Dual inhibition of mTOR and calcineurin pathways mitigates missing self–induced NK cell–mediated microvascular rejection

Significance

Our translational study focused on a mechanism of microvascular rejection driven by natural killer (NK) cells activated through “missing self” recognition, independent of traditional donor-specific antibody-mediated pathways. Our study identifies calcium and mTOR signaling as critical regulators of missing self-induced NK cell activation. Using both human and animal models, we demonstrated that dual inhibition with mTOR and calcineurin inhibitors synergistically suppresses vascular injury. In a pilot clinical study, this approach significantly improved graft outcomes. Our findings redefine the immunological landscape of graft rejection and introduce a targeted therapeutic strategy. By addressing a major unmet need in transplantation, our work paves the way for improving long-term graft survival through precision immunomodulation.

Abstract

The inability of graft endothelial cells to deliver HLA-I-dependent inhibitory signals to recipient natural killer (NK) cells (missing self, MS), drives donor-specific antibody-independent microvascular inflammation (MVI), leading to graft failure. This study aimed to elucidate the signaling pathways involved in MS-associated NK cell activation and explore therapeutic strategies. Analyses of kidney graft biopsies identified calcium signaling pathways and mTOR as a key regulator of MS-induced NK cell activation. Two experimental models were developed to mimic the pathological condition: in vitro cocultures of human NK cells with allogeneic microvascular endothelial cells and a murine heart transplantation model. These models showed that while calcineurin inhibitor (CNI) alone had a limited impact, combining CNI with mTOR inhibitors (mTORinh) synergistically reduced NK cell activation and endothelial damage. In a pilot clinical study involving 50 renal transplant recipients with MS-associated NK cell–mediated microvascular inflammation, patients who tolerated mTORinh introduced on top of CNI at diagnosis demonstrated reduced MVI lesions and improved graft survival compared to a historical cohort left on CNI and mycophenolate mofetil. This translational study identifies mTOR inhibition as a pivotal adjunct to CNI in mitigating MS-associated NK cell–mediated inflammation, potentially improving long-term graft outcomes.

Vital organ failure is a life-threatening condition resulting from the inability of a vital organ to perform its essential functions. In developed countries, increased life expectancy and the rising prevalence of hypertension, obesity, and diabetes, have significantly elevated the risk and incidence of vital organ failure, which is now recognized as a leading cause of disability and premature death (www.who.int).

Since its introduction in the mid-1950s, solid organ transplantation has rapidly become the primary—and often the only—viable treatment for patients with end-stage organ failure. Despite remarkable advancements in transplantation over recent decades, long-term success remains limited by the progressive loss of graft function, primarily due to a process known as “rejection” (1).

The prevailing immunopathological model attributes rejection to genetic disparities between donor and recipient, resulting in the recognition of alloantigens—primarily mismatched donor-specific HLA molecules—by the recipient’s adaptive immune cells (T and B lymphocytes) (2, 3). Distinct mechanisms of allorecognition activate different adaptive immune effectors (2, 3), leading to histological patterns thought to be characteristic of each rejection endotypes. In T cell mediated rejection (TCMR), alloreactive cytotoxic T cells infiltrate the graft, causing interstitial inflammation and attacking tubular epithelial cells, with minimal impact on graft vasculature (4). Conversely, the differentiation of alloreactive B cells into donor-specific antibody (DSA)-producing plasma cells, either in recipient secondary lymphoid organs or in tertiary lymphoid structures within the graft (5, 6), triggers antibody-mediated rejection (AMR), predominantly affecting graft vasculature. Due to their size, donor-specific antibodies (DSAs) are largely confined to the recipient’s circulation and primarily bind to accessible allogeneic targets on the surface of donor-derived graft endothelial cells (7). When present in sufficient quantities in recipient’s circulation, DSA binding on graft endothelial cells form hexameric platforms, which recruit C1q and activate the classical complement pathway (8). This leads to acute damage to the graft vasculature and rapid graft loss (9–11). Lower DSA titers, insufficient to trigger complement activation, can still cause subclinical vascular injury (12, 13). This is due to an alternative mechanism: CD16-mediated activation of recipient natural killer (NK) cells, driving antibody-dependent cell-mediated cytotoxicity (ADCC) (14–17). While NK cells are thought to play a critical role in initiating this pathophysiological cascade, they do not act alone. Through the release of proinflammatory mediators such as MIP-1β and IFN-γ, NK cells trigger an amplification loop that recruits additional immune cells, including monocytes and neutrophils, which further exacerbate the microvascular inflammation (MVI) characteristic of antibody-mediated rejection (AMR) (16, 18).

Reflecting this model, the Banff classification has historically associated interstitial inflammation (i and t scores) with TCMR, while microvascular inflammation (MVI, g and ptc scores) was considered a hallmark of AMR (19).

However, our group recently made an intriguing observation (20) later confirmed by independent studies (21, 22) that a significant subset of patients with microvascular inflammation (MVI) lack detectable circulating anti-HLA donor-specific antibodies (DSAs). While some cases may involve non-HLA DSA (23–25) for which no validated detection assay currently exists, we pursued an alternative hypothesis. Using experimental models, we demonstrated that the absence of self-HLA I-dependent inhibitory signals from graft endothelial cells—referred to as “missing self” (MS)—can independently drive MVI lesions, resulting in graft injury similar in severity to AMR (20, 26–28).

In this translational study, we aimed to externally validate our theory in an independent cohort of kidney transplant patients. By identifying the molecular mechanisms underlying MS-induced NK cell activation, we also sought to explore potential therapeutic strategies for managing MS-associated NK cell–mediated microvascular inflammation.

Results

DSA- and MS-Associated Microvascular Inflammation Have a Similar Detrimental Impact on Graft Survival.

The screening of the graft biopsies conducted between January 1, 2015, and December 31, 2018, in the cohort of renal transplant recipients followed at the Imperial College (London, UK) identified 1,621 biopsies from 982 patients, including 90 patients (9.2%) with MVI (Fig. 1A). Among these, 42 patients (46.6%) had detectable anti-HLA DSA in their circulation, meeting the Banff criteria for antibody-mediated rejection (AMR) (29). After excluding donor-recipient pairs with unavailable DNA (n = 1) or genetically predicted MS (n = 16), 25 patients formed the MVI+DSA+MS- (i.e., “pure AMR”) group.

Fig. 1.

Missing self-associated NK cell–mediated microvascular inflammation in kidney transplant recipients. (A) Flow chart detailing the selection process to form the 3 patient groups: MVI+DSA−MS+ (red, pure MS-associated MVI), MVI+DSA+MS− (orange, pure antibody-mediated rejection), and MVI−DSA−MS− (gray, controls). (B) The intensity of the MVI lesions [i.e., glomerulitis (g) and peritubular capillaritis (ptc) Banff scores] was compared between the MVI+DSA−MS+ (red) and the MVI+DSA+MS− (orange) group. ns: P > 0.05; Mann–Whitney test. (C) Renal graft survival from the time of the diagnosis of MVI were compared in the three groups (Log rank < 0.0001). ns: P > 0.05; *P ≤ 0.05, ****P < 0.0001; log-rank test. (D) The 770 genes from the BHOT consensus panel significantly dysregulated between MVI+DSA−MS+ (n = 12 biopsies) patients vs. MVI−DSA−MS− (n = 48 biopsies) are highlighted in red in the volcano plot. (E) GSEA was performed in MVI+DSA−MS+ patients vs. MVI−DSA−MS−, focusing on the selected relevant biological pathways. Normalized enrichment scores (NES) are plotted, with the corresponding false discovery rate (FDR) q-values represented using a color-coded scale.

Among the remaining 48 patients (53.3%) with MVI lesions but without detectable anti-HLA DSA, 5 had a positive antiendothelial cell crossmatch, suggesting non-HLA DSA-induced AMR. An additional 30 pairs were excluded due to unavailable DNA (n = 1) or absence of genetically predicted MS (n = 29), leaving 13 patients to form the MVI+DSA−MS+ (i.e., “pure MS”) group. The molecular definition of MS in these patients is shown in SI Appendix, Fig. S1 A and B.

Finally, 485 patients (49.4%) without MVI (g + ptc = 0) or other histological lesions were evaluated for circulating anti-HLA DSA (n = 17) or genetically predicted MS (n = 162), resulting in the identification of 306 patients who served as the negative control group (MVI−DSA−MS−).

The clinical characteristics of the 3 groups are presented in SI Appendix, Table S1. The severity of acute and chronic histological lesions was comparable between the MVI+DSA+MS− and MVI+DSA−MS+ groups (SI Appendix, Fig. S1C), including for the MVI (Fig. 1B).

Graft survival was similarly reduced in the MVI+DSA−MS+ and MVI+DSA+MS− groups compared to MVI−DSA−MS− (Fig. 1C; Log rank, P < 0.0001).

Importantly, not all patients with genetically predicted MS, developed microvascular inflammation. In fact, a large proportion (162/468, 34,6%) of patients without DSA had a genetically predicted MS but no MVI lesion on graft biopsy, and the latter maintained the same excellent graft survival as MVI-DSA-MS- patients without genetically predicted MS (SI Appendix, Fig. S1D; Log rank, P = 0.827).This observation, which was also made in another independent cohort (20), led us to put forward the hypothesis that MS risk factor may be partially controlled by the therapeutic immunosuppression.

MS-Associated Microvascular Inflammation Correlates with Intragraft NK Cell Activation.

To explore the molecular markers associated with MS-associated MVI, a transcriptomic analysis of graft biopsies was conducted. A substantial proportion (128, 16,6%) of the 770 genes in the Banff Human Organ Transplant (B-HOT) consensus panel (30, 31) were differentially expressed in the MVI+DSA−MS+ group (Fig. 1D) as well as in the MVI+DSA+MS− (SI Appendix, Fig. S1E), compared to MVI−DSA−MS−. However, in line with previously published work (21, 22) no difference could be evidenced when biopsies of MVI+DSA−MS+ and MVI+DSA+MS− groups were directly compared (SI Appendix, Fig. S1F), suggesting that MS and DSA trigger the same final common immunopathological pathway (26). Gene set enrichment analysis (GSEA) identified key biological processes associated with MS-associated microvascular inflammation, highlighting NK cell cytotoxicity, and interferon-gamma response as central mechanisms responsible for graft inflammation (Fig. 1E). Strikingly, calcium signaling pathway was among the most upregulated gene set in grafts from MVI+DSA−MS+ patients (Fig. 1E). These findings were further corroborated using an alternative analytical approach (SI Appendix, Fig. S1G).

Establishing an In Vitro Model to Investigate MS-Induced NK Cell Activation.

Since transcriptomic analyses were conducted on whole tissue samples, it remains unclear whether the upregulated calcium signaling pathway in MS-associated microvascular inflammation directly involves NK cells.

To address this issue, an in vitro model replicating MS-induced NK cell activation was developed. In this model, purified primary human NK cells were cultured on a layer of human glomerular endothelial cell line (32), which lacks expression of HLA-I molecules of the C2 type. NK cell donors were selected as a subset of their NK cells expressed the educated (C2-specific) inhibitory receptor KIR2DL1, simulating the MS condition with endothelial targets (Fig. 2A). NK cells expressing KIR2DL1 cultured alone served as an external negative control, while NK cells expressing other educated inhibitory receptors (C1-specific: KIR2DL2/KIR2DL3 or Bw4-specific: KIR3DL1), for which endothelial targets presented the corresponding HLA-I ligand, were used as internal negative controls (Fig. 2A).

Fig. 2.

CNI impedes MS-induced NK cell–mediated activation in vitro. (A–C) An in vitro model of MS-induced NK cell activation was developed. Purified NK cells from healthy donors, selected to express HLA molecules from the C2 group, were cocultured with CiGENC, a human glomerular endothelial cell line expressing HLA-Bw4 and HLA-C1 but lacking HLA-C2. This created a missing self-scenario for KIR2DL1+ NK cells, while KIR2DL2/3+ and KIR3DL1+ NK cells served as internal negative controls. KIR2DL1+ NK cells cultured alone served as external negative control. (A) Schematic representation of the model. (B) Gating strategy of the flow cytometry analyses performed after 4H of coculture on live NK cells expressing a single type of inhibitory KIR. (C) Effector functions were monitored based on the expression of CD107a (degranulation), MIP-1β (chemokine) and IFNγ (inflammatory cytokine) and compared between cocultured KIR2DL1s+ (red dots) vs. resting KIR2DL1s+ (open red circles), or vs. cocultured KIR2DL2/3s+ and KIR3DL1s+ (gray dots) NK cells. Individual data from 18 donors and mean ± SD. *P ≤ 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; Wilcoxon test or Kruskal–Wallis test. (D–F) Analysis of kinase-mediated signaling in NK cells stimulated by MS. (D) Schematic representation of the phosphorylation cascade. (E) The list of kinases whose activity was significantly dysregulated (kinase scores >1,3) in sorted KIR2DL1s+ NK cells in MS conditions vs. cultured alone (negative control) is shown for phosphotyrosine kinases (PTK, Left, n = 1 donor) and serine-threonine kinases (STK, Right, n = 2 donors). (F) Kinome tree showing all differentially expressed kinases between MS and control conditions. (G–I) Imaging flow cytometry was used to assess calcineurin pathway activation in NK cells stimulated by MS. (G) Representative images are shown for the 3 groups: resting KIR2DL1s+ NK cells (Upper row, open red circles), KIR2DL1s+ NK cells stimulated by MS without (Middle row, red dots) or with (Lower row, blue dots) CsA. The translocation of NFAT to the nuclei of KIR2DL1s+ NK cells was analyzed using two metrics: (H) the median similarity score, which quantifies the proportion of NFAT staining colocalizing with DAPI nuclear staining, (I) and the percentage of NK cells with a median similarity score >1, individual data from 7 different donors, mean ± SD. *P ≤ 0.05, ***P < 0.001; Friedman test. (J) Effector functions were monitored by flow cytometry and compared between KIR2DL1s+ NK cells resting (open red circles) or stimulated by MS without (red dots) or with CsA (blue dots). Individual data from 16 donors and mean ± SD. *P ≤ 0.05, **P < 0.01, ****P < 0.0001; Friedman test. (K and L) An in vitro model for MS-induced NK cell–mediated rejection was developed to test the therapeutic impact of CsA. NK cells were exposed to the drug overnight before coculture. (K) Schematic representation of the model. (L) The viability of CiGENC target cells was monitored by real-time Impedancemetry. Data are normalized to the control condition without MS (dotted gray line). Left: Impedance curve of induced MS (red) and induced MS+CsA (blue) conditions of the same representative donor are shown. Right: Viability of target cells, evaluated by the area under the impedance curve, was compared in induced MS (red dots) vs. Induced MS+CsA (blue dots) conditions. Individual data from 6 donors and mean ± SD. Wilcoxon test.

Educated NK cells expressing only KIR2DL1 (KIR2DL1s+; Fig. 2B and SI Appendix, Fig. S2A) in MS conditions with endothelial targets consistently exhibited higher expression of CD107a (a marker of degranulation and cytotoxicity), MIP-1β (a monocyte chemoattractant), and IFN-γ (a key cytokine in microvascular inflammation identified in our previous analyses, Fig. 1E) compared to all negative control conditions (Fig. 2C). These results demonstrate that this in vitro model is well suited to study MS-induced NK cell activation.

MS Triggers Calcium-Mediated Signaling in NK Cells.

The activity of phosphotyrosine kinases (PTKs) and serine-threonine kinases (STKs) were compared between sorted KIR2DL1s+ NK cells cultured alone (negative control) vs. cultured with human glomerular endothelial cells lacking C2-type HLA-I molecules (MS conditions). As expected, MS induced the activation of the upstream Src family kinases (Fyn, Lyn, Hck), which are involved in ITAM motif phosphorylation, as well as the critical mediators of intracellular signal propagation: ZAP70 and Syk kinases, (Fig. 2 D–F). Additionally, and in line with the human transcriptomic data presented above (Fig. 1E), MS also triggered the activation of ACG family kinases (PRKCA, PRKACA, PRKCD, PRKCB, PRKG2, PRKG1) and CAMK family kinases (CAMK2A, CAMK2D, CAMK4, CAMK2B, CAMK2G), which are linked to the calcium-mediated signaling (Fig. 2 D–F).

Beyond kinase activation, calcium signaling also involves the phosphatase calcineurin, which is undetectable by the kinase screening assay that was used above. To assess calcineurin activity in NK cells exposed to MS, we evaluated NFAT transcription factor translocation using imaging cytometry (SI Appendix, Fig. S2 B and G). NFAT translocation, defined by colocalization of NFAT labeling with nuclear DAPI staining (i.e., similarity score; Fig. 2H) and the proportion of NK cells with a similarity score >1 (Fig. 2I) were both significantly higher in KIR2DL1s+ NK cells activated by MS than in the same cells cultured alone.

Calcineurin Inhibition Partially Mitigates MS-Induced NK Cell Inflammation In Vitro.

Calcineurin inhibitors (CNI) are the cornerstone of maintenance immunosuppression. Indeed, all MVI+ patients in the Imperial College cohort were receiving either cyclosporin A (CsA) or tacrolimus. Given the role of the calcium signaling pathway in MS-induced NK cell activation and the observation that only a subset (estimated ~10%) of patients with genetically predicted MS develop MS-associated microvascular inflammation, we hypothesize that CNI may help controlling MS-associated NK cell–mediated inflammation. To test this hypothesis, human purified NK cells were exposed to therapeutically relevant concentrations of CsA before being cocultured with human glomerular endothelial cells deficient in C2-type HLA-I molecules. As anticipated, CsA inhibited the MS-induced translocation of the NFAT transcription factor in KIR2DL1s+ NK cells (Fig. 2 G–I), leading to a significant reduction in the acquisition of effector functions by CsA-treated KIR2DL1s+ NK cells exposed to MS (Fig. 2J).

To determine whether the inhibition of effector function acquisition by NK cells activated by MS in the presence of CsA translates into less damage to endothelial targets, we designed another in vitro model emulating MS-induced NK cell–mediated inflammation. Unlike the model used to analyze MS-induced NK cell activation (in which only the minor fraction of NK cells expressing KIR2DL1 were activated by MS, Fig. 2A), all NK cell donors were selected based on HLA-I compatibility with the human glomerular endothelial cells in this rejection model. MS was then induced (or not) in all NK cell subsets in the coculture by adding blocking anti-KIR mAbs (Fig. 2K). The condition with isotype control mAbs (no MS) was used for normalization across donors (gray dotted line; Fig. 2 L, Left), assessing the survival of endothelial targets when NK cells were pretreated (blue curve; Fig. 2 L, Left) or not (red curve; Fig. 2 L, Left) with CsA. Comparing the normalized areas under the curve, which serves as a surrogate for endothelial cell survival, showed a trend toward partial CsA-mediated protection of endothelial cells from MS-induced NK cell activation damage (Fig. 2 L, Right).

Establishing an In Vivo Model to Investigate the Impact of CNI on MS-Induced NK Cell Activation.

To further investigate the therapeutic impact of CNI in preventing MS-associated NK cell–mediated microvascular inflammation, we transition to a more representative preclinical in vivo model. Hearts from either wild-type (WT) or β2-microglobulin knockout (β2mKO) C57BL/6 mice were transplanted into C57BL/6 recipients (Fig. 3A). In this model, no alloantigens are introduced, as both donors and recipients share the same genetic background, thus excluding adaptive alloimmune responses. However, because β2mKO donors lack MHC-I expression, these grafts are expected to induce MS-mediated microvascular inflammation. Consistent with this hypothesis, MVI lesions, resembling those observed in kidney graft biopsies from MVI+DSA−MS+ patients, were found in heart grafts from β2mKO donors but not in WT controls (Fig. 3B). This rejection model is fully dependent on the presence of recipient’s NK cells since depletion of this subset before the transplantation completely abrogated the development of MVI lesions in β2mKO grafts. RNA sequencing analysis of β2mKO murine heart grafts revealed upregulation of all key biological pathways identified in kidney graft biopsies of MVI+DSA−MS+ patients (Fig. 3 C and D). To further validate that this murine model reliably reflects human MS-associated microvascular inflammation, we ranked the changes in gene expression between β2mKO and WT heart grafts for the 568 murine orthologues of the B-HOT panel. These data were then compared to those from Nanostring analyses of kidney graft biopsies from MVI+DSA−MS+ and MVI−DSA−MS− controls. In addition to the striking similarity in the expression profiles shown in the heatmap Fig. 3E, statistical analysis confirmed a strong correlation between the two datasets (Fig. 3F; Spearman’s rank correlation test: Rho = 0.65; P < 0.0001, SI Appendix, Fig. S3A).

Fig. 3.

CNI monotherapy has limited efficacy in a murine model of MS-induced NK cell–mediated microvascular inflammation. (A–F) A murine in vivo preclinical model of MS-induced NK cell–mediated microvascular inflammation was developed. WT C57BL/6 (B6) mice were transplanted with either a WT B6 (gray; control, Ctl group) or a C57BL/6 β2-microglobulin KO (red; missing self, MS group) heart graft. Some recipients were treated with an anti-NK1.1 depleting mAb (aNK1.1). (A) Schematic representation of the model. (B) CD45 staining of the heart graft was performed to compare the immune cell infiltration in controls (gray dots, n = 8), MS group (red dots, n = 8), and aNK1.1 group (purple dots, n = 4). Left: Representative immunohistochemical findings are shown. (Scale bar: 100 μm.) Right: The intensity of the immune cell infiltration was graded by a blinded trained pathologist on a semiquantitative scale (score 0 to 3). Individual values, as well as box and whiskers (min to max) are shown. *P ≤ 0.05, **P < 0.01; Kruskal–Wallis test. (C–E) RNA sequencing was performed on the heart grafts from MS (n = 5) vs. Ctl (n = 5) group. (C) Genes significantly dysregulated in MS vs. Ctl are highlighted in red in the volcano plot. (D) GSEA was performed on genes dysregulated in MS vs. Ctl, heart grafts focusing on the same relevant biological pathways as in patients’ biopsies (Fig. 1). (E) Heat map comparing the FC ranking of the 568 BHOT panel genes with identified mouse orthologs between: (Right column) Nanostring data from MVI+DSA−MS+ vs. Ctl patients, and (Left column) RNA-seq data from MS vs. Ctl murine heart grafts. (F) Correlation of FC ranks for the 568 BHOT panel genes with identified mouse orthologs. Nanostring data from MVI+DSA−MS+ vs. Ctl patients (x-axis) and RNA-seq data from MS vs. Ctl murine heart grafts (y-axis). ***P < 0.001, Rho = 0.646; Spearman test. (G–I) The murine preclinical model of MS-induced NK cell–mediated microvascular inflammation was used to test the therapeutic potential of CsA in vivo. (G) Schematic representation of the experiments. (H) GSEA performed in MS+CsA (n = 5) vs. MS (n = 5) heart grafts focusing on calcium signaling pathways and NK cell activation. (I) CD45 staining of the heart graft was graded on a semiquantitative scale (score 0 to 3) to compare the immune cell infiltration in MS (red dots, n = 8) and MS+CsA group (blue dots, n = 5), Left: Representative immunohistochemical findings are shown. (Scale bar: 100 μm.) Right: Individual values, as well as box and whiskers (min to max) are shown. Mann–Whitney test.

However, despite the promising in vitro results shown Fig. 2 and a reduction in the expression of the genes associated with calcium-mediated signaling and NK cell activation in vivo (Fig. 3 G and H), CsA alone demonstrated only a modest beneficial effect on the development of MS-induced MVI lesions in the preclinical murine model (Fig. 3I). The limited therapeutic effect of CNI on MS-induced microvascular inflammation may arise from its selective inhibition of the calcineurin pathway, leaving other calcium-dependent signaling pathways in NK cells activated in response to missing-self-recognition (Fig. 2). This observation aligns with evidence that some patients with genetically predicted MS still develop MVI despite CNI treatment (Fig. 1), highlighting the urgent need for more effective therapeutic strategies.

Inhibiting mTOR Pathway Alleviates MS-Induced NK Cell–Mediated Microvascular Inflammation.

In addition to calcium-mediated signaling, the mTOR pathway was also consistently identified as being upregulated in both the biopsies from patients with MS-associated microvascular inflammation (Fig. 1E and SI Appendix, Fig. S1E) and in β2mKO heart grafts (Fig. 3D). To validate these molecular findings and determine whether mTOR pathway was operating in MS-induced NK cell activation, multiplex immunofluorescence staining was performed on a subset of MVI+DSA−MS+ patient graft biopsies (n = 10). Representative results shown in Fig. 4A confirm the presence in the peritubular capillaries (CD34⁺) of kidney grafts of NK cells (CD3⁻ NKp46⁺ Eomes^+/−), which were also positively stained for the phosphorylated form of S6rp (pS6rp), a downstream target of the mTORC1 complex.

Fig. 4.

Inhibition of the mTOR pathway improves experimental MS-induced NK cell–mediated microvascular inflammation. (A) Representative images of paraffin-embedded kidney graft biopsy sections from MVI+DSA−MS+ patients labeled against NKp46 (orange), CD68 (magenta), CD3 (cyan), Eomes (yellow), P-S6rp (red), and CD34 (green) are shown. DAPI was used to stain the nucleus. Arrows indicate NK cells (CD3⁻ CD68⁻ NKp46⁺ Eomes^+/−) in graft vasculature (CD34⁺) expressing the phosphorylated form of S6sp (pS6rp), a target of mTORC1 complex. Dotted scale bar: 500 μm, full scale bar: 10 μm. (B and C) The in vitro model presented Fig. 2 was used to analyze the impact of MS on mTOR signaling in NK cells. Purified human NK cells (labeled with Cell Trace Violet, blue) and CiGENC (labeled with Cell Trace Far Red, purple) were cocultured for 1 h. KIR2DL1 (green) and pS6rp (red) were stained and confocal microscopy was used to perform 3D reconstitution of the surface interactions of NK cells with endothelial cells. (B) Representative images are shown. (Scale bar: 20 μm.) (C) The median fluorescence intensity of pS6rp staining was measured across all NK cells in 3 to 4 randomly selected fields per well. The histogram compares four NK cell groups: NK cells cultured alone (black open circles), NK cells cocultured with endothelial cells but not in contact with the target (gray open circles), NK cells in contact with endothelial cells but negative for KIR2DL1 (gray dots), and KIR2DL1+ NK cells in contact with endothelial cells (red dots, MS condition). Individual data from 8 donors and mean ± SD. *P ≤ 0.05, **P < 0.01, ***P < 0.001; one way-ANOVA. (D) The murine in vivo preclinical model of MS-induced NK cell–mediated microvascular inflammation was used to test the therapeutic potential of the mTOR inhibitor rapamycin. Left: Schematic representation of the experiments. Middle: Representative immunohistochemical findings are shown. (Scale bar: 100 μm.) Right: CD45 staining of the heart graft was graded on a semiquantitative scale (score 0 to 3) to compare the immune cell infiltration in MS (red dots, n = 8, positive controls), MS+aNK1.1 (purple dots, n = 4, negative controls), MS+CsA (blue dots, n = 5), and MS+rapa (green dots, n = 6). Individual values, as well as box and whiskers (min to max) are shown. *P ≤ 0.05, **P < 0.01; Kruskal–Wallis test. (E–G) RNA sequencing was performed on the heart grafts from MS+rapa (n = 6) and MS (n = 5) group. (E) Genes significantly dysregulated in MS+rapa vs. MS are highlighted in red in the volcano plot. MS. (F) GSEA was performed in MS+rapa (n = 6) vs. MS (n = 5) heart grafts focusing on the same relevant biological pathways as in patients’ biopsies (Fig. 1). (G) Left: RNA-seq data were analyzed to assess gene expression changes in MS vs. control (y-axis) and MS+rapamycin vs. control (x-axis). Genes upregulated under pathological conditions and most significantly downregulated by mTOR inhibition are highlighted in purple in the top-left red triangle. Right: Emapplot showing enriched Gene Ontology (GO) terms (biological process BP) of the selected genes. The top 50 enriched GO terms are displayed, excluding terms associated with more than 200 genes. Links between GO terms indicate that they share over 60% of the same genes. (H–L) The 2 in vitro models presented Fig. 2 were used to test the therapeutic potential of the mTOR inhibitor rapamycin, alone and in combination with CsA. (H) Schematic representation of the experiments conducted to analyze MS-induced NK cell activation. (I) The median fluorescence intensity of phosphorylated-S6rp (pS6rp) staining was measured for KIR2DL1 NK cells in contact with CiGENC. The impact of a pretreatment of NK cells with rapamycin is shown. Individual data from 8 donors and mean ± SD. **P < 0.01, Wilcoxon test. (J) Effector functions were monitored by flow cytometry and compared between KIR2DL1s+ NK cells stimulated by MS in the absence of treatment (red dots, positive controls) or after exposure to CsA alone (blue dots), rapamycin alone (light green dots), or the combination of CsA+rapa (dark green dots). Individual data from 15 donors and mean ± SD. *P ≤ 0.05, **P < 0.01, ***P < 0.001 ****P < 0.0001; Friedman test. (K) Schematic representation of the experiments conducted to analyze MS-induced microvascular inflammation. (L) The viability of CiGENC target cells was monitored by real-time impedancemetry. Data are normalized to the control condition without MS (dotted gray line). Left: Impedance curve of induced MS (red curve), induced MS + CsA (blue curve), induced MS + rapa (light green curve), and induced MS + CsA and rapa (dark green curve) conditions of the same representative donor are shown. Right: Viability of target cells, evaluated by the area under the impedance curve, was compared in induced MS (red dots), induced MS + CsA (blue dots), induced MS + rapa (light green dots), induced MS + CsA and rapa (dark green dots) conditions. Individual data from 6 donors and mean ± SD. *P ≤ 0.05, **P < 0.01; Friedman test.

In line with these clinical results, we observed that donor NK cells that formed close contacts (SI Appendix, Fig. S4A) with human glomerular endothelial cells deficient in C2-type HLA-I molecules, had increased signaling through the mTOR pathway in vitro compared to NK cells not in contact. This signaling was significantly further enhanced in educated KIR2DL1+ NK cells, representing the MS condition (Fig. 4 B and C).

These results led us to test mTOR inhibitors, an established class of immunosuppressive agents for preventing organ rejection (33), in the murine preclinical model described in Fig. 3. Rapamycin treatment fully inhibited MS-induced microvascular inflammatory lesions, showing efficacy comparable to that of NK cell depletion (Fig. 4D).

Previous studies have reported that mTORC1 pathway also operates in endothelial cells in physiological (34) and pathological (including rejection) conditions (35). To shed light on the mechanism of protection of mTOR inhibitor against MS-induced microvascular inflammation, we first compare the transcriptomic profiles of β2mKO heart grafts from rapamycin-treated vs. control recipients. The results indicate that rapamycin’s therapeutic benefit may stem from modulation of the NK cell response (Fig. 4 E–G).

This conclusion is further supported by in vitro experiments (Fig. 4H), demonstrating that rapamycin treatment reduces MS-induced phosphorylation of S6rp in NK cells (Fig. 4I), resulting in diminished acquisition of effector functions by stimulated NK cells (Fig. 4J). Rapamycin’s therapeutic effect is not mediated through endothelial cells, as pretreating endothelial targets alone had no impact on NK cell activation, and treating both NK and endothelial cells showed no additive effect compared to treating NK cells alone (SI Appendix, Fig. S4 B and C).

Finally, we took advantage of the in vitro model to test the synergistic potential of a dual inhibition of mTOR and calcineurin pathways on the MS-induced microvascular inflammation. The results demonstrate that the dual therapeutic inhibition offers better control on MS-induced acquisition of effector functions by NK cells (Fig. 4 H and J), which in turn resulted in better preservation of endothelial targets in the in vitro model of MS-induced microvascular inflammation (Fig. 4 K and L).

Introduction of an mTOR Inhibitor on Top of CNI at Diagnosis of MS-Associated Microvascular Inflammation Improve Clinical Outcomes.

Two large clinical studies indicate that reduced-exposure to CNI combined with an mTOR inhibitor (mTORinh) can yield long-term outcomes comparable to the standard mycophenolate mofetil (MMF) with standard-exposure CNI maintenance immunosuppression (36, 37). Building on these data and the original results presented above, all MVI+DSA-MS+ kidney transplant recipients on CNI+MMF-based maintenance immunosuppression identified at Lyon University Hospital after Oct 2016 were switched from MMF to an mTOR inhibitor (CNI+mTORinh group, n = 16; Fig. 5A). Consistent with previous reports, late mTOR inhibitor initiation faced tolerance issues, with 56% of patients discontinuing treatment within 12 mo (SI Appendix, Fig. S5 A and B).

Fig. 5.

Dual inhibition of mTOR and calcineurin pathways mitigates missing self-associated NK cells-mediated microvascular inflammation. (A) Flow chart of the pilot study conducted in Lyon University Hospital. A historical cohort of MS-associated microvascular inflammation diagnosed before Oct 2016 in patients on CNI-based maintenance immunosuppression (n = 34; control group, red) was compared to 16 patients in whom an mTOR inhibitor was introduced on top of the CNI at diagnosis of MS-associated microvascular inflammation (CNI+mTORinh group, green). A second follow-up biopsy was performed for 11 patients from the control group (dotted red line) and 10 from the CNI+mTORinh group (dotted green line). (B) The intensity of the MVI lesions [i.e., glomerulitis (g) and peritubular capillaritis (ptc) Banff scores] was compared between control (red) and CNI+mTORinh groups (green) group. ns: P > 0.05; *P ≤ 0.05; Mann–Whitney test. (C) The differences between the mean eGFR calculated before and after diagnosis for patients from the control (red) and the CNI+mTORinh (green) group are plotted. Violin plot showing individual data and median for each group. ns: P > 0.05, **P < 0.01; Wilcoxon test for individual paired comparison of the mean eGFR before vs. after diagnosis. (D) Renal graft survival from the time of the diagnosis of MVI were compared in control (red line) and CNI+mTORinh patients (green line). *P ≤ 0.05; log-rank test. (E) Representative images of Masson’s trichrome stain from paraffin-embedded sections showing the MVI lesions: glomerulitis (g, Upper row) and peritubular capillaritis (ptc, Lower row) observed in a patient from the control group (Left) and the CNI+mTORinh group (Right) at diagnosis of MS-associated microvascular inflammation (Left column) and on a follow-up biopsy (Right column). (Scale bar: 100 μm.) (F) The individual difference in g (Left), ptc (Middle) Banff scores and microvascular inflammation (g + ptc, Right) between the diagnosis and follow-up biopsies are compared for patients of the Ctl (red) and CNI+mORinh (green) groups. Violin plot showing individual data. ns: P > 0.05, *P ≤ 0.05; Wilcoxon test for individual paired comparison of the score at diagnosis vs. on follow-up biopsy.

The clinical outcomes of the MVI+DSA−MS+ patients from CNI+mTOR group were compared to a historical cohort of MVI+DSA−MS+ patients, diagnosed between Jan 2004 and Oct 2016 where no therapeutic adjustments were made (CNI+MMF, Control group, n = 34; Fig. 5A). The two groups had similar histological lesions at diagnosis (SI Appendix, Fig. S5C) except for the g lesions that were slightly more intense for patients from the CNI+mTOR group (Fig. 5B and SI Appendix, Table S2). These two groups were largely comparable regarding their molecular definition of MS (SI Appendix, Fig. S5 D and E) and their clinical characteristics (SI Appendix, Table S2). Both groups had also similar proteinuria and estimated glomerular filtration rate (eGFR) at diagnosis of MS-associated microvascular inflammation (SI Appendix, Table S2). However, while eGFR worsened in the control group, it stabilized in the few months following mTOR inhibitor introduction (Fig. 5C). Correspondingly, per-protocol graft survival was markedly improved in the CNI+mTOR group on the long term (Fig. 5D, Log rank, P = 0.05).

To investigate whether the improved clinical outcomes in the CNI+mTOR group were associated with direct effects on MS-associated microvascular inflammation, we analyzed a subset of 21 patients from both groups who underwent follow-up biopsies at 17.1 ± 16.1 and 10.8 ± 4.8 mo (P = 0.72) postdiagnosis for the control and CNI+mTOR group respectively (Fig. 5A). The clinical characteristics and baseline histological severity of these subgroups were similar (SI Appendix, Table S2). Supporting our hypothesis, a significant reduction in Banff g and ptc scores was observed in the CNI+mTOR group, while the microvascular inflammation (MVI) burden remained stable in the control group (Fig. 5 E and F).

Discussion

This study demonstrates that the inability of donor-derived graft endothelial cells to deliver HLA I-dependent inhibitory signals to a subset of the recipient’s circulating NK cells—a condition termed “missing self” (MS)—is sufficient to drive DSA-independent NK cell activation, leading to MVI and impacting graft survival in a manner comparable to AMR. Along with our previous seminal work (20) and recent corroborative studies (21, 38), these findings are reshaping our understanding of graft rejection immunopathology. Unlike traditional rejection involving recognition of “new” donor-specific alloantigens by the recipient’s adaptive immune cells (T and/or B lymphocytes), MS-associated MVI arises from the activation of the recipient’s innate NK cells due to the absence of a self-antigen. Of note, although MS or DSA (through CD16 signaling) alone can both drive MVI, the combination of the two stimuli can enhance MS-induced NK cell activation (39), leading to more severe MVI and worse graft survival (21, 39). Collectively these findings support a broader view of MVI as a histological phenotype arising from multiple (possibly overlapping) causes, that converge on NK cell activation (26). Though new to transplant immunology, this concept parallels fields like thrombotic microangiopathy, where endothelial cell activation through various stimuli leads to a shared pathological pathway and consistent histological pattern (40). This vision aligns well with the last revision of Banff classification (29), which introduced the “MVI+DSA−C4d−” category that was a posteriori validated by a recent international epidemiological study (41).

A key goal of the Banff community in introducing the MVI+DSA−4d− category was to stimulate further research and, ultimately, to improve clinical outcomes for this population, whose prognosis is as poor as that of patients with AMR. To investigate the molecular mechanisms of MS-associated MVI and identify potential therapeutic targets, we applied a translational approach, integrating transplant patient sample analysis with custom in vitro and in vivo experimental models. Our findings indicate that although activated NK cells constitute a minority of cells within MS-associated MVI lesions, they play a critical role. In vivo murine models confirmed this, as NK cell depletion via monoclonal antibody (mAb) treatment completely prevented the development of MVI lesions. This is especially relevant given recent evidence of felzartamab, a depleting anti-CD38 mAb, showing transient efficacy in treating AMR-related (DSA-induced) MVI lesions (42). Anti-CD38 targets NK cells (43) and our data (Fig. 3A) suggests this approach could also be beneficial for MS-associated MVI, reinforcing the concept of a common MVI histological pathway despite different initial triggers presented in the previous paragraph.

In the absence of NK-cell depleting options available for this indication in the clinic, we turned our focus to our findings that MS-induced NK cell activation depends on calcium and mTOR signaling. This led us to explore CNI and mTORinh, which are both established classes of immunosuppressive agents for preventing organ rejection, as a potential treatment strategy. Our data demonstrate that while CNI exert some beneficial effects on MS-induced NK cell activation in vitro—potentially explaining why only a small fraction of renal transplant recipients with genetically predicted MS develop microvascular inflammation—these agents alone may not fully prevent the development of MVI lesions in all patients. This limitation likely arises because CNIs target only a subset of the calcium signaling pathways involved in MS-induced NK cell activation (Fig. 2 E–H). In contrast, mTOR inhibition effectively prevented MVI development in murine models and showed synergistic therapeutic potential on MS-induced NK cell activation when associated with CNI in vitro. Based on these promising preclinical results we conducted a single-center pilot clinical study. In this study, initiating mTOR inhibition on top of CNI upon MS-associated MVI diagnosis stabilized graft function and reduced MVI lesions in subsequent biopsies, resulting in improved long-term graft survival in patients who tolerated the treatment.

A significant limitation of this therapeutic approach was the poor tolerance profile of late mTORinh introduction, with >50% of patients discontinuing treatment due to adverse effects, as previously reported (36, 37, 44, 45). It is therefore impractical to expose all transplant recipients with genetically predicted MS [approximately 30% of transplant recipients (27, 39)] to the side effects of mTORinh without clear benefit. One potential alternative is to reserve mTORinh treatment for patients diagnosed with MS-associated microvascular inflammation, though this approach presents several challenges. MS-associated microvascular inflammation is often subclinical, diagnosed upon the presence of MVI lesions on protocol biopsies, which are performed (at best) only once or twice and mostly within the first 12 mo posttransplantation (46). As a result, many cases of MS-associated microvascular inflammation are diagnosed late, during biopsies performed for clinically (e.g., decreased eGFR) or biologically (e.g., proteinuria) evident symptoms, which arise as a consequence of chronic, irreversible histological lesions that ultimately limit treatment efficacy. Furthermore, in the case of kidney transplantation, nephron reduction increases filtration pressure in the remaining glomeruli, prompting them to recruit their cytoskeleton for adaptation. This process depends on the mTORC2 signaling pathway (47), which is affected by mTOR inhibitors (48). Late introduction of mTOR inhibitors could thus exacerbate adverse effects by increasing susceptibility to mTOR inhibitor-induced focal segmental glomerulosclerosis (FSGS) (47). In such cases, controlling MS-associated microvascular inflammation would be counterbalanced by the drug’s direct toxic effect on podocytes, worsening clinical outcomes. This underscores the need for a biomarker that enables longitudinal, noninvasive monitoring of the activation status of circulating NK cell subsets in MS situations involving the graft—an objective that, while complex, appears achievable given the recent advancements in automated single cell isolation and dispensing technologies (49). Ideally, such a biomarker would be paired with one that tracks graft vasculature damage, such as donor-derived cell free DNA (ddcfDNA) (50), to identify graft recipients at risk of MS-associated MVI. This approach could lead the timing of early diagnostic biopsies, enabling timely initiation of mTOR inhibitor treatment.

This study is not without limitations. While the analysis of two independent kidney patient cohorts provides valuable insights, the conclusions of clinical studies should be interpreted with caution. The stringent selection of patients from the Imperial College cohort—focused on reducing potential confounders by including only those with MVI lesions on biopsy but without DSA—resulted in smaller subgroup sizes. Although the observation of a statistically significant difference despite limited power may suggest a strong association between the variable of interest (missing self) and graft survival, this approach inherently carries a risk of bias that must be acknowledged. This is precisely why we considered it crucial to complement our clinical findings with mechanistic in vitro studies, which provided additional support for a causal relationship. Likewise, while the encouraging clinical outcomes observed with the combination of mTOR inhibitors and CNIs in patients with MS-induced MVI align with our preclinical data, this preclinical study does not meet the methodological rigor of a prospective randomized controlled trial.

In conclusion, our study highlights MS-associated MVI as a distinct mechanism of graft rejection driven by NK cell activation. Although the dual inhibition with CNI + mTOR inhibitor has shown potential in preclinical and pilot clinical studies, the safety profile of this regimen poses challenges to widespread use. This underscores the urgent need for investigating alternative MS-induced NK cell activation pathways, which may pave the way for more tolerable therapeutic options. In this context, our validated murine model will serve as a valuable platform for testing therapies, underscoring the transformative potential of translational research in improving transplantation outcomes.

Materials and Methods

Clinical study.

Clinical cohorts.

At Imperial College, a retrospective cohort of adult kidney allograft biopsies from January 2015 to December 2017 was analyzed, excluding patients transplanted outside the United Kingdom or without tissue-typing data, with follow-up until April 2022. Samples were obtained from the Imperial College Healthcare Tissue Bank. In Lyon, biopsies with microvascular inflammation (MVI+) were retrospectively identified via the DIAMIC database (September 2004–December 2020), in accordance with French regulations and with informed consent (DIVAT); follow-up ended in March 2021. Longitudinal eGFR data were collected when available: 3 mo before the diagnostic biopsy (control group) or before CNI+mTORinh switch, and 6 mo after.

Histopathological features.

At Imperial College, biopsies were performed for allograft dysfunction, DSA detection, or before changes in immunosuppression. In Lyon, they were done routinely at 3 mo and 1-y posttransplant, or if rejection was suspected. All biopsies, whether indication-based or systematic, were reviewed by a single pathologist at each center (CR in London, MR in Lyon), blinded to genetic data, and scored per Banff criteria. Microvascular injury (MVI) was defined as g + ptc ≥ 2 without TCMR (i + t ≤ 3 if g = 0); the MVI diagnosis date was the first MVI+ biopsy. Control (MVI–) patients had no subsequent MVI+ biopsies during the study.

Detection of anti-HLA antibodies.

In London, de novo anti-HLA DSAs were screened twice in the first week posttransplant, then at 1, 3, 6, 12 mo, and annually; in Lyon, at 3, 12 mo, and annually. Both centers also collected samples during allograft dysfunction or for-cause biopsies. In London, DSAs were screened with LABScreen mixed beads and specified with single antigen beads (One Lambda). In Lyon, screening used LifeScreen flow beads and LSA single antigen beads (Immucor). Patients with antibodies detected in at least 2 samples and MFI >500 were considered DSA-positive.

HLA and KIR genotyping.

In London, recipients were typed using in-house PCR-SSP or LABType® PCR-SSO kits (One Lambda), and donor HLA typing came from local labs or NHS Blood and Transplant records. In Lyon, Luminex reverse PCR-SSO (One Lambda pre-2012, Immucor after) was used for recipients and living donors, and reverse PCR-SSO or PCR-SSP (Linkage Biosciences) for deceased donors. All patients were genotyped for four inhibitory KIR genes (2DL1, 2DL2, 2DL3, 3DL1) using either Lifecodes PCR-SSO or a PCR-SSP method developed by the Lyon HLA lab (at the Etablissement Français du Sang).

NanoString gene expression on biopsies.

Six 12 mm curls were cut from each FFPE block using a Leica microtome, with blades changed and equipment cleaned between samples. RNA was extracted using the RNeasy FFPE kit (Qiagen) and eluted in RNase-free water. RNA purity and concentration were measured with a NanoDrop 2000c. Gene expression was analyzed using the NanoString B-HOT panel (30) with 100 ng RNA hybridized at 65 °C for 18 h, then quantified via the nCounter SPRINT system. Genes with mean counts <10 were excluded. DESeq2 (1.44.0) was used for normalization (via housekeeping genes) and differential expression analysis [fold change (FC) ≥ 1, adjusted P ≤ 0.05].

Murine preclinical model.

Mice.

Wild-type C57BL/6 (H-2b) mice (8 to 15 wk) were purchased from Charles River Laboratories. β2-microglobulin knockout (KO) C57BL/6 mice were provided by Laurent Genestier (CIRI, Lyon). All mice were housed in germ-free conditions at the AniRA-PBES facility (Lyon, France).

Heart transplantation.

Cardiac allografts were transplanted into the subcutaneous space of the neck, connecting the graft’s aorta to the recipient’s carotid artery and the pulmonary artery to the jugular vein (7). Grafts were stored at 4 °C for 3 h to induce ischemia/reperfusion injury and harvested 60 d posttransplant for histology. CD45+ hematopoietic cells were stained (clone 30-F11; BD Biosciences) and semiquantitatively scored (0 to 3+) by a transplant pathologist (JPD). Where indicated, mice received intraperitoneally 100 μg anti-NK1.1 antibody (clone PB136; BioXcell) twice weekly from day −7; NK depletion was confirmed via NKp46+ cell quantification. Cyclosporin (20 mg/kg) or rapamycin (3 mg/kg) was administered every other day from day −7 to the experiment’s end.

RNA sequencing on transplanted hearts.

RNA of heart transplants was extracted using Rneasy microkit (Ambion) according to the manufacturer ‘instructions. Genomic DNA was digested with TurboDNase and RNA quality was assessed with Bioanalyzer (RINs were ranged from 8 to 9.6). Libraries were made with the Illumina trueseq total stranded library prep kit using 1 ng total RNA. The sequencing was performed by Genewiz society on a NovaSeq3000 (Genewiz) (2 × 150bp, single indexing).

Reads were aligned to mm10 mouse genome with Hisat2. Counted and produced counts table were generated with FeatureCounts (Rsubread). The normalization and differential expression analysis were performed using DEseq2 (1.44.0) with 1 and 0.05 as thresholds for FCs and adjusted P-values respectively. Normalization was done using default parameters of DESeq2. The shrinkage was done by enabling the “betaPrior” parameter of the DESeq2 function. Graph generation and data analysis were done in Rstudio (2023.12.1 + 402) using R (4.4.1) and the following packages: ggplot2 (3.5.1), ComplexHeatmap (2.20.0), RColorBrewer (1.1-3), circlize (0.4.16), ggvenn (0.1.10), enrichplot (1.24.0), ggbreak (0.1.2), DESeq2 (1.44.0), tidyverse (2.0.0), magrittr (2.0.3), stringr (1.5.1), data.table (1.15.4), “apeglm (1.26.1), biomaRt (2.60.1), org.Hs.eg.db (3.19.1), clusterProfiler (4.12.0), GSEA (1.2).

Human in vitro model.

Cells preparation.

The human glomerular endothelial cell line CiGEnC (32) (kind gift from Professor Simon Satchell, from the University of Bristol, United Kingdom) was cultured at 33 °C (passages 2 to 35) and differentiated for 7 d at 37 °C. On day 1, CiGEnC were trypsinized and plated on 1% gelatin-coated plates in endothelial growth medium (PromoCell). PBMCs were isolated from healthy donors (EFS Lyon) by Ficoll gradient. Donors were selected by HLA-I type to ensure NK cell education (C1/C2 or C2/C2 for KIR2DL1+; C1/C1 and Bw4 for KIR2DL2/3+ and KIR3DL1+). NK cells (>90% purity) were purified by negative magnetic selection(Stemcell) and cultured overnight at 37 °C with 50 ng/mL IL-15 (Peprotech). Cyclosporin (100 ng/mL) and/or rapamycin (10 ng/mL) were added overnight to CiGEnC or NK cells as indicated.

Microscopy phospho-S6rp staining.

On day 2, 0.6 × 10⁵ NK cells were added per well with CiGEnC cells (1:1 ratio) in µ-Slide 8 wells (Ibidi) and incubated in complete RPMI at 37 °C, 5% CO₂ after centrifugation. After 1 h, cells were fixed with 4% formaldehyde, washed, and blocked with 5% goat serum and 0.3% Triton X-100. Primary antibodies (rabbit anti-pS6rp, Cell signaling, 1:800; mouse IgM anti-KIR2DL1, BD Pharmingen, 1:500) were incubated overnight at 4 °C. After washing, secondary antibodies (goat anti-rabbit AF568 LIfe, goat anti-mouse AF488 Invitrogen, both 1:1,000) were applied for 2 h. After final washes, Fluoromount aqueous mounting (Sigma) was added, and images were acquired on a CQ1 confocal cytometer (Yokogama) (40×, 15 z-stacks) and analyzed with ImageJ and Imaris.

Flow cytometry.

On day 2, 0.5 × 10⁵ NK cells, pretreated or not with cyclosporin and/or rapamycin, were cocultured with CiGEnC cells (also pretreated or not with rapamycin) at a 1:1 ratio in 200 µL complete RPMI in 96-well plates. FITC-conjugated anti-CD107a antibody (clone eBioH4A3, 1/8, Invitrogen) was added, and plates incubated at 37 °C, 5% CO₂ after centrifugation. After 1 h, Golgi Stop (BD Biosciences) was added and coculture continued for 3 more hours. Cells were washed, trypsinized, washed again, pooled with pretrypsin cells, and stained extracellularly for 20 min at 4 °C. APC-H7-conjugated anti-CD3 (clone SK7, 1/25, BD Biosciences), PE-CF94-conjugated anti-CD56 (clone NCAM16.2, 1/50, BD Biosciences), APC-conjugated anti-KIR2DL1 (clone HP-3E4, 1/10, BD Biosciences), APC-conjugated anti-KIR2DL1/S1 (clone EB6B, 1/10, Beckman Coulter), PE-Cy5.5-conjugated anti-KIR2DL2/3 (clone GL183, 1/25, Beckman Coulter), BV605-conjugated anti-KIR3DL1 (clone DX9, 1/25, BD Biosciences), FITC-conjugated anti-CD107a (clone eBioH4A3, 1/8, Invitrogen), eF780-conjugated fixable viability dye (1/1,000, Invitrogen). Cells were then subsequently fixed and permeabilized (Foxp3/Transcription Factor Staining Buffer Set, eBioscience), stained with V450-conjugated anti-MIP-1β (clone D21-1351, 1/20, BD Biosciences) and PE-conjugated anti-IFN-γ (clone 4S.B3, 1/20, Invitrogen) and acquired by flow cytometry on Fortessa 4L (BD Biosciences). Analyses were performed with FlowJo software. Some healthy donors did not express the Bw4 or the C1 molecules, respective ligands of KIR3DL1 and KIR2DL2/3, in this case iKIR were not considered educated and were not plotted in the corresponding iKIR group.

Killing assay.

For the killing assay, 1 × 10⁴ CiGEnC were seeded in gelatin-coated E-plates and incubated in growth medium 12 h at 37 °C, 5% CO₂. Cell viability was monitored by impedance (xCELLigence RTCA). Then, 2 × 10⁵ NK cells (HLA-C C1/C1 donors), with or without cyclosporin and/or rapamycin, were added (effector:target 20:1) in 100 µL RPMI. Plates were incubated 10 h while monitoring CiGEnC viability. Blocking antibodies against KIR3DL1 (clone DX9; 1/500; BD Biosciences) and KIR2DL2/3 (clone GL183; 1/200; Beckman Coulter)) or isotype controls were added when indicated. Cell indices were normalized to values before NK addition and to control wells without NK cells or blocking antibodies.

Imaging flow cytometry.

On day 2, NK cells were washed with complete RPMI and 2.10⁵ NK cells, pretreated or not with cyclosporin, were added in a 24-well plate, along with CiGEnC cells (effector-to-target ratio of 1:1), in a final volume of 800 µL of complete RPMI. The plate was incubated at 37 °C in 5% CO₂ after a 2-min centrifugation at 100 g. After 1 h of coculture cells were washed twice with cold PBS, dissociated using trypsin, washed again with complete RPMI, and then pooled with the cells collected prior to trypsin dissociation. Cells were first incubated 30 min at 4 °C with extracellular staining panel: APC-H7-conjugated anti-CD3 (clone SK7, 1/25, BD Biosciences), PE-CF94-conjugated anti-CD56 (clone NCAM16.2, 1/50, BD Biosciences), APC-conjugated anti-KIR2DL1 (clone HP-3E4, 1/50, BD Biosciences), APC-conjugated anti-KIR2DL1/S1 (clone EB6B, 1/50, Beckman Coulter), PE-Cy5.5-conjugated anti-KIR2DL2/3 (clone GL183, 1/25, Beckman Coulter), BV605-conjugated anti-KIR3DL1 (clone DX9, 1/25, BD Biosciences), eF780-conjugated fixable viability dye (1/1,000, Invitrogen). Cells were then subsequently fixed and permeabilized (Foxp3/Transcription Factor Staining Buffer Set, eBioscience) and stained with AF488-conjugated anti-NFAT1 (clone D43B1, 1/10, Cell Signaling) 30 min at 4 °C. Sample acquisitions were made on an ImageStream X Mark II (Amnis-EMD Millipore, Darmstadt, Germany) with ×40 magnification after extemporaneous addition of DAPI (1/100), and analyzed with IDEAS software (v6.0). The similarity score, derived from the Pearson’s correlation coefficient of pixel intensities, quantifies the correlation between two spectrally distinct images of a single.

Statistics.

Graphs were generated using Prism (v10.4.0) or R (v4.4.1). Group differences were assessed using Mann–Whitney, Wilcoxon, Fisher’s exact test, or one-way ANOVA, depending on sample size and data distribution, with the test specified in figure legends. Renal graft survival was measured from rejection diagnosis to last follow-up or censoring (e.g., graft loss, mTORi discontinuation, or DSA). Kaplan–Meier curves were compared using the log-rank test. P-values < 0.05 were considered significant (*P ≤ 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Study Approval.

ICHTB is supported by the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) based at Imperial College Healthcare NHS Trust and Imperial College London. ICHTB is approved by Wales REC3 to release human material for research (22/WA/2836). For Lyon’s patient cohort, approval was obtained from Comité Consultatif sur le Traitement de l’Information en matière de Recherche dans le domaine de la Santé and Commission nationale de l’informatique et des libertés. All patients signed a written informed consent form authorizing the use of their personal data and biological material for research purposes, in accordance with applicable ethical guidelines. For mice model, all procedures were performed in accordance with EU guidelines and were approved by the local ethical committee for animal research (CECCAPP Lyon, registered by the French National Ethics Committee of Animal Experimentation under No.15, http://www.sfr-biosciences.fr/ethique/experimentation-animale/ceccapp).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All datasets used in this manuscript are listed in S_Datasets, available in the Supporting information. The R code used for the transcriptomic analyses has been deposited in a publicly accessible online repository and is available at: https://gitbio.ens-lyon.fr/ciri/nopab/sarah-h/hamada_and_al_2025 (51). In accordance with the PNAS data availability policy, the transcriptomic raw data supporting the findings of this study are available at: https://doi.org/10.57745/QBUY3M (52).