Multiplex Genetic Engineering Exploiting Pyrimidine Salvage Pathway-Based Endogenous Counterselectable Markers.

This work reports the discovery of a novel genetic toolbox comprising multiple, endogenous selectable markers for targeted genomic insertions of DNAs of interest (DOIs). Marker genes encode proteins involved in 5-fluorocytosine uptake and pyrimidine salvage activities mediating 5-fluorocytosine deamination as well as 5-fluorouracil phosphoribosylation. The requirement for their genomic replacement by DOIs to confer 5-fluorocytosine or 5-fluorouracil resistance for transformation selection enforces site-specific integrations. Due to the fact that the described markers are endogenously encoded, there is no necessity for the exogenous introduction of commonly employed markers such as auxotrophy-complementing genes or antibiotic resistance cassettes. Importantly, inactivation of the described marker genes had no adverse effects on nutrient requirements, growth, or virulence of the human pathogen Aspergillus fumigatus. Given the limited number and distinct types of selectable markers available for the genetic manipulation of prototrophic strains such as wild-type strains, we anticipate that the proposed methodology will significantly advance genetic as well as metabolic engineering of fungal species.

G enetic engineering commonly involves the introduction of DNA of interest (DOI) into a target cell, followed by its integration into the genome. However, the efficiency of such genetic transformations is typically very limited. Therefore, selection of successfully modified cells usually involves cotransformation of selectable marker genes together with the DOI to allow growth under selective conditions. Widely used marker cassettes either compensate for the inability to synthesize vital metabolites (auxotrophic selection markers) or confer resistance to growth inhibitory compounds such as antibiotics (dominant selectable markers) (1). While auxotrophic markers are restricted to auxotrophic recipients, dominant selectable markers can be used for virtually any prototrophic recipient cell that is susceptible to the antibiotic used for selection. Hence, dominant selectable markers play a crucial role in the genetic manipulation of most wild-type cells. Examples of selectable markers commonly employed for the genetic engineering of fungi include genes conferring resistance to hygromycin B, phleomycin, pyrithiamine, kanamycin, and nourseothricin (2)(3)(4). Their expression allows growth in the presence of the corresponding antibiotic, which classifies them as positive selectable markers. Negative selectable markers, in contrast, inhibit growth of the target cells during selective conditions. In addition to the herpes simplex virus 1 thymidine kinase gene (5,6), genes encoding cytosine deaminase (CD) and uracil phosphoribosyltransferase (UPRT) have been employed as negative selectable markers in diverse organisms (7)(8)(9)(10)(11). The presence of functional gene copies providing CD and UPRT activity renders host cells susceptible to prodrugs 5-fluorocytosine (5FC) and 5-fluorouracil (5FU). CD and UPRT, both enzymes of the pyrimidine salvage pathway, convert 5FC and 5FU into 5-fluorouridine monophosphate (5FUMP) (Fig. 1a). Further metabolization of 5FUMP into toxic ribo-and deoxyribonucleotides blocks cellular growth (12).
Here, we characterized the pyrimidine salvage pathway in the human fungal pathogen Aspergillus fumigatus and present its application for fungal genetic engineering. In the described technology, endogenous genes encoding 5FC uptake, CD, and UPRT serve as counterselectable markers for targeted, genomic introduction of multiple DOIs. Homologous recombination-driven replacement of marker genes by DOIs results in their inactivation, which can be selected via 5FC/5FU resistance. In addition to the individual use (e.g., integration of reporter cassettes as well as the 17-kb penicillin biosynthetic cluster), the potential sequential use of the three loci is demonstrated by the insertion of three different fluorescent-protein-encoding genes for multicolor imaging of three cellular compartments. Demonstrating its versatile applicability, the described technology was implemented in the industrial work-horse Penicillium chrysogenum and the plant pathogen Fusarium oxysporum.

RESULTS
Cytosine deaminase FcyA and uracil phosphoribosyltransferase Uprt are crucial for the metabolic activation of 5FC in Aspergillus fumigatus. Metabolization of 5FC has been well-studied in the model yeast Saccharomyces cerevisiae: 5FC is converted by the CD Fcy1p to 5FU (13,14) and subsequently phosphoribosylated to 5FUMP by the UPRT Fur1p (15). Inactivation of each of these steps resulted in 5FC resistance, whereby inactivation of Fur1p also conferred 5FU resistance (15). Orthologous proteins from S. cerevisiae (Fcy2p), A. fumigatus, and Aspergillus nidulans (FcyB) have been identified as major cellular 5FC importers (16)(17)(18).
Among other fungal species, A. fumigatus is susceptible to 5FC (19,20) and is therefore anticipated to harbor genes encoding CD and UPRT activities in addition to 5FC uptake. BLASTP-based in silico predictions revealed A. fumigatus FcyA (AFUB_005410) and Uprt (AFUB_053020) as putative orthologs of S. cerevisiae Fcy1p and Fur1p, respectively. To analyze their role in 5FC as well as 5FU metabolism, we inactivated fcyA and uprt in the A. fumigatus strain A1160Pϩ (21), termed wild type (wt) here, using hygromycin and phleomycin resistance-based deletion cassettes. Due to the interdependency of 5FC activity and environmental pH (16,19), we investigated the contribution of both enzymes, as well as FcyB, to 5FC and 5FU activity at both pH 5 and pH 7.
Plate growth-based susceptibility testing revealed that, similar to previous work (16), 5FC levels Ն 10 g/ml blocked wt growth at pH 5, while 100 g/ml 5FC was required at pH 7 (Fig. 1b). Although FcyB is the major 5FC uptake protein, at 100 g/ml 5FC, the ΔfcyB mutant strain was not able to grow at pH 5 and showed severe growth inhibition at pH 7. In contrast to the ΔfcyB mutant, the ΔfcyA and Δuprt mutants displayed full resistance to 5FC up to 100 g/ml, regardless of the pH. Furthermore, 100 g/ml 5FU (a) After uptake, 5FC is converted to 5FU by the enzyme cytosine deaminase (CD). Subsequently, 5FU is phosphoribosylated to 5FUMP by uracil phosphoribosyltransferase (UPRT). 5FUMP is further metabolized into RNA or DNA nucleotides that interfere with DNA, RNA, as well as protein metabolism (12). (b) Inactivation of genes encoding components involved in uptake (ΔfcyB), CD (ΔfcyA), and UPRT (Δuprt) activity in A. fumigatus leads to different degrees of 5FC and 5FU resistance. For plate growth-based susceptibility testing, strains were point inoculated on solid AMM containing different levels of 5FC and 5FU. Images were acquired after 48 h of incubation at 37°C. Dashed lines indicate several enzymatic steps.
Genetic Engineering Exploiting Endogenous Marker Genes ® blocked growth of wt, ΔfcyB, and ΔfcyA strains at pH 5 as well as pH 7, while the Δuprt mutant was resistant.
Our data confirm the role of FcyB as a major 5FC cellular importer but indicate the presence of additional uptake mechanisms. Similar to the orthologous proteins in S. cerevisiae, our findings reveal the essential roles of FcyA and Uprt for 5FC activity and of Uprt for metabolic activation of 5FU in A. fumigatus.
The fcyB, fcyA, and uprt loci can be used for 5FC/5FU-based transformation selection. Lack of FcyB, FcyA (CD activity), or Uprt (UPRT activity) confers resistance to 5FC (ΔfcyB, ΔfcyA, and Δuprt mutants) or 5FU (Δuprt mutant) (Fig. 1b), which suggested the utilization of the genes encoding these proteins as counterselectable markers for positive selection of cells with targeted integration of DOIs. Moreover, the different degrees in 5FC resistance observed for ΔfcyB and ΔfcyA indicated that 5FC can be used for selection of loss of FcyB at low 5FC concentrations (10 g/ml) and loss of FcyA at high 5FC levels (100 g/ml) (Fig. 1b). Selection for loss of Uprt was conducted with 100 g/ml 5FU.
For proof of principle, both green fluorescent protein (GFP, the sGFP[S65T] variant, here abbreviated as sGFP) (37) and ␤-galactosidase (LacZ) expression cassettes were used to replace fcyB, fcyA, as well as uprt. To achieve homologous recombinationmediated replacement of these loci with the reporter cassettes, approximately 1-kb 5= and 3= nontranslated regions (NTRs) of the respective gene were linked to each cassette via fusion PCR (Fig. 2a). The resulting knock-in constructs were transformed into wt protoplasts, which underwent selection for resistance to 5FC and 5FU (see above; Fig. 2b). Southern blot analyses confirmed site-specific integration of the DOIs into each of the three loci (see Fig. S1 in the supplemental material). In agreement, all knock-in strains displayed resistance phenotypes according to the respective mutation in the pyrimidine salvage pathway (compare Fig. 1b and Fig. S2). Fluorescence imaging and ␤-galactosidase staining confirmed the functionality of the knock-in cassettes (Fig. 2c). To analyze the transformation efficiency using individual selectable marker genes, we employed the corresponding LacZ knock-in constructs for each locus. For fcyB, fcyA, and uprt, 10, 27, and 13 transformants, respectively, were recovered on the corresponding selective media (Fig. S3). Out of these 10 (100%), 26 (97%) and 12 (92%) displayed LacZ activity. Southern blot analysis of 10 LacZ-positive transformants for each locus confirmed correct integrations.
Taken together, 5FC-and 5FU-mediated selection allowed replacement of each of the three loci by either a GFP or lacZ expression cassette, which demonstrates the suitability of fcyB, fcyA, and uprt as selectable markers for targeted, integrative transformation in A. fumigatus. fcyB, fcyA, and uprt can be consecutively used for multiple genomic integrations. Due to the fact that inactivation of fcyB and fcyA led to different levels of resistance to 5FC and inactivation of uprt caused resistance to 5FU (Fig. 1b), we investigated whether these marker genes can be sequentially employed for transformation selection in a single strain. As an exemplary application, we aimed to generate a strain expressing three fluorescent proteins for multicolor imaging: GFP, red fluorescent protein (RFP, mKate2) and blue fluorescent protein (BFP, mTagBFP2).
The strategy pursued and order of markers used for selection was based on the following results (Fig. 1b). (i) In contrast to the wt, the ΔfcyB mutant can grow in the presence of 10 g/ml 5FC at pH 5. (ii) In contrast to ΔfcyB, ΔfcyA can grow at 100 g/ml 5FC, which allows discrimination of ΔfcyB ΔfcyA from ΔfcyB. (iii) ΔfcyB and ΔfcyA are still able to import and metabolize 5FU, which is expected to allow discrimination of ΔfcyB ΔfcyA and ΔfcyB ΔfcyA Δuprt in the presence of 100 g/ml 5FU.
In the first step, we integrated an expression cassette encoding mKate2 carrying a C-terminal peroxisomal targeting sequence (PTS1, tripeptide SKL) (22) in the fcyB locus, yielding strain RFP PER (ΔfcyB::mKate2 PER ). In this strain, a second expression cassette encoding sGFP containing an N-terminal mitochondrial targeting sequence derived from citrate synthase (23) was targeted into the fcyA locus, yielding RFP PER GFP MIT strain (ΔfcyB::mKate2 PER ΔfcyA::sGFP MIT ). In the last step, an expression cassette encoding mTagBFP2 with expected cytoplasmic localization was integrated into the uprt locus yielding strain RFP PER GFP MIT BFP CYT (ΔfcyB::mKate2 PER ΔfcyA::sGFP MIT Δuprt:: mTagBFP2 CYT ). Multicolor laser scanning confocal microscopy confirmed expression of all three fluorescent proteins in RFP PER GFP MIT BFP CYT and localization to distinct subcellular compartments ( Fig. 3 and Fig. S4). A noteworthy finding was that the simultaneous lack of FcyB, FcyA, and Uprt (RFP PER GFP MIT BFP CYT strain) affected neither growth (Fig. 3), virulence in a murine model of invasive aspergillosis, nor adaptation to various stress environments (Fig. S5).
Collectively, these data demonstrate the feasibility of sequential use of endogenously encoded counterselectable markers fcyB, fcyA, and uprt for integration of up to three DOIs without adverse effects on A. fumigatus growth and virulence.
Endogenous counterselectable markers can be used for the integration of biotechnologically relevant, large DNA fragments. Fungi play important roles as cell factories in food industry as well as medicine. Pursuing a potential biotechnological approach using the described selection method, we tested whether we can integrate the 17-kb penicillin biosynthetic cluster (PcCluster) of P. chrysogenum into the genome of A. fumigatus. Therefore, a knock-in plasmid was constructed comprising the PcCluster as well as 5= and 3= fcyB NTRs. Linearization of the plasmid with PmeI allowed double crossover homologous recombination-mediated replacement of fcyB by the PcCluster, as illustrated in Fig. 4a and Fig. S6a.
Subsequent to transformation of this construct in the wt (selection, 10 g/ml 5FC, pH 5), we validated its site-specific integration at the fcyB locus (strain fcyB PENG ; Fig. S1). Northern blot analysis confirmed expression of penicillin biosynthetic genes pcbAB, pcbC, and penDE in three transformants (Fig. 4b). Concomitant to successful expression of this heterologous gene cluster in A. fumigatus, we detected penicillin G and its degradation product penillic acid in culture supernatants using nano-scale liquid chromatography (nanoLC) mass spectrometry (Fig. 4c). Both substances have the same molecular weight of 335.1060 g/mol but can be differentiated due to structure-specific fragmentation during tandem mass spectrometry (MS/MS) (Fig. S6b).
Taken together, these results demonstrate that pyrimidine salvage-based selectable markers can be applied for the integration of large-size DNA fragments, including whole gene clusters.
Pyrimidine salvage-based selectable markers can be utilized in Penicillium chrysogenum and Fusarium oxysporum. To identify encoded CD and UPRT activities in other fungal species, we searched for A. fumigatus FcyA and Uprt orthologs in biotechnologically (Aspergillus niger, Aspergillus oryzae, P. chrysogenum, Komagataella phaffii [previously Pichia pastoris], S. cerevisiae, and Trichoderma reesei) and pathologically relevant fungal species (Candida albicans, Cryptococcus neoformans, and F. oxysporum). In silico inspection of the annotated genomes of these species revealed that 8 out of the 10 species analyzed harbor a putative ortholog of A. fumigatus FcyA and that all species possess a putative ortholog of A. fumigatus Uprt with overall sequence identities of Ն40% (see Table S1A in the supplemental material). Notably, all species encoding an FcyA ortholog were also found to encode an FcyB ortholog. The genetic coupling of these two features might indicate that their main function is the utilization of extracellular pyrimidines.
To confirm CD and UPRT activities in the species analyzed, we monitored 5FC and 5FU susceptibility profiles following a broth microdilution-based method according to EUCAST (24). In agreement with our homology search, species with predicted FcyA orthologs (CD activity) were susceptible to 5FC, while those without these orthologs were resistant to the drug (Table S1B). All strains were susceptible to 5FU, which is in accordance with in silico-predicted Uprt orthologs (UPRT activity).
In the next step, we tested the applicability of the described selection strategy in P. chrysogenum and F. oxysporum. In agreement with the genomic data and 5FC/5FU susceptibility, P. chrysogenum expresses both CD (P. chrysogenum FcyA [Pc-FcyA], EN45_039280) and UPRT (Pc-Uprt, EN45_060980), while F. oxysporum lacks CD but expresses UPRT (F. oxysporum Uprt [Fo-Uprt], FOXG_01418). Employing the same protocol as used for A. fumigatus enabled the integration of GFP expression cassettes flanked by the 5= and 3= NTRs of the respective P. chrysogenum genes in both the Pc-fcyA and the Pc-uprt loci. In F. oxysporum, the same strategy enabled the targeting of a GFP expression cassette at the Fo-uprt locus. The presence and functionality of the GFP reporters were visualized as described above (Fig. 5). As observed for A. fumigatus knock-in mutants, the resistance profiles of P. chrysogenum and F. oxysporum knock-in strains were in accordance with the absence of individual salvage activities. Genetic Engineering Exploiting Endogenous Marker Genes ® Taken together, these data indicate high evolutionary conservation of the described pyrimidine salvage enzymes within the fungal clade and demonstrate the suitability of the loci encoding these enzymes as markers for transformation selection in P. chrysogenum and F. oxysporum.

DISCUSSION
In this study, we report the characterization and exploitation of multiple, endogenously encoded selectable markers for targeted genetic engineering. The marker genes encode activities mediating 5FC uptake (fcyB) and metabolization (fcyA and uprt) of 5FC or 5FU into cell-toxic nucleotides (Fig. 1a) (12). Genomic replacement of these genes by DOIs results in loss of the corresponding activities and can therefore be selected via 5FC or 5FU resistance. We validated the applicability of orthologous marker genes encoding pyrimidine salvage activities for their use as selectable markers in three fungal species (A. fumigatus, P. chrysogenum, and F. oxysporum) by targeted insertion of various fluorescent or enzymatic reporter genes (Fig. 2, 3, and 5). Mimicking a biotechnological application, we further introduced the 17-kb penicillin biosynthetic gene cluster from P. chrysogenum into A. fumigatus, which naturally is not capable of producing this secondary metabolite. In addition to their single use, we demonstrate the possibility of consecutive use of these markers in a single strain by generating an A. fumigatus mutant expressing three fluorescent proteins (mKate2, sGFP, and mTagBFP2) targeted to different subcellular compartments ( Fig. 3a; see also Fig. S4 in the supplemental material). Importantly, these endogenous genes can be used in addition to traditional selectable markers, and the absence of all three genes (ΔfcyB ΔfcyA Δuprt) affected neither growth nor virulence (Fig. 3b and Fig. S5) in A. fumigatus, which represents a prerequisite for downstream use of engineered strains. Similarly to fcyB, fcyA, and uprt, the orotidine-5=-decarboxylase-encoding gene pyrG illustrates a counterselectable marker, since loss of the corresponding gene function confers resistance to 5=fluoroorotic acid. However, PyrG is essential for the de novo biosynthesis of pyrimidines, and its loss renders cells uracil auxotrophic (25,26), which represents a major drawback in comparison to the markers described here.
Inspection of genomic sequences combined with 5FC/5FU susceptibility profiling of different fungal species playing important roles in biotechnology, medicine, or agriculture revealed the presence of uprt orthologs in all 10 species analyzed and fcyB as well as fcyA orthologs in 8 out of 10 species (see Table S1 in the supplemental material), indicating the broad applicability of the proposed selection method. Most likely, the number of endogenous counterselectable markers for DOI integration allows expansion beyond the genes characterized here. For instance, genes coding for components involved in 5FU uptake, such as orthologs of A. nidulans furD or S. cerevisiae FUR4 (27,28), represent potential candidates.
Employing pyrimidine salvage pathway-based endogenous markers, virtually any DOI can be inserted into a specific site in the target genome without apparent effects on growth phenotypes, demonstrating their huge potential for diverse applications in basic as well as applied research. The sequential use of marker genes enables the establishment of modular toolboxes, e.g., allowing site-directed insertion of multiple reporter genes for colocalization studies. Furthermore, the possibility of equipping strains with multiple DNA building blocks makes this technology particularly attractive for synthetic biology. Notably, the fact that marker genes have to be inactivated to allow selection based on 5FC/5FU resistance enforces targeted integration of DOIs. As the strategy described here avoids the use of foreign selection markers, the described genetic toolset further enables so-called "self-cloning." Strains engineered this way are not considered genetically modified organisms (GMO) in some countries (29,30). For this, only endogenously encoded DOIs can be used; examples include the increase of gene dosage or engineering of sequences such as promoter swapping. A further advantage of endogenous markers is the dispensability of antibiotic resistance genes, the use of which is generally discouraged in the manufacturing of medical or foodrelated products (31,32), as it may promote horizontal transfer of resistance genes.
In summary, we characterized a powerful genetic toolbox enabling multiple, targeted genomic insertions of DOIs. Evolutionary conservation of the pyrimidine salvage pathway suggests broad applicability of the described marker genes. Thus, we anticipate that this technology will significantly advance genetic and metabolic engineering of diverse organisms.

MATERIALS AND METHODS
Growth conditions and fungal transformation. Plate growth assays analyzing A. fumigatus and P. chrysogenum were conducted using solid Aspergillus minimal medium (AMM) (ammonium tartrate was used as nitrogen source, glucose as carbon source) (33), and for F. oxysporum, solid PDA was employed. Therefore, 10 4 spores of each strain were point inoculated on agar plates. Low-pH medium contained 100 mM citrate buffer (pH 5), and neutral-pH medium contained 100 mM morpholinepropanesulfonic acid (MOPS) buffer (pH 7). If not stated in the text or images, plate growth assays were performed at pH 5. For strains carrying xylP promoter (PxylP)-driven reporter genes (sGFP, mKate2 PER , sGFP MIT , lacZ), the medium was supplemented with 0.5% xylose to induce gene expression. Environmental and drugrelated stress was assayed by either adding high concentrations of H 2 O 2 , copper, cobalt, or sorbitol, omitting iron or zinc micronutrient, incubating at 48°C, or supplementing medium with caspofungin, voriconazole, or amphotericin B (see Fig. S5 in the supplemental material).
For fungal manipulations, 2 g DNA of each construct was transformed into protoplasts of the respective recipient. For the regeneration of transformants, solid AMM (A. fumigatus and P. chrysogenum) or PDA (F. oxysporum) supplemented with 342 g/liter or 200 g/liter sucrose, respectively, were used. Selection procedures using conventional selectable marker genes (hph, ble) were conducted as described previously for A. fumigatus (16).
Deletion of A. fumigatus fcyA and uprt. The strains and primers used in this study are listed in Tables S2 and S3 in the supplemental material. Coding sequences of fcyA and uprt were disrupted in the wild type (wt) (A1160Pϩ) using hygromycin B and zeocin resistance cassettes, respectively. Therefore, deletion constructs comprising approximately 1 kb of 5= and 3= nontranslated regions (NTRs) linked to the central antibiotic resistance cassette were generated using fusion PCR as previously described (21). Correct integration of constructs was confirmed by Southern blot analyses (Fig. S1).

Generation of A. fumigatus knock-in strains.
Knock-in constructs for A. fumigatus fcyB, fcyA, and uprt loci, P. chrysogenum loci Pc-fcyA and Pc-uprt, as well as F. oxysporum Fo-uprt were generated similarly to the gene deletion fragments described above using fusion PCR. Here, instead of the antibiotic resistance cassettes, DNAs of interest (DOIs) (for reporter templates, see also Fig. S7) were connected to approximately 1-kb 5= and 3= NTRs of the respective locus (Fig. 2a).
Images of RFP PER GFP MIT BFP CYT were taken using an HC PL APO CS2 63ϫ/1.30 glycerol objective on an SP8 confocal microscope (Leica Microsystems, Wetzlar, Germany) equipped with a 80-MHz pulsed white light laser (WLL) and a 405-nm CW diode laser (405-nm diode) according to Nyquist sampling.
Expression analysis of penicillin biosynthetic genes and detection of penicillin G in culture supernatants. Northern blot analysis was conducted as described previously, using digoxigenin-labeled probes (35).
To detect the potential production of penicillin G, strains were grown in AMM for 48 h at 25°C, and 2 ml of culture supernatant was extracted with 1 volume of butyl acetate. The organic phase was collected in a new reaction tube and dried using a centrifugal vacuum concentrator (speed-vac). Nano-scale liquid chromatography-mass spectrometry (nanoLC-MS)-based detection of penicillin G was conducted using an UltiMate 3000 nano-scale high-performance liquid chromatography (HPLC) system coupled to a Q Exactive HF mass spectrometer (Thermo Scientific, Bremen, Germany). The samples were separated on a homemade fritless fused-silica microcapillary column (100-m inner diameter [i.d.] by 280-m outer diameter [o.d.] by 19-cm length) packed with 2,4-m reversed-phase C18 material (Reprosil). Solvents for HPLC were 0.1% formic acid (solvent A) and 0.1% formic acid in 85% acetonitrile (solvent B). The gradient profile was as follows: 0 to 4 min, 4% solvent B; 4 to 57 min, 4 to 35% solvent B; 57 to 62 min, 35 to 100% solvent B; and 62 to 67 min, 100% solvent B. The flow rate was 300 nl/min.
Mass spectra were acquired in positive ion mode applying a precursor scan over the m/z range 50 to 500 in the FT analyzer. The ions at m/z ϭ 335.1060 were selected from this precursor scan for tandem MS (MS/MS) fragmentation in the linear ion trap.
Murine infection model. Specific-pathogen-free female outbred CD-1 mice (18 to 20 g; 6 to 8 weeks old; Charles River, Germany) were housed under standard conditions in individually ventilated cages and supplied with normal mouse chow and water ad libitum. All animals were cared for in accordance with the European animal welfare regulations, and experiments were approved by the responsible federal/ state authority and ethics committee in accordance with the German animal welfare act (permit no. 03-027/16). Mice were immunosuppressed with cortisone acetate and intranasally infected with 2 ϫ 10 5 conidia in 20 l phosphate-buffered saline (PBS) as described before (36). Infected animals were monitored twice daily and humanely sacrificed if moribund (defined by a score, including weight loss, piloerection, behavior, and respiratory symptoms).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.   A patent related to this work has been filed (38). We declare that we have no competing nonfinancial interests.