An Unbiased Linkage Approach Reveals That the p53 Pathway Is Coupled to NK Cell Maturation

Natural killer cells are innate lymphoid cells whose potent effector functions are important for both tumor immunosurveillance and viral clearance. Indeed, NK cells exhibit cytotoxic activity and the ability to rapidly produce and secrete vast amounts of chemokines and cytokines. NK cells acquire these properties in a sequential manner involving multiple stages of development. The first stage is defined as the pre-NK cell precursor, giving rise sequentially to NK cell precursor, immature NK cells, and finally mature NK (mNK) cells expressing the surface molecule CD49b (1, 2). Subsequently, mNK cells undergo a four-stage maturation process defined by the regulated expression of CD11b and CD27. Specifically, the first mNK cell functional maturation subset expresses low levels of both CD11b and CD27 [hereafter referred to as double negative (DN)]. DN mNK cells then follow a CD11b^lowCD27^high (named CD11b^low) → CD11b^highCD27^high [double positive (DP)] → CD11b^highCD27^low (CD27^low) differentiation process (3). Notably, Chiossone et al. (3) found that the least mature (DN and CD11b^low) and most mature (DP and CD27^low) mNK cells largely differ by their specialization in proliferation and effector functions, respectively.

Various lines of evidence indicate that transcription factors and microRNAs (miRNAs) can influence mNK differentiation. Indeed, deficiency in Gata3, Id2, Nfil3, miR150, or miR-15a/16 leads to a decrease in the proportion of CD27^low mNK cells (4–9). Eomes, Tbx21 (T-bet), Myb, Zeb2, and miR155 also influence NK cell differentiation at specific stages of functional maturation (3, 8, 10–13). Although Eomes, Myb, and Zeb2 play a more prominent role during the later stages (3, 8, 10, 13), Tbx21 is believed to play a role at both the early and later stages of functional maturation (10, 12). In contrast, miR155 inhibits NK cell maturation as miR155-deficient NK cells exhibit a more mature phenotype with an increase in the frequency of the CD27^low subset (14).

In addition to transcription factors and miRNAs, other genes also influence NK cell differentiation (15). As a follow-up to our study identifying genetic loci linked to the proportion of NK cells (16), in this study we aimed to identify the key genetic determinants involved in mNK cell functional maturation, more specifically to identify the molecular pathways defining the transition between the DN, CD11b^low, DP, and CD27^low stages of mNK cell development. To this end, we used an unbiased genetic approach that indiscriminately addresses the contribution of all genetic determinants to variations in mNK cell functional maturation. We performed a linkage analysis that allowed for the identification of six significant loci on chromosomes 2, 4, 7, 10, 11, and 18, suggesting a multigenic regulation of NK cell functional maturation. To restrict the list of candidate genes located within these intervals, we performed an in silico analysis based on data available through the Gene Expression Omnibus (GEO) repository. Our analysis allowed us to highlight various molecular pathways and candidate genes regulating the specific stages of NK cell functional maturation. Lastly, we validated Trp53 as a gene regulating the functional maturation of NK cells. This study brings not only insight into the biology of NK cell maturation, but also unveils novel candidate genes regulating the functional maturation of NK cells, which is of relevance to both tumor immunosurveillance and viral clearance.

C57BL/6 (B6), NOD, B6.Rag1^−/−, NOD.Rag1^−/−, and B6.Trp53^−/− mice were purchased from the Jackson Laboratory. All of these strains were subsequently maintained at the Maisonneuve-Rosemont Hospital animal house facility (Montreal, Canada). F1.Rag (B6.Rag1^−/− × NOD.Rag1^−/−) and F2.Rag (F1.Rag × F1.Rag) mice were bred in house from the parental strains. For phenotypic analyses, 6–10 wk old mice were used. The Maisonneuve-Rosemont Hospital ethics committee, overseen by the Canadian Council for Animal Protection, approved the experimental procedures.

Spleens were treated with collagenase (1 mg/ml in PBS, Type V from Clostridium histolyticum; Sigma-Aldrich) for 15 min at 37°C and passed through a 70 μM cell strainer (BD Biosciences) to yield single-cell suspensions prior to staining with Abs. CD11b (clone M1/70), CD122 (clone TM-B1), CD27 (clone LG.3A10), CD3 (clone 17A2), B220 (clone RA3-6B2), CD19 (clone 6D5), CD45.1 (clone A20), CD45.2 (clone 104), Annexin V, and Zombie Aqua Dye Amcyan Abs were purchased from BioLegend; CD49b (clone DX5), NKp46 (clone 29A1.4), T-bet (clone eBio4B10), and Eomes (clone Dan11mag) from eBioscience; Ki-67 (clone B56) from BD Biosciences; LIVE/DEAD Fixable Yellow from Thermo Fisher Scientific; and p53 (clone 1C12) from Cell Signaling Technology. The Foxp3/Transcription Factor Staining Buffer Set was used for intracellular staining according to the manufacturer’s instructions (eBioscience). Annexin V staining was performed in buffer containing HEPES 10 mM, NaCl 150 mM, KCl 5 mM, MgCl₂ 1 mM, and CaCl₂ 1.8 mM. All samples were acquired using FACSCanto I (BD Biosciences), Fortessa 1 (BD Biosciences), or Fortessa x-20 (BD Biosciences), and were analyzed using the FlowJo software (Tree Star, Ashland). After excluding doublets, CD3ε⁻CD19⁻ cells were excluded for Rag1-sufficient mice, followed by the application of a strict size exclusion backgate to the forward light scatter/side scatter profile corresponding to live CD122⁺CD49b⁺ NK cells for both Rag1-sufficient and -deficient mice. NK subsets were then separated based on their expression of CD11b and CD27.

Genomic DNA was isolated from the tails of F2.Rag male and female mice by using the DNeasy blood and tissue kit from Qiagen. Single nucleotide polymorphisms were then detected from the F2.Rag mice DNA using the Illumina mouse low density linkage panel serviced through The Centre for Applied Genomics at the Hospital for Sick Children, ON, Canada. Marker location (in megabases) was determined using the National Center for Biotechnology Information Build m37. The logarithm of odds (LOD) scores were obtained through a single- or two-dimensional quantitative trait locus model using the R/qtl package (17, 18) for the R software (version 2.11.1) with the Haley–Knott algorithm to increase single nucleotide polymorphism (SNP) resolution. LOD scores higher than 3.54 were significant for single-dimensional analysis according to permutation tests (n = 10,000, p = 0.05), and LOD scores between 2.14 and 3.54 were considered suggestive. A Pearson Chi–Square for allele frequencies confirmed that all the significant and suggestive loci did not differ from the Hardy–Weinberg equilibrium. For the NK functional maturation (Nkfm) 2 locus, the distal part of chromosome 7 starting at 100.84 Mb was excluded from the Bayes interval analysis because of the bimodal LOD score distribution.

Data for the various experiments were tested for significance using a nonparametric Mann–Whitney U test with a minimal threshold of 0.05, or a paired Student t test with a minimal threshold of 0.05 in the analysis of p53 expression. Estimation of the interval coordinates was obtained using a 95% Bayes interval test. Significance for the genotype distribution differences was tested with an ANOVA and Games–Howell post hoc test. Normality of distributions was determined using a Shapiro–Wilk test. All statistical analyses and the F₂ distribution were obtained using the SPSS 19.0 software.

Fold changes (>1.5) in microarray datasets for B6.Rag2^−/− mNK cell subsets from Chiossone et al. (3) were tabulated and uploaded into the Ingenuity-based pathway analysis (IPA) software (Ingenuity Systems, http://www.ingenuity.com). To predict the activation state of transcription factors and other regulators based on their effect on expression of downstream target genes, we performed an upstream regulator analysis (URA). The IPA regulation z-score algorithm was used to predict the direction of change for a given function (increase or decrease). A z-score >2 or <−2 means that a function is significantly increased or decreased, respectively, for all significantly enriched groups of genes (p <0.05).

NOD mice are known to possess NK cells that exhibit multiple functional defects (16, 19–24). Consequently, we first set out to determine whether NOD NK cells displayed any disparities in the four-stage model of functional maturation relative to those from B6 mice. To select for NK cells in both B6 and NOD mice, we performed flow cytometry and gated on CD3ε⁻CD19⁻B220⁻CD122⁺CD49b⁺ cells (Fig. 1A). Addition of NKp46 to this gating strategy did not improve our ability to select for NK cells and was thus not included in our subsequent analyses (Supplemental Fig. 1). To assess the functional maturation of NK cells, we subsequently verified the expression of CD27 and CD11b on the CD3ε⁻CD19⁻B220⁻CD122⁺CD49b⁺ NK cells (Fig. 1A). When comparing B6 and NOD mice, in addition to the total mNK cell proportion (16), we observed conspicuous differences in the maturation profiles of the mNK cells. Relative to B6 mNK cells, NOD mNK cells exhibited an increased proportion of the least mature DN and CD11b^low subsets and a reduced proportion of the most mature CD27^low subset (Fig. 1B). The impaired mNK cell functional maturation in NOD mice relative to B6 is also reflected in the absolute numbers of the CD11b^low and CD27^low NK cell subsets, denoting a block in the functional maturation process for NOD mNK cells. To confirm that this block is not secondary to a defect in the regulation of CD11b or CD27 expression in mNK cells from NOD mice, we also determined the level of expression of T-bet and Eomes. T-bet expression is enhanced with mNK cell functional maturation, whereas Eomes is more highly expressed in the least mature mNK cell subsets (25). Accordingly, we find that mNK cells from NOD mice express lower levels of T-bet and higher levels of Eomes than mNK cells from B6 mice (Fig. 1C). This is further evidenced by defining the ratio of Eomes/T-bet expression, quantified by the mean fluorescence intensity (MFI), which is highest in NOD relative to B6 mice (Fig. 1C). Together, these data support the view that mNK cell functional maturation is impaired in NOD mice.

As the NOD mouse is a well-established model of spontaneous autoimmune diabetes, we wished to determine if the inflammation associated with the underlying pathology was causal to the observed block in NK cell functional maturation. To do so, we took advantage of mice deficient for Rag1, where Rag1-deficient NOD mice do not develop autoimmune diabetes (26), eliminating any potential underlying inflammation that might bias our observed phenotype. Moreover, the deletion of Rag1 leads to a complete ablation of peripheral T and B cells while still allowing NK cell development (3, 27, 28). Due to the absence of T and B cells, we gated directly on live cells expressing CD49b and CD122 to select for mNK cells (Fig. 2A). Using Rag1-deficient mice of the B6 and NOD backgrounds, we found that, similar to the wild-type counterparts (Fig. 1B), mNK cells from NOD.Rag1^−/− mice display an increase in the proportion and absolute number of the CD11b^low subset as well as a decrease in the proportion and absolute number of the most mature CD27^low subset (Fig. 2B). Therefore, in comparison with the B6 genetic background, mNK cells from the NOD genetic background display a distinct block in their functional maturation process, which cannot be attributed to the ongoing autoimmune process or the presence of B and T cells. As a result, the differences observed in the functional maturation of NK cells are most likely to be caused by genetic differences between the two inbred strains.

To identify genetic factors contributing to the regulation of NK cell functional maturation, we performed a genetic linkage analysis. Recent publications from our team (16, 29, 30) have demonstrated that Rag1 deficiency increases the sensitivity for identifying significant loci linked to an observed phenotype for rare lympho-myeloid cell populations. As a result, we generated a cohort of 181 F₂ (B6.Rag1^−/− × NOD.Rag1^−/−) mice, hereafter termed F2.Rag mice, and quantified the proportions of mNK cells comprising the four stages of functional maturation, namely the DN, CD11b^low, DP, and CD27^low mNK cell subsets. The proportion of the four mNK cell subsets each followed a Gaussian distribution (Fig. 3A), which is suggestive of a multigenic regulation of the phenotypes. We then undertook linkage analyses for these traits, where we subjected the DNA from each F2.Rag mouse to the Illumina Golden Gate low-density platform and executed a genome-wide SNP genotyping. Using this approach we were able to identify six major loci on chromosomes 2, 4, 7, 10, 11, and 18 (Figs. 3B, 4), underlining the multigenic regulation of NK cell functional maturation. Accordingly, we have named the loci Nkfm.

We then generated a high-resolution map of the significant loci. Specifically for the DN mNK cell subset, we identified a single significant linkage on distal chromosome 18 with a LOD score of 5.76 (Fig. 4A), where we named this interval Nkfm1.1. The 95% Bayes confidence interval analysis, depicted by the gray shaded area, delimits the Nkfm1.1 locus between 47.48 and 67.48 Mb (Fig. 4A, top right). For the CD11b^low mNK cell subset, we identified three loci significantly linked to the trait. Notably, an overlapping segment of Nkfm1 was also significantly linked to the proportion of the CD11b^low mNK subset and was consequently named Nkfm1.2 (Fig. 3B). Nkfm1.2, which is positioned at 59.48–71.48 Mb on chromosome 18, presented with a LOD score of 8.02 (Fig. 4B). The two additional loci, on chromosomes 7 and 11, significantly linked to the CD11b^low mNK subset, showed respective LOD scores of 4.70 and 9.06 (Fig. 4B). We termed these loci Nkfm2 (chromosome 7, Bayes interval 43.46–83.46 Mb) and Nkfm3.1 (chromosome 11, Bayes interval 57.68–89.93 Mb). These results portray a differential genetic regulation of the two least mature mNK subsets, namely the DN and CD11b^low mNK cell subsets, where one and three loci are respectively linked to each subset. However, these data also suggest a potential common regulation of both DN and CD11b^low mNK cell subsets, as Nkfm1 is linked to both of these subsets (Fig. 3B).

The size of the DP mNK subset was independent of the Nkfm1, 2, or 3 loci. Instead, we identified a locus on proximal chromosome 10, which we termed Nkfm4, with a significant linkage for the DP mNK cells (Fig. 4C) as well as the CD27^low mNK cells (Fig. 4D). Consequently, we subdivided this locus into two separate loci, which we termed Nkfm4.1 and Nkfm4.2 (Fig. 3B). Nkfm4.1, which is positioned at 20.15–36.27 Mb, is the sole locus linked to the regulation of the DP mNK cell subset and presented with a LOD score of 7.67, whereas Nkfm4.2, which is positioned at 24.15–44.15 Mb, is linked to the regulation of the CD27^low mNK cell subset and presented with a LOD score of 8.02 (Fig. 4C, 4D). Along with Nkfm4.2, three additional significant loci were identified for the regulation of the most mature CD27^low mNK cell subset. These loci are located on chromosomes 2, 4, and 11 with LOD scores of 5.51, 5.41, and 3.90, respectively (Fig. 4D). These loci, which we named Nkfm5 (chromosome 2), Nkfm6 (chromosome 4), and Nkfm3.2 (chromosome 11), are located at 31.21–64.17 Mb, 23.03–47.63 Mb, and 73.68–110.98 Mb, respectively. Interestingly, as for DN and CD11b^low mNK cell subsets, although the DP and CD27^low mNK cell subsets are linked to one and four loci respectively, potential common regulation of both DP and CD27^low mNK cell subsets is suggested by a common linkage to the Nkfm4 locus (Fig. 3B).

To determine the impact of the specific genotype on each mNK cell subset, we subsequently segregated F2.Rag mice based on their haplotype at the SNP corresponding with the highest LOD score within each locus. This segregation of F2.Rag mice is, at times, in agreement with the parental NOD and B6 phenotypes, where NK cells from the NOD genetic background depict a more immature phenotype relative to the B6 background (Figs. 1, 2). Notably, F2.Rag mice bearing the NOD genotype within Nkfm1.1 and Nkfm1.2 exhibit a higher proportion of DN and CD11b^low mNK cells (Fig. 4A, 4B, lower right), and F2.Rag mice with NOD alleles at the Nkfm4.2 locus present with a lower proportion of CD27^low mNK cells (Fig. 4D, lower right). Conversely, mice homozygous for the NOD alleles within Nkfm3.1 presented with a decreased proportion of the CD11b^low mNK subset (Fig. 4B, lower right), and those with NOD alleles within the Nkfm3.2, Nkfm5, and Nkfm6 loci an increased proportion of CD27^low mNK cells (Fig. 4D, lower right), both of which contrasts with the mNK cell phenotype found in the NOD.Rag1^−/− parental strain (Fig. 2B). These results reveal complex interactions between loci, suggesting some degree of genetic epistasis in the regulation of NK cell functional maturation (31), and where the Nfkm4.2 locus likely plays a dominant contribution over the aforementioned loci in defining the proportion of CD27^low mNK cells.

We next generated a complete list of all genes encoded within the six identified Nkfm loci (Supplemental Table I). Several genes located in the Nkfm intervals caught our attention as they have previously been associated with NK cell development (Supplemental Table I). Indeed, Tbx21-deficient mice present with a deficit in CD27^low mNK cells (10), in line with our finding that Tbx21 is encoded within the Nkfm3.2 locus linked to the regulation of the CD27^low mNK cell subset. The Myb transcription factor encoded within the Nkfm4.1 locus linked to the DP subset also plays a role in mNK cell functional maturation (8, 10). Moreover, the Nkfm6 locus linked to the CD27^low mNK cell subset contains Tgfbr1, which is involved in the signal transduction pathway for TGF-β (32, 33), known to influence the final stage of mNK cell maturation (34). The Nkfm3 locus also reveals Stat3, Stat5a, and Stat5b as interesting potential candidate genes for NK cell functional maturation, where various cytokines, hormones, and growth factors activate Stat3 and Stat5 via their signaling pathway, including IL-2, IL-7, IL-15, and Flt3 (35, 36), which influence NK cell development (37–39). Interestingly, Flt3l is a candidate gene located within Nkfm2, where Flt3L is involved in the generation of common lymphoid progenitors from which NK cells originate (39). Altogether, the loci linked to NK cell functional maturation encode genes that have been previously associated to this trait. This supports the validity of the linkage analysis for identifying candidate genes and pathways regulating each differentiation step during NK cell functional maturation.

To delimit the list of candidate genes encoded within the Nkfm loci, we aimed to identify differentially expressed genes (DEGs) between the CD11b^low versus DP and DP versus CD27^low mNK cell maturation stages that were also located within the loci of interest. Correspondingly, in 2009, Chiossone et al. (3) published their findings that validated the four-stage developmental pathway of mNK cells during functional maturation for which they also performed microarray analyses on the CD11b^low, DP, and CD27^low mNK cell subsets from a B6.Rag2^−/− mouse. Importantly, these microarray data were submitted to, and thus accessible from, the GEO repository. Therefore, we took advantage of this microarray data and listed the DEGs that were found within Nkfm1.2, -2, -3.1, -3.2, -4.1, -4.2, -5, and -6 (Supplemental Table II). Of the DEGs listed, seven genes, namely Arg1, Cdc6, 1110036O03Rik, Myb, Pkp4, Ptgs1, and Zeb2, exhibited at least a 2-fold difference in their expression between maturation stages. Interestingly, the gene expression of Arg1, Cdc6, Myb, and Pkp4 decreases, whereas that of 1110036O03Rik, Ptgs1, and Zeb2 increases during mNK cell functional maturation, where the role of Zeb2 in functional maturation of mNK cells has recently been validated (13). Note that although the majority of the DEGs listed exhibited less than a 2-fold change in expression, small variations in expression in a set of genes may have a compound effect on a given biological phenotype and should not be readily dismissed. Altogether, this analysis highlighted multiple candidate genes from our linkage analysis that are also differentially expressed between the functional maturation stages of mNK cells.

To provide additional insight into the relevant genes affecting the specific steps of NK cell functional maturation, we next aimed to identify the genes located within the Nkfm loci whose differential gene expression between mNK cell subsets correlated with a change in the activity of their associated upstream regulators. To do so, an unrestricted Ingenuity-based URA was performed. This analysis predicts the activation state of upstream regulators based on their effect on the expression of all downstream target genes, where upstream regulators are not limited to transcription factors but can be any gene or small molecule that has been observed experimentally to affect gene expression. We applied a URA that was not restricted to genes within the Nkfm loci, but took into account all genes that are differentially expressed between the mNK cell subsets. We used two metrics to identify the most important upstream regulators: activation z-score and p value (dotted lines and red dots, respectively, in Fig. 5). The p value, calculated with the Fischer exact test, provides the statistical significance of the difference, whereas the activation z-score depicts the amplitude of the difference between NK cell subsets. The URA predicted that various pathways were differentially activated between the CD11b^low versus DP and DP versus CD27^low mNK cell subsets (Fig. 5). This approach positions the relative importance of specific genes in discrete functional maturation subsets. For instance, Il7 is an upstream regulator of the CD11b^low subset, whereas Il21 is an upstream regulator of the CD27^low subset, with respective p values of 3.83e−14 and 1.18e−08. The URA also suggests that mNK cells undergo major transcriptional changes as they transition from one functional maturation stage to the next, where almost twice the number of upstream regulators are listed during the transition between the DP to CD27^low stage relative to the CD11b^low to DP stage (compare Fig. 5A, 5B). Interestingly, some genes were common to both the CD11b^low to DP and the DP to CD27^low maturation stages (Fig. 6). Moreover, the activity of all but one of these common genes, namely Tlr3, followed a CD11b^low→ DP → CD27^low mNK cell development program as they either followed a CD11b^low > DP > CD27^low or CD11b^low < DP < CD27^low hierarchy (Fig. 6). Importantly, some of the genes predicted to be differentially activated in this analysis are also upstream regulators for DEGs located within the Nkfm loci (Tables I, II). For instance, our analysis revealed that Trp53 is a common upstream regulator predicted to be differentially activated between the CD11b^low and DP mNK cell subsets (Figs. 5A, 6, Table I) as well as between the DP and CD27^low mNK cell subsets (Figs. 5B, 6, Table II). In addition, our analysis revealed that Ifng is a common upstream regulator predicted to be differentially activated between the CD11b^low and DP mNK cell subsets (Fig. 5A), based on the differential gene expression of many genes, yet is not differentially activated between the DP and CD27^low mNK cell subsets (Fig. 5B). This suggests that Ifng may affect the early stages of NK cell functional maturation. On the contrary, Tnf may regulate later steps as it is a common upstream regulator predicted to be differentially activated between the DP and CD27^low mNK cell subsets, but not the CD11b^low and DP mNK cell subsets (Fig. 5, Table II). Altogether, this analysis revealed upstream regulators of DEGs that are encoded within Nkfm loci, thereby highlighting these downstream targets as interesting candidate genes for specific stages of NK cell functional maturation.

Although our understanding of the role for miRNAs in the regulation of NK cell development is still limited (40), they nevertheless remain important potential candidates for the regulation of mNK cell functional maturation. Based on associations with their putative target genes, miRNAs can be predicted to be activated or repressed during distinct biological conditions by using microarray data (41). Therefore, we executed an additional Ingenuity-based URA that would predict the activation state of upstream miRNAs based on their effect on the expression of downstream target genes. The analysis predicted that several miRNAs were differentially activated between the CD11b^low versus DP and DP versus CD27^low mNK cell subsets (Fig. 7). Indeed, the URA revealed a general increase in predicted miRNA activity during the transition from the CD11b^low to the DP stage (Fig. 7A), followed by a general decrease in predicted miRNA activity during the transition to the final CD27^low stage of mNK cell functional maturation (Fig. 7B). Altogether, this analysis predicts miRNAs as upstream regulators of DEGs between the functional maturation stages of mNK cells, with the DP subset presenting the most active miRNA activity.

Our analysis positioned the function of many genes and miRNAs along the NK cell functional maturation process. It also revealed genes that had not previously been associated with NK cell functional maturation. For instance, Trp53 is a candidate gene located within Nkfm3.1, associated with the CD11b^low mNK cell subset (Supplemental Table I). In addition, Trp53 is a common upstream regulator predicted to be differentially activated between mNK cell subsets based on the global gene expression changes (Figs. 5, 6). It is also highlighted as an upstream regulator for various candidate genes encoded within the specific Nkfm loci (Tables I, II). Together, these results point to Trp53 as a highly interesting candidate gene. As a result, we aimed to further characterize how Trp53 and its gene product, p53, influence mNK cell functional maturation. We first comparatively assessed the expression of p53 in mNK cell functional maturation stages from B6 and NOD mice. We reproducibly observed a lower level of p53 expression in DN, CD11b^low, and DP mNK cells from NOD mice relative to B6 mice (Fig. 8A, Supplemental Fig. 2). In addition, preliminary quantitative PCR analysis suggests that mRNA expression of relevant p53 target genes was lower in NK cells from NOD mice relative to B6 mice (data not shown). We next evaluated the impact of Trp53 deletion on NK cell functional maturation by comparing the proportion of the DN, CD11b^low, DP, and CD27^low mNK cell subsets between B6 and B6.Trp53^−/− mice. As for NOD mice, we observed a significant increase in the proportion of the immature CD11b^low mNK cell subset, and a respective significant decrease in the most mature CD27^low mNK cell subset in B6.Trp53^−/− mice relative to B6 mice, indicative of a defect in functional maturation of NK cells in Trp53-deficient mice (Fig. 8B). Therefore, these results demonstrate that Trp53, a candidate gene linked to the regulation of the CD11b^low mNK cell subset, is indeed relevant to NK cell functional maturation.

Trp53, although mostly investigated for its role as guardian of the genome and its implication in cancer, is also involved in the regulation of many cellular processes (42–44). To determine how Trp53 deletion influences NK cell functional maturation, we examined the proliferation of NK cells using the Ki-67 proliferation marker. Interestingly, a higher proportion of B6.Trp53^−/− NK cells expressed Ki-67 than those from B6 mice (Fig. 9A, top panels). Expectedly, within the different NK cell functional maturation stages, we confirm that the CD11b^low subset represents the stage with the highest proliferative rate, with an average of 24% of Ki-67⁺ cells in B6 mice (Fig. 9A). Trp53 deficiency substantially increased the proportion of Ki-67⁺ CD11b^low mNK cells to an average of 37%. The proportion of Ki-67⁺ cells was also considerably enhanced in B6.Trp53^−/− DP NK cells relative to those from B6 mice (Fig. 9A). The high level of proliferation of both CD11b^low and DP NK cell functional maturation stages may contribute, at least in part, to the apparent defect in NK cell functional maturation in B6.Trp53^−/− mice. In addition to the cell cycle, we quantified the proportion of apoptotic NK cells. We found that, as well as the enhanced cell cycle, total NK cells from B6.Trp53^−/− mice exhibited a higher level of apoptosis (Fig. 9B, top panel). This was mostly attributed to the later stage of NK cell functional maturation, namely the CD27^low mNK cell subset. Together, the enhanced cell cycle at the early and the enhanced apoptosis at the late NK cell functional maturation stages agree with the global phenotype observed in B6.Trp53^−/− mice, wherein we find an increase in the CD11b^low and a decrease in the CD27^low mNK cell subsets.

A four-stage model of NK cell functional maturation has been established based on the expression of CD11b and CD27, separating mNK cells into four distinct subsets that follow a DN → CD11b^low → DP → CD27^low differentiation process (3). Although various genes have been demonstrated to play a role in the regulation of NK cell maturation, the precise stage at which these genes are involved remains to be fully elucidated. In this study, by exploiting the B6 and NOD inbred strains, we identified genetic determinants that regulate the transition of mNK cells between the DN, CD11b^low, DP, and CD27^low stages of functional maturation. Indeed, we demonstrate that NOD mice exhibit a block in NK cell functional maturation in comparison with B6 mice in both a Rag1-sufficient and -deficient setting. Subsequently, we used an unbiased linkage analysis to reveal six novel loci, which we named Nkfm1 to Nkfm6, associated with variations in the proportion of the four distinct mNK subsets. Notably, none of these loci coincide with those linked to the proportion of mNK cells from our recent study, suggesting that mNK cell functional maturation and size of the mNK cell pool are regulated by distinct molecular pathways (16). More specifically, we found that the six loci were independently associated with distinct subsets, suggesting that the block in mNK cell maturation occurs at multiple stages during the four-stage maturation process, where each transition is regulated by different factors. To delimit the list of candidate genes encoded within each of the Nkfm loci, we took advantage of Affymetrix microarray data submitted to the GEO repository by Chiossone et al. (3), for which the CD11b^low, DP, and CD27^low mNK cell subsets from B6.Rag2^−/− mice were analyzed. We performed various in silico analyses that allowed us to uncover candidate genes that are not only linked to a block in NK cell functional maturation but are also differentially expressed between the functional maturation stages themselves. Altogether, the combination of our linkage analysis and the in silico analyses using data submitted to the GEO repository allowed us to highlight various genes, including transcription factors, receptors, proteins, enzymes, and kinases, as well as miRNAs as candidate genes regulating the specific stages of NK cell functional maturation.

Our analysis not only uncovered novel candidate genes, but also highlighted genes that are known to directly or indirectly influence NK cell maturation and positioned their role across the specific NK cell functional maturation steps. For instance, Flt3l located within Nkfm2, which is linked to the CD11b^low mNK cell subset, is crucial for the development and homeostasis of dendritic cells (DCs), which are known to have a close connection to NK cells through a process termed NK-DC cross-talk (38). Castillo et al. (45) have shown that, during this interaction, DCs can trans-present IL-15 to NK cells, where DC-mediated IL-15 trans-presentation plays a role in the development of mNK cells during the early stages of the maturation process, in particular the development of CD11b^low mNK cells. Therefore, the administration of recombinant Flt3L, which leads to the drastic expansion of DCs, consequentially gives rise to an NK cell expansion in an IL-15–dependent fashion (46). Importantly, our linkage analysis not only supports the role for Flt3l in the maturation of NK cells, but also identifies at which stage it plays its role. Indeed, our analysis reveals Flt3l to be a candidate gene located within the Nkfm2 locus linked to the proportion of the CD11b^low mNK cell subset, in accordance with findings by Castillo et al. (45). By taking advantage of the GEO repository data, our analysis revealed that the Flt3l gene is in fact differentially expressed between CD11b^low and DP mNK cells, whereas IL-7 and Tlr3, which are both upstream regulators of Flt3l, are also predicted to be differentially activated between CD11b^low and DP mNK cells. Altogether, these results suggest that Flt3l may regulate mNK cell functional maturation specifically at the CD11b^low stage.

Myb is a transcription factor that has been shown to negatively regulate the terminal stages of functional maturation (3, 8). Our linkage analysis uncovers Myb as a candidate gene located within Nkfm4.1 linked to the DP mNK cells. Moreover, by taking advantage of the GEO repository, we uncovered a predicted decrease in the activity of Myb, as well as a 5.602-fold decrease in expression of the Myb gene, between the DP and CD27^low functional maturation stages of mNK cells in the Rag2-deficient B6 mouse. Our analysis correlates with previous findings demonstrating that the expression of Myb is lowest in the CD27^low subset of mNK cells (3, 8), in agreement with a role for Myb in the later stages of mNK cell functional maturation.

Tgfbr1, which encodes for a TGF-β receptor, is a candidate gene located within the Nkfm6 locus linked to the most mature CD27^low mNK cell subset. Indeed, in the absence of TGF-β receptor signaling, the mNK cell population is comprised of a higher proportion of CD27^low mNK cells, which is thought to be a result of the effect of TGF-β on T-bet and GATA-3 expression (34), both of which are involved in NK cell maturation (3, 4, 10, 12). Interestingly, our linkage analysis also revealed Tbx21, which encodes for T-bet, as a candidate gene located within the Nkfm3.2 locus on chromosome 11 that, like Nkfm6, is linked to the CD27^low mNK cell subset. These findings correlate with those of Chiossone et al. (3), who found Tbx21 expression to be highest in the CD27^low mNK cell subset. T-bet is a transcription factor crucial for multiple stages of NK cell development including the developmental stability of immature NK cells as well as the functional maturation of mNK cells, where T-bet–deficient mNK cells do not mature to the final CD27^low stage (10, 12). In addition to Tgfbr1, Mbd2, which is located within Nkfm1.2 linked to the CD11b^low mNK cell subset, is yet another candidate gene revealed by our linkage analysis that is known to regulate T-bet expression (47). Altogether, these findings support an important role for the TGF-β signaling pathway and Tbx21 in the regulation of the CD27^low mNK cell subset during functional maturation.

Additionally, our analysis provides compelling evidence that Trp53 is an important regulator of NK cell maturation. Trp53, also known as tumor suppressor p53, is crucial in preserving genomic stability following cellular damage or stress, including DNA damage, hypoxia, and oncogene activation, such that p53 provides protection from tumor development (48). More specifically, p53 functions as a transcription factor to modulate the expression of various genes that regulate the major defenses against tumor growth, including cell cycle arrest, apoptosis, maintenance of genetic integrity, inhibition of angiogenesis, and cellular senescence (48). Our analysis revealed Trp53 as a candidate gene located within Nkfm3.1, which is also associated with the CD11b^low mNK cell subset. In this study, we find that p53 expression is lower in NOD than in B6 mice. Moreover, we demonstrate that Trp53 is indeed associated with the regulation of NK cell functional maturation, where Trp53-deficient mice exhibit a decrease in functional maturation of NK cells. The decrease in NK cell functional maturation is associated with heightened proliferation of CD11b^low mNK cells and increased apoptosis of CD27^low mNK cells, suggesting that Trp53 deficiency regulates NK cell functional maturation in an NK cell–intrinsic manner. Still, these data do not exclude an NK cell–extrinsic role for Trp53 in modulating NK cell functional maturation. Future experiments exploiting either competitive bone marrow chimeras or mice bearing NK cell lineage-specific Trp53 deletion will delineate the cell-intrinsic and -extrinsic role of Trp53 for this phenotype.

Interestingly, Trp53 is the most frequently mutated gene in human cancer, with greater than half of all tumors exhibiting a mutation at this locus (48). Indeed, there is growing evidence that these Trp53 mutations can result in a loss of tumor suppressor activity as well as the acquisition of functions that contribute to the malignant progression (49). Given the role of Trp53 as a tumor suppressor gene and its associated mutations’ causative role in cancer, it is possible that patients harboring specific polymorphisms in Trp53 may also have alterations in NK cell functional maturation and subsequent defects in tumor immunosurveillance. Hence, investigation of the proportion of mNK cell subsets in patients harboring Trp53 polymorphisms would be of great interest. To that effect, various polymorphisms in both validated and putative p53 response elements, which would alter p53 binding and consequently the transcription of the given gene, have been identified in humans (50).

miRNAs are also emerging as regulators of NK cell functional maturation. Indeed, recent work in animals lacking either global miRNAs (51, 52) or specific miRNAs, namely miR-150, miR-155, and miR-15a/16 (8, 9, 14), has highlighted miRNAs as regulators of NK cell functional maturation. During our in silico analysis, we uncovered multiple miRNAs as upstream regulators of differentially expressed downstream target candidate genes, where four of these miRNAs are also expressed by NK cells, namely let-7, mir-21, miR-339-5p, and miR-142-3p (53). Accordingly, members of the let-7 family, namely let-7f, let-7g, and let-7a, were found to be the three most abundant miRNAs in human NK cells (54), whereas the expression of miR-339-5p increases by almost 3-fold following IL-15 stimulation of murine NK cells (53). Furthermore, mir-21, which is the immature form of miR-21, is the most abundant miRNA expressed in both resting and activated murine NK cells (53). Based on our analysis, let-7 exhibits a predicted increase in activity during the transition from CD11b^low to DP mNK cells, whereas miR-339-5p and miR-142-3p exhibit a predicted decrease during the transition from the DP to CD27^low mNK cell stage. Moreover, miR-21 exhibits a predicted increase in activity during the transition from the DP to the most mature CD27^low mNK cell stage. Interestingly, mice deficient for either Dicer or Dcgr8, which are miRNA processing enzymes, display a reduction in both the relative expression of miR-21 in NK cells as well as the frequency of the most mature CD27^low mNK cell subset (51, 52). Together, these results suggest that miR-21 regulates the functional maturation of mNK cells to the final CD27^low stage. It will therefore be relevant to verify exactly how these miRNAs influence the functional maturation of NK cells.

Interestingly, many of the candidate genes highlighted by our linkage approach are associated with cellular apoptosis. For instance, Myb oncogenic variants suppress apoptosis, predominantly by enhancing the expression of its direct target, the anti-apoptotic gene Bcl2 (55, 56), whereas TGFβ signaling induces apoptosis through Smad-mediated expression of DAP-kinase (57). p53, which is associated with regulating apoptosis following stress signals (48), also modulates the expression of downstream target genes that can be linked to apoptosis. Accordingly, Pmaip1 has been previously demonstrated to act together with Bim to induce apoptosis of NK cells following IL-15 withdrawal in vitro (58), suggesting that Pmaip1 influences NK cell survival. Another candidate gene, Flt3l, is crucial for the development and homeostasis of DCs whose IL-15 trans-presentation plays a role in the development of CD11b^low mNK cells (45). As IL-15 is known to support the survival of NK cells in the peripheral lymphoid organs (59), this suggests that Flt3l may regulate NK cell functional maturation by promoting IL-15 trans-presentation. Lastly, Yamanaka et al. (60) recently showed that inhibition of miR-21 in a human NK cell line led to increased apoptosis associated with the upregulation of proapoptotic downstream targets, suggesting that miR-21 is a key regulator of NK cell survival. Altogether, these results suggest that several candidate genes may influence NK cell functional maturation via the regulation of apoptosis. Interestingly, this may be in part due to the model system exploited. Indeed, Rag deficiency influences NK cell functional maturation and increases their rate of apoptosis (61, 62). A linkage analysis in Rag-sufficient mice would likely yield additional loci and biological pathways linked to the regulation of NK cell functional maturation. Nevertheless, we show that Trp53, even if revealed in a Rag-deficient setting, clearly influences NK cell functional maturation in a Rag-sufficient setting. This suggests that, as for other traits (16, 30, 63), Rag^−/− mice are useful in linkage analyses to identify trait-relevant loci, biological pathways, or even genes.

There are several additional genes located within the Nkfm loci that exhibit a known association with NK cells and thus remain likely candidate genes for the regulation of NK cell functional maturation. For instance, Arrb2 (Nkfm3.1) (64), Lgals9 (Nkfm3.1/3.2) (65), Ccl3 (Nkfm3.1/3.2) (66), Ccl4 (Nkfm3.1/3.2) (67), Ccl5 (Nkfm3.1/3.2) (68), Mpo (Nkfm3.1/3.2) (69), and Cd72 (Nkfm6) (70) play a role in NK cell function, whereas CD74 (Nkfm1.1/1.2) (71) has been associated with NK cell numbers. Moreover, Arg1 (Nkfm4.1/4.2) has been associated with both NK cell function and numbers (72). Nevertheless, the role of other candidate genes that do not yet have any known association to NK cells will need to be further explored.

In conclusion, maintenance of a pool of mNK cells with optimal effector function is essential for host defense against both pathogens and cancerous tumor formation, where each mNK cell subset, namely DN, CD11b^low, DP, and CD27^low mNK cells, exhibits a unique combination of phenotype and effector function. This study brings to light six novel loci associated with a multi–check-point block in the functional maturation of mNK cells, where in silico analyses further delimit the list of potential candidate genes associated with this phenotype. Importantly, our analysis not only uncovered novel candidate genes, but also highlighted genes that are known to influence NK cell maturation and positioned their role across the specific NK cell functional maturation steps. In the future, it will be relevant to verify the impact of the candidate genes in additional mouse strains. In this study, we also validate Trp53 as a gene regulating mNK cell functional maturation as B6.Trp53^−/− mice fail to maintain normal mNK cell subset proportions. Given the causative role of Trp53 mutations in cancer as well as the identification of polymorphisms within the p53 response elements, it will be of great interest to verify whether these patients exhibit any alterations in NK cell functional maturation, which may lead to subsequent defects in tumor immunosurveillance. We speculate that two populations would be particularly informative in this regard: 1) elephants, which have 20 copies of TRP53 and a low incidence of cancer (73), and 2) patients with Li–Fraumeni syndrome, most of which have a germline TRP53 mutation and a highly penetrant familial cancer syndrome (74). Genes identified in this study may eventually serve as molecular targets to permit the modulation of mNK cell functional maturation, more specifically the alteration of the frequency of a given mNK cell subset, to potentiate both tumor immunosurveillance and viral clearance.

We thank Marie-Josée Guyon, Fany De Wilde, and the animal house staff for curating the mouse colonies and the center of Applied Genomics at Sick Kids Hospital in Toronto for SNP genotyping. We thank Dr. Tarik Moröy for providing some B6.Trp53-deficient mice, Martine Dupuis at the core flow cytometry facility at Maisonneuve-Rosemont Hospital for assistance, as well as Dr. Nathalie Labrecque and Dr. Heather Melichar for critical review of the manuscript. This work benefited from data assembled by the Walzer Laboratory (3).

The authors have no financial conflicts of interest.