Conventional NK cells and ILC1 are partially ablated in the livers of Ncr1 iCreTbx21 fl/fl mice

Mouse liver contains both Eomes-dependent conventional natural killer (cNK) cells and Tbet-dependent liver-resident type I innate lymphoid cells (ILC1). In order to better understand the role of ILC1, we attempted to generate mice that would lack liver ILC1, while retaining cNK, by conditional deletion of Tbet in NKp46+ cells. Here we report that the Ncr1 iCreTbx21 fl/fl mouse has a roughly equivalent reduction in both the cNK and ILC1 compartments of the liver, limiting its utility for investigating the relative contributions of these two cell types in disease models. We also describe the phenotype of these mice with respect to NK cells, ILC1 and NKp46 + ILC3 in the spleen and small intestine lamina propria.


Introduction
Mouse liver contains two NK cell populations. Conventional NK cells (cNK) are defined by their expression of CD49b (DX5) 1 , depend on the transcription factor Eomes 2 , and circulate freely 1,3 . The other NK cell population, which expresses CD49a 1 , depends on the transcription factor Tbet 2,3 and is unable to leave the liver 1,3 . There is still some dispute over whether these cells should properly be considered liver-resident NK cells or non-NK type I innate lymphoid cells (ILC1) 4,5 : here, we call them "liver ILC1". In mice, cNK and liver ILC1 are distinct lineages that cannot crossdifferentiate 6 .
The factors involved in the lineage specification of liver ILC1 are already well-understood, but the function of these cells is not yet clear. They produce IFNγ and TNFα, as expected of ILC1, as well as high levels of GM-CSF 1-3,6 , but it is unclear whether the production of these cytokines specifically by liver ILC1, as opposed to by cNK, have any role in health and disease. Tissue-resident ILC1 in some other organs have physiological functions 7,8 , so it is also possible that liver ILC1 have some asyet-undiscovered physiological role.
To answer these questions, we sought to generate mice that would lack liver ILC1 while retaining cNK. Tbet knockout (Tbx21 -/-) mice fulfill these criteria 2,3 , but also have alterations in the T cell compartment that would complicate the analysis. Therefore, we crossed Tbx21 fl/fl onto Ncr1 iCre mice to produce conditional knockout (Ncr1 iCre Tbx21 fl/fl ) animals, in which Tbet is lost only in cells expressing Ncr1, whose protein product is the NK cell activating receptor NKp46. Here, we report that these mice have a roughly equivalent reduction in both the cNK and ILC1 compartments of the liver, limiting their utility for investigating the relative contributions of these two cell types in disease models. We also note that the loss of Tbet differentially impacts NKp46 + ILC populations in the spleen, liver and small intestine, suggesting that Ncr1 iCre Tbx21 fl/fl mice could have potential as a tool for understanding how and when Tbet is required for NKp46 + ILC development and trafficking.
Although we chose to use floxed-only, rather than iCre-only, littermate controls, we do recognise that iCre transgene expression itself can have an effect on phenotype. Ncr1 iCre mice are known to have slightly reduced expression of NKp46 on NK cells, although the total number of NK cells (identified as CD3-NKp46+) in these mice is normal 9 . We confirmed these observations in Ncr1 iCre mice, compared to Ncr1 WT littermate controls in our own colony 10 . Further, we identify NK cells as Lin-NK1.1+, rather than Lin-NKp46+, to avoid potential confounding effects of reduced NKp46 expression.
Mice were sacrificed between 6.5 and 9 weeks of age, using rising carbon dioxide followed by cervical dislocation. Spleen, liver and intestines were dissected out of each of the 12 mice for cell isolation. Animal husbandry and experimental procedures were performed according to UK Home Office regulations and institute guidelines, under project license 70/8530.

Cell isolation
Dissected livers (a total of 12) were minced finely with opposing scalpel blades. The tissue was collected in HBSS with Ca 2+ Mg 2+ (Life Technologies, Paisley, UK) supplemented with 0.01% collagenase IV (Life Technologies) and 0.001% DNase I (Roche, distributed by Sigma-Aldrich, Dorset, UK) and passed through a 70 μm cell strainer. The suspension was spun down (500 ×g, 4°C, 10 minutes) and the pellet resuspended in RPMI 1640 medium (Life Technologies). The cell suspension was then layered over 24% Optiprep (Sigma-Aldrich) and centrifuged without braking (700 ×g, RT, 20 minutes). The interface layer was taken and washed in HBSS without Ca 2+ Mg 2+ (Lonza, distributed by VWR, Lutterworth, UK) supplemented with 0.25% bovine serum albumin (Sigma-Aldrich) and 0.001% DNase I.
Small intestine lamina propria lymphocytes were isolated using a protocol adapted from Halim and Takei 11 . Briefly, dissected intestines (a total of 12) were placed in ice-cold PBS supplemented with 2% fetal calf serum (FCS; Life Technologies) and the bulk of fecal matter removed by flushing the intestines using a syringe and 18G blunt end needle. The intestines were cut longitudinally and vortexed briefly 3x in ice-cold PBS/2% FCS to remove residual fecal matter. Tissue sections were incubated in PBS supplemented with 1 nM EDTA (shaking at 120 rpm, 37°C, 20 minutes) followed by 3x washes with ice-cold PBS/2% FCS before being minced finely with opposing scalpel blades. The homogenized tissue was digested in DMEM (Life Technologies) supplemented with 10% FCS, 50 μM 2-mercaptoethanol (Life Technologies), 250 U/mL collagenase IV and 50 U/mL DNase I (shaking at 120 rpm, 37°C, 20 minutes) and passed through a 70μm cell strainer. The cell suspension was centrifuged (400 ×g, 4°C, 5 minutes) and the cell pellet resuspended in 40% Percoll (GE Healthcare, distributed by Sigma-Aldrich) before centrifugation without braking (600 × g, 4°C, 10 minutes). The resultant pellet was washed in PBS/2% FCS (400 × g, 4°C, 5 minutes).
Dissected spleens (a total of 12) were passed through a 40 μm cell strainer. Red blood cells were lysed by 5 minute incubation in ACK lysing buffer (Life Technologies).

Flow cytometry
The antibodies used are displayed in Table 1.

Statistical analysis
Groups were compared using Mann-Whitney U Tests. Analysis was carried out using Vassarstats (RRID SCR_010263).

Results
See Figure 1.

Discussion
We observed a modest (~4-fold) reduction in splenic NK (defined as Lin-NK1.1+) in the Ncr1 iCre Tbx21 fl/fl conditional knockouts, compared to Ncr1 WT Tbx21 fl/fl littermate controls ( Figures 1A  and B). This is comparable to the 2-to 4-fold reduction in splenic cNK that has previously been reported in Tbx21 knockout, compared to wild-type, mice 2,3,12 and is likely to be a result of reduced survival 12 or bone marrow egress 13 of NK cells in the absence of Tbet.
We also observed a reduction in the absolute number of cNK (defined as Lin-NK1.1+ CD49b+) in the liver ( Figures 1C and D). We had expected that this might be similar to the reduction of cNK in the spleen, but, at ~10-fold, it was more pronounced, potentially pointing towards a differential requirement for Tbet in cNK survival in or recruitment to the liver, compared to the spleen. Although the absolute number of liver ILC1 (defined as Lin-NK1.1+ CD49a+) was also reduced (~10-fold), a substantial residual population was present ( Figures 1C and E), in contrast to the Tbx21 knockout, in which these cells are almost completely eliminated 2,3 . Unlike the cNK in the conditional knockout, the residual liver ILC1 all expressed Tbet ( Figure 1C). This supports the proposal that Tbet is absolutely required for continued survival of these cells 2 , since no ILC1 in which Tbet was not expressed persisted. We were surprised to note that Tbx21 excision seemed to be less efficient in liver ILC1  than cNK, because fate mapping of iCre activity under the Ncr1 promoter using R26R eYFP has previously shown that iCre activity is higher in ILC1 than cNK 9 . Given the absolute requirement of Tbet for the survival of liver ILC1 2,3 , it seems likely that even if the excision failed in only a few cells, these would have benefited from a selection advantage that might have resulted in their expansion. Whatever the cause of the unexpectedly large reduction in cNK and the unexpectedly small reduction in ILC1, the finding that both of these were reduced by equivalent amounts in the liver of conditional knockouts compared to controls limits the utility of Ncr1 iCre Tbx21 fl/fl mice for dissecting the relative contributions of the two cell types in disease models.
Rankin et al. have also recently generated Ncr1 iCre Tbx21 fl/fl mice, and report a severe reduction in ILC1 (defined in their paper as Lin-RORγt+ NKp46-) and NKp46 + ILC3 (defined as Lin-RORγt+ NKp46+) in the small intestine lamina propria 14 . We were able to isolate fewer Lin-CD127+ ILC from the small intestine than has previously been reported, but even with this suboptimal cell isolation procedure we made findings similar to those of Rankin et al., observing a ~6-fold reduction in ILC1 (defined here as Lin-CD127+ RORγt+ NKp46-) and a ~24-fold reduction in NKp46+ ILC3 (defined here as Lin-CD127+ RORγt+ NKp46+) compared to littermate controls ( Figures 1F-H).
In summary, conditional deletion of Tbet in NKp46 + cells, where Tbx21 excision has been successful, differentially affects cNK and ILC1 in different organs. In the liver, a residual population of ILC1, in which Tbx21 has not been excised, persists. We conclude that the Ncr1 iCre Tbx21 fl/fl mouse is therefore unlikely to be useful for investigating the relative contributions of liver cNK and ILC1 to pathogenesis in disease models, but could still have potential as a tool for understanding how and when Tbet is required for the development and trafficking of NKp46 + ILC.

Competing interests
No competing interests were disclosed.  It remains unclear why NCRiCre controls are not reported in this study as they appear to have been analysed. Nevertheless the authors have added a comment to the rationale of their exclusion from the dataset. This is an interesting small study in which the objective was to delete Tbet and hopefully generate mice that lacked ILC1. This turned out not to be the case. This is a very brief report that depends solely on the results shown in Figure 1. These data appear to support the claims of the authors that loss of T-bet is not sufficient to differentially delete ILC1. It is indicated that 6 mice have been used, however, it is not clear whether this represents a single experiment, or alternately, the data are pooled from several experiments. It is essential to clarify this point.

References
The authors have used NcrWTTbx21fl/fl control mice for their experiments. Ncr1iCre mice have lower levels of expression of NKp46 compared with either wildtype or floxed control mice. Given only floxed control mice are indicated to have been used, how was this reduction in expression accommodated, or was this considered in the experiments? It is not clear what gating strategy was used for enumerating the NK cell frequency or number.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Partly
No competing interests were disclosed.

Competing Interests:
Referee Expertise: Immunology

I have read this submission. I believe that I have an appropriate level of expertise to confirm that I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 03 Jul 2017 , Royal Free and UCL School of Medicine, UK

Victoria Male
This is an interesting small study in which the objective was to delete Tbet and hopefully generate mice that lacked ILC1. This turned out not to be the case. This is a very brief report that depends solely on the results shown in Figure 1. These data appear to support the claims of the authors that loss of T-bet is not sufficient to differentially delete ILC1. It is indicated that 6 mice have been used, however, it is not clear whether this represents a single experiment, or alternately, the data are pooled from several experiments. It is essential to clarify this point.

Six littermate pairs (a total of 12 mice) were used, with spleen, liver and small intestine taken from each mouse. We have clarified this point in the Methods of version 2:
"Ncr1 mice were crossed onto Tbx21 and the resultant F1 generation was backcrossed onto Tbx21 to produce Ncr1 Tbx21 conditional knockouts (n = 6) and Ncr1 Tbx21 littermate controls (n = 6)… Spleen, liver and intestines were dissected out of each of the 12 mice for cell isolation." The authors have used NcrWTTbx21fl/fl control mice for their experiments. Ncr1iCre mice have lower levels of expression of NKp46 compared with either wildtype or floxed control mice. Given only floxed control mice are indicated to have been used, how was this reduction in expression accommodated, or was this considered in the experiments? It is not clear what gating strategy was used for enumerating the NK cell frequency or number.
In version 2, we discuss the reported phenotype of Ncr1 mice and the steps we have taken to avoid the potential confounding effect of slightly lower NKp46 expression in NKp46+ cells in these mice; in particular, that we identify NK cells using NK1.1, rather than NKp46. We have also made the data from our own characterisation of Ncr1 compared to WT littermate control mice available, linked to this publication. "Although we chose to use floxed-only, rather than iCre-only, littermate controls, we do recognise that iCre transgene expression itself can have an effect on phenotype. Ncr1 mice are known to have slightly reduced expression of NKp46 on NK cells, although the total number of NK cells (identified as CD3-NKp46+) in these mice is normal . We confirmed these observations in Ncr1 mice, compared to Ncr1 littermate controls in our own colony . Further, we identify NK cells as Lin-NK1.1+, rather than Lin-NKp46+, to avoid potential confounding effects of reduced NKp46 expression." Gating strategies are as shown in the figure. For greater clarity, in version 2 we also give the gating strategies used to define each population in the text.
No competing interests were disclosed. In the current climate of developing our understanding of mechanisms using conditional KO mice, basic data on how models work has clear value to the immunological community and data indicating the success of problems with this type of model can really benefit other researchers. It is also important for researchers to fully assess how well different conditional KO mouse models work with robust controls -it is now apparent that one cannot simply assume either specific or efficient cre-mediated deletion simply because of the intentioned design of the mouse model. Within the data presented here, it is striking that tissue specific effects are observed, particularly given the interest in tissue residency of ILC populations and the potential for tissue-specific differentiation. Much of the data is clearly presented and appears robust in terms of the analysis of NK/ILC1 populations. There are some modest concerns with manuscript as it stands -some in terms of presentation of the data and some in terms of the experimental setup and methodology. Experimental design: The control mice used here are 'floxed only' controls rather than 'cre only' controls. The latter is better as it enables the impact of expression of the bacteriophage protein (cre), which is clearly a 'foreign' element expressed in the mouse. A number of studies have demonstrated that cre expression can impact on aspects of the immune system (e.g. Lck in T cell development in the thymus). This should be acknowledged within the description of the data.
The authors should have fate-mapped cre expression in their mice in their hands to enable the efficiency of cre-expression to be properly addressed. This would clarify the extent to which NK1.1 expressing cells show evidence of cre mediated deletion, informing the understanding of NK1.1+ cells that retain T-bet expression. Whilst beyond the scope of this study now, it again could be noted.

Methodology:
The SI LP prep is clearly sub-optimal for ILC populations -the fraction of the Lineage negative cells expressing IL-7Rα is substantially less than seen with other methods and there is clearly a substantial range in terms of numbers of cells isolated. Thus the data from the SI LP prep is of limited value given the spread of data and the obvious issues in cell isolation. Thus this data appears very preliminary and suboptimal -this should be recognised in the description of this data.
Presentation of data: In Figures B,D, E, G and H it would be more appropriate to show the median rather than the mean (which is what I assume the bar represents -it is not described in the figure legend). One cannot assume the data is normally distributed and plotting the median would eliminate bias driven by clear outliers -for example the data in Figure H, absolute number clearly suffers from this, as does the similar data in 'G'. The data in all these graphs should show median, not mean and this should the similar data in 'G'. The data in all these graphs should show median, not mean and this should be articulated in the Figure Legend.
It is not clear to me in the SI LP analysis (in part 'F') why Lin-IL-7Ra+ cells are not gated on -there is a clear IL-7Ra+ population, but the gate includes IL-7Ra-cells. Given that the aim of this analysis is to look at ILC populations (but with no reference to conventional NK cells which are IL-7Ra-), a more specific gate would seem sensible -this would be my recommendation.
Also in part F, the plots of T-bet expression would benefit from an indication of where the authors consider +ve and -ve T-bet staining. This would seem reasonably uncontentious given the populations observed.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed.

Competing Interests:
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 03 Jul 2017 , Royal Free and UCL School of Medicine, UK

Victoria Male
The control mice used here are 'floxed only' controls rather than 'cre only' controls. The latter is better as it enables the impact of expression of the bacteriophage protein (cre), which is clearly a 'foreign' element expressed in the mouse. A number of studies have demonstrated that cre expression can impact on aspects of the immune system (e.g. Lck in T cell development in the thymus). This should be acknowledged within the description of the data.
In version 2, we acknowledge that Cre expression itself can have a phenotypic effect. We discuss the reported phenotype of Ncr1 mice and the steps we have taken in this study to avoid the potential confounding effect of slightly lower NKp46 expression in NKp46+ cells in these mice, in particular that we identify NK cells using NK1.1, rather than iCre NKp46+ cells in these mice, in particular that we identify NK cells using NK1.1, rather than NKp46. We have also made the data from our own characterisation of Ncr1 compared to WT littermate control mice available, linked to this publication. In "Mice": "Although we chose to use floxed-only, rather than iCre-only, littermate controls, we do recognise that iCre transgene expression itself can have an effect on phenotype. Ncr1 mice are known to have slightly reduced expression of NKp46 on NK cells, although the total number of NK cells (identified as CD3-NKp46+) in these mice is normal . We confirmed these observations in Ncr1 mice, compared to Ncr1 littermate controls in our own colony . Further, we identify NK cells as Lin-NK1.1+, rather than Lin-NKp46+, to avoid potential confounding effects of reduced NKp46 expression." The authors should have fate-mapped cre expression in their mice in their hands to enable the efficiency of cre-expression to be properly addressed. This would clarify the extent to which NK1.1 expressing cells show evidence of cre mediated deletion, informing the understanding of NK1.1+ cells that retain T-bet expression. Whilst beyond the scope of this study now, it again could be noted.
Although we have not done this ourselves, fate mapping of the Ncr1 mice was done as part of the characterisation when they were first developed. We have added this to the Discussion of version 2: "We were surprised to note that Tbx21 excision seemed to be less efficient in liver ILC1 than cNK, because fate mapping of iCre activity under the Ncr1 promoter using R26R has previously shown that iCre activity is higher in ILC1 than cNK ." The SI LP prep is clearly sub-optimal for ILC populations -the fraction of the Lineage negative cells expressing IL-7Rα is substantially less than seen with other methods and there is clearly a substantial range in terms of numbers of cells isolated. Thus the data from the SI LP prep is of limited value given the spread of data and the obvious issues in cell isolation. Thus this data appears very preliminary and suboptimal -this should be recognised in the description of this data.
We acknowledge this in version 2. In the Discussion: "We were able to isolate fewer Lin-CD127+ ILC from the small intestine than has previously been reported, but even with this sub-optimal cell isolation procedure we made findings similar to that of Rankin , observing a ~6-fold reduction in ILC1 (defined here et al. as Lin-CD127+ RORγt+ NKp46-) and a ~24-fold reduction in NKp46+ ILC3 (defined here as Lin-CD127+ RORγt+ NKp46+) compared to littermate controls ( )." Figures 1F-H In Figures B,D, E, G and H it would be more appropriate to show the median rather than the mean (which is what I assume the bar represents -it is not described in the figure legend). One cannot assume the data is normally distributed and plotting the median would eliminate bias driven by clear outliers -for example the data in Figure H, absolute number clearly suffers from this, as does the similar data in 'G'. The data in all these graphs should show median, not mean and this should be articulated in the Figure Legend