Multi-parametric analysis of host response to murine CMV in MHC class I disparate mice reveals primacy of D^k-licensed Ly49G₂+ NK cells in viral control^,

Abstract

MHC I D^k and Ly49G₂ inhibitory receptor-expressing NK cells are essential to murine (M)CMV⁵ resistance in MA/My mice. Without D^k, Ly49G₂+ NK cells in C57L mice fail to protect against MCMV infection. As a cognate ligand of Ly49G₂, D^k licenses Ly49G₂+ NK cells for effector activity. These data suggested that D^k-licensed Ly49G₂+ NK cells might recognize and control MCMV infection. However, a role for licensed NK cells in viral immunity is uncertain. We combined classical genetics with flow cytometry to visualize the host response to MCMV. Immune cells collected from individuals of a diverse cohort of MA/MyxC57L offspring segregating D^k were examined before and after infection, including Ly49+ NK subsets, receptor expression features and other phenotypic traits. To identify critical NK cell features, automated analysis of 110 traits was performed in R using Pearson’s correlation followed with a Bonferroni correction for multiple tests. Hierarchical clustering of trait-associations and principal component analyses were used to discern shared immune response and genetic relationships. The results demonstrate that Ly49G₂ expression on naïve blood NK cells was predictive of MCMV resistance. However, rapid Ly49G₂+ NK cell expansion following viral exposure selectively occurred in D^k offspring; this response was more highly correlated to MCMV control than all other immune cell features. We infer that D^k-licensed Ly49G₂+ NK cells efficiently detected missing-self MHC cues on viral targets, which elicited cellular expansion and target cell killing. MHC polymorphism therefore regulates licensing and detection of viral targets by distinct subsets of NK cells required in innate viral control.

INTRODUCTION

NK cells confer essential innate immunity and host defense against viral infection (1). Human NK cell deficiency results in severe susceptibility to viral infections, especially herpesviruses (2, 3). Likewise, NK cell deficiency in the mouse due to selective immunodepletion, mAb-blocking or genetic mutation of NK cells results in susceptibility and much less efficient viral control. NK cell detection and killing of infected cells requires highly efficient and precise recognition of viral targets. Regulation of NK cell cytokines and cytotoxicity involves numerous different cell surface stimulatory and inhibitory receptors that are needed to scrutinize cellular targets for evidence of viral infection or transformation (4). Whereas the NK receptors are highly polymorphic, genetic approaches to investigate the mechanistic bases of NK-mediated resistance to infection and viral control have yielded significant insight.

Many NK cell receptors reportedly enhance recognition of viral targets and occasionally elicit viral control (5–14). Prominent examples are Ly49H and Ly49P, two stimulatory receptors that are known to bind the viral protein m157 and MHC I D^k on MCMV-infected targets, respectively (9, 15–18). Inhibitory receptors, on the other hand, bind class I MHC molecules as ligands and render NK cells self-tolerant by preventing lytic attack against autologous cells. Cellular changes that result in missing-self MHC I display, as sometimes occurs due to viral infection or cell transformation, also evoke NK detection and target cell lysis (19). Indeed, viruses have evolved intriguing and rather selective strategies to evade NK cell detection via missing-self class I MHC cues (6, 20–23).

Inhibitory receptors also serve to educate NK cells through ‘licensing’ or ‘arming’ (24, 25). Whereas licensed NK cells can be readily stimulated ex vivo via activating receptor cross-linkage, unlicensed NK cells display hyporesponsiveness. Less severe disease features due to chronic viral infection observed in patients with combined genes for inhibitory KIR and HLA class I molecules hints that human licensed NK cells might also contribute viral control (26, 27). In support of this concept are several human studies demonstrating expansion and/or activation of select inhibitory receptor-expressing NK cells following exposure to certain viral infections (28–30). The results suggest that self-licensed NK cells may be poised to efficiently respond to missing-self-MHC cues following viral exposure, though direct evidence of a role of human inhibitory NK receptors in viral control is lacking.

Genetic analysis of MCMV resistance in MA/My and C57L hybrid mice showed that NK cells with the inhibitory Ly49G₂ (G₂) receptor provide critical protection in offspring with MHC I D^k, a major MCMV resistance factor and G₂ licensing-ligand (31–34). Because D^k-resistance is heightened in C57L compared to MA/My (35), C57L alleles outside the MHC may enhance NK-mediated viral clearance. Altogether the results suggested that self-licensed NK cells, possibly regulated by background genetic effects, contribute to highly efficient detection and elimination of viral targets. However, several reports have shown that Ly49G₂+ NK cells in B6 mice rapidly expanded in response to proliferation cues (36) or exposure to one of several different viruses (37–39). Moreover, licensed Ly49C/I+ NK cells were since shown to diminish viral control in MCMV-infected B6 mice (40). Thus, the precise role of self-licensed NK cells and inhibitory receptors in viral control and clearance is still uncertain.

To address this question, we combined classical genetics with flow cytometry and examined distinct NK cell subset features before and after MCMV exposure to visualize the NK cell response to infection in a large cohort of MHC I D^k disparate mice. The results demonstrate that Ly49G₂+ NK cells underwent highly selective expansion in D^k mice, which directly corresponded with significant viral protection. We infer that licensed NK cells are essential to efficiently detect and then eliminate MCMV-infected targets; a response undeniably shaped by class I MHC polymorphism.

Materials and Methods

Mice

MA/My and C57L breeders were purchased from The Jackson Laboratory. C57L-derived strains known as R2, R2-NKC^het, R2-NKC^mamy, R7, R7- NKC^het, R7-NKC^mamy, R12 and Tg-D^k and MA/My-derived strains known as M.H2^b and M.Tg1 have been previously reported (32, 35). We bred and studied 38 (R7×M. H2^b)F₂, 62 (MA/My×C57L)F₂, 72 [(MA/My×C57L)F₁×M. H2^b]N₂ and 61 [(MA/My×C57L)F₁×C57L]N₂ hybrid offspring. Mice were managed with the Jackson Laboratory's Colony Management System (Version 4.1.2). All mice used in this study were maintained in a dedicated animal care facility under specific pathogen-free conditions and treated in accordance with the regulations and guidelines of the Animal Care and Use Committee of the University of Virginia.

Genotyping

Spleen and liver genomic (g)DNA samples were prepared using a Gentraprep kit. To ensure sample identity, all mice were 'footprinted' by genotyping MHC, NKC and chromosome 19 loci using gene-specific PCR and high-resolution melting (HRM) analysis as described (41) prior to genome-wide single nucleotide polymorphism (SNP) typing by DartMouse (Dartmouth, NH) using the Illumina medium density SNP panel. Footprint analysis was 100% concordant with the genome-wide SNP analysis for all mice (data not shown).

Antibodies, NK cell receptor analysis, and flow cytometry

mAbs and isotype controls used for flow cytometry were purchased from BD Biosciences, Biolegend, and R&D Systems and included anti-CD16/32 (2.4G2), Ly49G2 (4D11), Ly49I/U (14B11), CD3 (145-2C11), CD19 (6D5) and NKp46 (29A1.4). These were conjugated to FITC, PE, PerCP, PerCP-Cy5.5, APC, or biotin followed by streptavidin conjugated to allophycocyanin-Cy7 (Biolegend). Dead cells were excluded using a Live / Dead Fixable Violet Dead Cell stain kit (Invitrogen) with compensation based on the ArC Amine Reactive Compensation Bead Kit (Invitrogen). Compensation for tandem dyes on antibodies that stain rare cell subpopulations used the AbC anti-Rat/Hamster Bead Kit (Invitrogen) with uninfected blood leukocytes and single stain controls on spleen leukocytes after infection. Blood (4 d pre-infection) and spleen (3.5 d post-infection) leukocytes were stained as described previously (34). Live splenocytes were calculated after manual counting of cells excluding trypan blue. Flow cytometric analysis of stained cells was performed on a FACSCanto II (BD Biosciences). Data acquired using FACSDiva software (BD Biosciences) were analyzed with FlowJo software (Versions 8.0 and 9.2; Tree Star, Ashland, OR). NK cell subpopulation frequencies and NK receptor geometric means (MFI values) were calculated in FlowJo and exported into Excel tables.

Primary flow cytometry data files, grouped by experimental cohort, have been made publicly available through the FlowRepository.org (http://flowrepository.org/) as identified by the following public experiment tags: FR-FCM-ZZ4W, FR-FCM-ZZ4J, FR-FCM-ZZ4K, FR-FCM-ZZ4L, FR-FCM-ZZ4N, FR-FCM-ZZ4X, FR-FCM-ZZ4V, FR-FCM-ZZ5Z, FR-FCM-ZZ5Y, FR-FCM-ZZ52, FR-FCM-ZZ53, FR-FCM-ZZ54, FR-FCM-ZZ5C, FR-FCM-ZZ5B, FR-FCM-ZZ5D, FR-FCM-ZZ5E, FR-FCM-ZZ5F, FR-FCM-ZZ5H, FR-FCM-ZZ5W, FR-FCM-ZZ5J, FR-FCM-ZZ5K, FR-FCM-ZZ5L, FR-FCM-ZZ5M, FR-FCM-ZZ5N, FR-FCM-ZZ5X, FR-FCM-ZZ5P, FR-FCM-ZZ5S, FR-FCM-ZZ5R. The primary data include pre- and post-infection analyses for F₂ cohorts 1, 2, 4–6 and backcross cohorts 1–8. F₂ cohort 3 was excluded from the FACs analysis due to experimental error associated with loss of post-infection splenocyte samples. Note that public identification tags FR-FCM-ZZ5B and FR-FCM-ZZ5W include analysis of only one experimental animal. Individual sample details (e.g. source, treatment, mAb staining panel) are also available on selection of ‘Flow Sample / Specimen Details’.

MCMV infection and quantification

All mice were weighed and bled 4 d prior to i.p. MCMV infection (10⁵ PFU Smith strain salivary gland virus stock) and then weighed again 3.5 d after infection, just before euthanization. Approximately ¾ of each individual spleen was used for flow cytometric analysis. Small spleen and liver tissue fragments were processed for gDNA genotyping and quantitative real-time PCR (qPCR) analysis of MCMV genome level as described previously (42).

Data validation, analysis and statistics

To ensure flow cytometric data integrity, all FlowJo cell frequencies and NK receptor display values exported into Excel tables were verified against raw data. To further qualify the data set, ¼ of all mice, chosen at random, were re-typed for spleen MCMV and genetic footprint using qPCR and HRM, respectively. Another ¼ of randomly selected individual mice were retyped for liver MCMV. In addition, H-2 class I D exon 5 genotypes were verified using intron 8-specific PCR followed by HRM analysis (41). ‘Atypical’ mice were additionally screened with H-2 D intron 1- and 3-specific markers (41). All H-2 D gene makers were fully concordant in all mice.

Statistical analyses included paired and unpaired Student T tests, Pearson correlations and multiple linear regression tests performed using the R (versions 2.15.0 and 2.15.2) statistical computing environment with select plots drawn using ggplot2 (43) and Corrplot (44) packages. P-values were corrected for multiple comparisons using both the Benjamini-Hochberg false discovery rate (45) and the Bonferroni Correction.

RESULTS

Multigenic control of Ly49+ NK cells in C57L- and MA/My-derived strains

Strain-specific variance in viral control and splenocyte recovery following MCMV infection (35) prompted our genetic analysis of the NK cell response to MCMV. We first analyzed peripheral blood and spleen NK cells (NKp46+, CD3-, CD19-) from naïve mice with different MHC and NKC genotypes (Fig 1A and B, Table I). Ly49G₂+ (G₂+) and Ly49I/U+ (I/U+) NK cell subsets were analyzed with mAbs 4D11 and 14B11, as described (17, 32, 34, 46). Without a monospecific-staining reagent, Ly49P+ NK cells were not examined. As expected of D^k-licensed G₂+ NK cells (33, 34), each of our C57L-derived strains with D^k had less frequent G₂+ NK cells with significantly reduced G₂ receptor display intensity (MFI) in comparison to C57L (Fig 1C, Table I), which supports that the subset was licensed. The effect of D^k was specific to G₂+ NK cells as the frequency of I/U+ NK cells was unaffected in C57L- and M.H2^b-derived Tg-D^k mice (Table I, Supplemental Table I). Nonetheless, a reduction of I/U+ NK cells in R2 and R7, compared with C57L, revealed further MHC control of NK cell subsets (Table I, Supplemental Table I). The data confirmed that H-2^k regulates the homeostatic composition of NK subsets and receptor display features in mice with relevant Ly49 receptors, as shown previously for NK receptor expression in other strains (47, 48).

Figure 1. Genetic regulation of distinct subsets of Ly49+ NK cells in naïve and MCMV-infected C57L- and MA/My-derived congenic strains of mice.

A. The map depicts a 35-Mb genetic interval of chromosome 17 with low- (left) and higher-resolution (right) cross-over boundaries defined for the indicated MHC congenic strains. Key SNP markers used to genotype the strains and hybrid offspring are shown. Several MHC and non-MHC genes that reside in the genetic interval are also shown. B. Representative dot plots showing the gating strategy and NK cell frequencies for naive peripheral blood and MCMV-infected splenocytes. C. Representative dot plots for I/U+ and G₂+ subsets of live NK cells, gated as in (B), detected in naive blood and infected spleen cells of the indicated strains using mAbs 14B11 and 4D11, respectively.

Table I.

MHC- and NKC-dependent regulation of Ly49+ NK cell features in C57L-derived strains

			%G2+ NK cells		Ly49G2 MFI		%I/U+ NK cells		Ly49I/U MFI
Strain1–3	MHC4	NKC4	Blood	Spleen	Blood	Spleen	Blood	Spleen	Blood	Spleen
Naïve mice
C57L	H-2b	c57l	55.8	59.4	3116	3526	63.8	63.4	2556	2408
R25	Kk, Db	c57l	52.8	54.1†	2482	2511	49.6	48.4*	4522	4249†
R2-NKChet	Kk, Db	Het	58.8	60.2†	2592	2691	48.9***	47.1***	4147***	3908**†
R2-NKCm	Kk, Db	mamy	65.5	65.6†	2834	3089	46.4***	44.4***	3706*	3520*†
R12het	Kb, Dk/b	c57l	54.7	52.8	2435	2501	55.1**	53.7*	2057	1992
R12	Kb, Dk	c57l	48.3	48.1	1893*	2143	60.2	60.0	1922*	2150
Tg3-Dk	H-2b, Dk	c57l	46.0	46.2	1787***	2048*	59.8	60.6	2039	2172
Infected mice
C57L	H-2b	c57l	63.4	69.5	4271	4734	66.2	66.3	1368	1566
R12het	Kb, Dk/b	c57l	65.6	61.9	2469	2637	61.0	55.1	1238	1542
Tg3-Dk	H-2b, Dk	c57l	63.9	68.2	2147	1999	58.2	60.6	1326	1566

¹The data shown are the average values in naïve mice of the indicated strains based on analysis of 3 or more mice except for R2 (n=2). For infected mice, the average blood values are for 2 animals and average spleen values are for 3 mice except for R12 (n=2).

²Statistical analysis performed using Bonferroni-corrected Student t-test: *p_B≤0.05; **p_B≤0.01; ***p_B≤0.005 in comparison to C57L (naïve and infecteds analyzed separately). Blood and spleen NK cell features were not significantly different in any of the strains.

³T_g-D^k mice likely integrated four MHC I D^k genomic fragments [39] and display slightly higher D^k than R12 mice.

⁴MHC genotypes, including class I K and D genes, and NKC Ly49g genotypes were determined as described [39].

⁵Statistical analysis of NK cells collected from R2-derived NKC congenic strains performed using Bonferroni-corrected ANOVA. (^† p_B<0.05)

Likewise, NK receptor polymorphism is known to affect NK cell features and their role in MCMV resistance (49). Analysis of R2-NKC congenic strains revealed that both the frequency of G₂+ NK cells and I/U MFI were impacted by NK gene complex (NKC) polymorphism (Table I). Interestingly, a lower percentage of G₂+ NK cells in MA/My than R7-NKC^m mice suggested that a genetic factor(s) outside the MHC and NKC regions also shapes Ly49+ NK subsets and receptor display (Supplemental Table I). A difference in the frequency of I/U+ NK cells in M.H2^b, M.Tg1 and C57L mice further supports non-MHC, non-NKC genetic control of NK subset composition. Whereas D^k-mediated MCMV resistance has been shown more effective in C57L-derived than MA/My mice (35), the above results hinted that NK-mediated viral control may depend on quantitative, as well as qualitative NK receptor expression features.

Genetic analysis of the NK-cell response to MCMV

Previous studies implicated both Ly49P+ and Ly49G₂+ NK cells with MCMV resistance (16, 32). Disentangling the contribution of either subset, however, has been difficult without a monospecific reagent to detect Ly49P. A genetic approach to assess the role of the different NK receptors is equally challenging due to the complexity of disrupting a singular Ly49 gene (50) and because of the rare frequency of cross-over events (51) within the Ly49 gene cluster. Moreover, NKC haplotypes associated with MCMV resistance in D^k-expressing MA/My and C57L-derived mice encode both Ly49P and a D^k-binding G₂ allele, while NKC types in less resistant strains (e.g. BALB.K) lack both of the relevant Ly49 receptor genes/alleles.

In lieu of the challenges, we attempted to clarify the role of distinct NK subsets by analyzing a variety of immune response traits in a cohort of genetically diverse mice segregating D^k. While Ly49p is expressed in G₂+ and G₂- NK cells (33), we reasoned that both subsets may contribute MCMV protection and that key NK cell features must coincide with low viral burden in animals with the essential genetic make-up. We established a basis for the analysis by assessing NK cells in our D^k disparate strains after MCMV infection. As these were essentially equivalent in peripheral blood and spleen pre- and post-infection (Fig 1B, Table I), we concluded that the subset composition of naïve spleen NK cells could be estimated and then used to gauge MCMV responsiveness.

We next examined 200+ MA/My×C57L hybrid offspring segregating D^k. MHC and NKC congenic blocks were included in the genetic crosses to help resolve non-MHC, non-NKC genetic effects. Before infection, the mice were weighed and bled. Blood leukocytes were stained with fluorochrome-conjugated mAbs for CD3, CD19, and NK receptors and then analyzed by flow cytometry. Lymphocyte and NK cell subset light-scattering features, absolute number, frequency, and receptor display intensity were determined and recorded as individual pre-infection traits. Afterward, the mice were infected with MCMV. Spleen and liver tissues were collected 3.5 d later for analysis of post-infection traits, including genomic genotypes (not shown) and viral load. Spleen leukocytes from infected animals were analyzed as for pre-infection traits. The change in trait values were calculated from the pre- (blood) and post- (spleen) infection trait values. In total, 110 distinct traits were assessed (Supplemental Fig 1).

All trait-to-trait comparisons were performed in the R computing environment using an automated correlation analysis followed with a strict Bonferroni correction for multiple tests (Supplemental Table II). From a heatmap depicting the results of all correlations, we observed at least 15 different hierarchical clusters that revealed trait-association relationships (Supplemental Fig 1). For example, post-infection forward scatter (FSC) (cluster 2) and side scatter (SSC) (cluster 3) traits were clustered separately but also generally positively corresponded to one another. On the other hand, clusters 2 and 3 were generally negatively associated with traits grouped together in cluster 1. To reduce the complexity of the analysis, 40 representative traits for the 15 clusters were selected and re-analyzed separately (Fig 2).

Figure 2. Multi-trait hierarchical cluster analysis of representative pre- and post-infection NK cell features and clinical traits.

The heat map depicts the hierarchical cluster output following comparison of representative multi-trait associations obtained from automated Pearson correlations performed in R. Correlation values (ρ) are indexed to the heat map at bottom. Traits that were highly significantly negatively (red bold) and positively (blue bold) correlated to log spleen MCMV are highlighted. Arrows designate pre- (red) and post- (black) infection traits most highly correlated to log spleen MCMV. Numerals on the heat map indicate significant (1 = p < 10⁻¹⁶; 2 = 10⁻⁸ > p ≥ 10⁻¹⁶; 3 = 10⁻⁴ > p ≥ 10⁻⁸; 4 = 0.01 > p ≥ 10⁻⁴) correlation values after the Bonferroni correction (see also Supplemental Table II).

Detailed trait associations, including those that best explain viral control, were then ascertained. Remarkably, the post-infection percentage of both total G₂+ (ρ = −0.8, p < 10⁻¹⁶) and G₂+ I/U- (G₂ single positive (SP); ρ=−0.75, p<10⁻¹⁶) NK cells was highly negatively correlated with log spleen MCMV (Fig 2, Supplemental Table II). In fact, of all traits analyzed, only the change in %body weight had a higher inverse correlation with MCMV in spleen (Supplemental Fig 1, Supplemental Table II). Although the post-infection percentage of total I/U+ NK cells also negatively correlated (ρ=−0.43, p_Bonferroni (p_B)<5×10⁻⁸), I/U+ G₂- (I/U SP) NK cells showed no significant association (Fig 2). The results suggest that the I/U+ NK correlation with spleen MCMV likely was due to overlapping G₂ expression. In contrast, post-infection G₂- (ρ=0.8, p<10⁻¹⁶) and G₂- I/U-(DN; ρ = 0.75, p<10⁻¹⁶) NK cells directly corresponded with spleen viral load better than any other single trait (Fig 2, Supplemental Table II). Thus, we infer that G₂+ NK cells were inextricably linked with viral control, more so than any other lymphocyte feature, and that high viral load exacerbated malaise in mice with an abundance of G₂-NK cells.

To examine the effect of class I MHC polymorphism on NK cell features, a principal component analysis (PCA) was performed (Fig 3). As shown, all offspring, marked by MHC genotype, were plotted based on differences in the first two principal components, distilling all 110 features down to two dimensions to enable visualization (Fig 3A and B). The primary principal component (horizontal axis), which explains the majority of the variation among all 110 features, very clearly separates D^k and D^k/b mice from D^b offspring, which were generally marked by higher values in the first principal component. Interestingly, three ‘atypical’ D^b mice clustered separately from other D^b offspring, primarily due to a major difference in the first principal component. Altogether the PCA results clearly demonstrated the essential impact of MHC genotype on traits that distinguished the offspring.

Figure 3. Principal components analysis of the NK cell response to MCMV by MHC genotype.

To extract the essential information from the analysis of all 110 traits, a principal components analysis was performed. A. The two-dimensional graph depicts the PCA for the first two components segregated by MHC genotype. Arrows indicate ‘atypical’ D^b offspring clustered separately from other D^b animals. B. The graphs depict PCA parameter loadings with representative traits plotted for the first two principal components.

Primacy of the Ly49G₂+ NK response to MCMV coincides with self-expression of the G₂ licensing-ligand D^k

The above results suggested that G₂+ NK expansion after infection was defined by MHC genotype. To examine this, we focused on those NK features most significantly (Bonferroni-corrected) associated with viral control. Before infection, naïve NK cell features varied broadly in the genetically diverse cohort (Fig 4A and data not shown). As expected and consistent with a D^k-licensing effect, G₂+ NK cells were less frequent (ρ = −0.44, p_B<1.5×10⁻⁸) with lower Ly49G₂ receptor display (ρ = −0.38, p_B< 5×10⁻⁶) in naïve D^k offspring, whereas DN NK cells were slightly more frequent (Supplemental Fig 1). In contrast, MHC type had no impact on the percentage of I/U+ NK cells or Ly49I/U MFI in the hybrid offspring (Supplemental Fig 1).

Figure 4. Ly49G₂+ NK cells selectively expand after MCMV exposure in mice with self-D^k.

The matrices depict the relationships amongst naïve blood (A) and infected spleen (B) NK cell features (i.e. continuous variable traits) and spleen MCMV load. The univariate distribution of each variable is shown (top and right) via a histogram with a kernel density overlay (red curve). Each of the center panels shows a scatter plot for each pair of continuous variables, color-coded by MHC D genotype (homozygous D^b = blue; heterozygous and homozygous D^k = red), with the trend displayed by a lowess overlay (i.e., a locally-weighted polynomial regression shown as red curve in each scatter plot). C. The graphs depict the average percentage composition of Ly49G₂+ and Ly49I/U+ NK cells segregated by MHC D type (see above) for naïve and MCMV-infected offspring.

Remarkably, only Ly49G₂ MFI on peripheral blood NK cells in naïve mice significantly predicted (ρ = 0.33, p_B=0.002) viral load in spleen after infection (Figs 2 and 4A). This result suggests that G₂+ NK cells were pre-equipped to handle MCMV in mice with a G₂ licensing-ligand D^k. Interestingly, NKp46 MFI on naïve NK cells was somewhat higher in H-2^k offspring (Supplemental Fig 1). Pre-infection NKp46 by itself, however, was not a predictive index of viral control (Fig 4A). Altogether, the results from naïve mice suggested that D^k-licensed G₂+ NK cells marked by relatively lower G₂ MFI were essential to provide protection following MCMV infection.

We further addressed the question by comparing the composition and features of NK cells in the D^k-disparate offspring after infection. The percentage of post-infection G₂+ NK cells was much higher in D^k offspring, while DN NK cells were more abundant in infected offspring without D^k (Fig 4B). Selective expansion of the G₂+ NK subset in response to MCMV likely explains the dichotomy since the average G₂+ NK frequency in infected spleen was ~45% higher than in naïve blood, while the average I/U+ NK frequency revealed little, if any, response to MCMV (Fig 4B and C and data not shown). A severe reduction in G₂+ NK frequency was especially striking in offspring lacking D^k, though the percentage of I/U+ NK cells also declined in the same animals (Fig 4C). We infer from the results that in addition to shaping NK subsets at rest in naïve mice, MHC class I polymorphism had a definite effect on NK subset composition and effector function during the response to MCMV. The most parsimonious interpretation of the data is that G₂+ NK cells provided critical viral control in the spleen of mice with a self-D^k ligand, but failed to elicit the same protection in mice without D^k where DN NK cells instead were most abundant.

MCMV infection highlights the plasticity of NK subsets and the profound role of MHC polymorphism in shaping innate immunity and viral control. We next used multiple regression analysis to investigate if any traits other than MHC type significantly affected viral load in spleen. Each continuous variable trait plus MHC type was tested (full model) and compared to test results with MHC type alone (reduced model), which accounted for ~63% of the variation in spleen viral load (Supplemental Table III). Full models based on post-infection percentage of total G₂+ (77%) and G₂ SP (70%) NK cells, but not Ly49G₂ MFI, explained significantly more viral load variation than the reduced MHC-type alone model (Fig 5). MHC type plus the post-infection percentage of I/U+ NK cells also explained 71% variation in viral load (statistically significantly better prediction than MHC alone), likely due to overlap with G₂+ NK cells since any difference in the reduced and full model with I/U SP NK cells was insignificant (Fig 5). Full models with the post-infection percentage of G₂- or DN NK cells explained slightly more than 77% of viral load variation (Fig 5, Supplemental Table III). Both subsets, however, significantly correlated with high viral load (Fig 2, Supplemental Table II). Hence, as with Figure 4 results, multiple regression analysis confirms that the post-infection balance of G₂+ and DN NK cells significantly correlated with MCMV protection. Multi-regression analysis further implicates that another genetic factor(s) beyond the MHC regulates the frequency/composition of NK cells in the response to MCMV. It is unlikely that the effect was due to NKC polymorphism since the Ly49g genotype was not significantly correlated (Supplemental Fig 1). The results demonstrate that the best index of MCMV control involved a balance of more frequent G₂+ NK cells, which coincided with less frequent DN NK cells; together with MHC type these traits explained the majority of viral load variation.

Figure 5. MHC I D-independent MCMV control corresponds with distinct NK cell subsets and post NKp46 expression.

The graph depicts the −log₁₀(p) (filled) and %trait variance (full model; empty) values obtained in multiple linear regression analysis (see Supplemental Table III) to ascertain which continuous NK cell traits, beyond MHC type, significantly affected viral load variance in the spleen.

Regulated NKp46 display on distinct NK cell subsets essential in viral control

Beyond the role of G₂+ NK cells in viral control, we unexpectedly found that post-infection NKp46 MFI on NK cells negatively correlated with spleen MCMV (Figs 2 and 4B). The results implied that NKp46 might help to limit MCMV spread. To test this, we first examined its relationship to different NK cell subsets after infection. As expected, NKp46 MFI corresponded with more frequent G₂+ NK cells (ρ=0.48, p_B<7×10⁻¹¹). However, the post-infection percentage of I/U+ NK cells showed an even stronger (ρ=0.56, p<10⁻¹⁶) relationship (Figs 2 and 4B). In contrast, post NKp46 MFI negatively correlated (ρ=−0.54, p<10⁻¹⁶) with the percentage of DN NK cells (Figs 2 and 4B, Supplemental Table II). The results demonstrated that the expansion of DN NK cells best corresponded with lower NKp46 MFI.

Given its relationship to NK subset composition, we next examined post NKp46 MFI on defined NK subsets. As expected, its expression on G₂ SP NK cells was inversely related to spleen viral load (Fig 6A). However, post NKp46 MFI on I/U SP and DN NK cells corresponded even more favorably to viral control (Fig 6A). The results indicated that distinct NK subsets must variably express NKp46 after MCMV exposure. In fact, DN NK cells in infected D^k mice had much higher expression than either subset of Ly49+ NK cells (Fig 6B). In contrast, spleen NK subsets from infected offspring without D^k displayed substantially lower NKp46 than the D^k counterpart (Fig 6B). Similar NKp46 disparity amongst G₂− and G₂+ NK cells was observed in infected R7 mice (Fig 6C). Altogether the results suggest that a robust G₂+ NK response to MCMV in D^k mice was essential to regulate higher NKp46 MFI, and this was most pronounced on G₂- NK cells.

Figure 6. Differential NKp46 display by distinct NK cell subsets corresponds with licensed NK cell-mediated viral control.

A. The scatter plots show post NKp46 MFI versus log spleen MCMV for each of the indicated NK cell subsets for all offspring. Each data point for a single mouse is color-coded by MHC D genotype as in Figure 4. B. The graphs depict average post NKp46 MFI values for the indicated NK subsets, segregated by MHC I D genotype (D^k includes hets and homozygous D^k mice). Significant differences (***p < 0.002) compared with NKp46 MFI on DN NK cells are indicated. The average post NKp46 MFI was significantly higher on each subset in D^k mice compared to D^b mice (not depicted). C. A representative dot plot and histograms show gated NK cells from MCMV-infected R7 spleen stained for I/U+ (14B11) and G₂+ (4D11) NK cell subsets and respective NKp46 receptor display, by quadrant. The graph at right shows the average post NKp46 MFI (+sd) for the indicated NK cell subsets in R7 spleen. Data are representative of three experiments with 3–5 mice per group. ** p<0.01, *** p<0.005.

The above results led us to question whether NKp46 expression, apart from the MHC effect, might also contribute to any of the variance in viral control. We assessed post NKp46 MFI on NK cells by multiple regression analysis comparing full (MHC plus post NKp46 MFI) and reduced (MHC alone) models. The full model significantly explained 75% of spleen viral load variance (Fig 5, Supplemental Table III); a result equivalent in scale to a full model based on the post G₂+ NK cell frequency. A full model based on NKp46 MFI on DN NK cells proved even better, whereas full models restricted to NKp46 MFI on G₂+ or G₂ SP NK cells accounted for less of the viral control variance (Fig 5). Thus, the post NKp46 MFI on NK cells explained significantly more of the variance in viral control than MHC alone. MHC-independent genetic regulation of post NKp46 MFI was most significantly evident on DN NK cells, whose frequency also explained significantly more viral load variance than MHC alone. We infer that D^k-licensed G₂+ NK cells were able to efficiently detect infected targets, which led to their rapid activation and expansion. MHC-independent regulation further affects the balance of G₂+ and G₂- NK cells, as well the differential NKp46 expression on distinct NK subsets after MCMV exposure (Fig 7).

Figure 7. A model for D^k-licensed Ly49G₂+ NK cell-mediated MCMV resistance.

The diagram depicts G₂+ and G₂− NK subset responses in D^k disparate mice following MCMV infection. The observed percentage of G₂+ NK cells in D^k mice increased significantly after MCMV exposure and these cells were needed to deliver critical viral control. Although G₂− NK cells failed to expand in the same mice, they retained higher NKp46 receptor display, a major NK feature that also corresponded with enhanced viral control. G₂+ NK cells in non-D^k mice, however, failed to expand or to elicit viral control. Instead, the percentage of G₂- NK cells increased significantly and NKp46 receptor display remained substantially lower than on NK cells in mice with self-D^k.

DISCUSSION

The combined use of classical genetics and flow cytometric analysis of immune cells, collected first from naïve and then infected animals, is an innovative and powerful approach to interrogate the immune response following virus exposure. A concern with multiple testing in large sets of immunological data, however, is the possibility for random associations to occur by chance (52). With 110 different trait values, multiple trait comparisons yielded more than 6,000 correlation values. Thus, we applied the statistically conservative Bonferroni correction as well as calculation of the false discovery rate (45) to exclude false-positive associations. Related trait associations (e.g. post %Ly49G₂+ and post %Ly49G₂− NK cells or pre Ly49I/U MFI and post Ly49I/U MFI) were excluded. The report emphasizes presentation of significant data (i.e. p_B<0.01) only after the more conservative Bonferroni correction, so that some true associations may have been disregarded. Nonetheless, significant trait associations with viral load implicate either a direct or indirect effect on viral control. Significant variance in immune responsiveness and MCMV restraint facilitated our search to uncover key NK cell features and the responsible gene(s). Despite the focus on viral immunity, the approach is amenable to other settings where genetic diversity has an influence on host immunity.

Most remarkable, the data herein show that both the percentage and numbers of splenic G₂+ NK cells rapidly expanded in offspring with self-D^k, at the expense of G₂-NK cells, within days after MCMV infection. The response profile was completely reversed in non-D^k mice where innate control mechanisms were readily overwhelmed. Whereas no other single NK feature had a stronger or more significant correlation with viral control, we infer that licensed G₂+ NK cells were required to efficiently detect and elicit rapid killing of viral targets. This interpretation, however, contrasts with the results of Orr et al. (40), which showed that licensed NK cells attenuated MCMV control in neonatal B6 mice. While NK cells in B6 use the Ly49H activation receptor to detect MCMV m157 displayed on the surface of infected cells and render target cell killing (15), an H-2^b factor(s) is not known to significantly affect their capacity for MCMV recognition. In fact, Ly49C/I+ NK cells are not known to detect or respond to MCMV targets. Thus, the discrepancy might be explained by a difference in the primary mode of target recognition by NK cells in the different experimental settings. In the current work, licensed G₂+ NK cells were competent to detect missing-self MHC I cues on viral targets and responded more aggressively than all other NK cells examined.

That G₂+ NK cells rapidly respond and undergo significant expansion in virus-infected B6 mice has been reported (36–39). However, depletion of G₂+ NK cells from B6 mice had no effect on viral control in spleen (38), and only a relatively modest effect in liver and salivary gland by d 7 after MCMV infection (40). Moreover, unlike MHC independent G₂+ NK subset expansion following hematopoietic stem cell transplantation or in vivo cytokine stimulation in B6 mice (36), herein G₂+ NK cells selectively expanded in offspring bearing self-D^k. This could have been due to a collapse of splenocytes and immune response capability in more MCMV-sensitive offspring due to relatively high viral dose. However, similar differences in G₂+ NK responsiveness and expansion following lower dose (10⁴ PFU / mouse) MCMV infection were observed in a separate cohort of MA/My×C57L offspring (data not shown). Together, the results establish that it was not an intrinsic proliferative potential of the NK subset itself, but an extrinsic D^k ligand that corresponded with subset expansion and viral control. When the frequency of licensed G₂+ NK cells increased, G₂- NK cells were less frequent with higher NKp46 MFI, spleen MCMV was lower and morbidity was reduced. Whether these effects are directly related to NK cell licensing, enhanced detection of viral targets or both is an open question.

Curiously, a rare fraction of three ‘atypical’ D^b mice with aberrantly low spleen MCMV and very little weight loss were clustered separately by PCA from the remainder of D^b mice. Naïve G₂+ NK cells and Ly49G₂ MFI in the mice was very similar to other D^b offspring, consistent with the genetic typing (data not shown). Moreover, there was no evidence for allelic chimerism in MHC I D ((41) and data not shown). However, unlike other D^b offspring, the aberrant mice retained G₂+ NK cells following MCMV exposure. Whereas a similar result has been observed before in related genetic crosses (53, 54) and weight loss and lymphocyte light-scatter features were not unusual in the mice, we cannot verify if inadequate infection, a ‘cage effect’ for two of the three mice or enhanced MCMV resistance explains the outcome. Interestingly, Ly49G¹²⁹ allele, which is identical to the C57L Ly49G allele, has been shown to bind MHC I D^b tetramers, but no evidence for NK licensing due to the interaction has been observed (Andrew Makrigiannis, personal communication). So, it is tempting to speculate that G₂+ NK cells may, on rare occasion, be licensed by D^b or a nonclassical MHC I (55) and competent to detect MCMV targets via missing-self recognition. In fact, three additional D^b mice with aberrantly low spleen MCMV that were not distinguished by the PCA actually increased the G₂+ NK subset in response to MCMV (data not shown). Together, the data highlight the importance of D^k-licensed G₂+ NK cells in the response to MCMV and further hint that variation in the extent of NK licensing and capacity for detection of missing-self cues on viral targets might significantly influence NK-mediated viral control.

The natural cytotoxicity (stimulatory) receptor (NCR) NKp46 is displayed on human and mouse NK cells; high-level expression corresponds with NK-mediated tumor killing (56, 57). NKp46 and other NCRs are known for binding viral hemagglutinins in cells infected with influenza or poxviruses (5, 13, 29). Recent work further claims NKp46 detection of an unknown ligand manipulated by HCMV infection (13, 29). Lower expression, as in NKp46^dull NK cells, corresponds with persistence in some chronic viral infections, including HCV, HIV and HCMV (28, 58–61) and in patients with pulmonary TB (62). Diminished NK functionality due to repressed NCR display in HIV-infected patients may explain reduced viral control (58, 59). However, low NKp46 display is not correlated with dysfunction in all NK subpopulations or in all HIV-infected patient groups (60) and might even help to resolve HCV infection (61). This, however, is an active area of investigation since NKp46^high cells with high cytotoxicity and cytokine potential have been shown to control HCV infection (63). Thus, NKp46 is a critical marker to assess NK responsiveness in acute and chronic viral infections.

Interestingly, we found that high post NKp46 MFI corresponded with reduced weight loss and lower viral load in the spleen after MCMV infection. To our knowledge, this is the first report linking NKp46 with viral control and morbidity in MCMV-infected mice. Diminished NKp46 display was an unmistakable effect of acute MCMV exposure, in accordance with human CMV infection (28). The result was most conspicuous in MCMV-sensitive D^b mice, but also evident even in more resistant D^k mice. High expression amongst MCMV resistant D^k mice suggested that post NKp46 MFI might simply represent a marker of activated NK cells. Yet, NK cells in all mice exhibited features consistent with cellular activation after MCMV exposure. More reasonably, NKp46 display is modulated in a way that corresponds to viral control. We envision two scenarios: 1) NKp46 may help to detect MCMV targets; 2) NKp46 may be an additional effector of anti-MCMV immunity, one that requires a prior signal to unleash its full potential.

Whereas marked MCMV control was observed in offspring with post-NKp46^high G₂- and DN NK cells, the relationship was still limited to offspring with self-D^k, which also had more abundant G₂+ NK cells. Moreover, post NKp46 MFI plus MHC type accounted for ~76% of the variation in viral load after infection due to NKp46^high DN NK cells, which most significantly explained the association. These data illustrate the importance of G₂- NK cells, in addition to G₂+ NK cells, in offspring with self-D^k and suggest that licensed NK detection of MCMV enhanced defined NK cell subsets needed to efficiently control MCMV.

In summary, we conclude that NK cells provided vital antiviral immunity in mice with the G₂ licensing-ligand D^k. Their expansion following viral exposure closely marked significant immune cell responsiveness, reduced morbidity, increased MCMV control, and viral clearance. Moreover, detection of licensed NK features (i.e. Ly49G₂ MFI) prior to viral infection actually significantly predicted viral control, which implies that related human studies may likewise uncover predictive markers of NK-mediated viral control. The results foreshadow that predictive modeling is an important and promising approach to examine how protective human genotypes may correspond to NK cell-mediated responsiveness, immune enhancement and viral clearance.