All animal experiments were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Canadian Council for Animal Care. This study was approved by the animal ethics board at the University of Alberta (Protocol # AUP00001021 and AUP00002737).

Male and female BALB/c mice were purchased from the Charles River Institute. All animals were maintained and bred under pathogen-free conditions within the animal care facility at the University of Alberta. For the infection-related studies, adult mice were infected with E. coli (1 x10⁶ CFU) intraperitoneally (i.p.) as we have described elsewhere (26, 27).

B16-F10 melanoma tumor cells (1×10⁵), originally obtained from the ATCC, were injected subcutaneously on the left flank of mice under anesthesia conditions. Palpable tumors formed 5-7 days after injection and tissues were harvested 16 days after tumor inoculation.

Fresh blood was collected from human subjects infected with SARS-CoV-2 and the peripheral blood mononuclear cells (PBMCs) were isolated by gradient separation using Ficoll-Paque Premium (GE Healthcare, Chicago, USA) according to routine protocols (28, 29). The research ethics boards at the University of Alberta approved blood collection from human subjects with protocol # Pro00099502. Written informed consent was obtained from all the participants in this study.

Spleens from adult and neonate BALB/c or adult C57BL/c mice were collected. Single-cell suspensions were obtained by grinding tissue between sterile frosted glass slides in 7 mL of 1x RBC lysis buffer and filtering through a 40 µm cell strainer. Splenocytes were resuspended in RPMI with L-glutamine and sodium bicarbonate (Sigma-Aldrich) supplemented with 10% fetal bovine serum (Sigma-Aldrich) and 1% penicillin/streptomycin (Sigma-Aldrich) as we have described previously (30). Similarly, lung and liver samples were ground and resuspended in a culture medium and tissue-residing immune cells were isolated by gradient separation using Ficoll-Paque Premium (GE Healthcare, Chicago, USA). Blood samples from adult BALB/c mice were also subjected to Ficoll-Paque Premium gradient separation. Tumor tissues were dissected aseptically, washed with Hanks’ Balanced Salt Solution two times, cut in small pieces in lysis buffer (DNase [20 U] and Collagenase type IV [2000 U]). Tumor samples were then transferred to 15 ml conical tubes and incubated for 25 mins at 37° C inside a shaking incubator. Then tumor samples were washed, filtered, centrifuged on Ficoll^®-Paque PREMIUM 1.084.

Isolated cells from the spleen, lung, liver, bone marrow, blood, and tumor tissues were subjected to further analysis. In some experiments, the recombinant human Galectin-9 (rGal-9, 0.25 μg/mL) (Gal Pharma) was added to NK cells in the presence/absence of 100 μg/mL of the anti-CD44 blocking antibody (BD Biosciences, clone KM114) for an hr, as we have reported elsewhere (31). Moreover, in other experiments, the rGal-9 at the same concentration was added to mice splenocytes for an hr before analysis the interaction of rGal-9 with CD71⁺ erythroid cells.

Antibodies with specificity to mouse cell surface antigens and cytokines were purchased mainly from BD Biosciences and ThermoFisher Scientific. The following antibodies were used for mice: anti-CD49b [DX5], anti-NK 1.1 (PK136), anti-Gal-9 [RG9-35], anti-IFN-γ [XMG1.2], anti-Perforin [eBioOMAK-D], anti-Granzyme B [NGZB], anti-TNF-α [MP6-XT22], anti-CD107a [1D4B], anti-CD25 (PC61), anti-CD69 (H1-2F3), anti-KLRG-1(2F1), anti-Ly-6A/E also known as SCa-1 (E13-161.7), anti-CD71 (R17217), anti-TER-119 (TER-119), anti-Tim-3 [5D12], anti-CD44 [IM7], anti-CD11b [M1/70], anti-B220 [RA3-6B2], anti-p-LCK [SRRCHA], anti-ERK [20A], anti-AKT [M89-61], anti-MAPK [36/p38 (pT180/pY182)], anti-mTOR [O21-404]. The following antibodies were used for human samples: anti-CD3 [HIT3a], anti-CD16 [B73.1], anti-CD56 [B159], anti-Gal-9 [9M1-3], anti-IFN-γ [45.B3], anti-GzmB [GB11], anti-perforin [dG9], and anti-CD44 [G44-26]. A LIVE/DEAD Fixable Aqua Dead cell stain kit (Thermo Fisher Scientific) was used to exclude dead cells. For intracellular markers, cells were fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences) followed by intracellular staining with the appropriate antibody per our protocols (32). Stained cells (1 x 10⁶) were fixed in 4% paraformaldehyde. Data were acquired on an LSRFortessa Cell Analyzer (BD Biosciences) or Cytek Aurora and analyzed using FlowJo version 10.8 software.

Surface-stained mouse splenocytes (1 x 10⁶) were fixed in 4% paraformaldehyde. An Amnis ImageStream Mark II (EMD Millipore, ON, Canada) was used to collect at least 10,000 images for each sample as we have previously described (31). Data were analyzed using the IDEAS analysis software package to select single-cell images that were in focus. Next, we performed gating for NK cells (CD49b⁺) followed by gating for cells that co-express Gal-9 and CD44. Colocalization of Gal-9 and CD44 was determined using the IDEAS colocalization wizard.

MHC-I negative target cells (RMA-s cells, kindly provided by Dr. Kevin Kane, University of Alberta) were labeled with CFSE dye and then cultured at ratio of 25:1 (splenocyte: target cell) either total splenocytes with Gal-9^-NK cells or with cells isolated from liver Gal-9⁺NK cells from BALB/c mice. In brief, the same number of spleen or liver cells were cultured to mimic the physiological condition with target cells. After 4 hr incubation the frequency of killed cells were assessed by flow cytometry as described elsewhere (33).

Statistical comparisons between various groups were performed by a nonparametric student’s t-test Mann-Whitney U test or Kruskal Wallis test for multiple comparisons was used using the Graphpad Prism software (version 9). Results are presented as mean ± standard error of the mean (SEM). P < 0.05 was considered statistically significant.

We have already reported that Gal-9⁺NK cells have a distinctive phenotype compared to their negative counterparts in the peripheral blood of human subjects (4). Therefore, we decided to investigate the presence/function of Gal-9⁺NK cells in mice. First, we quantified the proportion of NK cells in different compartments in adult (8-10 weeks old) and neonatal BALB/c mice. We found that the frequency of NK cells was significantly higher in the peripheral blood followed by the lung in adult BALB/c mice ( Figures 1A, B , and Supplementary Figure 1A ). Moreover, we quantified the percentages of NK cells in the spleen, lung, and liver of neonatal BALB/c mice (7-day old), which showed a different pattern compared to adult mice. We found that the liver was significantly enriched with NK cells compared to the lung/spleen in neonatal BALB/c mice ( Figure 1C ). However, it was impractical to collect blood and bone marrow from pups, and as such we were unable to quantify the frequency of NK cells in the blood circulation and bone marrow of neonatal mice. A similar pattern for the proportion of NK cells in the spleen, lung, and liver of adult C57BL/6 mice was observed when using the NK1.1 marker ( Figure 1D ). However, the percentages of NK cells in the lung of adult BALB/c (average ~7.5%) appeared to be much lower than their counterparts in C57BL/6 mice (~12.5%) Figures 1B, D . To ensure that the same population of NK cells being evaluated in both mouse strains, we assessed NK cell frequency in C57BL/6 mice based on the co-expression of NK1.1 and CD49b markers ( Supplementary Figures 1B, C ). However, these observations confirmed our previous conclusions that C57BL/6 mice possess a higher proportion of NK cells in their lungs.

When the expression of Gal-9 on NK cells was analyzed, we observed that a subpopulation of NK cells under normal physiological circumstances express Gal-9 in different tissues in adult BALB/c mice ( Figure 1E ). The frequency of Gal-9⁺NK cells was significantly more pronounced in the liver compared to the spleen, blood, and, bone marrow and lung tissues of adult BALB/c mice ( Figure 1F ). Although the proportion of Gal-9⁺NK cells was similar between the blood and bone marrow, they were more abundant in the spleen and lungs of adult mice ( Figure 1F ). We also quantified the frequency of Gal-9⁺NK cells in neonatal mice (Day-7 old), which showed the liver as the main niche for these cells ( Figure 1G ). However, these observations confirmed that the lung had the least proportion of Gal-9⁺NK cells compared to the spleen and liver in neonatal mice ( Figure 1G ). In consistent with our observations in adult BALB/c mice, we found that the Gal-9⁺NK cells were significantly higher in the liver compared with the spleen/lung of adult C57BL/6 mice ( Figure 1H ). We also compared the frequency of Gal-9⁺NK cells in different tissues between male and female mice but we did not find any difference either in their frequency ( Figures 1I, J ) or the intensity of Gal-9 expression ( Supplementary Figures 1D, E ).

Since the frequency of Gal-9⁺NK cells was lower in the blood and bone marrow our further studies were performed on NK cells in the spleen, liver, and lungs. Taken together, these observations confirm the presence of a substantial number of Gal-9⁺NK cells in different compartments under normal circumstances in both adult and neonatal mice.

To better characterize the effector functions of Gal-9⁺ versus Gal-9^-NK cells, we subjected them to further analysis. We found that Gal-9⁺NK cells express significantly a greater level of Granzyme B (GzmB) and perforin expression compared to their negative counterparts in all studied tissues and even in the blood ( Figures 2A–D ). Likewise, Gal-9⁺NK cells had significantly a higher intensity of IFN-γ expression compared with Gal-9^-NK cells in the spleen, lung, liver, and blood of adult mice ( Figures 2E, F ). Also, we assessed the expression of TNF-α in NK cells, which confirmed a similar trend as IFN-γ expression in Gal-9⁺NK cells ( Figures 2G, H ). Considering the role of Gal-9 in immune cell effector function impairments (19), We speculated that a higher cytokine and cytolytic molecules content in Gal-9⁺NK cells might be associated with a lower degranulation capability. Surprisingly, we found that Gal-9⁺NK cells displayed a more robust degranulation capacity compared to their negative counterparts ( Figures 2I, J ). Collectively, our observations imply that Gal-9⁺NK cells have a more activated phenotype regardless of their niche, animal strain, and animal age.

To gain insight into the underlying mechanism of hyperimmune activation in Gal-9⁺NK cells, we conducted further investigations. It is reported that Gal-9 binds/interacts with different receptors (e.g. TIM-3, CD45, CD44, Dectin-1, etc.) (16, 34–36). Therefore, we quantified the expression of TIM-3 on Gal-9⁺/Gal-9^-NK cells in different tissues of adult mice. Interestingly, we found negligible expression of TIM-3 on NK cells regardless of their niche ( Figure 3A ). As such, we did not observe any Gal-9/TIM-3 co-expression on NK cells under homeostatic conditions in mice ( Figure 3A ). These observations precluded the interaction of Gal-9 with TIM-3.

Our further studies suggested that CD44 might be the main partner of Gal-9 on NK cells. This was illustrated by a more pronounced expression of CD44 in Gal-9⁺NK cells compared to their Gal-9^- counterparts ( Figures 3B, C ). We further analyzed the co-expression of Gal-9 with CD44. First of all, we noted that NK cells can be divided based on the expression of CD44 into CD44^high and CD44^mid/low NK cells ( Figure 3D ). When the co-expression of Gal-9 was further assessed, we found that Gal-9 was preferentially co-expressed with CD44^high NK cells in the spleen and lung ( Figure 3D ). However, this was not the case for the liver NK cells as the majority of them express Gal-9 ( Figure 3D ). To better clarify this finding, we compared the percentages of Gal-9 expressing NK cells in CD44^high and CD44^mid/low cells. These analyses confirmed that Gal-9 was strongly associated with the CD44^high NK subset as illustrated by the significant enrichment of Gal-9⁺NK cells in the CD44^high subset of NK cells ( Figures 3E, F ). Furthermore, we quantified the intensity of Gal-9 expression in CD44^high and CD44^mid/low NK cells in the lung, which confirmed a greater intensity of Gal-9 expression in CD44^high compared with CD44^mid/low NK cells ( Figures 3G, H ). To better visualize the co-expression of CD44 with Gal-9 on NK cells, we performed image stream analysis according to our protocols (37, 38). These observations indicated that Gal-9 is col-localized with CD44 on NK cells ( Figure 3I , and Supplementary Figure 1F ). To further assess the interaction of Gal-9 and CD44, we treated NK cells with the anti-CD44 antibody (an hr) followed by treatment with the rGal-9 (0.25 μg for an hr). We observed that CD44 blockade abrogates the binding of rGal-9 to NK cells in vitro ( Figure 3J ). Alternatively, we used CD71⁺ erythroid cells (CECs) (39, 40), as proof of concept, considering that these cells have negligible surface Gal-9 and a small portion of them express CD44. We found that treatment of these cells with rGal-9 resulted in an increased proportion of Gal-9⁺CD44⁺ cells without any changes in the CD44^-CEC subset ( Supplementary Figure 2 ). It is worth mentioning that treatment with rGal-9 increased the proportion of CD44 expressing CECs in vitro ( Supplementary Figure 2 ). This observation suggests that rGal-9 may enhance the expression of CD44, however, further studies are required to determine this possibility. Although these cells are different than NK cells, our observations show the tendency of Gal-9 towards CD44⁺ cells. Since Gal-9 is ubiquitous and interacts with different immune cells, we decided to determine whether CD44 is a common target for Gal-9 in different immune cells in mice. For this, we compared the expression of CD44 in different splenic immune cells in adult mice. We found that CD11b⁺ cells followed by T cells and NK cells had the greatest levels of CD44 expression but B cells had the least ( Figures 4A, B ). In line with this hypothesis, we observed a higher expression of Gal-9 in CD11b⁺ cells and T cells followed by NK cells ( Figure 4C ). This was further confirmed as we found a strong positive correlation between Gal-9 with CD44 expression in different immune cells ( Figure 4D ). Collectively, our results suggest that CD44 is the major partner for Gal-9 in different immune cells, in particular, NK cells.

To gain further mechanistic insight into the Gal-9:CD44 interactions in NK cell function, we analyzed the activation of intracellular signaling pathways known to operate downstream of CD44 (41). In agreement, we found enhanced expression of Phspho-LCK in Gal-9⁺ versus Gal-9^-NK cells in the spleen and lung of adult mice ( Figures 4E, F ). Similarly, we noted a significant elevation in the expression of ERK in Gal-9⁺ compared to Gal-9^- NK cells ( Figures 4G, H ). Likewise, the expression of Akt, MAPK, and mTOR were significantly higher in Gal-9⁺ versus Gal-9^- NK cells, respectively ( Figures 4I–N ).

Moreover, we subjected NK cells from the spleen to further investigations for the expression of co-stimulatory/inhibitory markers. We found that Gal-9⁺NK cells were enriched with a higher proportion of CD69 but reduced KLRG-1 expressing NK cells compared to their negative counterparts ( Figures 5A–D ). Given the activation status of Gal-9⁺NK cells, we speculated that these cells may express a higher level of stem cell Ag1 (Sca-1), as a marker of activated NK cell, which plays a role in NK cell response to infection (42). Interestingly, we observed that Gal-9⁺NK cells had significantly a higher level of this activation marker that their negative counterparts ( Figures 5E, F ). Similarly, we observed a greater expression of CD25 in Gal-9⁺ than in Gal-9^-NK cells ( Figures 5G, H ). To further associate the hyperactivation status of Gal-9⁺NK cells with their cytotoxicity, we performed cytotoxicity assay at 25:1 ratio of spleen/liver cells with target cells (1 x 10⁶: 40,000 RMA-s cells) for 4 hr without any stimulation. We found that liver Gal-9⁺NK cells exhibited a greater specific killing of target cells ( Figures 5I, J ).

Together, these data suggest the relatedness of the CD44 and Gal-9 signaling in inducting NK cell activation.

Considering a higher activity of Gal-9⁺NK cells in terms of cytokine and cytolytic molecule expression, we decided to determine the impact of infection on this NK subset in mice. As proof of concept, adult mice were infected with E. coli (1 x10⁶ CFU, intraperitoneally). Two days later their splenocytes were subjected to further analysis. These studies showed a significant expansion of Gal-9⁺NK cells in the spleen of infected mice ( Figures 6A, B ). Therefore, our observations suggest that the expansion of Gal-9⁺NK cells might be an indicative of enhanced innate immunity against infection.

To determine the presence of Gal-9⁺NK cells in pathological conditions, we investigated the presence of NK cells in the spleen and tumor tissues of tumor-bearing mice (B16-F10 melanoma). We did not observe any significant difference in the frequency of total NK cells in the spleen of tumor-bearing mice compared to controls. However, we found a significant expansion of Gal-9⁺NK cells in the spleen of tumor-bearing mice versus controls ( Figures 6C, D ). In particular, we observed that the majority of tumor resident NK cells expressed Gal-9 ( Figures 6C, D ). These observations showed that Gal-9 was mainly expressed on CD44^high NK cells ( Figure 6C ). Gal-9⁺NK cells in tumor-bearing mice exhibited a more robust effector functions as we reported for their counterparts in healthy mice. These observations support the interaction of Gal-9 with CD44^high in pathological conditions. However, further studies are needed to determine the prolonged impact of these interactions on NK cells effector functions.

We have previously reported that the frequency of Gal-9⁺NK cells get expanded in HIV-infected individuals and patients with virus-associated solid tumors (4, 10). To determine whether our observations in mice were applicable to humans, we investigated the interaction of CD44 with Gal-9 in human NK cells. Since healthy individuals have negligible levels of Gal-9⁺NK cells, we studied the frequency of Gal-9⁺NK cells in a few SARS-CoV-2 infected individuals. These results confirmed our observations in mice that Gal-9 was preferentially co-expressed with CD44^high NK cells ( Figure 6E ). Overall, these results suggest that Gal-9 has the propensity of interacting with CD44 on NK cells in both mice and humans. Despite the interaction of Gal-9 with CD44 we found a dichotomy in terms of effector functions in human NK cells. We observed Gal-9⁺NK cells in humans display a greater IFN-γ expression ( Figures 6F–H ) without a significant difference in their cytolytic molecules expression. These observations suggest differences in Gal-9⁺NK cells effector functions between mice and humans in different pathological conditions.