FAM13A regulates maturation and effector functions of natural killer cells

The education or licensing process is essentially required for the proper anti-tumor function of natural killer (NK) cells. Although several models for education have been proposed, the genetic factors regulating these processes still remain largely elusive. Here we show that FAM13A (family with sequence similarity 13, member A), strongly linked to the risk of prominent death-causing lung diseases, i.e., lung cancer and chronic obstructive pulmonary disease, critically modulated NK cell maturation and effector functions. Fam13a depletion promoted NK cell maturation, KLRG1 (killer cell lectin-like receptor G1) expression in NK cells and NK terminal differentiation in homeostatic mice. NK cells from Fam13a-deficient mice had impaired IFN-γ production and degranulation. Strikingly, the number of lung metastases induced by B16F10 melanoma cells was increased in Fam13a-deficient mice. Collectively, our data reveal a pivotal role of FAM13A in slowing down NK maturation, but promoting NK cell effector functions and immune protection against tumor metastasis.


Introduction
The functions of natural killer (NK) cells are governed through a panel of germline-encoded activating (AR) and inhibitory (IR) receptors (Caligiuri 2008, Demaria et al. 2019. The balance of signals transmitted through these receptors determines the fate of the target cells, which are killed when the activation prevails over the inhibition but spared when the contrary is the case. However, this balance is not enough to render NK cells functional: they must, in addition, be "licensed" or "educated". This education process takes place when a developing NK cell expressing an IR specific for an autologous major histocompatibility complex (MHC) class I molecule in the microenvironment interacts with this structure Tian 2017, Boudreau andHsu 2018). The consequence is that a NK cell is able to efficiently detect diseased cells that have downregulated their MHC class I expression (such as cancer cells and infected targets), without killing healthy cells. In human, these IR are mainly the killer immunoglobulin-like receptors (KIR) and the C-type lectin CD94/NKG2A, whereas the mouse counterparts are the Ly49 family and the mouse orthologue of human NKG2A is also present. In parallel, several MHC class I-independent IR have likewise been described to educate NK cells (He and Tian 2017). A seminal work has demonstrated the licensing concept in mice by showing that Ly49C + NK cells, specific for the MHC class I ligands H-2K b and H-2D b present in C57BL/6 (B6) mice, are better licensed than the Ly49Ccells from the same individuals (Kim et al. 2005). The same work also reported that upon short activation with the TLR3 agonist poly I/C or stimulation by the cytokine interleukin (IL)-2, the difference between the two subsets becomes less pronounced but is still present (Kim et al. 2005).
We aimed to get further insight into this topic, because in immunotherapeutic approaches with human activated autologous or allogeneic NK cells, it might be of advantage to select those that are best educated and therefore most efficient in tumor cell killing. For this purpose, we included for simplicity the known MHC class I specific IR expressed by NK cells from the B6 (H-2 b ) background, namely Ly49C (ligands: H-2K b , H-2D b ), Ly49I (same ligands), and NKG2A (ligand: Qa-1 b , the mouse orthologue of HLA-E, presenting signal peptides derived from the classical MHC class I molecules, which are H-2D b and H-2K b in B6 mice) (Boudreau and Hsu 2018). In contrast to the other works mainly based on resting or short-term or IL-2-stimulated NK cells (Kim et al. 2005), we here investigated the education process of the four subsets defined by the combinatory expression of NKG2A and Ly49I/C using different stimulation conditions with various cytokine cocktails. Those cytokine cocktails versus (vs.) IL-2 alone better mimic the tumor microenvironment. Furthermore, since the involvement of the immune system, and particularly of NK cells, in the fight against tumor genesis and metastasis is crucial (Vesely et al. 2011, Zimmer 2014, Cerwenka and Lanier 2016, Janssen et al. 2017, Huntington et al. 2020, we sought to identify and investigate whether a novel candidate gene, which is known to be associated with the risk of lung tumors, regulates the education process of NK cells. This is rationale as profound phenotypic and functional alterations of NK cells have been shown in lung cancer (Al Omar et al. 2011, Platonova et al. 2011, Cong and Wei 2019, Hervier et al. 2019) and particularly, in nonsmall cell lung cancer, several specific KIR genes are associated with better treatment response and longer survival (Kusnierczyk 2013).
Meanwhile, to further narrow down the candidate list, we assessed the genes also linked with the risk of another major lung disease, i.e., chronic obstructive lung disease (COPD), the pathogenesis and progression of which NK cells also significantly contribute to (Eriksson Strom et al. 2018, Pascual-Guardia et al. 2020. To this end, the gene FAM13A (family with sequence similarity 13, member A), a well-recognized risk gene, nonetheless less-characterized in the immune system, for both COPD and lung cancer in human came to the center of the study. Several independent genome-wide association studies (GWAS) among various populations have shown the strong link between FAM13A, COPD and lung cancer (Cho et al. 2010, Guo et al. 2011, Young et al. 2011, Kim et al. 2015, Ziółkowska-Suchanek et al. 2015, Hirano et al. 2017, van der Plaat et al. 2017, Ziółkowska-Suchanek et al. 2017, Zhang et al. 2018, Castaldi et al. 2019, Yu et al. 2019. The polymorphism of FAM13A was also associated with cystic fibrosis (Corvol et al. 2018). Functional studies have revealed that Fam13a depletion reduces the susceptibility to COPD via inhibiting the WNT/β-catenin pathway in a mouse model (Jiang et al. 2016). In human lung tumor cell lines, knocking-down FAM13A reduced tumor cell proliferation but induced cell migration in vitro (Eisenhut et al. 2017). However, the exact in vivo physiological role of Fam13a in the complicated process of tumor onset and metastasis, where different types of cells are heavily involved, still remains mysterious. Furthermore, in the same study, the authors observed the upregulation of FAM13A in CD4 + CD25effector T cells but reduced expression in T regulatory cells (Tregs) of human blood. In contrast, in our previous study (He et al. 2012), the expression of FAM13A was increased in human Tregs vs. CD4 + effector T cells. Another work with the expression quantitative trait loci (eQTL) method has briefly investigated the functional effect of FAM13A on naïve CD4 + T cells in vitro using siRNA (Schmiedel et al. 2018). Meanwhile, microRNA-328 in M2 macrophage-derived exosomes has been demonstrated to regulate the progression of pulmonary fibrosis via Fam13a in an animal model (Yao et al. 2019). We therefore hypothesized that Fam13a might also play a vital role in regulating immune functions in vivo, although it is still unclear in which types of immune cells.
Here we utilized a Fam13a whole-body knockout (KO) mouse model to study the in vivo impact of Fam13a on NK cell phenotypes and functions, in addition to the cellular phenotypes of T and B lymphocytes. We found that Fam13a did not affect the homeostatic composition and activities of total B cells, CD4 + T cells, CD4 + Tregs and CD8 + T cells. Interestingly, Fam13a depletion changed the NK cell maturation profile and impaired the effector functions of NK cells by controlling IFN-γ production and degranulation. Notably, Fam13a depletion exacerbated lung metastasis induced by B16F10 melanoma cells in the B6 mouse model. Altogether, our results provide strong evidence that Fam13a is an important component regulating the effector functions of NK cells, mostly through modulating their maturation and IFN-γ production.

Characterization of the immune phenotpye of Fam13a KO mice
To identify the potential effect of Fam13a on the homeostatic phenotype of the immune system, we first analysed the composition of different major immune cell types in various lymphoid organs and tissues of Fam13a KO mice [Fam13a tm2a (KOMP)Wtsi ] obtained from the Knockout Mouse Project vs. the age-and gender-matched littermate controls (wildtype, WT). As a starting point, we compared the mRNA expression of Fam13a in the lung tissue between Fam13a KO and WT and that in NK cells isolated from spleen. As expected, Fam13a KO mice relative to WT littermates exhibited a significant reduction in the transcript expression of Fam13a in both the lung tissue (Fig. 1a) and NK cells (Fig. 1b). We checked the frequency of CD19 -CD3 -NK1.1 + NK cells in several relevant tissues including spleen (Fig.   1c, d), peripheral lymph nodes (pLNs) (Extended Data Fig. 1a), bone marrow (BM) (Extended Data   Fig. 1b) and lung (Extended Data Fig. 1c) and no significant difference between homeostatic Fam13a KO and WT mice was observed. We also could not see any significant differences in the frequencies of several other lymphocytes, such as CD3 + T cells (Fig. 1e), CD19 + B cells (Fig. 1f), CD8 + T cells (Fig.   1g), CD4 + T cells (Fig. 1h), as well as FOXP3 + CD4 + Treg cells (Fig. 1i). We further characterized the naïve and memory compartment of CD4 + and CD8 + T cells. Again, no significant difference was found for the percentage of naïve (CD44 low CD62L high ) and effector memory (CD44 high CD62L low , EM) CD4 + T cells (Fig. 1j, k) and naïve, EM and central memory (CD44 high CD62L high , CM) CD8 + T cells in the tested age (8-12 wks) (Fig. 1l, m). In conclusion, Fam13a depletion does not lead to spontaneous abnormalities in the development of various major immune cells and no obvious inflammatory symptoms or immunodeficiency were observed in homeostatic Fam13a KO mice.
To further evaluate whether Fam13a influences the basic functions of those immune cells, we analysed the homeostatic activation and proliferation markers of CD4 + and CD8 + T cells. No significant differences in the frequency of the CD69 + population among CD4 + T cells and CD8 + T cells in spleen (Extended Data Fig. 1d, e) exhibited between Fam13a KO and WT mice. For another activation or exhaustion marker, PD-1, we also couldn't find any significant difference in the percentages of PD-1 + CD4 + and PD-1 + CD8 + T cells between Fam13a KO and WT mice (Extended Data Fig. 1f, g). To analyze cell proliferation under homeostasis, we examined Ki-67 expression among CD4 + and CD8 + T cells. no difference in the percentages of Ki-67-expressing cells was noticed among total CD4 + T cells (Extended Data Fig. 1h) and among total CD8 + T cells (Extended Data Fig. 1i) between Fam13a KO and WT mice. For Tregs, we evaluated the suppressive function by co-culturing the CFSE-labelled CD4 + conventional T cells (Tconv), antigen-presenting cells (APCs) and Tregs. Loss of Fam13a did not compromise Treg suppressor function against Tconv proliferation (Extended Data Fig. 1j). In short, Fam13a depletion causes deficiency neither in the development, nor in the activation or proliferation of major lymphoid subsets such as CD4 + and CD8 + T cells, at least in mice under homeostatic conditions.  j, Representative FACS plot of naïve (CD62L high CD44 low ) and effector memory (EM) (CD62L low CD44 high ) T cells among total CD4 + T cells in spleen of Fam13a KO and WT littermates. k, Percentages of naïve and EM CD4 + T cells among total CD4 + T cells in spleen of Fam13a KO and WT littermates (KO, n=5, WT, n=5). l, Representative FACS plot of naïve (CD62L high CD44 low ), EM (CD62L low CD44 high ), central memory (CM) (CD62L high CD44 high ) CD8 + T cells among total CD8 + T cells in spleen of Fam13a KO and WT littermates. m, Percentages of naïve (CD62L high CD44 low ), EM (CD62L low CD44 high ), central memory (CM) (CD62L high CD44 high ) CD8 + T cells among total CD8 + T cells in spleen of Fam13a KO and WT littermates (KO, n=5; WT, n=5). Results represent two (a, b), three (d, e, f) and four (g, h, i, k, m) independent experiments. Data are mean ± standard deviation (s.d.). The p-values are determined by a two-tailed Student's t-test. n.s. or unlabeled, not significant, *p<=0.05, **p <=0.01 and ***p <=0.001.

Fam13a regulates NK cell maturation
Similar to T cells, no significance difference was observed in the activation marker CD69 in NK cells in spleen and pLNs between Fam13a KO and WT littermates (Extended Data Fig. 1k, l). To explore whether Fam13a affects other NK cell parameters, we investigated the maturation profile of NK cells by checking the co-expression of CD27 and CD11b. Interestingly, we noticed that Fam13a depletion led to a modestly reduced proportion of immature CD11b -CD27 + (Fig. 2a, b), but significantly increased percentages of mature CD11b + CD27 -NK cells (Fig. 2a, b). Since CD11b + CD27 -NK cells are terminally differentiated NK cells, representing the major KLRG1-expressing NK subset , Chiossone et al. 2009, Elpek et al. 2010, we also analyzed the expression of the inhibitory receptor and maturation marker KLRG1 among NK cells. In line with the data related to CD27 and CD11b subsets, we also observed a much higher percentage of KLRG1-expressing cells among total NK cells in both spleen (Fig. 2c, d) and pLNs (Fig. 2e, f).
We then analyzed another important aspect of the NK functions, i.e., NK cell education which is crucial for the acquisition of its effector properties (Höglund and Brodin 2010). Interestingly, the two Ly49C/I negative subsets among the four subsets defined by the combination of presence or absence of NKG2A and Ly49C/I, the two crucial IR related to education in the C57BL/6 (H-2 b ) genetic background, represented the major fractions of NK cells in the splenocytes of WT mice (Fig.2g, h). Fam13a depletion did not cause a significant change in the fraction of any of the four subsets defined by the combinatory expression of NKG2A and Ly49C/I (Fig. 2g, h). We also checked the expression of the maturation marker KLRG1 among the four subsets of NK cells in WT spleen and found that KLRG1 was mainly expressed among the two Ly49C/I + subsets (Fig. 2i). Notably, again in line with the effect of Fam13a on NK cell maturation, a significant upregulation of KLRG1 was observed on several NK subsets, including NKG2A + Ly49C/I + , NKG2A -Ly49C/I + and NKG2A -Ly49C/I -NK subpopulations (Fig. 2i), but not on the NKG2A single positive cells.
To get a more comprehensive picture, we further analyzed different IR and AR of NK cells, whose engagement is critical to regulate and balance NK cell activities. For the AR, the frequency of NKp46 + , Ly49H + and Ly49D + NK cells (Extended Data Fig. 2a-   i, Percentages of KLRG1 + cells in NK cell subpopulations in the spleen based on the combinatory expression of NKG2A and Ly49C/I in spleen of WT mice or Fam13a KO vs. WT mice (KO, n=7; WT, n=5). Results represent three (g, h, i) and four (b, d, f) independent experiments. Data are mean±s.d. The p-values are determined by a two-tailed Student's t-test. n.s. or unlabeled, not significant, *p<=0.05, **p<=0.01 and ***p<=0.001. For multiple comparisons, Turkey's correction one-way Anova was applied.

Fam13a controls NK-cell IFN-γ production
Different developmental stages of NK cells possess different phenotypic features and effector functions Smyth 2006, Fu et al. 2011). Since we observed a significant upregulation in the frequency of the terminally differentiated NK subset, namely, CD27 -CD11b + NK cells, we further checked other effector functions of NK cells, including IFN-γ production, one of the most critical roles of NK cells. We expanded NK cells by culturing total splenocytes isolated from Fam13a KO and WT littermates in the presence of a high concentration of IL-2 for 5 days (Fig. 3a). We found that Fam13a deficiency did not affect the survival and expansion of NK cells in vitro (Fig. 3b, c). Compared with the ex vivo frequency, following IL-2 expansion, the NKG2A -Ly49C/Iand NKG2A + Ly49C/Isubsets still occupied the major portion of the NK cells while the double negative cells represented the most dominant subset (Fig. 3d, e). Alike that in the homeostatic condition, the four NK-cell subpopulations based on the combinatory expression of the two major IR NKG2A and Ly49C/I, also presented no obvious difference in the expanded NK cells from Fam13a KO or WT mice (Fig. 3f). However, when stimulated with the cytokine cocktail including IL-2, IL-12 and IL-15, Fam13a-deficient NK cells produced a significantly lower amount of IFN-γ compared with WT counterparts (Fig. 3g, h). We further checked the production of IFN-γ among the four NK subsets defined by the combinatory expression of NKG2A and Ly49C/I. In WT mice, among the four subsets the NKG2A -Ly49C/Icells expressed the lowest amount of IFN-γ while the two NKG2A + subpopulations expressed much higher levels of IFN-γ ( Fig. 3i). The IFN-γ production of each of the four subpopulations defined by the combination of Likewise, even when stimulated with the strong-activation cytokine cocktail composed of IL-2, IL-12 and IL-18, Fam13a-deficient NK cells still exhibited impaired IFN-γ production capacity (Fig. 3k, l).
This impairment was mainly reflected in the NKG2A -Ly49C/I + and NKG2A -Ly49C/Isubpopulations ( Fig. 3m). Again similar to that stimulated by the cytokine cocktail with IL-15, the NKG2A -Ly49C/Icells expressed the least amount of IFN-γ even in WT NK cells following IL-18 cocktail stimulation ( Fig. 3n). Taken together, our data demonstrate not only that the two NKG2A + NK cells produce the highest amount of IFN-γ, representing the best 'educated' ones, but also Fam13a depletion impairs IFNγ production in NK cells under various cytokine-cocktail stimulations in vitro.

Fam13a modulates NK-cell degranulation and IFN-γ production against tumor cells
To further check the killing capacity of NK cells towards tumour cells, we pre-activated NK cells in vivo by injecting the TLR3 agonist Poly (I:C) into Fam13a KO and WT littermates (Fig. 4a). We then checked the expression of IFN-γ and of the degranulation marker CD107a in NK cells following the incubation with the target tumor cell line YAC-1. CD107a is a degranulation marker of NK cells and CD8 + T cells, which reflects their cytotoxic activity (Alter et al. 2004, Aktas et al. 2009). No clear difference was observed in the percentages of CD107a-expressing cells among total NK cells (Fig. 4b,   c) and in the absolute number (Fig. 4d) of CD107a + NK cells between Fam13a KO and WT mice following Poly (I:C) injection. However, loss of Fam13a caused a decreased degranulation capacity per individual NK cell against YAC-1, as indicated by the decreased MFI of CD107a in NK cells (Fig. 4e).

Fam13a deficiency promotes lung metastasis induced by melanoma cells in vivo
Having shown that Fam13a depletion impaired several effector functions of NK cells in vitro or ex vivo, we sought to further investigate its in vivo effects. Since NK cells are critical in the control of metastasis (López-Soto et al. 2017), we induced lung metastasis by transplanting B16F10 melanoma cells. We intravenously injected Fam13a KO and WT littermates with B16F10 melanoma cells and evaluated the lung metastases 16 days post inoculation (Fig. 5a). Compared with Fam13a WT mice, Fam13a KO mice had developed much more tumor metastases in the lung (Fig 5b, c). Since NK cell-mediated control in melanoma metastasis is critical, we first analysed NK cells in the spleen and lung tissues. The frequency and absolute number of NKp46 + NK1.1 + NK cells among CD3 -CD19splenocytes was much lower in melanoma treated Fam13a KO mice compared with those in WT littermates (Fig. 5d-f).
Likewise, Fam13a KO mice also had a lower percentage of infiltrated NKp46 + NK1.1 + NK cells among CD3 -CD19cells in lung in comparison to WT littermates (Fig. 5g). Similar to the homeostatic phenotype of Fam13a KO NK cells (Fig. 2d), a much higher frequency of KLRG1 + cells was also observed among total NK cells in both spleen (Fig. 5h, i) and lung (Fig. 5j) of Fam13a KO mice in comparison with WT littermates. In addition, all the four subpopulations of NK cells based on the expression of the inhibitory receptors NKG2A and Ly49C/I in Fam13a KO mice showed a much higher percentage of KLRG1 + cells (Fig. 5k). KLRG1 is also regarded as a terminal differentiation marker of NK cells ). In the chronic infection model, KLRG1 + NK cells represent an exhausted phenotype and KLRG1 inhibits NK cell IFN-γ production , Müller-Durovic et al. 2016, Alvarez et al. 2019). This indicates that both the decreased number and dysfunction of NK cells in Fam13a KO mice aggravates tumour metastasis and that the dysfunction might be regulated via the enhanced expression of KLRG1.
Since dysregulation of NK cells might also affect T cell-mediated adaptive immunity against tumors, we further checked CD4 + T cells and cytotoxic CD8 + T cells. No big difference was observed for the frequency of CD4 + T cells in spleen between melanoma-treated Fam13a KO and WT littermates (Extended Data Fig. 3a). However, we found higher percentages of both FOXP3 + CD4 + Tregs and CD8 + T cells in spleens of Fam13a KO mice, the former of which is known to deteriorate tumor progression while the latter of which might ameliorate the tumor development (Extended Data Fig. 3b,   c). For naïve and memory compartments of T cells, we observed a higher frequency of effector memory (EM) but a lower percentage of naïve CD4 + T cells in Fam13a KO mice (Extended Data Fig. 3d, e).
However, there were significantly higher percentages of PD-1 + CD4 + T cells (Extended Data Fig. 3f), indicating more exhausted or activated T cells in Fam13a KO mice. Similarly, more EM and central memory (CM) CD8 + T cells as well as PD-1 + CD8 + T cells, although to a lesser extent, but fewer naïve CD8 + T cells were found in Fam13a KO mice (Extended Data Fig. 3g-i). k, Percentage of KLRG1 + cells among four NK subpopulations based on NKG2A and Ly49C/I expression in spleen of Fam13a KO and WT littermates following B16F10 cell inoculation (KO, n=3, WT, n=4). Results represent two (e, f, g, i, j, k) and four (c) independent experiments. Data are mean ± s.d. The p-values are determined by a two-tailed Student's t-test. n.s. or unlabeled, not significant, *p<=0.05, **p <=0.01 and ***p <=0.001. For multiple comparisons, Turkey's correction one-way Anova was applied.

Discussion
In this work, we showed that (i) cytokine-activated NK cells keep a trace of their initial educational profile, and licensing through NKG2A appears as more efficient than that through Ly49C/I, (ii) Fam13a inhibits NK cell maturation, as shown through the up-regulation of CD27 -CD11b + cells in the Fam13A-KO animals, (iii) Fam13a is required for an optimal NK cell IFN-γ production and degranulation, (iv) in vivo, the KO mice develop significantly more lung metastases after intravenous injection of a melanoma cell line, and (v) the absence of Fam13a does not seem to have a phenotypic impact on B and T cell numbers and subset distribution, nor on Treg function (at least under homeostasis).
The NK cell education concept is based on the observation that developing NK cells become functional upon recognition of autologous MHC class I molecules via specific IR. Several models for education have been proposed (Boudreau and Hsu 2018), such as the licensing model (Kim et al. 2005), the (dis)arming model (Elliott and Yokoyama 2011), the rheostat model (the more specific IR the NK cell expresses, the better educated and more functional the cell is) (Brodin and Hoglund 2008), the cis-trans interaction model (IR molecules are engaged with their MHC class I ligand in cis, so that the quantity of IR available for trans interactions is reduced and the activating threshold of the NK cell decreased) (Zimmer et al. 2001, Doucey et al. 2004, Chalifour et al. 2009). Finally, the confinement model was also proposed, taking into account the adhesion molecules (He and Tian 2017). More recently, from a mechanistic point of view, the presence of dense-core secretory lysosomes containing the cytolytic effector protein granzyme B distinguishes educated from uneducated resting NK cells (Goodridge et al. 2019, Pfefferle et al. 2020. Previous observations by Kim et al. have revealed that the IR from the Ly49 family can efficiently license NK cells (Kim et al. 2005). Initially neglected, NKG2A was later demonstrated to likewise educate NK cells in mouse and man (Lisovsky et al. 2015, Meyer et al. 2017, Kristensen et al. 2018. Our data reveal that the proportion of IFN-γ-producing NKG2A + NK cells (either single positive or associated with Ly49C/I) is significantly higher in both WT mice and Fam13A-KO mice than that in Ly49C/I single positive and NKG2A -Ly49C/Idouble negative subsets, after a strong cytokine-mediated activation. We chose the cocktail IL-2/IL-12/IL-15 to discriminate potentially differential effects on the four subpopulations. Interleukin-2 alone served as negative, while the cocktail IL-2/IL-12/IL-18 as positive control, as it gives maximal IFN-γ production. An imprint of the initial education profile was still present in a large fraction of the licensed NK cells even after activation, and the expression of NKG2A seemed to confer a higher frequency of IFN-γ-producing cells than the presence of Ly49C/I. The Fam13a-KO NK cells followed the same curve but with significantly lower frequencies of NK cells that accumulated IFN-γ in their cytoplasm. This effect of Fam13a was even observable in NK cells strongly activated with IL-2/IL-12/IL-18 on a global level, whereas its influence was only clear among the two NKG2Asubpopulations. Furthermore, after an in vivo pre-activation of NK cells via the TLR3 agonist Poly (I:C), the MFI of CD107a as well as the percentage of IFN-γ + NK cells followed the same pattern as the education process, which might be less surprising as the cells were harvested after only one night of in vivo stimulation. Here again, (i) NKG2A was the most efficient licensing IR, and (ii) the KO NK cells were less functional than the WT mice. Our finding that NKG2A "educates" better than the Ly49 family of IR is somewhat contrasting with the observation of another group . In that work, following CRISPR/Cas9-mediated deletion, NKG2A seems to have only a mild effect as opposed to a "moderate" one after the KO of Ly49C and Ly49I, while both groups of IR strongly synergize for an optimal licensing process. However, we observed the most prominent effect of NKG2A after five-day culture in IL-2 and then overnight stimulation with IL-12 and IL-15, whereas Zhang et al. mainly used the Poly (I:C) experimental system. In mice deficient in the CD94 chaperone molecule necessary for the surface expression of the IR NKG2A and the AR NKG2C and NKG2E, no major effects on NK cell functions are observed (Orr et al. 2010), but some doubts exist about the genetic background of those animals ).
The Fam13A-KO animals also displayed a different maturation profile compared to their WT counterparts, as reflected in a much higher frequency of mature CD27 -CD11b + and KLRG1 + NK cells ex vivo, which suggests that Fam13a has an inhibiting or regulatory effect on NK cell maturation, or preventing to some extent premature entry into the senescent state. As the CD27 -CD11b + NK cells are less proficient in cytokine production and cytotoxic activity than CD27 + CD11b + cells (Hayakawa and Smyth 2006), the simple change in proportion might already explain the functional differences between WT and KO mice, at least in the Poly (I:C) experiments. Furthermore, the extracellular matrix glycoprotein fibronectin is able to selectively keep CD11b + NK cells viable in vitro through a direct interaction with CD11b and with downstream Src and β-catenin, resulting in nuclear translocation of βcatenin and ERK activation (Zhang et al. 2009). Interestingly, β-catenin is inhibited by Fam13a (Jiang et al. 2016), so that in WT mice, a useless accumulation of mature and hypofunctional CD27 -CD11b + NK cells might be avoided via this pathway.
Even more striking is the difference between WT and Fam13a-KO mice in terms of KLRG1 expression. This is an IR of the C-type lectin superfamily able to inhibit NK cell functions (cytotoxicity and cytokine production) upon recognition of its non-MHC class I ligands, which are E-, N-and R-cadherins (Ito et al. 2006, Banh et al. 2009, Li et al. 2009). The latter are adhesion molecules downregulated on cancer cells. In mice, KLRG1 is expressed on roughly a third of NK cells, but it increases upon infections (Banh et al. 2009, Li et al. 2009). In Fam13a-KO mice, particularly in spleen, this value was almost doubled in most of the animals, which might again explain, at least in part, the functional results in the Poly (I:C) stimulation conditions. Importantly, the interaction between KLRG1 and E-cadherin is bidirectional and triggers signaling pathways in both the KLRG1-and the E-cadherin-expressing cells, a process called reverse signaling (Banh et al. 2009). Cadherins are again in turn linked to Wnt/β-catenin, so that one might speculate about the outcome of the interaction for the cadherin-expressing immune cells in Fam13a-KO mice, for example macrophages and dendritic cells. On one hand, cadherin + cells would be quite well protected from NK cell-mediated killing. On the other hand, those cells could have increased functional capacities to interact with the adaptive immune system, as Fam13a depletion would potentially lead to higher levels of Wnt/β-catenin signaling at least in epithelial cell lines. This hypothesis goes beyond the scope of the present work, but will have to be addressed in future experiments.
In our in vivo B16F10 melanoma models, we observed a strong reduction in the number of NK cells in spleen and lung of the KO mice compared to their WT counterparts, but a significantly higher frequency of KLRG1 + NK cells. This might suggest at first sight that more NK cells were inhibited, but this would depend on the expression level of the cadherin ligands on the tumor cells. In any case, Fam13a-KO mice displayed a significantly higher number of lung metastases compared to the WT animals, which is in accordance with our in vitro findings, showing a hypofunctional state of KO NK cells. In line with our observation of a higher expression of KLRG1 among NK cells in Fam13a-KO mice, KLRG1 neutralization antibody plus anti-PD-1 combinatory treatment vs. the anti-PD-1 therapy alone has achieved better survival rate and response to tumor volume in B16F10 melanoma models (Greenberg et al. 2019). On the other hand, CD27 -KLRG1 + NK cells control pulmonary metastasis in the CT26 colon carcinoma model (Renner et al. 2014). Meanwhile, in human studies, a higher expression of KLRG1 was associated with a better survival of the patients with lung adenocarcinoma or skin melanoma (Huntington et al. 2020). According to those two reports, we should have observed fewer metastatic nodules in the KO mice. However, the higher percentage of KLRG1 + NK cells in the Fam13a-KO mice would be a compensatory factor to better activate the anti-tumor mechanisms in the situation of a decreased number of NK cells, which otherwise would not work as well as in the WT counterparts. More realistically, KLRG1 + NK cells possibly produced, as previously described, less IFN-γ than KLRG1 -NK cells ). As we also observed some effects of Fam13a on T cells in the induced lung metastasis experiments, further cell-type specific knockout animal models are required to figure out whether our observation is only driven by NK cells or together with other cell types. In short, here we demonstrated a previously unrecognized critical role of FAM13A in the maturation and several effector functions of NK cells using both in vitro and in vivo models. As FAM13A is strongly associated with the risk of several common lung diseases, our discovery paves the way to develop a novel potential target to mediate NK cell functions in the fight against complex lung diseases. and Fam13a +/+ (WT) mice used in the experiments were age-and gender-matched littermates generated from Fam13a +/heterozygous breeding pairs. All mice were bred and housed in a specific pathogen-free animal facility. All animal experimental protocols were performed following the approval of the Animal Welfare Structure (AWS) of the University of Luxembourg and the Luxembourg Institute of Health, and in accordance with National and European guidelines about the care and use of laboratory animals.

Flow cytometric analysis of immunophenotype
Cell suspensions were obtained by the mechanical disruption of mouse spleen, peripheral lymph nodes, bone marrow and lung. Red blood cells were then lysed in 1x lysing buffer (555899, BD Bioscience).

Primary murine NK cell in vitro expansion and cytokines activation (NK cell education)
Five million of splenocytes per well in 12-well plate were cultured in 2.5 ml of complete medium with

Statistical analysis
P-values were calculated by Graphpad Prism software with non-paired two-tailed Student t test as presented in the corresponding figure legends. If multiple comparison to be considered, we applied Turkey's correction one-way Anova. P-values under 0.05 were considered as statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001; n.s. or non-labelled between the compared groups, not significant).