of M. musculus domesticus origin (21); wild mice; or in primate (including human) spillovers in the United States or western Europe, and were thus consistent with M. musculus domesticus mouse origin (22). LCMV lineage IV consists only of strains isolated from woodmice (Apodemus sylvaticus) from Spain. Given these observations, we hypothesized that host specificity could be a better predictor of LCMV genetics than the place or time of LCMV strain isolation.
In this study, we test the hypothesis that LCMV phylogenetic clustering reflects specificity to its host reservoirs by investigating the diversity of LCMV in central Europe across the house mouse hybrid zone (HMHZ). We also update the phylogenetic analysis of LCMV from Albariño et al. (16) by complementing their dataset with LCMV genomes sequenced in the last decade and with our data.

Sampling and Mouse Genotyping
A total of 748 house mice (410 M. musculus domesticus and 338 M. musculus musculus) from 179 localities  Note that house mice may not be found throughout the complete extent of some areas (e.g., subarctic regions, the Sahara Desert, and the Amazon rainforest). The tan, purple, and gray areas indicate regions of hybridization. Red arrows indicate inferred routes of historical migrations and recent movements in association with humans. Adapted from (7,8). Copyright ©2012 Springer-Verlag. All rights reserved. Adapted with permission from Springer Science and Business Media and Michael Nachman.
(100 for M. musculus domesticus and 79 for M. musculus musculus) were trapped in farms during 2008-2019 across a 145-km by 110-km belt stretching from northeastern Bavaria (Germany) to western Bohemia (Czech Republic), a region in which these mouse subspecies meet and form the HMHZ (23) (Figure 2; Appendix 1 Table 1, https://wwwnc.cdc.gove/EID/ article/27/10/21-0224-App1.xlsx). Tissue samples were preserved in liquid nitrogen and later stored at −80°C as described in Goüy de Bellocq et al. (24). Mice were identified on the basis of a set of diagnostic markers as in Macholán et al. (23) or on the basis of 1,401 single-nucleotide polymorphism (SNP) markers (25) or 0.62 million SNP markers (26) (Appendix 1 Table 1). Each individual mouse's hybrid index (HI) was estimated as the proportion of M. musculus musculus alleles. We considered all mice with HIs <0.5 as M. musculus domesticus-like and those with HIs >0.5 as M. musculus musculus-like.

LCMV Serologic and Molecular Screening
We screened 291 blood plasma samples collected from 100 localities during 2008-2011 for LCMV antibodies by using the ELISA kit IM-698 C-EB (Xpress-Bio, https://xpressbio.com). We used 100 μL of 1:50 diluted serum for the reaction according to the manufacturer's instructions. In addition, we extracted RNA from 616 spleen or salivary gland samples by using RNeasy Mini kit (QIAGEN, https://www.qiagen.com). We reverse-transcribed the RNA samples collected in 2008-2013 by using the Applied Biosystems High-Capacity RNA-to-cDNA Kit (Ther-moFisher Scientific, https://www.thermofisher. com) in 10 μL final volume. We screened for LCMV by targeting a 340-nt fragment of the large gene by using primers from Vieth et al. (27), because these primers detected LCMV in a previous study (28). Samples were screened with the Multiplex PCR kit (QIAGEN) in a final volume of 15 μL by using 2 μL of cDNA and following the manufacturer's instructions. To increase assay sensitivity, we also designed primers for a nested PCR assay on the basis of LCMV sequences available in GenBank and targeting 442 nt in a part of the large gene partially overlapping with the region described previously. We tested 96 samples with both assays and results showed the same number of positive samples. However, the first assay (i.e., Vieth et al. primers) showed higher sensitivity (stronger band in 1.5% agarose gels); therefore, we selected that assay to screen the complete dataset. However, we used the second assay for Sanger sequencing of all positive samples to obtain longer final large fragment (659-665 nt resulting from merging both assay outputs). We  genes (Appendix 2 Table, https://wwwnc.cdc.gov/ EID/article/27/10/21-0224-App2.pdf). We purified PCR products and Sanger sequenced in both directions by using Eurofins Genomics (https:// eurofinsgenomics.com).

Whole-Genome Sequencing and Assembly of LCMV Viruses
We selected 2 positive samples from localities 10 km apart: sample SK1042 from Kryry, Czech Republic (KRY1) and sample SK1194 from Nepomyšl, Czech Republic (NEPO1), for whole-genome sequencing. We extracted RNA from lung and liver specimens by using the viral enrichment protocol described in Goüy de Bellocq et al. (29). The cDNA synthesis, library preparation, and sequencing (BGI Genomics, https://www. bgi.com) were carried out as described in Goüy de Bellocq et al. (30). After read demultiplexing, quality filtering, and trimming, 48,209,592 paired-end reads were available for SK1042, and 39,228,040 pairedend reads were available for SK1194. We used only 10,000,000 paired-end reads for a de novo assembly by iterative mapping with Geneious Mapper in Geneious 11 (Geneious, https://www.geneious.com). We enriched for LCMV reads in silico by removing all reads that mapped to mouse reference genome GRCm38. The LCMV iterative mapping was seeded with the 340 nt of the large gene obtained by Sanger sequencing and a 74-nt sequence conserved among LCMV strains for the Z gene. For the small segment, we generated 2 small seed reference sequences of ≈150 nt in the glycoprotein and nucleoprotein by first mapping the paired-end reads to LCMV strain Traub (from M. musculus domesticus mice). We confirmed the sequence of the intergenic region of the large segment by Sanger sequencing designing primers in the neighboring coding regions (Appendix 2 Table). After assembly, we ensured the seeding had not influenced the output. Finally, as part of a viral metagenomic study of digestive tract samples taken from mice in the HMHZ (J. Goüy de Belloq, unpub. data), we detected 229-nt and 458nt contigs that matched via BLAST (https://blast.ncbi. nlm.nih.gov/Blast.cgi) with the large and glycoprotein gene of LCMV and Dandenong virus in a pooled sample of 3 mice coming from Buškovice (BUS2) collected in 2014. We included these 2 sequences in the current study.

Phylogenetic Analyses
LCMV nucleotide sequences were aligned with the sequence coding parts of the nucleoprotein, glycoprotein, and large genes of other strains available in GenBank (Appendix 1  data) suggest that LCMV is locally endemic in M. musculus musculus mouse territory, persisting within farms over several years.

Characterization of the Full Genomes of LCMV from the Czech Republic
We obtained LCMV whole-genome sequences from 2 mice samples. Because the partial large sequences of the 4 samples from NEPO1 were identical, we characterized the genome of only 1 LCMV sample (SK1194

Phylogenetic Analysis
We analyzed the large, glycoprotein, and nucleoprotein genes separately and highlighted the position of the new LCMV variants found in M. musculus musculus mice from the Czech Republic and of the variants known to have been isolated from wild M. musculus domesticus mice. For the large nucleotide tree (Figure 3), the topology of the phylogeny is similar to that of Albariño  In the phylogenetic tree constructed on the basis of large amino acid sequences, the position of lineage III is again well supported but is basal to lineages I and II, both with highly supported monophyly (PP = 1) (Appendix 2 Figure, panel A).
The phylogenetic position of the sequences from Czech Republic M. musculus musculus and wild M. musculus domesticus mice in our nucleoprotein and glycoprotein gene trees corresponds to that in the large gene tree. An additional clade, clade IV, is composed of strains isolated from the woodmouse (Apodemus sylvaticus). All 4 glycoprotein lineages based on amino acid sequences were highly supported (PP = 1) (Appendix 2 Figure, panel B), whereas the phylogenetic signal at the nucleotide level seems to be compromised by homoplasy, resulting in trichotomy between lineages I, II, and III (Figure 4). A similar pattern can be seen in the phylogenetic trees based on the nucleoprotein gene but with low support. Phylogenetic relationships between lineages are not resolved, demonstrating differences with regard to the type of data. The basal position of lineage IV (woodmouse) to other lineages is well supported (PP = 1) on the basis of amino acid sequences (Appendix 2 Figure, panel C). By contrast, nucleotide sequences show lineage IV as sister group to lineage I (PP = 0.92) and lineage III clustering with lineage II (PP = 1), whereas the strain from Bulgaria is basal to all other ingroup lineages (PP = 1), suggesting that homoplasy at the nucleic acid level affects the phylogenetic signal ( Figure 5).

Discussion
We found LCMV at low prevalence in wild mice in central Europe, and all genetically confirmed cases clustered within a small geographic region in the M. musculus musculus mouse side of the HMHZ. This low prevalence prevents direct inference of the zone as a barrier to LCMV exchange between the mouse subspecies in nature. However, our phylogenetic analyses, which included new LCMV variants from the Czech Republic, 3 variants sequenced from wild M. musculus domesticus mice, other LCMV variants sequenced during the last decade, and supplemented with published data, support the hypothesis that LCMV lineage I harbors viruses originating from M. musculus domesticus mice and lineage II includes viruses primarily found in M. musculus musculus mice. The low prevalence of LCMV observed in central Europe is not uncommon. In wild mice, this prevalence has been shown to be variable, ranging from 0 to 25% (2), but most studies have reported low prevalence and patchy distribution. For example, Ackermann et al. (34) found an overall prevalence of 3% in wild mice from Germany, with 65 LCMVpositive specimens from 44 localities, but despite extensive sampling efforts in Bavaria as a whole (380 mouse samples over 70,000 km 2 ), no LCMV-positive mice were found there (35). We also failed to detect any positive LCMV samples in Bavaria (M. musculus domesticus mouse region). The low prevalence of LCMV is comparable to other mammarenaviruses (e.g., Gairo virus and Morogoro virus in Mastomys natalensis mice in Tanzania) (13,15).
We reported LCMV infection in Buškovice in 2008 and 2014; however, we were unable to demonstrate genetic turnover during that period. Commensal mouse populations are usually structured to local Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 27, No. 10, October 2021 subpopulations or demes, with a dispersal scale of ≈1 km 2 (36,37). Because LCMV can spread both horizontally and vertically, maintenance of the virus within a deme over several years seems plausible. Whether LCMV variants are still present in the 12 km 2 area is not certain. If so, targeted rodent control measures could feasibly decrease or eliminate LCMV risk for humans in this geographic area.
Albariño et al. (16) described 4 main LCMV lineages. Our results suggest that >3 of these lineages correspond to different host subspecies: lineage I to M. musculus domesticus, lineage II to M. musculus musculus, and lineage IV to Apodemus sylvaticus. We make no claim regarding the origin of lineage III, a single isolate from a human in Georgia (USA) (i.e., theoretically M. musculus domesticus mouse territory). We suggest more highly divergent lineages are likely to be discovered corresponding to rodent species, subspecies, and cryptic taxa. A new LCMV strain was recently reported from human serum in southern Iraq (38), but its phylogenetic position cannot be resolved; only a short fragment of the large gene (395 nt) is available in GenBank. This new LCMV strain is likely to cluster in clade I because M. musculus domesticus is the expected house mouse subspecies in southern Iraq (39)(40)(41). Uncertainty persists with respect to 4 LCMV strains clustered within lineage I of expected M. musculus domesticus mouse origin; JX14, JX4, and JX31 were isolated from ticks in 2015 from a coastal area in Jinxin, Jilin Province, northeastern China, and strain OQ28 was sequenced in 1990 from a wild mouse (M. musculus) captured in Osaka, Japan (42,43). In both regions, mice of subspecies other than M. musculus domesticus were reported. M. musculus musculus mice occur in northern China (44), whereas in Japan, mice are generally identified as M. musculus castaneus or M. musculus molossinus (45). However, the M. musculus domesticus mouse is known to be a successful invasive species because of ancient and recent human mobility, and its introduction to new areas is regularly reported, particularly in port cities, coastal areas, and islands (6). This expansion might explain the presence of M. musculus domesticus LCMV strains in Osaka and Jinxin, both coastal areas. LCMV can take a severe toll on human health, particularly in immunosuppressed persons. Cases of death after organ transplant have been reported involving strains from both lineages I and II (3,18,46). Takagi et al. (41) showed that 3 LCMV strains-OQ28, WE, and BRC-differ in pathogenicity in mice, concluding that strains OQ28 and BRC were genetically classified within the same cluster but exhibited very different pathogenicity. In this study, we demonstrate that the OQ28 strain clusters to M. musculus domesticus lineage I and the BRC strain clusters to M. musculus musculus lineage II; thus, we propose the 2 lineag-es have different host origins. From this perspective, the differences observed in strain pathogenicity by Takagi et al. (41) seem less surprising. Nevertheless, the variation of pathogenicity of LCMV strains corresponding to other host taxa is currently unknown.
In conclusion, our results suggest that the evolutionary diversity of LCMV might reflect rodent expansion history. When a human LCMV infection is diagnosed, sampling efforts should be applied to any synanthropic rodents. This effort could help clarify LCMV evolutionary history and elucidate whether different lineages differ in their spillover ability. Figure 5. Phylogenetic analysis performed on nucleic acid sequences of nucleoprotein gene of lymphocytic choriomeningitis virus (LCMV) sequences using Bayesian inference. Bayesian posterior probabilities were used to assess node support. Lunk virus from Mus minutoides (Africa) was used as outgroup. All sequences obtained in this study were submitted to GenBank (accession numbers: MZ568450-7, MZ558311-3, MZ568449). Names of LCMV strains are composed of GenBank number, strain name, host species, and place and country of origin (if known) or isolation. Country code is defined as ISO code (https://countrycode.org). Colors indicate LCMV strains isolated from wild rodents where there is a match between expected mouse subspecies on the basis of geographic region and sampling area: blue, Mus musculus domesticus; red, M. musculus musculus. Arrows indicate known origin of mice subspecies on the basis of genetic data, asterix indicates LCMV strains from this study, and lineages are indicated by roman numerals. LCMV strains isolated from Apodemus sylvaticus are indicated in green (lineage IV). Scale bar indicates nucleotide substitutions per site. Mmd, M. musculus domesticus; Mmm, M. musculus musculus; Mmm_lab, laboratory mouse strain derived from M. musculus musculus; Mm_lab, laboratory mouse strain; Mm_sp, Mus musculus spp.
Asworth for the warm welcoming of A. Fornuskova at the University of Edinburgh during her internship and introducing her to the world of genomic analyses. We are indebted to Anne Lavergne for providing us genetic information of positive mouse samples from French Guiana and to Vladislav Vergilov for his assistance with translating a Bulgarian article and hence with locating the geographic origin of LCMV-Bulgaria. This work was supported by the Czech Science Foundation (grant no. 16-20049S)

About the Author
Dr. Fornůsková is a research assistant at the Institute of Vertebrate Biology, Czech Academy of Sciences, Czech Republic. Her research is focused on host-pathogen interactions with special attention on small mammals as reservoir hosts.