Tissue resident NK cells mediate ischemic kidney injury and are not depleted by anti-Asialo GM1 antibody

Mice

B6 (C57BL/6J), NOD (NOD/ShiLtJ), BALB (BALB/cJ), 129 (129S1/SvImJ), and CD1dKO (B6.129S6-Del(3Cd1d2-Cd1d1)1Sbp/J, formerly named B6.129S6-Cd1d1/Cd1d2^tm1Spb/J) mice were purchased from the Jackson Laboratory (Bar Harbor, ME) and briefly housed in a specific pathogen-free colony at the University of Colorado at AMC. CD1dKO (a gift from Dr. Laurent Gapin, University of Colorado, Anschutz Medical Campus) (78), B6, and BALB/c mice were also obtained from breeding colonies at the University of Colorado, Anschutz Medical Campus and maintained in a pathogen-free facility and were of 8–12 wk of age for experimental use. Mice were age- and sex-matched for experiments, with male mice uniformly used for general screening and IRI studies. Female mice were solely used for the analysis of trNK distribution in multiple organs. All experiments were approved by the Institutional Animal Care and Use Committee of the University of Colorado Denver, in accordance with the National Institutes of Health guidelines for use of live animals. The University of Colorado Denver, Anschutz Medical Campus is accredited by the American Association for Accreditation of Laboratory Animal Care. Tbet^−/− (formal gene name Tbx21) mice were purchased from Jackson Laboratory and NFIL-3^−/− mice (from Paul B. Rothman, Johns Hopkins University (43, 48)), were bred and housed in a pathogen-free facility, with procedures performed in accordance with the animal protocol approved by the Washington University School of Medicine (WUSM) Animal Studies Committee.

Induction of Ischemia reperfusion injury (IRI)

The hanging weight system was used as previously described (49). Briefly, mice were anesthetized with pentobarbital at 20mg/kg and body temperature was regulated using an anal probe during the entire surgery. For sham surgery, mice were subjected to a right kidney nephrectomy followed by recovery under a heat lamp. To induce IRI, nephrectomy of the right kidney was done, after which the left kidney was exposed and connective and adrenal tissue was gently separated. Next, the kidney was place in a Lucite cup and ischemia was induced for 30m by placing 6–0 nylon suture under the renal artery and applying hanging weights, to occlude blood flow. Following 30 minutes of ischemia, hanging weights were removed and the kidney was allowed to reperfuse. Mice were given saline, sutured closed and allowed to recover for the indicated time points of reperfusion (typically 4 and 24 hours of reperfusion).

Analysis of Renal Function

To assess kidney function, glomerular filtration rate (GFR) was measured as previously described (49). Briefly, mice were anesthetized with pentobarbital at 20mg/kg and mice were kept warm to maintain their body temp using an anal probe during entire surgery. Once mice were fully anesthesized, a catheter was inserted into the jugular vein with a second catheter inserted into the bladder to collect urine. Next, Inulin-FITC (Sigma) was infused at 800 microliters per minute. Blood and urine were collected every twenty minutes to measure GFR as described (49, 50).

Parabiosis

Parabiosis surgery was performed as previously described (42, 43). Briefly, a longitudinal skin incision was made on the flanks of both C57BL/6NCr (Ly5.2) and B6-Ly5.1/Cr age and weight matched female mice. Their elbows and knees were joined with dissolvable sutures and the incision closed with wound clips. Postoperative care included administration of buprenex for pain control, 5% dextrose and 0.9% sodium chloride for fluid replenishment. Nutritional gel packs were provided in each cage and Sulfatrim (antibiotic) in the drinking water for the duration of the experiment. Mice were sacrificed and analyzed fourteen days post surgery. The procedure was approved by the Animal Studies Committee at Washington University, St. Louis, MO.

Flow cytometric cell analysis

To quantify the number of NK and NKT cells, mice were perfused with PBS, kidney were harvested, and minced through a 70um filter. Digested tissue was washed with RPMI and then lymphocytes were purified using a 36% percoll gradient (GE Healthcare). Cell were stained in FACS buffer (PBS, 0.2% Fetal Bovine Serum, 0.01% sodium azide), anti-Fc receptor blockade CD16/32, and αCD3 PE-Cy7 (145-2C11), αNK1.1 PerCP-Cy5.5 (PK136), αNKp46 (29A1.4), and αCD45(30-F11). NK cells were defined as CD45+CD3−NK1.1+ events, with NKT cells defined as CD45+CD3+NK1.1+ events. For studies analyzing tissue-resident NK cells, mice were perfused with 15 mL of saline (excluding parabiotic studies and analysis of Tbet−/− and NFIL-3−/− mice) after which kidney, liver, lung, pancreas and uterus were harvested, minced and mechanically disrupted, followed by a 1 hr enzymatic digestion in collagenase D (1 mg/mL, Roche) at 37°C. Cells were then washed in complete RPMI 1640 (10% fetal calf serum, 1% penicillin/streptomycin). Cells were then stained with a cocktail of fluorescently-labeled antibodies from eBioscience: αCD3 e450 (145-2C11), αCD19 e450 (ebio1D3), αMHCII e450 (MS/114.15.2), αCD45 APC-eFluor780 (30-F11), αNK1.1 PerCP-Cy5.5 (PK136); or Biolegend: αNKp46 PE or PE-Cy7 (29A1.4), αCD49b PE-Cy7 (DX5), αCD49b (HMα2), αTRAIL PE, αCD44 FITC, αCD160 PE, αCD49a PE or APC, αAsGM1 AlexaFlour647 (poly21460); or BD Pharmingen αCD49a (Ha31/8). NK cells were identified as lymphocytes (by FSC and SSC) that were CD45+, Lineage negative (CD3−CD19−MHCII−), and either NK1.1+ or NKp46+. Tissue-resident NK (trNK) and cNK cells were identified as either CD49a+ or DX5+, respectively. Anti-AsGM1 antibody staining was done using 0.35 μg antibody per stain for 30 minutes. In a subset of studies, cNK cells were defined as NK cells that expressed CD49b+ CD49a− using the HMα2 clone, instead of the CD49b-reactive DX5 clone. Note that both antibodies recognize CD49b, though the DX5 clone may be less-sensitive than HMα2 for cells that express lower levels of CD49b (51). For NK, trNK versus cNK, and NKT cells composition after IgG or NK1.1 treatment, mice were treated with either IgG control antibody or NK1.1 on day −3 and −1. On day 0, mice were then subjected to hanging weight system for 30 minutes to induce IRI. After 1 day of reperfusion, kidney was harvested and lymphocytes were purified and analyzed using flow cytometry after stained as noted above with NKp46 being the NK cell marker. For parabiotic studies, cells were stained for CD45.1 (clone A20) and CD45.2 (clone 104), both from eBioscience. Gating strategies for flow cytometric analysis are indicated in figure legends. Samples were analyzed using a BD LSRII flow cytometer.

NK cell depletion

For NK1.1 depleting experiments, B6 and CD1d knock out mice were treated with either 200μg control α-mouse IgG2a (BioXCell C1.18.4, mouse IgG2a Isotype control) or α-NK1.1 (BioXCell PK136) on day −3 and −1. For Asialo-GM1 depleting experiments, B6 mice were treated with 800μg of either rabbit IgG (Southern Biotech) or AsGM1 (WAKO) on day −1. On day 0, mice were then subjected to hanging weight system for 30 minutes to induce IRI. After 1 day of reperfusion, kidney function was measured by GFR. NK cell depletion was verified by measuring NK cell percentages (CD45+CD3−NKp46+).

Detection of Cleaved Caspase-3

B6 mice were treated with IgG or NK1.1 or AsGM1 antibodies as above. Mice that received IgG or NK1.1 antibodies were exposed to sham (−I) or ischemic (+I) surgery and allowed to recover for 4 hours. AsGM1 treated mice only received ischemic surgery (+I). At 4 hours reperfusion, kidneys were harvested, fixed in 4% PFA and paraffin embedded. Five micron thick paraffin sections were deparaffinized, antigens unmasked and immunohistochemically stained for cleaved Caspase 3 (Cell Signaling, Danvers, MA; rabbit polyclonal; cat#: 9661, 1:1000 in TBST + 1% BSA w/v). Sections required antigen retrieval in BORG solution (Biocare Medical, Concord, CA; high pH; 10 minutes at 110°C) in the Decloaking chamber (Biocare). All incubations were performed on the autostainer (Benchmark XT; Ventana Medical Systems/Roche, Tucson, AZ) at an operating temperature of 37°C. Primary antibody was incubated for 32 minutes and detected with a modified I-VIEW DAB (Ventana) detection kit. The I-VIEW secondary antibody was replaced with a species specific polymer (Rabbit ImmPress; Full Strength; Vector Labs, Burlingame, CA; 8 minutes) and streptavidin-horseradish peroxidase replaced with diluted Rabbit ImmPress (1:2 dilution in PBS pH 7.6; 8 minutes). All sections were counterstained in Harris hematoxylin for 2 minutes, blued in 1% ammonium hydroxide (v/v), dehydrated in graded alcohols, cleared in xylene and coverglass mounted using synthetic resin. Quantitation of caspase-3 immuno-positive cells was determined by gathering the numerical total across the entire kidney section, at a 20X (0.26 mm²) magnification. Representative images were chosen based on the ratio of the number of caspase-3 positive cells in ischemic groups versus sham group. Differences between αNK1.1 and αAsGM1 groups were corroborated by a second, independent investigator who was blinded to the experimental groups.

Software and Statistical Analysis

Data analysis and plotting were done using Prism 5 (GraphPad Software). Flow cytometric data were analyzed using FlowJo version 10 (TreeStar), with data displayed as high-resolution pseudo-color dot plots showing outliers, plotted on a biexponential axis. Statistical analyses were performed using Prism 5, with unpaired student t-test or one-way ANOVA with Bonferroni’s post-test correction for multiple comparisons. Experimental values were considered to be statistically different if they were p<0.05.

Kidney conventional NK cells preferentially express Asialo-GM1 compared to tissue-resident NK cells

Recently, two subsets of NK cells were identified in the liver based on either tissue-retention (trNK) versus systemic circulation (cNK) (42). Further studies showed that these subsets could be identified as CD49a+DX5− (trNK) versus DX5+CD49a− (cNK) in the liver as well as multiple other organs (42, 43). While trNK cells were identified in multiple organs, the presence of trNK cells in the kidney has not been reported. To test this, we analyzed NK cells in the kidney and identified both CD49a+DX5− and CD49a−DX5+ NK subsets (Fig. 1A). In the unmanipulated kidney, CD49a+ NK cells were a minor, but readily detectable fraction (15–20%) of the NK cell pool (Fig. 1B). To demonstrate that CD49a+ NK cells in the kidney were tissue-resident we analyzed these populations in parabiotic pairs of mice. This analysis demonstrated that CD49a+ NK cells were present only in the host and not the other parabiont, indicating that CD49a+ NK cells did not systemically circulate between the two parabionts (Fig. 1C–D). In contrast, DX5+CD49a− NK cells were detectable in both parabionts, indicating that DX5+ NK cells were capable of systemic circulation, exchanging these cells between the two parabionts in the kidney (Fig. 1C–D). The percent chimerism in the parabiotic mice was determined by the expression of CD45.1 and CD45.2 in the spleen (Fig. 1E). These data demonstrate that the kidney contains both CD49a+ trNK cells and DX5+ cNK cells.

trNK cells and cNK cells have differential expression of multiple proteins (42, 43). To investigate the cell surface phenotype of these two subsets in the kidney, we performed an analysis of proteins involved in NK cell activation and trafficking. These studies showed that in the kidney, trNK cells had uniform increased expression of CD44 (an activation and adhesion protein), CD160 (an activation-associated protein) and TRAIL (an apoptosis-inducing cell surface protein) relative to cNK cells, with a subset of trNK cells also expressing the chemokine receptor CCR2 (Fig. 1F). In contrast, trNK cells had uniformly low expression of CD62L (an adhesion protein) relative to cNK cells (Fig. 1F). trNK cells also had reduced expression of two inhibitory receptors relative to cNK cells, uniformly lacking expression of KLRG1 with a prominent fraction of trNK cells that also failed to express CD244 (Fig. 1F). Strikingly, kidney trNK cells had reduced expression of asialo-GM1 (Fig. 1G), a glycolipid moiety that has been a target for antibody-mediated depletion of NK cells in vivo (52). These data identify that trNK cells have a more activated phenotype with potentially reduced susceptibility to inhibitory signals compared to cNK cells. They further indicate that trNK cells express variable levels of AsGM1 and may be differentially susceptible to AsGM1-antibody mediated depletion in vivo.

Kidney trNK cells develop independent of NFIL3 and Tbet

We next sought to determine the developmental requirements for kidney trNK cells. To do this, we investigated the relative frequency and abundance of trNK and cNK cell subsets in the kidney under baseline conditions, in mice deficient in either NFIL3, which is required for cNK cell development (53, 54), or Tbet (Tbx21), which contributes to NK cell maturation (38) and is required for the development of liver and skin but not uterine trNK cells (43). Kidney trNK cells were present at or above levels found in C57BL/6 mice both by total cell number and by frequency in NFIL3-deficient mice (Fig. 2A–B). Conversely, cNK cells were largely absent from the kidney in NFIL3-deficient mice, consistent with previous reports identifying NFIL3 as an essential regulator of cNK development (53, 54). trNK cells were also present at a normal frequency in Tbet-deficient mice, although Tbet-deficient mice had reduced number of NK cells, with ~50% the number of trNK cells and only ~20% the number of cNK cells relative to C57BL/6J mice (Fig. 2A–B). AsGM1 was expressed at lower levels in trNK cells relative to cNK cells in both NFIL3-deficient and Tbet-deficient mice, consistent with AsGM1 expression levels observed in trNK cells in wild-type mice (Fig. 2C–D). While differential AsGM1 expression was observed between trNK and cNK cells in all strains of mice, NFIL3-deficient mice have reduced AsGM1 expression in cNK cells and fewer AsGM1 low trNK cells than in wild-type mice (Fig. 2C–D). These studies show that kidney trNK cells develop independent of NFIL3 and Tbet.

Asialo-GM1 is preferentially expressed on cNK cells compared to trNK cells across multiple strains and organs

Given the broad use of AsGM1 NK depleting regimen on various strains of mice, we sought out to determine if preferential expression of AsGM1 on cNK cells was exclusive to the kidney and B6 mice. To determine whether trNK cells and cNK cells had differential expression of AsGM1 in additional tissues and genetic backgrounds, we analyzed these NK cell subsets in different strains of mice using NKp46, a common marker for NK cells (55). Notably, NKp46 is interchangeable with NK1.1 when identifying trNK and cNK cell subsets in the kidney of B6 mice (Supplemental Fig. 1).

When we analyzed the relative frequency of cNKs and trNKs across common inbred strains of mice (B6, NOD, BALB/c and 129), we found that trNK frequency was relatively comparable across different backgrounds, with NOD mice showing a slightly increased trNK frequency in some organs (Fig. 3A). While certain organs consistently had low frequencies of trNK cells (spleen, lung), other organs uniformly had elevated frequencies of trNKs (liver, kidney, and pancreas) regardless of background (Fig. 3A, Supplemental Fig. 1B–C). The frequency of uterine trNK cells varied dramatically between different strains of mice with BALB/c and 129 strains of mice having low frequencies of trNK cells compared to B6 and NOD mice (Fig. 3A).

Next, we analyzed the expression of AsGM1 on cNK and trNK cells in various organs and genetic backgrounds. cNK cells uniformly expressed high levels of AsGM1 across all organs analyzed, regardless of strain (Fig. 3B, Supplemental Fig. 1D). In contrast, trNK cells had differential expression of AsGM1, with the majority of trNK cells expressing only low levels of AsGM1 in the kidney, spleen, liver, and lung (Fig. 3B). While a subset of trNK cells had comparable expression of AsGM1 relative to cNKs (e.g. in the liver of B6 mice), the only instance where trNK cells uniformly expressed high levels of AsGM1 comparable to cNK cells was in the uterus of some strains of mice (B6, 129) (Fig. 3B). Analysis of NK cell subsets using AsGM1 as a primary parameter relative to either CD49a or DX5 further identified CD49a+ trNK cells with low, intermediate and high levels of AsGM1 expression, varying by organ and genetic background (Supplemental Fig. 2). Taken together, these data show that cNK cells uniformly express high levels of AsGM1 and that trNK cells contain a phenotypically heterogeneous population characterized by variable, but reduced expression of AsGM1 across multiple organs and strains of mice.

Asialo-GM1 antibody treatment preferentially depletes cNK cells in B6 and BALB/c mice and enhances the relative frequency of trNK cells

Anti-AsGM1 antibodies have been commonly used to deplete NK cells across diverse genetic backgrounds. Our data indicate that while AsGM1 is uniformly expressed at a high level in cNK cells, it is variably expressed in trNK cells, with a many trNK cells expressing low levels of AsGM1. We next sought to determine whether the presence of AsGM1^low trNK cells may reduce the susceptibility of trNK cells to be depleted by AsGM1 antibody treatment in B6 mice. AsGM1 antibody treatment resulted in a pronounced decrease in the frequency and number of total NK cells across multiple organs, consistent with the overall higher frequency of cNK cells relative to trNK cells (Fig. 4A–B, Supplemental Fig. 3A). While AsGM1 treatment reduced the number of both cNK and trNK cells (Supplemental Fig. 3B), the magnitude of depletion between cNK cells and trNK cells profoundly differed. Indeed, when we measured the relative frequency of cNK cells and trNK cells, we found that AsGM1 antibody treatment resulted in a pronounced shift in the relative frequency of cNK cells and trNK cells within the kidney, liver, lung, and spleen (Fig. 4A–B). These data indicate that in B6 mice trNK cells have reduced susceptibility to AsGM1-based antibody depletion.

While AsGM1 antibody depletion is used in B6 mice, it is often used as a method to validate results obtained by NK1.1 antibody depletion. In contrast, certain strains of mice (e.g. BALB/c) do not express the PK136-reactive NK1.1 allele, such that AsGM1 antibody is the primary method to test the role for NK cells in vivo (23). Given that BALB/c mice also contain AsGM1^low trNK cells, we next tested the selectivity of AsGM1 antibody depletion on cNK and trNK cells in this strain. AsGM1 antibody treatment again resulted in a large reduction in the frequency of NK cells and an overall decrease in cNK and trNK cell numbers (Supplemental Fig. 3A,C), with AsGM1 antibody treated mice showing a notable skewing in the relative frequency of trNK cells in the liver, lung and spleen (Fig. 4C–D). In total, these data indicate that AsGM1 antibody treatment preferentially depletes cNK cells, resulting in a relative sparing of trNK cells across multiple organs and two common inbred strains of mice.

Tissue-resident NK cells and conventional NK cells in IRI of the kidney

NK cells have been implicated in ischemia-reperfusion injury (IRI) in the kidney but the role of trNK and cNK cells is unclear (45–47). We sought to characterize the dynamics of trNK cells and cNK cells during ischemic injury in the kidney by employing the hanging weight system of IRI, a well-established model of acute kidney injury (49). In mice subjected to 30 minutes of ischemia, kidney function as measured by glomerular filtration rate (GFR) rapidly declined within 4 hours after reperfusion, with maximal loss of kidney function occurring by 24 hours (Fig. 5A). We next determined the total number of NK cells at 4, 8, and 24 hours post reperfusion, and found that there were no significant differences in the total number of NK cells present in the kidney following this acute ischemic injury (Fig. 5B). Given that NK cell numbers did not change dramatically, we next determined if there were changes in the relative distribution of trNK and cNK cells or their activation phenotype during this acute time course. At 4 and 24 hours post-reperfusion, there were no changes in the relative distribution of trNK cells and cNK cells in the kidney (Fig. 5C,E). Furthermore, when we analyzed the activation status of trNK cells and cNK cells during IRI, we found that trNK cells expressed higher levels of CD160, CD44, and TRAIL than cNK cells at both 4 and 24 hours, in both the naïve (sham) and ischemic kidney, with no discernable difference in either NK cell subset following ischemic injury (Fig. 5D,F). These distinct phenotypes mirrored differences observed in trNK and cNK cells in the steady state. Taken together, these data suggest that trNK cells and cNK cells have a relatively stable number and cell surface phenotype during acute ischemic kidney injury, with trNK cells showing evidence for a heightened activation state.

NK1.1 depleting antibody protects mice from kidney dysfunction during IRI

The use of anti-NK1.1 antibody as a depleting strategy for NK and NKT cells is well established in the field. While anti-NK1.1 depleting studies have been shown to be protective in some studies of ischemic kidney injury (47), the role of NK cells in the hanging-weight model of kidney injury remains to be elucidated. To test this, B6 mice were treated with either isotype or αNK1.1 depleting antibody (administered on day −3 and −1), subjected to IRI on day 0 and allowed to recover for 1 day. Depletion of NK cells (CD45+CD3−NKp46+) was confirmed in the kidney by flow cytometry (Fig. 6A), with NK1.1 antibody treatment resulting in an equivalent depletion of both cNK cells and trNK cells (Fig. 6B,C). Whereas isotype treated animals subjected to IRI had a pronounced impairment in kidney function as measured by GFR, αNK1.1 treated mice showed a clear improvement in kidney function with GFR values nearly restored to values measured from sham, non-ischemic mice (Fig. 6D).

The NK1.1 antigen is expressed by both NK and NKT cells, both which have been implicated in acute kidney injury (AKI) (45, 47, 56, 57). To test the role of NKT cells in this model, we quantified the number of NKT cells (CD45+CD3+NK1.1+), at 4, 8 and 24 hours post-IRI and found no significant changes in total numbers (Supplemental Fig. 4A–B). Importantly, when comparing the number of NK and NKT cells in the ischemic kidney, NK cells were much more abundant than NKT cells (Supplemental Fig. 4C).

Given that the NK1.1 antibody can deplete both NK and NKT cells, we sought to test the impact of NK1.1 antibody treatment on mice genetically deficient in NKT cells (e.g. Jα18- or CD1d-deficient mice). While mice deficient in the T cell receptor gene segment Jα18 have a deficit in NKT cells, recent studies have shown that these mice also have major alterations in T cell receptor repertoire (58). Based on this, we tested CD1d1 x CD1d2-deficient mice, which are defective in NKT cell development, without other known defects in lymphocyte development (59, 60). As expected, αNK1.1 antibody treatment profoundly depleted NK cells in CD1d-deficient mice (Supplemental Fig. 4D). Whereas isotype-treated CD1d-deficient mice subjected to ischemia showed decreased GFR, similar to wild-type mice, αNK1.1 antibody treatment significantly improved GFR in both CD1d-deficient, and wild-type, ischemic mice (Fig. 6D). Based on these data, we conclude that the primary protective effect of αNK1.1 antibody treatment results from depletion of NK cells, not NKT cells.

Asialo-GM1 preferentially ablates cNK cells, showing trNK cells promote AKI

While the above data identify NK cells as mediators of ischemic injury, they do not discriminate between the relative contributions of trNK cells and cNK cells, since NK1.1 antibody treatment equivalently depletes both NK cell subsets. Given the differential expression of AsGM1 on trNK cells and cNK cells in the kidney, and their differential susceptibility to AsGM1 antibody depletion, we hypothesized that AsGM1 antibody treatment may allow us to dissect the relative contributions of trNK cells and cNK cells to ischemic kidney injury. To test this, we treated B6 mice with αAsGM1 antibody depletion and subjected these mice to either sham (non-ischemia) or ischemic injury. Anti-AsGM1 antibody treatment resulted in a profound reduction of kidney NK cells in these mice, comparable to the magnitude of depletion observed following αNK1.1 antibody treatment (compare Fig. 6B, Fig. 7A). Consistent with our previous data, αAsGM1 antibody resulted in a prominent increase in the relative frequency of trNK cells in the ischemic kidney (Fig. 7A,B).

Next, we tested the consequence of antibody treatment of kidney injury and function. First, we examined the consequence of αNK1.1 and αAsGM1 antibody treatment on the early induction of apoptosis after ischemic injury, measured by active caspase-3 staining at 4 hours post-reperfusion. These studies showed that αNK1.1 antibody-injected mice had a significantly decreased number of apoptotic cells relative to αAsGM1 antibody-injected mice (Supplemental Fig. 4E–I). To interrogate the consequence of these antibodies on kidney function at 24 hours reperfusion, we measured kidney function by GFR, a gold standard for kidney function, comparing isotype, αAsGM1, and αNK1.1 antibody-injected mice. Strikingly, while αNK1.1 antibody treatment conferred protection, αAsGM1 antibody treatment failed to protect mice from IRI despite similar reduction in overall NK cells percentages (Fig. 7C). Given that αAsGM1 poorly depletes trNK cells, in contrast to αNK1.1 antibody treatment, these data identify that trNK cells are sufficient to promote IRI in the kidney.